Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Effects of component placement on patellar kinematics and loading in intraoperative and postoperative… Brimacombe, Jill Maureen 2006

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_2006-0160.pdf [ 34.53MB ]
Metadata
JSON: 831-1.0080712.json
JSON-LD: 831-1.0080712-ld.json
RDF/XML (Pretty): 831-1.0080712-rdf.xml
RDF/JSON: 831-1.0080712-rdf.json
Turtle: 831-1.0080712-turtle.txt
N-Triples: 831-1.0080712-rdf-ntriples.txt
Original Record: 831-1.0080712-source.json
Full Text
831-1.0080712-fulltext.txt
Citation
831-1.0080712.ris

Full Text

EFFECTS OF COMPONENT PLACEMENT ON PATELLAR KINEMATICS AND LOADING IN INTRAOPERATIVE AND POSTOPERATIVE LOADING CONFIGURATIONS by JILL MAUREEN BRIMACOMBE B.Sc.(Eng), The University of Alberta, 2003 A THESIS SUBMITTED IN PARTIAL F U L F I L L M E N T OF THE REQUIREMENTS FOR THE DEGREE OF M A S T E R OF APPLIED SCIENCE in THE F A C U L T Y OF G R A D U A T E STUDIES (Mechanical Engineering) THE UNIVERSITY OF BRITISH C O L U M B I A April 2006 © Jill Maureen Brimacombe, 2006 Abstract Postoperative patellofemoral complications and anterior knee pain contribute to suboptimal clinical outcomes following total knee replacement (TKR). Pain and complications are thought to be related to patellar tracking and loading, and the placement of the T K R implants can influence these biomechanical parameters. Surgeons aim to improve postoperative performance of the knee by optimizing intraoperative implant placement; however, they can only judge patellofemoral mechanics intraoperatively by observing passive, unloaded knee flexion. The main purpose of our study was to compare the patellar kinematics (tilt, shift and spin) and peak patellofemoral contact forces for cadaveric specimens in testing configurations designed to simulate both passive intraoperative and loaded postoperative flexion. The goal was to rank the effects of varying the placement of the TKR implants on kinematics and forces in both testing configurations. Zimmer NexGen posterior-stabilized implants were implanted in 8 fresh-frozen adult anatomic specimen knees. The knees were cycled dynamically through flexion and extension in rigs designed to simulate intraoperative and postoperative flexion. The femoral, tibial and patellar implants were modified to allow for the following changes in placement: 1) external and internal rotation of the femoral implant 2) external and internal rotation of the tibial implant 3) 3 patellar resection angles 4) 2 mediolateral patellar positions 5) additional patellar thickness. Changes in patellar tilt and shift were moderately- to well-correlated between the intraoperative and postoperative simulations (r = 0.70 to 0.85 for tilt, r = 0.54 to 0.77 for shift). The correlation indicates that changes made intraoperatively to optimize tracking will likely result in similar changes in the postoperative knee. Spin was not correlated between simulations. No changes in peak contact force were statistically significant in either testing rig and peak contact forces were not well-correlated between simulations. It is hoped that the results and protocol of this study will be used to guide a future clinical study comparing optimal intraoperative patellar tracking and loading to optimal postoperative mechanics. We also intend to use the results of this study to assist in the development of a computer assisted surgical technique for resurfacing the patella. ii Table of Contents Abstract » Table of Contents iii List of Tables vii List of Figures viii Acknowledgements xiii 1 Literature Review 1 1.1 Introduction 1 1.2 The Patellofemoral Joint 1 1.2.1 Anatomy 1 1.2.2 Mechanics 3 1.2.3 Kinematics, Contact Area, Force and Stress 5 1.3 Total Knee Arthroplasty 7 1.3.1 T K A Successes and Issues 7 1.4 The Etiology of Postoperative Pain and Complications 9 1.4.1 Cadaveric Studies 11 1.4.2 Clinical Studies 11 1.5 The Effects of Surgical Variables 13 1.5.1 Combined Femoral and Tibial Component Rotation 14 1.5.2 Femoral Component Rotation and Translation 15 1.5.3 Tibial Component Rotation and Translation 17 1.5.4 Patellar Resection Angle 18 1.5.5 Position of the Patellar Component 18 1.5.6 Patellar Thickness 19 1.5.7 Lateral Release 20 1.5.8 Recovery Time 21 1.5.9 Other Variables 21 1.6 Current Research Gaps and Limitations 22 1.6.1 Relationship between Intraoperative and Postoperative Biomechanics 22 1.6.2 Patellar Component Placement 22 1.6.3 Comparison of Wide Range of Variables 23 1.6.4 Relationship Between Tracking and Loading 24 1.6.5 Dynamic Measurement of Loads 24 1.7 Research Questions 25 2 Materials and Methods 26 2.1 Introduction 26 2.2 Kinematics 28 2.3 Force Measurement 29 2.3.1 Selection of a Measurement System 29 2.3.2 Description of the Tekscan I-Scan Pressure Measurement System 31 2.3.3 Conditioning and Calibration 32 2.3.4 Digitizing the Patellar Circumference and Landmarks 34 2.3.5 Other Details 35 2.4 Surgical Variables 35 2.4.1 Femoral Rotation 36 2.4.2 Tibial Rotation 37 2.4.3 Patellar Resection Angle 38 2.4.4 Mediolateral Patellar Position 38 2.4.5 Patellar Thickness 38 2.5.1 Femoral Rotation 39 2.5.2 Tibial Rotations 49 2.5.3 Patellar Placement 53 2.6 Testing Rigs 60 2.6.1 Horizontal Testing Rig 61 2.6.2 Vertical (Oxford) Rig 64 2.7 Integration of Measurement Systems, Knee Specimen and Testing Rigs 67 2.7.1 Adhering the Pressure Sensor to the Patella 67 2.7.2 Sensor Degradation and Loss of Output 68 2.8 Test Design 73 2.8.1 Specimen Details 73 2.8.2 Test Variables 74 2.9.1 Kinematic Analysis 76 2.9.2 Kinetic Analysis 82 iv 3 Validation of Calibration Techniques for Tekscan Pressure Sensors 85 3.1 Introduction 85 3.2 Methods 87 3.3 Results 90 3.4 Discussion 92 4 Results 95 4.1 Patellar Kinematics 95 4.1.1 Femoral Component Rotation 98 4.1.2 Tibial Component Rotation 103 4.1.3 Patellar Resection Angle 106 4.1.4 Patellar Medialization and Thickness 110 4.2 Ranking of Variable Effects on Kinematics 113 4.2.1 Tilt 113 4.2.2 Shift 113 4.3 Comparison of Kinematics between Horizontal and Vertical Rigs 117 4.4 Additional Tests: Kinematics 122 4.5 Patellar Contact Forces 125 4.6 Ranking of Variable Effects on Forces 128 4.7 Comparison of Forces between Horizontal and Vertical Rigs 129 4.8 Additional Tests: Forces 130 4.9 Contact Area and Pressure 132 4.10 Quadriceps Tension... 134 4.11 Summary 135 5 Discussion 138 5.1 Introduction 138 5.2 Tracking: Neutral Component Placement 138 5.3 Femoral Component Rotation 141 5.3.1 Femoral Component Rotation: Tracking 141 5.3.2 Femoral Component Rotation: Loading 148 5.4 Tibial Component Rotation 150 5.4.1 Tibial Component Rotation: Tracking 150 5.4.2 Tibial Component Rotation: Loading 151 v 5.5 Patellar Resection Angle 151 5.5.1 Patellar Resection Angle: Tracking 151 5.5.2 Patellar Resection Angle: Loading 154 5.6 Position of the Patellar Component 154 5.6.1 Position of the Patellar Component: Tracking 155 5.6.2 Position of the Patellar Component: Loading 156 5.7 Patellar Thickness 157 5.7.1 Patellar Thickness: Tracking , 157 5.7.2 Patellar Thickness: Loading 157 5.8 Ranking of Variables 158 5.8.1 Tracking 158 5.8.2 Loading 159 5.9 Comparison between Horizontal and Vertical Rigs 160 5.9.1 Tracking 160 5.9.2 Contact Forces 161 5.10 Relationship between Tracking and Loading 163 5.11 Strengths and Novelty of Study 163 5.12 Potential Confounding Factors 165 5.13 Limitations 166 5.14 Contributions and Future Work 172 6 Conclusions 175 References 178 Appendix 1: Adjustable Femoral Component 192 Appendix 2: Adhesives and Lubricant Selection 194 Appendix 3: Order of Tests Performed and Additional Tests 197 Appendix 4: Problems Experienced 198 Appendix 5: Excluded or Incomplete Data 204 Appendix 6: Patellar Rotation Results 206 Appendix 7: Statistical Data for Paired t-tests 209 VI List of Tables Table 1.1: Summary of studies correlating pain and tracking 12 Table 1.2: Summary of studies correlating complications and tracking 13 Table 1.3: Summary of clinical studies analyzing pain, component rotations, tracking 14 Table 1.4: Relationship between femoral rotation and postoperative tracking 16 Table 1.5: Relationship between patellar resection angle and tracking indices 18 Table 1.6: Relationship between patellar medialization and tracking indices 19 Table 1.7: Relationship between increased patellar thickness and tracking indices 20 Table 1.8: Relationship between preoperative patellar tilt and postoperative tilt 21 Table 2.1: Femoral rotations studied in vitro by other researchers 37 Table 2.2: Mediolateral patellar positions investigated in in vitro studies 38 Table 2.3: Patellar thicknesses studied in vitro 39 Table 2.4: Specimen details: sex, age, side of body, ideal component sizes 73 Table 3.1: Comparison of calibration accuracy 93 Table 4.1: P-values for Pearson's correlations comparing changes between testing rigs.. 118 Table 4.2: Average values of slopes and intercepts of the best-fit lines through horizontal and vertical tracking data 119 Table 4.3: P-values for Pearson's correlations comparing raw tracking data between testing rigs 121 Table 4.4: Changes in patellar tilt, shift and spin for static and dynamic flexion angles .. 124 Table 5.1: Maximum standard deviations of patellar tilt and shift in similar studies 140 Table 5.2: Summary of results of cadaveric studies on effects of femoral rotation on patellar tracking 142 Table 5.3: Number of specimens tested in cadaveric studies of the knee 165 vii List of Figures Figure 1.1: The anatomy of the knee joint 2 Figure 1.2: The bones, ligaments, tendon and menisci of the knee joint 2 Figure 1.3 : Lateral and medial capsular structures of the knee 3 Figure 1.4: Free body diagram (sagittal view) of the knee during open-chain flexion 4 Figure 1.5: Measurement of the Q-angle 5 Figure 1.6: Parameters of patellar tracking: shift, tilt and rotation 6 Figure 1.7: Reasons reported for revising total knee replacement procedures in Canada.... 9 Figure 1.8: Two main branches of research on patellofemoral issues 11 Figure 2.1: Schematic of knee specimens in horizontal and vertical testing rigs, with kinematic and kinetic measurement systems 27 Figure 2.2: Tekscan pressure sensor, model #5051 32 Figure 2.3: Tekscan sensor calibration apparatus 33 Figure 2.4: Typical cubic polynomial calibration curve for Tekscan pressure sensor 34 Figure 2.5: Tekscan software showing digitized circumference of patella 35 Figure 2.6: Modifications to the placement of the patellar implant: resection angle, mediolateral position, and additional thickness 36 Figure 2.7: Diagram of the modified femoral component implanted in the femur 42 Figure 2.8: Attachment of wedged steel block to femoral component using press-fit steel dowels, and proximally threaded steel rod extending from modified femoral component 42 Figure 2.9: The inferior aspect of the modified femoral component 43 Figure 2.10: Nuts, lockwasher, pointer on femoral rod and protractor on femoral tube.... 44 Figure 2.11: Protractor and aluminum pointer 45 Figure 2.12: Surgical towel clamps used to close medial incision 46 Figure 2.13: Stainless steel tube resting temporarily in intramedullary canal of femur, and the remaining femoral bone following surgery 48 Figure 2.14: Trial tibial tray, plastic tibial spacer, and aluminum tibial baseplate 49 Figure 2.15: Tibial baseplate secured to bone using 2 screws, and installed tibial tray 50 Figure 2.16: 3-part tibial construct and its attachment to the bone of the tibial plateau .... 51 Figure 2.17: Complete tibial construct secured in bone in neutral and external rotation... 51 Figure 2.18: Marking on tibia used to align tibial baseplate 53 Figure 2.19: Patellar construct in neutral alignment 54 Figure 2.20: Original NexGen patellar implant and modified trial patellar implant, anterior and posterior sides 54 Figure 2.21: A) Steel disk representing neutral placement. B) Steel wedge used to introduce 7.5° lateral bone cut angle. 55 Figure 2.22: A) Steel wedge used to introduce 15° lateral bone cut angle. B) Patellar component with 3mm thickness disk added 55 Figure 2.23: Patellar baseplate and neutral placement disk 56 Figure 2.24: Exposing the posterior surface of the natural patella 57 Figure 2.25: Measuring the original patellar thickness with calipers, and resecting the patellar bone 57 Figure 2.26: Using arched calipers to measure medial and lateral patella thicknesses 58 Figure 2.27: A) Patellar landmarks and K-wire inserted through patellar bone for marker array attachment. B) Patellar baseplate attached to bone 58 Figure 2.28: Marker array holder attached to patella via threaded K-wire, aluminum block and figure-eight wire twistings 59 Figure 2.29: The stainless steel tube cemented into the intramedullary canal of the femur 60 Figure 2.30: Knee specimen in horizontal rig 62 Figure 2.31: The anterior and posterior plates forming the quadriceps tendon clamp 63 Figure 2.32: Load cell inserted between slider and spring in horizontal rig for specimen 764 Figure 2.33: Vertical (Oxford-style) testing rig and the motor-clamp connection 66 Figure 2.34: Tekscan sensor glued to the patellar implant, inducing a fold in the sensor... 68 Figure 2.35: Wrinkles in sensor occurred if patellar implant was not glued several hours prior to testing. 68 Figure 2.36: Tekscan handle taped to tibia, causing the sensor tab to twist 71 Figure 2.37: The Tekscan handle holder attached to the tibial rod 72 Figure 2.38: Reduction in sensor degradation between earlier and later specimens 73 Figure 2.39: Reference frames to specify patellar and tibial tracking with respect to femur 77 Figure 2.40: Measurement of patellar tilt, shift and spin using axes of motion of patella... 78 Figure 2.41: Definitions of positive (lateral) tilt, shift, and spin 79 ix Figure 2.42: Calculation of scaling factors used to normalize data to first baseline measurement 83 Figure 3.1: The two Tekscan calibration methods: linear and power 86 Figure 3.2: Application of load to a Tekscan sensor using an Instron materials testing machine 88 Figure 3.3: Typical 10-point cubic polynomial calibration curve 89 Figure 3.4: Timeline of calibrations and loading 90 Figure 3.5: Typical sensor output calibrated using the 3 Tekscan calibrations and 2 user-defined calibrations 91 Figure 3.6: Average RMS errors of Tekscan and user-defined calibration algorithms 92 Figure 3.7: Repeatability of 10-point, cubic polynomial 92 Figure 4.1: Raw tilt, shift and spin data for baselines 1-4 with specimen 2 in the horizontal and vertical rigs.. 97 Figure 4.2: Effects of ±5° rotation of femoral component on absolute patellar tilt in the horizontal and vertical testing rigs 101 Figure 4.3: Effects of ±5° rotation of femoral component on relative patellar tilt in the horizontal and vertical testing rigs 101 Figure 4.4: Effects of ±5° rotation of femoral component on absolute patellar shift in the horizontal and vertical testing rigs 102 Figure 4.5: Effects of ±5° rotation of femoral component on relative patellar shift in the horizontal and vertical testing rigs. The data were averaged across specimens at each flexion angle 102 Figure 4.6: Effects of ±5° rotation of the tibial component on absolute patellar tilt in the horizontal and vertical testing rigs 104 Figure 4.7: Effects of ±5° rotation of the tibial component on relative patellar tilt in the horizontal and vertical testing rigs 104 Figure 4.8: Effects of ±5° rotation of the tibial component on absolute patellar shift in the horizontal and vertical testing rigs 105 Figure 4.9: Effects of ±5° rotation of the tibial component on relative patellar shift in the horizontal and vertical testing rigs 105 Figure 4.10: Effects of resection angles on absolute patellar tilt in the horizontal and vertical testing rigs 108 x Figure 4.11: Effects of resection angles on relative patellar tilt in the horizontal and vertical testing rigs 108 Figure 4.12: Effects of resection angles on absolute patellar shift in the horizontal and vertical testing rigs 109 Figure 4.13: Effects of resection angles on relative patellar shift in the horizontal and vertical testing rigs 109 Figure 4.14: Effects of medialization and increased thickness on absolute patellar tilt in the horizontal and vertical testing rigs I l l Figure 4.15: Effects of medialization and increased thickness on relative patellar tilt in the horizontal and vertical testing rigs I l l Figure 4.16: Effects of medialization and increased thickness on absolute patellar shift in the horizontal and vertical testing rigs 112 Figure 4.17: Effects of medialization and increased thickness on relative patellar shift in the horizontal and vertical testing rigs 112 Figure 4.18: Effects of all tested surgical variables on changes in patellar tilt at 3 flexion angles in the horizontal and vertical rigs 115 Figure 4.19: Effects of all tested surgical variables on changes in patellar shift at 3 flexion angles in the horizontal and vertical rigs 116 Figure 4.20: Correlation between changes in tracking in horizontal and vertical rigs 118 Figure 4.21: Slopes of the best-fit lines through the horizontal and vertical data (changes with respect to baseline) at each flexion angle (averaged across specimens) 119 Figure 4.22: The linear relationships between changes in tilt, shift and spin in the horizontal and vertical rigs at 45° flexion 120 Figure 4.23: Correlation between raw tracking data for horizontal and vertical rigs 121 Figure 4.24: Effects of additional variables on patellar tilt, compared to baseline, in the horizontal and vertical rigs 123 Figure 4.25: Effects of additional variables on patellar shift, compared to baseline, in the horizontal and vertical rigs 123 Figure 4.26: Effects of additional variables on patellar spin, compared to baseline, in the horizontal and vertical rigs 124 Figure 4.27: Patellar contact forces measured by the Tekscan pressure sensor for two consecutive baseline trials in the horizontal rig 125 Figure 4.28: Patellar contact forces measured by the Tekscan pressure sensor 126 XI Figure 4.29: Peak contact forces for first baseline trials in horizontal and vertical rig 126 Figure 4.30: Flexion/extension contact force for 3 different surgical variables 127 Figure 4.31: Degradation in reported force following 6 patellar resection angle tests 128 Figure 4.32: Effects of all tested surgical variables on peak patellar contact force (compared to neutral component placement) in the horizontal and vertical rigs 129 Figure 4.33: Linear relationship between changes in peak forces in the horizontal and vertical rigs 130 Figure 4.34: Comparison of static and dynamic measurements of contact force 132 Figure 4.35: Measured contact force, measured contact area, and calculated contact pressure for two baseline trials in the vertical rig 133 Figure 4.36: Medial and lateral contact areas displayed using the Tekscan software 134 Figure 4.37: Tension in the quadriceps tendon for four trials in the horizontal rig 135 Figure 5.1: Patellar translation (shift) during static lifting, reported by Zimmer (2004), and patellar shift in the vertical rig, as reported in the current study 139 Figure 5.2: Location of rotation axis of femoral component and tilt axis of patella 145 Figure 5.3: Expected change in shift due to external rotation of femoral component 147 Figure 5.4: Diagrams of patellofemoral contact at 10°, 45° and 90° flexion 150 Figure 5.5: Tilt and shift of bone and marker array due to change in resection angle 153 Figure 5.6: Medialization of patellar component results in lateral translation of bone 155 Figure 5.7: Quadriceps tension in the vertical rig, as measured by Michael Paice at UBC, and in the horizontal rig, as measured in our study 162 Figure 5.8: Geometrical relationship between the magnitude of the quadriceps force and the resultant patellofemoral contact force 163 xii Acknowledgements First and foremost, I'd like to thank my family. You have always encouraged me in all my pursuits, even the ones that scared you, and I know how lucky I am to have your support, trust and love. Carolyn, thank you for juggling the oft-conflicting roles of friend, coworker, mentor, and quasi-supervisor. Your passion for this project was inspiring, your help was invaluable, and your company made those late nights with the knee specimens seem much less interminable. Thanks to Tony for your valuable guidance and Dave for your advice throughout the project. Gracias to Mike Paice for showing us the ropes (and cables) during testing. I am extremely grateful to Dr. Jerome Tonetti, who performed all our surgeries and provided several creative solutions to challenging practical issues. Thanks to Drs. Nelson Greidanus and Bas Masri for their helpful insight and explanations and for supporting this project. Thanks to Wayne Schaffer and Joe Kudzin for their help with project logistics, and to Zimmer for the donation of the trial components and the use of their surgical equipment. Special thanks are owed to Steve Carey and Markus Fengler for fielding my endless questions in the machine shop and making sure I still have all my fingers. Thanks to Lisa and Rosie, my delightfully eccentric roommates who put up with my strange views on cleanliness, gave wine or hugs (not always in that order) when needed, and helped make Vancouver my home. Thanks to Candela for your easy friendship, to Em for listening (yes, I noticed), to Katie for reminding me to leave the house, and to Ed, John, Kay, Lindsay, Marg, Laura, Carol and Sara for being the best posse of friends. Lastly, thanks to Brady- you challenge and inspire me, even at a distance of hundreds of kilometers. Thanks for believing in me and for taking the risk to start this adventure in the first place! xiii 1 Literature Review 1.1 Introduction The placement of implant components during total knee arthroplasty is thought to influence postoperative anterior knee pain as well as patellofemoral tracking and loading. The key focus of this chapter will be the discussion of existing studies on postoperative patellofemoral pain, tracking and loading. This chapter commences with an overview of the anatomy and mechanics of the patellofemoral joint. It introduces the total knee replacement as a treatment for severe arthritis and details the current issues relating to the success of replacements. Finally, a thorough analysis will be given of studies on postoperative patellofemoral pain, tracking and loading, and the gaps and lack of consensus in the current field of postoperative patellofemoral biomechanics will be explored. 1.2 The Patellofemoral Joint 1.2.1 Anatomy The knee joint is composed of two separate joints, the tibiofemoral joint and the patellofemoral joint. The tibiofemoral joint is composed of the distal end of the femur and the proximal end of the tibia and enables the tibia to flex with respect to the femur. In the patellofemoral joint, the cartilaginous posterior surface of the patella articulates with the trochlea of the femur (Figure 1.1). The trochlea is the cartilage-covered, V-shaped groove on the anterior aspect of the distal femur (Figure 1.2). It is composed of the inner walls of the medial and lateral condyles. The lateral condyle is typically higher (i.e. protrudes more anteriorly) than the medial condyle; thus, the lateral wall of the trochlea is higher and helps prevent the patella from sliding laterally. The intercondylar fossa (intercondylar notch) is the open space between the femoral condyles. The anterior part of its roof is the distal portion of the trochlea. Although the anterior surface of the patella is entirely bone, three quarters of the posterior surface is cartilaginous (the distal quarter is nonarticular and is thus not surfaced with cartilage). Patellar articular cartilage is the thickest cartilage in the human body (Grelsamer 1998). 1 femur (thigh bone fibula patella (knee cap) patellar tendon tibia (shin bone) Figure 1.1: The anatomy of the knee joint. (http://www.aclsolutions.com/images/Seif_knee%20anatomy02.jpg) lemur anterior cruciate ligament posterior crudate ligament • 4V medial collateral ligament trochlea medial 111 i': fXi $>0ti§ patellar tendon patella Figure 1.2: The bones, ligaments, tendon and menisci of the knee joint. (http://www.clinicalsportsmedicine.com/_images/chapters/23/Bruckner-2e_23_l_a.jpg) The extensor mechanism of the knee is comprised of the bones, muscles and tendons which act to extend the knee. The four muscles of the quadriceps are the main extensor muscles (rectus femoris, vastus intermedius, vastus lateralis, and vastus medialis). These muscles extend the knee when the foot is off the ground and prevent the knee from buckling when the foot is in contact with the ground. The rectus femoris originates at the anterior inferior iliac spine, and the other quadriceps muscles originate on the femoral shaft. A l l four muscles merge into the quadriceps tendon. This tendon inserts at the anterior aspect of the superior pole of the patella. At flexion angles greater than approximately 90°, the quadriceps tendon articulates with the trochlea of the femur. The patellar ligament connects the patella to the tibia; it originates at the inferior pole of the patella and attaches to the tibial tuberosity at the proximal end of the tibia (Figure 1.1) The capsular structures of the knee play an important role in maintaining patellar alignment (Figure 1.3). The lateral retinaculum restrains the patella on the lateral side. This thick structure represents the convergence of several structures including the iliotibial (IT) band and the lateral patellofemoral ligament. It is counteracted on the medial side by the medial patellofemoral ligament and by the vastus medialis obliquus (VMO) muscle. The medial capsular structures are less tight than the lateral structures; the medial retinaculum is much thinner than the lateral retinaculum, and the medial patellofemoral ligaments are not always present (Conlan 1993, Reider 1981). If the V M O is weak, there may be a tendency for the patella to be pulled laterally. The restraints provided by the medial soft tissues and muscles and by the trochlear groove are not always sufficient to restrict the patella to the trochlear groove (Grelsamer 1998). Quadriceps muscle Quadriceps tendon p i — Patella Lateral patellar retinaculum Patellar tendon Quadriceps tendon Vastus medialis obliquus ) \ Medial paiellofemora ligament Patellar tendon Figure 1.3 : Lateral and medial capsular structures of the knee. (http://healthgate.partners.org/images/si55551499_ma.jpg and http://images.webmd.com/images/hw/media65/medical/hw/nr551501.jpg) 1.2.2 Mechanics The patella acts as the fulcrum of a lever in the knee and gives the extensor mechanism a greater mechanical advantage. The patella also functions as an eccentric pulley because it changes the direction of the force applied by the quadriceps. Studies have shown that throughout most of the flexion-extension cycle (at angles greater than approximately 50°), the tensile force in the patellar ligament is lower than that generated by the quadriceps tendon (Ahmed 1987, Bishop 1977, Buff 1988, Grelsamer 1998, Huberti 1984b). The patella increases the moment arm of the extensor mechanism (Ahmed 1987, Grood 1984, Heegaard 1993, Hirokawa 1991, Kaufer 1971, van Kampen 1990) by increasing the distance between the extensor mechanism and the flexion axis of the knee. As a result, the knee requires less quadriceps force to flex and extend. Figure 1.4 shows a two-dimensional force analysis of the knee during open-chain flexion. If one sums the moments about the patellofemoral contact point, the quadriceps force, F q , acting at a distance di is balanced by the tension in the patellar ligament, F p , acting at a distance d2. Similarly, i f one sums the moments about the tibiofemoral contact point, the weight of the leg, W, acting at a distance d4 is balanced by F p acting at a distance d3. If W»d4 represents the moment due to the weight of the leg, this moment is counteracted by the quadriceps force acting at a distance d i»d3 /d2 . This distance represents the effective moment arm of the quadriceps mechanism. Figure 1.4: Free body diagram (sagittal view) of the knee during open-chain flexion. (Grelsamer 1998) 4 The quadriceps angle, or Q-angle, is a measure of the direction of pull of the quadriceps muscles. It is the angle subtended by the patellar ligament and a line connecting the anterior superior iliac crest (the origin of rectus femoris) and the centre of the patella (Figure 1.5). Typical Q-angles range between 10° and 15° (Grelsamer 1998, Grelsamer 2005). The greater the Q-angle, the W\\ tuberosity Figure 1.5: Measurement of the Q-angle. (http://moon.ouhsc.edu/dthompso/namics/compose.htm) 1.2.3 Kinematics, Contact Area, Force and Stress Patellar tracking (or kinematics) refers to the sliding of the patella on the anterior surface of the distal femur during early knee flexion and its subsequent gliding within the intercondylar notch during later flexion. Three tracking indices are typically reported: shift (mediolateral displacement), tilt (rotation about the superior-inferior axis of the patella) and rotation (or spin, about the anteroposterior axis of the patella) (Figure 1.6). The two former measurements are thought to be of most clinical significance since they are associated with patellar subluxation and dislocation. larger the lateral force acting on the patella (Elias 2004b). • Midpoint of patella •Anterior sy peri or ii>iac spine 5 Lateral stdft (&MKal v\mr in Figure 1.6: Parameters of patellar tracking: shift, tilt and rotation (Katchburian 2003). When the knee is in extension, the patella sits just proximal to the trochlea. Although some researchers claim that there is no contact between the patella and the trochlea in extension (Aglietti 1975, Goodfellow 1976), others claim to have measured contact (Ahmed 1987, Huberti 1984b). As the knee flexes to approximately 15°, the patella comes into contact with the trochlea. The patella normally shifts medially for the first few degrees of flexion and then tracks straight relative to the centre of the femur (or laterally in global coordinates, due to the orientation of the femoral groove) (Grelsamer 1998, Katchburian 2003). It is pulled into the trochlear groove at approximately 20-30° of flexion (Hungerford 1979, Kaufer 1971), and then runs centrally in the trochlear groove during deep flexion. The contact area on the patella moves proximally as the knee flexes. At 90°, the superior portion of the patella is in contact with the trochlea, and as the knee continues to flex, the contact area moves toward the centre of the patella again. There is no consensus as to typical patterns of patellar tilt during flexion of the natural knee. In their review article, Katchburian (2003) found that existing in vitro studies measured medial tilt in early flexion followed by lateral tilt in later flexion. In vivo studies have found both increasing medial and lateral tilt. Patellar rotation seems to be extremely variable, and no agreement has been reached regarding "normal" rotation. Inferosuperior translation, anteroposterior translation and flexion of the patella are not commonly measured as they have little clinical relevance. During closed kinetic chain activities (such as squatting), the patellofemoral joint reaction force increases from 0° to 90° of flexion (Grelsamer 1998). The contact area also increases during this 6 Lateral Ul% phase; however, the rate of increase in area is less. As a result, the stress acting on the patella increases during early flexion. Between 90° and 120°, the joint reaction force decreases or plateaus, likely because the quadriceps tendon comes into contact with the trochlea and shares a portion of the load. Although contact area decreases slightly, the stress at the joint also decreases as the knee continues to flex. The maximum force acting on the patella has been shown to measure 0.5-1.5 times body weight during normal gait, or up to 7 or 8 times body weight during squatting (Dennis 1992, Huberti 1984a). 1.3 Total Knee Arthroplasty Total knee arthroplasty (TKA), also called total knee replacement (TKR), involves the replacement of the injured or damaged parts of the knee joint (either the femur and tibia alone, or the femur, tibia, and patella) with artificial (prosthetic) components. The prosthetic components resurface the ends of the tibia and femur, and the posterior surface of the patella. They replace the function of the joint cartilage, the anterior cruciate ligament (ACL), and often the posterior cruciate ligament (PCL). In Canada, the most common diagnosis resulting in the need for a first (primary) T K A is degenerative osteoarthritis (92%) (CIHI 2005). Other diagnoses include inflammatory arthritis (5%), post-traumatic osteoarthritis (2%) and osteonecrosis (1%). Osteoarthritis is a chronic disease characterized by destruction of cartilage, overgrowth of bone, bone spur formation and impaired function. This type of arthritis occurs when bone rubs against bone and occurs in most people as they age. In 2002-2003, the Canadian Joint Replacement Registry recorded 26,500 hospitalizations for TKAs in Canada (CIHI 2005). Between 1996/1997 and 2002/2003, the number of total knee replacements increased by 77%, approximately a 10% annual increase. Most of the TKAs performed in Canada are primary total knee replacements; 94% of all replacements are primary, and revision procedures account for the remaining 6%. 1.3.1 TKA Successes and Issues The rate of clinical success and patient satisfaction following T K A is typically reported to be greater than 90% (Barrack 1997, Healy 1995, Huo 1990, Ranawat 1986). Long term survival analyses have estimated 96% to 98%) prosthesis survival at 10- to 12-year follow-up (Rand 1991, 7 Ritter 1994). Primary T K A is generally very successful in treating pain and improving knee function for patients with severe arthritis (Abraham 1988, Aglietti 1988, Barrack 1997, L i 1999). (Success rates reported in the 1980s were poorer; however, this was due in large part to the dislocation and disassociation of the metal-backed patellar components (Barrack 2001). These have since been replaced by more successful one-piece polyethylene components.) Unfortunately, a subset of patients continues to experience pain, complications or poor knee function following T K A . Some patients even report more pain or reduced quality of life postoperatively than preoperatively (Anglin 2004). Studies have reported an incidence of anterior knee pain of as much as 20% after total knee arthroplasty (Barnes 1993, Barrack 1997, Healy 1995, Kim 1999). A recent review article assessed the patellofemoral complication rate to be approximately 4% (Rand 2004), and a recently reported series of T K A revisions stated that 12% of revisions were performed because of problems relating to the extensor mechanism (Sharkey 2002). The reasons for postoperative pain (often anterior or peripatellar) and patellofemoral complications are often unclear (Barnes 1993, Barrack 1997, Rand 1994). Preoperative variables such as preoperative pain, obesity, radiological parameters, age, and sex cannot predict postoperative pain, although weight may or may not be a factor (Barrack 2001, Peng 2003, Pellengahr 2002, Waters 2003, Wood 2002). According to Barrack (2001), "[the] patella is a reliable guide to the success or failure of a total knee replacement. Patients who do not experience peripatellar symptoms or a patellar complication usually have a successful result. Conversely, peripatellar symptoms or complications usually reflect an underlying problem with surgical technique, component designs, or both." Patellofemoral complications include: clicking, popping, crepitation (crackling sounds), persistent discomfort, subluxation, dislocation, fracture of the patella, extensor mechanism disruption, loosening or disassociation of the implant, wear, failure of the implant, extensor lag (the patient cannot actively extend to a completely straight position), giving way, and painful locking. In Canada, 11% of knee revisions are due to pain (location unspecified), 3% are due to unresurfaced patellae, 3% are due to patellar failure, 2% are due to patellar maltracking, and 1% is due to issues with the extensor mechanism (Figure 1.7) (CIHI 2005). Other more frequently reported grounds for revision in Canada include aseptic loosening (49%), polyethylene wear (36%>), osteolysis (22%) and instability (13%). These symptoms were not reported specific to a 8 particular component, but it may be assumed that a portion of these symptoms were related to the patella. Aseptic loosening Poly wear Osteolysis Instability Pain of unknown origin Infection- two stage Implant fracture Bone Fracture Unresurfaced patella Infection- single stage. Patella failure Patella maltracking Extensor mechanism Other N/A 13% 9% • 4% 13 3% 33 3% 3 3% 3 2% is 1% 3 6% 3 15% • 22% 0% 10% 20% 30% Percentage 39% • * B % 40% Figure 1.7: Reasons reported for revising total knee replacement procedures in Canada (2002 and 2003 fiscal years). Note that percentages do not add up to 100% because more than one reason for revision can be recorded. (CIHI 2005) Patellofemoral issues are difficult to treat postoperatively; an effective method of managing these problems may be to prevent them by optimizing the biomechanics of the knee during surgery. 1.4 The Etiology of Postoperative Pain and Complications Although it is thought that poor tracking and high loading of the patella may result in postoperative anterior knee pain (AKP) and patellofemoral complications, there is no consensus amongst orthopaedic surgeons and researchers as to the exact etiology of pain and complications. There is likely a host of factors responsible for patellofemoral problems. Parameters which may affect the success of a T K A include: component malpositioning (femoral, tibial or patellar), lack of soft-tissue constraints to stabilize tracking, inadequate repair of medial structures (failure to repair may be responsible for patellar dislocation), fibrous band formation, increased tension or pressure on soft tissues from altered kinematics, abnormal loads on the patella, a weak V M O muscle, increased Q-angle, patellar resurfacing, poor fixation, weak bone, increased strain in the 9 patella, soft tissue balancing (including lateral retinacular release), component design, and preoperative anxiety and depression (Barrack 2001, Brander 2003, D'Lima 2003, Hofmann 2003, Nizard 2005, Pagnano 2000, Parvizi 2005, Rhoads 1990, Sharkey 2002, Waters 2003, Wood 2002). A maltracking patella usually tends to travel too laterally in the groove, and the lateral side of the patella tilts downward (posteriorly) as the patella rides up the lateral condyle (Armstrong 2003). Maltracking can result in subluxation of the patella, during which the patella rides outside the femoral groove and almost dislocates. It can also result in frank dislocation of the patella. It is thought that poor patellar tracking increases the stresses on the patella because it results in decreased contact areas (Armstrong 2003). It is also commonly believed that most anterior knee pain and patellofemoral complications are either directly or indirectly related to high patellofemoral contact stresses (D'Lima 2003); reducing these contact stresses intra-operatively may therefore reduce the incidence of pain and complications following total knee arthroplasty. High stresses may result in fractures of the patellar bone and fractures and loosening of the patellar component (in resurfaced patellae) (Miller 2001a, Miller 2001b, Rand 2004). High contact stresses are also thought to be responsible for increased wear of the polyethylene components in resurfaced patellae (e.g. Miller 2001a, Miller 2001b). Studies on component positioning, patellar tracking and patellar loading are common because these parameters are quantifiable. In the ideal study, we would modify surgical variables (such as component placement) in a living human knee, wait for the knee to heal, have the patient walk around, and measure the effects of altering the surgical variables on the patient's pain and knee function. Of course, this sort of experimentation is not possible on living humans, therefore three branches of research have attempted to delve into patellofemoral issues: in vitro biomechanical models, clinical studies, and finite element (FE) or mathematical models. Mathematical models are currently oversimplified, and most FE studies focus on wear and fatigue of the tibiofemoral joint (D'Lima 2001a, Liau 2002b, Paganelli 1988, Sawatari 2005) or have only published validation results of their models (Godest 2002, Halloran 2005, Vil la 2004). To our knowledge, only 2 FE models have investigated the effect of component placement on patellofemoral mechanics (D'Lima 2003, Heegard 2001); the latter study did not validate the FE model with cadaveric testing. The following section details the relevant cadaveric and clinical studies; the results of the FE study by D'Lima are included in the section on cadaveric models. 10 1.4.1 Cadaveric Studies In vitro biomechanical studies are performed on cadaveric knee specimens and are usually models of dynamic, loaded knee flexion. Biomechanical investigations are particularly valuable because it is possible to dynamically measure the kinetics (patellar loading) and kinematics (patellar tracking) of the knee, or both simultaneously (Figure 1.8). These studies also allow researchers to compare the effects of changing several different surgical variables on the same knee. Obviously, cadaveric studies cannot make direct links between knee biomechanics and pain; however, they can suggest connections between the altered surgical variables and patellar tracking or loading, which are thought to be related to postoperative pain. A limitation of cadaveric studies is that the loads applied to the knee are not usually physiological- cadaveric knees cannot withstand the same loads as living legs. Cadaveric Studies Clinical Studies Figure 1.8: Two main branches of research on patellofemoral issues. 1.4.2 Clinical Studies Clinical studies of the postoperative patellofemoral joint involve inspecting the tracking of the patella on the femur using radiographs, computed tomography (CT) scans, or Magnetic Resonance Imaging (MRI) scans taken during the follow-up period after a T K A . The images of the knee joint are typically static and two-dimensional, and are taken while the knee is unloaded, often at only one angle. Clinical studies are useful to obtain information about postoperative patellar tracking over a population of T K A recipients; however, a two-dimensional view of the knee may be an inadequate representation of the biomechanics of the patellofemoral joint: the 11 tracking measurements depend on the technique used to obtain the view (Shih 2004), and it is difficult to obtain a consistent view (Bindelglass 1993). Also, clinical studies may be of limited use because the patient is not dynamically flexing and loading the knee joint; conclusions based on a static view of an unloaded knee (or a knee not loaded under body weight) may not be applicable to the condition of actual loaded gait with active quadriceps contraction. Patellar problems occur during loaded use of the knee, and are likely linked as much to the loads acting on the patella as to patellar tracking. Clinical studies do not investigate the links between loading and pain/complications. Although most clinical studies report the effects of altering various surgical variables (such as component placement) on patellar tracking parameters (such as patellar tilt, shift and rotation), some studies also attempt to correlate these variables with postoperative pain, knee function, and/or revision rates (Figure 1.8). 1.4.2.1 Pain, Complications and Tracking It is important to determine the connection between pain or complications and tracking; current surgical techniques assume that optimal intraoperative patellar tracking will result in reduced post-operative pain and complications. A special subset of clinical studies has attempted to directly link patellar tracking indices to reported postoperative pain (Barrack 2001, Bindelglass 1993, Gerber 1998, Shih 2004) and complications (Matsuda 2001, Shih 2004). The results of these studies are summarized in Table 1.1 and Table 1.2. Table 1.1: Summary of studies correlating pain and tracking. The flexion angle is the angle at which the images were acquired. All studies with resurfaced patellae investigated tracking of the patellar component (as opposed to the bone remnant). Resurfaced N Follow-Up Flexion Maltracking Patellae Angle Pain Barrack (2001) Both 102 5 years Unspec. No Bindelglass (1993) Yes 234 3.5 years 45° No Gerber (1998) No 30 2 years 20-30° Yes Shih (2004) No 227 8.5 years 45° No* There was no significant difference in knee scores; however, there was greater difficulty climbing stairs and rising from a chair (which might be related to pain). 12 Table 1.2: Summary of studies correlating complications and tracking. The study with resurfaced patellae investigated tracking of the patellar component (as opposed to the bone remnant). Resurfaced Patellae N Follow-Up Flexion Angle Maltracking Problems Matsuda (2001) 12/13 Yes 13 10 years 30° No Shih (2004) No 227 8.5 years 45° No** * There was no significant difference in function; however, there was greater difficulty climbing stairs and rising from a chair. Barrack measured postoperative tilt and shift in painful and control knees; they did not find a significant difference in these measures between the study and control groups. In addition, they did not find a correlation between knee pain and patellar shift or tilt. Bindelglass found that pain was not affected by patellar tilt or shift. In contrast to the results of Barrack and Bindelglass, Gerber found that knees with persisting pain or poor range of motion (ROM) showed lateral shift of more than 5mm or lateral tilt of more than 5°, whereas control knees did not. Shih studied the relationship between patellar tilt and pain/function (using the Knee Society Clinical Rating System) and found that abnormal postoperative tilt was not linked to pain scores. Matsuda did not discover a correlation between tracking parameters and clinical symptoms (measured using the Mean Knee Society score). Shih found that abnormal patellar tilt was not related to function (measured using the Knee Society Clinical Rating System); however, patients with abnormally tilting patellae had greater difficulty climbing stairs and rising from a chair. There does not seem to be a general consensus as to whether patellar tracking indices are correlated to postoperative pain or function. 1.5 The Effects of Surgical Variables Most existing clinical and cadaveric studies have focused on determining the effects of particular surgical variables on postoperative pain and complications, patellar tracking, or patellofemoral loading. The following sections discuss the existing clinical and cadaveric research on the following surgical factors: combined femoral and tibial component rotation, femoral component rotation, tibial component rotation, patellar resection angle, position of the patellar component, patellar thickness, lateral release, preoperative patellar tilt, and recovery time. 13 Each section describes the relevant clinical studies and cadaveric studies. The clinical studies focus mainly on pain and complications, component placement and patellar tracking. The cadaveric biomechanical investigations concentrate on component placement, patellar tracking, and patellar loading. A l l tracking studies detailed below discuss patellar tilt and shift; no studies have shown any correlation between component placement and patellar rotation (Armstrong 2003, Hsu 1996, Nagamine 1994, Rhoads 1990) or flexion (Hsu 1996), therefore patellar rotation and flexion are not discussed herein. 1.5.1 Combined Femoral and Tibial Component Rotation It has been shown that combined internal rotation of the femoral and tibial components is related to pain and complications (Barrack 2001, Berger 1998, Hofmann 2003). The details of these studies are summarized in Table 1.3. Table 1.3: Summary of clinical studies analyzing pain, component rotations and tracking. (F/T denotes femoral and tibial.) Resurfaced Patellae N Follow-Up Combined F/T Rotations-> Pain/Comps Individual F/T Rotations-> Pain/Comps Barrack (2001) Both 28 5 years Yes F No T Yes Berger (1998) Yes 50 Various Yes No Hofmann (2003) Yes 26 Prospective Yes Yes Barrack found a statistically significant difference in combined rotation between a study group of painful knees and a control group of pain-free knees. The study knees were rotated 4.7° internally on average and the controls were rotated 2.6° externally. Knees with combined component internal rotation were more than 5 times as likely to experience A K P as those with combined external rotation. Berger found that post-TKA knees with patellofemoral issues requiring revisions displayed combined internal rotation, whereas the control group knees (problem-free knees) showed combined neutral or external rotation. In the study group, the amount of excessive combined internal component rotation was directly proportional to the severity of the patellofemoral complication; as the rotation became more internal, the reported complications increased in seriousness, from lateral shifting and tilting to subluxation to prosthesis failure. Hofmann 14 discovered that 38% of knees requiring revision due to pain had clinically relevant internal malrotation of both the femoral and tibial components. No cadaveric studies have investigated combined rotation of the femoral and tibial components. 1.5.2 Femoral Component Rotation and Translation Although the placement of the femoral component has generally been shown to have an effect on the mechanics of the knee, the effect often depends on the research protocol or is interpreted differently between studies. 1.5.2.1 Clinical Studies: Pain and Complications The existing clinical studies have not reached a consensus on the effects of individual component rotations on pain and complications (Barrack 2001, Berger 1998, Hofmann 2003) (Table 1.3). Berger found no statistically significant difference between the individual femoral rotations for problematic and problem-free post-TKA knees. Hofmann determined that 25 out of 26 knees requiring revision due to pain had clinically relevant internal malrotation of the femur and/or tibia. This study concluded that individual component malrotations may lead to postoperative pain. The individual mean malrotations observed by Hofmann were very similar to the ranges reported by Berger. In contrast to the findings of Berger and Hofmann, Barrack compared femoral component rotations for painful study (AKP) and control knees and did not find a statistically significant difference between femoral component rotations for the study and control groups. Zihlmann (2005) conducted a literature review of studies focusing on the rotational malalignment of the femoral component. They stated that malalignment of a measurable degree occurs in approximately 10-30%) of patients; however, not all patients with noticeable malalignment report an unsatisfactory outcome. It should be noted that the definition of malalignment depends on surgical technique and the landmarks used to measure rotation. 1.5.2.2 Clinical Studies: Tracking Femoral component rotation was not linked to patellar shift (Akagi 1999, Kawano 2002, Matsuda 2001) (Table 1.4). Akagi and Kawano reported no correlation between femoral rotation and tilt, whereas Matsuda did find a correlation. 15 Table 1.4: Relationship between femoral component rotation and postoperative tracking. All studies investigated tracking of the patellar component (as opposed to the bone remnant). Resurfaced Patellae N Follow-Up Flexion Angle Femoral Rotation -> Tracking Akagi (1999) Yes 65 Unspec. 60° No Kawano (2002) Yes 62 1.5 years 60° No Matsuda (2001) 12/13 Yes 13 10 years 30° Yes (tilt) No (shift) 1.5.2.3 Cadaveric Studies: Tracking No consensus exists as to whether or not externally rotating the femur improves tracking. The cadaveric studies all agree that external rotation shifts the patella laterally (Anouchi 1993, Armstrong 2003, Miller 2001b, Rhoads 1990); although Miller stated that it merely moved with the femur and thus did not really affect tracking. Anouchi and Rhoads opined that this lateral shift is similar to the tracking in the natural knee, and is therefore a positive effect; Armstrong and Miller believed that the lateral shift is suboptimal and could result in subluxation or decreased contact area. The studies on the effects of femoral component rotation on patellar tilt are not in agreement: Rhoads found the effects of external rotation variable, although internal rotation tilted the patella medially. Miller (2001a) agreed that internal rotation tilts the patella medially but also found that external rotation resulted in lateral tilt (2001a, 2001b). Again, Rhoads stated that externally rotating produced optimal (and most similar to natural knee) kinematics. Miller thought that the resulting, lateral tilt would decrease the contact area and risk subluxation but would improve (reduce) the Q-angle and reduce the forces on the patella. The aforementioned conclusions regarding femoral component rotation and patellar tracking are in disagreement with most of the results of the clinical studies; the clinical studies found that femoral rotation was not linked to shift (Akagi, Kawano, Matsuda) or tilt (Akagi, Kawano). Only the clinical study by Matsuda found a correlation between tracking (tilt) and femoral component rotation. It is difficult to compare the results of clinical and cadaveric studies because the methodologies are vastly different. Armstrong and Rhoads both found that translating the femoral component medially or laterally resulted in a predictable, corresponding shift of the patella. Armstrong found that mediolateral 16 translation of the femoral component did not affect patellar tilt, whereas Rhoads found that medial displacement tilted the patella 3° medially between 0° and 60° flexion. This effect was not seen at larger flexion angles. Rhoads observed that lateral displacement of the femoral component resulted in variable changes in tilt. 1.5.2.4 Cadaveric Studies: Loading Controversy also exists as to the effects of femoral component rotation on patellofemoral loading. Anouchi (1993), D'Lima (2003) and Singerman (1997) concluded that external rotation reduces loading, and D'Lima, Miller (2001a) and Singerman agreed that internal rotation had negative effects on loading; however, in one study, Miller found that neutral rotation was optimal (2001a), and found no change in loading for various rotations in their subsequent study (2001b). 1.5.3 Tibial Component Rotation and Translation Tibial component rotation has been thought to relate to patellar tracking and loading because rotation of the tibia should result in the displacement of the tibial tubercle. This could in turn alter the Q-angle and change the direction of pull of the extensor mechanism. 1.5.3.1 Clinical Studies Similar to femoral component rotation, there is no consensus as to the effects of tibial component rotation on pain and complications (Barrack 2001, Berger 1998, Hofmann 2003) (Table 1.3). Berger found no statistically significant difference between the individual tibial rotations for problematic and problem-free post-TKA knees. Hofmann determined that 25 out of 26 knees requiring revision due to pain had clinically relevant internal malrotation of the femur and/or tibia. Barrack compared tibial component rotations for painful and control knees and found a significant difference in tibial component rotation between the study (AKP) and control groups; the tibial components of the study group were internally rotated whereas the components of the control group were externally rotated. Matsuda (2001) attempted to determine the relationship between tibial component rotation and tracking; they found no link between tibial rotation and shift at 30° yet they did determine a correlation between tibial rotation and tilt. 17 1.5.3.2 Cadaveric Studies The only cadaveric study on tibial rotation (Nagamine 1994) found that external malrotation shifted the patella medially; however, the effect of external rotation on shift was only statistically significant at 15° and the effect decreased at larger flexion angles. Tilt and rotation were not affected by external rotation of the tibial tray. Internal rotation did not affect patellar tracking. It is difficult to compare these results to the only other study on tibial rotation and tracking (Matsuda 2001) because it was a clinical study and tracking indices were only measured at one static angle. The clinical study by Matsuda (2001) found no link between tibial rotation and shift at 30°. This agrees with the results of Nagamine because shift was only found to be significant at 15° (and only for external rotation). The correlation between tibial rotation and tilt determined by Matsuda was not supported by the findings of Nagamine. 1.5.4 Patellar Resection Angle The effect of the patellar resection angle has only been studied in the clinical setting; to our knowledge, no cadaveric studies have investigated the effects of this- parameter on patellar mechanics. Chan (1999) and Kawano (2002) agreed that patellae with a more medial remnant tilted medially, however the results of Gomes (1988) differed (Table 1.5). This may be because Gomes did not separate out those knees which had lateral releases. Gomes and Kawano did not observe changes in shift due to different resection angles. Table 1.5: Relationship between patellar resection angle and tracking indices. All studies investigated tracking of the patellar component (as opposed to the bone remnant). N Follow-Up Flexion Angle Resection Angle (thicker med. side) -» Med. Tilt ^ Shift Chan (1999) 38 Unspec. 30° Yes --Gomes (1988) 40 6 weeks 30° No No Kawano (2002) 62 1.5 years 60° Yes No 1.5.5 Position of the Patellar Component 1.5.5.1 Clinical Studies Gomes (1988) and Kawano (2002) agreed that medialization of the patellar component increased lateral tilt; however, Nelissen (1995) found no correlation between medial placement and tilt. 18 These clinical studies did not agree on whether medialization shifted the button laterally during flexion (Gomes) or not (Kawano, Nelissen) (Table 1.6). Table 1.6: Relationship between patellar medialization and tracking indices. All studies investigated tracking of the patellar component (as opposed to the bone remnant). N Follow-Up Flexion Angle Patellar Component Medialization Lat. Tilt Lat. Shift Gomes (1988) 40 6 weeks 30° Yes Yes Kawano (2002) 62 1.5 years 60° Yes No Nelissen (1995) 72 2 years 45° No No 1.5.5.2 Cadaveric Studies: Tracking Both Miller (2001b) and Yoshii (1992) found that medializing the patellar implant lateralized the bony remnant. Lee (1999), Miller and Yoshii discovered that medializing the patellar implant also resulted in more lateral tilt, although Lee found that the changes were not statistically significant, and Yoshii found no significant changes in tilt at lower flexion angles (<45°). Lee did not observe any statistically significant changes in tilt when the patellar button was implanted proximally or distally relative to neutral placement. 1.5.5.3 Cadaveric Studies: Loading Patellar medialization was shown to decrease the lateral patellar shear while increasing the medial shear (D'Lima 2003, Lee 1999), and was also shown to decrease the mediolateral component of the patellofemoral loads (Miller 2001b). Lee found distally positioned patellar implants resulted in decreased patellar component loading at high knee flexion angles; however, a centralized component yielded the most evenly balanced patellar facet contact pressures. 1.5.6 Patellar Thickness 1.5.6.1 Clinical Studies No consensus exists on the effect of patellar thickness on tilt; Laughlin (1996) discovered that thicker patellae tilt more laterally, however Kawano (2002) found no effect (Table 1.7). In the only clinical study relating patellar thickness to shift, Kawano found that changes in thickness did not affect shift. 19 Table 1.7: Relationship between increased patellar thickness and tracking indices. - Resurfaced Patellae N Follow-Up Flexion Angle Thicker Patella Tilt Shift Kawano (2002) Yes 62 1.5 years 60° No No Laughlin(1996) Yes 89 Various 45° Lateral Tilt ~ 1.5.6.2 Cadaveric Studies: Tracking Hsu (1996) determined that increasing patellar thickness by 2 mm increased lateral tilt by approximately 2° and decreasing the thickness decreased tilt by the same amount. This result was in agreement with the clinical findings of Laughlin but not Kawano (who found no effect). Hsu showed that the thick patella shifts medially in comparison to the thin patella; however, the clinical study by Kawano found no effect on shift. 1.5.6.3 Cadaveric Studies: Loading Although Hsu (1996) and Star (1996) found that thinner patellae experienced decreased normal loads, Oishi (1996) found no change in the normal loads for 3 different patellar thicknesses. Instead, they found an increase in total, mediolateral and superoinferior shear forces with increased patellar thickness. 1.5.7 Lateral Release A lateral release is a surgical procedure during which the tight retinaculum (fibrous support) on the lateral side of the patella is cut vertically. This release is usually approximately an inch in length and is typically done to restore normal patellar tracking when the lateral retinaculum is too tight and is causing the patella to tilt abnormally. None of the clinical studies found any relationship between lateral releases and tracking at the flexion angle studied (Bindelglass 1993, Chan 1999, Gomes 1988, Kawano 2002, Laughlin 1996). This may be because lateral releases were only performed when required and were not studied as a separate, independent variable. In a cadaveric study, Singerman (1997) did not observe any differences in patellar loads following a lateral release. 1.5.7.1 Preoperative Tilt Preoperative tilt was found to be related to postoperative pain (Gerber 1998) and function (Shih 2004). Shih identified preoperative patellar maltracking (i.e. >5° tilt or >5 mm shift) as the only independent factor associated with an abnormal postoperative patellofemoral joint. No 20 consensus has been reached on the relationship between preoperative and postoperative tilt; in clinical studies, Bindelglass (1993) and Chan (1999) found a correlation, whereas and Kawano (2003) did not (Table 1.8). Chan found that every 2° increase in preoperative tilt increased postoperative tilt by approximately 1°. Table 1.8: Relationship between preoperative patellar tilt and postoperative tilt. All studies with resurfaced patellae investigated tracking of the patellar component (as opposed to the bone remnant). Resurfaced Patellae N Follow-Up Flexion Angle Preoperative Tilt-> Postoperative Tilt Bindelglass (1993) Yes 234 3.5 years 45° Yes Chan (1999) Both 38 Unspec. 30° Yes Gomes (1988) Yes 40 6 weeks 30° No (normal button) Yes (biconvex button) Kawano (2002) Yes 62 1.5 years 60° No 1.5.8 Recovery Time Using radiographs, Shih (2004) monitored patellar tilt in nonresurfaced patellae following T K A and showed that patellae tilt laterally as they deteriorate "slowly but relentlessly". Shih determined that 95% of postoperative knees tracked centrally immediately following surgery. By 6 weeks, 7% of those centrally tracking knees had tilted or shifted, and tracking continued to deteriorate for those knees. At the last follow-up (mean 8.5 years), only 59% of the patellae still tracked centrally. Laughlin (1996) observed deterioration in the tracking of resurfaced patellae following T K A ; similar to the findings of Shih, they observed that patellae tilted more laterally with time. Between the 3 week visit and the most recent follow-up (mean 3 years), the number of patellae tilting more than 5° laterally increased from 32% to 46%. It was also observed that medial tilting patellae moved towards neutral. The gradually increasing lateral tilt could be due to stretching or relaxation of the medial scar following surgery (Shih 2004). 1.5.9 Other Variables Other researchers have studied the effects of implant design (Matsuda 1995a, Szivek 1996), kneeling forces (Laman 2004), Q-angle (Elias 2004a) and loading level (Kessler 2004). Patellar 21 resurfacing substantially reduces the contact area and increases the contact stresses (Nizard 2005, Matsuda 1995a, Parvizi 2005). 1.6 Current Research Gaps and Limitations Based on the clinical and cadaveric studies summarized above, we believe that component placement (femoral, tibial and patellar) can have a significant effect on postoperative success and on the mechanics of the patellofemoral joint (kinematics and kinetics). In order to optimize the placement of each T K A component, it is necessary to fully characterize the effects of changing component placement intraoperatively on postoperative patellofemoral mechanics. This includes the effects of positioning the femoral, tibial and patellar components. The existing studies described above have investigated individual characteristics in a simulated postoperative configuration; however, these studies have been narrow in scope and their results have often been inconclusive or lack a consensus. We would like to compare our findings to the results of studies on individual surgical variables to determine if our results fit within the range reported by others. 1.6.1 Relationship between Intraoperative and Postoperative Biomechanics Currently, surgeons attempt to optimize postoperative knee function by observing and manipulating the knee intraoperatively. Although they assume that intraoperative and postoperative knee mechanics are closely related, a relationship has not yet been established. Bindelglass (1993) questioned whether the static, intraoperative tests used to improve patellar tracking are effective in optimizing dynamic, postoperative function. It would therefore be valuable to identify how intraoperative modifications alter postoperative performance of the knee. To our knowledge, no studies exist which have simulated the intraoperative setting. There are therefore no studies that have compared the effects of altering surgical variables on tracking and loading in both simulated intraoperative and postoperative loading configurations. 1.6.2 Patellar Component Placement Less clinical emphasis has been placed on the patella than the femur and tibia, despite it being the source of many problems. Many orthopaedic surgeons use computer assisted surgery (CAS) systems to guide their bone cuts and the placement of the tibial and femoral components; however, none of the CAS systems currently on the market include directions or guidelines to assist the surgeon in the preparation and installation of the patellar component. This may be 22 because researchers have not fully investigated the proper placement of the patella or the effects of malpositioning the patellar component (especially with regards to the patellar resection angle); it is unclear what the ideal placement of the component should be. Also, many surgeons, especially in Europe, choose not to resurface the patella. In personal communications with surgeons, we have found that most surgeons do not use the patellar cutting jigs supplied by the component manufacturers; they may find them awkward, cumbersome, time-consuming or even entirely unnecessary. Instead, surgeons often cut the patella freehand, which could result in large errors and considerable variability in placement or bone cut angle. It is currently unclear i f the placement of the patellar component is clinically important. If the positioning of the patellar component has a significant effect on the biomechanics of the knee joint, we should guide the development of CAS systems or more appropriate cutting jigs to promote optimal patellar component positioning. There is a particular paucity of knowledge regarding the optimal patellar resection angle. Although some clinical studies have characterized the effect of altering the patellar bone cut angle (Chan 1999, Gomes 1988, Kawano 2002, Pagnano 2000), this surgical factor has not been investigated in vitro. The advantage of studying the resection angle on cadaver knees is that several angles can be compared, and the resulting changes in loading as well as tracking can be measured. 1.6.3 Comparison of Wide Range of Variables No studies have compared and ranked the effects of modifying a wide range of surgical variables on patellar tracking and loading. A ranking of variables would give surgeons guidance with regard to where to focus their efforts and attention. The majority of cadaveric studies have focused on femoral component placement (Anouchi 1993, Armstrong 2003, D'Lima 2003, Kessler 2004, Miller 2001a, Miller 2001b, Rhoads 1990, Singerman 1997), despite clinical evidence that tibial component rotation may be more strongly related to postoperative pain or poor tracking than femoral rotation (Barrack 2001). Only one cadaveric study has investigated the effects of tibial component rotation on tracking (Nagamine 1994), and no studies have analyzed the consequences on patellar loads. It would be useful to 23 compare the effects of similar malrotations of the femoral and tibial components on both tracking and loading. Several biomechanical studies have analyzed the effects of patellar medialization (D'Lima 2003, Lee 1999, Miller 2001b, Yoshii 1992) and patellar thickness (Hsu 1996, Oishi 1996, Star 1996); however, no studies have compared the effects of modifying both variables on the same patella. A comparison of the effects of altering mediolateral placement, thickness, and bone cut angle would be valuable in developing a CAS approach to the patella and in improving surgical technique. 1.6.4 Relationship Between Tracking and Loading Although Anouchi (1993), Miller (2001a, 2001b), and Hsu (1996) measured both tracking and loading in their studies, none of these studies determined if kinematics and kinetics are related. Hsu assessed forces and kinematics in separate tests. It is unknown whether or not Anouchi measured kinematics and pressure simultaneously; however, I am inclined to believe that they did not because their testing protocol took 2-3 days. 1.6.5 Dynamic Measurement of Loads Of the studies that have investigated the effects of component position on patellar loads, most of these used a load cell (D'Lima 2003, Hsu 1996, Miller 2001a, Miller 2001b, Oishi 1996, Singerman 1997, Star 1996). Load cells are unable to measure contact area or pressure, and the weight and size of the transducer (up to approximately 400g, measuring 5 cm x 13 cm, according to Miller (2001b)) likely influence the motion of the patella and the loads acting on it. Other studies have measured contact pressures using Fuji pressure sensor film (Fuji Photo Film Company, Tokyo, Japan) (Anouchi 1993, Lee 1999); however, Fuji film is inappropriate for use in dynamic studies of the patellofemoral joint because it can only take a static snapshot of the loads at one instant in time. To dynamically measure patellar loads without influencing the mechanics of the knee joint, we chose to use Tekscan pressure sensors (Tekscan, Inc., South Boston, M A ) . The accuracy and appropriateness of the calibration routines supplied with the Tekscan system software were unknown. To use Tekscan pressure sensors in our study, it was necessary to investigate the accuracy of the calibration algorithms. 24 1.7 Research Questions Component placement in total knee arthroplasty has been shown to affect patellofemoral mechanics and patient outcome. The primary hypothesis motivating our research is that if a surgeon could optimize component placement during the surgery, this would lead to an improved clinical outcome and patient satisfaction. The first steps in optimizing component placement are to know the relative effects of each surgeon-controlled variable and to determine whether optimizing component placement intraoperatively will have the desired effect postoperatively. The objectives of this study were therefore to address the following research questions: 1. What are the effects of varying the orientations of the femoral and tibial components on patellar tracking and loading? 2. What are the effects of modifying the patellar component placement (mediolateral position, thickness of patellar construct, patellar bone cut angle) on patellar tracking and loading? 3. How do the effects of altering the position of the femoral, tibial and patellar components compare relative to each other (ranking) in terms of: a) kinematics, and b) kinetics. 4. How do the aforementioned effects compare in intraoperative and postoperative simulations? 5. Does a relationship exist between patellar kinetics and kinematics, and i f so, what is the relationship? 6. What is the accuracy of the calibration algorithms supplied in the Tekscan system software, and do user-defined calibration routines provide more accurately calibrated force output? 25 2 Materials and Methods 2.1 Introduction The primary purpose of this study was to quantify the effects of altering surgeon-controlled variables related to total knee arthroplasty (TKA) implant positioning on patellofemoral tracking and loading. The variables we selected represented either typical surgical errors or practical displacements and rotations that a surgeon might make to improve knee mechanics. To that end, we developed adjustable femoral, tibial and patellar components and used them in the testing of 8 cadaveric knee specimens. We altered the standard surgical technique as necessary to install the modified components in the knee specimens. We also investigated whether or not a simulated intraoperative assessment of patellar tracking and loading appropriately represented the postoperative weight-bearing situation. To simulate the intraoperative and postoperative loading conditions, we tested the specimens in two different experimental rigs. The kinematics of the patella were measured using optoelectronic sensors and the patellofemoral loads were recorded using pressure sensors. Figure 2.1 shows a schematic of a knee specimen in the two loading rigs with the attached optoelectronic and pressure sensors. 26 Horizontal Testing Rig Vertical Testing Rig Figure 2.1: Schematic of knee specimens in horizontal and vertical testing rigs, with kinematic and kinetic measurement systems. The horizontal rig emulated passive, intraoperative flexion, and the vertical rig emulated postoperative loading. 27 2.2 Kinematics The kinematics of the femur, tibia and patella were measured using an Optotrak 3020 optoelectronic tracking system (Northern Digital, Inc., Waterloo, ON). This system uses active markers with light emitting diodes. When the markers are 2.25 m from the capture system, the RMS accuracy of this system is 0.1 mm for the x and y coordinates and 0.15 mm for the z coordinate (NDI 2006). We mounted optoelectronic marker arrays with four infrared emitting diodes (IREDs) per array on the femur, patella and tibia to track their motion throughout the flexion-extension cycle of the knee. To measure the kinematics of the femur, patella and tibia, it was necessary to define the anatomical axes for each bone. Pairs of #4 x %" woodscrews were inserted into the femur and tibia and sewing pins were inserted into the patella to serve as landmarks during digitization (described in further detail in sections 2.5.1.4, 2.5.2.2, and 2.5.3.2). Prior to using the Optotrak digitizing pointer to digitize the landmark screws, we calibrated the pointer using an aluminum plate that had a tiny divot scored into it. We used the standard pointer calibration and verification procedure (NDI 1994), only accepting the calibration i f errors were less than 0.5 mm. As expected, the highest errors were in the z-direction, i.e. distance from the camera. We captured the static position of all three marker arrays to determine their locations in the global frame of reference. We then digitized the landmark screws that were inserted to represent the axes of the bones. Motion data were initially collected at 10 Hz (specimens 1 - 4) because this was the sampling rate used in previous knee studies at U B C ; we later increased the sampling frequency to 20 Hz (specimens 5 - 8) to improve resolution. The motions we investigated included patellar tilt about the long (superior-inferior) axis of the patella, patellar shift along the transepicondylar axis of the femur, and patellar rotation about the floating axis perpendicular to the two aforementioned axes. The digitization process, the definition of the bone axes, and the construction of the coordinate frames are explained in detail in section 2.9.1. Following the completion of a day of testing, the calibration and digitization processes were repeated; these repetitions were a precaution in case an error occurred during the first digitization 28 or in case any of the marker arrays were jarred during testing. We did not begin to perform 2 digitizations until specimen 5. 2.3 Force Measurement 2.3.1 Selection of a Measurement System To measure loads in the patellofemoral joint, it is desirable to use a technology that is accurate and reliable and permits both static' and dynamic measurements. The transducer must be usable in a wet environment and must not be damaged by the presence of the fatty fluids in the knee joint. Because peak loads in the patellofemoral joint may be in the range of 3-33 MPa (Ahmed 1983, Csintalan 2002, Fuchs 2000, Fuchs 2002, Lee 1997, Lee 1999, Lee 2001, Matsuda 1997, Matsuda 2000, Whiteside 2003), the transducer must be durable. Patellofemoral loads can be measured by implanting a load cell directly into the patellar component or by using thin pressure sensors such as Fuji film, Novel sensors, or Tekscan sensors. The main disadvantage of using uniaxial load cells (Hsu 1996, Star 1996), three degree of freedom load cells (D'Lima 2003, Oishi 1996) or six degree of freedom load cells (Miller 2001a, Miller 2001b, Singerman 1997) is that the weight and size of the transducer could be expected to influence the motion of the patella and thus the loads acting on it. This effect is especially noticeable at low flexion angles where the balancing quadriceps loads are small (Miller 2001b). Miller (2001b) reported using a load cell measuring 5 cm x 13 cm, with a mass of approximately 400g. Also, load cells cannot provide a spatial mapping of the loads and cannot calculate contact pressures or area. Load cells cannot reflect actual load distribution (such as mediolateral balance) or show bimodal contact patterns; they can only report the location of the centre of pressure. Hsu (1996) attempted to calculate contact areas using measured kinematics and digitized surface geometries; however, these indirect methods have not been validated for accuracy or repeatability. It was originally hoped that the load measurement system selected for this study would also be used in subsequent clinical studies and would possibly be incorporated into a computer-assisted surgical system to assist with component placement during TKAs. As such, we did not believe that load cells would be practical intraoperatively; to implant the transducer in the knee during surgery, the surgeon would be required to drill holes into the patellar bone, which would reduce 29 its strength and integrity. Also, load cells cannot be used to measure loads applied to non-resurfaced patellae. Fuji pressure sensor film (Fuji Photo Film Company, Tokyo, Japan) is inappropriate for use in dynamic studies of the patellofemoral joint because it can only take a static snapshot of the loads at one instant in time (e.g. Anouchi 1993). During cyclic loading, Fuji film shows the peak pressure over the entire flexion cycle. In addition, the thickness of Fuji film (0.3 mm) has been shown to cause changes in joint mechanics: a finite element model of the natural knee revealed that the additional film thickness results in the overestimation of contact pressures by 10-26% (Wu 1998). Using a finite element model of the artificial tibiofemoral joint, Liau (2002a) determined that joint contact pressures are underestimated by 8-14% upon introduction of the film. Harris (1999) and Matsuda (1995b) found that Fuji film underestimates contact areas by as much as 36%., Novel capacitive sensors (Novel, Munich, Germany) are flexible and elastic and can conform to three-dimensional surfaces. They were inappropriate for this study because they are thick (between 0.6 and 1 mm), have inadequate pressure ranges for the postoperative, weight-bearing setup (2 MPa maximum), and have poor spatial resolution (14 sensing elements per cm2) (Mortier 2004). We chose to use the Tekscan pressure measurement system for this study (Tekscan, Inc., South Boston, M A ) . The sensors calculate pressure using dynamic measurements of force and area. They are extremely thin (0.10 mm thick), have good resolution (62 sensing elements per cm2), and have acceptable measurement accuracy for force (6.5±4.4% RMS error, Wilson 2003) and contact area (Harris 1999, Matsuda 1995b). Tekscan sensors are disposable and are available with a wide variety of pressure ranges, ranging up to 174 MPa. Although Tekscan sensors have been shown to display considerable hysteresis and drift (Otto 1999), our loading application is not particularly subject to these effects. The load on the sensor shifts spatially and loading is typically discrete rather than ramped. The main disadvantage of Tekscan sensors is that they can only measure contact (normal) forces and do not account for shear loads; however, in their assessment of patellofemoral loads with an implanted force transducer, Singerman (1997) determined that the resultant contact force is nearly perpendicular 30 to the contact surface and changes very little in orientation. Tekscan sensors are not deformable or stretchable in-plane; therefore, folds are created when the sensors are forced to conform to three-dimensional curved surfaces such as patellar implants (see section 2.7.1). 2.3.2 Description of the Tekscan I-Scan Pressure Measurement System We measured the forces on the patellar component using a Tekscan I-Scan pressure measurement system with model #5051 sensors. A Tekscan sensor is a thin, flexible printed circuit which functions as a variable resistor. Rows of conductive ink are printed on one sheet of Mylar, and columns of ink are printed on a second sheet. When the rows and columns are overlaid, the resultant grid contains individual pressure sensing locations. These active sensing areas, where the rows and columns meet, are called sensels. Each sensel acts as a resistor; when a force is applied to the sensel, the distance between the row and column decreases and the resistance at that sensel decreases. The accompanying hardware scans the sensor to detect the resulting increase in current flow through the sensor. The current flow is converted to a digital raw output value ranging between 0 and 255. The Tekscan I-Scan system includes sensor-specific software to monitor data real-time and to analyze it post-testing. Each Tekscan pressure sensor is divided into two portions: a sensing matrix and a tab extending from the sensing matrix (Figure 2.2). The #5051 sensor has a square sensing area of 3136 mm 2 containing 1936 sensels arranged in 44 rows and 44 columns. The sensor resolution is thus 62 2 2 sensels per cm . Each sensel has an area of 1.61 mm , and the sensels are spaced 1.27 mm (0.05") apart. The sensor tab contains electrical leads and connections, and inserts into an electrical handle which connects the sensor to a personal computer (PC). 31 Overall WidlhOW) Maxttix Width (&1W) Overall Length (L) Figure 2.2: Tekscan pressure sensor, model #5051 2.3.3 Conditioning and Calibration We used two new I-Scan pressure sensors (model #5051) for each knee specimen: a 1200 psi (8.3 MPa) sensor for tests in the horizontal rig and a 2500 psi (17.2 MPa) sensor for tests in the vertical rig. The sensors were subjected to larger loads in the vertical rig which necessitated the larger range; the loads in the horizontal rig were much smaller and we were able to use a lower-pressure sensor to obtain better sensitivity. We did not begin using the more sensitive 1200 psi sensors in the horizontal rig until specimen 5; previous tests in the horizontal rig were performed using the 2500 psi sensors because we initially expected to be able to use the same sensor for both setups. I conditioned and calibrated each sensor using a materials testing machine (Instron 8874, Canton, MA) . These stages are described below in further detail. To simulate the loading in the 32 patellofemoral joint of a prosthetic knee, sensors were compressed between a flat polyethylene disk (D = 3.82 cm) and a larger aluminum plate (Figure 2.3). Sensors were coated with K - Y lubricant (Johnson & Johnson, Montreal, Canada) to reduce shear loads at the sensor surfaces. F v Instron Load Cell Metal Plate Tekscan Sensor Metal Plate PE Disk Figure 2.3: Tekscan sensor calibration apparatus Prior to calibration, I conditioned each sensor four times at a load corresponding to 120% of the maximum expected pressure, as recommended by the manufacturer. I increased the force steadily over 10s, held it for 5s, and decreased it to zero over 10s. Sensors remained unloaded for 120s between load applications. I used the same protocol throughout conditioning and calibration. I applied loads of 10.0 kN and 15.0 kN to the 1200 psi and 2500 psi sensors, respectively; these loads corresponded to pressures of 8.7 MPa and 13.1 MPa at the surface of I calibrated the sensors externally to the Tekscan software using a five-point, cubic polynomial. The development of the calibration polynomial is described in further detail in Chapter 3. To calibrate the 1200 psi sensors, I loaded them at five evenly-spaced loads between 0 and 8.3 kN. The 2500 psi sensors were loaded over a range of 12.5 kN. Figure 2.4 shows a typical calibration curve. I saved the coefficients of the calibration curve and then applied them to the raw data during data processing. the PE disk. 33 14 Raw Output Figure 2.4: Typical cubic polynomial calibration curve for Tekscan pressure sensor. Raw output is the sum of the raw force across all loaded sensels. 2.3.4 Digitizing the Patellar Circumference and Landmarks With the Tekscan sensor adhered securely to the patellar implant (described further in section 2.7.1), we recorded the medial, lateral, superior and inferior poles of the patella by pressing down at the specified locations. The medial and lateral poles of the implant were digitized by pressing a fingernail to the sensor at the edge of the patella, along the mediolateral axis of the implant. This axis was defined by the two pins protruding from the anterior surface, as described later in section 2.5.3.2. The superior and inferior poles were recorded by pressing on the edge of the implant, perpendicular to the mediolateral axis. This enabled us to later discern whether forces occurred on the medial or lateral sides of the patella. The circumference of the patella was digitized by running a finger around the rim of the patellar implant (Figure 2.5). This allowed us to discriminate between force readings located on the patella and extraneous readings occurring outside the patellar circumference. 34 Figure 2.5: Tekscan software showing digitized circumference of patella 2.3.5 Other Details A l l data was collected at 10 Hz (specimens 1 - 4) or 20 Hz (specimens 5 - 8). The Tekscan handle and sensor tab were wrapped in a plastic bag to protect them from exposure to fluids. 2.4 Surgical Variables To be able to simulate variations in component positioning and rotation and to emulate incorrect bone cuts, we modified trial Zimmer NexGen Legacy Posterior Stabilized (LPS) knee components (Zimmer, Inc., Warsaw, Indiana). The femoral, patellar and tibial components were altered to allow for the following modifications in component placement: Femur: 5° internal and external rotation in addition to neutral rotation. Neutral rotation was defined as 3° externally rotated with respect to the posterior condylar line, approximately in line with the transepicondylar (TE) axis (see section 2.5.1.4). This placement corresponded to the position typically selected intraoperatively. Tibia: 5° internal and external rotation in addition to neutral rotation. We defined neutral rotation as the position in which a central mark on the anterior surface of the baseplate was aligned with the mid-to-medial third of the tibial tubercle. This placement corresponded to the position typically selected intraoperatively. Patella: 4 resection (bone cut) angles* (0.0°, 7.5° medial, 7.5° lateral and 15° lateral), 3 positions (0.0, 2.5 and 5.0 mm medial), and 2 thicknesses (original and +3 mm) (Figure 2.6). Neutral 35 placement of the patellar component corresponded to the position which would maximize and centralize coverage of the bony surface. The 0° bone cut angle was achieved by ensuring equal bone thickness on the medial and lateral sides, and the original patellar thickness was equivalent to the measured preoperative thickness. Resection Angle Mediolateral Position Thickness Figure 2.6: Modifications to the placement of the patellar implant: resection angle, mediolateral position, and additional thickness. * Note that the bone cut angle denotes that an additional thickness was added to the stated side. The following sections explain the rationale behind the specific component positions that we tested. 2.4.1 Femoral Rotation We tested ±5° of femoral component rotation in addition to neutral placement because these rotations were surgically practical, represented typical choices a surgeon might make (intentional choice or unintentional malrotation), and have been tested in other studies. These rotations also reflected the possibility that a surgeon may choose to intentionally externally rotate the femoral component to improve tracking. Table 2.1 shows the angular rotations that have been studied in vitro by other researchers. We selected a range of ±5° rotation to enable comparisons between biomechanical studies performed by Anouchi (1993), Armstrong (2003), Kessler (2004), and Miller (2001a, 2001b). 36 In their study on 14 knees with anterior knee pain, Barrack (2001) found that symptomatic knees had femoral component rotations ranging between 5° internal and 7° external (with respect to the posterior condyles). Berger (1998) found that problematic knees were rotated between 2-8° externally (with respect to the TE axis). Our range of ±5° rotation therefore bracketed most of the surgical range of rotational error. Larger rotations would require the removal of even greater amounts of bone, and the amount of bone removed to allow 5° rotations was already quite substantial, though it did leave the ligament attachments intact. Table 2.1: Femoral rotations studied in vitro by other researchers (E denotes external rotation). Anouchi Armstrong D'Lima Kessler Miller Miller Rhoads Singerman (1993) (2003) (2003) (2004) (2001a) (2001b) (1990) (1997) 0° 0° 0° 0° 0° 0° 0° 0° ±5° 5°E ±3° ±5° ±2.5 ±5° 2.5° E ±10° ±10° ±10° 5°E * Note that Anouchi, Miller (2001b) and Singerman measured rotation relative to the posterior condyles. Miller (2001a) measured rotation with respect to both the TE and posterior condylar axes. Armstrong and Kessler measured rotation about the femoral shaft but did not state how they defined neutral rotation. D'Lima and Rhoads did not define neutral rotation. 2.4.2 Tibial Rotation We studied 5° internal and external rotations of the tibial component in addition to neutral placement because these rotations were surgically practical and represented typical choices a surgeon might make. The study by Nagamine (1994) is the only one to have investigated tibial rotation in vivo; they analyzed the effects of ±15° tibial component rotation on tracking. Barrack (2001) found that symptomatic knees had tibial component rotations ranging between 15° internal and 6° external. Our range of ±5° rotation therefore bracketed almost half of Barrack's surgical range of rotational error. By changing the tibial component rotation in 5° increments, we also hoped to compare similar rotations of the femoral component and tibial component. The other rationale for rotating ±5° was that we wanted the combined femoral and tibial rotation to bracket part of the range that Barrack found for symptomatic knees (17° internal to 4° external); we had initially planned to test combined tibiofemoral rotation in addition to individual rotations, but we later reduced the number of surgical variables in our final testing protocol. 37 2.4.3 Patellar Resection Angle No in vitro, biomechanical studies to date have studied the effect of different bone cut (resection) angles on patellar kinematics or kinetics. Pagnano (2000) defined asymmetric resection as a discrepancy in thickness greater than 2 mm between the medial and lateral facets of the patella. In personal communications with orthopaedic surgeons, we determined that an angle of 5° was not a large enough resection error to represent typical surgical variability, and that surgeons typically over-resect the medial side of the patella. We therefore chose to study 7.5° medial, 7.5° lateral and 15° lateral bone cut angles. 2.4.4 Mediolateral Patellar Position Miller (2001b) studied the effects of medializing the patellar component by 3.75 mm (Table 2.2); they stated that this placement replicates the average location of the patellar ridge and offers more balanced coverage of the resected surface. D'Lima (2003) analyzed the effects of 2 mm medial and lateral translations of the patellar component, and Lee (1999) studied 5 mm medialization and lateralization of the patellar component. Yoshii (1992) investigated 10 mm medialization of the patella. We investigated 2.5 mm and 5 mm of medialization on the advice of our consulting surgeons because a displacement greater than 5 mm seemed extreme and unrealistic. Table 2.2: Mediolateral patellar positions investigated in /// vitro studies D'Lima (2003) Lee (1999) Miller (2001b) Yoshii (1992) 0 mm 2 mm medial 2 mm lateral 0 mm 5 mm medial 5 mm lateral 0 mm 3.75 mm medial 0 mm 10 mm medial 2.4.5 Patellar Thickness Hsu (1996) and Star (1996) measured the effects of adding and removing 2 mm of thickness to the patellar construct (Table 2.3). Oishi (1996) analyzed the consequences of adding 2 mm and 4 mm of patellar thickness. We tested the addition of 3 mm of patellar thickness; based on these previous studies, 3 mm seemed to be a large enough change to observe a difference in tracking or loading, i f one existed. We could not subtract patellar thickness using our modified patellar components, described below in section 2.5.3.1. 38 Table 2.3: Patellar thicknesses studied in vitro Hsu (1996) Oishi (1996) Star (1996) 0 mm 0 mm 0 mm ± 2 mm + 2 mm ± 2 mm + 4 mm 2.5 Modified Components and Surgical Procedure The following three sections detail the design and development of the femoral, tibial and patellar adjustable components7. Each section is followed by a description of the modified surgical procedure needed to implant the components in the specimens. Additional surgical steps required to prepare the specimens for testing are also outlined below. A l l surgeries were performed by Dr. Jerome Tonetti, an experienced orthopaedic surgeon, using the instrumentation for Zimmer NexGen LPS total knee arthroplasty. 2.5.1 Femoral Rotation 2.5.1.1 Development of the Modified Femoral Component To generate reproducible rotations of the femoral component with respect to the femoral bone, I designed an adjustable femoral component. The 2 main functional requirements which the design was required to fulfill were: 1) Constrain the component in 6 degrees of freedom relative to the bone. 2) Be able to mount the component at 3 different rotation angles (0°, ±5°). 2.5.1.2 Existing Designs Several other researchers have modified the rotation of femoral components using various methods. Using a low melting-point alloy, Anouchi (1993) modified the internal dimensions of two trial femoral components so that the components themselves were wedged at ± 5° relative to the posterior condyles. They did not state how the components were affixed to the bone such that they could be replaced with other neutral or wedged components. Armstrong (2003) affixed a rod to the femoral component and allowed this rod to rotate in an intramedullary sleeve which was cemented into the femoral shaft. The rod/component assembly could be locked to the intramedullary sleeve at the proximal end. 1 I designed the modified femoral component and machined an initial prototype. Based on primary tests of the prototype in a pilot specimen, I modified the design of the component and had the final version machined and constructed by a professional machinist. Dr. Carolyn Anglin designed and built the modified tibial component and a prototype version of the modified patella component. The final modified patellar component was constructed by a machinist. 39 Miller (2001a) achieved femoral rotation by attaching a special device to the proximal femur that allowed the component to rotate about the long axis of the femur without opening the joint capsule. It was constructed using a precision worm gear drive, bearings, and an intramedullary shaft that extended out of the proximal end of the femur. In another study, Miller (2001b) used a pair of plates with matching holes to change and set femoral rotation. These plates were attached to the femur and the femoral component, and rotation was achieved by aligning different pairs of holes. In their study comparing 2 different T K A knee designs, Singerman (1997) used a different method with each knee design to achieve and measure femoral component rotation. For the Insall/Burstein Constrained Condylar Knee (CCK, Zimmer, Warsaw, Indiana), the femoral component was rotated about a cemented intramedullary stem. The change in angle was measured visually by comparing the plane of the posterior condyles of the implant to a rod marking the previous locations of the natural posterior condyles. The accuracy of this method was estimated to be only ±2°. For the Miller/Galante II knee (MG II, Zimmer, Warsaw, Indiana), component rotation was set using 10° plastic shims that were held in place with bone pins. It is not clear how the component was held in place during tests. Rhoads (1990) did not explain how they achieved femoral rotations in their study. Kessler (2004) used a Stryker Navigation System to accurately rotate the femoral component but did not specify how the component was attached to the bone. The modified component used by Armstrong (2003) seemed to be a very practical starting point for our design because it would provide a solid connection between the bone and the component despite the minimal remaining bone stock. Their design seemed similar to that used by Singerman (1997) for the Zimmer C C K knee. The wedge designs used by Anouchi (1993) and Singerman (1997) (for the M G II knee) did not offer a removable proximal/distal attachment method, and the precision worm gear drive used by Miller (2001a) seemed unnecessarily complicated. The design that we thus intended to implement was one that attached a trial femoral component to a rod and allowed the entire construct to rotate in a tube cemented in the centre of the 40 intramedullary canal. To keep the component seated in the bone at all positions (i.e. no rotation, no pull-out), we required a locking mechanism. I decided to locate the locking mechanism at the proximal end of the femoral rod for maximum accessibility. 2.5.1.3 Final Design Following feasibility tests and the construction of a prototype, as described in Appendix 1, a pair (right and left) of size E provisional (trial) femoral implants was modified to allow for internal and external rotation about the femoral canal (Figure 2.7: #7). Size E components were selected because they were the most appropriate fit for all but one of the cadaver legs used in a previous training course on T K A . The two pegs extending proximally from the inner surface of the provisional implant were removed. These pegs were unnecessary, given our design concept, and impeded further machining. A stainless steel block was machined to mate with the box structure on the proximal surface of the implant and was secured to the box structure of the implant using press-fit steel dowels (Figure 2.8). The anteroposterior extent of the steel block had to be limited to allow clearance for the cam on the tibial component. A 24.5 cm long stainless steel rod (R = 0.625 cm, Figure 2.7: #1) was threaded on one end and screwed into a hole that was drilled and tapped into the wedged steel block (Figure.2.7: #6). The rod was then arc welded to the block to ensure rigidity. The threaded rod was inserted through an 18 cm long stainless steel tube (ID = 1.0 cm, OD = 1.25 cm, Figure 2.7: #5), which was cemented into the intramedullary canal of the femur (Figure 2.7:#4). The component could thus rotate about the long axis of the femur. It was locked in place by threading a lockwasher and a nut onto the proximal end of the femoral rod and tightening them against the femoral tube (Figure 2.7: #2 and #3). The proximal 9.75 cm of the steel rod extending from the implant was threaded with 3/8" (0.95 cm) threads (Figure 2.8) to connect directly to the testing rigs, described later in sections 2.6.1.1 and 2.6.2.2. 41 1) Femoral rod of modified femoral component, attaches to testing rig 2) Nut 3) Lockwasher • 4) Femoral bone l ^ r ^ ^ ™ ^ ^ ^ ™ " 5) Femoral tube (in intramedullary canal) 6) 6° block to provide valgus angle 7) Modified femoral component Figure 2.7: Diagram of the modified femoral component implanted in the femur (side view). Figure 2.8: The attachment of the wedged steel block to the femoral component using press-fit steel dowels, and the proximally threaded steel rod extending from modified femoral component. 42 During surgery, the femoral component is implanted in line with the mechanical axis of the femur, which is 6° from the anatomical axis. This alignment is achieved by making the distal femoral cut at a 6° valgus angle with respect to a line perpendicular to the intramedullary canal (Zimmer 2004). To provide the valgus angle for each rotation of the femoral component, the proximal surface of the steel box was machined at a 6° angle prior to attachment to the trial component (Figure 2.9). Following this modification, the contact between the end of the femoral tube and the angled surface of the block imposed the required valgus angle. Figure 2.9: The inferior aspect of the modified femoral component. The final version of this design was machined and constructed by a professional machinist. I had the component machined professionally for several reasons: the trial femoral components are made from an extremely hard alloy, and as such they are very difficult to machine and require special drill bits to drill through them. The components are also very awkwardly shaped, and transfer drilling through the component and into the steel block would have been beyond my technical capabilities. Although the component/rod assembly was held in place in the femur by tightening the lockwasher and a nut against the femoral tube (Figure 2.10), we also enhanced this rotational constraint during testing by hammering steel wedges into place between the femoral component and the medial and lateral anterior aspects of the cut femoral bone. These wedges ensured that 43 appropriate rotational alignment was maintained during the larger loads sustained in deep flexion. Angle pointer Nut Lockwasher Femoral tube Protractor Aluminum block Figure 2.10: Nuts, lockwasher and pointer on femoral rod and protractor on femoral tube. To achieve repeatable rotation, we measured angular rotation with a modified steel protractor and a pointer. The protractor was attached to the tube and the pointer was connected to the implant, so their rotation with respect to each other was measured as the difference in angle indicated by the pointer on the protractor. The protractor was secured to a 1.7 cm x 2.3 cm x 3.9 cm aluminum block using 2 screws, and then a hole (D = 1.31 cm) was drilled in the centre of the protractor and block. We slid the protractor and block onto the proximal end of the steel tube in the femur (Figure 2.10), and the block was held in place using a setscrew. The protractor was held in place using a setscrew because it needed to be removable; the steel tube was cemented permanently into the femur, and thus a permanent connection between the protractor and the tube was not possible. When the implant was set at the correct rotation within the femoral tube and the entire construct was secured using the lockwasher and nut, a threaded aluminum pointer was screwed onto the implant rod. This pointer was set to align with an arbitrary angular marking (Figure 2.11), and was held in place between two nuts. 44 Figure 2.11: Protractor and aluminum pointer To adjust the rotation of the femoral component, we loosened the nut and lockwasher holding the tube/implant construct together and removed the wedges which maintained a stable angular rotation. The implant could then be rotated by 5° and the rotation measured using the scale on the protractor. Details on the method used to set neutral femoral component rotation are given below in the description of surgical preparation (section 2.5.1.4). This design met our functional requirements and allowed us to place the femoral component at the three desired rotation angles (0°, ±5°) while constraining the component in 6 degrees of freedom relative to the bone. 2.5.1.4 Surgical Preparation of the Femur To allow for 5° of internal and external component rotation about the intramedullary canal and to prepare our leg specimens for our testing protocol, we modified the standard surgical T K A procedure as described below. Initial Preparation Dr. Tonetti stripped away most of the soft tissues around the knee; he removed the posterior soft tissues except the joint capsule, and stripped the anterior soft tissues down to the muscles. The joint, tendons and ligaments were left intact. We chose to remove most of the soft tissues surrounding the joint because they were flaccid and played very little role in constraining the motion of the knee. The removal of these tissues for similar experiments has been described by other investigators (Elias 2004c, Miller 1998, Oishi 1996, Rhoads 1990, Singerman 1997, Star 45 1996) and may be preferable in instances where knee specimens undergo significant decomposition over several full days of testing. Dr. Tonetti ensured that enough tissue remained on the medial side of the joint to be able to close the medial incision (following the TKA) with a pair of surgical towel clamps (Figure 2.12). Other researchers have closed the medial incision using sutures following each trial (Armstrong 2003; Miller 2001a; Miller 2001b; Singerman 1997). We were unable to do this because we needed to leave space for the Tekscan sensor to exit the knee joint. During our study, we made at least 30 changes in femoral, tibial and patellar component placement per leg (15 per testing rig); repetitive suturing was not possible because it would have damaged the tissues and would also have been extremely time-consuming. Figure 2.12: Surgical towel clamps used to close medial incision The shaft of the femur was cut short enough to avoid the larger curvature in the bone that would make it difficult to insert the femoral tube into the intramedullary canal. ^ Marking the Kinematic Axes Dr. Tonetti located the transepicondylar axis prior to making the femoral cuts in order to use the axis as a guide for bone cutting and component placement. This axis was also used later to define the kinematic flexion axis of the knee. He palpated the epicondyles and marked their most medial and lateral aspects by inserting #4 x V" woodscrews into the bone at those locations. The long axis of the femur was also defined using two screws; we inserted these superior and inferior screws into the posterior surface of the bone proximal to the joint, parallel to the femoral tube and spaced approximately 10 cm from each other. The screws were tightened to the appropriate depth to ensure alignment of the axis with the femoral tube in the sagittal plane. K^^gfi^ ^ ^ ^ ^ ^ ^ ^ 46 Modified Femoral Arthroplasty Dr. Tonetti began the arthroplasty by drilling a hole in the medullary canal of the femoral shaft, approximately 1 cm anterior to the origin of the posterior cruciate ligament and 5-10 mm medial to the midline of the femur. He ensured that the hole was parallel to the shaft in the anteroposterior and mediolateral projections, although it was also possible to slightly alter the location of the hole later when it was enlarged to fit the stainless steel tube. Dr. Tonetti checked whether the anteroposterior position of the hole in the femoral shaft was at approximately the same location as the rod protruding from the femoral component (relative to the condyles), and whether the placement of the component would create significant anterior notching. Using the appropriate sizing guide, Dr. Tonetti measured and recorded the recommended femoral component size. Sizing is based on the anteroposterior dimensions of the condyles. If the guide indicated that the femur was between two sizes, we reported the smaller size, as recommended by the manufacturer. Although we always used a size E component, we wished to know the discrepancy from the recommended size. He made sure that the guide was in full contact with the bone, the anterior boom was not resting on a bump or in a notch in the bone, and the guide was not in flexion or extension. We compared the transepicondylar line to the sizing guide's indicated cutting line. The cutting guide was set to specify 3° of exterior rotation from the line joining the most posterior points of the posterior condyles. If we found a discrepancy between the guide and the transepicondylar line, both alignments were checked, and occasionally altered. The maximum final discrepancy was 2°. Although the TE axis is a better representation of the flexion axis (e.g. Churchill 1988), our definition of neutral rotation was based on the cutting jig's reference to the posterior condyles due to its repeatability. We checked the TE axis for two reasons: 1) To confirm the posterior condylar line (in one case the instrument did not contact the condyles) and 2) To know how different the cut axis was from the kinematic axis. We defined neutral rotation as 3° external rotation from the posterior condylar axis. In the one case with a previous T K A (specimen 4), the posterior condyles had already been removed and we used the transepicondylar axis to define neutral rotation. Dr. Tonetti inserted one medial and one lateral pin into the holes corresponding to 3° of external rotation on the sizing guide and set the alignment guide to 6° valgus alignment, as recommended 47 by the component manufacturer (Zimmer 2004). He attached the standard cutting block and made the distal cut. He then freehand-cut the anterior and posterior cuts of the femur. To make room for the adjustable femoral component, he resected extra bone in the box cut, leaving the ligament attachments intact. Using a drill, he enlarged the hole for the stainless steel femoral tube and ensured that it was in the correct location with respect to the rod protruding from the adjustable femoral component. He temporarily inserted the stainless steel tube (Figure 2.13) and slid the femoral component into the tube to test the fit of the component in neutral rotation. The placement of the component and tube were adjusted until the component was flush with the distal end of the femoral bone cut, and the steel tube was flush with the femoral component. Using a saw, he marked the tube where it projected from the distal bone so that he could reproduce its placement. He slid the protractor block onto the proximal end of the tube, and i f necessary, cut the proximal end of the femoral bone shorter to make room for the protractor block on the tube. Figure 2.13: The stainless steel tube resting temporarily in intramedullary canal of femur, and the remaining femoral bone following surgery. The femoral component was rotated to neutral alignment. Using a scalpel, he marked the neutral position of the component by carving a thin line into the anterior surface of the femur to match an etched mark on the metal femoral component. This 'landmark' ensured that approximate neutral placement could be reproduced upon removal and reinsertion of the component. Neutral alignment was based on the surgeon's perception. After marking neutral rotation, Dr. Tonetti resected additional bone from the anterior and posterior surfaces of the distal femur and in the box cut to allow for at least 5° of internal and 48 external rotation of the modified femoral component. Because this procedure eliminated the condyles defining neutral rotation, the neutral markings on the component and femoral bone were necessary references for accurate placement of the component. It was difficult to remove sufficient bone to allow for the rotation. What remained was often simply enough bone to maintain the ligament connections (Figure 2.13). 2.5.1.4.1 Marker Array Attachment We attached an Optotrak femoral marker array to a threaded Kirschner wire (K-wire) using a setscrew. For the right knee specimens, we mounted the array on the lateral aspect of the femoral shaft by drilling the K-wire into the bone. The femoral marker array was mounted on the medial aspect of the left knee specimen. The array extended posteriorly from the surface of the bone, approximately 15 cm from the joint centre. It was not possible to mount the marker array on the anterior surface of the femoral shaft because the array would interfere with the motion of the quadriceps tendon. 2.5.2 Tibial Rotations 2.5.2.1 Development of the Modified Tibial Component Similarly to the femoral component, we modified left and right trial tibial components such that we could rotate them 5° internally and externally while constraining them in 6 degrees of freedom relative to the bone. We could find no reports of other adjustable tibial components in the literature. The original surgical trial component is a 2-part construct consisting of a tibial tray and a 10 mm plastic tibial spacer (Figure 2.14); the tibial spacer is designed to snap into place in the tibial spacer. Size 4 trial components were the most appropriate fit for the range of knee specimens we planned to test, based on measurements of the cadaver legs used in a previous training course on T K A . Figure 2.14: The trial tibial tray (size 4), the plastic tibial spacer (10 mm thickness), and the aluminum tibial baseplate. 49 To attach the tibial tray and tibial spacer to the bone of the tibial plateau in 3 rotational positions, Dr. Anglin designed and machined a 3.2 mm thick aluminum baseplate (Figure 2.14) to which the tibial tray could be attached. The baseplate was machined with dimensions and contours matching those of the tibial tray and could be secured to the tibia medially and laterally using two 4 cm long screws (Figure 2.15). Three pairs of holes (D = 2.7 mm) were drilled and tapped into the tibial baseplate, corresponding to neutral rotation, 5° of internal rotation, and 5° of external rotation (Figure 2.15). Using 2 screws, the trial tibial tray could be attached to the aluminum baseplate in any of the three rotational positions (Figure 2.15). We originally intended to use a 3 r d screw posteriorly, but found that this was both difficult to access and unnecessary to secure the tibial tray. Figure 2.15: The tibial baseplate secured to bone using 2 screws, and the installed tibial tray. The tibial spacer was snapped into place in the tibial tray, and the entire 3-part construct was secured into the tibial bone using a 5 cm long screw and a 2.5 mm thick plastic washer (Figure 2.16). The screw passed through the spacer, the tray, and the plate and threaded into the bone. 50 Figure 2.16: The 3-part tibial construct and its attachment to the bone of the tibial plateau The oversized hole in the tibial spacer (Figure 2.17) accommodated the rotation of the tibial tray and spacer with respect to the base plate. Two different plastic washers were used when attaching the tibial spacer to the tibial bone: one with a central hole for the neutral rotation and one with an off-centre hole for the 5° rotations (Figure 2.17). The offset washer was needed to accommodate for the change in position of the threaded hole in the tibia with respect to the spacer. Figure 2.17: Complete tibial construct secured in bone in neutral rotation and in external rotation using offset washer. This design allowed us to place the tibial component at the three desired rotation angles (0°, ±5°) while constraining the component in 6 degrees of freedom relative to the bone. 51 2.5.2.2 Surgical Preparation of the Tibia To allow for 5° of internal and external tibial component rotation and to prepare our leg specimens for our testing protocol, we modified the standard surgical T K A procedure as described below. Initial Preparation Dr. Tonetti cut the shaft of the tibia short enough to avoid the larger curvatures in the bones which make it difficult to insert a rod into the intramedullary canal. Marking the Kinematic Axes The coordinate system for the tibia was based on medial and lateral screws (in line with the component) and superior and inferior screws (in line with the threaded rod). Similarly to the femur, we marked the long axis of the tibia by inserting superior and inferior screws into the posterior surface of the bone, approximately 10 cm from each other and in line with the tibial rod. These screws were located distal to the joint and were tightened to the appropriate depth to ensure alignment of the tibial axis with the tibial rod in the sagittal plane. Modified Tibial Arthroplasty Dr. Tonetti inserted a 32 cm length of 3/8" steel rod into the intramedullary canal of the tibia. Using the cutting guide, he resected enough bone from the tibial plateau to allow extension with 7° posterior slope of the plateau, as recommended by the implant manufacturer (Zimmer, 2004). It was necessary to remove extra bone from the tibial plateau to accommodate the 3.2 mm tibial baseplate that would be attached to allow for rotation of the tibial component. We recorded the tibial size recommended by the sizing guide. We placed the tibial baseplate on the proximal cut surface of the tibia and aligned a central mark on the anterior surface of the baseplate with the mid-to-medial third of the tibial tubercle (Figure 2.18). This placement corresponded to the position typically selected intraoperatively. 52 Figure 2.18: The marking on the tibia at the mid-to-medial third of the tibial tubercle, used to align the tibial baseplate. Marker Array Attachment We secured the tibial marker array to a threaded K-wire, which was drilled into the anterior surface of the tibia, approximately 15 cm from the joint centre. It protruded anteriorly from the bone in the sagittal plane; this plane was perpendicular to the optoelectronic tracking system and thus resulted in the most accurate kinematic measurements (see section 2.2). 2.5.3 Patellar Placement 2.5.3.1 Development of the Modified Patellar Component Dr. Anglin modified a trial patellar implant (button) such that we could alter its placement with respect to the remnant bone during testing. Using modular wedges and disks, we were able to constrain the patellar button in 6 degrees of freedom relative to the bone while altering its position in a repeatable manner. With the appropriate combination of wedges and disks, we were able to adjust the patellar component's mediolateral position (0.0 mm, 2.5 mm and 5.0 mm medial), simulate different bone cut angles (0.0°, 7.5° medial, 7.5° lateral and 15° lateral), and alter the thickness of the patellar construct (original thickness and +3 mm). The design made it possible to combine these variations if desired (e.g. additional thickness at a medialized location), although we chose not to do so due to the degradation of the pressure sensors (see section 2.7.2.1) and the time constraints on testing. The patellar component was attached to the bone via a baseplate and a modular disk or wedge that set it in neutral placement or altered its position. The modular wedges and disks all had the same central thickness. Figure 2.19 shows the neutrally aligned patellar construct, composed of the modified patellar button, a modular disk, and the baseplate. 53 Figure 2.19: Patellar construct in neutral alignment. Modified Patellar Implant To create the modified patellar component, the pegs which protrude from the anterior surface of a normal trial patellar component were removed; these pegs typically assist in the fixation of the implant in the bone during surgery (Figure 2.20). An additional 2.5 mm thickness was also removed from the anterior surface of the implant, and a cylindrical magnet (D = 6.4 mm) was press-fit into the centre (Figure 2.20). Two steel pegs (D = 1.6 mm) were press-fit at a distance of 8.0 mm and 6.5 mm from the centre of the patella. The placement of these pegs corresponded to the location of a pair of holes along the mediolateral axis of each modular wedge and disk. The magnet provided a normal force to connect the button to the wedge or disk, and the two steel pegs prevented translation or rotation of the button. Many design alternatives were considered; however, the magnet approach seemed the most robust and the modular components seemed the most practical for the number of patellar variables we tested. Components were machined for sizes 32, 35 and 38; however, we only performed tests using size 35 because that was the recommended size in all but one case (see section 2.8.1). Figure 2.20: The original NexGen patellar implant and the modified trial patellar implant, anterior and posterior sides 54 Modular Disks and Wedges A set of modular steel disk and wedges allowed us to connect the patellar implant to the baseplate and to represent different mediolateral positions, incorrect cut angles, and additional thickness (Figure 2.21 and Figure 2.22). To represent the neutral position, we used a 5 mm thick steel disk (D = 3.5 cm) (Figure 2.21A). To represent incorrect bone cut angles, we used three wedged disks, all of which had a constant central thickness of 5 mm. Two lateral wedges were machined at 7.5° and 15° angles, with thicker lateral sides (Figure 2.21B and Figure 2.22A, respectively). The 7.5° medial wedge had a thicker medial side. An additional thickness could be added to the patellar construct by attaching a steel disk of 3 mm thickness to the 'neutral' disk (Figure 2.22B). Figure 2.21: A) Steel disk representing neutral placement. B) Steel wedge used to introduce 7.5° lateral bone cut angle. Figure 2.22: A) Steel wedge used to introduce 15° lateral bone cut angle. B) Patellar component with 3mm thickness disk added. Patellar Baseplate The 2.7 mm thick, hexagonal, aluminum baseplate (3.3 x 4.3 cm) (Figure 2.23) was attached to the posterior surface of the patella using six shortened sewing pins that passed through holes (R 55 = 0.25 mm) along the edge of the plate at a 60° angle. On the medial side of the baseplate along the mediolateral axis, three 1.8 mm diameter holes were drilled at 2.5 mm intervals. A slot was machined along the lateral portion of the mediolateral axis. The holes and the slot were designed to mate with the two 1.6 mm pegs protruding from the anterior surface of each disk or wedge (Figure 2.23). We medialized the patellar button by selecting a different hole along the mediolateral axis (0.0 mm, 2.5 mm or 5.0 mm medial). A 6.4 mm diameter magnet was press-fit into the centre of the patellar baseplate to provide a connective force between the aluminum baseplate and the steel wedges and disks. Figure 2.23: Patellar baseplate and neutral placement disk. The pegs on the disk were designed to mate with the holes and the slot in the baseplate. The disk corresponding to an additional 3 mm thickness was attached to the neutral disk via a pair of pegs and holes along the superior-inferior axis of the patella. The new patellar construct enabled us to alter the placement of the patellar button in a repeatable manner. We were able to simultaneously constrain the patellar button in 6 degrees of freedom relative to the bone and adjust its position, angle and thickness. 2.5.3.2 Surgical Preparation of the Patella We modified the standard surgical T K A procedure to implant the modified patellar component as described below. Modified Patellar Arthroplasty Dr. Tonetti dissected through the pre-patellar bursa to expose the posterior surface of patella (Figure 2.24). He removed all osteophytes and synovial insertions from around the patella while 56 being careful not to damage the tendon insertions onto the bone. Using calipers, he measured and recorded the original patellar thickness (Figure 2.25) and recorded the recommended patella size. To resect the patellar bone, he applied the patellar saw guide in line with the patellar tendon and pushed the patella up between the jaws of the saw guide (Figure 2.25). He levelled the patella within the saw guide jaws and used the thumbscrew to tighten the guide. Figure 2.24: Exposing the posterior surface of the natural patella Figure 2.25: Measuring the original patellar thickness with calipers, and resecting the patellar bone. Dr. Tonetti measured the thickness of the patellar component (button, neutral angle disk, and baseplate) and resected additional bone (typically between 9-10 mm) until the total reconstructed patellar thickness was within 1 mm of the original thickness of the patella. To ensure equal bone thickness on the medial and lateral sides, he measured the medial and lateral thicknesses at approximately one quarter of the width of the patella using arched calipers (Figure 2.26) and resected extra wedges until the patellar cut was flat. 57 Figure 2.26: Using arched calipers to measure medial and lateral patella thicknesses Marking the Kinematic Axes It was necessary to establish the mediolateral axis of the patella both for correctly aligning the patellar baseplate setting landmarks for later digitization. Dr. Tonetti inserted 4 shortened sewing pins at the inferior and superior poles and medial and lateral borders of the patella. These landmarks were identified visually based on geometry of patella and the soft tissue attachments. The pins were inserted into the anterior surface of the patella, passing through the soft tissues into the bone (Figure 221 A). Figure 2.27: A) Patellar landmarks and threaded K-wire inserted through patellar bone for marker array attachment. B) Patellar baseplate attached to patellar bone using 6 shortened sewing pins. The baseplate was attached to the posterior surface of the patella using 6 shortened (-10 mm) sewing pins (Figure 2.27B). This provided a secure fixation. The medial and lateral landmark pins on the anterior surface of the patella were used as alignment guides as we simultaneously attempted to maximize and centralize coverage of the exposed bony surface. 58 Marker Array Attachment To secure a marker array to the patella, we attached the array to a K-wire which was screwed permanently into a small aluminum block. Using a drill, Dr. Tonetti inserted a second threaded K-wire through the anterior surface of the patella such that the two ends protruded from the bone by approximately 5 mm (Figure 2.27A). The two ends of this K-wire were used as anchors for securing the aluminum block to the patella using wire. We used two figure-eight twistings of wire around the anchors and across the aluminum block to attach the aluminum block (and thus the marker array) to the patella (Figure 2.28); this method was adequate to immobilize the marker array, based on manual tests. Figure 2.28: Marker array holder attached to patella via threaded K-wire, aluminum block and figure-eight wire twistings Final Surgical Steps We affixed a neutral angle disk to the patellar baseplate, and mounted the patellar component to the disk. With the adjustable patella, adjustable tibia and adjustable femur in neutral positions, Dr. Tonetti performed a trial reduction. He checked the ligament stability and the range of motion of the knee. If necessary, he performed soft tissue releases or recut bone surfaces. The femoral tube and tibial rod were cemented in place using bone cement (Figure 2.29). Prior to cementing the femoral tube in place in the femoral shaft, Dr. Tonetti inserted a cork into the end of the femoral tube to prevent cement from entering the tube. When the cement cured, we cut the tibial rod such that the distance from the tibial plateau (the proximal bone cut) to the end of the rod measured 32.5 cm. When the implants were in place and the knee specimen was 59 attached to the testing rig (see sections 2.6.1.1 and 2.6.2.2), the distance between the 'hip' and 'ankle' joints was approximately 86 cm, which corresponded to the average leg length for a 175 cm tall person (Winters 2005). Using the markings on the anterior surfaces of the bone and the component as guides, we rotated the femoral component to the neutral position in preparation for testing. Figure 2.29: The stainless steel tube cemented into the intramedullary canal of the femur 2.6 Testing Rigs Testing rigs are used by researchers to understand the mechanical behaviour of joints in simulated physiological circumstances. We employed two testing configurations, an intraoperative testing rig and a postoperative testing rig, to establish whether or not changes made during surgery affect the postoperative performance of the knee. Surgeons rely on intraoperative measurements, observations and manipulations to give them an indication of how a knee will function postoperatively; however, a link between intraoperative and postoperative knee mechanics has not yet been established. We performed our first tests in a simulated intraoperative configuration to mimic the kinematic and kinetic effects a surgeon would see by changing component placement during surgery. The intraoperative simulation was performed using a horizontal rig to emulate the patient's positioning. In the horizontal rig, each knee specimen was flexed passively. To link intraoperative measurements to postoperative performance, the knees were subsequently tested in a loaded configuration. Each knee was mounted in a vertical, Oxford-style testing rig and was 60 actively loaded during flexion and extension to simulate deep knee bends (squats) or stair-climbing. Both testing rigs are described in detail below. 2.6.1 Horizontal Testing Rig During surgery, the surgeon stabilizes the femur and manipulates the tibia while the patient lies supine. The surgeon tests joint laxity, monitors the flexion gaps, and checks the tracking of the patella. Based on their qualitative observations, the surgeon decides i f a lateral release is required or i f additional adjustments must be made to component sizing and placement. Ideally, the surgeon would make quantitative measurements of patellofemoral mechanics at this stage in order to predict and improve postoperative performance. To replicate the surgeon's intraoperative manipulations, Dr. Carolyn Anglin designed a horizontal testing rig . This 5 degree-of-freedom rig held the "ankle" joint of a knee specimen and allowed the "hip" joint to slide horizontally, which produced flexion in the knee. To our knowledge, no other similar device has been designed for biomechanical testing. Similar devices are sometimes used during surgery, but these are relatively expensive and would have needed to be adapted to suit knee specimens transected at the mid-tibia. 2.6.1.1 Rig Description The horizontal rig consisted of a 2.5 cm thick varnished wooden base, to which we attached a metal track with screws (Figure 2.30). A universal joint mounted at the end of the metal track acted as the ankle joint of the leg. The tibial rod protruding from the knee specimen was threaded into a sleeve, inserted into the universal joint and held in place with a setscrew. The femoral rod extending from the proximal portion of the knee specimen was also inserted into a threaded sleeve; we attached this sleeve to a slider which was allowed to slide along the track on the wooden base. The slider could also be locked into position at a specific knee position. The connection between the slider and the femoral rod permitted the simulated hip joint to flex and extend like a hinge, while the slider allowed the simulated hip joint to translate along the track. Thus, by pushing the slider back and forth along the track, the knee was forced to flex and extend (Figure 2.30). 2 Dr. Carolyn Anglin designed and built the horizontal rig, and I assisted in modifications of rig parts. 61 To test a knee specimen under simulated physiological circumstances, no constraints should be placed on the structures of the knee either by the apparatus used to hold the specimen or by the devices used to apply forces to it. Assuming that the knee behaves substantially as a single degree of freedom joint, the specimen initially has 7 degrees of freedom. Since both the femoral and tibial threaded connections were free to rotate, the hip applied 3 constraints and the ankle 3 constraints. There were therefore just enough constraints to allow the hip to slide along the track, thus allowing the knee to flex and extend in the sagittal plane. Figure 2.30: Knee specimen in horizontal rig 2.6.1.1.1 Quadriceps Clamp To connect the extensor mechanism of the knee to their testing rigs, other researchers have sutured the quadriceps tendon to a nylon strap (Lane 1994, Miller 2001b, Oishi 1996, Shoemaker 1993, Star 1996) or to steel cables (Armstrong 2003). We felt it was unlikely that sutures would hold during loaded flexion in the vertical testing rig; therefore, we connected the quadriceps tendon to the testing rigs using a specially-designed clamp similar to those used by Elias (2004c), Nagamine (1994) and Singerman (1997), and used previously in our research group. The clamp consisted of two plates with mating rows of teeth designed to compress the quadriceps tendon (Figure 2.31). 62 Figure 2.31: The anterior and posterior plates forming the quadriceps tendon clamp We stripped the extra flesh off the quadriceps tendon such that it would fit between the two plates and secured the plates together with four screws, sandwiching the tendon in the middle. The clamp was attached approximately 8 cm from the centre of the patellar baseplate, and any remnant of the quadriceps tendon extending past the top of the plates was trimmed. To prevent tearing of the tendon due to overtightened clamp screws, we tightened the Allen screws with the handle of the Allen wrench in the unextended position (to minimize the applied torque). A hook was screwed into the superior face of the anterior plate such that the quadriceps clamp could be easily detached and reattached to the testing rigs. Attachment of Quadriceps Clamp to Testing Rig To ensure that the quadriceps tendon was under some tension and that the patella would not slide off the femoral component, we attached the quadriceps clamp to the "hip" slider via a 9.5 cm light spring. The spring connection ensured that the tendon was kept in alignment with the femoral shaft throughout flexion and extension and also mimicked the passive resistance that would still be present in the slack yet intact muscles of a T K A patient. Measurement of Quadriceps Force To measure the tensile force in the quadriceps tendon of specimen 7, we attached a 200 lb (890 N) load cell (Model LCI01-200, Omegadyne, Stamford, CT) between the hip slider and the spring attached to the quadriceps clamp (Figure 2.32). The load cell remained in place throughout testing with this specimen. The tension across the spring was measured to range between approximately 11 N in full extension (5°) and 90 N in full flexion (125°). We used the load cell during one horizontal trial for specimen 8 and found the output to be so consistent with specimen 7 that we did not continue to record further loads. 63 Figure 2.32: Load cell inserted between slider and spring in horizontal rig for specimen 7 2.6.1.1.2 Range of Motion and Timing To maintain a consistent maximum angle of flexion, we flexed each knee to 125° (measured with a goniometer) and fastened a screw into the wooden base to prevent the slider from translating any further. We flexed the leg and extended it fully while ensuring that all the optoelectronic markers were visible for the whole flexion cycle. If any markers were not visible, we adjusted the position of the Optrotrak camera or the angle of attachment of the marker arrays. We positioned the camera such that it was approximately 2.1 m from the markers. During testing, the knee was flexed manually for 15 seconds and extended for 15 seconds using a pushrod. 2.6.2 Vertical (Oxford) Rig 2.6.2.1 Background The knee specimens were mounted in a vertical Oxford-style testing rig to simulate the flexed-knee stance (i.e. deep knee bending) which occurs when rising from a chair or climbing stairs (Miller 1998). The Oxford rig simulates closed kinetic chain knee extension; it is a motor-driven rig which flexes and extends the knee by pulling on and releasing the quadriceps tendon. The relative movements of the femur and the tibia are constrained by the passive structures of the knee specimen instead of by the rig (Zavatsky 1997). The Oxford rig was developed at the University of Oxford by O'Connor and colleagues (Bourne 1978). The Oxford research group has continued to use the original rig and modifications of it in various studies (Biden 1990; Miller 1998; Weale 2002). Several studies from the Scripps Clinic in La Jolla, California have used the Oxford rig in their knee investigations (Browne 2005, 64 D'Lima 2000, D'Lima 2001b, D'Lima 2001c, D'Lima 2003, Ezzet 2001, Oishi 1996, Patil 2005, Petersilge 1994, Star 1996). It has also been used in Pittsburgh (Miller 2001a, Miller 2001b) and Boston (Wilson 2003, Elias 2004c). Variations of the original rig have been used in other knee studies (Shoemaker 1993, Singerman 1997). 2.6.2.2 Rig Description In the Oxford rig, the knee specimen is connected to "hip" and "ankle" universal joints which allow spherical movement about the joint centres (i.e. flexion/extension, abduction/adduction, and internal/external rotation of the femur and tibia) (Figure 2.33). To guide the vertical movement of the hip relative to the ankle, the hip joint is mounted on a moveable tray which is allowed to slide up and down tracks on the vertical support of the rig. The tray is suspended from a pulley at the top of the vertical support and is counterbalanced by a weight hanging on the opposite side of the vertical support. For our testing, the effective tray weight with the counterbalance in place was 80 N (see section 2.6.2.2.2). The tibial rod extending from the knee specimen was held in the ankle joint on the base of the Oxford rig using a threaded connective sleeve and a setscrew. We threaded the femoral rod protruding from the knee specimen into the mounting frame of a small DC servo motor, which was connected to the hip joint on the movable tray. The quadriceps tendon clamp was attached to the motor via a 1.25 mm diameter bicycle brake cable and a double pulley (Figure 2.33). Lengthening the cable allowed the knee specimen to flex, and winding the cable actively extended the knee. The weight applied to the hip joint would cause the knee to buckle i f a load were not applied to the quadriceps tendon to extend the knee. 2.6.2.2.1 Range of Motion and Timing We allowed the knee to flex to approximately 95°-100° (measured with a goniometer). Each cycle of flexion and extension took approximately 60 seconds. To produce the same maximum flexion angle on each trial, we taped two wooden blocks to the vertical support to prevent the tray from sliding below that level (Figure 2.33). 65 Figure 2.33: Vertical (Oxford-style) testing rig and the motor-clamp connection 2.6.2.2.2 Applied Vertical Load We attempted to generate a knee flexion moment on the order of a physiological moment without damaging the knee specimens. The applied load of 80 N generated a maximum knee moment of approximately 26.5 Nm. Other researchers using the Oxford rig have reported applying vertical loads ranging between 18 N and 87 N (Elias 2004c, Lane 1994, Miller 2001a, Miller 2001b, Oishi 1996, Shoemaker 1993, Star 1996). Singerman (1997) applied loads such that the knee moment ranged between 1.5 Nm and 6 Nm. D'Lima (2001b) applied a load to the rig crosshead which generated a maximum flexion moment of approximately 40 Nm at the knee; they reported that this load was comparable with that reported during stair climbing after total knee arthroplasty. Morrison (1968) measured a knee extension moment of 35 Nm during walking, and Schuldt (1983) measured a moment of 70 Nm at 105° flexion when rising from a squat. The loads we applied to the knee may seem somewhat low compared to physiological loads; however, it was not possible to increase loading without damaging the specimens. 2.6.2.2.3 Attachment of Quadriceps Clamp to Testing Rig We attempted to use the Omegadyne load cell during testing to measure the quadriceps force for comparison to the quadriceps tension in the horizontal testing rig, but this proved to be 66 unfeasible without moving the quadriceps clamp directly beside the patella and removing the hook from the clamp, both of which we were reluctant to do for the entire experiment; there was not adequate space between the quadriceps clamp and the double pulley to insert the 9 cm long load cell. Since this load cell had been used in a separate experiment performed by Mike Paice & Dr. John Mountenay at UBC, we had information regarding the general shape and magnitude of the force curve. These data will be presented in section 5.9.2 of the Discussion. 2.6.2.2.4 Direction of Quadriceps Pull Initially, we attempted to simulate a physiological quadriceps angle while pulling on the quadriceps tendon (approximately 14°-22° (Rhoads 1990)); however, we found that this caused the patella to sublux laterally. The lateral subluxation was most likely due to the absence of constraining soft tissues on the medial side of the knee. To achieve more central tracking, we decided to align the direction of pull with the axis of the femoral shaft. Most researchers report using this practice (Anouchi 1993, Miller 2001a, Miller 2001b, Singerman 1997, Yoshii 1992). 2.7 Integration of Measurement Systems, Knee Specimen and Testing Rigs To assemble a functioning testing system, it was necessary to integrate the kinematic and kinetic measurement systems (Optotrak and Tekscan), the knee specimens, and the testing rigs. We adapted various aspects of our testing configuration and protocol to ensure that our measurement systems were providing reliable data. The following section outlines special modifications which were made to the system, primarily to the Tekscan pressure measurement sensors, in order to simulate and measure physiological motion of the patella. 2.7.1 Adhering the Pressure Sensor to the Patella To be able to track the loads on the patella, it was necessary to permanently affix the Tekscan pressure sensor to the patella. We used 3M Hi-Strength 90 Spray Adhesive (3M, St. Paul, MN) to attach the sensor to the patellar button. See Appendix 2 for details on the logic behind using an adhesive and on the adhesive selection process. When we glued the sensor to the patella, we aligned the sensor with the pegs on the anterior surface of the patellar button such that the sensor exited the joint through the medial incision of the knee specimen. To force the sensor to conform to the dome-shaped patellar implant, it was necessary to create a fold in the sensor (Figure 2.34). We decided to create the fold on the 67 inferior surface of the patella, because we predicted that contact would rarely i f ever inferior aspect of the patella. Figure 2.34: Tekscan sensor glued to the patellar implant, inducing a fold in the sensor We found that wrinkles appeared in the sensor i f it was not glued to the sensor several hours prior to testing (Figure 2.35). The sensor would shift relative to the patellar implant due to the high shear loads during loaded flexion. We obtained good stability of the sensor when it was attached to the patellar implant the day before testing. 2.7.2 Sensor Degradation and Loss of Output During tests with our first 2 specimens, we observed both gradual degradation of sensor output as well as the loss of entire rows and columns of output data. The loss of output made comparisons between variables very difficult and gave us very little confidence in our force data. 68 We hypothesized that these issues (degradation and loss of rows/columns) were independent of each other, and developed separate solutions to deal with each problem. 2.7.2.1 Sensor Degradation Tekscan sensors have become notorious for the degradation of their output over time (Brown 2004). The decrease in raw sensor output (between 0 and 255) occurs over time with repeated loadings of the sensor. Tekscan does not state in its literature the number of times a sensor can be used before it becomes unreliable due to gradual decrease in output; however, Tekscan warns that when loaded "under severe conditions, such as against hard surfaces, sharp edges, non-flat surfaces, sliding surfaces, or shear forces, a sensor may have a very limited life" (Tekscan 2001). Rough handling will also shorten the useful life of a sensor (Tekscan 2001). We did not originally believe that our testing setup would produce "severe conditions". In a personal communication with Dr. James Rudert from the University of Iowa, he concluded that shear loads enhanced sensor degradation; his research group witnessed the degradation of output with Tekscan sensors inserted into the tibio-talar joint of cadaver ankles. When asked for advice on how to minimize sensor degradation, Dr. Rudert made two valuable suggestions: limit the sliding (relative motion) to one surface of the sensor, and lubricate the joint. Sensor Lubrication We already minimized sliding by gluing the sensor firmly to the patella; to lubricate the joint, we used a Teflon-based dry film lubricant (DryTef, Walter Tool Company Inc., Norwell, MA) . We used this aerosol spray to coat the articular surfaces of the femur, patella and tibia with a thin, dry layer of Teflon. See Appendix 2 for the details of the selection of this lubricant. To enhance the sliding between the Tekscan sensor and the femur, we placed a Teflon-sprayed square of wax paper between the sensor and the femur. The piece of wax paper was replaced with a fresh piece whenever it became torn or crumpled. The combination of the Teflon lubricant and the wax paper significantly reduced degradation, and was used successfully for specimens 3-8. Because the wax paper was less than 0.05 mm thick, we do not believe that it affected joint mechanics. 2.7.2.2 Loss of Rows and Columns of Sensor Output While gradual degradation affected the entire sensing surface of the sensor, we also observed the loss of entire rows and columns of data during testing with our earlier specimens. The output of 69 intact cells is not affected by damage to individual cells, rows or columns (Tekscan 2001); however, the worst cases in our study involved the loss of as many as 13 rows or columns early in the testing. To test the theory that electrical leads were breaking (or nearly breaking) in the sensor tab or sensing matrix, we connected two sewing pins to an ohmmeter and measured the resistance between the pins when they were poked into the leads. The ohmmeter reported large resistances in the sensor tab where the leads made abrupt angular turns before entering the actual sensing area of the sensor. We deduced that the leads were breaking as the sensor tab was twisted and distorted during testing. Sensor Taping To prevent damage to the lead wires in the sensor tab, we coated the entire femoral side of the sensor with a layer of 0.05 mm thick packing tape. The tape added a layer of protection from shear loads and also served to prevent punctures and tears in the sensor. We do not believe that the tape was thick enough to affect the mechanics of the joint. We attempted to tape both sides of the sensor (specimen 2) but found that when the taped side of the sensor was glued to the patella, the adhesive on the tape would fail.. As a result, the tape would remain glued to the patella while the sensor began to shift. We also covered the sensor tab alone in several additional layers of packing tape to increase its stiffness and prevent it from twisting. The areas of the sensing part of the sensor that were not in contact with the patellar button were reinforced on both sides with thicker, fibre mailing tape (Figure 2.34). 2.7.2.3 Supporting the Tekscan Handle For the first 3 specimens, we wrapped the Tekscan handle in a Ziploc bag to protect it, and taped it to the tibia to hold it in place. Since the tibia is a consistent distance from the patella, this prevented pushing and pulling of the sensor. Unfortunately, this method caused the sensor to bend as it left the joint through the medial incision and was twisted to be aligned with the tibial shaft (Figure 2.36). When we determined that the loss of columns and rows was most likely due to the abuse of the lead wires in the sensor tab, we attempted to prevent the twisting of the tab. 70 Dr. Anglin designed and built a "handle holder" for the horizontal rig, and I designed a hanging system for the vertical rig. Figure 2.36: Tekscan handle taped to tibia, causing the sensor tab to twist Supporting the Tekscan Handle in the Horizontal Rig The Tekscan handle holder was designed to support the handle and allow it to extend medially from the joint without introducing bends or twists in the sensor tab. It was necessary for the supporting platform to swivel about the approximate flexion axis of the patella; otherwise, the sensor tab would twist with respect to the sensing matrix during flexion. The handle holder was attached to the tibial rod with a threaded connection arm. By moving the connection arm up the threaded rod, we were able to adjust the proximity of the supporting platform to the approximate flexion axis of the patella. The connection arm was attached to the supporting platform with a loose screw, which allowed the platform to swivel about the patellar flexion axis. The Tekscan handle was taped to the upper surface of the supporting platform (Figure 2.37). We hung a mass from a rod extending laterally from the connection arm (Figure 2.37) to counterbalance the moment created by the handle holder about the long axis of the tibia. This moment would otherwise cause the knee to twist to the medial side. We adjusted the placement of this mass along the rod until the knee no longer rotated medially. The mass was then taped in place to prevent it from sliding along the rod during motion. 71 Figure 2.37: The Tekscan handle holder attached to the tibial rod Supporting the Tekscan Handle in the Vertical Rig Instead of trying to balance the two moments created by using the handle holder in the vertical rig, we hung the Tekscan handle from the Oxford rig. We inserted the Tekscan handle in a plastic bag and attached a rope to the bag. We then passed the rope over the rig tray and attached it to the rig counterweight. As the knee was flexed, the counterweights moved upwards and the handle was allowed to descend with the knee. After we implemented the new methods of lubricating the sensor, applying protective tape, and supporting the Tekscan handle, we ceased to lose rows and columns of data, and the degree of degradation was reduced, though not eliminated (Figure 2.38). In the horizontal rig, sensor degradation between first and last baselines was decreased from an average of 36% for specimens 2-6 to 11% for specimens 7 and 8. In the vertical rig, average degradation was 57% for specimens 2-6 and 36% for specimens 7 and 8. 72 80 0 -i , , . . , 1 0 1 2 3 4 5 6 Baseline Trial Figure 2.38: Reduction in sensor degradation between earlier and later specimens. This figure shows the average of the reported forces for each pair of repeated baselines for specimens 2 and 8 in the horizontal rig. 2.8 Test Design 2.8.1 Specimen Details Dr. Jerome Tonetti performed total knee arthroplasties on 8 fresh-frozen cadaveric knees (2 males, 6 females, ages 51-80) using the modified Zimmer NexGen Legacy Posterior Stabilized (LPS) components described in section 2.5. Each specimen was thawed overnight before surgical preparation and then refrozen before testing. Horizontal testing and vertical testing each took one day; the specimen was sprayed, wrapped and kept in the fridge between testing days. The specimen details are given in Table 2.4. Table 2.4: Specimen details: sex, age, side of body (right/left), ideal component sizes Specimen 1 2 3 4 5 6 7 8 Sex F F M F F F F M . Age 80 73 67 78 66 61 66 51 Right/Left R R L R R R R R Femoral Size E E G E E E E F Tibial Size 4 4 5 4 4 4 4 5 Patellar Size 35 35 35 35 35 35 35 38 A l l knees were implanted with size E femoral components, size 4 tibial components, and size 35 patellar implants. Although these sizes were appropriate for most of the specimens, the two male specimens (specimens 3 and 8) were larger and ideally would have been implanted with larger 73 components (Table 2.4). We could not implant larger modified femoral and tibial components because we only machined one size due to the costs of machining a full set. As previously explained in section 2.5.1.3, we based our size selection on the fact that all but one of the cadaver legs used in a previous training course on T K A were size E/4. We possessed a size 38 modified patellar implant that could have been used with specimen 8; however, we did not change patellar sizes in order to compare results more directly with other specimens. 2.8.2 Test Variables We clustered the testing variables into four main testing groups and performed all tests in one group before beginning tests with a new set. We were primarily concerned with comparing the effects of related adjustments within groups (as opposed to between groups). Also, it was more practical in terms of testing logistics to perform all the related alterations and adjustments one after another. The following four groups of variables were tested in the same order in both the horizontal and vertical rigs: 3 femoral rotations: 0° and ±5.0° 3 tibial rotations: 0° and ±5.0° 4 patella bone cut angles: 0.0°, 7.5° medial, 7.5° lateral and 15° lateral (the bone cut angle denotes that an additional thickness was added to the stated side) 3 patella positions and 2 patella thicknesses: 0.0, 2.5 and 5.0 mm medial; original thickness and +3 mm Before testing each group of variables, we reset all components to neutral placement and tested the knee specimen in this neutral, 'baseline' configuration, for a total of 5 baseline measurements. We repeated the tests for each component variation twice, including baseline measurements. 2.8.2.1 Variable Randomization We tested the variable groups in a randomized order, and within each set of variables, the order of component placement was also randomized (see Appendix 3 for the order of tests performed). The femoral rotation testing group was an exception to this testing order; we performed the 74 femoral rotations either first or last because neutral placement was not perfectly repeatable. Thus, had we adjusted femoral rotation midway through testing, it would have been difficult to determine whether observed changes were due to femoral rotation or changes in other variables. A n incentive to perform the femoral rotations last was that they were the most demanding on the knee; they required the greatest disassembly of the setup and occasionally caused the knee to dislocate (see Appendix 4), which could damage the Tekscan sensor. Testing was always completed in the horizontal rig prior to testing in the vertical rig (instead of randomizing the rig testing order) for several reasons: • We attempted to replicate the order of measurements that would be taken in vivo (i.e. surgery is obviously performed prior to vertical loading of the healed knee). • If specimens were going to fail, they would likely fail in the vertical rig. We wanted to be sure to collect the horizontal data prior to failure of specimens. • We wanted to test the additional effect of performing a lateral release (see below in section 2.8.2.2), and we were more interested in the effect of the lateral release on postoperative tracking. We could not perform a lateral release in both testing rigs, so logically we performed tests in the vertical rig last. 2.8.2.2 Additional Tests Upon completion of the aforementioned tests, we performed additional tests to answer smaller research questions; these inquiries related to the effects of our testing protocol on the measured kinematics and kinetics. We completed the following additional tests: Effects of speed: In the horizontal rig, we performed fast trials to test the effect of speed on joint mechanics (not possible in the vertical rig). N = 6. Static measurements: We measured patellofemoral loads and kinematics at several static positions to determine whether static and dynamic measurements are correlated. N = 2. No sensor: To test the effect of the thickness of the sensor on joint kinematics, we inserted a clean patellar button without an attached sensor. N = 4 in the horizontal rig, N = 3 in the vertical rig. No towel clamps: To test the effect of the towel clamps (which closed the medial incision) on joint mechanics, we performed trials without the clamps in place. N = 4 in the horizontal rig, N = 3 in the vertical rig. 75 Lateral release: Following the completion of all other tests in the vertical rig, we performed a lateral release on the knee specimen and measured the patellofemoral loads and kinematics. The release was approximately an inch in length. N = 3. Appendix 3 shows which additional tests were performed on each specimen. 2.8.2.3 Other Testing Details After the components were inserted in the appropriate positions and the patellar component and Tekscan sensor were in place, we closed the medial side of the joint with surgical towel clamps. The locations at which the clamps grasped the tissue on either side of the medial incision were marked with a felt pen such that we could clamp the tissues together in approximately the same manner following each change in implant orientation. Throughout testing, we sprayed the knee specimen with water to keep the tissues moist. We chose to spray with water rather than a saline solution because depositing NaCl on the surface of the tissues over the two-day testing would promote further desiccation of the specimen. After every change in component placement, we checked the integrity of the wax paper between the patella and femur, and changed the paper if it was damaged or crumpled. Periodically, we lubricated the knee by spraying the femoral component, the tibial component and the exposed surface of the Tekscan sensor with the DryTef lubricant. We monitored the graphical output from the Tekscan sensor and when it seemed that rows or columns of data were missing, we checked the presence or absence of rows or columns by pressing on the entire surface of the sensor and recording the output. We were also able to detect the presence of punctures or tears to the sensor by visually inspecting the surface of the sensor after each change in component placement. 2.9 Data Analysis 2.9.1 Kinematic Analysis The raw kinematic data from Optotrak was first analyzed using a Matlab program (tkapfj.m) written by Dr. David Wilson for previous biomechanical experiments; details are described below. I modified this program to be able to read in Optotrak files without requiring the user to manually convert each file to ASCII and to delete the header of each data file. 76 Raw Data Analysis A crucial component of the program was to calculate the patellofemoral and tibiofemoral kinematics. The local coordinate systems of each bone (femur, tibia and patella) were defined using the landmark screws and pins, described in sections 2.5.1.4, 2.5.2.2, and 2.5.3.2, which were digitized during testing. The y-axis (i.e. longitudinal axis) of each bone was defined by the inferior and superior landmarks in the bone and the z-axis (i.e. mediolateral axis) of each bone was defined by the medial and lateral landmarks (Figure 2.39). The x-axis of each bone was calculated as the cross-product of the y-axis and the z-axis and therefore pointed nominally anteriorly (Figure 2.39). Figure 2.39: Reference frames used to specify patellar and tibial tracking with respect to the femur. Patellar tilt and rotation were defined by the rotations of the patellar frame (x p a „ y p a „ xpat) with respect to the femoral frame. Translation of the patella with respect to the femur (shift along the z-axis) was defined by the translation of the centre of the patella with respect to the femoral frame. (http://us.il.yimg.eom/us.yimg.com/i/edu/ref/ga/l/1240.gif) According to the definition of the non-orthogonal Joint Coordinate System (JCS) suggested by Grood (1983) and standardized by Cole (1993), the three-dimensional motion at the knee joint can be described using a coordinate system with one axis imbedded in the proximal limb segment (referred to as ei), one axis imbedded in the distal segment (referred to as £3), and a floating axis normal to the two body fixed axes (referred to as 62). In our study, we computed the motions of both the tibiofemoral and patellofemoral joints because it was necessary to 77 correlate the patellofemoral tracking measures with the flexion angle (i.e. tibiofemoral angle). The femur was defined as the proximal bone of each joint. The axes of the joint coordinate systems of the two joints of the knee were defined as follows: both the tibiofemoral and patellofemoral joints. • The y-axes (i.e. longitudinal axes) of the tibia and patella were selected as §3 for the respective joints. • For both joints, the floating axis, £2, was defined as the cross-product of §3 and §1 for that joint. Tibiofemoral flexion was measured as the angle of rotation about the axis 81. Patellar tilt was measured as the rotation about the 63 (longitudinal) axis of the patella (Figure 2.40). Patellar shift was calculated as the translation along the flexion axis, 61, of the femur. Patellar spin (patellar rotation) was measured as the rotation about the floating axis, e2. We defined lateral tilt, shift and rotation as positive, according to the convention suggested by Katchburian (2003) (Figure 2.41). ?jtcllar long jx.> >-x Figure 2.40: The measurement of patellar tilt, shift and spin using the axes of motion of the patella (Katchburian 2003). The z-axis (i.e. transepicondylar axis) of the femur was selected as §1, the flexion axis of l.iler.V; axis 78 (ftml-M mm m Figure 2.41: Definitions of positive (lateral) tilt, shift, and spin (Katchburian 2003). Because we tracked the motion of the Optotrak marker arrays and not the actual landmarks used to create the JCS, it was necessary to determine the relationship between the location of each marker array and the landmark screws or pins in the corresponding bone. We digitized the landmarks to obtain the location of each bone's coordinate system in global coordinates, and then captured the location of the marker arrays in the same static position to create a second set of (marker array) coordinate systems. The program then calculated the transformation matrices between the landmark and marker array coordinate systems. Given that the data from 4 IREDs were normally available and only 3 are needed to define the rigid body transformation, the program used the Veldpaus/Woltring/Dortmans technique (Veldpaus 1988) to calculate the best transformation in the presence of noise. When we collected dynamic data, we applied this transformation to the marker array locations to express results in the local (landmark) coordinate systems. For specimens 5-8, we performed two different digitizations of the bony landmarks in case markers shifted during testing or to average them for a more reliable estimate of tracking throughout testing. We analyzed the kinematic data for these specimens using both digitizations and visually inspected the output data to ensure that no noticeable abnormalities occurred in the data. We averaged the data from both digitizations to obtain the most accurate representation of the motion of the patella throughout the testing of a specimen. In one instance, for specimen 5 in the horizontal rig, the first digitization resulted in incomplete and erratic data; as a result, we only used the second digitization to analyze the data for this specimen in the horizontal rig. Lateral! slul ralWi (frorual view in CaxEaNi.ew)- extension) 79 Data Extraction I wrote a Matlab program to extract values of tilt, shift and spin at every 15° of the flexion and extension cycle. Data were analyzed between 15°-105° in the horizontal rig and between 15°-90° in the vertical rig. We could not investigate kinematics at angles past 105° in the horizontal rig and 90° in the vertical rig because we did not have kinematic data for more extreme flexion angles for all specimens. We analyzed data for only the extension phase because loads were higher in extension than flexion and because the soft tissues may fail to guide the patella into the proximal area of the trochlear groove during early, passive flexion (Katchburian 2003) (i.e. in the horizontal rig). Loads are generally higher in extension than flexion for typical daily activities such as stair-climbing and getting out of a chair. To compare each surgical variable across the set of 8 specimens, we set the average tracking indices of all baseline trials at 15° flexion as a reference point (i.e. 0° tilt, 0 mm shift and 0° spin). We could not use a smaller flexion angle because not all specimens could attain full extension. My program also calculated the average difference in each tracking index with each change in surgical variables for all specimens. Because we typically performed two repetitions of each test (e.g. two cycles for femoral external rotation), we averaged the output data for the two tests. Statistics To analyze the differences in tracking (tilt, shift and spin) resulting from each change in surgical variables, we performed a two-way analysis of variance (ANOVA) test with repeated measures (i.e. repeated measurements on each specimen) at each extracted flexion angle. The two factors of the A N O V A were testing rig (2 levels: horizontal rig and vertical rig) and surgical variable (11 levels: baseline and 10 changes in component placement). No vertical data were available for specimen 6 because this specimen broke during testing; therefore, the two-way A N O V A was performed using the data from 7 specimens. We did not perform a three-way A N O V A using flexion angle as a third factor because we expected the various tracking and loading parameters to vary as a function of flexion angle. We were more interested in determining the effects of component placement and rig type at several flexion angles throughout the range of motion. Furthermore, the tracking and loading parameters are likely to be continuous functions of flexion angle, so a regression analysis across flexion angles would be more appropriate for estimating this dependence than a three-way A N O V A . 80 When tracking was found to be statistically different between rigs and component placements (p < 0.05), we used paired t-tests with a Bonferroni correction to determine which specific surgical variables produced changes in tracking measures which were statistically different than the baseline. A Bonferroni correction is a multiple-comparison correction used to reduce the probability of erroneously reporting a difference; when several dependent or independent statistical tests, are performed simultaneously, the chance of reporting a spurious positive increases. To apply a Bonferroni correction, the alpha value for the entire data set is divided by the number of comparisons being made. We divided the alpha value of 0.05 by 10 because we compared the 10 changes in component placement (2 femoral rotations, 2 tibial rotations, 3 patellar resection angles, 2 patellar medializations, and an additional patellar thickness) to the baseline data. We performed paired t-tests at three flexion angles: 15° to represent early flexion, 45° to represent mid-flexion, and 90° to represent late flexion. Other multiple comparison techniques exist, such as Tukey's method and Scheffe's method. Tukey's method should be used to test all pairwise differences of means to determine if at least one difference is significantly different from zero. Scheffe's Method tests all possible contrasts at the same time, to see if at least one is significantly different from zero (NIST/SEMATECH 2006). We did not want to test all possible pairwise combinations, so we used the Bonferroni method; it allows one to test a pre-selected group of contrasts. For various specimens, we were missing data for certain surgical variables. These discrepancies were due to either incomplete kinematic data or choices we made during testing to avoid variables which would increase loads on an already damaged Tekscan pressure sensor (see Appendix 5 for details on missing data). To include these partial trials in the analysis, we replaced the missing data with the average value from the other specimens. We performed a sensitivity analysis using the minimum value from the other specimens (a more conservative approach than using average values because it uses the smallest change compared to baseline) and verified that the p-values and our conclusions were not greatly affected by this small change. Although the two-way A N O V A provided information regarding the lack of statistical differences in tracking between testing rigs, we also wished to determine the correlation between the horizontal and vertical data; this was done using a Pearson Product Moment Correlation. This 81 correlation value ranges between -1 and +1 and reflects the degree (strength and direction) of linear relationship between the horizontal and vertical measurements. A correlation of +1 indicates a perfect positive linear relationship between variables; a correlation greater than 0.8 is generally described as strong, and a correlation less than 0.5 is generally described as weak. The correlation represents the degree to which post-operative joint mechanics can be predicted by intra-operative mechanics for an individual patient. Pearson correlations and the corresponding p-values (for a = 0.05) were calculated for each specimen, and an average of these values was determined for 7 specimens. Another common measure of correlation is the coefficient of determination (r2). This value is a measure of the proportion of variation in one variable which can be explained by the other variable. We did not calculate coefficients of determination because we did not wish to predict the variability in postoperative tracking using intraoperative measures. Also, very high values of r 2 can arise even though the relationship between the two variables is non-linear. 2.9.2 Kinetic Analysis The aforementioned program written by Dr. Wilson also analyzed the kinetic data from the Tekscan pressure sensors and matched the contact data to the kinematic data. I added code to this portion of the program which enabled us to add our own calibration algorithm (instead of relying on the Tekscan calibrations); this will be described in detail in Chapter 3. The program summed the force output for the sensor over the sensing matrix and reported the summed contact force, area and pressure. I also added code that would exclude any force data recorded by the sensor that was located outside the circumference of the patellar button (based on the digitization we performed during testing, described in section 2.3.4). I wrote a program to extract the peak forces for each cycle of flexion and extension. As with the kinematic data, we averaged the data from the two trials for each surgical variable. Because the force data degraded over the course of testing, it was necessary to normalize the data so that forces could be compared over the course of the experiment: each subsequent baseline measurement was scaled by the ratio of final to initial peak forces to equal the original baseline values. The scaling factors for the measurements made between baseline tests (i.e. the tests of each surgical variable) were determined by approximating the force degradation between baseline measurements as a linear relationship (Figure 2.42). Force data were excluded from the analysis i f the scaling factor used to normalize the data was greater than 2 (i.e. there was more 82 than 50% degradation in reported force) because this data seemed to become more unusual and erratic (see Appendix 5 for details on discarded data). Force measurements were not available for specimen 1 because of the loss of several rows and columns of sensor output throughout testing (see Appendix 4 for a discussion of this issue). During the testing of specimen 3 in the vertical rig, the sensor was punctured, so results from this sensor could not be used. As a result, force data were available for specimens 2-8 in the horizontal rig and specimens 2, 4, 5, 7 and 8 in the vertical rig (specimen 6 broke during testing in the vertical rig). Baseline 1 Peak Force Assumed linear degradation between baselines I ' I I I Base2 V1 V2 Base3 Testing Timeline Scaling Factors (S) Baseline 2: SBaSe2 = Fi/F 2 Baseline 3: SBaSe3 = F-|/F3 Variable 1: S V i = 2/3*SB a s e 2 + 1/3*SBase3 Variable 2: S V 2 = 1/3*SB a s e 2 + 2/3*SBase3 Figure 2.42: Calculation of the scaling factors used to normalize all data to the first baseline measurement. We attempted to smooth the force data in order to produce more robust estimates of the forces at our 15° index points using the Generalized Cross-Validation (GCV) method; however, the G C V 83 algorithm could not find a clean dividing line between noise and signal content, so no filtering was performed. Statistics To analyze the differences in peak forces between changes in surgical variables in both testing rigs, I performed a two-way A N O V A with repeated measures (i.e. repeated measurements on each specimen). The Pearson correlations between horizontal and vertical data were also calculated on an individual specimen basis (for specimens 2, 4, 5, 7 and 8). These values were then averaged to obtain an overall correlation value to estimate the degree of linear relationship between changes in contact force in the horizontal and vertical rigs. 84 3 Validation of Calibration Techniques for Tekscan Pressure Sensors 3.1 Introduction Thin, flexible pressure sensors are often used in orthopaedic biomechanics to measure loads in the knee joint. Prior to using sensors in cadaver and clinical studies, it is important to show that the results they provide are accurate under the proposed testing conditions. One of the most commonly used pressure sensor systems is made by Tekscan (South Boston, M A ) . This validation paper addresses the accuracy of the calibration routines provided in the Tekscan system software, and investigates the appropriateness of an alternative, user-defined calibration method. Tekscan pressure sensor systems are sold with accompanying calibration software to "[address] the nonlinearity inherent in the sensor's (and system's) response" (Tekscan 2001). The software offers the user a choice between two simple calibrations: linear and power. To apply a linear calibration to a set of raw data collected using the sensor, the user applies a single known load to the sensor. The software calculates the equation of the line passing through this point and the origin (Figure 3.1). This linear calibration can then be applied to all sets of raw data collected with that sensor. Alternatively, the power calibration requires that the user apply two known loads to the sensor. Based on the two applied loads and the point (0,0), the software creates a power calibration curve of the form y = A x b (Figure 3.1). Many investigators using the pressure sensors clinically are primarily interested in qualitative measurements and therefore use the raw data or a linear calibration in their work (e.g., for seating or insole analyses). However, in the field of joint research, investigators are often concerned with quantitative force and pressure measurements. Previous investigators using Tekscan sensors to measure loads in the knee have used both the linear calibration (Harris 1999, Matsuda 1995a) and the two-point power calibration (Matsuda 1995b, Matsuda 1997, Stukenborg-Colsman 2002, Wilson 2003). 85 • — Linear o Power | Raw Sensor Output Figure 3.1: The two Tekscan calibration methods: linear and power Several researchers have devised their own external calibration routines of unknown form. Gil l (2004) calibrated Tekscan sensors by applying incremental loads and creating a nonlinear correlation equation to link known contact pressures to sensor output. Wallace (1998) performed calibration runs at five different loads, but it was not clear what calibration equation was used. None of the aforementioned studies reported their rationale for using the selected calibration curve. Other investigators have neglected to state which calibration curve was used in their study (Anderson 2003, Chapman-Sheath 2003, Kdolsky 2004, L i 2004, Ostermeier 2003, Wirz 2002). There is a paucity of information in the literature regarding the appropriateness of the built-in calibrations supplied in the program software. In a recent validation study using I-Scan 6900 Q U A D sensors (pressure range, 1100 psi (7.6 MPa)) to measure facet loads in the lumbar spine, Wilson (2006) determined that the linear calibrations were more accurate than power calibrations, although both methods overestimated actual sensor loads. They calibrated sensors using a linear calibration performed at 80% of the maximum load and a power calibration performed at 20% and 80% of the maximum load. At low loads, the 80% linear calibration overestimated forces by 50 ± 9%>, and the power calibration overestimated forces by 56 ± 10%). Because the sensors used in this study were only loaded over the lowest 15% of the sensor range, it remains unknown which calibration was most accurate over the entire range of the sensors. Since the Tekscan literature states that sensors are least accurate over the bottom 7%> of their 86 range (Tekscan, 2004), the findings of Wilson may not apply to all Tekscan sensors. Also, the appropriateness of specific calibration routines may vary depending on the interface materials, the sensor type and range, and the experimental loading method (Tekscan 2001). We performed the current validation study to determine the most appropriate calibration method for Tekscan pressure sensors in an environment simulating the resurfaced patellofemoral joint. In particular, we investigated I-Scan model #5051- 2500 psi sensors over their full range of 17.2 MPa. This study addresses the following three research questions: 1) Which calibration method supplied in the Tekscan program software provides the most accurate calibrated output for simulated patellofemoral loading? 2) Do user-defined calibration algorithms supply more accurately calibrated force measurements than the calibration routines provided in the system software? 3) What are the greatest errors encountered when using Tekscan calibration routines? 3.2 Methods To test the effects of different calibration algorithms on sensor output, three new, identical I-Scan pressure sensors (model #5051 - 2500 psi (17.2 MPa)) were conditioned, calibrated and loaded using a materials testing machine (Instron 8874, Canton, MA) . Each of these stages is described below in further detail. To simulate the loading in the patellofemoral joint of a prosthetic knee (with a resurfaced patella), sensors were compressed between a flat polyethylene disk (D = 3.82 cm) with deburred edges and a larger aluminum plate (Figure 3.2). Sensors were coated with K - Y lubricant (Johnson & Johnson, Montreal, Canada) to reduce shear loads at the sensor surfaces. Figure 3.2 shows a typical pressure distribution map at the surface of the PE disk. Throughout conditioning, calibration and loading, we increased the force steadily over 10s, held it for 5s, and decreased it to zero over 10s. Sensors remained unloaded for 120s between load applications. We saved all force data as raw (uncalibrated) values in order to be able to later calibrate the same data using several different algorithms. Tekscan recommends conditioning sensors at 120% of the maximum expected pressure prior to use (Tekscan 2001). We conditioned all sensors four times at a load of 20.0 kN, which produced a pressure of 17.4 MPa at the surface of the PE disk. This pressure corresponded to 120% of the maximum expected pressure of 14.5 MPa. 87 F Instron Load Cell Metal Plate PE Disk Tekscan Sensor Metal Plate Figure 3.2: Application of load to a Tekscan sensor using an Instron materials testing machine, and a graphical map of loads applied to Tekscan pressure sensor. The resulting pressure distribution is smooth in the radial direction without sharp discontinuities. To calibrate the sensors, we applied five different calibration methods to identical sets of raw data: 1. A linear calibration using the Tekscan system software, performed at 3.3 kN (20% of the maximum applied load of 16.7 kN). 2. A linear calibration using the Tekscan system software, performed at 13.3 kN (80% of the maximum applied load). 3. A power calibration using the Tekscan system software, performed according to the manufacturer's instructions at 3.3 kN and 13.3 kN (20% and 80%) of the maximum applied load). A l l Tekscan calibrations pass through (0,0). 4. A ten-point, cubic polynomial whose coefficients were calculated using a custom-written program rather than the Tekscan software. To obtain the calibration loads for the polynomial, we loaded the sensors at ten evenly-spaced loads between 0 and 16.7 kN. The maximum raw output from the sensor (between 0 and 255) was exported to Matlab and plotted versus the applied (Instron) load. Using the raw output and the ten known calibration loads, we fit a cubic polynomial through the data (Figure 3.3). The coefficients of this polynomial were determined using a non-linear least squares curve-fitting technique (Matlab, The Math Works, Inc., Natick, MA) . The lsqnonlin.m function was used to minimize the following cost function: 88 C S(Fcalculated Finstron) where Fcaicuiated was a function of the raw sensor output, and Fi n s t r 0 n was the measured applied load, reported by the Instron load cell. 5. A three-point, quadratic polynomial whose coefficients were also determined using the non-linear least squares curve-fitting technique described above. The calibration loads for this polynomial were chosen to be 1.67 kN, 8.33 kN and 16.7 kN, corresponding to 10%, 50% and 100% of the tested load range. (These loads were chosen because we determined that they provided the smallest root mean square testing errors.) 140 Raw Sensor Output (in thousands) Figure 3.3: Typical 10-point cubic polynomial calibration curve Following conditioning, we performed each of the three Tekscan calibrations: the 20% and 80% linear and the power calibrations. We then subjected each sensor to two loading cycles, each consisting of 10 loads between 0 and 16.7 kN applied in a random order. The first set of loads was used as the raw data to which all calibration algorithms were applied. The second set of loads was used to determine the calibration coefficients for the two user-defined algorithms. Figure 3.4 shows the timeline of the loading and the application of calibration routines. 89 Conditioning 4 times at 120% of maximum load I Tekscan Calibrations 20% Linear Calibration 80% Linear Calibration Power Calibration I Loading Cycle 10 static loads applied between 0 and 16.7 kN in random order I Loading Cycle 10 static loads applied between 0 and 16.7 kN in random order Calibration files saved to be applied to \. test data later Used to calculate calibration polynomials: 3-pt quadratic 10-pt cubic Calibrated test data using 5 algorithms: 3 Tekscan 2 user-defined Figure 3.4: Timeline of calibrations and loading. We saved all force data as raw (uncalibrated) values such that each of the five calibration algorithms (3 Tekscan and 2 user-defined) could be applied to the same raw output. We defined measurement accuracy as the root mean square (RMS) error between the calibrated Tekscan output and the Instron load cell measurements over the tested range. 3.3 Results Of the 3 Tekscan calibrations tested, the power calibration was the most accurate (Figure 3.5); the RMS errors of the measured forces for the 20% linear, 80% linear, and power calibrations were 4.07 ± 0.21 kN, 1.75 ± 0.09 kN, and 0.45 ± 0.16 kN, respectively, corresponding to 24.4%, 10.5%) and 2.7% of the tested sensor range (Figure 3.6). The user-defined polynomial calibrations were markedly more accurate; the quadratic and cubic calibrations had RMS errors 90 of 0.24 ± 0.15 kN (1.5%) and 0.10 ± 0.03 kN (0.6%), respectively. Although the Tekscan power calibration performed well over the mid-range of the sensor, it underestimated the applied force by 6.2% at the maximum applied load (Figure 3.5). The repeatability 10-point, user-defined calibration was extremely good (Figure 3.7); the standard deviation over the tested range was 0.2%. ' 3 Q. 3 o o £ ~o +-» <B i _ « o 18 16 14 12 10 8 6 4 2 0 • Linear (20%) • Linear (80%) a Power • 3-Pt Poly. • 10-Pt Poly. • 0 2 4 6 8 10 12 14 16 18 Applied Load (kN) Figure 3.5: Typical sensor output calibrated using the 3 Tekscan calibrations and 2 user-defined calibrations. 91 Figure 3.6: Average RMS errors of Tekscan and user-defined calibration algorithms. 18 -J z 16 XL *J 14 a 12 o i-o 10 to c Q) 8 w •a a; 6 TO 4 TO o 2 0 g . a .8a Br' O Cycle 1 • Cycle 2 A Cycle 3 8 ' ' 10 15 20 Applied Load (kN) Figure 3.7: Repeatability of 10-point, cubic polynomial. The same calibration algorithm was applied to three different sets of test data. 3.4 Discussion I-Scan force measurements may be accurate to within 0.6% when calibration algorithms use a least squares minimization technique. Although the 10-point cubic polynomial was the most 92 accurate algorithm, the loading process is less practical and considerably more time-consuming. The 3-point quadratic calibration requires only one additional calibration load compared to the Tekscan power calibration and decreased the error in force measurement nearly by half, from 2.7% to 1.5%. While most studies do not report their calibration procedure, a previous study which used the Tekscan power calibration found a higher, but comparable, RMS error of 6.5±4.4% (Wilson 2003). The Tekscan literature states that linear calibrations are suitable for tests over a limited loading range, and the power calibration is preferable for tests in which loads vary considerably (Tekscan 2001). The sensor is resistive; therefore, the response is not linear, although over a small range it appears more linear. Table 3.1 shows a comparison of the calibration accuracies reported by Wilson (2006) and by the current study. Both studies concluded that the 80% linear calibration dramatically overestimated the applied loads at low loading levels; however, our study showed that the 80%> linear calibration underestimated forces at high loading levels while Wilson observed that the linear calibration overestimated loads over the whole range of the sensor. These differences may be because Wilson performed calibrations and tests over the bottom 15% of the sensor range, which is much more linear than the entire sensor range. Also, linearity is a function of particular sensor design (Tekscan 2001), and Wilson used a different type and range of I-Scan sensor (I-Scan 6900 QUAD, range: 7.6 MPa). Wilson showed that the power calibration overestimated the applied loads throughout the tested loading range, whereas we found that the power calibration underestimated forces at low loading levels and slightly overestimated forces in the mid-range of the sensor. Table 3.1: Comparison of calibration accuracy. Wilson (2006) Current Study Linear (20% Fmax) — Underestimated loads by up to 38% Linear (80% Fmax) Overestimated loads by up to 50% Overestimated loads by up to 53%) Power Overestimated loads by up to 56 % Underestimated loads by up to 5.6% In tests with low-pressure F-Scan sensors, Zong-Ping (1998) discovered that stiffer contact surfaces yield higher outputs and greater variability in readings. They found that sensors compressed between two hard surfaces reported loads 3 times greater than sensors compressed between two soft surfaces. Wirz (2002) calibrated Tekscan sensors using both articular cartilage 93 & hard rubber and found that each material produced a different linear calibration curve. Since sensor output is extremely dependent on the stiffness of the contact materials (Tekscan 2001), researchers should always endeavour to calibrate using materials similar to those in the testing environment. A limitation of this work is that we only investigated one type of Tekscan sensor and one set of materials. Also, we did not test the effect of curved, non-conforming contact surfaces; geometry may have a considerable effect on sensor output. The accuracy of the system may be less than reported, since calibrations are static and most biomechanical tests are dynamic. It is expected that errors in individual sensels could be much larger than errors for the overall sensor; Tekscan acknowledges that the output from individual sensels can vary (Tekscan 2001). Since it is straightforward to export sensor output and calibrate data externally, we recommend that investigators design their own calibration curves. The raw sensor data is very nonlinear, thus the conclusions based on raw data may be highly inaccurate. The performance of a sensor may be significantly affected by the conditions under which it is loaded. Output (and sensor nonlinearity) depends on the experimental protocol (sensor type, interface shape and materials, sensor range in use, loading method, etc.) (Tekscan 2001); therefore, sensor behaviour should be investigated for each different application. 94 4 Results 4.1 Patellar Kinematics Figure 4.1 shows the patellar tilt, shift and spin, respectively, for four baseline trials (all components in neutral position) of specimen 2, taken over the course of testing in both the horizontal and vertical rigs. This specimen was selected because it displayed typical patellar tracking patterns. Zero tilt, shift and spin occurred when the coordinate systems of the patella and femur, described in section 2.9.1, were aligned. This definition of neutral tracking was arbitrary and of no clinical importance, although zero tilt in an x-ray view is often a surgical goal. For this particular specimen, tilt ranged between 0°-18° lateral in the horizontal rig and between 3°-14° in the vertical rig (Figure 4.1); however, it should be noted that the knees were flexed to a greater angle in the horizontal rig (105°) than in the vertical rig (90°). We did not investigate kinematics at angles past 90° in the vertical rig because we did not wish to load the specimen excessively. Between 15° and 90° flexion, tilt varied over a similar range. The shapes of the tilt curves were very similar between testing rigs. Patellae tilted more laterally in flexion than in extension in both testing rigs, although the difference in tilt between flexion and extension was greater in the horizontal rig. The measurements of patellar tilt were more variable between baseline trials in the horizontal rig than in the vertical rig. For specimen 2, shift ranged between 0.5 mm medial and 6 mm lateral in the horizontal rig and between 0 mm and 7 mm lateral in the vertical rig (Figure 4.1). Shift varied over a similar range between 15°-90° flexion and the shape of the shift curves were similar between testing rigs. Patellae shifted more laterally in flexion than in extension in both testing rigs. Shift was variable in early flexion but became more repeatable at approximately 35° flexion. The variability in early flexion was more noticeable in the horizontal rig than the vertical rig. In the horizontal rig, spin ranged between 4°-7.5° lateral for this particular specimen (Figure 4.1). In the vertical rig, spin ranged between 10°-13° lateral. The shapes of the spin curves were very similar between testing rigs. Patellae rotated more laterally in flexion than in extension in both testing rigs. The difference in rotation between flexion and extension was greater in the 95 horizontal rig, and rotation was more variable between baseline trials in the horizontal rig than in the vertical rig. The shapes and ranges of the tracking curves varied between specimens; however, the changes in tracking due to the changes in surgical variables were more repeatable. To compare the changes in tracking across all specimens, it was necessary to group the data from the 8 specimens together. We defined neutral (zero) tracking (tilt, shift and spin) for each specimen as the tracking at 15° flexion for the average baseline trials of that specimen. We then averaged the tracking data for all 8 specimens and extracted the kinematic data at 15° increments (between 15°-105° for the horizontal rig and 15°-90° for the vertical rig), as described in section 2.9.1. 96 Patellar Spin Lateral Baseline 1 Baseline 2 Baseline 3 Baseline 4 Lateral Spin 20 40 60 Flexion Angle (•) 80 100 120 14 13 12 *1 10 o- 9 W " Patellar Spin Lateral Baseline 1 Baseline 2 Baseline 3 Baseline 4 20 40 60 80 100 Flexion Angle (*] Figure 4.1: Raw tilt, shift and spin data for baselines 1-4 with specimen 2 in the horizontal and vertical rigs. Each flexion cycle shows an upper trace for the flexion phase and a lower trace for the extension phase. 97 Using a two-way A N O V A , we determined that changes in tilt and shift were statistically different across the range of tested surgical variables at all flexion angles (p<0.05). Changes in spin were only statistically significant at angles of 75° and greater. To determine which particular changes were statistically significant, we used paired t-tests with a Bonferroni correction factor (see section 2.9.1). The following sections outline the changes in patellar tilt and shift due to changes in the following surgical variables: 1) Femoral component rotation 2) Tibial component rotation 3) Patellar resection angle 4) Medialization of the patellar component 5) Increased patellar thickness We observed very few statistically significant differences in spin due to changes in component placement, so those results are presented in Appendix 6. 4.1.1 Femoral Component Rotation External Rotation In the horizontal testing rig, 5° of external rotation of the femoral component increased lateral patellar tilt by 4.0° in early flexion and 1.2° in late flexion (Figure 4.2, Figure 4.3). Statistical differences were analyzed in early flexion (15°), mid-flexion (45°) and late flexion (90°); in the horizontal rig, the change in tilt (with respect to baseline) was significant in early and mid-flexion (p < 0.05) but not in late flexion. A similar increase in tilt was seen in the vertical testing rig; the change in patellar tilt ranged between 4.8° lateral in early flexion and 1.7° in late flexion (Figure 4.2, Figure 4.3). The change in tilt in the vertical rig was significant in early, mid- and late flexion (p < 0.05). In the horizontal rig, the patella was shifted 2.0 mm laterally at low flexion angles and tracked medially until it was 1.2 mm medial at 105° flexion (Figure 4.4, Figure 4.5). A similar medial transition was observed in the vertical rig; the patella tracked between 3.4 mm and 1.1 mm lateral (Figure 4.4, Figure 4.5). The change in shift was not statistically significant in any phase of flexion in the horizontal rig (p > 0.05); however, the change in shift was statistically significant (p < 0.05) in all phases of flexion in the vertical rig. 98 Internal Rotation Internal component rotation of 5° largely had the opposite effect on patellar tilt; in the horizontal rig, the patella was tilted and shifted medially in early flexion but tracked laterally at greater flexion angles. In the vertical rig, the patella remained medially tilted at approximately the same angle throughout flexion. In the horizontal rig, the patella tilted an additional 3.5° medially (relative to neutral rotation) in early flexion and 1.4° medially in late flexion (Figure 4.2, Figure 4.3). In the vertical rig, the change in tilt remained fairly constant and averaged 3.7° medial (Figure 4.2, Figure 4.3). In the horizontal rig, the patella was initially shifted medially by 2.2 mm and tracked to a lateral position of 0.2 mm in full flexion (105°) (Figure 4.4, Figure 4.5). A similar range was observed in the vertical rig (1.9 mm medial to 0.7 mm lateral) (Figure 4.4, Figure 4.5). Internal rotation had a statistically significant effect on patellar tilt in mid- and late flexion in the horizontal rig (p < 0.05), and throughout flexion in the vertical rig (p < 0.005). Shift was not statistically different in the horizontal rig and was only statistically different in mid-flexion in the vertical rig (p < 0.001). External vs. Internal Rotation In both testing rigs, when comparing external and internal component rotation, there was a statistically significant difference in the change in patellar tilt throughout flexion (p < 0.05). The change in patellar shift was statistically different between external and internal rotation in early and mid-flexion (p < 0.05) but not late flexion in both testing rigs. Comparison between Testing Rigs The shape of the tilt curves were similar between testing rigs; medial tilt increased sharply between 15° and 45° and typically decreased gradually thereafter. Changes in tilt were typically comparable between testing rigs (see section 4.3). Differences existed between the shift curves for the horizontal and vertical rigs. In the horizontal rig, the neutral baseline curve decreased (became more medial) gradually, whereas it dropped sharply between 15° and 30° and then remained steady in the vertical rig. In both rigs, the internal rotation curves crossed the baseline curve in late flexion (i.e. the change in shift 'compared to the neutral baseline was zero at this point); in the horizontal rig, this occurred at 99 approximately 70°, whereas it occurred at 80° flexion in the vertical rig. In the horizontal rig, the external rotation curve also crossed the baseline curve (at 100° flexion); however, this crossing did not occur in the vertical rig. The differences in shift compared to the baselines at these crossing points and at all subsequent flexion angles were not statistically significant. For both tilt and shift measurements, there was greater variability across specimens for the horizontal rig than the vertical rig. 100 _ -4 Horizontal Rig Femoral Rotation 45 60 75 Flexion Angle (°) Vertical Rig Femoral Rotation Lateral External Internal Neutral 15 30 45 60 75 Flexion Angle (°) 90 105 Figure 4.2: Effects of ±5° rotation of the femoral component on absolute patellar tilt in the horizontal and vertical testing rigs. The data were averaged across specimens at each flexion angle, and the values were normalized such that zero tilt corresponded to the tilt in the neutral configuration at 15°. Minimum standard deviations in the horizontal and vertical rigs were 3.9° and 1.9°, respectively. Maximum standard deviations in the horizontal and vertical rigs were 10.6° and 3.9°, respectively. Lateral Horizontal Rig Femoral Rotation J ±±- External Internal 15 30 45 BO 75 90 Flexion angle (°) 105 _ro © -6 0 1 -B -10 -12 -14 -16 Lateral Vertical Rig Femoral Rotation -1 r -rzr External Internal _j i _ 15 3D 45 B0 75 90 105 Flexion angle (°) Figure 4.3: Effects (mean change ± SD) of ±5° rotation of the femoral component on relative patellar tilt (compared to neutral baselines) in the horizontal and vertical testing rigs. The data were averaged across specimens at each flexion angle. An asterix (*) indicates that a particular phase of flexion (early, mid- or late) showed statistically significant changes in tilt. A dagger (f) indicates that when comparing external and internal component rotation, there was a statistically significant difference in the change in patellar tilt. 101 Figure 4.4: Effects of ±5° rotation of the femoral component on absolute patellar shift in the horizontal and vertical testing rigs. The data were averaged across specimens at each flexion angle, and the values were normalized such that zero shift corresponded to the shift in the neutral configuration at 15°. Minimum standard deviations in the horizontal and vertical rigs were 2.4 mm and 1.1 mm, respectively. Maximum standard deviations in the horizontal and vertical rigs were 8.2 mm and 5.1 mm, respectively. 61 Lateral £ o I 2 0) I or -4 -a Horizontal Rig Femoral Rotation • TT^—1 • • • c L i i H I External • H Internal 0 15 3D 45 60 75 90 105 Flexion angle (°) L a t e r a l Vertical Rig Femoral Rotation 45 60 75 90 105 Flexion angle (°) Figure 4.5: Effects (mean change ± SD) of ±5° rotation of the femoral component on relative patellar shift (compared to neutral baselines) in the horizontal and vertical testing rigs. The data were averaged across specimens at each flexion angle. An asterix (*) indicates that a particular phase of flexion (early, mid- or late) showed statistically significant changes in shift. A dagger (f) indicates that when comparing external and internal component rotation, there was a statistically significant difference in the change in patellar shift. 102 4.1.2 Tibial Component Rotation External Rotation With the tibial component rotated 5° externally, the change in patellar tilt averaged 1.7° lateral in the horizontal rig and 0.2° medial in the vertical rig (Figure 4.6, Figure 4.7). The change in tilt was statistically significant in the horizontal rig throughout flexion (p < 0.05), but was not statistically significant in the vertical rig in any phase of flexion (p > 0.05). The change in patellar shift averaged 0.6 mm lateral and 0.4 mm medial in the horizontal and vertical rigs, respectively (Figure 4.8, Figure 4.9); this change was not statistically significant in either testing rig during any phase of flexion (p > 0.05). Internal Rotation When the tibial component was internally rotated by 5°, patellar tilt measured 0.6° lateral in the horizontal rig and 0.3° medial in the vertical rig, relative to the baseline (Figure 4.6, Figure 4.7). The corresponding average patellar shifts were 1.0 mm lateral and, 0.1 mm lateral for the horizontal and vertical rigs, compared to the baseline (Figure 4.8, Figure 4.9). The changes in patellar tilt and shift from baseline due to internal rotation were not statistically significant over the range of tested angles in both testing rigs (p > 0.05). External vs. Internal Rotation In the horizontal rig, when comparing external and internal tibial component rotation, the change in patellar tilt was only statistically different mid-flexion and the change in patellar shift was statistically significant during early and late flexion (p < 0.05). In the vertical rig, changes in patellar tilt and shift were not statistically different for external and internal component rotation. Comparison between Testing Rigs For both testing rigs, the changes in tilt and shift were typically smaller than the variability between specimens. For both tilt and shift measurements, there was greater variability across specimens for the horizontal rig than the vertical rig. A full comparison between rigs is given in section 4.3. 103 Horizontal Rig Vertical Rig Lateral Tibial Rotation Lateral 0 15 30 45 60 75 90 105 Flexion Angle (°) Tibial Rotation "0 15 30 45 60 75 90 105 Flexion Angle (°) Figure 4.6: Effects of ±5° rotation of the tibial component on absolute patellar tilt in the horizontal and vertical testing rigs. The data were averaged across specimens at each flexion angle, and the values were normalized such that zero tilt corresponded to the tilt in the neutral configuration at 15°. Minimum standard deviations in the horizontal and vertical rigs were 1.8° and 1.5°, respectively. Maximum standard deviations in the horizontal and vertical rigs were 10.6° and 3.9°, respectively. Horizontal Rig Vertical Rig Lateral Lateral Tibial Rotation Flexion angle (' Tibial Rotation External Internal I 1 T " i, 30 45 60 75 90 Flexion angle (°) 105 Figure 4.7: Effects (mean change ± SD) of ±5° rotation of the tibial component on relative patellar tilt (compared to neutral baselines) in the horizontal and vertical testing rigs. The data were averaged across specimens at each flexion angle. An asterix (*) indicates that a particular phase of flexion (early, mid- or late) showed statistically significant changes in tilt. A dagger (f) indicates that when comparing external and internal component rotation, there was a statistically significant difference in the change in patellar tilt. 104 Horizontal Rig Tibial Rotation Vertical Rig Tibial Rotation 30 45 60 75 90 105 Flexion Angle (°) 30 45 60 75 90 105 Flexion Angle (°) Figure 4.8: Effects of ±5° rotation of the tibial component on absolute patellar shift in the horizontal and vertical testing rigs. The data were averaged across specimens at each flexion angle, and the values were normalized such that zero shift corresponded to the shift in the neutral configuration at 15°. Minimum standard deviations in the horizontal and vertical rigs were 3.3 mm and 0.8 mm, respectively. Maximum standard deviations in the horizontal and vertical rigs were 8.2 mm and 5.1 mm, respectively. Lateral -t Horizontal Rig Tibial Rotation 30 45 60 75 Flexion angle (°) 90 105 W ? -2 JS d) a: -4 Vertical Rig Lateral Tibial Rotation External Internal 30 45 60 75 90 Flexion angle (°) 105 Figure 4.9: Effects (mean change ± SD) of ±5° rotation of the tibial component on relative patellar shift (compared to neutral baselines) in the horizontal and vertical testing rigs. The data were averaged across specimens at each flexion angle. An asterix (*) indicates that a particular phase of flexion (early, mid- or late) showed statistically significant changes in shift. A dagger (f) indicates that when comparing external and internal component rotation, there was a statistically significant difference in the change in patellar shift. 105 4.1.3 Patellar Resection Angle In the horizontal rig, a medial resection angle of 7.5° (i.e. thicker on the medial side) resulted in an average increase in lateral tilt of 6.0° in the horizontal rig and 6.6° in the vertical rig (Figure 4.10, Figure 4.11). A 7.5° lateral resection angle resulted in medial tilt measuring 4.5° in the horizontal rig and 6.3° in the vertical rig. The lateral angle of 15° increased the medial tilt by an average of 9.9° in the horizontal rig and 13.0° in the vertical rig. A l l changes in tilt were statistically significant throughout flexion in both testing rigs (p < 0.005). The average shifts resulting from the aforementioned resection angles (7.5° medial, 7.5° lateral, and 15° lateral) measured 1.8 mm lateral, 1.6 mm medial and 3.4 mm medial respectively in the horizontal rig (Figure 4.12, Figure 4.13). The shift due to the medial angle was statistically significant throughout flexion, the shift due to the 7.5° lateral angle was only significant in late flexion, and the shift due to the 15° lateral angle was significant in mid- and late flexion (p < 0.05). The corresponding average shifts in the vertical rig measured 1.7 mm lateral, 1.4 mm medial, and 3.9 mm medial (Figure 4.12, Figure 4.13). The shift due to the 7.5° medial resection angle was only statistically significant at 60° in the vertical rig, and the 7.5° lateral resection did not have a statistically significant effect on shift. In the vertical rig, the shift due to the medial angle was statistically significant throughout flexion, and the shift due to both lateral angles was significant in mid- and late flexion. Differences between Angles The changes in tilt for the three resection angles were all statistically different from each other throughout flexion in both testing rigs (p < 0.005). In the horizontal rig, the changes in shift for the three resection angles were all statistically different from each other throughout flexion (p < 0.05), with one exception: in early flexion, the changes in shift due to the 7.5° and 15° lateral wedges were not statistically different from each other. In the vertical rig, the changes in shift for the three resection angles were all statistically different from each other in mid- and late flexion, but were not statistically different in early flexion. Comparison between Testing Rigs The shape of the tilt curves were similar between testing rigs; medial tilt generally increased sharply between 15° and 45° and typically decreased gradually thereafter. Changes in tilt were generally comparable between testing rigs, although the lateral wedges resulted in greater 106 changes in tilt in the vertical rig than in the horizontal rig. The differences between the shift curves for the horizontal and vertical rigs were similar to those described above in section 4.1.1. Shift was more variable in the horizontal rig than in the vertical rig. Changes in shift were comparable between testing rigs; however, the 15° lateral wedge resulted in a greater change in shift in the vertical rig than in the horizontal rig. 107 Lateral Horizontal Rig Patelar Angle Lateral Vertical Rig Patelar Angle 0 15 30 45 60 75 90 105 Flexion Angle (°) 0 15 30 45 60 75 90 105 Flexion Angle (°) Figure 4.10: Effects of 7.5° medial, 7.5° lateral, and 15° lateral resection angles on absolute patellar tilt in the horizontal and vertical testing rigs. The data were averaged across specimens at each flexion angle, and the values were normalized such that zero tilt corresponded to the tilt in the neutral configuration at 15°. Minimum standard deviations in the horizontal and vertical rigs were 2.1° and 1.6°, respectively. Maximum standard deviations in the horizontal and vertical rigs were 10.6° and 4.2°, respectively. Lateral Horizontal Rig Patelar Angle > 30 45 60 75 90 Flexion angle (") Lateral Vertical Rig Patelar Angle 0 15 30 45 60 75 Flexion angle (°) 90 105 Figure 4.11: Effects (mean change ± SD) of 7.5° medial, 7.5° lateral, and 15° lateral resection angles on relative patellar tilt (compared to neutral baselines) in the horizontal and vertical testing rigs. The data were averaged across specimens at each flexion angle. An asterix (*) indicates that a particular phase of flexion (early, mid- or late) showed statistically significant changes in tilt. A dagger (|) indicates that the changes in tilt due to the 3 resection angles were statistically different from each other. 108 Lateral Horizontal Rig Patelar Angle 0 15 30 45 60 75 90 105 Flexion Angle (°) Lateral Vertical Rig Patelar Angle - i r~ --o- Med 7.5° -e- Lat 7.5° e- Lat 15° —e— Neutral -o 7.5 M -e o Neut -e © 7.5 L ^ e -e 15 L 0 15 30 45 60 75 90 105 Flexion Angle (°) Figure 4.12: Effects of 7.5° medial, 7.5° lateral, and 15° lateral resection angles on absolute patellar shift in the horizontal and vertical testing rigs. The data were averaged across specimens at each flexion angle, and the values were normalized such that zero shift corresponded to the shift in the neutral configuration at 15°. Minimum standard deviations in the horizontal and vertical rigs were 1.5 mm and 1.4 mm, respectively. Maximum standard deviations in the horizontal and vertical rigs were 8.2 mm and 5.1 mm, respectively. Horizontal Rig Vertical Rig Lateral Lateral Patelar Angle Patelar Angle Flexion angle (°) Flexion angle (°) Figure 4.13: Effects (mean change ± SD) of 7.5° medial, 7.5° lateral, and 15° lateral resection angles on relative patellar shift (compared to neutral baselines) in the horizontal and vertical testing rigs. The data were averaged across specimens at each flexion angle. An asterix (*) indicates that a particular phase of flexion (early, mid- or late) showed statistically significant changes in shift. A dagger (t) indicates that the changes in shift due to the 3 resection angles were statistically different from each other. 109 4.1.4 Patellar Medialization and Thickness Medialization: 2.5 mm A 2.5 mm medialization had a statistically significant effect on tilt throughout flexion in the horizontal rig, and in mid- and late flexion in the vertical rig (p < 0.05). It also had a significant effect on shift throughout flexion in both testing rigs (p < 0.05). Medializing the patellar component by 2.5 mm resulted in an average increase in lateral tilt of 2.6° in the horizontal rig and 1.7° in the vertical rig (Figure 4.14, Figure 4.15). This corresponded to average lateral shifts of 1.8 mm and 1.9 mm in the horizontal and vertical rigs, respectively (Figure 4.16, Figure 4.17). Medialization: 5 mm A 5 mm medialization of the patellar button increased the average lateral patellar tilt by 3.6° in the horizontal rig and 3.2° in the vertical rig. This effect was statistically significant in the horizontal rig in mid- and late flexion, and in the vertical rig throughout flexion (p < 0.05). The 5 mm medialization also shifted the patella an average of 3.9 mm laterally in both the horizontal and vertical rigs. This lateral shift was statistically significant over the entire flexion range in both rigs. The average changes in tilt and shift due to 5 mm medialization of the patellar component were approximately twice as great as the changes due to the 2.5 mm medialization. The changes in tilt due to the two different medializations were not statistically different from each other in the horizontal rig at any angle, but were statistically different throughout flexion in the vertical rig (p < 0.05). Changes in shift due to the two medializations were statistically different from each other in all phases of flexion (p < 0.01). Patellar Thickness Increasing the patellar thickness by 3 mm did not have a statistically significant effect on tilt or shift in either testing rig, with one exception: it altered tilt significantly in late flexion in the horizontal rig. Tilt averaged 1.1° lateral in the horizontal rig and 0.3° medial in the vertical rig. The patella was shifted an average of 0.3 mm laterally in the horizontal rig and 0.4 mm medially in the vertical rig. Comparison between Testing Rigs The shape of the tilt curves were similar between testing rigs; medial tilt generally increased between 15° and 45° and typically decreased gradually thereafter. The differences between the 110 shift curves for the horizontal and vertical rigs were similar to those described above in section 4.1.1. Changes in tilt and shift were comparable between testing rigs. Horizontal Rig Patelar Position and Thickness Vertical Rig Patelar Position and Thickness Lateral 5 mm 2.5 nhm Thick Neutj 15 30 45 60 Flexion Angle (°) —e-5 mm Med —e— 2.5 mm Med e Thick -e~ Neutral 75 90 105 8 4 0 -4 -8 -12 -16 -20 2.5 mm Neut Thick -e— 5 mm Med -e- 2.5 mm Med ••<> Thick -e- Neutral 15 30 45 60 75 Flexion Angle (°) 90 105 Figure 4.14: Effects of 2.5 mm medialization, 5 mm medialization, and 3 mm increased thickness on absolute patellar tilt in the horizontal and vertical testing rigs. The data were averaged across specimens at each flexion angle, and the values were normalized such that zero tilt corresponded to the tilt in the neutral configuration at 15°. Minimum standard deviations in the horizontal and vertical rigs were 2.2° and 1.9°, respectively. Maximum standard deviations in the horizontal and vertical rigs were 10.6° and 3.9°, respectively. Horizontal Rig Vertical Rig _ i i i i i i . 1 6 1 1 1 1 1 1 1 1 30 45 60 75 90 105 0 15 3D 45 BO 75 90 105 Flexion angle (°) Flexion angle f) Figure 4.15: Effects (mean change ± SD) of 2.5 mm medialization, 5 mm medialization, and 3 mm increased thickness on relative patellar tilt (compared to neutral baselines) in the horizontal and vertical testing rigs. The data were averaged across specimens at each flexion angle. An asterix (*) indicates that a particular phase of flexion (early, mid- or late) showed statistically significant changes in tilt. A dagger (f) indicates that the changes in tilt due to the 2 patellar medializations were statistically different from each other. I l l Horizontal Rig Vertical Rig Patelar Position and Thickness Patelar Position and Thickness o Lateral -e——— 5 mm Med 2.5 mm Med Thick Neutral 5 mm 2.5 mm Thick o Neut £ £ W -4 -6 Lateral 5 mm Med 2.5 mm Med Thick Neutral 15 30 45 60 75 90 105 Flexion Angle (") -8 o 5 mm 0 15 30 45 60 75 90 105 Flexion Angle (°) Figure 4.16: Effects of 2.5 mm medialization, 5 mm medialization, and 3 mm increased thickness on absolute patellar shift in the horizontal and vertical testing rigs. The data were averaged across specimens at each flexion angle, and the values were normalized such that zero shift corresponded to the shift in the neutral configuration at 15°. Minimum standard deviations in the horizontal and vertical rigs were 1.3 mm and 1.1 mm, respectively. Maximum standard deviations in the horizontal and vertical rigs were 8.2 mm and 5.1 mm, respectively. Lateral Horizontal Rig Patelar Positon and Thickness — i 1— 1 5 mm Med 2.5 mm Med Thickness 3D 45 60 75 Flexion angle (°) 9D 105 S a o; -4 S i l l -Vertical Rig , .Patelar Positon and Thickness H I 5 mrn Med • i 2 5 mm Med I I Thicl-ness 15 X 45 60 75 90 Flexion angle (°) 105 Figure 4.17: Effects (mean change ± SD) of 2.5 mm medialization, 5 mm medialization, and 3 mm increased thickness on relative patellar shift (compared to neutral baselines) in the horizontal and vertical testing rigs. The data were averaged across specimens at each flexion angle. An asterix (*) indicates that a particular phase of flexion (early, mid- or late) showed statistically significant changes in shift. A dagger (t) indicates that the changes in shift due to the 2 patellar medializations were statistically different from each other. 112 4.2 Ranking of Variable Effects on Kinematics In both testing rigs, the surgical variables that had the greatest effect on patellar shift and tilt varied according to the flexion angle. Throughout flexion, internal tibial component rotation and patellar thickness had very little effect on tilt or shift. 4.2.1 Tilt At 15° flexion, external femoral component rotation and the patellar resection angles (15° lateral, 7.5° lateral, and 7.5° medial) had the greatest effects on patellar tilt in both testing rigs (Figure 4.18). In mid-flexion, at 45°, the ranking of effects on tilt was similar: external femoral component rotation and the patellar resection angles had the greatest effects on tilt. In the horizontal rig, the effect of 5 mm medialization of the patellar implant was comparable in magnitude to the effects of external femoral component rotation and the 7.5° lateral resection angle. In the vertical rig, internal femoral component rotation had an effect which was opposite but comparable in magnitude to internal femoral component rotation. At 90° flexion, the ranking of effects on tilt was different than in early and mid-flexion: external femoral component rotation ceased to be an important effect. The patellar resection angles continued to have the greatest effects on tilt. The effect of 5 mm medialization of the patellar implant also had a visible effect on tilt in both testing rigs. In the vertical rig, internal femoral component rotation had a noticeable effect on tilt. In general, tibial component rotation and patellar thickness had comparatively little effect on changes in patellar tilt. The p-values for the paired t-tests comparing changes in tilt between neutral baseline measurements and each surgical variable appear in Appendix 7. Appendix 7 also lists the phases of flexion (early, mid-, or late) at which significant differences existed between variables. 4.2.2 Shift At 15° flexion, femoral component rotation and 5 mm button medialization had the greatest effects on patellar shift in both testing rigs (Figure 4.19). At 45° flexion, the ranking of effects on shift was quite different than in early flexion. The 15° lateral resection angle and the 5 mm button medialization had the greatest effects on patellar shift. At 90° flexion, the ranking of effects on shift was similar to mid-flexion. The 15° lateral resection angle and the 5 mm button medialization continued to have the greatest effects on patellar shift. The 7.5° lateral and medial resection angles and the 2.5 mm button medialization had effects of comparative magnitude on shift. The effects of femoral component rotation decreased with increasing flexion angle. In 113 general, tibial component rotation and patellar thickness had comparatively little effect on changes in patellar shift. Appendix 7 gives the p-values for the paired t-tests comparing changes in tilt between neutral baseline measurements and each surgical variable; Appendix 7 also lists the phases of flexion (early, mid-, or late) at which significant differences existed between variables. 114 10 Lateral 8 6 4 tr 2 ? 0 -2 -4 -6 -8 10 12 14 0) > re <u a: Medial -16 Femur Ext Int Relative Tilt, Horizontal Rig Tibia I Patella 9 TT it it it Ext Int 1 VL M Patella Med. it * T -r- * T H15° • 45° • 90° 2.5 Thick. Lateral r-0) > « o Medial 10 8 6 4 2 0 -2 -4 -6 -8 •10 •12 •14 16 Femur it it it Relative Tilt, Vertical Rig Tibia i Patella 9 S | I T ft it it it it it Ext Int 1 Ext Int 1 VL L M 1 2.5 Patella Med. ititit • 15° H45° • 90° Thick. Figure 4.18: Effects (mean ± SD) of all tested surgical variables on changes in patellar tilt (compared to neutral component placement) at 3 flexion angles in the horizontal and vertical rigs. The first two bars represent femoral component rotation: external and internal. The third and fourth bars represent tibial component rotation: external and internal. The next three bars represent the patellar resection angles: 15° lateral (VL, very lateral), 7.5° lateral (L), and 7.5° medial (M). The next two bars represent 2.5 mm and 5 mm medialization of the patellar component. The last bar represents the 3 mm additional patellar thickness. This pattern also applies to Figure 4.19 below. 115 Femur Relative Shift, Horizontal Rig Tibia . Patella 9 Patella Med. 4 E 2 E 0 w ive -2 re a> -4 01 Ext Int Ext Int •A1 it • 15° • 45° • 90° VL M 2.5 Femur Relative Shift, Vertical Rig Tibia . Patella 6 Patella Med. .Thick. Figure 4.19: Effects (mean ± SD) of all tested surgical variables on changes in patellar shift (compared to neutral component placement) at 3 flexion angles in the horizontal and vertical rigs. 116 4.3 Comparison of Kinematics between Horizontal and Vertical Rigs To test whether or not the results in the intraoperative simulation (horizontal rig) predicted those in the postoperative simulation (vertical rig), we compared both absolute tracking profiles and changes from baseline between the two rigs. We did not find any statistically significant differences (p > 0.05) between changes in patellar tilt, shift or spin in the horizontal and vertical rigs at any flexion angle based on a two-way repeated-measures A N O V A (described in section 2.9.1). To test whether changes in component placement had similar effects in both rigs, we also determined Pearson's correlation (the Pearson Product Moment Correlation) between the data from the horizontal and vertical rigs for each flexion angle, averaged across all specimens. First, we determined the correlation between the differences in tracking due to the changes in surgical variables (Figure 4.20). The changes in tilt and shift were highly correlated between testing rigs (p < 0.001); however, changes in spin were not highly correlated between rigs (p > 0.3). Table 4.1 lists the p-values for the correlation in differences in tracking due to altered surgical variables between testing rigs. Each p-value is the probability of calculating a correlation as large as the observed value by random chance, when the true correlation is zero. The high values of Pearson's correlation for tilt and shift reflects that there is a strong linear relationship between changes in those tracking indices for the two testing rigs (a correlation of +1 indicates a perfect positive linear relationship between variables). Figure 4.21 shows the average slopes of the best-fit lines through the horizontal and vertical tracking data (changes with respect to baseline) at each flexion angle. Table 4.2 lists the slopes and intercepts of the best-fit lines. 117 Correlation between Changes in Horizontal and Vertical Tracking 15 30 45 60 75 90 Flexion Angle (°) Figure 4.20: Correlation between changes in tracking (due to altered surgical variables) in the horizontal and vertical rigs. The correlation coefficients were averaged across specimens. Correlations for individual specimens ranged between 0.57 and 0.94 for tilt, 0.47 and 0.94 for shift, and -0.13 and 0.62 for spin. Bars marked with an asterix (*) indicate a statistically significant correlation. Table 4.1: The p-values for the Pearson's correlation coefficients comparing the changes in data between testing rigs. Bold values indicate statistical significance. p-value Flexion Angle Tilt Shift Spin 15° < 0.001 < 0.001 0.41 30° < 0.001 < 0.001 0.38 45° < 0.001 < 0.001 0.63 60° < 0.001 < 0.001 0.93 75° < 0.001 < 0.001 0.73 90° < 0.001 < 0.001 0.59 118 C <D m *-o o a o w a) a> ra i _ > < 15 30 45 60 Flexion Angle (°) 75 90 Figure 4.21: Slopes of the best-fit lines through the horizontal and vertical data (changes with respect to baseline) at each flexion angle (averaged across specimens). A slope of less than 1 indicates that changes in the vertical rig were smaller than in the horizontal rig. Table 4.2: Average values of the slopes and intercepts of the best-fit lines through the horizontal and vertical tracking data (changes with respect to baseline, averaged across all specimens and flexion angles). The ranges for individual specimens are given in parentheses. Tilt Shift Spin Average slope (across all angles) 0.76 (0.33, 1.06) 0.85 (-0.03,1.74) 0.61 (-0.52,6.51) Average horizontal intercept (across all angles) -0.7° (-5.0°, 2.1°) 0.5 mm (-1.7 mm, 3.4 mm) -0.2° (-4.2°, 4.3°) Average vertical intercept (across all angles) 1.4° (-2.5°, 15.2°) -5.2 mm (-184.5 mm, 16.2 mm) 3.4° (-49.0°, 364.0°) 119 Figure 4.22 shows examples of the linear relationships between the tracking indices in the two rigs at 45° flexion. re o t > E E £ "to o '€ > !E w < Shift •40-D) hi re o r Q> > 'a < A Tilt, Horizontal Rig (°) Spin A Shift, Horizontal Rig (mm) • -4 • -8 A Spin, Horizontal Rig (°) Figure 4.22: The linear relationships between changes in tilt, shift and spin in the horizontal and vertical rigs at 45° flexion. Data for all trials with each specimen have been plotted together, although the analysis was performed for individual specimens. We also determined the correlation between the raw tracking data collected using both testing rigs (Figure 4.23). Although tilt showed good correlation between testing rigs across all flexion angles, this correlation was only statistically significant at 15°, 30° and 60° (p < 0.05). Shift was less correlated between rigs at lower flexion angles than high flexion angles, and this correlation was only statistically significant between 45°-90° (p<0.05). Spin was not highly correlated between testing rigs, and the correlation was not statistically significant at any flexion angle (p<0.05). Table 4.3 shows the p-values for the Pearson's correlation coefficients comparing the raw tracking data between testing rigs. The average slopes of the best-fit lines through the raw 120 tracking data from the horizontal and vertical rigs were similar to the slopes for the changes in tracking, but slopes were more variable between specimens. Correlation between Absolute Values of Horizontal and Vertical Tracking -0.1 J— — 15 30 45 60 75 90 Flexion Angle (degrees) Figure 4.23: Correlation between raw tracking data for the horizontal and vertical rigs. The correlation coefficients were averaged across specimens. Correlations for individual specimens ranged between 0.470 and 0.95 for tilt, 0.52 and 0.95 for shift, and -0.17 and 0.65 for spin. Bars marked with an asterix (*) indicate a statistically significant correlation. Table 4.3: The p-values for the Pearson's correlation coefficients comparing the raw tracking data between testing rigs. Bold values indicate statistical significance. p-value Flexion Angle Tilt Shift Spin 15° 0.035 0.277 0.442 30° 0.044 0.081 0.557 45° 0.058 0.038 0.597 60° 0.048 0.020 0.417 75° 0.053 0.016 0.321 90° 0.057 0.017 0.303 121 4.4 Additional Tests: Kinematics In addition to the aforementioned surgical variables, we also tested the effects of additional variables at the end of testing to determine their influence on patellar mechanics. We did not perform these additional tests on all specimens (see Appendix 3). Due to the small number of tests performed, we did not perform statistical analyses of the extra tests described herein. To determine the dependency of kinematic measurements on flexion speed, we increased the speed of flexion substantially in the horizontal rig by flexing and extending the knee in approximately 3 seconds instead of 30 seconds. Increasing flexion speed had very little consistent effect on tilt, shift or spin (Figure 4.24-Figure 4.26, respectively). In the vertical rig, following all other tests, we performed a lateral release and measured its effects on tracking. The release did not have an effect on patellar tracking (Figure 4.24-Figure 4.26). We performed tests without the Tekscan sensor in place to determine its effect on tracking. The removal of the Tekscan sensor appeared to increase lateral tilt in both testing rigs (Figure 4.24). In the horizontal rig, this was also accompanied by medial shift; however, in the vertical rig, the patella was shifted laterally (Figure 4.25). In the horizontal rig, the absence of the sensor created lateral spin, whereas in the vertical rig this resulted in medial spin (Figure 4.26). In both testing rigs, we measured the effect of removing the surgical towel clamps which were used to close the medial incision. The absence of the clamps did not have a noticeable effect on tilt (Figure 4.24) or shift (Figure 4.25); however, the lack of clamps seemed to induce lateral spin in both the horizontal and vertical testing rigs (Figure 4.26). 122 12 10 8 C 6 *-i H 0) 4 > re a: 2 0 -2 -4 Tilt, Horizontal Rig Tilt, Vertical Rig Lateral • Fast • No Sensor • No Clamp ----- - -r-f 15 30 45 60 75 90 105 Flexion Angle (°) ^ 3^FIexfon Ang?e ( ° ) 7 5 ^ Figure 4.24: The effects of additional variables on patellar tilt, compared to baseline, in the horizontal and vertical rigs. In the horizontal rig, N = 6 for fast flexion and extension, N = 4 for trials with no Tekscan sensor in place, and N = 4 for trials performed without surgical towel clamps to close the medial incision. In the vertical rig, N = 3 for the lateral release, N = 3 for trials with no Tekscan sensor in place, and N = 3 for trials performed without surgical towel clamps to close the medial incision. See Appendix 3 for information on the specimens tested. 4 3 E 2 E, is 1 !c CO Jo re a> *-1 -2 -3 Shift, Horizontal Rig Lateral • Fast • No Sensor • No Clamp 15 30 45 60 75 90 105 Flexion Angle (°) 4 3 E 2 E, £ 1 !c CO I o re 0) Di.1 Shift, Vertical Rig -2 -3 Lateral • Lateral Release • No Sensor • No Clamp 4 . * 15 30 45 60 75 Flexion Angle (°) 90 Figure 4.25: The effects of additional variables on patellar shift, compared to baseline, in the horizontal and vertical rigs. 123 Spin, Horizontal Rig Lateral • Fast • No Sensor • No Clamp 15 30 45 60 75 90 105 Flexion Angle (°) Spin, Vertical Rig Lateral • Lateral Release • No Sensor • No Clamp 15 30 45 60 75 90 Flexion Angle (°) Figure 4.26: The effects of additional variables on patellar spin, compared to baseline, in the horizontal and vertical rigs. We also took static measurements of tracking in both rigs at several angles to determine whether static measurements are accurate assessments of patellar motion; our measures of tracking were not greatly affected by measuring tracking statically instead of dynamically. Table 4.4 shows the changes in tracking indices (averaged across the individual flexion angles) due to the static nature of the motion measurement. It appears that there was no difference in tracking: the standard deviations were always greater than the difference in each tracking index. Table 4.4: Changes in patellar tilt, shift and spin (compared to dynamic baseline measurements) when data was recorded at static flexion angles as opposed to dynamic flexion angles. Lateral tilt, shift and spin were positive. In the horizontal rig, we took a total of 15 static measurements with 5 specimens. In the horizontal rig, we took a total of 4 measurements with 4 specimens. A Tilt AS riift AS pin Horizontal Rig Vertical Rig Horizontal Rig Vertical Rig Horizontal Rig Vertical Rig -0.8° ±0.9° -0.2° ±0.5° 0.2 mm ± 0.4 mm -0.1 m m ± 1.2 mm 0.1° ±0.5° -0.2° ±0.9° 124 4.5 Patellar Contact Forces The loads acting on the patella were measured using a Tekscan pressure sensor glued to the patellar button, as described in section 2.7.1. A l l forces reported herein are from data that were calibrated using the least-squares algorithm described in section 3.2. Figure 4.27 shows the reported contact force for 2 baseline measurements in the horizontal rig. Forces were generally slightly higher during flexion than extension in the horizontal rig. 40 60 80 100 Flexion Angle (°) 140 Figure 4.27: Patellar contact forces measured by the Tekscan pressure sensor for two consecutive baseline trials in the horizontal rig. These data are for specimen 7. In the vertical rig, the loads were approximately 1.5-2 times larger in extension than flexion (Figure 4.28). Patellar contact forces in the vertical rig were typically 3-6 times higher in the vertical rig than in the horizontal rig (Figure 4.29). The average contact forces measured during the first set of baseline tests in the horizontal and vertical rigs were 52±22 N and 253±101 N , respectively. 125 300 40 60 Flexion Angle {") Figure 4.28: Patellar contact forces measured by the Tekscan pressure sensor, repetitions of a baseline measurement with specimen 7 in the vertical rig. 100 These data are for two 450 -r 400 z 350 o 300 o UL 250 u re •*-> c 200 o O 150 J * re o 100 CL 50 0 • Horizontal Rig • Vertical Rig 4 5 6 Specimen Number 8 Figure 4.29: Peak contact forces for the First baseline trials in the horizontal and vertical rig. Data from specimen 1 were not analyzed due to sensor degradation and loss of rows and columns of sensor output (see Appendix 5). Data from specimen 3 in the vertical rig were not analyzed because the sensor punctured, and we did not conduct tests with specimen 6 in the vertical rig because of problems with femoral dislocation (see Appendix 3). 126 Although the forces measured by the Tekscan sensor were generally repeatable between the two repetitions of each baseline or variable test (Figure 4.28), there was considerable variability in the shape of the loading pattern for different variables (Figure 4.30). Also, the location of the peak load fluctuated between tests; it did not consistently occur at high flexion angles. Because force magnitude showed very little visible correlation or trend with flexion angle, there was no logic in extracting force data at several angles for comparison to the kinematic data. We therefore extracted the peak force for each flexion/extension cycle. 80 Flexion Angle (°) Figure 4.30: Flexion/extension contact force for three different surgical variables (specimen 5, vertical rig). The shapes of the curves are visibly different, and the peak loads occur at different flexion angles. As described in section 2.7.2.1, Tekscan pressure sensors exhibited considerable degradation during testing. Figure 4.31 shows the degradation in reported force after one subgroup of test variables (patellar resection angle). The baseline measurements shown were separated by 6 flexion cycles (3 patellar resection angles x 2 trials each). The procedure used to normalize degraded output is described in section 2.9.2. Force data were excluded from the analysis i f the scaling factor used to normalize the data was greater than 2 (i.e. there was more than 50% degradation in reported force). 127 100 Flexion Angle (°) Figure 4.31: The degradation in reported force following 6 patellar resection angle tests (3 variables x 2 trials each). The red curve is a baseline measurement and the blue curve is a second baseline measurement taken following the subgroup of resection angle tests. These data are for specimen 5; trials with specimens 7 and 8 showed less degradation in sensor output than previous specimens. 4.6 Ranking of Variable Effects on Forces Using a two-way A N O V A with repeated measures, we determined that no changes in peak contact force were statistically significant in either testing rig (p > 0.05). In the horizontal testing rig, adding an additional 3 mm thickness to the patella had the greatest apparent effect, on average, on the peak force reported by the Tekscan pressure sensors; the peak force increased 18±22% (Figure 4.32). Internal tibial component rotation resulted in a 9±8% reduction in peak loads. A l l other variables had less than a 10% average effect on peak patellofemoral contact loads in the horizontal rig. The standard deviations in peak load (across all specimens) ranged between 8% and 27%. In the vertical testing rig, medializirig the patellar button by 5 mm had the greatest apparent effect, on average, on the peak force reported by the Tekscan pressure sensors; it decreased the reported peak load by 22±32% (Figure 4.32). The 15° lateral resection angle decreased the peak load by 12±25%, and externally rotating the femoral component increased the peak load by 11±1%. A l l other variables had less than a 10% effect on peak patellofemoral contact loads in 128 the vertical rig. The standard deviations in peak load (across all specimens) ranged between 1% and 32%. CD O i_ o u_ J * 03 O Q. 0) O) c re .c O 50 40 30 20 10 0 -10 -20 -30 g -40 I "50 -60 Femur Ext Int Tibia Ext Int Patella 6 I Horizontal I Vertical VL M Med. 2.5 Thick. Figure 4.32: Effects (mean ± SD) of all tested surgical variables on peak patellar contact force (compared to neutral component placement) in the horizontal and vertical rigs. The first two bars represent femoral component rotation: external and internal. The third and fourth bars represent tibial component rotation: external and internal. The next three bars represent the patellar resection angles: 15° lateral (VL, very lateral), 7.5° lateral (L), and 7.5° medial (M). The next two bars represent 2.5 mm and 5 mm medialization of the patellar component. The last bar represents the 3 mm additional patellar thickness. 4.7 Comparison of Forces between Horizontal and Vertical Rigs Using a two-way A N O V A with repeated measures, we determined that the changes in contact force measurements in the horizontal and vertical rigs were not statistically different from each other (p > 0.05); however, the power of this A N O V A was very low (0.050) due to the small number of specimens available for force comparisons between rigs (only 5). The average correlation coefficient (r) between changes in peak forces in the horizontal and vertical rigs was 0.59 (range 0.39 to 0.72), and this value was not statistically significant (p = 0.130). The average slope of the best-fit line through the horizontal and vertical data was 0.74 (range 0.26 to 1.09), the average horizontal intercept was 1.6% (range -3.4% to 11.4%), and the average vertical intercept was -2.4% (range -12.5% to 3.1%). Figure 4.33 shows a typical relationship between changes in peak force in the horizontal rig and changes in peak force in the vertical rig. 129 25~0~ 50r9~ 25.0 %A in Peak Contact Force, Horizontal Rig Figure 4.33: Linear relationship between changes in peak forces in the horizontal and vertical rigs for specimen 7. For both testing rigs, external and internal femoral component rotation increased peak force above baseline values. Tibial component rotation had opposite effects on contact forces in the two rigs; in the horizontal rig, external tibial component rotation increased forces and internal rotation decreased forces, whereas in the vertical rig, external rotation decreased forces and internal rotation increased forces. A consistent trend was observed in both rigs with respect to patellar resection angle: patellae with more medial resection angles (i.e. thicker medial sides) experienced greater peak forces. Medializing the patella had little effect on forces in the horizontal rig; however, medializing the patella in the vertical rig decreased forces. Increased patellar thickness increased forces in both testing rigs. None of the aforementioned effects on peak contact forces were statistically significant (p > 0.05). 4.8 Additional Tests: Forces We tested the effects of additional variables to determine their influence on patellar kinematics (section 4.4) and peak contact forces. As described in section 4.4, we did not perform statistical analyses of the extra tests. 130 To determine the dependency of force measurements on flexion speed, we increased the speed of flexion substantially in the horizontal rig by flexing and extending the knee in approximately 3 seconds instead of 30 seconds. Faster flexion cycles reduced the contact force by 16.6 ± 8.4% (N = 5). We were only able to obtain useful measurements of force following lateral release for one specimen: specimen 7 (see Appendix 5). A lateral release resulted in a 1.3% decrease in force for this specimen. A lateral release could only be performed following all other tests so as not to bias our other tests, and the sensor output was typically degraded at this point. To test the effect of the surgical towel clamps on kinetics and kinematics, we removed the clamps and performed two flexion cycles according to the usual testing protocol. In the vertical rig, the lack of clamps reduced contact forces, on average, by 6.4 ± 3.0% (N = 2). In the horizontal rig, the forces were reduced by 4.7 ± 8.3% (N = 4). To determine whether static measurements (such as those taken using Fuji film) are accurate assessments of patellar contact loads, we measured contact loads with the knee flexed and stationary at several flexion angles. We compared these static contact loads to the loads measured at the same flexion angles during dynamic flexion for specimens 6 and 7 in the horizontal rig. For the other specimens, the sensors were either too degraded or there was only one data point to compare (see Appendix 5). There was no consistent increase or decrease in measured force when comparing static and dynamic measurements for the two specimens (Figure 4.34). The maximum percentage differences between the static and dynamic measurements for specimens 6 and 7 in the horizontal rig were 20% and 22%, respectively. 131 Figure 4.34: Comparison of static and dynamic measurements of contact force for specimens 6 and 7 in the horizontal rig. 4.9 Contact Area and Pressure Although the Tekscan pressure sensors measured contact area and pressure as well as force, we could not report quantitative changes in area or pressure due to limitations in the measurements of contact area. The output data was very discrete and volatile due to the spatial resolution of the sensor (Figure 4.35); the contact areas were generally small (on the order of 5-10 mm2) compared to the area of one sensel (1.6 mm2). Often, only 10-20 sensels registered loads above the sensor's threshold. Figure 4.36 shows the small medial and lateral contact areas displayed using the Tekscan software. Also, the areas reported by the Tekscan sensors decreased over the course of testing as the sensor degraded. The pressure data was calculated using the measured force and area and was therefore very sensitive to small changes in area, since the contact areas were so small (Figure 4.35). As a result, the pressure data was extremely unreliable and could not be used. In general, increases in measured contact force were accompanied by increases in reported contact area (Figure 4.35). 132 Contact Pressure 6 0 i , 1 , 1 0 20 40 60 80 100 Flexion Angle (°) Figure 4.35: Measured contact force, measured contact area, and calculated contact pressure for two baseline trials in the vertical rig. Pressure is calculated using the measured force and area. The baseline trials were the first and last performed on specimen 5 and show the degradation in measured force and area over the course of testing. 133 Figure 4.36: Medial and lateral contact areas displayed using the Tekscan software. The force and contact areas reported by the sensor are the net force and area; i.e., the reported normal force would be the sum of Fmedial and F|ateral- It should be noted that the net force reported over the whole Tekscan sensor is the scalar sum of the medial and lateral forces and not a vector sum. 4.10 Quadriceps Tension We measured the quadriceps force in the horizontal rig for all trials with specimen 7 and two trials with specimen 8. The peak force occurred at maximum flexion and consistently measured approximately 86 N (Figure 4.37). Quadriceps tension was the same for flexion and extension in the horizontal rig. 134 90 80 70 _ 60 Baseline - 1 30 20 10 Lateral Angle 15° Lateral Angle 7.5° — Medial Angle 7.5° 0 0 2 4 6 8 Time (s) 10 12 14 Figure 4.37: Tension in the quadriceps tendon for four trials (one baseline and three changes in patellar resection angle) in the horizontal rig for specimen 7. Flexion did not occur at exactly the same rate between trials. 4.11 Summary Tracking: Effects of Surgical Variables 1. External rotation of the femoral component increased lateral patellar tilt. The patella shifted laterally in early flexion and more medially in late flexion. 2. Internal rotation of the femoral component tilted and shifted the patella medially. In the horizontal rig, the patella tilted more laterally as the flexion angle increased. 3. External and internal tibial component rotation did not result in consistent changes in patellar tilt or shift compared to neutral placement. 4. A medial resection angle of 7.5° tilted and shifted the patella laterally and increased lateral tilt by 6.0°-6.6°. 5. Lateral resection angles of 7.5° and 15° tilted and shifted the patella medially. The tilt and shift due to the 15° resection angle were approximately twice the values measured for the 7.5° angle. 6. Medializing the patellar component increased lateral tilt and lateral shift of the patella. 7. Increasing the patellar thickness had no effect on either tilt or shift. 8. There were very few statistically significant differences in spin due to changes in component placement, and these differences only occurred in late flexion. 135 Tracking: Ranking of Effects of Surgical Variables (Not Statistical) 1. In early and mid-flexion, external femoral component rotation and all patellar resection angles had the greatest effects on patellar tilt. In late flexion, the patellar resection angles had the greatest effects on tilt. 2. In early flexion, femoral component rotation and 5 mm button medialization had the greatest effects on patellar shift. In mid-flexion and late flexion, the 15° lateral resection angle and 5 mm button medialization had the greatest effects on patellar shift. 3. Considerable variability in tracking was observed between specimens. Tracking: Comparison between Testing Rigs 1. Changes in patellar tilt, shift and spin were not found to be statistically different in the horizontal and vertical rigs at all flexion angles. A strong linear relationship exists between changes in tilt and shift for the two testing rigs; however, the relationship is not highly linear for spin. 2. The average slopes of the lines through the horizontal and vertical data (changes with respect to baseline) were 0.76, 0.85, and 0.61 for tilt, shift and spin. 3. Absolute tracking data were correlated for tilt and shift but not spin; however, this correlation was not statistically significant across all flexion angles. Tracking: Effects of Additional Variables 1. Increasing the speed of flexion in the horizontal rig had little consistent effect on tilt, shift or spin. 2. Leaving the medial incision open (by removing the surgical towel clamps used to close it) the medial incision did not appear to affect patellar tilt or shift, but it increased lateral spin. 3. Performing flexion cycles without the Tekscan sensor glued to the patella tended to increase lateral tilt. In the horizontal rig, the absence of the sensor appeared to induce medial shift and lateral spin, whereas it induced lateral shift.and medial spin in the vertical rig. 4. The performance of a lateral release did not seem to have an effect on patellar tracking. 5. Static measurements of tracking were very similar to dynamic measurements at the same flexion angle. 136 Contact Forces 1. Patellar contact forces were slightly higher during flexion than extension in the horizontal rig; in the vertical rig the forces were much higher in extension than in flexion. Forces in the vertical rig were 3-6 times higher in the vertical rig than in the horizontal rig. 2. No changes in peak contact force were statistically significant in either testing rig. The addition of a 3 mm thickness to the patellar construct resulted in the greatest increase in peak forces in the horizontal rig, and internal tibial component rotation resulted in the greatest decrease in peak forces. In the vertical rig, externally rotating the femoral component resulted in the greatest increase in peak forces, and medializing the patellar button by 5 mm resulted in the greatest decrease in peak forces. 3,. Peak contact forces were highly variable between specimens. 4. Changes in peak contact force measurements in the horizontal and vertical rigs were not statistically different from each other; however, they were also not highly correlated. 5. Removing the surgical towel clamps used to close the medial incision tended to reduce patellar contact loads. 6. Fast flexion cycles appeared to reduce contact forces. 7. A lateral release had little effect on contact forces. 8. Static positioning of the knee did not have a consistent effect on the measurement of contact loads, compared to dynamic positioning at the same flexion angle. Quadriceps Tension 1. In the horizontal rig, the peak force in the quadriceps tendon measured approximately 86 N and was consistent between trials and specimens. 137 5 Discussion 5.1 Introduction Despite the increase in successful TKAs, the patellofemoral joint continues to be an important source of pain and complications following surgery. Surgeons attempt to optimize knee component placement for postoperative performance but can only judge patellofemoral mechanics by observing passive knee flexion intraoperatively. To our knowledge, this was the first study to compare patellar tracking and loading in passive intraoperative and loaded postoperative simulations. We assessed the relationship between positioning of the femoral, tibial and patellar components on patellar kinematics and kinetics in cadaver specimens. This study was the first to investigate the effects of modifying several surgical variables on the same specimen. The following sections compare our findings for each surgical variable to the results of other studies. 5.2 Tracking: Neutral Component Placement The shapes of the patellar tilt and shift curves for TKAs with all components positioned neutrally vary between studies. The studies by Lee (1999), Rhoads (1990) and Yoshii (1992) showed that patellar tilt remained steady or increased gradually throughout flexion, whereas Armstrong (2003) and Nagamine (1994) found that tilt decreased during flexion (positive tilt is defined as being lateral, where the lateral half of the patella tilts toward the femoral condyle). The studies performed by Chew (1997) and Miller (2001a, 2001b) showed a sharp increase in lateral tilt in early flexion followed by a decrease or plateau. We observed an increase in medial tilt in early flexion, followed by a gradually decreasing medial tilt. When the patella became well-seated within the femoral groove (between 30° and 45°), the patellar tilt stabilized as its motion was largely constrained by the sides of the femoral groove. Patellar tilt was similar in both the horizontal and vertical rigs. Despite differences in loading configurations, the shapes of the shift curves reported by Rhoads (1990) and Yoshii (1992) were similar to the shift curve we observed in the horizontal rig: the patella shifted medially very gradually or remained steady. One study by Miller (2001b) yielded a very erratic shift pattern which increased and decreased dramatically throughout flexion, and Armstrong (2003) observed steady medial shift during flexion. Chew (1997) reported increasing lateral shift in early flexion and decreasing lateral shift in later flexion. In the vertical rig, we 138 observed an initial medial shift followed by increasing lateral shift; this corresponded visually to the shift reported by Anouchi (1993) and Nagamine (1994). The shape of the shift curve in the vertical rig was very similar to the shift curve during static lifting reported in the Zimmer product literature (Figure 5.1) (Zimmer, 2004). Differences in the magnitude of shift between our findings and those in the product literature may be due to inter-specimen variability; it is not clear whether the Zimmer tests were performed on one specimen or several. Also, the product literature did not provide details of the testing procedure and did not specify i f the reported data was for flexion or extension. The medial shift in early flexion was likely the translation necessary for the patella to find the femoral groove; as the patella entered the groove, it mainly remained central as it followed the path of the femoral groove. Medial Translation (mm) < — — 15 10 5 0 -5 -10 Medial Translation (mm) o NexGen Posterior Stabilized Knee Figure 5.1: Patellar translation (shift) during static lifting as reported by Zimmer (2004), and patellar shift (baseline tests) in the vertical rig as reported in the current study. 139 The differences in baseline tilt and shift measurements across studies likely reflect the differences in loading configurations, the shape of the femoral and patellar implants (Yoshii 1999), and the direction of the femoral groove. The disparities may also be a function of inter-specimen variability; Table 5.1 shows the maximum standard deviations for tilt and shift across similar studies where it was reported or could be approximated from graphical results. The variability in tilt that we observed in the vertical rig was slightly less than the variability reported in other studies, and the variability in shift was comparable to that reported in other studies. The larger variability in tracking that we observed in the horizontal rig was a function of the low loads applied to the knee in an intraoperative simulation. Because we only applied enough tension to the quadriceps tendon force to keep the patella from sliding off the femur in extension and to simulate flaccid muscles, the tracking of the patella was not highly constrained and was thus more erratic than in the vertical rig. The difference in loading between rigs also explains the difference between the shapes of the shift curves. In the horizontal rig, there was no large quadriceps force pulling the patella laterally, so the patella did not have to shift medially to enter the femoral groove; in the vertical rig, the patella was initially more lateral due to the quadriceps force, and medial translation was necessary to enter the femoral groove. Table 5.1: Maximum standard deviations of patellar tilt and shift in similar post-TKA studies. N Std. Dev. Tilt (°) Std. Dev. Shift (mm) Applied Axial Load (N) Applied Quads Load (N) Our Study- Horizontal Rig 8 10.6 8.2 ~87 Our Study- Vertical Rig 7 3.9 5.1 80 ? Armstrong (2003) 7 ~8 ~6 — 80* Lee (1999) 10 ~8 — ? 200 Miller (2001a) 11 6.5 -- 44.5 ? Miller (2001b) 9 8.6 7.9 44.5 ? Nagamine (1994) 5 ~6 ^ 4 30 30 Yoshii (1992) 6 ~5 3.3 ? 100 * The load applied to the quadriceps was divided amongst the medial, lateral and central quadriceps groups. A portion of the variability in tracking between specimens may be attributed to the manner in which we registered the landmarks which defined the coordinate systems of the knee. It has 140 been shown that digitizing the lateral and medial epicondyles with a point probe such as the one used in our study results in some variability in identifying the transepicondylar (TE) axis (Bullock 2003, Fuiko 2003, Jerosch 2002, Kessler 2003). Bullock (2003) determined that the total variability in locating the centre of the TE axis using a point probe was 0.4 mm (standard deviation (SD)) -in the mediolateral direction and 2.8 mm (SD) in the anteroposterior direction. The total rotational variability in the location of the TE axis was 2.9° (SD). To estimate how much variance in tracking may be attributed to variance in digitizing the landmarks of the knee, we used the findings of Bullock (2003) to calculate approximate ratios of variance in tracking (the square of the SD) to variance in defining the axes of motion. The ratios of angular variance in defining the TE axis to the variance in patellar tilt were approximately 7% and 55% in the horizontal and vertical rigs, respectively. The ratios of variance in defining the TE axis in the mediolateral direction to the variance in patellar shift were approximately 0.2% and 0.6% in the horizontal and vertical rigs, respectively. The variance in tracking in the horizontal rig cannot be fully attributed to the variability in digitizing the landmarks of the knee; however, a large portion of the variability in patellar tilt in the vertical rig may possibly be attributed to intra-operator variability in defining the landmarks used to build the coordinate systems of the knee. 5.3 Femoral Component Rotation The patella has an intimate relationship with the femur; its motion is guided by the geometry of the femoral condyles and groove, and by the direction of pull of the quadriceps mechanism (i.e. the Q-angle). Rotating the femoral component changes both the Q-angle and the location of the track in which the patella moves. These changes are likely reflected in both patellar tracking and contact loads. Rotation of the femoral component has generally been shown to have an effect on patellar tracking and loading; however, the effect has been interpreted differently between studies. 5.3.1 Femoral Component Rotation: Tracking In a cadaver study using a closed-chain, vertical testing rig similar to ours, Anouchi (1993) determined that external rotation of the femoral component 5° relative to the posterior condyles caused lateral shift of the patella. External rotation also resulted in a shift pattern most similar to the natural knee. The knees were flexed with 135 kg of tension applied to the quadriceps tendon 141 and the applied axial loads were unknown. Despite differences in applied loads (they applied a smaller quadriceps tendon tension and did not apply an axial load), we found similar results comparing the effects of 5° external and internal rotation on patellar shift in early and mid-flexion (Table 5.2); however, we observed that the patella tracked medially in late flexion following internal rotation, whereas Anouchi reported that the patella remained laterally shifted. Our observed change in shift (from lateral to medial) is explained later in this section. Anouchi observed that external rotation resulted in smoother tracking with less sharp deviations during the first 60° of flexion; qualitatively, we did not observe differences in smoothness of tracking for any rotational position. The study by Anouchi was somewhat limited because they only tested 4 cadaveric specimens. They did not monitor the effects of component rotation on patellar tilt. Table 5.2: Summary of results of cadaveric studies on the effects of femoral component rotation on patellar tracking. Positive tilt and shift are lateral. 0 denotes the component rotation angle, and the reference column indicates what line was used to define neutral rotation (3° PC refers to 3° externally rotated from the posterior condylar line, PC refers to alignment with the posterior condylar line, and T E refers to alignment with the transepicondylar axis). N e Reference Change in Tilt Change in Shift External Rotation Internal Rotation External Rotation Internal Rotation Our study: vertical rig 8 5° 3° PC 2°-5° Avg: 3.5° - 3°-4° Avg: -3.7° 3.4 mm (early to mid-flex.) 1.1 mm (90° flex.) Avg: 2.1 mm -1.9 mm (early to mid-flex.) +0.7 mm (90° flex.) Anouchi (1993) 4 5° PC 1-2.5 mm (early to mid-flex.) 0.5 mm (90° flex.) -2 mm (early to mid-flex.) -0.5 mm (90° flex.) Armstrong (2003) 7 6° 3° PC - 3 ° ~ 4 mm Miller (2001a) 11 5° TE s 2°-3° ~ - 2°-3° — ~ Miller (2001b) 9 2.5°, 5° PC 2.5°, 5° — No effect No effect Rhoads (1990) 7 10° ? 3° -6-10° 3-5 mm - 2-7 mm 142 Armstrong (2003) used an open-chain testing rig to investigate the effects of external femoral component rotation on tracking. They secured the femur horizontally on a table and allowed the tibia to hang flexed over the edge of the table. Steel cables were sutured to the tendons of the 3 quadriceps muscle groups (vastus medialis, vastus lateralis, and rectus femoris/vastus intermedius) and the knee was extended using motors attached to these cables. The experimenters moved the tibia while the quadriceps loads were applied to simulate eccentric loading; however, the force applied to the tibia was not measured. The loads applied to the medial, lateral and central quadriceps groups were 21 N , 27 N , and 32 N , respectively. This setup does not emulate either of our testing rigs. Neutral femoral rotation was defined as 3° of external rotation (similar to our definition). Armstrong attempted to rotate the femoral component 10° externally with respect to the neutral placement; however, their achieved rotation measured 6.0±4.9°. Similar to our findings, they found that rotating the femoral component externally caused the patella to shift and tilt laterally (Table 5.2). Malrotation of the femoral component did not have significant effects on patellar rotation, consistent with the results of our study. They did not test the effects of internally rotating the femur. The authors acknowledged that a major limitation of their study was that the loads applied to the knee were much less than physiological loads. Kessler (2004) found no significant differences in patellar tracking for knees loaded at 100 N and 500 N ; however, there may be changes in tracking at loads lower than 100 N . Kessler did not specify which type of testing rig was used to flex the cadaveric knees. Miller (2001a, 2001b) measured the effects of femoral component rotation on patellar tracking in two studies. In one study (2001a), they loaded knees in an Oxford testing rig with the femoral components rotated ±2.5° and ±5° with respect to both the TE and posterior condylar lines. Components were placed using the posterior condyles but placement was measured with respect to the TE axis. They found that rotating the femoral component externally and internally resulted in respective increases in lateral and medial tilt between 10° and 90° flexion (Table 5.2). These differences were comparable to those reported by this study and to those reported by Armstrong. They found that the effects of femoral rotation were less reproducibly related to the posterior condyles than to the TE axis. In a similar study (2001b), they again observed an increase in lateral tilt with external rotation (with respect to the posterior condyles) (Table 5.2). Contrary to the findings of our study and those of Armstrong and Anouchi, they found that 143 patellar shift was insensitive to femoral component rotation. It is interesting to note that a comparison of the tilt curves for both studies performed by Miller shows that they were dissimilar in shape, despite identical testing protocol. Rhoads (1990) studied the effects of malrotation of the femoral component on patellar tracking using a rather primitive testing configuration: they fixed the tibia vertically and suspended a mass of 20 kg from the proximal femur to cause the knee to flex between 20° and 80°. The knee was brought from 140° flexion to full extension by pulling on a cable attached to the quadriceps tendon. They rotated the femoral component by 10° externally or internally with respect to neutral placement and found that external rotation tilted and shifted the patella laterally and internal rotation tilted shifted the patella medially. The patellar tilt and shift they reported due to 10° internal rotations were approximately twice as large as the values we measured due to 5° rotations (Table 5.2); however, they did not report as great a change in tilt following external rotation compared to internal rotation. Patellar rotation was not affected by either internal or external femoral component rotation. Rhoads supported the claims of Anouchi that external rotation yielded patellar tracking which was most similar to that of the intact, natural knee. Similar increases in lateral shift following external component rotation have been observed by Anouchi, Armstrong, Rhoads, and the current study; however, the aforementioned authors disagree as to the clinical implications of lateral shift. Anouchi and Rhoads opined that the lateral shift reproduced the tracking observed in the natural knee and was therefore a positive effect. Conversely, Armstrong found that neutral placement best reproduced the tracking in the natural knee and believed that the lateral shift would likely increase contact pressures. I believe that our study and all the studies described above are flawed in their measurement protocol: the reported changes in spatial position of the patella reflect the fact that we have rotated the groove in which the patella rides; we have not necessarily altered the way in which the patella tracks relative to the groove. Because we measured tracking relative to the femoral bone and not femoral groove, we have measured the effect of rotating the component on both the location of the femoral groove and on the location of the patella within the groove. It is therefore difficult to postulate what is actually occurring as we rotate the femur. The conclusion of clinical studies that femoral rotation is not linked to shift (Akagi 1999, Kawano 2002, Matsuda 2001) or tilt (Akagi, Kawano) appear to disagree with the cadaveric studies because they measured tracking relative to the femoral groove of the implant instead of the femoral bone. 144 We observed the greatest change in patellar tilt in early flexion, when the patellar tilt axis was almost parallel to the rotation axis of the femoral component (Figure 5.2); because the axes were almost aligned, a rotation about the femoral axis was reflected in the measurement of patellar tilt. At 90° flexion, the axis of patellar tilt was no longer parallel to the axis of rotation of the femoral component; therefore, rotations about the femoral component axis did not alter the measurement of patellar tilt as much. If the patellar tilt axis was approximately parallel to the rotation axis of the femoral axis in early flexion, one would expect a 5° change in tilt following external or internal rotation on the femoral component; instead, the change in patellar tilt typically measured 4°. The difference between the geometrically expected change in tilt and the measured tilt may indicate that external and internal rotations result in small increases in medial and lateral patellar tilt, respectively. This hypothesis is reasonable because external and internal rotations change the direction of pull of the extensor mechanism relative to the femoral implant and could therefore be expected to induce patellar tilt in the direction of pull. Also, medial or lateral soft tissues may pull the patella back towards neutral tilt. 90° Figure 5.2: Location of the rotation axis of the femoral component and the tilt axis of the patella. (Zimmer, 2004) The measured changes in patellar shift following femoral component rotation can largely be explained by the displacement of the femoral groove with respect to the femoral bone (shift is measured along the TE axis of the femoral bone, which remains stationary during rotation of the 145 femoral component). The patella is furthest from the axis of rotation of the femoral component in early flexion, and a rotation of the femoral component would translate the trochlea and the patella by a distance equal to XApsin(5°), where X A P is the anteroposterior distance between the centre of the patellar coordinate system and the axis of rotation of the femoral component. For X A P ~ 40 mm, we would expect a lateral or medial translation of approximately 3.5 mm following external or internal component rotation, respectively. We measured a change in shift between 2-3.5 mm, which indicates that the patella likely remained in roughly the same location relative to the femoral component. Miller (2001b) also postulated that the patella mainly moves with the femur during component rotation and patellar position is therefore insensitive to component rotation. In mid- to late flexion, the patella is closer to the axis of rotation of the femoral component in the anteroposterior direction (i.e. X A P has decreased); therefore, a rotation of the femoral component did not have as great an effect on measured shift as in early flexion, as expected. In late flexion, the patella is deep in the femoral groove, and the location of the most posterior portion of the groove has been translated in the opposite direction as the anterior portion of the trochlea (following component rotation). As a result, between 60° and 105° flexion, the patella crosses the rotation axis of the femoral component and the value of patellar shift changes sign (Figure 5.3). 146 Rotation axis of femoral component Tracking path, neutral rotation Tracking path, external rotation Transepicondylar (shift) axis 5° external rotation of femoral component Figure 5.3: Geometrically expected change in shift due to external rotation of the femoral component. The tracking paths of the patella have been approximated by straight lines. Optimal tracking does not necessarily replicate the tracking in the natural knee; the T K A condyles are differently shaped than those in the natural knee, and the patella should not be expected to regain its original path despite changes to the condyles and the trochlea. It is more likely that optimal tracking centralizes the patellar implant within the femoral groove. Internally rotating the femoral component increases the Q-angle because it displaces the patella medially. A n increased Q-angle increases the lateral component of the force in the quadriceps tendon and likely increases the risk of subluxation and excessive shear loading. Externally rotating the femoral component may reduce the Q-angle, but it also decreases the relative height of the lateral femoral condyle in the coronal plane. A reduced lateral ridge would decrease the effectiveness of the condyle in preventing patellar subluxation. It is difficult to predict which of the opposing effects would have a dominating effect on patellar tracking: a reduced Q-angle or a lower lateral ridge. Rhoads observed that with the femoral component in external rotation, patellar tracking was less predictable because the patella was less constrained. In a literature review focusing on the rotational malalignment of the femoral component, Zihlmann (2005) found that malalignment "of a measurable degree" occurs in approximately 10-30% of patients; however, not all patients with noticeable malalignment report an unsatisfactory 147 outcome. The definition of malalignment depends on surgical technique and the landmarks used to measure rotation. They concluded that excessive external rotation of the femoral component can cause abnormal tightness in the popliteus tendon complex, which could result in a loss of rotational laxity during late knee flexion. Any rotation (internal or external) may result in uneven flexion gaps and could cause tightness in the retinaculum surrounding the patella. Future studies on the effects of femoral rotation should attempt to measure tracking relative to the femoral groove such that the effect of rotating the femoral groove is not the major effect observed. 5.3.2 Femoral Component Rotation: Loading There is no consensus as to the effects of femoral component rotation on patellofemoral loading. Comparisons are difficult to make because different studies report different measures of force or pressure, depending on the method used to analyze forces. Anouchi (1993) and D'Lima (2003) concluded that external rotation reduced or balanced patellar loading. Although they did not quantitatively measure patellofemoral loads, Anouchi (1993) used Fuji film to assess the force balance at the medial and lateral patellar contact points. They observed a greater balance between medial and lateral patellar forces for femoral components in 5° external rotation than in neutral or internal rotation. They postulated that external rotation medialized the loads on the patella, resulting in equalized medial and lateral loads. D'Lima analyzed the effects of ±3° rotation of the femoral component on shear patellofemoral loads using a validated finite element model. They found that external rotation of the femoral component substantially reduced mediolateral shear, while internal rotation increased shear. Because we measured normal force and not mediolateral loads, we could not compare our results to the studies of Anouchi and D'Lima. Singerman (1997) investigated the effects of femoral component rotation on patellofemoral forces for two different knee designs. They found that femoral component rotation had no effect on the loads for knees implanted with a highly constrained tibiofemoral (TF) joint; however, all components of the patellofemoral force were significantly related to femoral component rotation in knees implanted with a lowly constrained TF joint. The normal force was significantly different between internal and external rotation but not for neutral rotation. In internal rotation, the normal force was 8% higher than in external rotation when averaged over all angles; 148 however, the difference was not uniform over all angles and became negative at high flexion angles. This finding did not correspond to our results; we found that peak forces occurred during late flexion and that the reported normal force increased in both external and internal rotation for both testing rigs. Singerman reported that the contact force reported by the 6 degree of freedom load cell was nearly perpendicular to the contact surface, which supports the use of pressure sensors that measure only normal loads such as Tekscan sensors. When averaged over all tested angles, the mediolateral patellar force measured by Singerman was 17% higher in internal rotation, but that difference was not uniform over all angles. This result corresponded to the findings of D'Lima. The authors hypothesized that internally rotating the femoral component also rotated the patella (medially) and increased the Q-angle, which increased mediolateral forces. The mediolateral load became negative at larger flexion angles, likely because the patella moved laterally and the reduced Q-angle resulted in a decrease in shear loading. Miller (2001a) implanted a 6 degree of freedom (DOF) load cell in the patella to measure the effects of femoral component rotation on patellofemoral loads. They sought to determine which rotation would minimize the mediolateral forces acting on the patella. They found that rotation parallel to the TE axis minimized patellofemoral shear forces early in flexion. Rotating the femoral component by 2.5° and 5° internally or externally increased shear forces early in flexion. Although they measured mediolateral forces instead of normal forces, this finding corresponded loosely with our observations for both testing rigs. The conclusion that neutral rotation optimized shear forces contrasted with the conclusions of Anouchi, D 'Lima and Singerman. In another study, Miller (2001b) found that contrary to their previous findings, there was no change in mediolateral loading due to 2.5° or 5° external femoral component rotations. The normal component of the force was not affected by either component placement. The effects of femoral rotation were less reproducibly related to the posterior condyles than to the TE axis (2001a). At small flexion angles, both Miller and Singerman observed a single central contact area on the patella, whereas the contact pattern was bimodal at high flexion angles as the patella traveled in the trochlear groove. We observed a similar phenomenon for all specimens in our study (see 4.36), and this pattern is supported by the Zimmer product literature (Figure 5.4). 149 Figure 5.4: Diagrams of patellofemoral contact at 10°, 45° and 90° flexion (Zimmer, 2004). It is difficult to make direct comparisons between other loading studies because the aforementioned studies all used loads cells or qualitative observations (Anouchi) to measure force, whereas we used Tekscan pressure sensors. Some studies mainly reported mediolateral shear, which we could not measure with our sensors because they only report normal force. Singerman and Miller reported normal force; our results agreed with those of Miller. Our measurements were greatly affected by specimen variability (the maximum standard deviations in peak load in the horizontal and vertical testing rigs were 17% and 9%, respectively) and may be reduced in accuracy because of sensor degradation. Also, no changes in peak force relative to the baseline were statistically significant. 5.4 Tibial Component Rotation Similar to femoral component rotation, tibial component rotation is thought to relate to patellar tracking and loading because rotation of the tibia results in changes to the Q-angle; a displacement of the tibial tubercle changes the direction of pull of the extensor mechanism. Internally rotating the tibial tray causes the bone to rotate externally because the bone surfaces of the femoral and tibial components are forced to conform to each other. Malrotation of the tibial tray may also affect strains in the tibial bone or in the soft tissues. 5.4.1 Tibial Component Rotation: Tracking We did not find statistically significant differences in patellar tracking with 5° external or internal rotation of the tibial component, with one exception: external rotation of the component in the horizontal rig resulted in lateral patellar tilt and medial spin. The only other in vitro study on tibial rotation (Nagamine 1994) found that 15° external tibial component rotation affected patellar shift but not tilt or rotation. External malrotation shifted the patella medially, but this effect was only statistically significant at 15° and decreased at larger flexion angles. Internal rotation did not affect patellar tracking. It can be concluded that rotations of the tibial component have small and inconsistent effects on patellar tracking in vitro. It is plausible that 150 rotations of the tibial tray do not greatly affect the Q-angle. The possibility also exists that the full rotation of the tibial bone could not be achieved due to soft tissue tensions. 5.4.2 Tibial Component Rotation: Loading Ours is the only study we are aware of that has looked at the consequences of tibial component rotation on patellar loading. The results from the horizontal and vertical rig were opposite to each other. In the horizontal rig, external rotation increased contact forces and internal rotation decreased contact forces; in the vertical rig, external rotation tended to decrease contact forces and internal rotation tended to increase contact forces, although these changes were not statistically significant. It is difficult to determine whether the differences in loading rigs were responsible for the opposite loading effects or whether the variability between specimens masked the changes or lack of changes in loading (maximum standard deviations in the horizontal and vertical rigs were 12% and 6%, respectively). One would expect that by internally rotating the tibial tray, the tibial tubercle would be translated externally, which would increase the Q-angle and increase the load applied by the lateral condyle on the lateral aspect of the patella (Elias 2004b). This hypothesis is supported by the changes observed in the vertical rig. Externally rotating the tibial component should have the opposite effect; it should decrease the Q-angle and reduce lateral patellar loading, as observed in the vertical rig. 5.5 Patellar Resection Angle Asymmetrical resection of the patella may lead to excessive loading and wear and anterior knee pain. Although the effects of resection angle have been investigated in clinical studies, surprisingly, asymmetrical resection has not been studied previously in vitro. 5.5.1 Patellar Resection Angle: Tracking We found that patellae with a thicker medial side tilted laterally, which agrees with the clinical findings of Gomes (1988) but disagrees with the findings of the clinical studies performed by Chan (1999) and Kawano (2002). Our finding that in the vertical rig, patellae with a medial resection angle of 7.5° (i.e. thicker on the medial side) experienced an average increase in lateral tilt of 6.6° suggests that the patella is largely constrained by the femoral groove. The implant probably remained at the same angle traveling in the femoral groove, and the change in tilt we 151 measured largely reflected the change in angle of the marker array (and bony remnant) due to the added angular wedge (Figure 5.5). Because the measured angle of tilt was slightly smaller than the added angle, we may actually have observed a very small amount of medial tilt (0.9°, on average). The 7.5° lateral resection angle resulted in an average medial tilt of 6.3°, which suggests that the modified patellar implant may have been tilted 1.2° laterally, on average. The lateral angle of 15° increased the medial tilt by an average of 13.0°; the 2° discrepancy may indicate slight lateral tilt of the patellar implant. The discrepancies between measured and geometrically predicted tilt may also be explained by the inter-specimen variability: the maximum standard deviation was 4.2°. Tilt of the implant is likely of more clinical significance than tilt of the bony remnant; it is related to patellar loads and risk of subluxation. Tilt of the bony remnant could be an indirect measure of tension in the soft tissues, which may be related to A K P . 152 Neutral Placement Altered Resection Angle Neutral Disk t = thickness of bony remnant 9 = tilt of bony remnant and marker array due to addition of angular wedae Marker Array Bony Remnant Patellar Baseplate Modified Patellar Component Wedged Disk x = tsin9 ~ shift of bony remnant and marker array due to addition of angular wedge Figure 5.5: Tilt and shift of bony remnant and marker array due to change in resection angle. Note that the central thickness of the angular wedge was the same as the thickness of the neutral disk. Gomes and Kawano did not observe changes in shift due to different resection angles in their clinical studies. In our study, the 7.5° medial and lateral resection angles resulted in lateral and medial shifts of 1.7 mm and 1.4 mm, respectively, in the vertical rig. This shift was likely not a measure of the patellar implant shifting in the femoral groove; it corresponded to the shift we would expect due to rotation of the optoelectronic marker array. Upon addition of the angular wedge to the patellar construct, the marker array was also rotated through an angle. In our study, the patellae had bony remnants measuring between 10 mm and 15 mm thick3; we would expect In order to compensate for the addition of a patellar baseplate to the patellar construct, we resected more bone than a surgeon would usually resect during surgery. A patellar bone remnant measuring 10-15 mm thick would typically be the minimum amount of bone remaining following surgery. 153 the corresponding shift due to a 7.5° angular wedge to measure between 1.3 mm and 2.0 mm (Figure 5.5). This range encompasses the observed shifts for the 7.5° resection angles. With a 15° lateral resection angle and patellae with similar bony remnants, we would expect medial shift between 2.6 mm and 3.9 mm. We measured an average medial shift of 3.9 mm; this value is within the expected range. Less change in shift was observed in early flexion than late flexion; this indicates that in early flexion, the marker array remained in a relatively similar position compared to baseline, and that the patellar implant may have shifted to compensate for the change in direction of pull of the extensor mechanism. In early flexion, the patella is not as constrained by the femoral groove, and its position is probably determined to a great extent by the lines of action of the quadriceps tendon and the patellar ligament. A lateral resection angle would slightly increase the Q-angle and shift the patella laterally. A medial resection angle would decrease the Q-angle and have the opposite effect on shift. The small changes in tilt and shift of the bone remnant may be larger in the presence of tight lateral or medial soft issues. 5.5.2 Patellar Resection Angle: Loading Although the effects of the resection angles on contact forces were extremely variable and not statistically significant, there seemed to be an overall trend in change in loading: in both testing rigs, as the resection angle became more medial, contact loads increased. It is unclear why this trend in loading occurred. A possible explanation for the low forces reported with the 15° lateral wedge is that for some specimens, the contact areas appeared to be outside the circumference of the patellar implant in late flexion and would therefore not be included in the sum of forces over the surface of the sensor. Because the standard deviations were extremely large for the resection angle force data (up to 25% in both testing rigs), it is difficult to draw conclusions about the effect of each patellar resection angle on loading. 5.6 Position of the Patellar Component Miller (2001b) stated that medializing the patellar implant by 3.75 mm best replicates the location of the patellar ridge in the natural patella. They also believed that this position offered the best coverage of the cut surface. It is a commonly held belief that medializing the patellar component results in lateral displacement of the patellar bone with respect to the femoral groove. 154 This lateral shift of the patellar bone should decrease the Q-angle, reduce mediolateral patellar loads and decrease the risk of patellar subluxation. 5.6.1 Position of the Patellar Component: Tracking Both Miller (2001b) and Yoshii (1992) found that medializing the patellar implant shifts the bony remnant laterally during flexion; this is in agreement with our findings. Miller medialized the patellar button by 3.75 mm and observed a lateral shift in the bony remnant of approximately 4 mm. We observed that medializing the patellar component by 2.5 mm and 5 mm resulted in average lateral shifts of 1.9 mm and 3.9 mm, respectively. The similarity between the medialized distance and the measured shift shows that the patellar implant is highly constrained by the femoral groove, and that the position of the implant in the groove is relatively insensitive to implant position in the patellar bone. The measured shift represents the displacement of the implant with respect to the bony remnant during implantation (Figure 5.6). Marker Array Bony Remnant Patellar Baseplate Neutral Disk Modified Patellar Component Posterolateral Component of Quadriceps Force d = medial displacement of patellar implant with respect to the bony remnant Figure 5.6: The medialization of the patellar component results in a lateral translation of the bony remnant. It also increases the moment arm of the posterior component of the quadriceps force, which results in lateral tilt of the patella. 155 Yoshii found that the effect of medializing the patellar button was altered by the shape of the femoral groove. For a symmetric femoral component similar to ours, with identical medial and lateral condyles, the patellar bone was shifted laterally by 9 mm following a 10 mm medialization of the implant. For a femoral component with a 3 mm build-up on the lateral condyle, the bony remnant shifted only 5 mm. It is possible that the bone remnant came into contact with lateral build-up and that this prevented the patella from shifting further. Miller observed very little change in shift at low flexion angles, similar to our findings. This is likely because in early flexion, the patellar button is not constrained by the trochlear groove. The position of the implant is therefore governed by the action of the quadriceps tendon, and the line of action of this force has not changed. Because the bony remnant remained in the same location as during baseline tests, the modified patellar implant would therefore have shifted medially. Lee (1999), Miller and Yoshii discovered that medializing the patellar implant results in more lateral tilt, as confirmed by our findings. In their clinical studies, Gomes (1988) and Kawano (2002) concurred that medialization of the patellar component increased lateral tilt. This lateral tilt is the result of a moment applied about the inferosuperior axis of the patella; the lateral translation of the line of force of the quadriceps relative to the centre of the patellar implant increases the moment arm at which the force acts (Figure 5.6). Yoshii found that tilt was greater for knees with symmetrical femoral components than for knees with a high lateral condyle. It is likely that the deeper groove resulting from the higher lateral ridge prevented the patella from tilting abnormally. 5.6.2 Position of the Patellar Component: Loading Patellar medialization has been shown to decrease the lateral patellar shear while increasing the medial shear (D'Lima 2003, Lee 1999), and was also shown to decrease the mediolateral component of the patellofemoral loads (Miller 2001b). Lee found that a centralized component yielded the most evenly balanced patellar facet contact pressures. None of these studies reported normal patellofemoral loads. We found that medializing the patellar component by 2.5 mm or 5 mm had little effect on loads in the horizontal rig and tended to decrease patellar loads in the vertical rig, although these results were not statistically significant and were extremely variable between specimens (maximum standard deviations of 27% and 32% in the horizontal and vertical rigs, respectively). It is likely that slight medialization of the implant pushes the patella 156 against the medial condyle with more force and results in a corresponding decrease in force at the lateral condyle. The overall mediolateral force (or normal force, as we measured) should decrease because the Q-angle is reduced following the lateral shift of the bony remnant. 5.7 Patellar Thickness Patellar thickness is thought to affect the loading and wear at the surface of the patellar implant. It may be related to probability of fracture (Josefchak 1987, Marmor 1988, Reuben 1991) and may also affect joint range of motion (Marmor 1988). 5.7.1 Patellar Thickness: Tracking Although Hsu (1996) determined that increasing patellar thickness by 2 mm increases lateral tilt while decreasing the thickness decreases tilt, we found no change in tilt due to increased thickness. Clinically, Laughlin observed changes in tilt whereas Kawano did not. Hsu showed that the thick patella shifts medially in comparison to the thin patella; however, we observed no changes in shift due to altered thickness. The clinical study by Kawano found no effect of thickness on shift. The hypothesis that tracking is not affected by changes in patellar thickness is supported by the argument that the Q-angle and patellofemoral joint geometry have not been altered. The possibility exists that an increase in volume in the joint capsule could change the tensions in the soft tissues and alter patellar kinematics, although we did not find evidence to support this theory. 5.7.2 Patellar Thickness: Loading Oishi (1996) found an increase in total, mediolateral and superoinferior shear forces with increased patellar thickness, but reported no change in the normal loads for 3 different patellar thicknesses. Hsu (1996) and Star (1996) found that thicker patellae experience increased normal loads; our findings, although not statistically significant, support those of Hsu and Star. Both a decrease and an increase in patellar loads due to thicker patellae could be explained with biomechanical arguments: 1) An additional patellar thickness could decrease the quadriceps force by increasing the effective moment arm of the quadriceps mechanism; the quadriceps tension would decrease to compensate for the larger moment arm. Decreased quadriceps tension should result in smaller patellar contact loads. 157 2) The addition of patellar thickness may increase soft tissue tensions. It may also increase patellofemoral loads because it would cause the lines of action of the quadriceps mechanism to be directed more posteriorly. As a result, the reaction force (normal force) at the surface of the patella and femur would increase. In a study of knees with and without patelloplasties, Masri (2005) determined that reducing the patellar thickness increased the tension in the quadriceps tendon in late flexion. The increase in quadriceps tension reflects the change in the magnitude of the effective moment arm (effect 1). In our testing rigs, we saw the forces increase with thicker patellae, most likely because of increased soft tissue tensions or the change in the direction of pull (effect 2). It should be noted that the quadriceps force was constant in the horizontal rig; therefore, effect 1 could not apply to the horizontal loading configuration. In a postoperative loading situation, effect 1 would be expected to dominate the mechanics of the patella because the quadriceps loads are much greater than the tensions in the peripatellar soft tissues. A thick patella may overstuff the joint capsule and increase the strain in the surrounding soft tissues. This phenomenon could plausibly be related to A K P . Bone conserving resections that increase the patellar composite thickness above the precut thickness should probably be avoided. Care must also be taken to not over-resect bone during T K A ; Josefchak (1987) reported that significant weakening of the patella occurred if the subchondral bone was resected, and Reuben (1991) found that decreasing the bony patellar thickness to less than 15 mm or the patellar composite to less than 25 mm increased the anterior patellar surface strain (which may be related to A K P or fracture of the patella). 5.8 Ranking of Variables 5.8.1 Tracking To optimize patellar tracking relative to the femur, surgeons can likely pay less attention to tibial rotation and patellar thickness than femoral rotation, patellar resection angle, and patellar medialization. Tibial rotation and patellar thickness had very little impact on patellar tracking, compared to the latter three surgical variables. However, it is difficult to discern how important the effects of femoral rotation were on tracking relative to the femoral component itself because tilt and shift were measured relative to the femoral bone, as discussed in section 5.3.1. Changes to component placement largely had the expected geometric effect on the tracking of the patellar 158 bone remnant. It can therefore'be inferred that the patellar component maintained a constant position and angle within the femoral groove to within a few degrees and a few millimeters, especially post-operatively. Discrepancies between the geometrically predicted tilt or shift and the measured values likely reflect the contribution of soft tissues in guiding patellar kinematics. Also, the discrepancies suggest changes in soft tissue tension, which may be linked to anterior knee pain. In both the horizontal and vertical testing rigs, the surgical variables which had the greatest effect on patellar tilt and shift were similar in mid-and late flexion but different in early flexion. Femoral component rotation had a great effect on tracking in early flexion but not in mid- to late flexion. The change in the magnitude of the effect was probably due to the rotation of the femoral groove; in early flexion, the patella is located at the greatest distance (in the transverse plane) from the centre of the femoral shaft, and a rotation about this axis results in a large change in the position of the patella. In mid- and late flexion, the patella is seated within the femoral groove and is thus closer to the long axis of the femur (in the transverse plane); a rotation of the groove does not result in as great a change in position of the patella. The resection angles did not have large effects on shift in early flexion because the location of the patella was largely dictated by the pull of the quadriceps mechanism. In late flexion, the position of the patellar implant was dictated by the geometry of the femoral condyles. 5.8.2 Loading The variables which had the greatest effect on patellar loading differed between rigs. In the horizontal rig, the addition of a 3 mm thickness to the patellar construct had the greatest non-significant effect on loading, whereas medializing the patellar component by 5 mm had the greatest non-significant effect on loads in the vertical rig. A qualitative conclusion that may be drawn from the highly variable force measurements is that changes in placement of the patellar component (angle, position and thickness) have visible effects on patellar contact loads; however, the effects are highly variable between specimens. Clearly, the surgeon must take care to optimize patellar positioning, yet it is not clear which intraoperative position would minimize postoperative contact forces; intraoperative and postoperative forces were only moderately correlated. 159 Peak loads did not occur at the same flexion angles for different surgical variables. Also, there was a great deal of variability in the shapes and magnitudes of the force curves between specimens and between surgical variables. As mentioned previously in section 2.7.2.12.7.2.1, the degradation of sensor output throughout testing made comparisons between variables difficult, despite our normalization procedure. Another possible, but likely less significant, explanation for the variability in the shape of the force output may be that the adhesive used to attach the Tekscan sensor to the patellar implant created a layer of material with varying stiffness. The sensors are known to be sensitive to material stiffness (Tekscan 2001); in tests with low-pressure F-Scan Tekscan sensors, Zong-Ping (1998) discovered that stiffer contact surfaces result in higher force outputs and greater variability in readings. In our study, i f a change in component position resulted in a change in location of the contact area, and the new contact area had less glue than the original location (i.e. increased stiffness), the sensor would report an increase in force. The uneven surface created by the glue may also have affected the number of sensels that were activated at low loads. 5.9 Comparison between Horizontal and Vertical Rigs 5.9.1 Tracking Changes in patellar tilt and shift (relative to baseline) were statistically similar and were moderately- to well-correlated in the horizontal and vertical rigs at all flexion angles. The slopes of the best-fit lines through the data were less than 1, which indicates that changes in postoperative tracking could be expected to be smaller in magnitude than changes in intraoperative tracking. It is likely that the changes in tracking in the horizontal rig were larger because the low quadriceps tension resulted in more laxity of the quadriceps mechanism. The good correlation between changes in tilt and shift between loading rigs is extremely significant to the study of knee mechanics; it indicates that changes made intraoperatively to optimize tracking will likely result in very similar changes in the postoperative knee, i.e. optimizing patellar tracking intraoperatively should result in improvements in postoperative tracking. For example, i f a surgeon chose to increase lateral tilt and shift of the bony remnant by medializing the patellar implant, they can be reasonably confident that such a geometric change will have similar effects (although of smaller magnitudes) on the post-operative joint. This new link between changes in intraoperative and postoperative tracking increases the applicability of 160 existing studies which have mimicked postoperative loading conditions in vertical rigs; their findings may now be more confidently extrapolated to the intraoperative setting. The similarity in relative tracking indices between rigs validates the horizontal rig for use in further studies. Although changes in tracking were similar between testing rigs, the absolute measurements of tracking were not as well correlated between rigs. Measurements of intraoperative tracking may not be accurate predictors of postoperative tracking. The poorer correlation of tracking indices between rigs is likely due to differences between loading methods (see section 5.13, below). A l l measurements of tracking indices were more variable in the horizontal testing rig than the vertical testing rig, particularly in early flexion. This variability was due to the more flaccid state of the muscles and the lower quadriceps tension in the horizontal rig; without greater muscle tension, the patella tracked more erratically. This effect was most noticeable in early flexion because the patella merely rested on the anterior surface of the trochlea and was not constrained by the walls of the femoral groove. 5.9.2 Contact Forces There is no reason to suggest that surgeons should not take care to reduce patellar loading during surgery, but the lack of correlation between intraoperative and postoperative forces makes it difficult to recommend component positions that would minimize postoperative loads. The similarity between relative measurements of tracking in the two rigs did not extend to force data; for several surgical variables, the changes in force were opposite between rigs. The force measurements obtained in the vertical rig were generally more variable than those taken in the vertical rig. The differences between measured forces could be the result of different loading configurations. It is possible that the number of specimens with usable force data in the vertical rig was inadequate: we were only able to collect complete force data for 5 specimens in the vertical rig, whereas we had acceptable data from 7 specimens in the horizontal rig. Peak patellar contact forces were 3-6 times greater in the vertical rig than in the horizontal rig. This is logical because the vertical rig actively 'contracted' the quadriceps muscle and applied large axial loads to the femur and tibia. In the horizontal rig, just enough force was applied to the quadriceps tendon to keep it taut, and the knee was flexed passively. Based on our measurements of quadriceps tension in the horizontal rig and the measurements taken by 161 Michael Paice in a similar experiment in the vertical rig (Figure 5.7), the tension in the quadriceps tendon appeared to be between 5-10 times greater in the vertical rig than in the horizontal rig. (We could not measure quadriceps tension in the vertical rig because the sensor would not fit between the quadriceps clamp and the rig pulley, as described in section 2.6.2.2.3) 15 Figure 5.7: Quadriceps tension in the vertical rig, as measured by Michael Paice at UBC, and in the horizontal rig, as measured in our study. Note that the maximum flexion angle for Michael Pake's data was 60°, whereas our data is for flexion angles up to approximately 120°. The quadriceps tension in our horizontal rig at 60° was approximately 70 N. In the horizontal rig, the quadriceps tension was identical for flexion and extension but patellar contact forces were generally higher during flexion than extension (although the difference was not great- approximately 5-10%). The difference in contact load between flexion and extension phases may be explained by the difference in speed at which the knee was flexed. We unintentionally flexed the knee slightly faster than we extended it, and additional tests on the effect of speed (see section 4.8) showed that faster flexion increased reported load by 16.6 ± 8.4%. In the vertical rig, contact forces were higher during extension than flexion, and force in the quadriceps tendon was also greater (Figure 5.7). During flexion, the quadriceps resisted the buckling of the knee and counteracted the weight of the Oxford rig tray. As the cable attached to the tendon was gradually released, the force in the tendon decreased and the weight of the rig tray was large enough to induce flexion of the knee. In extension, the quadriceps tension had to overcome the weight of the rig tray to actively extend the knee. The increase in quadriceps tension during extension resulted in a greater normal force applied to the patella (Figure 5.8). 162 Friction in the system would increase the amount of quadriceps tension required to extend the knee. Figure 5.8: Geometrical relationship between the magnitude of the quadriceps force and the resultant patellofemoral contact force, (veggie.org/run/chondromalacia/images-knee-uorg/comprel.gif) In both the horizontal and vertical rigs, the magnitudes of the peak contact forces were extremely variable between specimens. The variability may have resulted from differences in soft tissue tensions between specimens. 5.10 Relationship between Tracking and Loading We had hoped to determine a relationship between patellar tracking and loading; however, due to the great variability in force measurements and the differences in loading between testing rigs, we were unable to find such a relationship. Parameters that do not affect patellar kinematics might still affect patellofemoral contact force. For example, patellar thickness had no effect on patellar tracking, but seemed to increase contact loads in both testing rigs. Tibial rotation also had no effect on tracking, and resulted in opposite effects on loading in the horizontal and vertical rigs. Other patellar tracking and loading studies have not attempted to determine a link between tracking and loading (Anouchi 1993, Miller 2001a, Miller 2001b, Hsu 1996). 5.11 Strengths and Novelty of Study A l l previous studies of component placement that we are aware of have tested specimens in a simulated post-operative setup. As a result, the relationship between intra-operative observations and post-operative kinematics was unknown. This study was the first, to our knowledge, to 163 compare intraoperative and postoperative patellar mechanics. The similarity we found in relative tracking indices between rigs validates our horizontal rig for use in further studies. Previous researchers have studied the effects of one or two component placement changes on patellar tracking and loading; however, to our knowledge, our study is the only study to have ranked or compared a range of different factors. Every change in component placement was tested on each knee in both testing rigs. This study was the first cadaver study to investigate the impact of different bone cut angles on patellar tracking and loading; it was also the first study to focus on patellar loading following tibial component rotation, and the second study to investigate the effects of tibial component rotation on tracking (Nagamine (1994) was the first). Ours was the first study to report changes in patellofemoral load due to altered component positioning using Tekscan sensors; other researchers have used Tekscan sensors in the unresurfaced patellofemoral joint to study changes in contact stress due to differing implant designs (Whiteside 2003). Instead of using the calibration software supplied with the sensors, we designed our own more accurate calibration routines, as described in Chapter 3. Both kinematics and kinetics were measured simultaneously during dynamic flexion. We tested 8 specimens, and this sample size is comparable or greater than the number of specimens tested in other cadaveric studies of the knee (Table 5.3). (Six of the 19 studies we reviewed tested more specimens, 3 other studies also tested 8 specimens, and 11 studies tested less than 8 specimens.) We strove to create simulations of intraoperative and postoperative flexion that would reasonably approximate the in vivo cases. To further qualify the appropriateness of our models, we attempted to determine the potentially biasing effects of certain aspects of our testing protocol (see section 5.12 below). 164 Table 5.3: Number of specimens tested in cadaveric studies of the knee. Study Number of Specimens Singerman (1997) 13 Miller (2001a) 11 Lee(1997) 10 Oishi (1996) 10 Chew (1997) 9 Miller (2001b) 9 Our Study 8 Arizono (1999) 8 L i (2004) 8 Miller (1998) 8 Armstrong (2003) 7 Hsu (1996) 7 Rhoads (1990) 7 Star (1996) 7 Szivek(1996) 7 Kessler (2004) 6 Yoshii (1992) 6 Lee(1999) 5 Nagamine (1994) 5 Anouchi (1993) 4 D'Lima (2003) 3 5.12 Potential Confounding Factors We tested the effects of additional variables that we hypothesized may affect the accuracy of tracking and force measurements. In the horizontal rig, we found that the speed had very little consistent effect on tilt, shift or spin; however, faster flexion cycles reduced the patellofemoral contact force. Although we attempted to perform each flexion cycle within one minute, slight differences in timing between cycles may have resulted in inaccurate differences in contact forces. Differences in timing were less likely to occur in the vertical rig because the motion of the knee was controlled by a motor. In both testing rigs, we measured the effect of removing the surgical towel clamps which were used to close the upper portion of the medial incision (above the Tekscan sensor). We hypothesized that the presence of the towel clamps simulated a surgical closure such as sutures, and that removing the towel clamps would reduce the tension in the medial soft tissues. The reduced tension would likely increase lateral tilt and shift, and reduce patellar contact loads. We found that the absence of the clamps did not have a noticeable affect on tilt or shift; however, the lack of clamps seemed to induce lateral spin in both the horizontal and vertical testing rigs. This 165 would indicate a difference in soft tissue tension. As predicted, the lack of clamps reduced contact forces in both testing rigs; the clamps were therefore useful in simulating medial sutures. We hypothesized that the presence of the thin Tekscan sensor in the patellofemoral joint would not affect the kinematics of the joint. We performed tests without the Tekscan sensor in place and found that, in general, the removal of the Tekscan sensor had little effect on kinematics. The changes in tracking were typically less than the variability of the changes between specimens. To determine the validity of measurement systems that determine tracking or loads at static, fixed flexion angles (such as radiographs or Fuji film) instead of dynamically, we took static measurements of tracking in both rigs at several angles. In contrast to the conclusions of Brossmann (1993), we found that static measurements are adequate assessments of patellar tracking; tilt and shift were not greatly affected by measuring tracking statically instead of dynamically. Contact force was different between static and dynamic flexion (maximum 20% and 22% difference in the horizontal and vertical rigs, respectively); static and dynamic force appear to differ when measured using Tekscan pressure sensors. Following our tests of possible confounding factors in the vertical rig, we performed a lateral retinacular release. The lateral release did not have an effect on patellar tracking or loading; this conclusion has been supported by clinical tracking studies (Bindelglass 1993, Chan 1999, Gomes 1988, Kawano 2002, Laughlin 1996) and one in vitro study on loading (Singerman 1997). The conclusion of the clinical studies may be misleading, however, because in all studies, lateral releases were only performed when required and were not studied as a separate, independent variable. It was not possible to measure the changes in tracking due to a lateral release on the same patient. Our kinetic conclusion must be interpreted with caution because we were only able to obtain force measurements following lateral release for one specimen. 5.13 Limitations In vitro studies such as ours have a common set of limitations. Although they can suggest relationships between surgical variables and patellar mechanics, in vitro studies cannot determine direct links between knee mechanics and pain. It is hoped that the protocol and results of this study will guide a future clinical study investigating the relationship between pain and the most 166 relevant surgical variables we studied, such as patellar resection angle, femoral component rotation, and patellar medialization. In vitro studies are often performed with a relatively small population of specimens, compared to clinical studies. Our study included 8 specimens, which seems like a small number but is typical for knee studies. We observed a great deal of variability in patellar mechanics, especially contact forces, between specimens. This variability, which has also been reported by other researchers (see section 5.2), makes it difficult to create general guidelines for optimal component placement that apply to all T K A patients. It is interesting to note that Miller (2001a, 2001b) reported different patellar tracking in two studies that had identical testing protocol. In our study and all the other in vitro studies we reviewed, soft tissue releases were not done, even i f they were warranted by a change in component placement, because they would influence kinematics and bias the results. As a result, there may have been abnormal tension in the soft tissues. We could not suture the medial opening in the knee closed because the Tekscan sensor exited the joint through that space; instead, we used a pair of surgical towel clamps to temporarily reattach the medial tissues to the quadriceps muscles above the opening. This may have resulted in decreased soft tissue tensions and could have led to more lateral tracking. Proponents of open-chain (hanging knee) testing rigs such as those used by Armstrong (2003) have critiqued the Oxford rig because it applies tension to only the quadriceps tendon instead of dividing the load between the individual quadriceps muscle groups. They claim that the separation of loads may result in more physiologically accurate patellar motion. Although the individual load vectors can be summed to create a vector pulling in just one direction, this vector may be located at a distance from the centre of the patella and cause the patella to spin. To our knowledge, the differences in tracking or loading caused by dividing the quadriceps load into the muscle groups have not been determined. If the division of load into individual vectors does in fact affect the spin of the patella, we think this effect is minimized in open-chain rigs because the direction of pull for each muscle group is mainly proximal (via pulleys at the proximal end of the femur). The significant disadvantage of open-chain rigs is that they do not load the knee at a physiological load. The Oxford rig applies physiological axial forces to the knee and simulates 167 postoperative loading, whereas the loads applied to the knee in an open-chain rig are via the quadriceps group only and do not include the compressive loads across the joint due to weight-bearing. Hanging knee, open-chain rigs may be appropriate to simulate certain clinical tests; however, they do not apply loads on the order of body weight. Changes to tissue properties may occur during freezing and thawing of specimens, although Foutz (1992) and Woo (1986) concluded that careful freezing had little or no effect on the biomechanical properties of tissues. In our study, the specimens were first thawed when Dr. Tonetti performed surgery on them. They were refrozen following surgery and were thawed a second time prior to testing. As our testing took place over two days, they were refrigerated between testing days. We could not test the effects of resurfacing the patella due to time limitations. To test the effects of resurfacing in both testing rigs, we would need to switch testing rigs an additional 2 times, and we would also need to perform the patellar resection during testing instead of on a separate day prior to testing. Resurfacing the patella may increase contact pressure by 1.1-4 times (Fuchs 2000, Hsu 1996, Matsuda 1997, Stukenborg-Colsman 2003). Contact areas are larger with the natural patella than with the resurfaced patella; the forces measured in this study cannot likely be generalized to the case of the natural patella. The large pressures resulting from the small contact areas with the resurfaced patella may have contributed to sensor degradation (see section 2.7.2.1); the high loads concentrated over small surface areas may have created the "severe conditions" and "rough handling" that reduce the useful life of a sensor (Tekscan 2001). Tekscan sensors may be more appropriate for studies of the natural patella. Due to time limitations (each specimen could only be put through a limited number of cycles) and due to Tekscan sensor degradation, we could not test effects of pairing several surgical variables together as we had initially intended. For example, we would have liked to determine the changes in kinematics and kinetics following simultaneous 5° internal rotation of both the femoral and tibial components. Patellar height (patella alta or patella baja) appears to have a significant impact on patellar tracking and loading; however, this parameter is not often changed as part of standard surgical technique: the height of the patella is greatly affected by the height of the tibial cut. 168 We only tested one rotation angle for both femoral and tibial component rotation; it would be interesting to test the effects of a wider range of rotations at several angular increments. It may be that a tibial component rotation of 5°, although clinically realistic, is not large enough to induce noticeable changes in patellar tracking. During testing with problematic specimens, data were not collected for some variables. This was often because we were concerned that an extremely malpositioned component (such as increased patellar thickness) would increase loads substantially and damage the Tekscan sensor. During statistical analysis, it was necessary to fill the data gaps with the averages of the other specimen data to perform A N O V A s . A sensitivity analysis determined that this had very little effect on our conclusions regarding statistically significant changes in tracking. We were not able to collect data for specimen 6 in the vertical rig because the knee had deteriorated and the modified components would not remain fixed in the poor bone (see Appendix 4). As a result, specimen 6 could not be included in the two-way A N O V A comparing horizontal and vertical kinematics. We performed post-hoc paired t-tests with a Bonferroni correction factor when statistical differences were found using the two-way A N O V A . The Bonferroni method is more conservative than other methods such as Tukey's or Sheffe's methods (IOA 2006), which may have resulted in an under-reporting of significant results; however, the Bonferroni method allowed us to select which pairwise comparisons we wished to evaluate instead of forcing us to compare each possible pair of means in the data set. We did not study the statistical interaction between testing rig and change in component placement. An interaction indicates that the dependent variables (tracking indices) depend not only on the independent variables (factors) considered separately, but also on how they are combined. It is possible that tracking may have been affected by the combination of rig and change in component placement. Kinematic data were not smoothed or filtered. We do not believe that this resulted in very large errors or differences in findings because the data were not very noisy. Filtering of kinematic data is not typically reported in other studies of patellofemoral kinematics and kinetics. We had difficulties with dislocating femurs and torn quadriceps tendons; the details of these problems and the methods we used to deal with them are found in Appendix 4. 169 The results of our study may be somewhat limited to the particular design of components that we used. Although Chew (1997) found no differences in tracking for 3 different knee designs, Yoshii (1999) found significant differences in tracking for 2 different knee designs. It has been shown that loading differs significantly between knee designs (Singerman 1997). We would have liked to measure the quadriceps tension in the vertical rig; however, this was not possible due to spatial constraints in our loading setup (see section 2.6.2.2.3). Because Tekscan pressure sensors can only measure loads normal to their surface, we could not measure shear loads acting on the patella. These loads are low compared to the normal forces (Singerman 1997) but are still large enough to damage Tekscan sensors and may therefore affect tracking. A n ideal pressure sensor would also be able to measure shear loads. Although Novel claims that it manufactures sensors capable of measuring shear forces, it does not produce sensors which measure both normal and shear forces simultaneously. No ideal force or pressure measurement system exists. Load cells likely affect patellar tracking and loading due to their size and weight, and they cannot calculate contact area or pressure. Fuji film is relatively thick (0.3 mm), which may alter the mechanics of the knee and result in over- or underestimation of contact pressures by up to 26% (Liau 2002a, Wu 1998). Fuji film underestimates contact areas by as much as 36% (Harris 1999, Matsuda 1995b) and cannot take dynamic measurements. Novel pressure sensors are twice as thick as Fuji film (0.6 mm - 1 mm) and have poor resolution and inadequate pressure ranges. As detailed in section 2.7.2.1, Tekscan sensors are subject to significant output degradation. Other pressure sensors exist (e.g. X-sensor), but they have not been widely used and their accuracy is unknown. A custom-designed system would require significant development. We attempted to account for degradation during our force analysis; however, the data (especially from the vertical rig) was highly variable which made it difficult to draw conclusions. Even during testing of specimens 7 and 8, when the issue of degradation had been substantially resolved, we still observed a great deal of variability in force output. Only 5 specimens were available for force comparisons in both testing rigs; degradation of sensor output and one punctured sensor limited the number of variables available in the first specimens (see Appendix 170 5). It would have been useful to recalibrate the sensors following testing to determine i f sensor response changed over time and if the initial calibration was still accurate. To deal with the issue of degradation, we normalized the force data to the first set of baselines, as described in section 2.9.2. We assumed that degradation was linear between baseline tests. In the horizontal rig, force output decreased by an average of 8% between baseline tests and in the vertical rig, force output decreased by an average of 15% between baseline tests. Using these values of degradation as conservative guidelines, we may assume that errors due to the linear interpolation scheme could result in approximately 8-15% error in reported force. Because we did not observe statistically significant changes in peak contact force over the range of tested variables, it may have been prudent to perform a power analysis to determine how many additional specimens would be required to observe a significant effect. However, this was deemed unnecessary because we believed that Tekscan force data was too unreliable, erratic and variable to yield significant results for a = 0.05, even with an increased number of specimens. Although the data was unreliable and the results were not statistically significant for this conventional confidence interval, we believe that a discussion of the apparent trends in force data was warranted. The selection of an extremely tight confidence interval may be inappropriate for extremely variable force data; many of the apparent trends may have been statistically significant for a less conservative alpha value. Changes in postoperative tracking can generally be predicted by changes in intraoperative tracking; however, the term 'postoperative' must be qualified: tracking may change in the follow-up period after surgery. Laughlin (1996) observed deterioration in tracking over time. They looked at the effect of follow-up time on patellar tilt; similar to the findings of Shih (2004), they observed that patellae tilted more laterally with time. Between the 3 week visit and the most recent follow-up (mean 3 years), the number of patellae tilting more than 5° laterally increased from 32% to 46%. It was also observed that medially tilting patellae moved towards neutral. This may be due to medial repairs stretching over time or to scar contracture if a lateral release was performed. Although we believe that our intraoperative and postoperative loading simulations are good models of in vivo loading, some differences likely exist between the loading characteristics. 171 During in vitro testing, the loads applied to the knee are not usually physiological because cadaver knees cannot withstand the same loads as living legs. Also, the lack of muscle tone during in vitro studies (the muscle is mostly flaccid) may limit the applicability of the results to in vivo situations. It was important to simulate the differences in loading between intraoperative and postoperative configurations. We could not simulate a physiological Q-angle during testing because this caused the patella to sublux laterally, likely due to the lack of medial soft tissue tension (see section 2.6.2.2.4); however, the direction of pull (in line with the femoral shaft) was identical between testing rigs, as it is between configurations in vivo. The reduced lateral angle of pull of the quadriceps tendon likely reduced the lateral patellar contact forces. We would like to validate other conclusions regarding our intraoperative and postoperative models by comparing kinematic data in vivo both intraoperatively and postoperatively (see section 5.14 below). 5.14 Contributions and Future Work This study is a stepping stone for future clinical studies relating optimal intraoperative patellar tracking and loading to optimal postoperative mechanics. It is hoped that the results of this study will focus attention on the most relevant surgical factors influencing patellar mechanics. "None of the existing [computer-assisted surgery] systems deals with the functioning of the patellofemoral joint except through the determination of the rotational positioning of the femoral and tibial components" (Nizard 2002). We plan to use the results of this study to collaborate with Praxim Medivision (Grenoble, France) and to develop a computer assisted surgical (CAS) technique for resurfacing the patella. Future clinical studies should attempt to link postoperative pain to intraoperative tracking or loading measures. Using a CAS approach to the patella, it would be possible to measure the kinematics and kinetics during surgery, and these measures could be studied in conjunction with results from patient surveys on pain and function. 172 Our study was the first step in forming a link between intraoperative and postoperative patellar mechanics; the next step must be to compare intraoperative and postoperative mechanics in the clinical setting. Praxim Medivision is currently collecting intraoperative tracking data in vivo; it would be valuable to also collect data postoperatively, both immediately following surgery and at two different follow-up periods (since tracking may change with time). A comparison of intraoperative and postoperative tracking could be used to validate our intraoperative and postoperative testing rigs for use in future biomechanical studies. Future studies of patellar tracking should attempt to measure the location and angle of the patellar component relative to the femoral component; current studies have measured the tracking of the patellar bone remnant with respect to the femoral bone, which does not reveal the true changes in tracking. High-speed magnetic resonance imaging (MRI) may be a useful tool to compare tracking in the two loading configurations, although the presence of metal in the joint could be a large issue. We hope that the protocol of this study will help guide the collection of kinematic and kinetic data in the clinical setting. We discovered that user-defined calibration algorithms yield more accurate measurements of force than the algorithms supplied in the Tekscan system software (see Chapter 3). We have also determined that Tekscan sensors would be unsuitable for measuring patellofemoral forces for the resurfaced patella in vivo (the sensors may be more reliable with unresurfaced patellae). Future in vivo studies should investigate the measurement differences between Tekscan pressure sensors and implanted load cells. It would be interesting to attach a sensor to a patellar load cell and measure forces in the knee in vitro with both instruments simultaneously. A better method of measuring force needs to be developed for the patella in vivo (load cells cannot be used in vivo because large holes must be drilled through the patella for attachment purposes). It would be interesting to measure tightness in the lateral retinaculum during testing because it has been hypothesized that soft tissue tension may be related to anterior knee pain. The data from this study are also being used by Dr. Carolyn Anglin to determine the change in location of the centre of pressure on the patella throughout flexion for each change in component placement. This relates to the mediolateral force balance, which is also a surgical goal. It is less sensitive to sensor degradation because it is essentially a relative measure. The results are 173 therefore smoother, less variable and more reliable, providing a kinetic match to the kinematic data. 174 6 Conclusions In this study, we assessed the effect of T K A component placement on patellar tracking and contact forces in the patellofemoral joint. We studied patellar kinematics and kinetics in both intraoperative and postoperative loading configurations. Our research questions are listed below with a summary of our findings relating to each question. Question 1: What are the effects of varying the femoral and tibial components on patellar tracking and loading? External rotation of the femoral component tilted and shifted the patella laterally in early and mid-flexion; however, the patella shifted medially in late flexion. Internal component rotation had the opposite effect on patellar tilt and shift. The observed changes in tilt and shift were mainly due to rotation of the groove in which the patella travels. The measured tilt and shift were similar to the values that were predicted due to geometrical changes. Patellar spin was only affected by changes in component rotation in late flexion. Both external and internal femoral component rotation increased patellar contact loads, although this increase was not statistically significant. Tibial component rotation had a minimal and variable effect on kinematics. Contact loads were affected by tibial tray rotation; however, this effect was opposite in the horizontal and vertical testing rigs and the trends were not statistically significant. Question 2: What are the effects of modifying the patellar component placement (patellar bone cut angle, mediolateral position, thickness of patellar construct) on patellar tracking and loading? As expected (due to geometry), patellar bone cut angle had a large influence on the tilt of the patellar bone remnant; however, it had a relatively small effect on the position and orientation of the patellar component within the femoral groove. Medial patellar angles increased patellofemoral forces, and lateral angles decreased contact forces, although these changes were not statistically significant. 175 Medializing the patellar implant had a large effect on both tilt and shift of the patellar bone remnant; it resulted in lateral shift of the bony patella, as expected, and lateral tilt of the entire patellar construct, which was unexpected. We observed very little shift of the patellar implant within the femoral groove following medialization. Medializing the patellar component had little effect on loads in the horizontal rig and decreased patellar loads in the vertical rig. Changes in contact forces were non-significant extremely variable in both testing rigs. Patellar thickness had a minimal effect on kinematics but tended to increase normal patellar loads non-significantly. Patellar spin was not affected by changes in patellar component placement. Question 3: How do the effects of altering the position of the femoral, tibial and patellar components compare relative to each other (ranking) in terms of: a) kinematics, and b) kinetics. Tibial rotation and patellar thickness had very little impact on patellar tracking, compared to other surgical variables. Patellar resection angle, femoral component rotation and patellar medialization were observed to have the greatest effects on the tracking of the patellar bone remnant relative to the femur. It is difficult to discern the effects of these three variables on tracking of the patellar implant relative to the femoral component because we measured tilt and shift of the bony remnant relative to the femoral bone instead of the patellar implant relative to the femoral groove; however, it can be inferred from geometrical relationships that the patellar component maintained a relatively constant position and angle within the femoral groove. The variables which had the greatest effect on patellar loading differed between intraoperative and postoperative simulations. Intraoperatively, patellar thickness increased contact forces; postoperatively, medializing the patellar component reduced contact loads. Neither change was statistically significant. Changes in placement of the patellar component affected patellar contact loads; however, the effects were highly variable between specimens. No changes in peak force were statistically significant. Question 4: How do the aforementioned effects compare in intraoperative and postoperative simulations? 176 Changes in patellar tilt and shift were well-correlated between the two testing configurations; however, absolute values of tilt and shift were not as well-correlated. Both absolute and relative measures of spin were not correlated between testing rigs. Changes in intraoperative and postoperative forces were not well-correlated. Question 5: Does a relationship exist between patellar kinetics and kinematics, and i f so, what is the relationship? We were not able to determine a relationship between tracking and contact loads. Changes in tracking did not always correspond to changes in loading, and vice versa. Question 6: What is the accuracy of the calibration algorithms supplied in the Tekscan system software, and do user-defined calibration routines provide more accurately calibrated force output? The Tekscan power calibration is the most accurate calibration provided in the system software; it measured force to within 2.7% accuracy, whereas linear calibrations were between 10.5% and 24.4% accurate. User-defined polynomial calibrations were markedly more accurate; 3-point quadratic and 10-point cubic calibrations were accurate to within 1.5% and 0.6% of the sensor range, respectively. Sensor output and nonlinearity depend on the experimental protocol (sensor type, interface shape and materials, sensor range in use, loading method.); therefore, sensor behaviour should be investigated for each different application. 177 References Abraham W, Buchanan JR, Daubert H, Greer RB 3rd, Keefer J. (1988) Should the patella be resurfaced in total knee arthroplasty? Efficacy of patellar resurfacing. Clinical Orthopaedics and Related Research 236, 128-134. Aglietti P, Buzzi R, Gaudenzi A . (1988) Patellofemoral functional results and complications with the posterior stabilized total condylar knee prosthesis. Journal of Arthroplasty 3, 17-25. Aglietti P, Insall JN, Walker PS, Trent P. (1975) A new patella prosthesis. Clinical Orthopaedics and Related Research 107, 175-187. Ahmed A M , Burke DL , Hyder A. (1987) Force analysis of the patellar mechanism. Journal of Orthopaedic Research 5, 69-85. Ahmed A M , Burke DL, Y u A . (1983) In-vitro measurement of static pressure distribution in synovial joints- Part II: Retropatellar surface. Journal of Biomechanical Engineering 105(3), 226-36. Akagi M , Matsusue Y , Mata T, Asada Y , Horiguchi M , Iida H, Nakamura T. (1999) Effect of rotational alignment on patellar tracking in total knee arthroplasty. Clinical Orthopaedics and Related Research 366, 155-163. Anderson IA, MacDiarmid A A , Harris M L , Gillies R M , Phelps R, Walsh WR. (2003) A novel method for measuring medial compartment pressures within the knee joint in-vivo. Journal of Biomechanics 36, 1391-1395. Anglin C, Hodgson A J , Masri B A , Greidanus N V , Garbuz DS. (2004) Quality of life before and after total knee arthroplasty. Canadian Arthritis Network Annual Scientific Conference, Vancouver, BC, Oct 27-29, 2004. Anouchi YS , Whiteside L A , Kaiser A D , Milliano MT. (1993) The effects of axial rotational alignment of the femoral component on knee stability and patellar tracking in total knee arthroplasty demonstrated on autopsy specimens. Clinical Orthopaedics and Related Research 287, 170-177. Arizono T, Emoto G, Whiteside L. (1999) Effect of malrotation on tibiofemoral articular contact pressure and kinematics with a conforming tibial articulation in total knee replacement. Transactions of the Orthopaedic Research Society 24, 146. Armstrong A D , Brien HJC, Dunning CE, King GJW, Johnson JA, Chess DG. (2003) Patellar position after total knee arthroplasty. Journal of Arthroplasty 18(4), 458-465. Barnes C L , Scott RD. (1993) Patellofemoral complications of total knee replacement. Instructional Course Lectures 42, 303-307. 178 Barrack RL , Schrader T, Bertot AJ , Wolfe M W , Myers L. (2001) Component rotation and anterior knee pain after total knee arthroplasty. Clinical Orthopaedics and Related Research 392, 46-55. Barrack RL , Wolfe M W , Waldman DA, Milicic M , Bertot AJ , Myers L. (1997) Resurfacing of the patella in total knee arthroplasty: a prospective, randomized, double-blinded study. Journal of Bone and Joint Surgery 79A, 1121-1131. Berger R A , Crossett LS, Jacobs JJ, Rubash HE. (1998) Malrotation causing patellofemoral complications after total knee arthroplasty. Clinical Orthopaedics and Related Research 356, 144-153. Berger R A , Rubash HE, Seel MJ , Thompson WH, Crossett LS. (1993a) Determining the rotational alignment of the femoral component in total knee arthroplasty using the epicondylar axis. Clinical Orthopaedics and Related Research 286, 40-47. Berger RA, Seel MJ , Schleiden M , Britton CA, Rubash HE. (1993b) Computerized tomographic determination of the normal tibiofemoral rotational angle: A guide to tibial component rotational alignment in T K A . Transactions of the Orthopaedic Research Society 17(3), 1174. Biden E N , O'Connor JJ. (1990) Experimental methods used to evaluate knee ligament function. In Daniel D M , Akeson WA, O'Connor J J (eds) Knee Ligaments: Structure, Function, Injury and Repair. New York, Raven Press, 135-151. Bindelglass DF, Cohen JL, Dorr LD. (1993) Patellar tilt and subluxation in total knee arthroplasty: relationship to pain, fixation, and design. Clinical Orthopaedics and Related Research 286, 103-109. Bishop RED, Denham RA. (1977) A note on the ratio between tensions in the quadriceps tendon and infrapatellar ligament. Journal of Engineering in Medicine 6, 53-54. Bourne RB, Goodfellow JW, O'Connor JJ. (1978) A functional analysis of various knee arthroplasties. Transactions of the Orthopaedic Research Society 24, 160. Brander V A , Stulberg DS, Adams A D , Harden R N , Bruehl S, Stanos SP, Houle T. (2003) Predicting total knee replacement pain: a prospective, observational study. Clinical Orthopaedics and Related Research 416, 27-36. Brossmann J, Muhle C, Schroder C, Melchert U H , Bull CC, Spielmann RP, Heller M . (1993) Patellar tracking patterns during active and passive knee extension: evaluation with motion-triggered cine M R imaging. Radiology 187, 205-212. 179 Brown TD, Rudert MJ , Grosland N M . (2004) New methods for assessing cartilage contact stress after articular fracture. Clinical Orthopaedics and Related Research 423, 52-58. Browne C, Hermida JC, Bergula A , Colwell CW, D'Lima DD. (2005) Patellofemoral forces after total knee arthroplasty: effect of extensor moment arm. Knee 12(2), 81-88. Buff H U , Jones L C , Hungerford DS. (1988) Experimental determination of forces transmitted through the patello-femoral joint. Journal of Biomechanics 21, 17-23. Bullock SJ. (2003), Development and system integration of a computer-assisted total knee replacement technique [Masters thesis]: University of British Columbia, Vancouver, 135 p. Canadian Institute for Health Information (CIHI). Canadian Joint Replacement Registry (CJRR). (2005) 2005 Report: Total Hip and Total Knee Replacements in Canada. Ottawa: CIHI. Chan K C , Gi l l GS. (1999) Postoperative patellar tilt in total knee arthroplasty. Journal of Arhtroplasty 14(3), 300-304. Chapman-Sheath PJ, Bruce WJM, Chung WK, Morberg P, Gillies R M , Walsh WR. (2003) In vitro assessment of proximal polyethylene contact surface areas and stresses in mobile bearing knees. Medical Engineering and Physics 25(6), 437-443. Chew JTH, Stewart NJ, Hanssen A D , Zong-Ping L, Rand JA, An K - N (1997) Differences in patellar tracking and knee kinematics among three different total knee designs. Clinical Orthopaedics and Related Research 345, 87-98. Churchill DL, Incavo SJ, Johnson CC, Beynnon BD. (1998) The transepicondylar axis approximates the optimal flexion axis of the knee. Clinical Orthopaedics and Related Research 356, 111-118. Cole GK, Nigg B M , Ronsky JL, Yeadon MR. (1993) Application of the joint coordinate system to three-dimensional joint attitude and movement representation: a standardization proposal. Journal of Biomechanical Engineering 115, 344-349. Conlan T, Garth WP, Lemons JE. (1993) Evaluation of the medial soft-tissue restraints of the extensor mechanism of the knee. Journal of Bone and Joint Surgery 75 A , 682-693. Csintalan RP, Schulz M M , Woo J, McMahon PJ, Lee TQ. (2002) Gender differences in patellofemoral joint biomechanics. Clinical Orthopaedics & Related Research 402, 260-269. Dennis DA. (1992) Patellofemoral complication in total knee arthroplasty. American Journal of Knee Surgery 5, 156-166. 180 D'Lima DD, Chen PC, Colwell CW. (2001a) Polyethylene contact stresses, articular congruity, and knee alignment. Clinical Orthopaedics and Related Research 392, 232-238. D'Lima DD, Chen PC, Kester M A , Colwell CW. (2003) Impact of patellofemoral design on patellofemoral forces and polyethylene stresses. Journal of Bone and Joint Surgery 85A (Suppl. 4), 85-93. D'Lima DD, Poole C, Chadha H, Hermida JC, Mahar A , Colwell CW. (2001b) Quadriceps moment arm and quadriceps forces after total knee arthroplasty. Clinical Orthopaedics and Related Research 392, 213-220. D'Lima DD, Trice M , Urquhart A G , Colwell CW. (2000) Comparison between the kinematics of fixed and rotating bearing knee prostheses. Clinical Orthopaedics and Related Research 380, 151-157. D'Lima DD, Trice M , Urquhart A G , Colwell CW. (2001c) Tibiofemoral conformity and kinematics of rotating-bearing knee prostheses. Clinical Orthopaedics and Related Research 386, 235-242. Elias JJ, Cech JA. Weinstein D M , Cosgrea AJ . (2004a) Reducing the lateral force acting on the patella does not consistently decrease patellofemoral pressures. American Journal of Sports Medicine 32(5), 1202-1208. Elias JJ, Mattessich SM, Kumagai M , Mizuno Y , Cosgarea AJ , Chao E Y . (2004b) In vitro characterization of the relationship between the Q-angle and the lateral component of the quadriceps force. Journal of Engineering in-Medicine 218(1), 63-67. Elias JJ, Wilson DR, Adamsonc R, Cosgrea AJ . (2004c) Evaluation of a computational model used to predict the patellofemoral contact pressure distribution. Journal of Biomechanics 37, 295-302. Ezzet K A , Hershey A L , D'Lima DD, Irby SE, Kaufman KR, Colwell CW. (2001) Patellar tracking in total knee arthroplasty - Inset versus onset design. Journal of Arthroplasty 16 (7), 838-843. Foutz TL, Stone EA, Abrams CF. (1992) Effects of freezing on mechanical-properties of rat skin. American Journal of Veterinary Research 53 (5), 788-792. Fuchs S, Schutte G, Witte H , Rosenbaum D. (2000) Retropatellar contact characteristics in total knee arthroplasty with and without patellar resurfacing. International Orthopaedics 24, 191-193. 181 Fuchs S, Schutte G, Witte H, Rosenbaum D. (2002) What influence do size and placement of patella resurfacing have on knee endoprosthesis? Unfallchirurg 105(1), 44-48. Fuiko R, Kotten B, Zettl R, Ritschl P. (2003) Landmarks for navigated implantation of total knee arthroplasty (TKA). In Langlotz F, Davies BL , Bauer A (Eds): 3rd Annual Meeting of CAOS-International Proceedings, 108-109. Gerber BE, Maenza F. (1998) Shift and tilt of the bony patella in total knee replacement. Orthopade 27, 629-636. Gi l l T, DeFrate L, Carey C, Zayontz S, Golden D, Zarins B, L i G. (2003) Variation of patello-femoral joint contact pressures after posterior cruciate ligament reconstruction. 49th Annual Meeting of the Orthopaedic Research Society. New Orleans, Louisiana. Gil l TJ, DeFrate L E , Wang C, Carey CT, Zayontz S, Zarins B, L i G. (2004) The effect of posterior cruciate ligament reconstruction on patellofemoral contact pressure in the knee joint under simulated muscle loads. American Journal of Sports Medicine 32, 109-115. Godest A C , Beaugonin M , Haug E, Taylor M , Gregson PJ. (2002) Simulation of a knee joint replacement during a gait cycle using explicit finite element analysis. Journal of Biomechanics 35(2), 267-275. Gomes L S M , Bechtold JE, Gustilo RB. (1988) Patellar prosthesis positioning in total knee arthroplasty. Clinical Orthopaedics and Related Research 236, 72-81. Goodfellow JW, Hungerford DS, Zindel M . (1976) Patello-femoral mechanics and pathology I: functional anatomy of the patello-femoral joint. Journal of Bone and Joint Surgery 58B, 287. Grelsamer RP, Bazos A N , Proctor CS. (1993) Radiographic analysis of patellar tilt. Journal of Bone and Joint Surgery 75B, 822-824. Grelsamer RP, Dubey A , Weinstein CH. (2005) Men and women have similar Q angles. Journal of Bone and Joint Surgery 87B, 1498-1501. Grelsamer RP, McConnell J. (1998) The Patella: A Team Approach. Aspen Publishers, Inc.: Gaithersburg, Maryland. Grood ES, Suntay WJ. (1983) A joint coordinate system for the clinical description of three-dimensional motions: application to the knee. Journal of Biomechanical Engineering 105, 136-144. Grood ES, Suntay WJ, Noyes FR, Butler DL. (1984) Biomechanics of the knee-extension exercise. Journal of Bone and Joint Surgery 66A, 725-734. 182 Halloran JP, Petrella AJ , Rullkoetter PJ. (2005) Explicit finite element modeling of total knee replacement mechanics. Journal of Biomechanics 38(2), 323-331. Harris M L , Morberg P, Bruce WJM, Walsh WR. (1999) An improved method for measuring tibiofemoral contact area in total knee arthroplasty: a comparison of K-scan sensor and Fuji film. Journal of Biomechanics 32, 951-958. Healy WL, Wasilewski SA, Takei R, Oberlander M . (1995) Patellofemoral complications following total knee arthroplasty: correlation with implant design and patient risk factors. Journal of Arthroplasty 10, 197-201. Heegaard JH, Leyvraz PF, Hovey CB. (2001) A computer model to simulate patellar biomechanics following total knee replacement: the effects of femoral component alignment. Clinical Biomechanics 16, 415-423. Heegaard JH, Leyvraz PF, van Kampen A, Rakotomanana L, Rubin PJ. (1993) Patellar stability as a function of joint anatomy. Journal of Bone and Joint Surgery 75B(suppl), 140-141. Hirokawa S. (1991) Three-dimensional mathematical model analysis of the patellofemoral joint. Journal of Biomechanics 24, 659-671. Hofmann S, Romero J, Roth-Schiffl E, Albrecht T. (2003) Rotational malalignment of the components may cause chronic pain or early failure in total knee arthroplasty. Orthopade 32(6), 469-476. Hofmann A A , Van Gorp C. (1998) Patellofemoral resurfacing: the value of instrumentation and design. Journal of Arthroplasty 13(4), 459-463. Hsu H-C, Zong-Ping L, Rand JA, An K - N . (1996) Influence of patellar thickness on patellar tracking and patellofemoral contact characteristics after total knee arthroplasty. Journal of Arthroplasty 11(1), 69-80. Huberti H H , Hayes WC. Patellofemoral contact pressures. (1984a) Journal of Bone and Joint Surgery 66A, 715-724. Huberti H H , Hayes WC, Stone JL, Shybut GT (1984b) Force ratios in the quadriceps tendon and ligamentum patellae. Journal of Orthopaedic Research 2, 49-54. Hungerford DS, Barry M . (1979) Biomechanics of the patellofemoral joint. Clinical Orthopaedics and Related Research 144, 9-15. Huo M H , Sculco TP. (1990) Complications in primary total knee arthroplasty. Orthopaedic Review 19, 781-788. 183 Insall IN, Binazzi R, Soudry M , Mestriner L A . (1985) Total knee arthroplasty. Clinical Orthopaedics and Related Research 192, 13-22. IOA. (2006) Institute on Aging, Portland State University. http://vvww.ioa.pdx.edu/newsom/dal/ho_posthoc.doc. Accessed April 8, 2006. Jerosch J, Peuker E, Philipps B, Filler T. (2002) Interindividual reproducibility in perioperative rotational alignment of femoral components in knee prosthetic surgery using the transepicondylar axis. Knee Surgery, Sports Traumatology, Arthroscopy, 10(3), 194-197. Josefchak RG, Finlay JB, Bourne RB, Rorabeck CH. (1987) Cancellous bone support for patellar resurfacing. Clinical Orthopaedics and Related Research 220, 191-199. Katchburian M V , Bull A M J , Shih Y-F, Heatley FW, Amis A A . (2003) Measurement of patellar tracking: assessment and analysis of the literature. Clinical Orthopaedics and Related Research 412 , 241-259. Kaufer H. (1971) Mechanical function of the patella. Journal of Bone and Joint Surgery 5 3 A , 1551-1560. Kawano T, Miura H , Magamine R, Urabe K, Matsuda S, Takaaki M-O, Iwamoto Y . (2002) Factors affecting patellar tracking after total knee arthroplasty. Journal of Arthroplasty 17(7), 942-947. Kdolsky RK, Arabid B A , Fuchs M , Schabus R, Vecsei V . (2004) Measuring the pressure pattern of the joint surface in the uninjured knee. Wiener Klinische Wochenschrift 116 (5-6), 196-200. Kessler OC, Moctezuma de la Barrer JL, Streicher R M , Stockl B, Krismer M , Nogler M . (2003) Variation in the entry point selection for the transepicondylar axis and its quantitative effects. In Langlotz F, Davies B L , Bauer A (Eds): 3rd Annual Meeting of CAOS-International Proceedings, 176-177. Kessler O, Widmer K H , Moctezuma J. (2004) Influence of femoral component rotation on patella kinematics. 4 t h Annual Meeting of the International Society for Computer Assisted Orthopaedic Surgery. June 16-19, Chicago, Illinois. Kienapfel H , Springorum H-P, Ziegler A , Klose K-J , Georg C, Griss P. (2003) Effect of rotation of the femoral and tibial components on patellofemoral malalignment in knee arthroplasty. Orthopade 32, 312-318. Kim BS, Reitman RD, Schai PA, Scott RD. (1999) Selective patellar nonresurfacing in total knee arthroplasty: 10 year results. Clinical Orthopaedics and Related Research 367, 8 1 - 8 8 . 184 Laman B L , White JT, Thompson MT, Conditt M A , Noble PC. (2004) Tibial and patellar component forces during kneeling after total knee replacement. 50th Annual Meeting of the Orthopaedic Research Society. March 7-10, San Francisco, California. Lane JG, Irby SE, Kaufman K, Rangger C, Daniel D M . (1994) The anterior cruciate ligament in controlling axial rotation. American Journal of Sports Medicine 22(2), 289-293. Laskin RS. (1995) Flexion space configuration in total knee arthroplasty. Journal of Arthroplasty 10, 657-660. Laughlin RT, Werries B A , Verhulst SJ, Hayes JM. (1996) Patellar tilt in total knee arthroplasty. American Journal of Orthopedics 23, 300-304. Lee TQ, Budoff JE, Glaser FE. (1999) Patellar component positioning in total knee arthroplasty. Clinical Orthopaedics and Related Research 366, 274-281. Lee TQ, Gerken AP, Glaser FE, Kim WC, Anzel SH. (1997) Patellofemoral joint kinematics and contact pressures in total knee arthroplasty. Clinical Orthopaedics and Related Research 340, 257-266. Lee TQ, Yang B Y , Matthew DS, McMahon PJ. (2001) The effects of tibial rotation on the patellofemoral joint: assessment of the changes in in situ strain in the peripatellar retinaculum and the patellofemoral contact pressures and areas. Journal of Rehabilitation Research and Development 38(5). L i G, DeFrate L E , Zayontz S, Park SE, Gi l l TJ. (2004) The effect of tibiofemoral joint kinematics on patellofemoral contact pressures under simulated muscle loads. Journal of Orthopaedic Research 22, 801-806. L i PLS, Zamora J, Bentley G. (1999) The results at ten years of the Insall-Burstein II knee replacement: clinical, radiographical and survivorship studies. Journal of Bone and Joint Surgery 81B, 647-653. Liau JJ, Cheng CK, Huang CH, Lo WH. (2002a) Effect of Fuji pressure sensitive film on actual contact characteristics of artificial tibiofemoral joint. Clinical Biomechanics 17, 698-704. Liau JJ, Cheng CK, Huang C H , Lo WH. (2002b) The effect of malalignment on stresses in polyethylene component of total knee prostheses - a finite element analysis. Clinical Biomechanics 17(2), 140-146. Mantas JP, Bloebaum RD, Skedros JG, Hofmann A A . (1992) Implications of reference axes used for rotational alignment of the femoral component in primary and revision knee arthroplasty. Journal of Arthroplasty 7, 531-535. 185 Marmor L. (1988) Technique for patellar resurfacing in total knee arthroplasty. Clinical Orthopaedics and Related Research 230, 166-167. Masri B, Mountey J, Paice M , Greidanus N , Wilson D. (2005) A kinetic and kinematic analysis of a new trabecular metal augmentation patella. 57 t h Annual Meeting of the Association of Bone and Joint Surgeons. June 8-12, Carmel, CA. Matsuda S, Ishinishi T. (1995a) The patellofemoral joint after total knee arthroplasty: the effect on contact area and contact stress. Transactions of the Orthopaedic Research Society 19(2), 337-338. Matsuda S, Ishinishi T, White SE, Whiteside L A . (1997) Patellofemoral joint after total knee arthroplasty: effect on contact area and contact stress. Journal of Arthroplasty 12(7), 790-797. Matsuda S, Ishinishi T, Whiteside L A . (2000) Contact stresses with an unresurfaced patella in total knee arthroplasty: the effect of femoral component design. Orthopedics 3, 213-218. Matsuda S, Matsuda H, Miyagi T, Sasaki K, Iwamoto Y , Miura H. (1998) Femoral condyle geometry in the normal and varus knee. Clinical Orthopaedics and Related Research 349, 183-188. Matsuda S, Miura H , Nagamine R, Urabe K, Hirata G, Iwamoto Y . (2001) Effect of femoral and tibial component position on • patellar tracking following total knee arthroplasty: 10-year follow-up of Miller-Galante I knees. American Journal of Knee Surgery 14, 152-156. Matsuda S, Williams V G , White SE, Whiteside L A . (1995b) A comparison of pressure sensitive film and digital electronic sensors to measure contact area and contact stress. Transactions of the Orthopaedic Research Society 19(4), 920. Miller M C , Berger RA, Petrella A J , Karmas A , Rubash HE. (2001a) Optimizing femoral component rotation in total knee arthroplasty. Clinical Orthopaedics and Related Research 392, 38-45. Miller M C , Zhang A X , Petrella A J , Berger RA, Rubash HE. (2001b) The effect of component placement on knee kinetics after arthroplasty with an unconstrained prosthesis. Journal of Orthopaedic Research 19 (4), 614-620. Miller R K , Goodfellow JW, Murray DW, O'Connor JJ. (1998) In vitro measurement of patellofemoral force after three types of knee replacement. Journal of Bone and Joint Surgery 80B(5), 900-906. Moreland JR, Hungerford DS, Insall JN, Scott RD, Whiteside L A . (1988) Symposium: Total knee instrumentation. Contemporary Orthopaedics 17, 93-97. 186 Morrison JB. (1968) Bioengineering analysis of force actions transmitted by the knee joint. Biomedical Engineering 3, 164. Mortier J, Zichner L. (20.04) Computer-assisted pressure measurement in the patellofemoral joint with electronic pressure sensors. In: Navigation and Robotics in Total Joint and Spine Surgery, eds. Stiehl JB, Konermann WH. Ch 29, 204-208. Springer-Verlag, Berlin. Nagamine R, Whiteside L A , White SE, McCarthy DS. (1994) Patellar tracking after total knee arthroplasty: the effect of tibial tray malrotation and articular surface configuration. Clinical Orthopaedics and Related Research 304, 263-271. Nelissen R G H H , Weidenheim L, Mikhail W E M . (1995) The influence of the position of the patellar component on tracking in total knee arthroplasty. International Orthopaedics 19, 224-228. NIST/SEMATECH. (2006) NIST/SEMATECH e-Handbook of Statistical Methods, www.itl.nist.gov/div898/handbook/. Accessed April 8, 2006. Nizard RS, Biau D, Porcher R, Ravaud P, Bizot P, Hannouche D, Sedel L. (2005) A meta-analysis of patellar replacement in total knee arthroplasty. Clinical Orthopaedics and Related Research 432, 196-203. Northern Digital Inc. (NDI). (1994) Optotrak Digitize Guide, Manual Version 1.0. NDI, Waterloo, ON. Northern Digital Inc. (NDI). (2006) Optotrak Technical Specifications. Accessed Jan. 13, 2006 from www.ndigital.com/optotrak-techspecs.php. Oishi CS, Kaufman KR, Kenton R, Irby SE, Colwell CW. (1996) Effects of patellar thickness on compression and shear forces in total knee arthroplasty. Clinical Orthopaedics and Related Research 331, 283-290 Olcott CW, Scott RD. (1999) Femoral component rotation during total knee arthroplasty. Clinical Orthopaedics and Related Research 367, 39-42. Ostermeier S, Nowakowski A , Stukenborg-Colsman C. (2003) Dynamic in vitro measurement of pressure and movement with the LCS prosthetic system. Orthopade 32, 292-295. Otto JK, Brown TD, Callaghan JJ. (1999) Static and dynamic response of a multiplexed-array piezoresistive contact sensor. Experimental Mechanics 39(4), 317-323. Paganelli JV, Skinner HB, Mote CD Jr. (1988) Prediction of fatigue failure of a total knee replacement tibial plateau using finite element analysis. Orthopaedics 11, 1161-1168. 187 Pagnano M W , Trousdale RT. (2000) Asymmetric patella resurfacing in total knee arthroplasty. The American Journal of Knee Surgery 13(4), 228-233. Parvizi J, Rapuri VR, Saleh KJ , Kuskowski M A , Sharkey PF, Mont M A . (2005) Failure to resurface the patella during total knee arthroplasty may result in more knee pain and secondary surgery. Clinical Orthopaedics and Related Research 438, 191-196. Patil S, Colwell CW, Ezzet K A , D'Lima DD. (2005) Can normal knee kinematics be restored with unicompartmental knee replacement? Journal of Bone and Joint Surgery 87A(2), 332-338. Pellengahr C, Maier M , Muller PE, Diirr HR, Schulz C, Zysk S, Trouillier H , Lindhorst E, Jansson V , Refior HJ. (2002) Surgical and anatomic parameters influencing femoropatellar pain in total knee arthroplasty. European Journal of Trauma 28(4), 242-246. Peng CW, Tay B K , Lee BPH. (2003) Prospective trial of resurfaced patella versus non-resurfaced patella in simultaneous bilateral total knee replacement. Singapore Medical Journal 44(7), 347-351. Petersilge WJ, Oishi CS, Kaufman KR, Irby SE, Colwell CW Jr. (1994) The effect of trochlear 'design on patellofemoral shear and compressive forces in total knee arthroplasty. Clinical Orthopaedics and Related Research 309, 124-30. Poilvache PL, Insall JN, Scuderi GR, Font-Rodriguez DE. (1996) Rotational landmarks and sizing of the distal femur in total knee arthroplasty. Clinical Orthopaedics and Related Research 331, 35-46. Ranawat CS. (1986) The patellofemoral'joint in total condylar knee arthroplasty: pros and cons based on five to ten-year follow-up observations. Clinical Orthopaedics and Related Research 205, 93-99. Ranawat CS, Flynn WF Jr, Saddler S, Hansraj K K , Maynard MJ . (1993) Long-term results of the total condylar knee arthroplasty; a 15-year survivorship study. Clinical Orthopaedics and Related Research 286, 94-102. Rand JA. (1994) Current concepts review: the patellofemoral joint in total knee arthroplasty. Journal of Bone and Joint Surgery 76A, 612-620. Rand JA. (2004) Extensor mechanism complications following total knee arthroplasty. Journal of Bone and Joint Surgery 86A, 2062-2072. Rand JA, Ilstrup D M . (1991) Survivorship analysis of total knee arthroplasty: cumulative rates of survival of 9200 total knee arthroplasties. Journal of Bone and Joint Surgery 73A, 397-409. 188 Reider B, Marshall JL, Koslin D, Ring B, Girgis FG. (1981) The anterior aspect of the knee joint. Journal of Bone and Joint Surgery 63A, 351-356. Reuben JD, McDonald CL, Woodard PL, Hennington LJ. (1991) Effect of patellar thickness on patella strain following total knee arthroplasty. Journal of Arthroplasty 6, 251-258. Rhoads DD, Noble PC, Reuben JD, Mahoney O M , Tullos HS. (1990) The effect of femoral component position on patellar tracking after total knee arthroplasty. Clinical Orthopaedics and Related Research 260, 43-51. Ritter M A , Herbst SA, Keating E M , Faris P M , Meding JB. (1994) Long-term survival analysis of a posterior cruciate-retaining total condylar total knee arthroplasty. Clinical Orthopaedics and Related Research 309, 136-45. Sawatari T, Tsumura H , Iesaka K, Furushiro Y , Torisu T. (2005) Three-dimensional finite element analysis of unicompartmental knee arthroplasty - the influence of tibial component inclination. Journal of Orthopaedic Research 23(3), 549-554. Schuldt K , Ekholm J, Nemeth B, Arborelius UP, Harms-Ringdahl K. (1983) Knee load and muscle activity during exercise in rising. Scandinavian Journal of Rehabilitation Medicine 9(Suppl), 174. Sharkey PF, Hozack WJ, Rothman RH, Shastri S, Jacoby SM. (2002) Why are total knee arthroplasties failing today? Clinical Orthopaedics and Related Research 404, 7-13. Shih H-N, Shih L - Y , Wong, Y - C , Hsu RW-W. (2004) Long-term changes of the nonresurfaced patella after total knee arthroplasty. Journal of Bone and Joint Surgery 86A, 935-939. Shoemaker SC, Adams DA, Daniel D M , Woo SL. (1993) Quadriceps/anterior cruciate graft interaction. Clinical Orthopaedics and Related Research 294, 379-390. Singerman R, Pagan HD, Peyser A B , Goldberg V M . (1997) Effect of femoral component rotation and patellar design on patellar forces. Clinical Orthopaedics and Related Research 334, 345-353. Star MJ , Kaufman KR, Irby SE, Colwell CW. (1996) The effects of patellar thickness on patellofemoral forces after resurfacing. Clinical Orthopaedics and Related Research 322, 279-285. Stiehl JB, Cherveny P M . (1996) Femoral rotational alignment using the tibial shaft axis in total knee arthroplasty. Clinical Orthopaedics and Related Research 331, 47-55. Stukenborg-Colsman C, Ostermeier S, Burmester O, Wirth CJ. (2003) Dynamic in vitro measurement of retropatellar pressure after knee arthroplasty. Orthopade 32(4), 319-322. 189 Stukenborg-Colsman C, Ostermeier S, Hurschler C, Wirth CJ. (2002) Tibiofemoral contact stress after total knee arthroplasty: Comparison of fixed and mobile-bearing inlay designs. Acta Orthopaedica Scandinavica 73(6), 638-646. Szivek JA, Anderson PL, Benjamin JB. (1996) Average and peak contact stress distribution evaluation of total knee arthroplasties. Journal of Arthroplasty 11(8), 952-963. Tekscan. (2001) Tekscan I-Scan Pressure Measurement System User's Manual, Version 5.0. Tekscan Inc., South Boston, M A . Tekscan. (date unknown, obtained 2004) Tekscan Medical Sensors - Standard Pressures. Tekscan Inc., South Boston, M A . van Kampen A , Huiskes R. (1990) The three-dimensional tracking pattern of the human patella. Journal of Orthopaedic Research 8, 372-382. Veldpaus FE, Woltring HJ, Dortmans LJ . (1988) A least-squares algorithm for the equiform transformation from spatial marker co-ordinates. Journal of Biomechanics 21(1), 45-54. Vil la T, Migliavacca F, Gastaldi D, Colombo M , Pietrabissa R. (2004) Contact stresses and fatigue life in a knee prosthesis: comparison between in vitro measurements and computational simulations. Journal of Biomechanics 37(1), 45-53. Wallace A L , Harris M L , Walsh WR, Bruce WJ. (1998) Intraoperative assessment of tibiofemoral contact stresses in total knee arthroplasty. Journal of Arthroplasty 13, 923-927. Waters TS, Bentley G. (2003) Patellar resurfacing in total knee arthroplasty: a prospective, randomized study. The Journal of Bone and Joint Surgery 85 A , 212-217. Weale A E , Feikes J, Prothero D, O'Connor JJ, Murray D, Goodfellow J. (2002) In vitro evaluation of the resistance to dislocation of a meniscal-bearing total knee prosthesis between 30 degrees and 90 degrees of knee flexion. Journal of Arthroplasty 17(4), 475-483. Whiteside L A , Arima J. (1995) The anteroposterior axis for femoral rotational alignment in valgus total knee arthroplasty. Clinical Orthopaedics and Related Research 321,168-172. Whiteside L A , Nakamura T. (2003) Effect of femoral component design on unresurfaced patellas in knee arthroplasty. Clinical Orthopaedics and Related Research 410,189-198. 190 Wilson DC, Niosi CA, Zhu QA, Oxland TR, Wilson DR. (2006) Accuracy and repeatability of a new method for measuring facet loads in the lumbar spine. Journal of Biomechanics 39(2), 348-53. Wilson DR, Apreleva M V , Eichler MJ , Harrold FR. (2003) Accuracy and repeatability of a pressure measurement system in the patellofemoral joint. Journal of Biomechanics 36, 1909-1915. Winter DA. (2005) Biomechanics and Motor Control of Human Movements. 3 r d ed. John Wiley and Sons, Inc., Hoboken, NJ. Wirz D, Becker R, Feng L i S, Friedrich NF, Muller W. (2002) Validation of the Tekscan system for static and dynamic pressure measurements in the human femorotibial joint. Biomedizinische Technik 47, 195-201. Woo SL-Y, Orlando CA, Camp JF, Akeson WH. (1986) Effects of postmortem storage by freezing on ligament tensile behavior. Journal of Biomechanics 19(5), 399-404. Wood DJ, Smith AJ , Collopy D, White B, Brankov B, Bulsara M K . (2002) Patellar resurfacing in total knee arthroplasty: a prospective, randomized trial. Journal of Bone and Joint Surgery 84A(2), 187-193. Wu JZ, Herzog W, Epstein MJ. (1998) Effects of inserting a pressensor film into articular joints on the actual contact mechanics. Journal of Biomechanical Engineering 120(5), 655-9. Yoshii I, Whiteside L A , Anouchi YS. (1992) The effect of patellar button placement and femoral component design on patellar tracking in total knee arthroplasty. Clinical Orthopaedics and Related Research 275, 211-219. Zavatsky A B . (1997) A kinematic-freedom analysis of a flexed-knee-stance testing rig. Journal of Biomechanics 30(3), 277-280. Zihlmann MS, Stacoff A , Romero J, Kramers-de Quervain I, Stiissi E. (2005) Biomehcanical background and clinical observations of rotational malalignment in T K A : Literature review and consequences. Clinical Biomechanics 20, 661-668. Zimmer. (date unknown) NexGen System Design Rationale (product literature). Obtained 2004. Zong-Ping L , Berglund LJ , An K - N . (1998) Validation of F-Scan pressure sensor system: a technical note. Journal of Rehabilitation Research and Development 35(2), 186-191. 191 Appendix 1: Adjustable Femoral Component Feasibility Tests During the initial design stage, I considered using plates similar to those used by Miller (2001b); we would attach a baseplate to the femoral bone and align the plate with holes and screws or pegs and slots on a second plate attached to the femoral component. To assess the feasibility of this design, it was necessary to determine whether the amount of bone remaining following surgery would be adequate for the attachment of a baseplate. I drew sketches of the femoral bone cuts which would be necessary to obtain ±5° of rotation of the femoral component. This included rotation of both the anterior and posterior flanges of the component as well as the rotation of the central box structure. I then performed surgeries on Sawbones (Pacific Research Laboratories, Inc., Vashon, Washington) using the Zimmer surgical tools for T K A , and removed additional bone to accommodate 5° of internal and external component rotation. The sketches and surgeries helped determine that the remaining bone stock would not be adequate to attach baseplates such as those used by Miller (2001b). Figure A 1.1 shows the actual femoral bone remaining following the extra resections required to enable ±5° of rotation of the component. Figure A l . l : Bone remaining following extra resections to allow ±5° of rotation of the femoral component Design and Construction During the first design iteration, I designed and built a prototype of the modified component. At this stage, we were under the assumption that we had to return the components to Zimmer and we could not permanently damage the femoral component by drilling holes into it. 192 To connect 25 cm of threaded steel rod to the femoral implant, I machined a slotted aluminum block to mate with the box structure on the proximal surface of the implant. I then drilled and tapped a hole in the aluminum block and threaded the rod into the hole. The threaded connection between the block and rod was strengthened using Loctite 430 Super Bonder- Metal Bonding (Henkel Technologies, Munchen, Germany) and the block was glued to the proximal surface of the implant using the same adhesive. The threaded rod was inserted through an 18 cm long aluminum tube (ID = 1.0 cm, OD = 1.25 cm), which would be cemented into the intramedullary canal of the femur (see Figure 2.7, Chapter 2). It was locked in place by threading a lockwasher and a nut onto the proximal end of the femoral rod and tightening them against the femoral tube. To provide the required 6° valgus angle at each rotation of the femoral component, I machined the proximal surface of the aluminum block at a 6° angle. When we inserted the prototype femoral assembly into a pilot knee, we discovered that the Loctite 430 metal bonder dissolved in the greasy joint. Although the geometry of the design seemed appropriate, we did not believe that any adhesive would provide a strong enough connection to withstand the loads applied to the assembly during testing. We also worried about bending of the femoral rod, because the experience of Michael Paice, another researcher at UBC, showed that the steel rods cemented in the femoral shafts gradually bent during testing. This was acceptable when the rods were permanently affixed in the bone and new rods were used with each specimen; however our components were designed to be reused with each of 8 specimens and could not be damaged. Final Design Modifications Although the second version of this design was very similar to the original prototype, I made some modifications to improve the robustness of the assembly. We were informed that we could in fact permanently damage the components, thus all connections between parts were made permanent. To increase the stiffness of the assembly and prevent bending of the femoral rod and tube, I decided to use a stainless steel femoral tube and to machine the connection block and the threaded rod out of steel. See section 2.5.1.3 for a description of the final design. 193 Appendix 2: Adhesives and Lubricant Selection Adhesives In their in vivo study, Anderson (2003) inserted a sensor loose into the natural tibiofemoral joint. Matsuda (1997) inserted sensors loose into the natural and resurfaced patellofemoral joints. We could not insert the sensor loose into the patellofemoral joint because we needed to measure loads relative to the patellar axes. We wanted to be able to classify whether loading occurred medially or laterally on the patella; this would not be possible if the sensor was not attached securely to the patellar implant. Also, the sensor would probably slide around in the joint and be subject to additional shear forces, which most likely contribute to the degradation of sensor output. Some researchers have sutured a Tekscan sensor to the soft tissues surrounding the natural patella such as the patellar tendon (Gill 2003, Gi l l 2004, L i 2004). Suturing the sensor to soft tissues seemed impractical for two reasons: the sensor would probably still shift around on the patella and cause small wrinkles, and suture holes in the sensor would most likely tear when the sensor was pulled during loaded flexion or disrupt the electrical connections, making the sensor useless. Harris (1999) used double-sided adhesive tabs and Wallace (1998) used a "thin adhesive plastic membrane" (tape) to attach the sensor to the tibiofemoral joint of a T K A . We could not use adhesive tape because the bond would not be strong enough to withstand the large shear loads acting on the patella during loading. Other researchers have cemented (Wilson 2003) or glued (Elias 2004c) the sensor to the natural patella to compare medial and lateral patellar loads. Initially, we attempted to glue the patellar button to the pressure sensor using Loctite Prism 401 Instant Adhesive (Henkel Technologies, Miinchen, Germany). The main advantage of this adhesive is that it is sold with a solvent, X -N M S Clean Up Solvent (Henkel Technologies, Miinchen, Germany), which facilitated removal of the sensor from the patella. While Tekscan sensors are disposable, we needed to reuse the same patellar implant for each specimen. 194 Adhesive Selection The Loctite 401 adhesive seemed to hold well until the sensor was placed in the greasy environment of the knee. The adhesive then began to gradually dissolve, and the sensor slowly peeled away from the patella, beginning at the edges. Following several attempts using this adhesive, including in combination with a primer (Loctite 770 Primer, Henkel Technologies, Miinchen, Germany) to strengthen bonding, we resolved to conduct qualitative tests on several adhesives to find an optimal bond between the Mylar sensor and the patellar implant. We tested the following five adhesives: • Zapagap cyanoacrylate (Super Glue Corporation, Rancho Cucamonga, CA) • Loctite Prism 401 Instant Adhesive (Henkel Technologies, Miinchen, Germany) • 3M Scotch-Grip Plastic Adhesive 1099 (3M, St. Paul, MN) • 3M Super 77 Spray Adhesive (3M, St. Paul, MN) • 3M Hi-Strength 90 Spray Adhesive (3M, St. Paul, MN) We found that the Hi-Strength 90 glue formed the strongest bond, and used it to attach the sensors to the patellar button for specimen 2. We did not experience any other problems with the sensors peeling off or with glue dissolving. The Hi-Strength 90 adhesive seemed to be much more resistant to oil and grease than the Loctite 401 adhesive. Lubrication In their experiments with the tibio-talar joint, Dr. Rudert and colleagues performed 30-40 load applications per joint. Initially, they would destroy a sensor after 3-4 tests; however, they found that by using better lubrication in the joint, they could prolong the life of the sensor and perform the whole series of tests with just 1-2 sensors. We already minimized sliding by gluing the sensor firmly to the patella, but the issue of lubrication was one with which we struggled for the first 2 specimens. Most studies using Tekscan sensors in human joints do not specify i f a lubricant was used. Wilson (2003) used K Y Jelly as a lubricant between the natural patella and the T K A femoral component. Dr. Rudert experimented with several fluids such as synovial fluid and Vaseline, and his final recommendation was to use Vaseline and recoat the sensor every 7-8 tests as the lubricant is wiped off or pushed out of the way. 195 We began our tests using ViperLube High Performance Synthetic Grease (Henkel Technologies, Miinchen, Germany) because we required a thicker, more lasting lubricant than K Y Jelly. This lubricant was very greasy and seemed to stay in the joint for several tests. However, we still noticed considerable degradation of sensor output, and decided to try a Teflon-based dry film lubricant (DryTef, Walter Tool Company Inc., Norwell, MA) . To enhance the sliding between the Tekscan sensor and the femur, we placed a Teflon-sprayed square of wax paper between the sensor and the femur. The combination of the Teflon lubricant and the wax paper significantly reduced degradation, and was used successfully for specimens 3-8. 196 Appendix 3: Order of Tests Performed and Additional Tests Table A3.1 shows the order in which the surgical variables were tested for each specimen. Table A3.2 shows the additional tests which were performed on each specimen. Table A3.1: Order of tests performed (numbers divided by a slash indicate horizontal/vertical) SI S2 S3 S4 S5 S6 S7 S8 Femoral Rotation Internal 4b 4b 4a 4b lb / a 4a 4a 4b External 4a 4a 4b 4a la/b 4b 4b 4a Tibial Rotation Internal 3b/a lb/3b 3a 3a 2a lb 3a 3a External 3a lb la/3a 3b 3b 2b la 3b 3b Patella Bone Cut Angles 7.5° med 2a 2c lc 2c 4c 2b lc 2a 7.5° lat 2b 2b l a 2b 4b 2c l a 2c 15° lat 2c 2a lb 2a 4a 2a lb 2b Patella Position / Thickness 2.5 mm la 3a/ la 2a la 3a 3b 2a lb 5.0 mm lb 3b / lb 2c lc 3c 3a 2b lc Thickness lc 3c / l c 2b lb 3b 3c 2c la Table A3.2: Additional tests performed on each specimen. H and V indicate horizontal and vertical rigs. Data which could not be analyzed are described in Appendix 5. SI S2 S3 S4 S5 S6 S7 S8 H V H V H V H V H V H V H V H V Fast X X \ X X X X \ Static X X X X X X X X X X X No sensor X X X X X X X No clamps X X X X X X X Lat. release X X \ \ X \ 197 Appendix 4: Problems Experienced Previous TKA Specimen 4 had a previous T K A prior to being used in our study. When we removed the existing components, a considerable amount of bone that was affixed to the bone cement was lost. Dr. Tonetti rebuilt the femur and tibia with bone cement; however, the 'neutral position' of the femur was hard to define without the presence of intact posterior condyles. Re-cementing Femoral Tube During the adjustment of the femoral component, it was very common to notice that the femoral tube had loosened in the bone. For specimens 1, 2, 5, 7 and 8, we were forced to re-cement the femoral tubes because the bond between the metal and the cement had broken. We hypothesize that freezing the specimens and using cement in such a greasy environment undermined a solid bond between the cement and the stainless steel tube. Lack of Tissues Several specimens were missing some soft tissues around the knee, which limited our ability to constrain the motion of the patella and to simulate in vivo loading conditions. The quadriceps tendon on specimen 1 was partially severed and was repaired using sutures. We observed that the medial tissue on specimen 1 had loosened from the bone during testing in the vertical rig; this made it difficult for us to close the medial incisions adequately, and the patella most likely tracked abnormally laterally as a result. Possibly due to the previous T K A it experienced, specimen 4 had very little lateral soft tissues remaining. The medial tissues had already been removed; we improvised the effect of medial tissues by suturing a 6 cm x 1 cm band of fabric to the remaining medial patellar tissue (Figure A4.1). This may have affected patellar tracking as well as patellofemoral loading. The posterior capsule of specimen 8 became weak during testing in the horizontal rig and tore during the first tests in the vertical rig. This may have contributed to the slackness of the joint which caused the knee to dislocate repetitively. 198 Figure A4.1: Band of fabric sutured to medial tissues to help close medial incision Dislocating Knees During testing, we experienced problems with knee dislocations with 5 of the 8 specimens (2, 4, 6, 7, and 8). The femur dislocated by sliding posteriorly off of the tibia. Specimen 2 was the only knee in which this phenomenon occurred in the horizontal rig; all other dislocations were observed during loading in the vertical rig. Specimen 4 was the first knee which we attempted to repair in order to continue testing. It seemed to dislocate under 'extreme' testing conditions, such as with the addition of additional patellar thickness or with the 15° lateral wedge. We discontinued testing for these surgical variables to avoid further dislocations. To salvage the knee, we sutured 2 ties from a surgical gown to the femur and tibia to act as additional collateral ligaments. We sutured one end of the lateral band into the patellar ligament and screwed the other end into the lateral epicondyle of the femur. The medial band was screwed into the tibial bone and the medial epicondyle of the femur. Because the joint was very loose, we replaced the 10 mm tibial spacer with a 12 mm spacer to endeavor to fill the joint space and create a tighter joint. When the sutures ripped out of the lateral fabric band in the patellar ligament, we screwed the band into the bone. The knee continued to dislocate and the fabric strip tore, so we replaced the fabric strip with a figure 8 of copper wire (Figure A4.2). We were forced to tighten the figure 8 continuously, which prevented dislocation but probably affected joint motion. 199 Figure A4.2: Copper wire attached to femur and tibia with screws and used to prevent dislocation of femur. When we experienced similar dislocation with specimen 6, we used copper wire and screws on the lateral side to correct the problem. When the wire was too tight, it constrained the motion of the knee to such an extent that the knee could not attain full flexion. Nevertheless, the knee continued to dislocate. For that reason and others listed below, we discontinued testing on specimen 6. When specimen 7 dislocated in the vertical rig, we added wire and screws on the lateral side to successfully prevent dislocation. This repair did not function as effectively when we applied it to the dislocating specimen 8. Most likely due to weak bone, the screws in the tibia and lateral femoral condyle continuously became loose or completely pulled out of the bone. Although we added several loops of wire to the medial side of the knee as well, those loops did not seem to be carrying any load. Preventing dislocation became a process of trial and error in which we adjusted the position of the screws in the bone and the length of wire connecting them. We believe that the aforementioned knee specimens dislocated because the knee components were too small for the joint space. We noted that specimens 6, 7 and 8 seemed very loose and that it was thus very easy to adjust the rotation of the femoral component. For example, the bones of specimen 8 were larger than the other femurs tested, yet we used the same set of components throughout all tests; the femur was a size F (instead of E), the tibia was a size 5 200 (instead of 4) and the patella was a size 38 (instead of 35). By under-sizing the components, we did not properly constrain the motion of the femur on the tibia, nor did we create an appropriate amount of tension in the surrounding soft tissues. The stability and range of motion appeared to be appropriate at the time of specimen preparation; however, it is likely that the extension gap was more accurately recreated than the flexion gap. In the earlier specimens, Dr. Tonetti had difficulty achieving adequate femoral rotation. For later specimens, he decided to place the femoral component slightly more distally; as a result, it was much easier to achieve the 10° range of femoral rotation. Unfortunately, this probably contributed to the uneven flexion gap. Loose Tibial Screw During the loading of three specimens (6, 7, and 8), the screw attaching the tibial spacer to the tibia became loose in the bone. As a result, during flexion past approximately 45°, a popping noise could be heard as the anterior edge of the tibial spacer lifted out of the tibial tray (which remained firmly attached to the bone). The loosening of the screw in the bone was most likely due to poor trabecular bone quality (low strength) and large anteroposterior loads during flexion in the vertical rig. We attempted to solve this problem by using a longer tibial screw (specimens 6 and 8 in the vertical rig, specimen 7 in the horizontal rig). When this failed to hold, we poured bone cement into the threaded hole in the bone and formed threads in the cement by rotating the screw in the hole as the cement dried (specimens 6 and 7). While this technique worked well for specimen 7, it was not successful for specimen 6. We also cemented the tibial baseplate of specimen 7 in place using bone cement. Dislodged Patella Baseplate The patellar baseplate implanted in specimen 6 was dislodged during testing in the vertical rig. This occurred in tandem with the loosening of the tibial screw, which suggests that the bone quality of the specimen was very poor. When the sewing pins were torn from the bone, they left gaping holes where they had been. Since we could not simply insert new pins into the remaining craters, we attempted to cement the baseplate to the patellar bone. Unfortunately, the baseplate tore out again and also ripped out some bone as it was disconnected. As noted above, we discontinued testing on specimen 6; this was due to the combination of the dislocating femur, the loose tibial screw and the dislodged patellar baseplate. 201 Torn Quadriceps Tendon When specimen 5 was mounted in the vertical rig and preconditioned, we initially neglected to add the counterweight to the sliding tray. This caused the motor to work harder and also loaded the quadriceps tendon and the knee with an additional 53 N (133 N compared to 80 N). This extra loading may have damaged the quadriceps tendon. Following this error, the tendon tore across its width, below the edge of the clamp. The tear may also have been due to over-tightening of the clamp. We moved the clamp further down the tendon and did not fasten the plates together as tightly. In the vertical rig, the quadriceps clamp on specimen 7 began to slide up the tendon (proximally) while also causing the tendon to tear. Although the screws on the clamp were tightened, the clamp continued to slide. We moved the clamp distally down the tendon to a location where the tendon was thicker and of better integrity. Hardware Interfering with Motion of Quadriceps Clamp We experienced problems with the motion of the quadriceps clamp during the testing of specimens 5 and 7. In the horizontal rig, the quadriceps clamp on specimen 5 collided with the wedges we used to hold the femur in the correct rotation. This occurred close to maximum flexion, as the clamp slid down the femur and approached the femoral component. We addressed this problem by moving the clamp further up the quadriceps tendon. This adjustment most likely reduced the force in the connecting spring, and thus reduced the loads on the patella. When the quadriceps tendon on specimen 5 ripped in the vertical rig, we were forced to move the clamp down the tendon (closer to the femoral component). Because it was inevitable that the clamp would then impact the femoral component, we built a cement ramp to allow the clamp to slide up and along the anterior face of the femoral component. This transitional ramp caused the line of action of the quadriceps to move anteriorly by approximately 1 cm. We experienced a similar problem testing specimen 7 in the vertical rig. To avoid hitting the component with the clamp, we reduced the maximum flexion angle of the knee (by adjusting the locations of the wooden stopping blocks on the vertical support, see section 2.6.2.2.1). 202 Dislocating Patella During testing in the vertical rig, the patella of specimen 1 dislocated when the 7.5° lateral wedge was added. Because patellar dislocations are abusive to the Tekscan sensor, we did not continue tests with the 7.5° lateral wedge with this specimen, nor did we insert the 15° lateral wedge into the joint. The lateral wedge was likely causing the patella to ride on its lateral side, and as a result it was not well-seated within the femoral groove. Ripped Tekscan Sensors In the vertical rig, the Tekscan sensors in specimens 1, 2 and 3 tore or were punctured. To prevent the entrance of fluids into the sensor and to maintain sensor sensitivity at the post-puncture level, we covered the puncture with a layer of packing tape, and reapplied tape as it was torn or pushed away during testing. The sensor in specimen 2 continued to provide good readings, but the sensors in specimens 1 and 3 were irreparably damaged. Fluids leaked into the sensor in specimen 3 and shorted out rows and columns in the area of the puncture. To prevent sensor punctures in later specimens, the sensors were covered in tape, as described in section 2.7.2.2. Tekscan Sensor Degradation and Loss of Rows and Columns of Output We witnessed the loss of columns and rows of Tekscan sensor output for specimens 1 -3. We hypothesized that by minimizing the abuse to the sensor tab, we could reduce the loss of column and row output. We designed and built new apparatuses to hold the Tekscan handle and minimize the twisting and distortion of the sensor tab (see section 2.7.2.3). They also helped to support the Tekscan sensor while we made changes to component placement. We also became more careful in handling the Tekscan sensor as our testing progressed. In particular, we did not insert the sensor into the joint during preconditioning cycles or any additional test cycles. Originally, we had planned to perform 3 flexion cycles for each surgical variable; as a result of the sensor damage, we reduced the repetitions to 2 cycles. After we implemented the new methods of supporting the Tekscan handle and the protective taping measures (see sections 2.7.2.2 and 2.7.2.3), we ceased to lose rows and columns of data, and the degree of degradation was reduced. 203 Appendix 5: Excluded or Incomplete Data Kinematic Data Table A5.1 shows the surgical variables for which the kinematic data could not be analyzed because the data were not available. The reasons for the missing data are listed below the table. Table A5.1: Surgical variables for which kinematic data was missing or incomplete. SI Horizontal SI Vertical S2 Horizontal S4 Vertical S5 Horizontal External tibial rotation 1 External 2 and internal 3 femoral rotation Baseline 5 4 15° lateral patellar resection angle 5 Baseline 2 b Both lateral patellar resection angles 3 External tibial rotation 5 Patellar thickness 5 Reasons for Missing Data: 1: Did not perform test because believed it would damage Tekscan sensor 2: Discontinued testing due to abnormal axial twisting of the joint (tibia and femur rotating in opposite directions) 3: Patellar subluxation or risk of subluxation 4: Motion data missing for at least one marker array 5: Dislocation of femur 6: Outliers (possibly because we moved the quadriceps clamp proximally up the patellar tendon so that it would not hit the wedges keeping the femoral component from rotating (see section 2.5.1.3)) Force Data Kinetic data from specimen 1 were not analyzed because several rows and columns of data were missing. No force data were available for specimen 6 in the vertical rig because of the many problems encountered (see Appendix 4). For specimen 3 in the vertical rig, the sensor punctured during the testing of the second surgical variable. Table A5.2 shows the surgical variables which were not analyzed because the sensors were too degraded (more than a 50% reduction in force from the first baselines). 204 Table A5.2: Surgical variables which were not analyzed because of sensor degradation. S2 Vertical S3 Horizontal S4 Vertical S5 Vertical Main Surgical Variables - Baselines 4 & 5 - External and internal femoral rotation - External and internal tibial rotation - Baseline 5 - Baseline 5 - Internal femoral rotation - Baselines 4 & 5 - A l l patellar resection angles - Both patellar medializations - Patellar thickness Additional Variables - N o clamps - Lateral release - N o clamps - Fast - N o clamps - Lateral release - N o clamps - Lateral release For specimen 2 in the horizontal rig, we discarded the data for external and internal femoral rotation as well as the fast additional trial because the scaled peak forces were unreasonably large. These variables were the last ones tested and the sensors were fairly degraded; the large scaling factors applied to normalize the data to the first set of baselines (see section 2.9.2) may have falsely enlarged the peak forces. We also discarded the data for external tibial rotation because the femur dislocated during testing. For specimen 4 in the vertical rig, we did not analyze the data for the 15° lateral patellar resection angle or the patellar thickness because the femur dislocated during those trials. We began taking a set of static data points (see section 2.8.2.2) during the testing of specimen 6. We were only able to obtain complete sets of static force data (for more than one flexion angle) for specimens 6 and 7 in the horizontal rig. 205 Appendix 6: Patellar Rotation Results We observed very few statistically significant differences in patellar rotation (spin) due to changes in component placement. Using a two-way A N O V A with repeated measures, we determined that statistically significant changes in spin occurred only in late flexion, at flexion angles of 75° and greater (p<0.05). Figures A6.1-A6.4 show the effects (mean change ± standard deviation) of changes in component placement on relative spin (compared to neutral baselines) in the horizontal and vertical testing rigs. The data were averaged across specimens at each flexion angle. For all figures included in this appendix, an asterix (*) indicates statistically significant changes in spin. Femoral Component Rotation Externally rotating the femoral component resulted in medial patellar spin, compared to baseline, which was statistically significant at flexion angles greater than 75° in both testing rigs (Figure A6.1). Internally rotating the femoral component only resulted in a statistically significant change in spin in late flexion in the vertical rig; internal rotation induced lateral spin. Figure A6.1: Effects (mean change ± SD) of ±5°rotation of the femoral component on relative spin in the horizontal and vertical testing rigs. 206 Tibial Component Rotation Externally rotating the tibial only resulted in a statistically significant change in spin in late flexion (angles greater than 75°) in the horizontal rig; it resulted medial patellar spin (Figure A6.2). Internally rotating the tibial component did not have a statistically significant effect on spin. Horizontal Rig Tibial Rotation 30 45 60 75 90 106 Flexion angle (") C o Vertical Rig Tibial Rotation Lateral 30 45 60 75 Flexion angle (°) External Internal Figure A6.2: Effects (mean change ± SD) of ±5°rotation of the tibial component on relative spin in the horizontal and vertical testing rigs. Patellar Resection Angle Altering the patellar resection angle did not have a statistically significant effect on spin at any flexion angle in either testing rig (Figure A6.3). Horizontal Rig Vertical Rig Flexion angle (•) Flexion angle (•) Figure A6.3: Effects (mean change ± SD) of patellar resection angle (7.5° medial, 7.5° lateral, and 15° lateral) on relative spin in the horizontal and vertical testing rigs. 207 Patellar Medialization and Thickness Medializing the patellar implant did not have a statistically significant effect on spin, with one exception: in the horizontal rig at 75° flexion, the 5 mm medialization increased lateral spin by 2° (Figure A5.4). Increasing patellar thickness by 3 mm did not have a statistically significant effect on spin at any flexion angle in either testing rig. Horizontal Rig Patellar Position and Thickness 30 45 60 75 90 Flexion angle (°) a w i m Vertical Rig Patellar Position and Thickness Lateral _ j 1 L _ L . _ i . _ i _ . • ! i i i 1 _ H i 5 m m Med H 2.5 m m Med -• 9 Thickness 45 60 75 Flexion angle (•) Figure A6.4: Effects (mean change ± SD) of mediolateral patellar placement (2.5 mm medial, 5 mm medial) and 3 mm additional patellar thickness on relative spin in the horizontal and vertical testing rigs. 208 Appendix 7: Statistical Data for Paired t-tests Tables A7.1 and A7.2 list the p-values for the paired t-tests comparing changes in tilt and shift, respectively, between neutral baseline measurements and each surgical variable. Table A7.3 shows the phases of flexion (early, mid-, or late) at which significant differences existed between variables. Table A7.1: Summary of p-values for paired t-tests (using a Bonferroni correction factor against a single control) between baseline tilt measurements and each change in tilt due to component placement. Tilt Horizontal Vertical 15° 45° 90° 15° 45° 90° Femoral Rotation External 0.024 0.013 0.066 <0.001 <0.001 0.039 Internal 0.074 0.026 0.034 0.0027 <0.001 <0.001 Tibial Rotation External <0.001 0.004 0.004 0.139 0.987 0.622 Internal 0.208 0.485 0.405 0.733 0.523 0.165 Patellar Resection Angle 15° lateral <0.001 <0.001 <0.001 0.001 <0.001 <0.001 7.5° lateral 0.001 <0.001 0.001 <0.001 <0.001 <0.001 7.5° medial <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 Patellar Medialization 2.5 mm 0.026 0.009 0.004 0.074 0.002 0.047 5 mm 0.050 0.008 0.001 0.010 0.007 0.031 Patellar Thickness 3 mm 0.219 0.220 0.034 0.059 0.165 0.812 Table A7.2: Summary of p-values for paired t-tests (using a Bonferroni correction factor against a single control) between baseline shift measurements and each change in shift due to component placement. Shift Horizontal Vertical 15° 45° 90° 15° 45° 90° Femoral Rotation External 0.099 0.212 0.531 <0.001 <0.001 0.026 Internal 0.297 0.228 0.729 0.051 <0.001 0.217 Tibial Rotation External 0.642 0.512 0.736 0.078 0.357 0.397 Internal 0.272 0.301 0.194 0.650 0.620 0.125 Patellar Resection Angle 15° lateral 0.117 0.073 0.049 0.556 <0.001 <0.001 7.5° lateral 0.034 0.012 0.014 0.049 <0.001 <0.001 7.5° medial 0.268 0.012 0.022 0.420 <0.001 <0.001 Patellar Medialization 2.5 mm 0.014 0.001 0.005 0.029 <0.001 0.002 5 mm 0.002 <0.001 <0.001 <0.001 <0.001 <0.001 Patellar Thickness 3 mm 0.249 0.589 0.705 0.343 0.286 0.227 209 Table A7.3: Summary of angles at which statistically significant differences (p < 0.05) were found in changes in tilt and shift between surgical variables. Tilt Shift Variab les Horizontal Vertical Horizontal Vertical Femoral Rotation External -Internal All phases All phases Early and mid-flexion Early and mid-flexion Tibial Rotation External -Internal Mid-flexion - Early and late flexion -Patellar Resection Angle Between all angles All phases All phases All phases* Mid- and late flexion Patellar Medialization 2.5 mm -5mm - All phases All phases All phases *Except between lateral wedges in early flexion ) 210 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0080712/manifest

Comment

Related Items