@prefix vivo: . @prefix edm: . @prefix ns0: . @prefix dcterms: . @prefix dc: . @prefix skos: . vivo:departmentOrSchool "Land and Food Systems, Faculty of"@en ; edm:dataProvider "DSpace"@en ; ns0:degreeCampus "UBCV"@en ; dcterms:creator "Orphanidou, Charitini Ioannou"@en ; dcterms:issued "2009-01-15T18:45:53Z"@en, "1995"@en ; vivo:relatedDegree "Master of Science - MSc"@en ; ns0:degreeGrantor "University of British Columbia"@en ; dcterms:description """The most commonly described psychological abnormality in Anorexia Nervosa (AN) is a distorted perception of body weight and shape. Anorexia nervosa patients typically fear that weight gain is accompanied by preferential fat deposition in the abdomen, hips and thighs; and this fear may contribute to their resistance in gaining weight even when this is of medical necessity. One objective of this study was to investigate body composition and body fat distribution changes that accompany short-term weight gain in AN patients. Another objective was to assess the level of agreement in the measurement of change in percentage body fat in patients with A N pre- and post-weight gain, as determined by two body composition assessment methods: bioelectrical impedance analysis (BIA) and dual energy X-ray absorptiometry (DEXA). Twenty six female subjects, 28+7 years of age (mean + SD), initial BMI17+2 kg/m^, who met the diagnostic criteria for AN completed the study. Subjects were recruited from the inpatient and outpatient Eating Disorders Clinics, St. Paul's Hospital, Vancouver, BC. Body composition and body fat distribution changes were assessed by skinfold (SKF), circumference (CIRC), and DEXA methods. Bioelectrical impedance analysis was used to measure the change in percentage body fat pre- and post-weight gain, and this change was compared to that obtained by DEXA. Skinfold and CIRC measurements were performed at 9 body sites; DEXA was used to quantify body fat mass in the subscapular, waist and thigh regions. Measurements by all methods were performed at baseline, and at the point of maximum weight gain. Results of body composition changes included a highly significant weight gain of 6.7+5.3 kg (p < .001). This weight gain was achieved by significant gains in body fat (p < .001), lean body mass (p < .05), and bone mineral content (p < .01). Total body fat was, however, the component which increased to the greatest extent. Analysis of absolute and relative changes pre- and post-weight gain as assessed by SKF and CIRC indicated a greater fat deposition in the central regions (chest, abdomen, hip and thigh) than in the extremities (arm and calf). However, comparison of body fat mass change (kg) in the subscapular, waist and thigh regions as measured by DEXA indicated no significant differences among these 3 central regions (subscapular: 1.7+1.2, waist: 1.8+1.3, thigh: 1.5+1.0; p = .10). Comparison of measurement of change in percentage body fat upon weight gain between BIA and DEXA indicated poor agreement between the two methods. It appears that single-frequency BIA may not be sensitive enough to reliably quantify changes in body composition in AN patients. Overall, the preliminary findings of this study suggest that although weight gain in A N patients is accompanied by greater fat deposition in the central regions than in the extremities, there is no preferential fat accumulation in any of the central regions. Therefore, the gynoid fat distribution pattern in these patients is preserved despite renourishment and subsequent weight gain. This also implies that significant weight gain does not predispose these patients to the health risks associated with central body fat distribution."""@en ; edm:aggregatedCHO "https://circle.library.ubc.ca/rest/handle/2429/3672?expand=metadata"@en ; dcterms:extent "7871446 bytes"@en ; dc:format "application/pdf"@en ; skos:note "BODY COMPOSITION AND BODY FAT DISTRIBUTION CHANGES AFTER SHORT-TERM WEIGHT GAIN IN ANOREXIA NERVOSA PATIENTS by CHARITINIIOANNOU ORPHANIDOU B.Sc, The University of Alberta, 1990 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCES in THE FACULTY OF GRADUATE STUDIES (School of Family and Nutritional Sciences) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA April 1995 © Charitini Ioannou Orphanidou, 1995 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department The University of British Columbia Vancouver, Canada Date DE-6 (2/88) A B S T R A C T The most commonly described psychological abnormality in Anorexia Nervosa (AN) is a distorted perception of body weight and shape. Anorexia nervosa patients typically fear that weight gain is accompanied by preferential fat deposition in the abdomen, hips and thighs; and this fear may contribute to their resistance in gaining weight even when this is of medical necessity. One objective of this study was to investigate body composition and body fat distribution changes that accompany short-term weight gain in A N patients. Another objective was to assess the level of agreement in the measurement of change in percentage body fat in patients with A N pre- and post-weight gain, as determined by two body composition assessment methods: bioelectrical impedance analysis (BIA) and dual energy X-ray absorptiometry (DEXA). Twenty six female subjects, 28+7 years of age (mean + SD), initial BMI17+2 kg/m^, who met the diagnostic criteria for A N completed the study. Subjects were recruited from the inpatient and outpatient Eating Disorders Clinics, St. Paul's Hospital, Vancouver, BC. Body composition and body fat distribution changes were assessed by skinfold (SKF), circumference (CIRC), and D E X A methods. Bioelectrical impedance analysis was used to measure the change in percentage body fat pre- and post-weight gain, and this change was compared to that obtained by D E X A . Skinfold and CIRC measurements were performed at 9 body sites; D E X A was used to quantify body fat mass in the subscapular, waist and thigh regions. Measurements by all methods were performed at baseline, and at the point of maximum weight gain. Results of body composition changes included a highly significant weight gain of 6.7+5.3 kg (p < .001). This weight gain was achieved by significant gains in body fat (p < .001), lean body mass (p < .05), and bone mineral content (p < .01). Total body fat was, however, the component which increased to the greatest extent. Analysis of absolute and relative changes pre- and post-weight gain as assessed by SKF and CIRC indicated a greater ii fat deposition in the central regions (chest, abdomen, hip and thigh) than in the extremities (arm and calf). However, comparison of body fat mass change (kg) in the subscapular, waist and thigh regions as measured by D E X A indicated no significant differences among these 3 central regions (subscapular: 1.7+1.2, waist: 1.8+1.3, thigh: 1.5+1.0; p = .10). Comparison of measurement of change in percentage body fat upon weight gain between BIA and D E X A indicated poor agreement between the two methods. It appears that single-frequency BIA may not be sensitive enough to reliably quantify changes in body composition in A N patients. Overall, the preliminary findings of this study suggest that although weight gain in A N patients is accompanied by greater fat deposition in the central regions than in the extremities, there is no preferential fat accumulation in any of the central regions. Therefore, the gynoid fat distribution pattern in these patients is preserved despite renourishment and subsequent weight gain. This also implies that significant weight gain does not predispose these patients to the health risks associated with central body fat distribution. i i i TABLE OF CONTENTS Abstract ii Table of Contents iv List of Tables vi List of Figures vii Acknowledgement viii Dedication ix CHAPTER ONE: INTRODUCTION 1 A) Rationale 1 B) Null Hypotheses Tested 3 C) Study Objectives 3 CHAPTER TWO: LITERATURE REVIEW 4 A) Physiology of Starvation 4 B) Normal Body Composition 6 C) Body Composition Characteristics of Anorexia Nervosa 10 1) Body Fat 10 2) Fluid and Electrolyte Status 11 3) Skeletal System 12 4) Skin Thickness 13 D) Body Composition and Body Fat Distribution Changes after Weight Gain in Anorexia Nervosa 14 E) Measurement Techniques of Body Composition and Body Fat Distribution 21 1) Skinfolds 24 2) Circumferences 27 3) Dual Energy X-ray Absorptiometry 28 4) Bioelectrical Impedance Analysis 33 CHAPTER THREE: METHODOLOGY 36 A) Pilot Study 36 B) Experimental Design 37 C) Subject Recruitment 38 D) Measurements of Body Composition and Body Fat Distribution 40 1) Anthropometry 40 2) Dual Energy X-ray Absorptiometry 42 3) Bioelectrical Impedance Analysis 44 E) Meal Support and Renourishment Protocol 44 F) Assessment of Dietary Intake and Exercise Behavior 46 1) Dietary Intake Assessment 46 2) Exercise Behavior 48 G) Statistical Analysis of the Data 49 iv CHAPTER FOUR: RESULTS 51 A) Study Participants 51 B) Baseline Data 53 C) Dietary Intake and Exercise Behavior 54 D) Body Composition Changes 56 E) Body Fat Distribution Changes 57 1) Skinfolds 58 2) Circumferences 62 3) Circumference Ratios 66 4) Dual Energy X-ray Absorptiometry Analysis 66 F) Estimation of Change in Percentage Body Fat upon Refeeding: Comparison Between Bioelectrical Impedance Analysis and Dual Energy X-ray Absorptiometry 68 CHAPTER FIVE: DISCUSSION 70 A) Major Findings 70 B) Body Composition Changes 71 C) Body Fat Distribution Changes 77 1) Skinfolds 78 2) Circumferences 79 3) Dual Energy X-ray Absorptiometry Analysis 81 D) Estimation of Change in Percentage Body Fat upon Refeeding: Comparison Between Bioelectrical Impedance Analysis and Dual Energy X-Ray Absorptiometry 85 CHAPTER SIX: CONCLUSIONS AND RECOMMENDATIONS 88 A) Conclusions 88 B) Limitations of the Study 89 C) Future Research 90 D) Recommendations for Practice 91 References 92 Appendices 103 Appendix 1: Pilot Study Publication 104 Appendix 2: Ethics Approval Certificates 109 Appendix 3: Subject Criteria 112 Appendix 4: Subject Recruitment Advertisement 114 Appendix 5: Informed Written Consent 116 Appendix 6: Anthropometric Measurement Standards - Skinfolds 120 Appendix 7: Anthropometric Measurement Standards - Circumferences 125 Appendix 8a: D E X A Whole Body Image Computer Print-out 130 Appendix 8b: Pre-determined D E X A Regions 133 Appendix 8c: D E X A Regions Used in Study 135 Appendix 9: BIA Electrode Placement 137 Appendix 10: Food Intake Record Form 139 Appendix 11: Data Analysis of Inpatients 144 v LIST OF TABLES 1. Summary of studies investigating the composition of regained weight in AN patients 20 2. Hospitalization profile and length of study enrollment of A N patients 52 3. Baseline subject characteristics of patients with A N 53 4. Estimated post-weight gain energy intake and purging and exercise behaviours of patients with A N 55 5. Comparison of body composition parameters pre- and post-weight gain in patients with AN. 57 6a. Measurement of subcutaneous fat at 9 sites in A N patients - Pre, post, and absolute change 59 6b. ANCOVA fully repeated measures of absolute subcutaneous body fat change with initial body weight as the covariate 60 7a. Relative subcutaneous fat change at 9 sites in A N patients 61 7b. ANCOVA fully repeated measures of relative subcutaneous body fat change with initial body weight as the covariate 62 8a. Measurement of circumference change at 9 sites in A N patients - Pre, post, and absolute change 63 8b. ANCOVA fully repeated measures of absolute circumference change with initial body weight as the covariate 64 9a. Relative circumference change at 9 sites in A N patients 65 9b. ANCOVA fully repeated measures of relative circumference change with initial body weight as the covariate 65 10. Selected circumference ratios pre- and post-weight gain in patients with A N 66 11a. DEXA measurement of fat mass at 3 regions in A N patients - Pre, post, absolute change, and relative change 67 l i b . ANCOVA fully repeated measures of absolute fat mass change as measured by DEXA with initial body weight as the covariate 68 11c. ANCOVA fully repeated measures of relative fat mass change as measured by DEXA with initial body weight as the covariate 68 VI LIST OF FIGURES 1. Major body composition compartments 9 2. Experimental protocol 39 3. Differences in the measurement of change in percentage body fat between DEXA and BIA 69 vii ACKNOWLEDGEMENT First and foremost, I would like to thank my thesis supervisor, Dr. Linda McCargar, for her excellent guidance, inspiration, and tremendous encouragement, support and patience throughout this project. I would also like to thank my thesis committee members: Dr. C. Laird Birmingham for his assistance with the initiation and funding of this project, for his assistance with subject recruitment, and for sharing his expertise in Eating Disorders; Dr. Alan Martin for helping me enhance my understanding of the complexities and challenges associated with Body Composition; Dr. Gwen Chapman and Dr. Susan Barr for their valuable contributions to the project; and Dr. Allen Belzberg for making available the D E X A densitometer and his department facilities for body composition measurements. A special thank you to Ruth Foster, nuclear medicine technologist, St. Paul's Hospital, for her excellent technical assistance with D E X A body composition testing, and her flexibility with patients' appointments. I would like to thank the multidisciplinary health care team and nursing staff, Eating Disorders Clinic, St. Paul's Hospital, for their interest and support in this research. Their commitment to helping individuals with Eating Disorders in their struggle for recovery was truly inspiring. Also a special word of thanks to Linda Watts and Kosa Matic-Smyrnis, Eating Disorders dietitians, for their assistance with dietary intake information. A big thank you to my fellow graduate students: Trish Fitzpatrick, Lynn Farrales, Eileen Guillen, Dorcas Lwanga, Fabio Valentini, Sabina Parkes, and Jaye Bulani for their friendship and support. They were always there to cheer me and help me persevere. And to my friend Janet Le Patourel for introducing me to the world of graduate studies. Special thanks to my friends who became my family during all the years away from my home-country: Laura Di Laurenzio, Francoise Arnaud, Sophia Zhou, Bia Boleti, Keti Nikolarea and Roxanne Chabun. Their friendship has been a most precious gift. And a very big thank you to my family in Cyprus, for all their love, support, and for always believing in me. I also wish to thank all the faculty and staff at the School of Family and Nutritional Sciences for their support and encouragement during my studies in the Master's program. Finally, but most importantly, I would like to thank the study participants. I have been amazed and touched by their strength in fighting against Anorexia Nervosa. Their life experiences have taught me lessons that I would not have otherwise learned. I gratefully acknowledge the British Columbia Medical Services Foundation, for providing financial support for this project. viii This thesis is dedicated to my parents Yiannakis and Maro, and to my brother Costas. It is also dedicated to the memory of D. F., one of the study participants, who died in the midst of her fight against anorexia nervosa. ix CHAPTER ONE INTRODUCTION A) Rationale In recent years, there has been an increasing interest among the scientific community in the area of regional fat distribution, particularly in relation to obesity. It is now well-established that many of the health risks of human obesity are due to the distribution of body fat rather than the absolute amount of body fat (Wadden et al., 1988; Vansant, Den-Besten, Westrate, & Deurenberg, 1988). Android (upper body or central) fat distribution which is most commonly seen in men, has been strongly associated with certain metabolic aberrations such as diabetes mellitus, hypertriglyceridemia and hypertension (Lapidus et al., 1984; Ohlson, Larsson, & Svardsudd, 1985). In contrast, these metabolic disturbances have been found to be less common in gynoid (lower body or peripheral) fat distribution most commonly found in women, whose fat predominates in the gluteal-femoral region (Vansant et al., 1988). In recent decades, there has been a dramatic increase in the incidence of the Eating Disorders (ED) Anorexia Nervosa (AN) and Bulimia Nervosa (BN). These disorders are of complex etiology and are characterized by a relentless pursuit of thinness manifested by strict dieting, obsessive exercising, vomiting or laxative abuse which subsequently leads to significant weight loss, often to the point of emaciation (Davis, 1988; King, 1991; Garfinkel & Garner, 1982; Woodside & Garfinkel, 1989). The majority of A N patients are typically dissatisfied with their body weight and shape regardless of their actual weight. Their dissatisfaction may focus on specific areas of the body such as the abdomen, hips, and thighs, or it may encompass contempt against the entire body. They also view their bodies in a distorted way, and can identify areas where they are convinced there is too much fat l accumulation, even though objective evidence does not indicate this to be the case (Davis, 1988). Unpublished observations (Drs Birmingham & Goldner, Eating Disorders Clinic, St. Paul's Hospital) have further substantiated that fear of weight gain by A N patients is enhanced by their belief that this gain is accompanied by preferential fat distribution in the abdominal area. As a consequence, they are reluctant to accept appropriate nourishment, and they avoid gaining weight even when this is of medical necessity. Despite the fact that this fear of \"fatness\" is a central feature in A N , there has been surprisingly little investigation and only minimal understanding of the biology of body composition in patients with this condition (Mitchell & Truswell, 1987). As well, to date there has been no study conducted to examine fat distribution before and after weight gain in A N patients. Thus, no empirical data are available to substantiate or refute these patients' claim that upon refeeding body fat accumulation occurs preferentially in specific areas. The goal of this research was to better understand the nature of body composition and fat distribution changes that accompany short-term weight gain (up to 6 months) in A N patients. It was anticipated that the results of this study would provide relevant information for use in the development of future nutrition intervention strategies for the treatment of this patient population. 2 B) Null Hypotheses Tested 1. There will be no difference in the composition of body weight (fat vs lean tissue) in patients with A N pre- and post-weight gain. 2. There will be no difference in the distribution of body fat in patients with A N pre-and post-weight gain. 3. There will be no difference in the measurement of change in percentage body fat in patients with A N pre- and post-weight gain, as determined by bioelectrical impedance analysis (BIA) and dual energy X-ray absorptiometry (DEXA). C) Study Objectives 1. To determine whether there are changes in body composition pre- and post-weight gain in patients with A N , as determined by dual energy X-ray absorptiometry (DEXA). 2. To determine whether there are changes in body fat distribution pre- and post-weight gain in patients with A N , as determined by skinfolds (SKF), circumferences (CIRC), and dual energy X-ray absorptiometry (DEXA). 3. To determine whether there are differences in the measurement of change in percentage body fat in patients with A N pre- and post-weight gain, as determined by bioelectrical impedance analysis (BIA) and dual energy X-ray absorptiometry (DEXA). 3 C H A P T E R T W O L I T E R A T U R E R E V I E W A review of the literature was done to gain an understanding of body composition patterns that characterize A N , the changes in body composition and body fat distribution that occur upon weight gain in these patients, and the methodologies used to assess such changes. This chapter first provides an overview of the physiology of starvation. This is followed by a description of body composition patterns in normal-weight individuals, and definitions of body composition terminology are provided. The body composition characteristics of A N are then discussed, followed by a review of the body composition and body fat distribution changes that accompany weight gain in A N . Lastly, measurement techniques for body composition and body fat distribution assessment are outlined. A) Physiology of Starvation During starvation, the human body undergoes a series of hormonal and metabolic changes which lead to modifications in body composition. The supply of calories is selectively drawn from adipose tissue, in an attempt to spare the breakdown of vital proteins (Mitchell & Truswell, 1987). Each gram of either protein or glycogen is associated with 3-4 grams of water which are excreted during energy utilization. In contrast, fat is not water-bound (Mitchell, & Truswell 1987). Fuel is required by the body primarily to supply energy for vital functions. A primary fuel is glucose, two-thirds of which (100-145 g) is required by the brain, and one-third by muscles and red blood cells. Glucose is mainly stored as glycogen in the liver. This reserve amounts to approximately 100 g of glucose, which would be enough to sustain the needs of the brain for only 12-16 hours (Cahill, 1970). As a result, gluconeogenesis occurs and 4 protein in skeletal muscle is broken down for the production of glucose by the liver. At the same time, triglycerides in adipose tissue are broken down, to provide muscles with energy (Mitchell & Truswell, 1987). During the early stages of total starvation, protein breakdown occurs at a high rate. Since 1 g of protein is bound to 3-4 g of water, this protein loss accompanied by water loss causes the initial weight reduction to be rapid (Mitchell & Truswell, 1987). As the starvation process continues, more of the weight loss is accounted for by body fat mobilization. Since fat is a richer energy source than protein, each unit of body fat consumed provides relatively more energy, therefore there is a slowing down in the rate of weight loss as starvation continues. At this time, metabolic rate is also reduced as a biological adaptation in preserving energy sources and reducing energy needs (Keys, Brozek, Henschel, Mickelsen, & Taylor, 1950). After 3-4 days, the process of ketosis commences, in which fatty acids released from fat depots are oxidized in the liver to acetoacetic acid and its derivatives. During this time, the brain and other tissues become accustomed to using ketones as energy substrates instead of glucose (Mitchell & Truswell, 1987). However, even after this period, the breakdown of body protein is not completely eliminated in fasted subjects. In fact, the body continues to require some glucose from protein breakdown. Nitrogen, in the form of urea and ammonia continues to be excreted in the urine (Young & Scrimshaw, 1971). The physiology of semistarvation which closely resembles that of the anorexic state, has been studied less intensively than total starvation. One important physiological distinction between total and semistarved states is that there is no evidence of ketonuria in semistarvation (Keys et al., 1950). Similarly ketonuria is uncommon in patients with A N , although increased plasma concentrations of ketoacids are often present (Garfinkel & Garner, 1982). Elevated levels of low density lipoprotein (LDL) cholesterol are commonly seen in A N (Blendis & Crisp, 1968), possibly as a consequence of the enhanced mobilization of adipose tissue (Russell & Beumont, 1987). 5 In summary, starvation and semistarvation appear to progressively select fat as a body fuel. In total starvation, ketogenesis protects against the excess utilization of proteins; however, negative nitrogen balance and protein loss still occur. In contrast, during semistarvation the carbohydrate intake usually provides sufficient glucose so that ketonuria does not develop. Some protein loss persists, which appears to decrease as semistarvation continues (Mitchell & Truswell, 1987). B) Normal Body Composition To explore body composition changes that accompany starvation in general, and A N in particular, one needs to have a basic understanding of body composition in normal healthy individuals. The study of human body composition has a relatively recent history, spanning mainly the last one hundred years (Wang, Pierson, & Heymsfield, 1992). Body composition analysis can provide a tool for characterizing populations; a reference basis for nutritional counselling; and an aid in identifying patterns characteristic of various disease entities (Buskirk, 1987). Thus, body composition measurements can have wide applicability in the fields of clinical nutrition, medicine, and physical education (Buskirk, 1987). Despite the advances that have been made in the acquisition of knowledge in body composition, there are two principal problems that make the study of body composition difficult: First, terminology used to define body composition components/compartments is often unclear. An example of this is the frequent interchangeable use of terms like fat-free mass and lean body mass when in fact these represent different body compartments. This lack of clear definitions can result in errors in published body composition equations and models (Wang et al., 1992). Second, the only method that allows direct measurement of body composition is cadaveric analysis, and to date, very few cadaveric studies have been undertaken (Brodie, 1988). Thus, the majority of the body composition methods commonly 6 used are indirect in nature. As such, a number of assumptions are made that may not necessarily be valid (Martin, Ross, Drinkwater, & Clarys, 1985). Taking into consideration these inherent shortcomings of body composition measurement, it became imperative to define the key body compartments that were relevant to the present research study. Fat is the term used to describe the chemical substance which consists of glyceryl esters of fatty acids. Fat is one component of adipose tissue (Gurr & Harwood, 1991). Lipid is not synonymous with fat. Lipid refers to a group of chemical compounds that are insoluble in water but very soluble in organic solvents. There are fifty different lipids recognized in humans. Glyceryl esters of fatty acids which are referred to as fat are only a subcategory of total lipid (Gurr & Harwood, 1991). Adipose tissue (AT) describes the anatomical tissue which is comprised of fat cells (adipocytes), as well as the cellular matrix made up of protein, water and minerals. Adipose tissue can be either subcutaneous (beneath the skin), or visceral (surrounding body organs) (Snyder et al., 1984). Lean body mass (LBM) is the sum of all tissues in the body excluding AT. In addition to water, protein, and mineral, L B M also consists of \"essential\" lipid; lipid which is found in the marrow of bones, in cell membranes, within the central nervous system and in internal organs (Behnke, Osserman, & Welham, 1953; McArdle, Katch, & Katch, 1991). Lean body mass can be expressed as total mass minus adipose tissue. Fat-free mass (FFM) is not synonymous with L B M . Fat-free mass excludes all lipid, that is, all chemically extractable lipid from a corpse or animal carcass (Mitchell & Truswell, 1987). Thus, unlike L B M , F F M does not include the \"essential\" lipid component (Mitchell & Truswell, 1987). Fat-free mass can be expressed as L B M minus essential lipid. Total body water (TBW) constitutes the largest body compartment. In an average person, it can comprise from 50 to 65% of the total body weight (Moore, 1963). Total body water consists of two fractional components: a) intracellular fluid (ICF), fluid found inside the cell, and b) extracellular fluid (ECF), the non-metabolizing fluid surrounding cells which 7 provides a medium for gaseous exchange, nutrient transfer, and excretion of metabolic end products (Wang et al., 1992). Extracellular fluid is in turn distributed into two main compartments: i) plasma, which represents the fluid portion of blood, and ii) interstitial fluid (ISF), fluid surrounding body tissues. A simplified form of the major body composition compartments is depicted in Figure 1. 8 Figure 1: Major body composition compartments MINERAL PROTEIN INTRACELLULAR FLUID BODY WEIGHT EXTRACELLULAR FLUID FFM LBM FAT ADIPOSE TISSUE 9 C) Body Composition Characteristics of Anorexia Nervosa A N has the following effects on body compartments: 1) Body Fat The few studies that have quantified percentage body fat in the wasted anorexic patient have utilized a variety of body composition methods. In a study by Russell et al. (1983), 6 A N patients, with a mean age of 23.8 years, and with clinically severe malnutrition of four year duration, were found to have 14% body fat. To determine percentage body fat, anthropometric evaluation was done by measuring triceps and iliac SKF thicknesses. In a group of 15 patients with A N , with a mean age of 24 years, Mayo-Smith et al. (1989) measured triceps, biceps, subscapular, and suprailiac SKF, and calculated percentage body fat using the Durnin and Womersley equation (1974). They obtained a value of 13% body fat, in comparison to 29% for their 39 healthy female volunteers. In the same study, computed tomography (CT) was used to assess subcutaneous and visceral fat. In the A N patients subcutaneous fat was 20% and visceral fat was 50% that of normal-weight volunteers. Mazess, Barden & Ohlrich (1990b) used dual-photon absorptiometry (DPA) to measure body fat in 11 A N patients, with a mean age of 21 years, with more than a one year history of A N , and a history of amenorrhea ranging from 1 to 8 years. They obtained a value of 7.8% body fat, in contrast to a value of 26% seen in the 22 control subjects. Although the above researchers utilized different methodologies, they all observed significantly lower body fat values in malnourished A N patients compared to a range of 25-35% body fat commonly seen in normal-weight females (Mitchell & Truswell, 1987). There has been significant interest in the relationship of body fat to menstrual status and endocrine factors in A N . From studies of adolescent women with primary and secondary amenorrhea, Frisch and Mac Arthur (1974) proposed that the onset and maintenance of 10 regular menstrual function are both dependent upon maintenance of a minimum weight for height. This hypothesis, however, has been rejected. A more widely accepted theory is that menstruation is determined by multiple factors related to the processes of growth and development which are regulated by central nervous system maturation, rather than specifically to body fat percentage (Mitchell & Truswell, 1987). Nevertheless, this hypothesis has provided a framework for exploring the relationship between gonadotrophin secretion and changes in body fat in A N patients (Jeuniewic, Brown, Garfinkel, & Moldofsky, 1978). 2) Fluid and Electrolyte Status Edema has been recognized as a common consequence of severe starvation (Mitchell & Truswell 1987). The cause of this \"famine edema\" is still not well understood. Under normal conditions, the body has the ability to closely regulate the amount of total body fluid (Reiff & Reiff, 1992). However, this is not the case in A N patients, who often experience rebound fluid retention, manifested by thirst, decreased urinary output, or puffiness in the fingers, ankles and face. This phenomenon may occur for a variety of reasons including low albumin levels, excessive periods of time spent standing, bingeing on large quantities of salty foods (Kaplan, 1990), self-induced vomiting, or laxative and diuretic abuse (Reiff & Reiff, 1992). The body's adaptive response to these phenomena is to retain more total fluids than normal. It is thus not uncommon for A N patients to experience 2-5 kg weight gain following rehydration. Since these patients find weight gain so frightening, they are unable to distinguish between long-term and temporary water weight shifts, and tend to view all weight increases as changes in body fat related to caloric intake. They respond by further restricting, exercising, bingeing and purgeing, creating a vicious cycle that is resistant to treatment. With respect to the specific impact of edema on body composition, studies in 11 undernourished populations and A N patients have found that E C F space accounts for an increased proportion of total body weight (Moore, McMurray, Parker, & Magnus, 1956; Dempsey et al., 1984; Vaisman, Corey, Rossi, Goldberg, & Pencharz, 1988b). As well, ICF as measured by the difference of T B W and E C F has been shown to decrease, reflecting the decrease in cellular tissue (McCance & Widdowson, 1951; Moore et al., 1956). 3) Skeletal System There has been increasing concern over the long-term effects of A N on bone structure. In fact, a number of reports have found significant osteoporosis and pathologic bone fractures in A N patients (Biller, Saxe, & Herzog, 1989; Rigotti, Neer, Skates, Herzog, & Nussbaum, 1991). The pathophysiology of osteoporosis is multifactorial and not well understood. Several metabolic consequences of A N could adversely affect skeletal mass, including estrogen deficiency, secondary hyperparathyroidism due to low calcium intake or vitamin D deficiency, Cortisol excess and protein-energy malnutrition (Rigotti et al., 1991; Salisbury & Mitchell, 1991; Seeman, Szmukler, Formica, Tsalamandris, & Mestrovic, 1992). Mazess et al., (1990b), assessed bone mineral content (BMC) and bone mineral density (BMD) of the total skeleton, B M D of the lumbar spine and the proximal femur, and total body soft-tissue composition using DPA in 11 female patients with A N . Then-findings confirmed those of previous studies that A N is accompanied by decreased percentage body fat, a reduction in B M C by 10%, and preferential spinal osteopenia. The first longitudinal study on changes in bone mass in A N was reported by Rigotti et al. (1991). This study investigated whether any therapeutic intervention, particularly weight gain would lead to increased bone mass in 27 women with A N , who were followed for an average period of 27 months. No significant change in cortical bone 12 density of the radius as assessed by single-photon absorptiometry was observed during the follow-up period, even though most of the patients had gained some weight, exercised regularly, resumed menses, and increased their calcium intake. As well, no difference was observed between those subjects who had recovered to within 15-20% of their ideal body weight and those who had not. These results were unexpected and contrasted with studies in which estrogen-deficient women had increased cortical mass following treatment and resumption of menses (Klibanski & Greenspan, 1986). Although the results of the Rigotti et al. study are preliminary, there is concern that bone mass remains low despite recovery from A N . It further suggests that the occurrence of A N in young individuals may adversely affect skeletal integrity during adulthood resulting in a predisposition to osteoporotic fractures (Rigotti et al., 1991). To date, no controlled trials of estrogen replacement or calcium supplementation in A N have been reported, thus the proper treatment for decreased bone mineral density in these patients remains unknown. Until further research is undertaken in this area, the most important interventions appear to be medical stabilization and weight gain (Salisbury & Mitchell, 1991). 4) Skin Thickness One of the assumptions underlying the use of SKF is constancy of skin thickness. However, results of a cross-sectional study by Sawas et al. (1989) who investigated the effects of A N on skin thickness do not support this assumption. Thirty six women with A N were compared with a control group of 33 female volunteers of comparable age. The median forearm skin thickness, assessed by a radiological technique developed by one of the authors, was significantly higher (0.88 mm) in the control group compared to the A N patients (0.70 mm). This finding implies that although the contribution of skin to total SKF thickness is generally not large, it may lead to overestimation of body fat in underweight A N patients.. 13 D) Body Composition and Body Fat Distribution Changes after Weight Gain in Anorexia Nervosa While numerous studies have examined the psychological and neuroendocrine aspects of A N , few have looked at body composition. Some of these studies have utilized techniques which have questionable validity in starved populations. As well, the majority of research has focused on changes that occur in total body composition with refeeding. As a result, when SKF measurements are performed, only sums of SKF thicknesses are reported as indicators of total body fat. It thus becomes impossible to evaluate the differential site responses and the distribution patterns of adipose tissue that may accompany refeeding (Himes, 1988). A well-known study which examined body composition changes during nutritional rehabilitation (Keys et al., 1950), focused on refeeding after prolonged experimental semistarvation. When starvation is used as a model for A N , caution must be used as there are clinical differences between starved and anorexic populations. In order to lose weight, A N patients frequently engage in excessive exercise, and may also resort to purgeing and diuretic abuse. In contrast, the experimentally starved subjects voluntarily reduce their exercise levels during the period of weight loss (Mitchell & Truswell, 1987). As well, semistarvation studies normally attempt to replicate the food intake of populations in famine areas by using male subjects whose average daily intake is approximately 1,600 kcal (Keys et al., 1950). This is discrepant from the intake of A N patients which is usually between 700-800 kcal per day (Beumont, Chambers, Rouse, & Abraham, 1981). These figures clearly demonstrate that in terms of caloric intake the starvation of A N patients can be more severe than semistarvation reported in the experimental groups. Furthermore, semistarvation studies have primarily used male subjects, whereas anorexia studies typically use female subjects. Gender differences may affect body composition in ways that make comparisons difficult. Despite these differences, studies of semistarved populations are a useful analogue for AN. 14 The classic study of semistarvation by Keys et al. (1950) was undertaken at the University of Minnesota towards the end of the Second World War. The 32 male subjects participated in a program of 3 months controlled diet, 6 months of semistarvation, and 3 months of nutritional rehabilitation. During the semistarvation period, the male subjects displayed emotional and personality changes also seen in A N patients. They exhibited intense preoccupation with food, they spent prolonged periods eating their meals, and some had episodes of binge eating. Their mood became irritable, their concentration was poor, their sleep was disturbed, and they became socially withdrawn. At the end of the first 3 months of rehabilitation, subjects had regained 86% of their control weight. During the first 2 weeks of ad libitum eating, most subjects achieved food intakes as high as 7,000-10,000 kcal/day; however, eventually intakes levelled off to 3,200-4,500 kcal/day. By 5 months of rehabilitation when weight had been regained to equal previous control levels, subjects were approaching normal food intakes. In the Minnesota study a number of body composition measurements were performed. Body density was used to measure body fat, and extracellular fluid volume was determined by thiocyanate dilution. The fluid findings along with assumed constants for the density of the edema fluid and constants for the maintenance of bone minerals determined by bone density, were used to correct the final determination of body density and thus body fat. \"Active tissue\" mass (mainly skeletal muscle, organs and red celis) was calculated from the difference between body weight, and the sum of fat, thiocyanate space and bone minerals. Lastly, body protein was measured by nitrogen balance and muscle mass was determined by creatinine excretion. During refeeding, a more rapid recovery of adipose tissue compared to that of muscle was observed. At the stage of complete weight regain, the fat content was approximately 120% that of the control values. However, the active tissue mass was still 8% below the control. In terms of body fat distribution changes during nutritional rehabilitation, the male subjects regained their body CIRC to levels that equalled or exceeded the prestarvation basal 15 values. At the end of the rehabilitation period, abdominal CIRC had the greatest rate of increase, followed by the thigh, arm and calf CIRC. Results from the Minnesota study indicate that nutritional rehabilitation of severely undernourished adults is accompanied by a more rapid increase in A T compared to muscle. As well, CIRC measurements revealed preferential deposition of fat in the abdominal and thigh regions rather than the extremities. Although these observations are significant, the anthropometric measurements were performed at a limited number of body sites. Also, as previously mentioned, fat deposition patterns during refeeding in males cannot be extrapolated to anorexic women, as fat distribution patterns during weight gain may be different between the sexes. Russell & Mezey (1962) investigated the changes in body composition upon refeeding in A N patients. Their objectives were to determine the composition of the tissues synthesized when A N patients were placed oh a high-calorie diet, and the energy intake required to yield a given weight gain. Four A N patients, 18-22 years of age, with an initial weight range of 30-44 kg were treated for 5-6 weeks in a Metabolic Unit, where they were very closely supervised. During this time, the patients were provided with a liquid diet which was gradually increased to intakes of 3,500-5,500 kcal daily. Energy balance was measured by keeping a daily record of the patients' activities, the metabolic cost of which was estimated twice weekly by indirect calorimetry. Nitrogen balance was also measured daily. As well, SKF thicknesses were measured at 5 sites (triceps, scapula, midaxilla, abdomen and upper thigh), to determine the distribution of subcutaneous fat deposits. At the end of 6 weeks, weight gain in the range of 6-12 kg was achieved. An average energy requirement of 7,500 calories for a 1 kg weight gain was found. The results of the energy and nitrogen balance studies indicated that 1 kg of deposited tissue contained 77% fat, 7% protein and 16% water in chemical form. In other words, a weight gain of 1 kg was due to an increase of 770 g of fatty tissue and 230 g of cellular matter, the latter including an expansion of the intracellular space by 160 g (assuming that 70% of cellular matter is water). According to 16 the authors, the extracellular space remained unaltered. It is unclear, however, why they assumed that the 16% water gain was entirely intracellular fluid. When changes in SKF thickness were expressed either in absolute or relative terms, no preferential fat deposition patterns were observed. However, when mean gains in SKF thickness (mm of subcutaneous fat) were expressed per kg of weight gained, the following results were obtained: abdomen (0.44 mm/kg), thigh (0.39 mm/kg), axilla (0.28 mm/kg), triceps (0.07 mm/kg), and scapula (0.05 mm/kg). These indicate a greater deposition of subcutaneous fat in the abdomen and thigh regions. Results from the Russell & Mezey study are thus in agreement with the Minnesota study in terms of the preferential deposition of subcutaneous fat that appears to occur in the abdomen and thigh regions upon refeeding. Russell et al. (1983) examined changes in muscle function in relation to body composition changes, during an 8-week refeeding period, in 1 male and 5 female severely depleted A N patients. Anthropometric evaluation included weight, height, midarm muscle CIRC, wrist diameter, forearm diameter, iliac SKF (in females), pectoral SKF (in males), and triceps SKF. From these parameters, L B M and percentage body fat were calculated. Patients received an average of 2,100 kcal per day during the 8-week period. A weight gain of approximately 7.5 kg was achieved, with L B M and body fat accounting for 68% and 32% of the weight gain respectively. Forbes et al. (1984) investigated the effect on body composition of two dietary regimes consisting of different protein content, during a 40-day rehabilitation of 2 male and 10 female A N patients. Patients were fed a hypercaloric diet providing 20% of energy from protein (n=5), or 10% of energy from protein (n=7). The initial caloric intake ranged from 1,600-1,800 kcal/day; this was increased by 400 kcal increments as needed to ensure a weight gain of at least 150 g/day. Nitrogen balance tests and potassium-40 assays were implemented. Body composition measurements included SKF thickness at the biceps, triceps and subscapula, and CIRC at the abdomen, buttocks, mid-arm and forearm. The investigators found no advantage in using high protein diets in the rehabilitation of A N 17 patients. In terms of anthropometric changes during recovery, both groups had the greatest increase in the abdomen and buttocks CIRC with minimal changes in the arm CIRC, or in the biceps, triceps and subscapula SKF. Lean body mass was estimated by the potassium-40 method which assumes that F F M contains 68.1 mEq of potassium/kg in males, and 64.2 mEq/kg in females (Forbes, Gallup & Hursh, 1961). It was found that 70% of the 5 kg weight gain achieved was due to L B M in the high protein group and 61% in the low protein group. Due to the variability among subjects, this difference was not statistically significant. Pirke et al. (1986) examined the body composition changes upon weight gain in 16 female patients with A N , as compared to 11 female normal-weight controls. The potassium-40 method was used to determine body cell mass (defined as the mass of organs, muscles, and ICF). Refeeding of the A N patients was accompanied by a weight gain of 7.9 kg; 55% of the regained weight was fat, and 45% was L B M . The authors noted that body cell mass was reduced in underweight anorexics when compared to normal-weight controls, but that it was normalized rapidly during weight gain, even before ideal body weight was achieved. However, despite significant gains in body weight and body fat during refeeding, body fat mass in the A N patients was still significantly lower than that of controls. Vaisman et al., (1988b), investigated the changes in body composition before and during 2 months of refeeding in 13 adolescent girls with A N . Body fat mass and F F M were derived from the sum of triceps, biceps, subscapula and suprailiac SKF. Intracellular fluid was assessed by the potassium-40 assay, and E C F was measured as the bromide space after oral bromide administration. The authors noted a gradual increase in weight, body fat mass, F F M and total body potassium (TBK) during refeeding. Despite the fact that none of the subjects had clinically detectable edema, E C F was expanded on admission and increased in all patients during the first 5 weeks of treatment, however it later subsided. Overall, refeeding was accompanied by a weight gain of approximately 6 kg, and percentage body fat increased from 16% to 21%. In terms of the composition of the regained weight, 50% was fat and 43% was L B M ; approximately 7% of the regained weight was unaccounted for in 18 terms of its composition. The authors concluded that most of the changes in F F M over the first weeks of refeeding could be accounted for by an expansion in ECF. Melchior, Rigaud, Rozen, Malon, & Apfelbaum (1989), followed 11 female A N patients for a period of 5 months. Body composition measurements included SKF at 4 body sites. A weight gain of approximately 8 kg was observed, 52% of which was comprised of fat mass and 48 % of L B M . Lastly Russell et al. (1994) examined the effects of weight restoration on body composition in A N patients, using anthropometry and in vivo neutron-capture analysis (IVNCA). Thirty two females with A N and 29 matched control subjects were recruited for this study. Total body protein was obtained from measurements of total body nitrogen by IVNCA; total body fat and L B M were derived from the sum of 4 SKF. The initial mean weight of the A N patients was 73% of that of control subjects, increasing to 90% of mean weight of control subjects after refeeding. Overall, a 10 kg weight gain was achieved in the A N subjects. Compared with the control group, A N patients' nitrogen was initially depleted by 25%, increased by 18%, but remained 11% below control values. Body fat was depleted by 58%, increased by 90%, but remained 22% below control values. Therefore, despite a greater initial depletion and a subsequently greater net gain, body fat remained relatively more depleted after refeeding, than did nitrogen and protein. It appears that protein stores are less affected during depletion, and thus remain closer to control values at the end of rehabilitation. Table 1 summarizes the findings of these studies. 19 Table 1: Summary of studies investigating the composition of regained weight in A N patientsa Authors # of Subjects Weight Gain (kg) Method Body Fat (%)b L B M (%)b Russell & Mezey (1962) 4 Females 8.0 SKFC 77.0 23.0 Russell et al. (1983) 5 Females & 1 Male 7.5 SKF 32.0 68.0 Forbes et al. (1984) 10 Females & 2 Males 5.0 SKF T B K d 61.0* 70.0f 39.0e 30.0f Pirke et al. (1986) 16 Females 7.9 T B K 55.0 45.0 Vaisman et al. (1988b) 13 Females 5.4 SKF T B K 50.0 43.0 Melchior et al. (1989) 11 Females 8.4 SKF 52.0 48.0 Russell et al. (1994) 32 Females 9.9 SKF 57.6 42.4 a adapted from Russell, J. D., Mira, M . , Allen, B. J., Stewart, P. M . , Vizzard, J., Arthur, B., & Beumont, P. J. V. (1994). Protein repletion and treatment in anorexia nervosa. American Journal of Clinical Nutrition, 59, 98-102 b % of regained weight c skinfold thickness d total body potassium e value for low protein-fed group f value for high protein-fed group This review of the literature indicates that the results reported by various studies on the composition of tissue deposited during weight gain are discrepant. These discrepancies may be due to differences in sample size, initial state of nutrition, sex, rate of weight gain, composition of the diet, degree of physical activity, presence of edema, and the varying body composition methods used, some of which may give false interpretations when used in starved subjects (Mitchell & Truswell, 1987). For example, there are problems with the use of the potassium-40 assay in assessing body composition in A N patients. Hannan, Cowen, 20 Freeman, & Shapiro (1990b) demonstrated that the relationship between T B K and F F M is not linear in patients at low weight. Use of the assumed constants for the relationship between T B K and F F M would underestimate the latter at low body weights, leading to overestimation of the difference before and after refeeding and an erroneously higher proportion of F F M in the regained tissue. Also, a number of investigators have made the incorrect assumption that L B M contains a constant amount of potassium, when in fact it is F F M which is related to T B K . As has been previously mentioned, L B M and F F M are not synonymous, thus interchangeable use of these terms can lead to errors. In terms of the body fat distribution component of these studies, there is general agreement that preferential A T deposition in the abdomen and thighs may occur during the refeeding of A N patients, as assessed by SKF and CIRC. However, it is important to note that the number of subjects used in these studies was small, thus limiting statistical power. Furthermore, the selection of body sites for the measurement of anthropometric parameters was limited. As well, only SKF and CIRC were used; and to date, none of the more reliable technologies including dual energy X-ray absorptiometry (DEXA) have been utilized to explore fat distribution changes in refed A N patients. For the above reasons more carefully designed studies are warranted to further explore this area. E) Measurement Techniques of Body Composition and Body Fat Distribution This century has seen the advent of a number of techniques for measuring body compositon and A T distribution. Traditionally, hydrodensitometry was one of the most commonly utilized research techniques for measurement of body composition. In fact, it was and it is still considered by some to be the gold standard in body composition despite its limitations as a quantitative index of fat content (Lohman, 1981). Hydrodensitometry uses underwater weighing (UWW) to measure whole-body density. Percentage body fat is then 21 mathematically derived based on the assumption that the body can be divided into 2 compartments: fat mass which has a density of 0.9 g/cm^ and F F M which has a density of 1.1 g/cm^ (Siri, 1956). This method, however, does not take into account differences in the densities and proportions of F F M constituents. This can lead to fat-free densities different from l.lg/cm^ and thus to potentially large errors. Bone mineral, although a relatively small compartment, has the highest density, and the percentage contribution of bone to L B M can vary with age, exercise habits, and various disease states (eg. osteoporosis and AN) (Wang et al., 1992b; Johansson et al., 1993). Further disadvantages associated with this method are that U W W requires specialized equipment and experienced personnel; it is dependent on the cooperation of the subject (full underwater submersion is required which may be difficult for some subjects); and it is affected by the air content of the lungs (pulmonary residual volume) and the gas content of the gastrointestinal tract (Craig & Ware, 1967; Bedell, Marshall, Dubois, & Harris, 1956). Skinfold and CIRC measurements have also been used routinely. The combination of SKF and CIRC measurements provide valuable information on the relationship between adipose and lean tissues (Sjostrom, 1988). The utility of SKF has been twofold: first, they provide a relatively simple and non-invasive method of estimating general fatness, and second, they can be used in the characterization of the distribution of subcutaneous A T (Harrison et al., 1988). Similarly, CIRC and selected CIRC ratios have been used as indices of the patterning of subcutaneous A T (Callaway et al., 1988). In subsequent years, T B K and TBW methods were developed for the determination of L B M , and T B N for the measurement of body protein (Mitchell & Truswell, 1987). Total body potassium is used as an indication of F F M because more than 90% of potassium is found within non-fat cells (Brodie, 1988). Total body potassium can be estimated by measuring 4 0 K (the radioactive form of potassium which is naturally present in the body in a constant proportion of the TBK) in individuals placed in a suitable shielded whole body counter. Thus, F F M is estimated on the assumption that each kg of F F M consists of 63 mmol 22 of potassium (Fidanza, 1980). However, there has been increased controversy over the value of the ratio of body potassium to F F M . In fact, it has been demonstrated that the relationship between T B K and F F M is not linear in various disease states including A N (Hannan et al., 1990b). Total body water can be determined with an isotope dilution technique using deuterium or tritium (Gibson, 1990). This method is based on the assumption that 73.2% of F F M consists of water. The major limitation of this technique is the assumption made that the F F M of an adult contains a constant percentage of water (Fidanza, 1980; Gibson, 1990). Total body nitrogen is used to assess protein status, and it is based on the principle that nitrogen bears a fixed ratio to the mass of protein (1 g N : 6.25 g protein). However, muscularity, body hydration, and the protein and mineral content of the skeleton can affect the relationship of T B N to L B M , thus compromising the validity of this method (Brodie, 1988). Since the beginning of the 1940s, several investigators have used radiological techniques to measure subcutaneous A T thickness. During the 1980's and 1990's, computed tomography (CT) has been used for measurements of A T . Computed tomography is currently considered to be the most reproducible body fat technique available, and it is the new gold standard of body composition (Sjostrom, Kvist, Cederblad, & Tylen, 1986; Grauer, Moss, Cann, & Goldberg, 1984; Armellini et al., 1993). Although C T has the distinctive advantage of measuring both subcutaneous and visceral fat, the amount of irradiation involved and the high cost of utilization, limit its use in field surveys and routine clinical assessments (Ferland et al., 1989; Kuczmarski, Fanelli, & Koch, 1987). The usefulness of ultrasonography (US) for measurement of body composition is also currently being investigated. This technique allows sonographic imaging of the internal body. The operation of US is based on the pulse-echo principle: a beam of high frequency sound waves is emitted, which penetrates the skin surface, passes through the A T and is then reflected off the fat-muscle interface. The time taken for the echo to return to the US receiver is converted to a distance score, and in this manner subcutaneous A T thickness can be measured (van der Koy 23 & Seidell, 1993). Ultrasound overcomes some of the limitations of CT in that it is portable, it does not emit radiation, and it is less expensive (Kuczmarski et al., 1987). However, several sources of measurement error are possible with this technique. These include the inevitable compression created as good contact between the skin and the transducer is required to form an image, as well as the difficulty in identifying which of the reflecting surfaces represent the boundaries of subcutaneous fat (Orphanidou, McCargar, Birmingham, Mathieson, & Goldner, 1994). The development of dual energy X-ray absorptiometry (DEXA) has added further insights into total and regional body fat composition, and skeletal physiology (Lohman, 1992). Finally, bioelectrical impedance analysis (BIA) is increasingly being used in total body fat, T B W and L B M determinations (Brylowski, 1992). In the remainder of this section, the underlying principles of SKF, CmC, D E X A , and BIA will be outlined, and the usefulness of each in assessing body composition and fat distribution in A N patients will be discussed. 1) Skinfolds One of the most practical and most widely used approaches to the assessment of body fat has been the use of SKF (Lohman, 1992). The fact that 50-70% of body fat is located subcutaneously, has given SKF a preeminent position in the quantification of body fatness (Garn, Sullivan, & Hawthorne, 1987; Lohman, 1992). Skinfolds provide a fast, inexpensive, and convenient way of measuring subcutaneous fat in clinical and field settings (Gibson, 1991). Skinfold calipers are used to measure the thickness of a double layer of skin and subcutaneous A T . Various combinations of sites for SKF measurements have been suggested as the basis for equations for predicting total body fat content. The most widely used combination is that of four SKF, utilizing the sum of biceps, triceps, subscapula and 24 suprailiac, proposed by Durnin & Womersley (1974). The authors produced linear regression equations for the estimation of body density (and therefore body fat) from the logarithm of the total of the four SKF thicknesses, for males and females, seventeen to seventy two years of age. The major hmitation of this choice of SKF sites is that it encompasses only the upper body. There was subsequent development of other body fat prediction equations, (Jackson, Pollock & Ward, 1980; Jackson & Pollock, 1978), which utilize up to seven SKF sites, with inclusion of both the upper and lower body. There are two controversies regarding the use of SKF equations (Lohman, 1992). One is the concept that SKF equations are often population specific, and are thus not applicable to other populations or even other samples of the same population. A second controversy relates to the accurate measurement of SKF, and the various technical errors associated with it. This procedure requires training and skill, since choice of an incorrect site can lead to significantly false results (Lohman, 1981). For skilled observers performing SKF in individuals with normal body composition, the standard error of estimation is approximately 3-4% of the body weight (Lohman, 1981; Durnin & Womersley, 1974). In unskilled investigators, however, the error is expected to be much greater. A number of assumptions underlying the use of SKF have also been questioned, particularly when examined in terms of direct evidence from cadaver studies (Martin et al., 1985). These authors specifically challenged five of these assumptions. First, they argue that constant compressibility cannot be assumed, as there is a decline in caliper reading over the first minute of caliper application to the SKF. The cause of this variation may be a varying proportion of interstitial water in the A T (Brans, Summers, Dweck, & Cassidy, 1974). As well, similar thicknesses of subcutaneous A T may yield different caliper readings due to different degrees of tissue compressibility. Second, constancy of skin thickness cannot be assumed, as there is skin thickness variability in different body sites. Although it is recognized that the contribution Of skin to total SKF thickness is generally not large, it may lead to significant errors in lean subjects, including underweight A N patients. Third, the 25 patterning of subcutaneous A T exhibits very wide inter-individual variation. This warrants the selection of body sites from both the upper and lower body, and emphasizes the potential hazards of using fat prediction formulas that do not contain lower limb sites. Fourth, it has been shown that as the level of adiposity increases, its water content decreases, whereas fat content increases. Thus two identical thicknesses of A T may contain different fat concentrations. This implies that a constant fraction of fat in the A T cannot be assumed. And finally, SKF calipers can only measure subcutaneous AT. Thus, in order to estimate total body fat, assumptions must be made about the relationship between internal and subcutaneous fat. Most studies have employed SKF thicknesses in investigations of total body composition. Their use as indicators of fat distribution has been reported in obese subjects during weight reduction (Gray et al., 1990; Sjostrom et al., 1986; Wadden et al., 1988). Their use in studies involving A N patients has been minimal, and has primarily concentrated on total body composition (Forbes et al., 1984; Vaisman et al., 1988a). An exception is the study by Russell & Mezey (1962), in which the authors examined body fat distribution after weight gain in patients with A N . One shortcoming of these anorexia nervosa studies has been that SKF measurements were performed only at a limited number of body sites, primarily in the upper body. Body composition changes that commonly occur in A N may affect the accuracy of measurement by SKF. For example, edema in skin and subcutaneous tissues which occurs secondary to the relative increase in the E C F space in these patients, may mask the full extent of the subcutaneous tissue lost. However, the decreased amount of subcutaneous fat observed in A N patients allows for easier anatomical identification and access to body sites. Despite these inherent shortcomings, SKF measurements have the potential to provide valuable fat distribution information, if they are performed by experienced observers, and a wide selection of body sites is utilized. 26 2) Circumferences Circumferences are useful in measuring the size of cross-sectional and circumferential dimensions of the body (Callaway et al., 1988). Ratios between selected CIRC of the trunk and limbs can provide indices of the patterning of subcutaneous A T (Callaway et al., 1988). Waist CIRC is an index of deep AT. When used in a ratio with the thigh or hip CIRC, waist CIRC is an indicator of the degree of android distribution of AT; the higher the waist:hip (WHR) or the waistrthigh ratio (WTR), the greater the degree of upper body obesity. A WHR greater than 0.95 in males and greater than 0.80 in females is considered to represent android body fat distribution (Ross, Leger, Marliss, Morris, & Gougeon, 1991). In contast, hip and thigh CIRC are associated with gynoid, or lower body obesity (Lohman, 1992). The relationship between WHR and various diseases has been widely investigated. In a large prospective study of middle-aged men (Larsson et al., 1984), a significant association was observed between coronary heart disease and WHR (waist measured at the level of the umbilicus, and hip measured at the level of the iliac crest). This association was not as significant, however, when body mass index (BMI), SKF or CIRC were examined individually. In another prospective study, Lapidus et al. (1984) found the WHR measured in women (waist measured midway between the lower rib margin and iliac crest, and hip measured at the widest point between hip and buttock) to be a stronger predictor of cardiovascular disease than BMI or subscapular and triceps SKF. A limitation in the use of CIRC measurements is that they do not allow evaluation of the potential contribution of visceral or subcutaneous fat, or L B M (Ross et al., 1991). An additional challenge is that various studies have used different anatomical definitions for the waist CIRC. Some authors consider the waist to be the narrowest part of the torso, others consider it to be at the point of the umbilicus (Ross et al., 1991). These discrepancies imply that CIRC location as well as fat depots are confounded in CIRC ratios (Lohman, 1992). 27 Only one study (Forbes, 1990) has utilized WHR in the investigation of a spectrum of diseases. This included 2 males and 30 females with A N . Although the actual CIRC data obtained were not included, the author stated that these patients had WHR which were normally distributed despite the marked loss in weight. Overall, the use of CIRC ratios can be very helpful in observing changes in fat distribution patterns in A N patients, provided that sites from both the upper and lower body are used, and various CIRC ratios are measured. 3) Dual Energy X-ray Absorptiometry Dual photon absorptiometry and more recently, D E X A were the first techniques available to study variation in the B M C of the body (Lohman, 1992). Recently, the application of D E X A has been expanded to estimate the fat and lean contents of the soft tissue of the body (Lohman, 1992). In 1987, the first commercially available dual energy X-ray absorptiometer was designed to estimate spine B M C with greater precision and less radiation exposure than the dual photon absorptiometer. Comparisons between DPA and the new X-ray technique have shown good agreement; however, D E X A has shorter scanning times and increased precision (Shephard, 1991). The fundamental principle on which D E X A is based, involves the differential attenuation by tissues of transmitted X-rays (Mazess, Barden, Bisek, & Hanson, 1990a). A high stability X-ray generator produces photons over a broad spectrum of energy levels. This photon output is filtered to produce two distinct peaks which distinguish bone from soft tissue (XR-Series Bone Densitometer Operator's Guide, 1992), with soft tissue being comprised of fat and non-skeletal fat-free mass (Heymsfield, Wang, Heshka, Kehayias, & Pierson, 1989). Total body scans with D E X A take only 14-20 minutes. Also the radiation 28 dose is extremely low: 0.02 - 0.05 mRem is needed to scan the entire body, in contrast to a typical chest X-ray which emits 40 mRem (Lohman, 1992). Dual energy X-ray absorptiometry is based on a three-compartment body composition model, and is thus able to quantify B M C (regional or whole body), total body fat content and lean tissue mass directly (Hart, Wilkie, Edwards, & Cunningham, 1992). This implies that D E X A measurement, unlike U W W and SKF which employ the two-compartment model for estimating total body fat and F F M , does not rely on the assumptions that B M C represents a fixed fraction of L B M (Hart et al., 1992). In a review of techniques and assumptions in measuring body fat, Martin & Drinkwater (1991) indicated that variation in bone density may influence the accuracy of fat values estimated from UWW. These authors proposed that alterations in bone density may underestimate the F F M component by 34% or overestimate it by up to 168%. Although bone density is affected by age, gender, menopausal status, and certain disease states, U W W cannot account for the variation in B M C , as it uses a constant value for the density of F F M (Pritchard et al., 1993). This points to the potential usefulness of D E X A in making improved estimates of F F M due to its capability of measuring B M C more precisely. The short-term precision of D E X A has been described by Mazess et al. (1990a), for whole-body and regional estimates of lean tissue, bone, mineral, and fat based on 10 repeated measurements in each of 12 subjects (2 measurements per day) over a five to seven-day period. The authors reported the coefficients of variation for B M C and B M D of the total body to be 1.5% and 0.6% respectively. For percentage body fat in the soft tissue, the precision of measurement (standard deviation) was 1.2%. When bone mineral mass (g) was expressed as a percentage of fat-free soft tissue mass, a mean of 5.8% was found with a precision of 0.13% (coefficient of variation = 2.3%). This small amount of variation offers a degree of precision in estimating B M C never before available (Lohman, 1992). Heymsfield et al. (1989), correlated percentage body fat of 13 subjects measured with DPA with the mean percentage body fat estimated by four criterion methods (UWW, TBW, 29 TBK, and neutron activation analysis). The correlation coefficient was 0.95 and the standard error of estimation was equal to 2.5%. This is one of the lowest prediction errors found between two methods. Going et al. (1993) studied the ability of D E X A to detect small changes (<2 kg) in body composition in 17 men and women during a dehydration-rehydration protocol, and compared this to U W W and deuterium dilution methods. Since water is found solely in lean tissue, the resulting changes in body weight were expected to reflect equal changes in L B M with estimates of bone and fat remaining unchanged. As expected, bone mass and fat estimates were unaffected by changes in hydration, however D E X A had a somewhat limited capacity to resolve small changes in soft-tissue mass secondary to fluctuations in hydration in L B M . In comparison to D E X A , UWW was found to be more affected by changes in F F M . Despite the somewhat limited ability of D E X A to resolve small changes in body weight into its lean and fat components, the changes induced in some subjects were not much greater than the reported limits of precision for this technique. On average, D E X A correctly attributed 98% of the change in body weight with dehydration to a change in F F M , whereas the change in F F M by U W W was only 65% of the change in body weight. As well, the change in F F M by D E X A was significantly correlated with the change in body weight (r=0.67), in contrast to a non-significant correlation of r=0.28 in UWW. The authors concluded that D E X A has potential for providing accurate estimates of changes in body composition in studies of longer duration in which greater changes in body mass occur. D E X A may also be useful in assessing body fat distribution and regional body composition. For the estimation of trunk composition and abdominal fat, this method may offer a practical way to validate anthropometric dimensions as indices of abdominal fat (Lohman, 1992). Ley et al. (1992) used D E X A to investigate the effects of gender and menopausal status on body fat distribution and body composition in nonobese, healthy male and female volunteers. Dual energy X-ray absorptiometry was used to estimate total fat and total lean tissue. Also, default software lines were positioned to divide body measurements 30 into the following four regions or \"boxes,\" all of which had the same height: a) android waist region located between the upper part of the dorsal vertebra #12 and the iliac crest, b) android subscapular region which had its lower border at the level of the upper part of the dorsal vertebrae #12, c) gynoid hip region which had its upper aspect through the most superior point of the inner pelvis, and d) gynoid thigh region with its superior border at the level of the inferior border of the gynoid hip boxed region. One important region that was omitted, however, was the abdominal area surrounding the umbilicus. The researchers observed that their boxed regional measurements provided the opportunity to examine specific regions of fat distribution. Their findings, which corresponded to what was expected (in terms of the regional differences between men and women, and between pre- and post-menopausal women) supported the potential usefulness of D E X A in fat distribution investigations. Because D E X A is two dimensional, it cannot differentiate between visceral and subcutaneous AT, in contrast to CT which permits such differentiation. However, CT has the major draw-backs of being expensive, time-consuming, and requiring a relatively high radiation dose. Svendsen, Hassager, Bergmann, & Christiansen (1992) used CT to develop a model which could predict visceral A T in postmenopausal women, as measured by D E X A combined with anthropometry. The percentage body fat determined by D E X A and the mean attenuation of soft tissue by C T were found to be highly correlated, with an error of estimation of 3%. The authors concluded that D E X A measurements of abdominal fat appear to be valid, referring to the potential that D E X A has in estimating regional body composition and abdominal fat. In summary, the precision of D E X A in its primary role of measuring bone is excellent (Mazess et al., 1990a). In addition, measurements done with D E X A are both safe and quick, requiring minimal cooperation from the subject, and the equipment is much less expensive than for other techniques, especially CT. Such ease means that populations of great interest, such as the very young, the very old, and the sick can often be studied by D E X A when other methods cannot be applied (Roubenoff, Kehayias, Dawson-Hughes, & Heymsfield, 1993). 31 Despite these capabilities of D E X A , there are limitations associated with it. Although D E X A is based on a three-compartment model, it is still not free of the assumption of uniform hydration (Roubenoff et al., 1993). Abnormal hydration which is often seen in A N patients, can alter an individual's ratio of mass attenuation coefficient for lean tissue mass, which in turn can lead to errors in the estimation of lean tissue. Because lean and fat mass are summed to give total body weight, an error in lean mass is propagated to the fat compartment, where it may be proportionally much larger (Roubenoff & Kehayias, 1991). Also, D E X A measurements of bone are sensitive to the anteroposterior thickness of the body, so that results may be systematically different between thin and obese individuals (Roubenoff et al., 1993). Another consideration with D E X A technology is its precision in regional measurements of soft tissue body composition. It has been indicated, for example, that D E X A cannot distinguish clearly between soft tissue and bone in the thorax, because the arrangement of the ribs and spine prevents the X-ray beam from finding much bone-free soft tissue mass. This can subsequently lead to imprecise estimates of thoracic composition (Roubenoff et al., 1993). In view of these observations, the precision of D E X A in the determination of regional body composition needs to be further explored. The above account describes both the capabilities and limitations of D E X A as a soft-tissue measurement of body composition. Nevertheless, due to its ability to provide direct measurement of bone tissue and skeletal mass, it is ideally suited for measuring, among other parameters, the bone mass in A N patients, which has been shown, in various studies, to be reduced (Rigotti et al., 1991; Biller et al., 1989; Salisbury & Mitchell, 1991). Dual energy X -ray absorptiometry also has potential as both a body composition and a regional fat distribution measurement technique. A unique feature of the present study is that it is the first one reported to utilize D E X A in examining both whole body composition and body fat distribution upon refeeding in A N patients. 32 4) Bioelectrical Impedance Analysis Bioelectrical Impedance Analysis (BIA) is a relatively new body composition method, which is becoming increasingly popular in the laboratory, clinical and field assessment of human body composition, as it is safe, non-invasive, rapid, portable, inexpensive and easy to operate (Lukaski, 1990; Kushner, 1992). The underlying principle of BIA is that the impedance (Z) of a geometrically isotropic conductor is related to its length (L) and configuration, its cross-sectional area (A), and applied signal frequency (Kushner, 1992). In the body, lean tissues represent a low resistance, high conductance electrical pathway, because they contain large amounts of water and conducting electrolytes. In contrast, fat and bone which contain small amounts of fluid and conducting electrolytes are low conductance, high resistance electrical pathways. Impedance is a function of resistance (R) and reactance (X) across biological tissues (Twyman, & Liedtke, 1987). Resistance is equal to the pure opposition to flow of an alternating current. Reactance is the opposition to flow of an electrical current caused by capacitance produced by tissue interfaces and cell membranes. By definition, a capacitor consists of two or more conducting plates separated from one another by an insulating non-conductive material used to store the electrical charge. In the body, cell membranes theoretically act as capacitors with a bilayer of polar proteins and phospholipids separated by a core of non-conductive lipid (Kushner, 1992). As stated above, the underlying principle of BIA is that the body is an isotropic conductor with a uniform length and cross-sectional area. However, this assumption is not entirely true. The geometrical shape of the human body more closely approximates a series of five cylinders (two arms, two legs and trunk) excluding the head. Since R is inversely proportional to the cross-sectional area, the upper and lower extremities (which have the smallest cross-sectional area) will have the most influence on whole-body R measurements. Conversely, the trunk which contains approximately 50% of the body mass, will contribute <5-12% of total body R (Patterson, 1989). 33 Bioelectrical Impedance Analysis assumes that F F M is linearly related to body heights/body resistance or body heights/impedance (Van Loan, 1990), and that F F M consists of a constant proportion of water (ie. TBW=0.732 FFM). However, in conditions of altered fluid and electrolyte status, the T B W : F F M ratio, and the intracellular to extracellular fluid volume may be altered (Lukaski, 1990). Thus, although it might be reasonable to speculate that BIA may be useful in estimating TBW, extrapolating this to F F M , probably yields spurious results (Cohn, 1985). The most frequently used BIA analyzers are of single frequency, which measure impedance at a fixed 50 kHz frequency with a constant 800 uA alternating current (Chumlea & Guo, 1994). However, these analyzers do not allow the fractionation and quantification of T B W into ICF and E C F components (Chumlea & Guo, 1994). Current applications of the tetrapolar BIA method have focused primarily on healthy individuals for the development and validation of prediction models (Lukaski, 1990). A critical and unresolved question is the validity of BIA in estimating body composition among patients with altered hydration state, including A N patients. To date, very few studies have explored the potential of BIA as a measure of body composition in A N . Hannan et al. (1990b) evaluated BIA in 38 anorexic women with a wide range of BMIs. Fat-free mass was obtained from the mean of three methods: TBW, T B K and prompt neutron activation of nitrogen and was then regressed against impedance to establish the prediction equation with the smallest standard error. This prediction equation used the body habitus parameters of weight, height^/R and shoulder width. They also derived a BIA prediction equation for F F M estimation. Their results indicated that BIA compared favourably with the other methods even in patients with very low BMI. One major shortcoming of the prediction equation derived was, however, that it was based on three methods that also make body composition assumptions which may not be valid in A N patients. In a more recent study by Hannan and co-investigators (1993a), the ability of BIA and anthropometry to measure percentage body fat and its changes in patients with E D , was assessed by determining their agreement with 34 D E X A which was used as the reference method. The authors concluded that in these patients, BIA offers no significant improvement over equations based on anthropometry alone, for the prediction of either percentage body fat or its changes. It appears that the marked shifts in fluid balance observed in A N may limit the usefulness of BIA which relies on the estimation of total body water as a predictor of lean body mass (LBM) (Krahn, Rock, Dechert, Nairn, & Hasse, 1993; Rock & Curran-Celentano, 1994). Calculation of L B M from T B W may in fact overestimate L B M in this patient population (Hannan, Cowen, Freeman, Mackie & Shapiro, 1990a). More long-term studies are required to further explore the clinical application of BIA in A N , possibly through the utilization of segmental and multifrequency BIA, which have the added advantage of assessing ICF and E C F compartments. This review of the literature suggests that despite the fact that loss of body weight and concern about fatness are essential features of A N , there has been little investigation and minimal understanding of the biology of body composition in this condition. Only a few studies have examined the body composition changes which occur upon weight gain in A N patients. Furthermore, the results of these studies have been discrepant, primarily due to the utilization of varying body composition methodologies which may give false interpretations when used in starved subjects. As well, to date there has been no formal examination of changes in fat distribution during weight gain in these patients. No empirical data exist to substantiate or refute the patients' claim that body fat accumulation occurs preferentially in specific body regions upon refeeding. Thus the objective of the present investigation was to address unanswered questions regarding body composition and body fat distribution changes which occur during refeeding in A N , through the utilization of traditional as well as modern body composition methods. 35 CHAPTER THREE METHODOLOGY A) Pilot Study Prior to the commencement of the present study a pilot study was undertaken to determine an appropriate method for the measurement of subcutaneous body fat (please see Appendix 1 for publication). Computed tomography can differentiate between subcutaneous and visceral fat, and is thus considered to be the gold standard for quantifying body fat depots particularly in the abdominal region (Ashwell, Cole & Dixon, 1985, van der Koy & Seidell, 1993). However, the dose of radiation and the high cost associated with its use, limit its utilization in routine body composition studies. Ultrasonography, which involves a more sophisticated technology than SKF, overcomes some of the limitations of CT in that it is portable, it does not emit radiation, and it is less expensive to operate (Kuczmarski et al., 1987). The purpose of this pilot study was to compare US and SKF measurement of subcutaneous fat at 3 abdominal sites with CT which was considered to be the gold standard. If a good correlation was obtained between US and CT measurements, then US would subsequently be used to assess body fat gain in A N patients. This was a cross-sectional study. Twenty two subjects (13 men and 9 women), ages 24-81 volunteered to participate in the study. Al l subjects were recruited from a group of patients who had been scheduled for an abdominal or pelvic scan by CT for diagnostic purposes in the Department of Radiology, St. Paul's Hospital, Vancouver, BC. Body composition assessment included subcutaneous fat measurement by CT, US, and SKF at 3 distinct abdominal sites located 5 cm to the left of midline at 6 cm (site 1), 9 cm (site 2), and 12 cm (site 3) caudal to the inferior tip of the xiphoid process. The relationships 36 among the 3 methods were analyzed by determining Pearson correlation coefficients. A graphical method described by Bland & Altman (1986) was also used to assess agreement among the 3 methods. Significant correlation coefficients were observed between SKF and CT at all 3 abdominal sites (site 1, r = .60, p = .003; site 2, r = .70, p = .0001; site 3, r = .73, p = .0001). Ultrasound and C T methods only showed a significant correlation at site 3 (r = .54, p = .009). As well, the graphical method revealed that the variation in the US measurements was much greater than that of SKF when compared to CT values. The results of this study indicated that relative agreement in the measurement of subcutaneous body fat between SKF and CT was superior to that exhibited between US and CT. As a result of this pilot study, US was not used to investigate body fat distribution changes following weight gain in patients with A N . Instead, D E X A was used to assess body composition and body fat distribution changes in this patient population. B) Experimental Design The present study followed an observational cohort research design. Twenty six malnourished, underweight females, ages 18-45, who met the diagnostic criteria for A N completed the study. Subjects were either inpatients or outpatients of the Eating Disorders Clinic, St. Paul's Hospital, Vancouver, BC. They all followed an individualized weight gain regime, and at the time of their recruitment to the study, they were receiving intensive psychiatric and psychological treatment. Patients were followed for a period of 6 months, or until maximum weight was achieved, whichever occurred first. Assessments of dietary intake and measurements of body composition and body fat distribution were performed at baseline and at the point of maximum weight. Final measurements were performed once the weight agreed upon by the Clinic's treatment team and the patient was achieved and maintained for two consecutive 37 weeks (see Section E of this chapter for details). Monthly follow-up visits were scheduled with the patients, as a means of maintaining contact, developing rapport, and monitoring their progress. Food intake was assessed by 4-day food records. Body composition measurements included: weight (wt), height (ht), elbow and wrist breadths (for frame size determination), SKF and CIRC at 9 sites, D E X A and BIA. All measurements and monthly follow-ups took place at St. Paul's Hospital. This study received ethics approval by the University of British Columbia and St. Paul's Hospital Clinical Screening Committees for Research and Other Studies Involving Human Subjects (Appendix 2). Figure 2 outlines the experimental protocol. C) Subject Recruitment Al l subjects recruited for the study were under the medical care of the attending internist of the Eating Disorders Clinic at St. Paul's Hospital (Dr. C L . Birmingham). Patients were eligible for enrollment in the study if they met the diagnostic criteria for A N as outlined in the Diagnostic and Statistical Manual of Mental Disorders, 3rd. ed. revised (American Psychiatric Association, 1987) (Appendix 3). Only patients who appeared motivated to follow the treatment protocol and gain weight were recruited. Preliminary information regarding the objectives of the study was provided to potential participants in the form of a handout (Appendix 4) by Dr. Birmingham. Once they indicated a willingness to participate in the study, they received a detailed explanation of the study and were asked to sign a consent form (Appendix 5). Once written consent was obtained, the patient was enrolled in the study and measurements commenced. A target of 30 patients was established for recruitment. 38 •2 x vo o < C/i H W w CD < W 2 •M r j to Q « QJ O § 8 c (ti o o o £ o £ CO O s C N o O T to s o a. o 0 8 1 c £ •c X W CS § o £ to CN cn o C/3 H m S w « < p 2 pa cn < > 3 . £ X cn -Si, c c 8 £ . isO cn 4-> cn T3 s O r t f c 2 > < l > C T3 « ai o ^, £ OJ CJCU I-l -a £ g H H IS C o d CJ cn g TS cy i-i *- 8 g 3 S « o z: c o o cn c CU ft) & 3 O 39 D) Measurements of Body Composition and Body Fat Distribution 1) Anthropometry A weight history is crucial in the initial evaluation of patients with A N (Kaplan, 1990). Baseline measurements included the adult premorbid weight which is the weight prior to any attempt by the patient to alter body weight or shape. It is commonly used as a crude approximation of a patient's individual set point, the weight which is genetically determined for physiological homeostasis (Kaplan, 1990). For patients who develop A N in early adolescence, prior to the closing of skeletal epiphyses there is no reliable way of establishing a premorbid adult weight (Kaplan, 1990). In such instances, an approximation of this weight was obtained by considering the average weight for age and height. The weight history also included the menstrual weight threshold (the weight at which amenorrhea or oligomenorrhea occurs) and the patient's desired weight. The discrepancy between the patient's desired weight and the premorbid weight was used as a rough measure of the patient's drive for thinness and degree of body image distortion. A l l anthropometric measurements were performed by the same experienced investigator, to eliminate inter-observer variation. Patients with A N are unreliable in their self-estimation of body weight (Kaplan, 1990), thus baseline and final weights were established by objective measurement rather than by self-report. The patients' weight was obtained using D E X A . Height measurements were determined to the nearest cm and were taken while subjects stood erect without shoes. Body mass index was then calculated as weight in kg/height in m^. Frame size was determined by measuring elbow breadth (the distance between the epicondyles of the humerus) and wrist breadth (the distance between the ulnar and the radial styloid processes), using a sliding caliper. The procedure oulined in the Anthropometric Standardization Reference Manual (Wilmore et al., 1988) was followed. Measurements were 40 taken to the nearest 0.1 mm and results were compared to standards adapted from Frisancho &Flegel(1983). Edema status was assessed by the internist physician, by applying thumb pressure on the anterior tibial plateau. Edema was graded on a scale from 0 to 3 (0 = trace edema, 1 = definite edema, 2 = moderate edema, and 3 = marked edema). An edema of 3 was used as an exclusion criterion, as marked edema has been shown to confound the accuracy of SKF (Martin et al., 1985; Rock & Curran-Celentano, 1994) and BIA measurements (Hannan et al., 1990b; Kushner, 1992; Krahn et al., 1993). Skinfold measurements were performed using Harpenden calipers (Creative Health Products, Plymouth, MI). These precision calipers are designed to measure the compressed double fold of subcutaneous fat and skin. They exert a defined and constant pressure of 10 g/mm^ throughout the range of measured SKF and have a standard pinch area of 20-40 mm^ (Brozek, & Henschel, i963). For all the SKF measurements, the subjects stood erect with feet together and arms at the side. Caliper readings were read 3 seconds after the calipers were applied to the selected body site (Gibson, 1991). Three caliper measurements were taken at each site, and the median value was recorded to the nearest 0.1 mm. One measurement was done at all body sites, and this pattern was repeated for two more times to obtain a total of three measurements. This technique was employed in order to allow the tissue at each site to return to its original form after being compressed by the caliper. Skinfold measurements were taken on the right side of the body at 9 sites: triceps, chest, subscapula, axilla, abdomen a (5 cm to the left of the umbilicus), and abdomen b (3 cm lateral to the midpoint of the umbilicus and 1 cm inferior to it), iliac, thigh and calf. A diversity of sites located in both upper and lower body was chosen to enable a more comprehensive examination of fat distribution changes with time. Skinfolds were measured at two abdomen sites, since a central aspect of this study was to determine whether preferential deposition of fat occurs in the abdomen region. The prodecures outlined in the Anthropometric Standardization Reference Manual (Harrison et al., 1988), were used as 41 guidelines for the anatomical identification of body sites, and for the appropriate SKF measurement techniques. Appendix 6 includes diagramatic representations of the various SKF sites utilized in the study. In the present study D E X A was the principal method used to determine body composition changes which occurred upon weight gain, in contrast to other studies which have used SKF to quantify these changes. Thus, in order to make direct comparisons with previous studies, body composition changes were also assessed by SKF. The equation for predicting body density in women developed by Jackson et al. (1980) was used. This equation utilizes 7 body sites: triceps, chest, subscapula, axilla, abdomen, iliac and thigh. The value obtained for body density was then used to determine percentage body fat by the Siri formula (Siri, 1961). Circumferences were measured using a flexible, nonstretch tape. Measurements were done in triplicate, and the median was recorded to the nearest 0.1 cm. Circumferences were measured at 9 sites: arm, forearm, chest, waist a (at the level of the smallest circumference to the torso), waist b (at the level of the umbilicus), hip, thigh a (mid-thigh), thigh b (proximal thigh), and calf. A l l circumference procedures were performed according to the Anthropometric Standardization Reference Manual (Callaway et al., 1988). A detailed depiction of CIRC measurements is included in Appendix 7. Ratios between selected CIRC of the trunk and limbs were used to provide indices of body fat distribution. The following ratios were calculated: waist a:hip, waist b:hip, waist a:thigh a, waist a:thigh b, waist b:thigh a, and waist b:thigh b. 2) Dual Energy X-ray Absorptiometry All D E X A measurements were performed by an experienced technologist at the Department of Nuclear Medicine, St. Paul's Hospital, at baseline and at the end of the study. 42 An XR-26 M A R K II/HS (high speed) D E X A scanner (Norland Corporation, Fort Atkinson, Wl), was used. During the D E X A measurement, patients were dressed in hospital gowns with all metal objects removed. The patient lay in a supine position, with the scanner placed directly above the top of the centre of the patient's head. The entire body was scanned from head to toes within 14 minutes, and an image was generated on an adjacent computer screen (Appendix 8a shows a whole body image computer print-out). This body image is divided into regions pre-determined by the XR-Series Control/Analysis computer software program (Norland Corporation, Fort Atkinson, Wl) (Appendix 8b). One of the limitations of this program is that the pre-determined regions cannot be removed. However, a maximum of three new regions can be created provided that they are given names different from the existing ones. Appendix 8c, depicts the three regions that were introduced in an attempt to determine the fat distribution changes that occurred upon refeeding in the upper and lower body. These three regions were named as follows: a) subscapular region: located between the anterior axillary fold and the lowest point of the rib cage, b) waist region: between the lowest point of the rib cage and the uppermost point of the trochanter, c) thigh region: between the uppermost point of the trochanter and the mid-thigh. The D E X A data utilized were changes observed with time in: i) total body weight (kg), total body fat (kg), total percentage body fat, and whole body L B M (kg), and ii) body fat (kg) in each of the three regions. Changes in bone mineral content (kg) observed upon refeeding were also measured. It is important to note that in the literature, android (upper body or central) fat distribution is defined as excess fat in the abdomen (McArdle et al., 1991). In contrast, gynoid (lower body or peripheral) fat distribution is associated with fat located in the hips and thighs (McArdle et al., 1991). However, for the purposes of the present study, the chest, abdomen, hips and thighs were all defined as central regions, whereas the arm and calf were classified as extremities. 43 3) Bioelectrical Impedance Analysis A tetrapolar impedance plethysmograph (RJL Systems, Detroit, MI) with a 800 uA current at a frequency of 50 kHz, was used to determine body resistance (R) and body reactance (Xc) (Lukaski, Johnson, Bolonchuk, & Lykken, 1985). Measurements were made within 30 minutes of voiding, with the subject lying on a bed in a supine position, with socks and shoes removed. Spot electrodes were placed on the right hand dorsal surface at the distal metacarpal and on the right foot at the distal metatarsal. Electrodes were also placed between the distal prominence of the radius and ulna, and between the medial and lateral malleoli at the ankle (Lukaski, Bolonchuk, Hall, & Siders, 1986). A diagrammatic depiction of electrode placement is included in Appendix 9. Three measurements of R and Xc respectively were taken, and the median of each was recorded. The RJL BIA 101 Version 1.0 software (RJL Systems, Detroit, MI) was used to calculate percentage body fat. The above software package has been designed to estimate body fat percentage in individuals with normal body composition, and thus its applicability in A N patients is unknown. In an attempt to make more accurate estimations of percentage body fat in this study sample, a more updated BIA software package called Fluid Analysis 1.0a (RJL Systems, Detroit, MI) was also utilized. According to the manufacturer, this software version is more sensitive to hydration changes. Thus, it was anticipated that this analysis could provide further understanding of hydration changes in A N patients who experience considerable edema upon refeeding. E) Meal Support and Renourishment Protocol The critical factor for recovery in A N is the restoration of a healthy weight (Casper, 1982). Various approaches to weight restoration are used in clinics world-wide. Most advocate the promotion of normal eating in A N patients, without the use of liquid 44 supplements, enteral tube feeds, or parenteral nutrition (Williams, Touyz & Beumont, 1985). In fact, some clinicians consider these artificial methods of nourishment as intrusive and anti-therapeutic (Garner, 1985). In contrast, there are those who propose that the use of liquid supplements and tube feeding allow easier weight recovery under certain conditions (Kennedy & Abbas, 1993; Williams, 1958). Anorexic individuals who require a large caloric intake to achieve weight restoration sometimes find it acceptable to eat a normal amount of food at meals and snacks and to increase their caloric intake by drinking liquid supplements. Supplementation may be considered a medicine that can be discontinued when the individual's goal weight has been achieved (Fairburn & Cooper, 1989). Occasionally, anorexic patients prefer exclusive nasogastric tube feeding for weight restoration. Tube feeding may allow these individuals to distance themselves from their ambivalence over actively eating and gaining weight. However, the resumption of normal eating behaviours should be established as early as possible during the course of treatment. To achieve weight restoration in A N patients, the Eating Disorders Clinic at St. Paul's Hospital has implemented a meal support and renourishment protocol, which allows the assessment and documentation of patient eating behaviour and dietary intake. Upon hospital admission, low-weight patients are expected to gain 1 kg/week, and 1% body fat. This is accomplished by an appropriate meal selection plan developed and agreed upon by the inpatient dietitian and the patient. The patient's caloric intake is gradually increased from 1,200 kcal to a level that is more appropriate for age and height (usually in the range of 2,400-3,500 kcal). Meals are supervised by nursing staff, in order to provide patients with encouragement and support, to monitor the amount of food consumed, and to prevent hiding or disposal of food by patients. Food that is not eaten by the patient during each meal, is replaced in the form of a liquid supplement, which the patient is required to consume after the meal. Patients are required to spend 1-2 hours on supervised bed rest after their main meals (lunch and supper) to prevent overexercising or self-induced vomiting. If a patient has not made progress in gaining weight and increasing body fat, the possibility of tube feeding 45 is discussed with the patient, and a combination of oral and nasogastric tube feeding is implemented. As medical and nutritional stability impoves, patients are encouraged to become more physically active. Once the agreed weight and percentage body fat are achieved (usually between 22-25% body fat), a maintenance diet is instituted. Patients, however, are advised to stay 1-2% above the maintenance percentage body fat as a safety margin. Following discharge, patients are followed as oupatients once a week, through the Outpatient Nutrition Clinic conducted by the attending internist and a dietitian. The aim of the present study was not the provision of nutritional intervention during the rehabilitation of A N patients. Instead, the objective was to monitor the changes in body composition and body fat distribution which accompanied the weight gain accomplished as a result of the established refeeding protocol followed by the clinic. F) Assessment of Dietary Intake and Exercise Behaviour 1) Dietary Intake Assessment The dietary intake of subjects who received outpatient treatment was assessed by two four-day food records (Appendix 10) completed at baseline and at the end of the study. Intake of inpatients was monitored by calorie counts performed by the attending inpatient dietitian. One of the major issues in assessing nutritional status in individuals with chronic diseases including A N , is how to accurately measure usual intake of nutrients (Potosky, Block, & Hartman, 1990; Schoeller, 1990). Despite the importance of measuring dietary intake, the existing dietary assessment methods have never been fully validated (Block, 1982). The validity of a dietary intake assessment method describes the degree to which it measures what it is intended to measure (Block, 1982). It is difficult to validate food records because the \"truth\" about dietary intake is never known with absolute certainty (Gibson, 46 1991). Even if actual food intake, monitored by unobstrusive weighed observations compares favorably with results obtained from food records maintained during the same period, there is no guarantee that the records represent the usual food intake (Block, 1982). Precision of a dietary intake method in terms of reliability and reproducibility is another important consideration. A dietary assessment method is considered precise if it gives very similar results when used repeatedly in the same situation (Gibson, 1991). A seven-day food record has been found to be appropriate for estimating the average intake of individuals. However, its use can place a considerable burden on respondents and problems with compliance may arise (Barrett-Connor, 1991; Gibson, 1991). In one study (Gersovitz, Madden, & Smiciklas-Wright, 1978) it was reported that food records completed at the early stages of record-keeping (days 1 and 2) were more valid than those from days 5, 6 and 7. As well, a 4-day food record has been shown to estimate usual protein intake within 10%, with 95% confidence for a small group of individuals (Basiotis, Welsh, Cronin, Kelsay, & Mertz, 1987). For the above reasons, the use of 4-day food records seemed appropriate. Patients were asked to record the time of day, location and amounts of all foods and beverages that they consumed for a period of four days. Three consecutive weekdays and one weekend day were included to account for potential day-of-the-week-effect on food and nutrient intake (Gibson, 1991). Patients were provided with written guidelines on how to complete the food records. They were instructed on how to describe cooking methods, foods eaten and how to estimate portion sizes. Those patients who exhibited purging behaviour were asked to record the foods that they had purged. Once the completed food records had been reviewed with the patient, the energy content (kcal) of the food intake reported, was determined using the Food Processor II Software Package (ESHA Research, Salem, OR). The food intake record form was adapted from the food intake form used by the Dietary Department, University Hospital, U B C site, and that developed for a previous clinical nutrition research study (Reid, 1990). 47 2) Exercise Behaviour Excessive physical activity is a common feature of A N (Kron, Katz, Gorzynski, & Weiner, 1978; Falk, Halmi, & Tryon, 1985). Although many authors have commented on hyperactivity in their A N patients, it is only recently that the importance of this behaviour in the symptomatology of A N has been recognized. Epling, Pierce & Stefan (1983), and Epling & Pierce (1991) have suggested that the association of increasing activity and decreasing food intake is crucial for the pathogenesis of a subset of A N known as \"activity anorexia.\" It has long been known that exercise is accompanied by a paradoxical decrease in appetite in A N patients (Beumont, Arthur, Russell, & Touyz, 1994). In fact, it is common among severely emaciated patients to engage in increased activity as their weight decreases, until they become unable to perform basic tasks of daily living (Beumont et al., 1994). In contrast to most healthy individuals who exercise for the purposes of fitness, health, enjoyment, and recreation, A N patients view exercise as a means of losing weight, disposing calories and gaining feelings of self-worth. Their exercising regimes are rigid, inflexible, and addictive, and the frequent admission among patients of feelings of chronic fatigue, further confirm that their level of activity is excessive (Beumont et al., 1994). Typical exercise patterns in A N include continously standing or being in motion, and spending a considerable part of the day pacing, jogging, powerwalking, or doing calisthenics (Kaye, Gwirtsman, Obarzanek, & George, 1988). This is compounded by restless hyperactivity, characterized by a patient's difficulty in keeping still even for short time periods, becoming distressed if constrained to do so, and performing tasks in an anxious and hurried manner (Beumont et al. 1994). Since the current diagnostic criteria for A N (American Psychiatric Association, 1987) make no reference to activity anorexia, the Eating Disorders Clinic, St. Paul's Hospital uses the following common features encountered in overexercising A N patients (Beumont et al., 1994; Thompson & Trattner-Sherman, 1993) as criteria: a) insistence on exercising daily, as if life depends on it; b) experiencing \"withdrawal\" symptoms such as irritability, anxiety, and 48 depression when restrained from exercising and c) continuing to exercise even when medically contraindicated. Presence of all of the above features is used as an indication of overexercising behaviour. In the present study, information on the exercise behaviour of patients was based on these criteria. G) Statistical Analysis of the Data The Systat PC statistical program (Systat Inc, Evanston, II) was used to perform all statistical tests. The acceptable level of significance was set at p < .05. The difference between initial and final body weight was analyzed using a one-factor repeated measures A N O V A . Differences between pre- and post-weight gain body composition parameters (ie. body fat, percentage body fat, L B M , and bone mineral content) were assessed using a repeated measures A N C O V A design, with initial body weight as the covariate. Body fat distribution changes which accompanied weight gain were assessed using a repeated measures A N C O V A with initial body weight as the covariate. Both the absolute and the relative change which occurred at each body site (for SKF, CIRC), or body region (for DEXA) following weight gain, were analyzed. Post-hoc contrasts on the absolute and relative change among means for the various body sites and regions were conducted by the Scheffe method for complex contrasts. A graphical method described by Bland & Altman (1986) was used to assess agreement between BIA and D E X A in the measurement of change in percentage body fat. These authors propose that this method is useful when the true values of certain parameters (eg. percentage body fat) remain unknown. Under such circumstances a new method has to be evaluated by comparison with an established technique, rather than with the true quantity. Thus when two methods are compared, neither provides an unequivocally correct measurement, however, it is still possible to assess the degree of agreement between them. According to these authors, the use of the Pearson correlation coefficient is inappropriate 49 when measurement techniques are compared. In other words, a high level of correlation does not necessarily signify that two methods produce similar results. One assumption made when this graphical method was used to assess agreement between D E X A and BIA was that D E X A was the more accurate, and thus the more established technique for the measurement of change in percentage body fat. 50 CHAPTER FOUR RESULTS A) Study Participants Twenty six of the 30 patients enrolled in the study completed both baseline and final measurements. Of the 4 who did not complete the study, one patient died of cardiac failure and the remaining 3 patients chose not to participate in the final measurements. Seventeen of the 26 patients had both sets of measurements performed while they were inpatients. Three of the patients had one set of measurements done as inpatients, and the other set as outpatients. The remaining 6 subjects completed all measurements while followed on an outpatient basis, and they did not receive inpatient treatment during the study. The 20 subjects who received inpatient treatment had an average of 1.7 hospitalizations during the study, with a mean length of hospitalization of 11 weeks. These patients were enrolled in the study for an average of 20 weeks. In contrast, outpatients were followed for an average of 48 weeks to provide them with a longer time interval during which they could achieve weight gain, as they did not receive the vigorous refeeding protocol that the inpatients received. Table 2 provides a detailed hospitalization profile and time period of enrollment in the study of the 26 subjects. The data analyzed were based on the 26 patients who completed both baseline and final measurements. In addition, the data of inpatients only were analyzed and these results are included in Appendix 11. No differences were observed in the results of the overall data as compared to the inpatient data, with the exception of one relative circumference and one circumference ratio. These differences are discussed in pertinent sections in Chapter 5. 51 Table 2: Hospitalization profile and length of study enrollment of A N patients Subject # # of Hospitalizations During Study Total Length of Hospitalizations (weeks) Total Length of Time Enrolled in Study (weeks) Inpatients 1 1 12 12 2 2 12 12 3 4 20 32 4 1 8 8 5 1 6 36 6 2 6 40 7 2 11 12 8 1 5 5 9 1 12 12 10 2 3 36 11 2 12 28 12 2 15 24 13 2 16 16 14 1 10 10 15 1 8 8 16 1 12 12 17 2 12 12 18 2 12 12 19 1 12 32 20 2 23 44 Mean 7.7 11 20 Outpatients 21 48 22 — — 48 23 — — 48 24 — — 48 25 — — 48 26 — — 48 52 B) Baseline Data Only the data of the 26 patients who completed the study were included in the analysis. Al l data are presented as mean ± SD. Baseline subject characteristics are shown in Table 3. The low baseline values in weight and BMI, and the long history of A N in these patients reflect the chronicity and severity of their malnourished state. Table 3: Baseline subject characteristics of patients with A N a Characteristic Baseline Patients (n) 26 Sex F Age (yr) 27.6 ±6.6 Disease duration (yr) 10.6 ±6.4 Weightb (kg) 43.6 ±5.4 Height (cm) 162.3 ± 5.7 B M I C (kg/m2) 16.5 ±1.9 a mean ± SD D weight (kg) obtained by DEXA c BMI = body mass index (weight [kg]/height [m2]) 53 C) Dietary Intake and Exercise Behaviour Table 4 depicts the estimated post-weight gain maximum energy intake (kcal/day) for the 26 subjects during the course of their participation in the study. Subjects 1-20 received inpatient hospitalization, and calorie counts for these patients were performed by the attending inpatient dietitian. In 16 of the 20 inpatients, refeeding was achieved by a combination of oral intake and tube feeding; the remaining 4 inpatients were renourished solely by means of oral intake. Subjects 21-26 were outpatients, and they were asked to complete 4-day food records, which were analyzed using the Food Processor II software package (ESHA Research, Salem, OR). Overall, the post-weight gain maximum energy intake of hospitalized patients was 3,450 kcal, which was greater than that of 1,260 kcal seen in the outpatients. This difference in caloric intake is due to the rigorous refeeding protocol implemented by the inpatient clinic. Seventeen of the 26 subjects engaged in purging behaviours, whereas the remaining 9 patients restricted their food intake without purging. Table 4 also provides information on the exercise behaviour exhibited; 23 of the 26 subjects were overexercisers. 54 Table 4: Estimated post-weight gain energy intake and purging and exercise behaviours of patients with A N Subject # Inpatientsa Mode of feeding Estimated Post-Weight Gain Energy Intake (kcal/day) Purging Behaviour Overexerciser 1 oral 2000 no yes 2 oral + tube-feed 4500 no yes 3 oral + tube-feed 4650 yes yes 4 oral + tube-feed 3750 no yes 5 oral + tube-feed 4100 yes yes 6 oral + tube-feed 3910 yes no 7 oral + tube-feed 3300 yes no 8 oral 3500 yes yes 9 oral + tube-feed 4700 yes yes 10 oral + tube-feed 3500 yes yes 11 oral + tube-feed 4200 no yes 12 oral + tube-feed 3300 yes yes 13 oral + tube-feed 2700 no yes 14 oral 2000 no yes 15 oral 3190 no yes 16 oral + tube-feed 3500 yes yes 17 oral + tube-feed 3850 yes yes 18 oral + tube-feed 2600 yes yes 19 oral + tube-feed 2640 yes yes 20 oral + tube-feed 3140 yes yes Outpatients0 21 oral 2000 yes yes 22 oral 1000 yes yes 23 oral 1040 yes yes 24 oral 1330 no yes 25 oral 2000 no yes 26 oral 200 yes no a for inpatients, energy intake estimated by attending dietitian for outpatients, energy intake estimated by 4-day food records completed by the subjects 55 D) Body Composition Changes The mean values for body composition parameters pre- and post-weight gain are shown in Table 5. The weight gain of 6.7+5.3 kg achieved by the subjects was found to be highly significant (p = .000), as well as the increase in BMI of 2.5±2.0 kg/m 2 (p = .000). An increase of 8.2+5.4 in percentage body fat and an increase of 5.3+3.5 kg in total body fat were both found to be highly significant (p = .000). Significant increases in total L B M (1.4+2.9 kg; p = .03) and total bone mineral content (0.057±0.089 kg; p = .005) were also observed. Measurement of percentage body fat using SKF indicated pre- and post-weight gain percentage body fat values of 10.0+2.2 and 14.4+2.6 respectively. An increase of 4.5+2.2 in percentage body fat as determined by SKF was found to be highly significant (p = .000). It can be seen that the pre- and post-weight gain values of percentage body fat as determined by SKF are discrepant from those obtained by D E X A . 56 Table 5: Comparison of body composition parameters pre- and post-weight gain in patients with A N Characteristic Pre Post Weightb (kg) 43.6 + 5.4a 50.3 ± 5.9*** Height (cm) 162.3 ± 5.7 BMIC (kg/m2) 16.5 ± 1.9 19.0 ± 1.7*** Percentage body fatb 18.8 ± 6 . 2 27.0 ± 5 . 1 * * * Total body fatb (kg) 8.3 ± 3 . 2 13.6 ± 3 . 3 * * * Total lean body massb (kg) 33.1 ± 4 . 3 34.5 ± 5.0* Total bone mineral contentb 2.1 ± 0 . 2 2.2 ± 0.2** (kg) a mean ± SD b values obtained by D E X A c BMI = body mass index (weight [kg]/height [m2]) * p < .05 pre vs post weight gain ** p < .01 pre vs post weight gain *** p < .001 pre vs post weight gain E) Body Fat Distribution Changes Body fat distribution changes that accompanied weight gain in the 26 patients were analyzed separately for each of the 3 methods (ie. SKF, CIRC and DEXA) . Both the absolute and the relative change which occurred at each body site (for SKF, CIRC) or body region (for DEXA) following weight gain, were analyzed. In the post-hoc statistical analysis for contrasts, all possible comparisons between individual body sites or combinations of these were analyzed. However, only those contrasts which had body fat distribution relevance, and which were found to be statistically significant are presented. 57 1) Skinfolds Tables 6a and 6b describe the results obtained from the analysis of the absolute SKF change. The dependent variable, Absolute SKF Change was measured as the amount of change in mm of subcutaneous body fat, subtracting the initial from the final measurement. The greatest mean absolute SKF change (mm) was observed at the thigh followed by abdomen b, abdomen a, iliac, triceps, calf, axilla, subscapula, and chest (see Table 6a). A repeated measures A N C O V A was performed on the dependent variable Absolute SKF Change. The within-subjects independent variable was body site, with 9 body locations. The covariate was initial body weight. There was a significant effect of body site, F(8,200) = 18.272, p < .05. The full table of results from the A N C O V A are shown in Table 6b. Post-hoc contrasts on Absolute Skinfold Change among selected subsets of means for the various body sites were conducted by the Scheffe method for complex contrasts. All contrasts were judged against a Scheffe F ' c n t of 15.8 for a = .05 and 200 degrees of freedom. Four contrasts were found to be significant. The first significant contrast was between the composite of the mean absolute SKF changes in the central body sites by contrast to the composite of the mean absolute SKF changes in the extremities. For the purpose of SKF measurements in the present study, the subscapula, axilla, abdomen a, abdomen b, iliac, and thigh were defined as central body sites, while triceps, chest, and calf were considered as extremity sites. The chest SKF was classified as an extremity site due to its location slightly below the shoulder (see Appendix 6). The composite mean absolute SKF change in the central body sites (4.3 mm) was significantly higher than the mean absolute SKF change in the extremities (2.4 mm), F (1,200) = 55.9,/J < .05. The second significant contrast was between the composite of the mean absolute SKF changes in the subscapula and axilla, by contrast to the composite of the mean absolute SKF changes in abdomen a and abdomen b. The composite mean absolute SKF change in the 58 abdomen a and abdomen b (5.3 mm) was significantly higher than that of the composite of the subscapula and axilla (2.5 mm), F (1,200) = 59.5, p < .05. The third significant contrast was between the composite of the mean absolute SKF changes in the subscapula and axilla, by contrast to the mean absolute SKF change at the iliac site. The mean absolute SKF change at the iliac site (4.5 mm) was significantly higher than that of the composite of the subscapula and axilla (2.5 mm), F (1,200) = 20.5, p < .05. Lastly, the fourth significant contrast was between the composite of the mean absolute SKF changes in the subscapula and axilla, by contrast to the mean absolute SKF change at the thigh site. The mean absolute SKF change at the thigh (5.7 mm) was significantly higher than that of the composite of the subscapula and axilla (2.5 mm), F (1,200) = 50.2, p<. 05. Table 6a: Measurement of subcutaneous fat at 9 sites in A N patients - Pre, post, and absolute change Body site Pre Post Absolute Change (mm) Triceps 7.0 + 2.9a 10.4 ± 2 . 8 3.4 ± 2 . 1 Chest 2.8 ± 0.9 4.0 ± 1 . 2 1.2 ± 0 . 9 Subscapula 5.6 ± 1.5 8.1 ± 2 . 2 2.5 ± 1.4 Axilla 4.2 + 1.2 6.7 ± 2.3 2.5 ± 2 . 1 Abdomen a 5.2 ± 1 . 6 10.5 ± 3.8 5.3 ± 3.4 Abdomen b 5.2 ± 1.5 10.5 ± 3.8 5.3 ± 3 . 5 Iliac 4.5 ± 1.8 9.0 ± 4 . 0 4.5 ± 3 . 1 Thigh 9.4 ± 4 . 1 15.1 ± 3 . 9 5.7 ± 3.2 Calf 6.3 ± 3.2 9.0 ± 3 . 2 2.7 ± 1.5 a mean + SD 59 Table 6b: A N C O V A fully repeated measures of absolute subcutaneous body fat change with initial body weight as the covariate SOURCE SS DF MS F P B O D Y SITE ERROR 498.143 681.560 8 200 62.268 3.408 18.272 0.000 Tables 7a and 7b describe the results obtained from the analysis of the relative SKF change (expressed as percentage change). The dependent variable, Relative SKF Change was measured as the amount of change in mm of subcutaneous body fat, subtracting the initial from the final measurement, dividing by the initial value, and multiplying by 100. The greatest relative SKF change was observed at abdomen b, followed by abdomen a, iliac, thigh, axilla, triceps, calf, subscapula, and chest (see Table 7a). A repeated measures A N C O V A was performed on the dependent variable Relative SKF Change, with the within-subjects independent variable being body site with 9 body locations, and initial body weight used as the covariate. A significant effect of body site, F (8,200) = 9.129, p < .05 was found. The full table of results from the A N C O V A are shown in Table 7b. Post-hoc contrast analysis by the Scheffe method, with F ' c r i t of 15.8 for a = .05 and 200 degrees of freedom indicated 3 contrasts of statistical significance. The first significant contrast was between the composite of the mean relative SKF changes in the central body sites by contrast to the composite of the mean relative SKF changes in the extremities. The central body sites were the subscapula, axilla, abdomen a, abdomen b, iliac, and thigh, and the extremities were triceps, chest, and calf. The composite mean relative SKF change in the central body sites (88.2%) was significantly higher than the mean relative SKF change in the extremities (55.1%), F (1,200) = 29.2, p < .05. The second significant contrast was between the composite of the mean relative SKF changes in the subscapula and axilla, by contrast to the composite of the mean relative SKF changes in the abdomen a and abdomen b. The composite mean relative SKF change in the 60 abdomen a and abdomen b (111.9%) was significantly higher than that of the composite of the subscapula and axilla (57.7%), F (1,200) = 39.2, p < .05. The third significant contrast was between the composite of the mean relative SKF changes in the subscapula and axilla, by contrast to the mean relative SKF change at the iliac site. The mean relative SKF change at the iliac site (109.2%) was significantly higher than that of the composite of the subscapula and axilla (57.7%), F (1,200) = 23.6, p < .05. Overall, the contrasts found to be significant with absolute SKF changes were the same as those when relative SKF changes were analyzed, with the exception of the comparison between the composite of the subscapula and axilla, and the thigh; this contrast was not found to be significant with relative SKF change. Table 7a: Relative subcutaneous fat change at 9 sites in A N patients Body site Relative Change (%) Triceps 65.8 ± 62.6a Chest 45.8 ± 38.4 Subscapula 48.1 ± 2 8 . 4 Axilla 67.3 ± 57.6 Abdomen a 111.5 ± 9 3 . 8 Abdomen b 112.3 ± 8 6 . 9 Iliac 109.2 ± 7 5 . 7 Thigh 80.8 ± 65.6 Calf 53.8 ± 4 5 . 1 a mean + SD 61 Table 7b: ANCOVA fully repeated measures of relative subcutaneous body fat change with initial body weight as the covariate SOURCE SS DF MS F P BODY SITE ERROR 14.272 39.084 8 200 1.784 0.195 9.129 0.000 2) Circumferences The same statistical procedures used for SKF changes were utilized in the analysis of the CIRC data. The only difference in the CIRC analysis was that the chest CIRC was considered to be a central body site, due to its anatomical location. This is in contrast to the chest SKF which, due to its anotomical location, was defined as an extremity. Tables 8a and 8b indicate the results obtained from the analysis of the absolute CIRC change. The dependent variable, Absolute CIRC Change was measured as the amount of circumference change in cm, subtracting the initial from the final measurement. The greatest mean absolute CIRC change (cm) was observed at waist a followed by hip, waist b, thigh b, thigh a, chest, arm, calf, and forearm (see Table 8a). A repeated measures ANCOVA was performed on the dependent variable Absolute CIRC Change, with the within-subjects independent variable being CIRC body site with 9 body locations, and initial body weight used as the covariate. A significant effect of body site, F (8,200) = 26.059, p < .05 was found. The full table of results from the ANCOVA are shown in Table 8b. Post-hoc contrast analysis by the Scheffe method, with F' c ri t of 15.8 for a = .05 and 200 degrees of freedom indicated 3 contrasts of statistical significance. The first significant contrast was between the composite of the mean absolute CIRC changes in the central body sites by contrast to the composite of the mean CIRC changes in the extremities. The central body sites were the chest, waist a, waist b, hip, thigh a and thigh b and the extremities included the arm, forearm, and calf. The composite mean absolute 62 CIRC change in the central body sites (5.9 cm) was significantly higher than the mean absolute CIRC change in the extremities (2.0 cm), F (1,200) = 179.3, p < .05. The second significant contrast was between the mean absolute CIRC change at the chest, by contrast to the composite of the mean absolute CIRC change at waist a and waist b. The composite mean CIRC change at waist a and waist b (6.7 cm) was significantly higher than that at the chest (4.2 cm), F (1,200) = 25.2,/? < .05. The third significant contrast was between the mean absolute CIRC change at the chest, by contrast to the mean absolute CIRC change at the hip. The mean absolute CIRC change at the hip (6.6 cm) was significantly higher than that at the chest (4.2 cm), F (1,200) = 16.8,/?<.05. Table 8a: Measurement of circumference change at 9 sites in A N patients - Pre, post, and absolute change Body site Pre Post Absolute Change (cm) Arm 19.6 ± 2. ia 22.3 ± 1.7 2.7 ± 1.7 Forearm 20.3 ± 1.2 21.6 ± 0 . 9 1.3 ± 1.0 Chest 74.4 ± 4 . 2 78.6 ± 4 . 2 4.2 ± 2 . 8 Waist a 59.3 ± 4.2 66.2 ± 5 . 0 6.9 ± 4 . 2 Waist b 63.9 ± 5 . 1 70.5 ± 6.3 6.5 ± 4 . 6 Hip 79.0 ± 4 . 9 85.5 ± 5 . 1 6.6 ± 4 . 3 Thigh a 39.2 ± 3 . 6 43.9 ± 2.6 4.8 ± 3.4 Thigh b 42.5 ± 4.8 49.1 ± 3 . 4 6.5 ± 4 . 5 Calf 30.5 ± 1 . 5 32.5 ± 1.4 2.0 ± 1 . 4 a mean + SD 63 Table 8b: A N C O V A fully repeated measures of absolute circumference change with initial body weight as the covariate SOURCE SS DF MS F P B O D Y SITE ERROR 929.135 891.391 8 200 116.142 4.457 26.059 0.000 Tables 9a and 9b indicate the results obtained from the analysis of the relative CIRC change (expressed as percentage change). The dependent variable, Relative CIRC Change was measured as the amount of circumference change in cm, subtracting the initial from the final measurement, dividing by the initial value, and multiplying by 100. The greatest mean relative CIRC change was observed at thigh b, followed by the arm, thigh a, waist a, waist b, hip, forearm, calf, and chest (see Table 9a). A repeated measures A N C O V A was performed on the dependent variable Relative CIRC Change, with the within-subjects independent variable being CIRC body site with 9 body locations, and initial body weight used as the covariate. A significant effect of body site, F (8,200) = 12.619, p < .05 was found. The full table of results from the A N C O V A are shown in Table 9b. Post-hoc contrast analysis by the Scheffe method, with F Crit of 15.8 for cc=.05 and 200 degrees of freedom indicated 3 contrasts of statistical significance. The first significant contrast was between the mean relative CIRC change at the chest, by contrast to the composite of the mean relative CIRC change at waist a and waist b. The composite mean relative CIRC change at waist a and waist b (11.0%) was significantly higher than that at the chest (5.0%), F (1,200) = 20.7,/? < .05. The second significant contrast was between the mean relative CIRC change at the chest, by contrast to the composite of the mean relative CTRC change at thigh a and thigh b. The composite mean relative CIRC change at thigh a and thigh b (15.0%) was significantly higher than that at the chest (5.0%), F (1,200) = 57.7, p < .05. 64 The third significant contrast was between the mean relative CIRC change at the hip, by contrast to the composite of the mean relative CIRC change at thigh a and thigh b. The composite mean relative CIRC change at thigh a and thigh b (15.0%) was significantly higher than that at the hip (8.8%), F (1,200) = 22.0, p < .05. Overall, the contrasts found to be significant with absolute CIRC changes were different from those when relative CIRC changes were analyzed, with the exception of the comparison between the chest and the composite of waist a and waist b. This contrast was significant when analyzed both in absolute and relative terms. Table 9a: Relative circumference change at 9 sites in A N patients Body site Relative Change (%) Arm 14.2 + 10.4a Forearm 6.5 ± 5.7 Chest 5.0 ± 3 . 9 Waist a 1 1 . 5 ± 7 . 6 Waist b 10.4 ± 7 . 7 Hip 8.8 ± 5 . 9 Thigh a 13.1 ± 1 0 . 2 Thigh b 16.9 ± 12.8 Calf 6.2 ± 5 . 1 a mean + SD Table 9b: A N C O V A fully repeated measures of relative circumference change with initial body weight as the covariate SOURCE SS DF MS F P B O D Y SITE 0.298 8 0.037 12.619 0.000 ERROR 0.591 200 0.003 65 3) Circumference Ratios In addition to individual CIRC, certain CIRC ratios were measured and then comparisons of these were made pre- and post-weight gain in order to determine whether changes in body fat distribution had occurred. Table 10 depicts these results. An initial mean waist a:hip ratio of 0.75, reflects a gynoid fat distribution at baseline. Although the waist a:hip ratio significantly increased with weight gain (p < .001), a post waist a:hip ratio of 0.77 indicates preservation of the gynoid body shape. Significant were also, the pre- and post-weight gain changes in the waist b:hip and the waist brthigh b ratios (p < .05). The changes in the remaining ratios were not statistically significant. Table 10: Selected circumference ratios pre- and post-weight gain in patients with A N Circumference Ratioa Pre Post Waist a:Hip 0.75 ± 0.04b 0.77 ± 0.04*** Waist b:Hip 0.81 ± 0 . 0 5 0.82 ± 0 . 0 5 * Waist a:Thigh a 1.52 ± 0 . 1 3 1.51 ± 0 . 1 1 Waist b:Thigh a 1.64 ± 0 . 1 9 1.61 ± 0 . 1 4 Waist a:Thigh b 1.41 ± 0 . 1 3 1.35 + 0.11 Waist b:Thigh b 1.52 ± 0 . 1 8 1.44 ± 0 . 1 4 * a circumferences measured in cm b mean + SD * p < .05 pre vs post weight gain *** p < .001 pre vs post weight gain 4) Dual Energy X-ray Absorptiometry Analysis Tables 11a - 11c describe the results obtained from analysis of the data measured by D E X A . Data of only 24 of the 26 subjects were utilized in this body composition analysis. 66 For two subjects the baseline amount of body fat in one of the three regions was too low to be quantified by the D E X A computer software. Table 11a depicts the results obtained from the analysis of absolute and relative body fat mass change as measured by D E X A . The greatest mean absolute body fat mass change (kg) was observed at the waist region, followed by the subscapular and thigh regions. The greatest mean relative body fat mass change (expressed as percentage change) was obtained at the subscapular region, followed by the waist, and thigh regions. The relative body fat mass change observed in the 3 regions was very large, particularly that obtained in the subscapular region (1375.4%). An explanation for this is that in some of the patients, initial values of body fat were very low, leading to very large relative changes. A repeated measures A N C O V A was performed on the dependent variables Absolute Body Fat Change and Relative Body Fat Change respectively. The within-subjects independent variable was body region, with 3 body locations, and the covariate was initial body weight. No significant effect of body region was found either in terms of absolute change, F (2,46) = 2.377, p > .05 (see Table 1 lb) or relative change, F (2,46) = 2.131, p > .05 (see Table 1 lc). For this reason subsequent statistical analysis by post-hoc contrasts was not executed. Table 11a: D E X A measurement of fat mass at 3 regions in A N patients - Pre, post, absolute change, and relative change Body region Pre Post Absolute Relative Change Change (kg) (%) Subscapular 0.822 + 0.760a 2.532 ± 1 . 1 9 4 1.710+ 1.159 1375.4 ± 3 3 5 9 . 1 Waist 1.286 ± 0 . 8 0 6 3.106 ± 1.183 1.820 ± 1.264 617.1 ± 1211.7 Thigh 2.212 ± 1.041 3.685 + 1.083 1.473 ± 0 . 9 6 3 114.8 ± 156.6 a mean + SD 6 7 Table l ib: A N C O V A fully repeated measures of absolute fat mass change as measured by D E X A with initial body weight as the covariate SOURCE SS DF MS F P B O D Y REGION 1.396 2 0.698 2.377 0.104 ERROR 13.505 46 0.294 Table 11c: A N C O V A fully repeated measures of relative fat mass change as measured by D E X A with initial body weight as the covariate SOURCE SS DF MS F P B O D Y REGION 1430.772 2 715.386 2.131 0.130 ERROR 15444.550 46 335.751 F) Estimation of Change in Percentage Body Fat: Comparison Between Bioelectrical Impedance Analysis and Dual Energy X-rav Ahsorptiometrv The standard RJL BIA 101 Version 1.0 software (RJL Systems, Detroit, MI) generated percentage body fat values for both baseline and final measurements in 14 of the 26 patients. In the remaining 12 patients, the baseline value for percentage body fat was too low to be detected by the BIA method. Thus, only data of the 14 patients are presented. The change in percentage body fat determined by BIA was calculated by subtracting the initial from the final value generated by the BIA software. This change was compared to the change in percentage body fat determined by D E X A , to assess agreement between the two methods. Figure 3 shows the difference in the measurement of change in percentage body fat between D E X A and BIA plotted against the change in percentage body fat as determined by 68 D E X A . The mean+2 SD of the differences are presented. A mean difference of -1+7.2 in the measurement of change in percentage body fat was found between the two methods. Therefore the change in percentage body fat varied by approximately 3.5% between the two methods, a value that is significant. In addition to the standard BIA software program, a more updated BIA software package called Fluid Analysis 1.0a (RJL Systems, Detroit, MI) was used in the analysis, which according to the manufacturer, is more sensitive to hydration changes. However, this program was able to provide both baseline and final values of percentage body fat in one patient only. In the remaining patients, a value of 0% body fat was generated, labelling these patients as edematous. Thus, data obtained from Fluid Analysis 1.0a were not used. Figure 3: Differences in the measurement of change in percentage body fat between D E X A and BIA . S 3 ox SO B Q OJ 2 ar 2 SS * Q 9 7 + 5 3 • 1 •3 •5 •7 + + + 4-4 6 8 10 Change in % body fat (DEXA) + 2 SD Mean 2 SD 1 2 1 4 69 CHAPTER 5 DISCUSSION A) Major Findings Anorexia Nervosa is a complex disease entity, characterized by alteration and impairment of mental and somatic functions. The most perplexing, yet most commonly described perceptual abnormality in A N , is the patient's inability to assess body size accurately. Some patients exhibit an extraordinary loathing for all parts of their body; others view selected areas such as the stomach, hip or thighs as disproportionate to the rest of the body. Thus, these patients become convinced that dieting and excessive physical activity are essential in order to eliminate these \"protruding\" areas, and they refuse nourishment even when this is medically necessary. In recent years, the association of upper body and abdominal fat with increased risk of several chronic diseases has resulted in a renewed interest in body fat distribution. Although to date, numerous investigators have examined the fat distribution changes which accompany weight loss, very few studies have been done to assess the patterns of body fat following weight gain. Thus there are no data available to support or refute the claims made by A N patients that renourishment leads to preferential fat deposition in specific areas of the body. The objectives of this study were three-fold: to determine the body composition changes which accompany weight gain in A N patients; to assess the changes in the distribution of body fat upon weight gain in patients with AN; and to compare measurements of change in percentage body fat pre- and post-weight gain in A N patients between D E X A and BIA techniques. The main findings of this study are as follows. The rigorous refeeding regime that characterized the hospitalization of A N subjects led to the achievement of a significant 70 increase in body weight. This weight gain included statistically significant gains in total body fat, total L B M , and total B M C . Total body fat was, however, the component which increased the most. In terms of the fat distribution changes which accompanied weight gain, there was general agreement among the results obtained by SKF, CIRC, and D E X A methods. As assessed by SKF and CIRC, there was a significantly greater deposition of body fat in the central regions (ie. chest, abdomen, hip and thigh), than in the extremities (ie. arm and calf). However, when body fat deposition changes pre- and post-weight gain in the abdomen, hip and thigh were compared using the 3 methods, no significant differences were observed among the regions. That is, there was no preferential deposition of body fat in any one of these 3 central regions. Body fat was deposited proportionately in these regions, with no change observed in fat distribution patterns. This finding was confirmed by the analysis of data obtained from CIRC ratios. Although there was a significant increase in the WHR pre-and post-weight gain, the gynoid body fat distribution pattern was preserved. Lastly, the agreement obtained in the measurement of change in percentage body fat upon renourishment between BIA and D E X A was poor., Tt appears that single-frequency BIA may not be sensitive enough to adequately quantify changes in body composition parameters in patients with A N . B) Body Composition Changes Recently, there has been increasing emphasis on the use of body composition analysis in the nutritional assessment of malnourished patients. One objective of this study was to determine the changes in body composition parameters upon normalization of A N patients' clinical and nutritional status following refeeding. A highly significant weight gain as measured by D E X A (6.7+5.3 kg), and a highly significant increase in BMI (2.5 kg/m^) was observed in the 26 A N patients who completed the study. This weight gain was the outcome of the rigorous refeeding protocol followed by 71 the inpatient Eating Disorders Clinic, which consisted primarily of a combination of oral and nasogastric tube feeding. The composition of the weight gain as assessed by D E X A was 78% fat mass, 21% L B M , and 1% bone mineral. Compared to other studies, the amount of fat mass expressed as a percentage of weight gained in the present study is the highest ever reported. There could be a number of reasons for the above discrepancy. First, the degree of initial malnutrition and the amount of weight gain achieved by patients are variable among studies. Second, various body compartments are given different definitions by different investigators. For example some use the term L B M synonymously with F F M , while others view L B M as being comprised of an additional component known as \"essential\" lipid. To further complicate this issue, L B M measured by D E X A refers to bone mineral-free fat-free mass. And third, the body composition measurement methods used in these studies (eg. SKF and TBK) are indirect in nature, and as such a number of assumptions are made which may not be valid, especially when body composition changes in A N are being examined. The body composition results obtained in the present study are in close agreement with those of Russell & Mezey (1962). The initial weight, and the amount of weight gain achieved in 4 A N patients, was similar to that obtained in the present study. Using energy and nitrogen balance studies, these investigators found that 1 kg of deposited tissue contained 77% fat, 7% protein, and 16% water. They also indicated that the 16% water gain was entirely ICF, but it is unclear how they made this assumption. The small sample size used is a definite limitation in generalizing their results. Some studies have examined body composition changes in refed A N patients using the T B K technique (Forbes et al., 1984; Pirke et al., 1986; & Vaisman et al., 1988b). The composition of regained weight as determined in the above studies was 55-70% fat mass, and 30-45% L B M . Thus, the amount of L B M regained upon refeeding in the present study was less than that observed in the literature. This discrepancy can be partly explained by the variability in the amount and type of physical activity that A N patients were permitted to 72 engage in during treatment. It appears that in most of the above studies, patients were allowed to participate in supervised aerobic and nonaerobic activities, contingent on a satisfactory rate of weight gain and reasonable compliance. In contrast, patients in this study were more constrained in terms of their physical activity, and they were encouraged to become more physically active only as their medical and nutritional stability improved. As well, the limitations associated with TBK, which was the method used in these studies must be recognized. Hannan et al. (1990b) demonstrated that the relationship between T B K and F F M is not linear in patients at low weight. Use of the assumed constants for the relationship between T B K and F F M would underestimate the latter at low body weights, leading to overestimation of the difference before and after refeeding and an erroneously higher proportion of F F M in the regained tissue. A number of studies have examined body composition changes following weight gain in A N patients using SKF (Russell et al., 1983; Forbes et al., 1984; Vaisman et al., 1988b; Melchior et al.; 1989; Russell et al., 1994). According to these studies, the composition of regained weight ranged from 32-70% fat mass and 30-68% L B M . Thus, there appears to be a significant discrepancy between the D E X A results of the present study and those from the SKF studies. In order to determine whether this discrepancy was due to methodological differences between D E X A and SKF, or the result of differences in refeeding protocols among studies, body composition parameters of.the present study were also estimated using SKF measurements. The equation for predicting body density in women developed by Jackson et al. (1980), which is based on 7 SKF measurements was used. Measurement of percentage body fat using SKF indicated pre- and post-weight gain percentage body fat values of 10% and 14.4% respectively (versus pre- and post-weight gain values of 18.8% and 27.0% estimated by DEXA) . Although an increase of 4.5 in percentage body fat as determined by SKF was highly significant, it was only one-half of the increase estimated by D E X A . The composition of the weight gain as assessed by SKF was 43% fat mass and 57% L B M , in 73 contrast to the D E X A values of 78% and 21% respectively. Overall, these results are discrepant from the D E X A data, however, they are in agreement with previous SKF studies. Thus these findings imply that the discrepancies in body composition observed between the present study and the literature are primarily due to methodological differences between D E X A and SKF. The validity of using SKF equations to predict percentage body fat in various disease entities including A N is questionable (Lohman, 1992). Skinfold equations, including the one proposed by Jackson et al. (1980) are population specific, and cannot be extrapolated to other populations since the relationship between density and percentage body fat changes (Lohman, 1992). As well, SKF are based on the two-compartment model for estimating total body fat and F F M , and rely on the assumption that B M C represents a fixed fraction of L B M (Hart et al., 1992). However, constancy of B M C in A N cannot be assumed, since significant osteoporosis and pathologic bone fractures have been well documented in these patients (Biller et al., 1989; Rigotti et al., 1991; Salisbury & Mitchell, 1991; Seeman et al., 1992). Dual energy X-ray absorptiometry, on the other hand, is based on a three-compartment body composition model, and is able to quantify B M C directly. Thus despite its shortcomings, D E X A appears to be a more valid method than SKF in quantifying body composition changes and weight gain constitution in refed A N patients. In a more recent study, Russell et al. (1994) investigated the effects of weight restoration on body composition in A N patients, using anthropometry and IVNCA. Thirty two females with A N and 29 matched control subjects were recruited for this study. Total body protein was obtained from measurements of total body nitrogen by IVNCA. Total body fat and L B M were derived from the sum of 4 SKF. Overall, a 10 kg weight gain was achieved in the A N subjects which, according to SKF measurements, was comprised of 58% fat and 42% L B M . These findings are in agreement with those of previous SKF studies. However, the authors proceeded to compare the results from the A N patients with those of the control group, an analysis that yielded different results. Compared with the control 74 group, A N patients' nitrogen was initially depleted by 25%, increased by 18%, but remained 11% below control values. Body fat was depleted by 58%, increased by 90%, but remained 22% below control values. Therefore, despite a greater initial depletion and a subsequently greater net gain, body fat remained relatively more depleted after refeeding, than did nitrogen and protein. No comparisons could be made with the present study as there was no control group, an acknowledged limitation of the present study (see Chapter 6, Section B). These authors emphasized the usefulness of IVNCA in measuring TBN, which in their opinion permits direct estimation of the protein component of the body. In contrast to these authors' claims, it has been indicated that IVNCA has certain limitations (Brodie, 1988; Ryde, Laskey, Morgan, & Compston, 1993). These include the utilization of the factor of 6.25 to relate nitrogen to protein in all body tissues. As well, body hydration, and the protein and mineral content of the skeleton, factors affected by the anorectic state, may influence the relationship of T B N to L B M thus leading to erroneous results. An important question is how the body composition of renourished A N patients compares with that of healthy, normal-weight females of similar age. To address this issue, the findings of studies which investigated body composition in normal-weight premenopausal females using D E X A , were compared with the results of this study. Only 2 pertinent studies were found in the literature, as the majority of research has focused on body composition in groups of individuals with heterogeneous age and weight ranges. Hansen et al. (1993) compared 4 methods for predicting body composition in 100 premenopausal females, ages 28-39. One of the 4 methods used was D E X A . Exclusion criteria pertaining to subject recruitment included participation in any regular exercise program for 2 years prior to the study; experiencing less than 10 normal menses in the previous year, and the use of oral contraceptives or other drugs or medications known to affect bone metabolism. The subjects had a BMI of 22.3 kg/m^. Body composition measurements as assessed by D E X A indicated that fat comprised an average of 30% of body weight, and bone mineral content totalled 2.6 kg (or 4% of body weight). No numerical values were provided for L B M ; however, it 75 appears reasonable to assume that L B M was the difference between body weight and the sum of body fat and bone mineral content. This gives a L B M value of 40 kg (or 66% of body weight). In our study, the renourished A N patients had a BMI of 19 kg/m 2. Twenty seven percent of body weight was comprised of body fat, L B M accounted for 34.5 kg (or 69% of body weight), and bone mineral content was 2.2 kg (or 4% of body weight). These results suggest that the body composition of renourished A N patients is comparable to that of normal-weight premenopausal women. Rico, Revilla, Villa, Alvarez del Buergo, & Ruiz-Conteras (1994) investigated the principal determinants of bone mass in 50 postpubertal women, ages 14-18 yrs using D E X A . Exclusion criteria included a history of menstrual disorders or findings suggestive of gynecologic problems. In addition to determinations of bone mineral content, the investigators examined fat mass and L B M . Subjects had a BMI of 21.7 kg/m 2 . Body composition measurements using D E X A indicated that body weight consisted of 27% fat, 68% L B M and 5% bone mineral content. As with the Hansen et al. study, comparison of these results with the present study suggest that the body composition of renourished A N patients is similar to that of postpubertal females. A final comment is related to the measurement of percentage body fat using D E X A . There are 2 ways in which percentage body fat measured by D E X A can be calculated. First, by knowing the mass of the fat, lean and bone compartments, fat mass can be calculated as a percentage of the total body weight. The second method involves the use of the physical densities of the 3 compartments to calculate the average tissue density. Siri (Siri, 1961) or Brozek equations (Brozek, Grande, Anderson, & Keys, 1963) can then be utilized to compute percentage body fat. The first method avoids the assumption made by the Siri and Brozek equations that bone mass is a fixed fraction of total nonfat mass. However, this straightforward D E X A calculation does not give percentage body fat values which agree with UWW. In the present study, a percentage body fat value of 27% was obtained in the renourished A N patients using method 1, in contrast to 20.5% using method 2 - a discrepancy 76 which is significant. In view of this disagreement, it is necessary to determine which of the techniques contains the error and why. It appears that although the percentage body fat values obtained by D E X A and U W W differ considerably, the 2 methods agree very closely in the measurement of body tissue density (Nord & Payne, 1994). It has been suggested that this discrepancy may be due to the currently used Brozek and Siri equations (Nord & Payne, 1994). Thus, the principal assumptions underlying the derivation of these equations, such as density of lean tissue and fraction of bone mineral in the lean compartment, need to be continuously challenged. C) Body Fat Distribution Changes Body fat distribution after weight gain in A N patients was investigated for two reasons. First, an upper body fat distribution has been shown to be a significant predictor of cardiovascular disease mortality in both men (Larsson et al., 1984); and women (Lapidus et al., 1984). As well, increased amounts of abdominal fat have been associated with elevated triglyceride and L D L subtraction levels, reduced H D L cholesterol levels (Terry, Stefanick, Haskell, & Wood, 1991; Despres et al., 1989) and impaired glucose metabolism in premenopausal females (Peiris, Mueller, Strove, Smith & Kissebah, 1987; Evans, Hoffman, Kalkhoff, & Kissebah, 1984). One of the aims of this study was to determine whether aggressive refeeding in A N patients results in preferential deposition of body fat in the abdomen area which, over time, could potentially lead to atherogenic metabolic profiles for these women. Second, A N patients typically refuse to accept appropriate nourishment during treatment, as they believe that this will lead to excessive fat accumulation in certain body areas - specifically the abdomen, hips and thighs. One objective of this study was to provide empirical data to either substantiate or refute these patients' beliefs; and to provide a basis for developing strategies to help patients cope with the body changes which accompany weight gain. 77 1) Skinfolds Results obtained from SKF measurements indicated that in terms of absolute change, there was a significantly greater deposition of body fat in the central regions (chest, abdomen, hip and thigh) than in the extremities (arm and calf). This finding comes as no surprise. Preferential deposition of body fat in the central regions rather than the extremities under conditions of weight gain constitutes the norm (Lapidus et al., 1984; Ohlson et al., 1985). The greatest absolute SKF change was observed at the thigh, followed by the 2 abdomen sites, the iliac, triceps, calf, axilla, subscapula, and lastly the chest. Individual or combinations of body sites were then compared to determine specific patterns of body fat distribution. One important finding of this analysis was that when the absolute SKF changes in the abdomen, iliac and thigh sites were compared, no significant differences among these 3 sites were found. This suggests that there was no preferential deposition of body fat in any of these 3 central regions; rather, body fat appears to be deposited proportionately in the 3 regions. Significant differences were obtained, however, when the composite of the subscapula and axilla sites both located in the thoracic region, were compared with each of the abdomen, iliac and thigh sites. This finding indicates that during renourishment of A N patients, less body fat is deposited in the upper central sites (subscapula and axilla) in comparison to the lower central sites (abdomen, iliac and thigh). Similar results were obtained when SKF data was analyzed in relative terms. A significantly greater deposition of body fat was observed in the central regions than in the extremities. The greatest relative change was observed at the 2 abdomen sites, followed by the iliac, thigh, axilla, triceps, calf, subscapula and chest. Although the sequence of sites in terms of the relative SKF increase achieved varies from that of the absolute change, this variation is slight, and the results essentially lead to the same conclusion - that there is greater deposition of body fat in the abdomen, iliac and thigh compared to the remaining sites. As well, when the relative SKF changes in the abdomen, iliac and thigh sites were compared, no 78 significant differences among these 3 sites were found. When the composite of the subscapula and axilla sites were compared with each of the abdomen, iliac and thigh sites, significant differences were obtained with the abdomen and iliac, but not with the thigh. Overall, results from the absolute and relative SKF change analysis were in agreement. There is only one study in the literature that the above results can be compared to, and this study was conducted by Russell & Mezey (1962). These investigators performed SKF measurements at 5 body sites to determine the distribution of subcutaneous fat in 4 AN patients undergoing refeeding. These patients were 18-22 years of age, and received intensive treatment for 5 weeks, during which they achieved an average weight gain of 8 kg. In terms of absolute SKF gains, the greatest increase was observed at the thigh, followed by the abdomen, subscapula, axilla, and upper arm. The relative SKF gains occurred, in descending order, at the abdomen, thigh, axilla, subscapula and upper arm. These findings are in close agreement with the results obtained in this study. It is important to note, however, that there were only 4 patients in the Russell & Mezey study, thus generalizability of their results must be viewed with caution. 2) Circumferences CIRC measurements were also performed in order to obtain further information on the fat distribution changes that occur upon weight gain in AN patients (although it is recognized that CIRC measurements do not permit evaluation of the individual contributions of visceral fat, subcutaneous fat and LBM). As with SKF results, a significantly greater absolute CIRC increase was observed in the central regions (chest, waist, hip and thigh) compared to the extremities (arm, forearm and calf). The greatest absolute CIRC gain was observed at waist a, followed by the hip, waist b, thigh b, thigh a, chest, arm, calf and forearm. Again, no significant differences were obtained when the waist, hip and thigh CIRC were compared with each other, indicating that no preferential deposition of body fat had 79 occurred in any of these 3 central sites. Comparisons between the chest CIRC and each of the waist, hip and thigh CIRC, revealed a significant difference between the chest and waist, and between the chest and hip, but not with the thigh CIRC. This finding supports the observations made with absolute SKF results, namely that during the refeeding of A N patients, less body fat is deposited in the upper central region (chest) compared to lower central ones (waist and hip). A possible explanation as to why no significant difference was obtained between the chest and thigh, was because the change in the chest CIRC was compared to the composite of the change in thigh a and thigh b CIRC. The change that was observed in thigh a was approximately 1.8 cm less than that observed in thigh b, thus reducing the combined mean for both the thigh CIRC, and subsequently leading to a non-significant difference when compared to the change in the chest CIRC. The results obtained from the analysis of relative CIRC change are more variable, and thus more difficult to interpret. The greatest relative CIRC change was observed at thigh b, followed by the arm, thigh a, waist a, waist b, hip, forearm, calf and chest. An unexpected finding was the relative increase in the arm CIRC, which ranked second in magnitude following the change in thigh b. Being an extremity site, it would have been expected that less change would have been obtained at the arm in comparison to the central sites. A possible explanation for this discrepancy is the fact that pre-weight gain, the arm CIRC was small in magnitude and any amount of post-weight change achieved would result in a magnified relative change (since division by a small initial value leads to a large number). The same explanation can be given in terms of the greater relative CIRC change observed at the forearm and calf compared to the chest. For this reason, no significant differences in relative CIRC increase were observed in the central regions (chest, waist, hip and thigh) compared to the extremities (arm, forearm and calf). When data of inpatients were analyzed separately (see Appendix 11), results were similar to those of the overall data. However, one additional comparison was found significant, and that was between the relative CIRC change of the composite of waist a and b 80 and that of the composite of thigh a and b. In inpatients then, it appears that in terms of relative change, there was a significantly greater body fat deposition in the thigh regions than the waist. Selected CIRC ratios were also determined pre- and post-weight gain, and these were used as indices of body fat distribution change. A significant change in WHR from 0.75 to 0.77 was observed. This increase in the WHR is expected, as increased levels of body fatness invariably lead to an increased WHR. It is important to note that the post-weight gain WHR value of 0.77 is below 0.8 - the cut-off point that characterizes upper body fat distribution. This indicates that despite a significant weight gain, these patients were able to retain a lower-body fat distribution shape. Additionally, a decrease was observed between pre- and post-weight gain in all the measured WTR; however, only the change in the waist b to thigh b ratio was statistically significant. This general trend of a decrease in the WTR suggests that there was a greater increase in the thigh CIRC in comparison to the waist CIRC. This confirms the observation that the gynoid body fat distribution in the A N patients was maintained post-weight gain. An additional statistically significant decrease was observed in inpatients in the waist a to thigh b ratio pre- and post-weight gain. This finding substantiates the general trends observed in the overall data described above. There are no studies cited in the literature examining the effect of weight gain on CIRC change in A N patients; thus no comparisons could be made with the results obtained in this study. 3) Dual Energy X-ray Absorptiometry Analysis The overall results obtained by D E X A were in agreement with those by SKF and CIRC. Examination of the absolute change in fat mass in the 3 D E X A regions revealed no significant differences. Despite the lack of statistical significance, the greatest absolute 81 change was observed at the waist region, followed by the subscapular, and thigh. It should be noted that there was a large degree of variability (presented in terms of standard deviation) in the amount of fat mass gained after refeeding. Noteworthy is the lack of significance obtained when the subscapular region was compared to the waist and thigh regions respectively. This finding was not in agreement with observations made with SKF and CIRC, both of which indicated significantly less deposition of body fat in the thorax in comparison to the waist and thigh. A possible explanation for this discrepancy is that the chest SKF and chest CIRC are located in the upper end of the thoracic region, whereas the D E X A subscapular region spans the entire thoracic region. It appears that the chest SKF and chest CIRC are located closer to extremity sites rather than central sites, and this could possibly explain these differences. Similarly, no significant differences were obtained when the relative changes in fat mass in the 3 D E X A regions were compared. This again can be attributed to the large amount of variability in the results; in fact the relative changes in fat mass for all 3 D E X A regions had standard deviations which exceeded the mean value. The greatest relative change was observed in the subscapula, followed by the waist and thigh. The relative change in the subscapular region was 1375%, a value which is strikingly high. This can be explained by the fact that 15 of the 24 subjects had baseline subscapular fat mass values which were less than 1 kg - with one patient having a value of only 13 g. When the difference between initial and final fat mass is divided by a very small baseline value to determine the relative change, the result obtained is very large, and this explains the magnified value obtained with the subscapular region. Thus, although the absolute change in fat mass in the 3 regions was not large, the low initial values in some patients led to very large relative changes. It is also important to note that the D E X A data of 24 of the 26 patients were utilized, because in 2 subjects the baseline amount of body fat in the subscapular region was too low to be quantified by the D E X A computer software. It has been indicated that D E X A cannot distinguish clearly between soft tissue and bone in the thorax, because the arrangement of the 82 ribs and spine prevents the x-ray beam from finding bone-free soft tissue mass, so that estimates of thoracic composition tend to be imprecise (Roubenoff et al., 1993). This problem would become more pronounced in malnourished A N patients, who, as has been shown, have very little fat mass in the thorax. It is thus unclear as to whether the results obtained for the subscapular region are accurate, and no information is available in the literature to make any comparisons. Overall, the results obtained by all 3 methods are in general agreement. Although direct comparisons among the 3 methods cannot be made as each evaluates a different body composition component (ie. subcutaneous fat in SKF, subcutaneous fat and muscle in CIRC, fat mass in DEXA) , all 3 methods are suggestive of the same observation. It appears that there is no preferential deposition of body fat in any one of the 3 central regions during the refeeding of malnourished A N patients. Instead, body fat appears to be deposited proportionately in the 3 regions, thus causing no alterations in fat distribution patterns. This was confirmed by the data from CIRC ratios. It thus appears that the gynoid body fat distribution was preserved in the refed A N patients. This is desirable because this type of body fat distribution is reported to be protective against atherogenic risk. These preliminary findings also substantiate the claims of A N patients that weight gain leads to greater deposition of body fat in the abdomen, hips and thighs, compared to other parts of the body. One interesting question that remains to be answered is how the body fat distribution of renourished A N patients compares to that of healthy, premenopausal females. Such comparisons are difficult to make, as the number of studies done in this area are limited, and the methodologies used are variable. Only 2 pertinent studies were found in the literature and the findings of these are discussed below. Ley et al. (1992) used D E X A to investigate the effects of gender and menstrual status on body fat distribution in nonobese, healthy male and female volunteers. In addition to male and post-menopausal female subjects, the study included 61 women, ages 19-51, who were premenopausal, as judged by a regular menstrual cycle and no menopausal symptoms. None of these subjects were obese (mean BMI was 22 83 kg/m2), or receiving sex-steroid therapy, and they were all in good health. To assess body fat distribution, the investigators used D E X A default software lines and divided body measurements into the following four regions or \"boxes,\" all of which had the same height: android subscapular region; android waist region; gynoid hip region; and gynoid thigh region. One important region that was omitted, however, was the abdominal area surrounding the umbilicus. The amount of fat mass (kg) in each of the 4 regions was: 2.1 subscapular fat, 2.3 waist fat, 3.6 hip fat, and 3.3 thigh fat. Comparison of these results with those of the present study is difficult because the D E X A compartments chosen were different. It appears, however, that these subjects had significantly greater fat mass in the hip and thigh regions than the renourished A N patients. Whether the fat distribution seen in these 61 females is reflective of the fat distribution patterns in the general premenopausal female population is unknown. Thus, comparisons must be made with caution. Williams et al. (1993) examined the relationship between fat distribution and dehydroepiandrosterone sulfate (an abundant steroid in humans thought to have an antiobesity effect) in premenopausal females. Ninety six healthy females, ages 28-39 completed the study. Regional fat distribution was assessed by circumference ratios. As well, D E X A was used to measure fat mass in the trunk and leg regions. A mean WHR of 0.73 and WTR of 1.49 were obtained respectively. In the renourished A N patients of our study, a value of 0.77 for WHR and 1.35 for WTR were obtained respectively. Although the WHR results obtained in the 2 studies appear to be in agreement, the WTR values are discrepant. The authors do not indicate whether they utilized mid- or proximal thigh in then-measurements; the location of measurements may thus be a possible explanation for this discrepancy. 84 D) Estimation of Change in Percentage Body Fat upon Refeeding: Comparison between Bioelectrical Impedance Analysis and Dual Energy X-rav Absorptiometry The results of this study indicate poor agreement between D E X A and BIA for the measurement of change in percentage body fat upon refeeding in A N patients. This is reflected by the fact that the change in percentage body fat varied by approximately 3.5% between the two methods, a value that is significant. The refeeding of A N patients is accompanied by significant body composition changes (Vaisman et al., 1988b). For example, TBW has been reported to increase, with an expansion of E C F at the expense of the ICF compartment. The marked shifts in fluid balance that accompany refeeding in A N may limit the usefulness of BIA which relies on the estimation of T B W as a predictor of L B M (Krahn et al., 1993; Rock & Curran-Celentano, 1994). Calculation of L B M from T B W has been shown to overestimate L B M in this patient population (Hannan et al., 1990a). Hannan et al. (1990b), compared F F M in anorexic subjects as determined by BIA, TBW, TBK, and nitrogen neutron activation. As well, they derived a BIA prediction equation for F F M estimation using weight, height, resistance, and shoulder width. Their results indicated that BIA compared favourably with the other methods even in patients with very low BMI. However, all the methods used in the study are indirect in nature, and make assumptions regarding body composition that may not be valid in A N . In a more recent study, Hannan and co-investigators (1993b) evaluated the use of D E X A for the assessment of body composition in 37 anorexic females. They obtained a high correlation between F F M measured by D E X A and that measured by prompt neutron activation analysis and TBW. They concluded that D E X A was a suitable method of assessing F F M and that the reproducibility of the results obtained was sufficiently good to allow changes in body composition to be followed during the course of the patients' 85 treatment. Hannan, Cowen, Freeman, Wrate & Barton (1993a) assessed the ability of BIA and anthropometry to measure percentage body fat and its changes in 24 refed A N patients, by determining their agreement with D E X A which was used as the reference method. The authors concluded that when compared to D E X A , BIA offers no significant improvement over equations based on anthropometry alone, for the prediction of either percentage body fat or its changes in A N patients. Various investigators have explored the clinical application of BIA in other disease entities which are characterized by malnutrition and altered hydration state including acquired immunodeficiency syndrome (ADDS) (Wang et al., 1992b); Crohn's disease (Royall et al., 1994); chronic hemodialysis (Rammohan & Aplasca, 1992); and intensive care patients (Robert et al., 1993). The findings of these studies vary in terms of the validity and reliability of BIA, due to differences in methodology, and the use of different prediction equations. Bioelectrical impedance analysis body composition equations are derived by regression analysis and thus tend to be population specific. This implies that body composition estimates are adjusted toward the group mean, leading to under- or overestimation of body fatness in patients whose body composition differs from the mean. It also appears that the assumption made by whole-body BIA that the body is an isotropic conductor with uniform length and cross-sectional area may not be valid (Kushner, 1992). As well, the use of a 50-kHz current in single-frequency bioelectrical impedance analyzers does not allow the fractionation of TBW into intra- and extracellular components (Chumlea & Guo, 1994). In the present study, use of the standard RJL BIA 101 Version 1.0 software (RJL Systems, Detroit, MI) was able to generate percentage body fat values for both baseline and final measurements in only 14 of the 26 patients. Thus an attempt was made to estimate percentage body fat more accurately by using, in addition to the manufacturer's BIA formula, a more updated BIA software package - Fluid Analysis 1.0a (RJL Systems, Detroit, MI), which according to the manufacturer, is more sensitive to hydration changes. However, this 86 program was able to provide both baseline and final values of percentage body fat in one patient only. In the remaining 25 patients, the program produced a value of 0% body fat, labelling these patients as edematous. These results show that BIA has limited ability to assess changes in percentage body fat, following refeeding in patients with A N . Also as indicated by a graphical method, BIA compares poorly to D E X A in the measurement of changes in percentage body fat. More long-term studies are warranted to further explore the clinical application of BIA in A N , possibly through the use of segmental and multifrequency BIA. 87 CHAPTER SIX CONCLUSIONS AND RECOMMENDATIONS A) Conclusions One of the fundamental clinical features of A N is an extreme concern for body weight and shape. Anorexia nervosa patients are dissatisfied with their bodies, and they fear that if they gain weight, that body fat will be deposited preferentially in the abdomen, hips and thighs. Subsequently this fear may contribute to their resistance in gaining weight even when they are obviously emaciated. The primary objective of this study was to address the patients' concerns by investigating body composition and body fat distribution changes that characterize short-term weight gain in A N . Results of body composition changes included a highly significant weight gain which consisted of significant gains in body fat, L B M , and B M C . Total body fat was found to be the component which increased the most. The discrepancies in body composition changes observed between the present study and the literature are primarily due to methodological differences. Analysis of absolute and relative changes in body fat distribution by SKF and CIRC revealed a greater fat deposition in the central regions than in the extremities. Comparison of fat mass change among the 3 central regions as measured by D E X A , however, indicated no significant differences. These results were also supported by CIRC ratios. It appears that the gynoid fat distribution pattern (believed to be protective against the health risks associated with android body fat distribution) is preserved in these patients, despite the significant weight gain achieved during refeeding. It is recognized, however, that promoting an increase in body fat improves morbidity and mortality, and overides the potential risk of cardiovascular disease in these patients over the long term. 88 Lastly, a poor agreement was obtained between BIA and D E X A for the measurement of change in percentage body fat. Single-frequency BIA may thus not be sensitive enough to quantify changes in body composition in patients with A N . More studies are needed to further validate the use of sophisticated body composition methods in A N . In the meantime, the findings of the present study can constitute an important component of nutrition education strategies for the treatment of A N . B) Limitations to the Study Twenty six of 30 subjects successfully completed both baseline and final measurements. However, only 20 of these, who received at least one inpatient treatment achieved significant gains in body weight and body fat. Six of the subjects who were recruited as outpatients either refused inpatient treatment or were medically stable, and thus did not require urgent hospitalization. Whether or not outpatients would consent to inpatient treatment was beyond the investigators' control, but the fact that they did not receive aggressive inpatient refeeding and were subsequently unable to make significant gains in weight was a limitation to the study. This was an observational cohort study following a group of A N patients and determining body composition and body fat distribution changes upon weight gain. It would have been beneficial if age- and height-matched female controls were also recruited, in order to determine how body composition parameters in A N compare with those in a normal-weight female population. The limitations of all body composition methods used in this study, including D E X A , are recognized as these methods are indirect in nature, and thus make assumptions that may not be valid in A N . To date, there is no information available in the literature regarding the optimal selection of regions for body fat distribution measurements using D E X A . Thus it is uncertain whether the selection of D E X A regions in this study was fully appropriate. In 89 addition, regional tissue variability might have had an effect on the body fat distribution results obtained. Hydration status was not uniform during the various stages of the study. Patients were typically dehydrated prior to renourishment, but were euhydrated once they had reached a maintenance weight. Since fluid gains have an effect on L B M , the big fluid shifts that accompany refeeding in A N could have introduced error in the measurement of L B M . As well, inter-observer reliability for D E X A was not tested in this study, thus the reliability of D E X A in measuring body composition and body fat distribution was assumed without being directly assessed. Lastly, the pre-morbid body fat distribution in these patients is unknown. It is possible that following weight gain, body fat distribution changed to its pre-morbid form, rather than being the direct outcome of nutrition intervention. C) Future Research The focus of the present study was to determine the body composition and body fat distribution changes that occur upon maximum weight gain after renourishment in A N patients. A question which remains unanswered is whether body composition and/or body fat distribution are altered as A N patients continue to maintain an increased body weight. Currently, the Eating Disorders Clinic at St. Paul's Hospital operates two day programs designed to provide outpatient therapy to those A N patients who have received inpatient treatment and then proceed to a weight-maintenance nutritional regime. These patients would be ideal for recruitment in such a study. As emphasized previously, there are a number of challenges associated with the field of body composition assessment, the primary one being the indirect nature of the currently used body composition methods. Similar challenges exist in the area of body fat distribution. Since various investigators have used different indices to assess fat distribution, no concensus has yet been reached on the best approach to assess fat distribution in relation to health and disease. It is imperative that future research directions in the area of fat distribution include 90 the validation of D E X A in various disease states, including A N . Such a development would thus clarify the association of fat distribution versus total body fatness in relation to disease risk. D) Recommendations for Practice Body image distortion and a morbid fear of becoming fat are two of the most common features of A N . In fact, weight and shape concerns have a central role in governing the anorexic patient's evaluation of her sense of self-esteem. Some patients express dissatisfaction towards their entire body, while others view selected body sites such as the stomach, hips and thighs as \"protruding\" and disproportionate to the rest of the body. Thus a central aspect of treatment is, through various strategies including nutrition intervention, to facilitate the patient in overcoming these misconceptions. The preliminary findings of the present research study could provide a framework for educating A N patients about the body composition and body fat distribution changes that accompany weight gain, and could help in alleviating the patients' concerns regarding these changes. For example, it would be beneficial for the anorexic individual to be forewarned that during the initial stages of weight restoration, the abdomen appears to be disproportionately large due to intestinal distension, but that this subsides as physical health improves. As well, patients could be reassured that despite weight gain the gynoid body fat distribution pattern is preserved, which in turn provides protection against the health risks associated with an android body shape. It could also be emphasized that the body composition patterns achieved following renourishment resemble those of a normal-weight female, and that this constitutes a foundation for recovery. Lastly, exploring the essential physiological functions of a gynoid body shape during various stages of the lifecycle, could help the anorexic patient to face her fears of gaining weight and encourage her to aspire towards a healthier body and an improved sense of self-worth. 91 REFERENCES American Psychiatric Association. (1987) Diagnostic and Statistical Manual of Mental Disorders (3rd. ed. revised). Washington, DC: Author. Aoki, T. T., Muller, W. A. , Brennan, M . R., & Cahill, G. F. Jr. (1975). The metabolic effects of glucose in brief and prolonged fasted man. American Journal of Clinical Nutrition, 28, 507-511. Armellini, F., Zamboni, M . , Robbi, R., Todesco, T., Rigo, L . , Bergamo-Andreis, I.A., & Bosello, O. (1993). Total and intra-abdominal fat measurements by ultrasound and computed tomography. International Journal of Obesity, 17, 209-214. Ashwell, M . , Cole, T. J., & Dixon, A . K. (1985). Obesity: new insight into the anthropometric classification of fat distribution shown by computed tomography. British Medical Journal, 290,1692-1694. Barrett-Connor, E . (1991). Nutrition epidemiology: how do we know what they ate? American Journal of Clinical Nutrition, 54, 182S-187S. Basiotis, P. P., Welsh, S. O., Cronin, F. J., Kelsay, J. L . , Mertz, W. (1987). Number of days of food records required to estimate individual and group nutrient intakes with defined confidence. Journal of Nutrition, 117, 1638-1641. Bedell, G. N., Marshall, R., Dubois, A. B., & Harris, J. H. (1956). Measurement of the volume of gas in the gastrointestinal tract. Values in normal subjects and ambulatory patients. Journal of Clinical Investigation, 35, 336-345. Behnke, A. R., Osserman, E. F., & Welham, W. E. (1953). Lean body mass. Archives of Internal Medicine, 91, 585-601. Beumont, P. J. V. , Arthur, B., Russell, J. D., & Touyz, S. W. (1994). Excessive physical activity: proposals for a supervised exercise program. International Journal of Eating Disorders, 15, 21-36. Beumont, P. J. V. , Chambers, T. L . , Rouse, L . , & Abraham, S. F. (1981). The diet composition and nutritional knowledge of patients with anorexia nervosa. Journal of Human Nutrition, 35, 265-273. Biller, B. M . , Saxe, V. , & Herzog, D. B. (1989). Mechanisms of osteoporosis in adult and adolescent women with anorexia nervosa. Journal of Clinical Endocrinology and Metabolism, 68, 548-551. Bland, J. M . , & Altman, D. G. (1986). Statistical methods for assessing agreement between two methods of clinical measurement. Lancet, 1, 307-310. Blendis, L . M . , & Crisp, A. H . (1968). Serum cholesterol in anorexia nervosa. Journal of Postgraduate Medicine, 44, 327-330. Block, G. (1982). A review of validations of dietary asessment methods. American Journal of Epidemiology, 115,492-505. 92 Brans, Y. W., Summers, J. E . , Dweck, H. S., & Cassidy, G. (1974). A non-invasive approach to body composition in the neonate: dynamic skinfold measurements. Pediatric Research, 8,215-222. Brodie, D. A. (1988). Techniques of measurement of body composition. Part I. Sports Medicine, 5, 11-40. Brozek, J., Grande, F., Anderson, J. T., & Keys, A. (1963). Densitometric analysis of body composition: revision of some quantitative assumptions. Annals of the New York Academy of Sciences, 110, 113-140. Brozek, J., Grande, F., Taylor, H . L . , Anderson, J. T., Buskirk, E . R., & Keys, A. (1957). Changes in body weight and body dimensions in men performing work on a low calorie carbohydrate diet. Journal of Applied Physiology, 10,412-420. Brozek JF, & Henschel A. (1963). Techniques of Measuring Body Composition. National Academy of Sciences, National Research Council, Washington, D.C. Brylowski L . M . Body composition analysis in the clinical setting. (1992). Frontiers in Clinical Nutrition 1, 1-7. Buskirk, E . R. (1987). The 1986 C.H. McCloy Research Lecture. Body composition analysis: the past, present and future. Research Quarterly for Exercise and Sport, 58,1-10. Cahill, G. F. Jr. (1970). Starvation in man. New England Journal of Medicine, 282, 668-675. Callaway, C. W., Chumlea, W. C , Bouchard, C , Himes, J. H. , Lohman, T. G., Martin, A. D., Mitchell, C. D., Mueller, W. H. , Roche, A. F., & Seefeldt, V. D. (1988). Circumferences. In T. G. Lohman, A. F. Roche, R. Martorell, (Eds.), Anthropometric Standardization Reference Manual, (pp. 39-54). Illinois: Human Kinetics Books. Casper, R. C. (1982). Treatment principles in anorexia nervosa. Adolescent Psychiatry, 10, 431-454. Chumlea, W. C , & Guo, S. S. (1994). Bioelectrical impedance and body composition: present status and future directions. Nutrition Reviews, 52,123-131. Cohn SH. (1985). How valid are bioelectrical impedance measurements in body composition studies? American Journal of Clinical Nutrition, 42, 889-890. Craig, A. B., & Ware, D. E . (1967). Effect of immersion in water on vital capacity and residual volume of the lungs. Journal of Applied Physiology, 23,423-425. Davis, R. (1988). Clinical Assessment of the Eating Disorders. In P. E . Garfinkel (Ed.), Anorexia Nervosa and Bulimia Nervosa: Current Update (pp. 11-18). Toronto: Upjohn. Dempsey, D. T., Crosby, L . O., Lusk, E . , Oberlander, J. L . , Pertshuk, M . J., & Muller, J. L . (1984). Total body water and total body potassium in anorexia nervosa. American Journal of Clinical Nutrition, 40, 260-269. Despres, J. P., Moorjani, S., Ferland, M . , Tremblay, A., Lupien, P. J., Nadeau, A., Pinault, S., Theriault, G. , & Bouchard, C. (1989). Adipose tissue distribution and plasma lipoprotein levels in obese women: importance of intra-abdominal fat. Arteriosclerosis, 9,203-210. 93 Durnin, J. V. G. A., & Womersley, J. (1974). Body fat assessed from total body density and its estimation from skinfold thicknesses: measurements on 481 men and women aged 16 to 72 years. British Journal of Nutrition, 32, 77-97. Epling, W. F., & Pierce, W. D. (1991). Solving the anorexia puzzle. A scientific approach. Toronto: Hogrefe and Huber. Epling, W. F., Pierce, W. D., & Stefan, L . (1983). A theory of activity based anorexia. International Journal of Eating Disorders, 3, 7-46. Evans, D. J., Hoffman, R. G. , Kalkhoff, R. K. , & Kissebah, A. H. (1984). Relationship of body fat topography to insulin sensitivity and metabolic profiles in premenopausal women. Metabolism, 33, 68-75. Fairbura, C. G. , & Cooper, P. J. (1989). Eating disorders. In K. Hawton, P. M . Salkovskis, J. Kirk, & D. M . Clark (Eds.), Cognitive Behaviour Therapy for Psychiatric Problems: A Practical Guide (pp. 277-314). Oxford University Press. Falk, J. R., Halmi, K. A., & Tryon, W. W. (1985). Activity measures in anorexia nervosa. Archives of General Psychiatry, 42, 811-814. Ferland, M . , Despres, J. P., Tremblay, A., Pinault, S., Nadeau, A., Moorjani, S., Lupien, P. J., Theriault, G. , & Bouchard, C. (1989). Assessment of adipose tissue distribution by computed tomography in obese women: association with body density and anthropometric measurements. British Journal of Nutrition, 48, 1351-1361. Fidanza, F. (1980). Effects of starvation on body composition. American Journal of Clinical Nutrition, 33, 1562-1566. Fohlin L . (1978). Exercise performance and body dimensions in anorexia nervosa before and after rehabilitation. ActaMedica Scandinavica, 204, 61-65. Forbes, G. B. (1990). The abdomemhip ratio normative data and observations on selected patients. International Journal of Obesity, 14,149-157. Forbes, G. B., Gallup, J., & Hursh, J. B. (1961). Estimation of total body fat from potassium-40 content. Science, 133,101-102. Forbes, G. B., Kreipe, R. E. , Lipinski, B. A., & Hodgman, C. H. (1984). Body composition changes during recovery from anorexia nervosa: comparison of two dietary regimes. American Journal of Clinical Nutrition, 40, 1137-1145. Frisancho, A. R., Flegel, P. N. (1983). Elbow breadth as a measure of frame size for US males and females. American Journal of Clinical Nutrition, 37, 311-314. Frisch, R. E . , & McArthur, J. (1974). Menstrual cycles: fatness as a determinant of minimum weight for height necessary for their maintenance or onset. Science, 185,949-951. Garfinkel, P. E . , & Garner, D. M . (1982). Anorexia Nervosa: A Multi-Dimensional Perspective . New York: Brunner/Mazel. Garfinkel PE, Goldbloom DS. (1988). Anorexia nervosa and bulimia nervosa - introduction. In Anorexia Nervosa and Bulimia Nervosa: Current Update. Toronto: Upjohn eds, p. 3-9. 94 Garn, S. M . , Sullivan, T. V. , & Hawthorne, V. M . (1987). Differential rates of fat change relative to weight change at different body sites. International Journal of Obesity, 11, 519-525. Garner, D. M . (1985). Iatrogenesis in anorexia nervosa and bulimia nervosa. International Journal of Eating Disorders, 4, 701-726. Gersovitz, M . , Madden, J. P., & Smiciklas-Wright, H . (1978). Validity of the 24-hr. dietary recall and seven-day record for group comparisons. Journal of the American Dietetic Association, 73, 48-55. Gibson, R. S. (1991). Principles of Nutritional Assessment, New York: Oxford University Press. Going, S. B., Massett, M . P., Hall, M . C , Bare, L . A. , Root, P. A. , Williams, D. P., & Lohman, T. G. (1993). Detection of small changes in body composition by dual-energy-X-ray absorptiometry. American Journal of Clinical Nutrition, 57, 845-850. Grauer, W. O , Moss, A. A., Cann, C. E . , & Goldberg, H. I. (1984). Quantification of body fat distribution in the abdomen using computed tomography. American Journal of Clinical Nutrition, 39, 631-637. Gray, D. S., Bray, G. A., Bauer, M . , Kaplan, K., Gemayel, N., Wood, R., Greenway, F., & Kirk, S. (1990). Skinfold thickness measurements in obese subjects. American Journal of Clinical Nutrition, 51, 571-577. Gurr, M . I, & Harwood, J. L . (1991). Lipid Biochemistry 4th ed. London: Chapman and Hall. Hannan, W. J., Cowen, S., Freeman, C. P., Mackie, A., & Shapiro, C. M . (1990). Assessment of body composition in anorexic patients. In S. Yasamura, J. E . Harrison, K G. Mc Neill, A. D. Woodhead, F. A. Dilmanian (Eds), In Vivo Body Composition Studies. Recent Advances (pp. 149-154). New York: Plenum Press. Hannan, W. J., Cowen, S., Freeman, C. P., & Shapiro, C. M . (1990). Evaluation of bioelectrical impedance analysis for body composition measurements in anorexia nervosa. Clinical Physics and Physiological Measurement, 11,209-216. Hannan, W. J., Cowen, S., Freeman, C. P., & Wrate, R. M . (1993a). Can bioelectrical impedance improve the prediction of body fat in patients with eating disorders? European Journal of Clinical Nutrition, 47, 741-746. Hannan, W. J., Cowen, S. J., Freeman, C. P., Wrate, R. M . , & Barton, J. (1993b). Evaluation of dual energy X-ray absorptiometry for the assessment of body composition in anorexic females. In K. J. Ellis, J. D. Eastman (Eds), Body Composition. In Vivo Methods, Models and Assessment (pp. 169-172). New York: Plenum Press. Hansen, N. J., Lohman, T. G. , Going, S. B., Hall, M . C , Pamenter, R. W., Bare, L . A., Boyden, T. W., & Houtkooper, L B. (1993). Prediction of body composition in premenopausal females from dual-energy X-ray absorptiometry. Journal of Applied Physiology, 75, 1637-41. 95 Harrison, G. G. , Buskirk, E. R., Carter, J. E . L . , Johnston, F. E. , Lohman, T. G., Pollock, M . L . , Roche, A. F., & Wilmore, J. (1988). Skinfold thickness and measurement technique. In T. G. Lohman, A. F. Roche, R. Martorell, (Eds.), Anthropometric Standardization Reference Manual, (pp. 55-70). Illinois: Human Kinetics Books. Hart, P. D., Wilkie, M . E . , Edwards, A. , & Cunningham, J. (1993). Dual energy X-ray absorptiometry versus skinfold measurements in the assessment of total body fat in renal transplant recipients. European Journal of Clinical Nutrition, 47, 347-352. Heymsfield, S. B., Wang, J., Heshka, S., Kehayias, J. J., & Pierson, R. N. (1989). Dual photon absorptiometry: comparison of bone mineral and soft tissue mass measurements in vivo with established methods. American Journal of Clinical Nutrition, 49,1283-1289. Himes J. H. (1988). Alteration in distribution of adipose tissue in response to nutritional intervention. In C. Bouchard, & F. E . Johnston, (Eds.), Fat Distribution During Growth and Later Health Outcomes (pp. 313-332). New York: Alan R. Liss, Inc. Jackson, A. S., & Pollock, M . L . (1978). Generalized equations for predicting body density of men. British Journal of Nutrition, 40, 497-504. Jackson, A. S., Pollock, M . L. , & Ward, A. (1980). Generalized equations for predicting body density of women. Medicine and Science in Sports and Exercise, 12,175-182. Jeuniewic, N., Brown, G. ML , Garfinkel, P. E . , & Moldofsky, H . (1978). Hypothalamic function as related to body weight and body fat in anorexia nervosa. Psychosomatic Medicine, 40, 187-198. Johansson, A. G. , Forslund, A. , Sjodin, A. , Mallin, H . , Hambraeus L . , & Ljunghall, S. (1993). Determination of body composition - a comparison of dual energy X-ray absorptiometry and hydrodensitometry. American Journal of Clinical Nutrition, 57, 323-Kaplan A. S. (1990). Biomedical variables in the eating disorders. Canadian Journal of Psychiatry, 35,745-753. Kaye, W. H . , Gwirtsman, H . E . , Obarzanek, E . , & George, D. T. (1988). Relative importance of calorie intake needed to gain weight and level of physical activity in anorexia nervosa. American Journal of Clinical Nutrition, 47,989-994. Kennedy, S. H. , & Abbas, A. (1993). Inpatient treatment for anorexia nervosa and bulimia nervosa. In S. H. Kennedy (Ed.), University of Toronto Handbook of Eating Disorders (pp. 30-36). Toronto: University of Toronto Press. Keys, A., Brozek, J., Henschel, A., Mickelsen, O., & Taylor, H. L . (1950). The Biology of Human Starvation.. Minneapolis: University of Minnesota Press. King, M . (1991). Update: Eating disorders. Comprehensive Therapy, 17,35-40. Klibanski, A. , & Greenspan, S. L . (1986). Increase in bone mass after treatment of hyperprolactinemic amenorrhea. New England Journal of Medicine, 315, 542-546. Krahn, D. D., Rock, C , Dechert, R. E . , Nairn, K. K., & Hasse, S. A. (1993). Changes in resting energy expenditure and body composition in anorexia nervosa patients during refeeding. Journal of the American Dietetic Association, 93, 434-438. 96 Kron, L . , Katz, J. L . , Gorzynski, G. , & Weiner, H . (1987). Hyperactivity in anorexia nervosa: a fundamental clinical feature. Comprehensive Psychiatry, 19,433-440. Kuczmarski, R. J., Fanelli, M . T., & Koch, G. G. (1987). Ultrasonic assessment of body composition in obese adults: overcoming the limitations of the skinfold caliper. American Journal of Clinical Nutrition, 45, 717-724. Kushner, R. F. (1992). Bioelectrical impedance analysis: a review of principles and applications. Journal of the American College of Nutrition, 11, 199-209. Lapidus, L . , Bengsston, C., Larson, B., Pennert, K., Rybo, E . , & Sjostrom, L . (1984). Distribution of adipose tissue and risk of cardiovascular disease and death: a 12 year follow-up of the participants in the population study of women in Gothenberg, Sweden. British Medical Journal, 289, 1257-1261. Larsson, B., Svardsudd, K., Welin, L . , Wilhehnsen, L . , Bjorntorp, P., & Tibblin, G. (1984). Abdominal adipose tissue distribution, obesity, and risk of cardiovascular disease and death: 13 year follow-up of participants in the study of men born in 1913. British Medical Journal, 288, 1401-1404. Ley, C. J., Lees, B., & Stevenson, J. C. (1992). Sex- and menopause-associated changes in body-fat distribution. American Journal of Clinical Nutrition, 55, 950-954 Lohman T. G. (1992). Dual energy radiography: total body and regional composition. In Advances in Body Composition Assessment. Current Issues in Exercise Science (pp. 25-36). Monogram Number 3, Illinois: Human Kinetics Publishers. Lohman T. G. (1981). Skinfolds and body density and their relation to body fatness: a review. Human Biology, 53,181-225. Lukaski H . C. (1990). Applications of bioelectrical impedance analysis: a critical review. In S. Yasamura, J. E . Harrison, K G. Mc Neill, A. D. Woodhead, F. A. Dilmanian (Eds.), In Vivo Body Composition Studies. Recent Advances (pp. 365-373). New York: Plenum Press. Lukaski, H . C., Bolonchuk, W. W., Hall, C. B., & Siders, W. A. (1986). Validation of tetrapolar bioelectrical impedance method to assess human body composition. Journal of Applied Physiology, 60, 1327-1332. Lukaski, H. C., Johnson, P. E . , Bolonchuk, W. W., & Lykken, G. I. (1985). Assessment of fat-free mass using bioelectrical impedance measurements of the human body. American Journal of Clinical Nutrition, 41, 810-817. Martin, A. D., Ross, W. D, Drinkwater, D. T., & Clarys, J. P. (1985). Prediction of body fat by skinfold caliper: assumptions and cadaver evidence. International Journal of Obesity, 9S, 31-39. Martin, A. D., & Drinkwater, D. T. (1991). Variability in the measures of body fat. Sports Medicine, 11,277-288. Mayo-Smith, W., Hayes, C. W., Biller, B. M . K., Klibanski, A., Rosenthal, H., & Rosenthal, D. I. (1989). Body fat distribution measured with CT: correlations in healthy subjects, patients with anorexia nervosa, and patients with cushing syndrome. Radiology, 170, 515-97 Mazess, R. B., Barden, H. S., Bisek, J. P., & Hanson, J. (1990a). Dual energy X-ray absorptiometry for total body and regional bone-mineral and soft-tissue composition. American Journal of Clinical Nutrition, 51,1106-1112. Mazess, R. B., Barden, H. S., & Ohlrich, E . S. (1990b). Skeletal and body-composition effects of anorexia nervosa. American Journal Clinical Nutrition, 52,438-441. McArdle W. D, Katch F. I, Katch V. L . (1991). Body composition, energy balance and weight control. In Exercise Physiology. Energy, Nutrition and Human Performance, 3rd ed (pp. 599-633). Philadelphia: Lea and Febiger. McCance, R. A. , & Widdowson, E . M . (1951). A method of breaking down the body weights of living persons into terms of extracellular fluid, cell mass and fat, and some applications of it to physiology and medicine. Proceedings of the Royal Society of Biology, 1 oo 1 t c i o n Melchior, J. C , Rigaud, D., Rozen, R., Malon, D., & Apfelbaum, M . (1989). Energy expenditure economy induced by decrease in lean body mass in anorexia nervosa. European Journal of Clinical Nutrition, 43, 793-799. Mitchell, P. B., & Truswell, S. Body composition in anorexia nervosa and starvation. (1987). In P. J. V. Beumont, G. D. Burrows, & R. C. Casper (Eds.), Handbook of Eating Disorders Part 1: Anorexia and Bulimia Nervosa (pp. 45-77). New York: Elsevier Science Publishing. Moore, F. D., McMurray, J. D., Parker, H . V., Magnus, T. C. (1956). Symposium: water and electrolytes: body composition: total body water and electrolytes: intravascular and extravascular phase volumes. Metabolism, 5,447-467. Moore, F. D. (1963). The Body Cell Mass and its Supporting Environment. Body Composition in Health and Disease. Philadelphia: WB Saunders. Nord, R. H . , & Payne, R. K. (1994). Body composition by D X A - A review of the technology. Asia Pacific Journal of Clinical Nutrition, 3 (suppl), 1-6. Norland Corporation. (1992). XR-Series Bone Densitometer Operator's Guide. Fort Atkinson, Wisconsin. Ohlson, L . O., Larsson, B., & Svardsudd, K. (1985). The influence of body fat distribution on the incidence of diabetes mellitus: a 13.5 year follow-up of the participants in the study of men bom in 1913. Diabetes, 34,1055-1058. Orphanidou, C , McCargar, L . , Birmingham, C. L . , Mathieson, J., & Goldner, E . (1994). Accuracy of subcutaneous fat measurement: Comparison of skinfold calipers, ultrasound, and computed tomography. Journal of the American Dietetic Association, 94, 855-858. Patterson, R. (1989). Body fluid determinations using multiple impedance measurements. Engineering in Medicine and Biology, 8, 16-18. Peiris, A. N., Mueller, R. A. , Struve, M . F., Smith, G. A. , & Kissebah, A. H. (1987). Relationship of androgenic activity to splanchnic insulin metabolism and peripheral glucose utilization in premenopausal women. Journal of Clinical Endocrinology and Metabolism, 64, 162-169. 98 Pirke, K. M . , Pahl, J., Schweiger, U. , Munzing, W., Lang, P., & Bull, U. (1986). Total body potassium, intracellular potassium and body composition in patients with anorexia nervosa during refeeding. International Journal of Eating Disorders, 5, 347-354. Potosky, A. L . , Block, G. , & Hartman, A. M . (1990). The apparent validity of diet questionnaires is influenced by number of diet-record days used for comparison. Journal of the American Dietetic Association, 90, 810-813. Pritchard, J. E . , Nowson, C. A., Strauss, B. J., Carlson, J. S., Kaymakci, B., & Wark, J. D. (1993). Evaluation of dual energy X-ray absorptiometry as a method of measurement of body fat. European Journal of Clinical Nutrition, 47,216-228. Rammohan, M . , & Aplasca, E . C. (1992). Caliper method vs bioelectrical impedance analysis for determining body fat in patients undergoing chronic dialysis and in healthy individuals. Journal of the American Dietetic Association, 92, 1395-1397. Reid D. (1990). Folate and zinc status of chronic hemodialysis patients. MSc Thesis. University of British Columbia. Reiff, D. W., Reiff, K. K. (1992). Eating disorders. Nutrition therapy in the recovery process (pp. 236-242). Maryland: Aspen Publishers. Rico, H. , Revilla, M . , Villa, L . F., Alvarez del Buergo, M . , & Ruiz-Conteras, D. (1994). Determinants of total-body and regional bone mineral content and density in postpubertal normal women. Metabolism: Clinical & Experimental, 43,417-422. Rigotti, N. A. , Neer, R. M . , Skates, S. J., Herzog, D. B., & Nussbaum, S. R. (1991). The clinical course of osteoporosis in anorexia nervosa. A longitudinal study of cortical bone mass. Journal of the American Medical Association, 265, 1133-1138. Robert, S., Zarowitz, B. J., Hyzy, R., Eichenhorn, M . , Peterson, E . L . , & Popovich, J., Jr. (1993). Bioelectrical impedance assessment of nutritional status in critically ill patients. American Journal of Clinical Nutrition, 57, 840-844. Rock, C. L . , & Curran-Celentano, J. C. (1994). Nutritional disorder of anorexia nervosa: a review. International Journal of Eating Disorders, 15,187-203. Ross, R., Leger, L . , Marliss, E . B., Morris, D. R., & Gougeon, R. (1991). Adipose tissue distribution changes during weight loss in obese adults. International Journal of Obesity, 15, Roubenoff, R., Kehayias, J: J. (1991). The meaning and measurement of lean body mass. Nutrition Reviews, 46, 163-175. Roubenoff, R., Kehayias, J. J., Dawson-Hughes, B., & Heymsfield, S. B. (1993). Use of dual-energy X-ray absorptiometry in body composition studies: not yet a \"gold standard.\" American Journal of Clinical Nutrition, 58, 589-591. Royall, D., Greenberg, G. R., Allard, J. P., Baker, P., Harrison, J. E . , & Jeejeebhoy, K. N. (1994). Critical assessment of body-composition measurements in malnourished subjects with Crohn's disease: the role of bioelectric impedance analysis. American Journal of Clinical Nutrition, 59, 325-330. 99 Russell, D. McR., Prendergast, P. J., Darby, P. L . , Garfinkel, P. E . , Whitewell, J. , & Jeejeebhoy, K. N. (1983). A comparison between muscle function and body composition in anorexia nervosa: the effect of refeeding. American Journal of Clinical Nutrition, 38, 229-237. Russell, G. F. M . , & Mezey, A. G. (1962). An analysis of weight gain in patients with anorexia nervosa treated with high calorie diets. Clinical Science, 23,449-461. Russell, J., & Beumont, P. J. V . (1987) The endocrinology of anorexia nervosa. In P. J. V. Beumont, G. D. Burrows, & R. C. Casper (Eds.), Handbook of Eating Disorders Part 1: Anorexia and Bulimia Nervosa (pp. 201-232). New York: Elsevier Science Publishing. Russell, J. D., Mira, M . , Allen, B. J., Stewart, P. M . , Vizzard, J., Arthur, B., & Beumont, P. J. V. (1994). Protein repletion and treatment in anorexia nervosa. American Journal of Clinical Nutrition, 59, 98-102. Ryde, S. J. S., Laskey, M . A. , Morgan, W. D., & Compston, J. E . (1993). A comparison between neutron activation and dual energy X-ray absorptiometry for the measurement of body fat. In K. J. Ellis, & J. D. Eastman (Eds.), Body Composition. In Vivo Methods, Models and Assessment (pp. 367-370). New York: Plenum Press. Salisbury, J. J., Mitchell, J. E . (1991). Bone mineral density and anorexia nervosa in women. American Journal of Psychiatry, 148, 768-774. Sawas, M . , Treasure. J., Studd, J., Fogelman, T., Moniz, C. & Brincat, M . (1989). The effect of anorexia nervosa on skin thickness, skin collagen and bone density. British Journal of Obstetrics and Gynecology, 96,1392-1394. Schoeller, D. A. (1990). How accurate is self-reported dietary energy intake? Nutrition Reviews, 48, 373-379. Seeman, E . , Szmukler, G. I., Formica, C , Tsalamandris, C , & Mestrovic, R. (1992). Osteoporosis in Anorexia Nervosa: The influence of peak bone density, bone loss, oral contraceptive use, and exercise. Journal of Bone and Mineral Research, 7,1467-1474. Shephard, R. J. (1991). Estimation of bone mass. In Body Composition in Biological Anthropology (pp. 108-114). Cambridge: Cambridge University Press. Siri, W. E . (1956). The gross composition of the body. Advances in Biological and Medical Physics, 4,239-280. Siri, W. E . (1961). Body composition from fluid spaces and density: Analysis of methods. In Techniques for Measuring Body Composition (pp. 223-244). Washington, DC: National Academy of Science - National Research Council. Sjostrom, L . , Kvist, H. , Cederblad, A., & Tylen, U. (1986). Determination of total adipose tissue and body fat in women by CT, 4 0 K , and tritium. American Journal of Physiology, 250, 736-745. Sjostrom, L. Measurement of fat distribution. (1988). In Fat Distribution During Growth and Later Health Outcomes (pp. 43-61). New York: Alan R. Liss Inc. 100 Snyder, W. S., Cook, M . J., Nasset, E . S., Karhausen, L . R., Howells, G. P., & Tipton, I. H . (1984). Report of The Task Group on Reference Man.. Oxford: Pergamon Press. Svendsen, O. L . , Hassager, C , Bergmann, I., & Christiansen, C. (1993). Measurement of abdominal and intra-abdominal fat in postmenopausal women by dual energy X-ray absorptiometry and anthropometry: comparison with computerized tomography. International Journal of Obesity, 17,45-51. Terry, R. B., Stefanick, M . L . , Haskell, W. L . , & Wood, P. D. (1991). Contributions of regional adipose tissue depots to plasma lipoprotein concentrations in overweight men and women: possible protective effects of thigh fat. Metabolism, 40, 733-740. Thompson, R. A. , & Trattner-Sherman, R. (1993). Helping Athletes with Eating Disorders. Champaign II: Human Kinetics Publisher. Twyman, D. L . , & Liedtke, R. J. (1987). Bioelectrical Impedance Analysis of Body Composition Assessment Techniques. Detroit MI, RJL Systems. Vaisman, N. , Rossi, M . F., Goldberg, E . , Dibden, L . J., Wykes, L . J., & Pencharz, P. (1988a). Energy expenditure and body composition in patients with anorexia nervosa. Journal of Pediatrics, 113,919-924. Vaisman, N., Corey, M . , Rossi, M . F., Goldberg, E. , & Pencharz, P. (1988b). Changes in body composition during refeeding of patients with anorexia nervosa. Journal of Pediatrics, 113,925-929. van der Koy, K., & Seidell, J. C. (1993). Techniques for the measurement of visceral fat: a practical guide. International Journal of Obesity, 17, 187-196. Van Loan M D . (1990). Bioelectrical impedance analysis to determine fat free mass, total body water and body fat. Sports Medicine, 10, 205-217. Vansant, G. , Den Besten, C , Westrate, J., & Deurenberg, P. (1988). Body fat distribution and prognosis for weight reduction: preliminary observations. International Journal of Obesity 12,133-140. Wadden, T. A, Stunkard, A. J., Johnston, F. E , Wang, J., Pierson, R. N, Van Itallie, T. B., Costello, E . , & Pena, M . (1988). Body fat deposition in adult obese women II - Changes in fat distribution accompanying weight reduction. American Journal of Clinical Nutrition, 47, Wang, J., Heymsfield S. B., Aulet, M . , Thornton, J. C , & Pierson, R. N. (1992a). Body fat from body density: underwater weighing vs. dual-photon absorptiometry. American Journal of Physiology, 256, E829-E834. Wang, J., Koetler, D. P., Russell, M . , Burastero, S., Mazariegos, M . , Thornton, J., Dilmanian, F. A. , & Pierson, R. N. Jr. (1992b). Body-fat measurement in patients with acquired immunodeficiency syndrome: which method should be used? American Journal of Clinical Nutrition, 56,963-967. Wang, Z. M , Pierson, R. N., & Heymsfield, S. B. (1992). The five-level model: a new approach to organizing body composition research. American Journal of Clinical Nutrition, 101 Williams, D. P., Boyden, T. W., Pamenter, R. W., Lohman, T. G. , & Going, S. B. (1993). Relationship of body fat percentage and fat distribution with dehydroepiandrosterone sulfate in premenopausal females. Journal of Clinical Endocrinology & Metabolism, 11, 80-85. Williams, E . (1958). Anorexia nervosa: a somatic disorder. British Medical Journal, ii, 190-194. Williams, H. , Touyz, S. W., Beumont, P. J. V. (1985). Nutritional counselling in anorexia nervosa. In S. W. Touyz, & P. J. V. Beumont (Eds.), Eating Disorders: Prevalence and Treatment (pp. 23-31). Balgowlah: Williams & Wilkins. Wilmore, J. H . , Frisancho, R. A. , Gordon, C. C., Ffimes, J. FL, Martin, A. D., Martorell, R., & Seefeldt, V. D. (1988). Body breadth equipment and measurement techniques. In T. G. Lohman, A. F. Roche, R. Martorell, (Eds.), Anthropometric Standardization Reference Manual, (pp. 36-38). Illinois: Human Kinetics Books. Woodside, D. B., & Garfinkel, P. E . (1989). An overview of the eating disorders anorexia nervosa and bulimia nervosa. National Institute of Nutrition Review, 8,1-4. Young, V. R., & Scrimshaw, N. S. (1971). The physiology of starvation. Scientific American, 225, 14-21. 102 APPENDICES 103 APPENDIX 1 PILOT STUDY PUBLICATION 1 0 4 Reprinted from JOURNAL OF THE AMERICAN DIETETIC ASSOCIATION, Vol. 94, No 8, Augutl I W » Copyright The American Dietetic Asportation, Printed In U.S.A. RESEARCH Accuracy of subcutaneous fat measurement: Comparison of skinfold calipers, ultrasound, and computed tomography CHARITINIORPHANIDOU, RDN; LINDA McCARGAR, PhD, RDN; C. LAIRD BIRMINGHAM, MD; JOHNMATHIESON, MD; ELLIOT GOLDNER, MD ABSTRACT Objective The purpose of this study was to compare skinfold caliper and ultrasound measurement of subcutaneous body fat at three abdominal sites with computed tomography, which is considered to be the gold standard. Design This was a cross-sectional study in which computed-tomography, ultrasound, and skinfold caliper measurements were made at three distinct abdominal sites. All body composi-tion and anthropometric measurements were performed on each subject on one occasion. Subjects Twenty-two subjects were recruited (13 men and 9 women). Mean ages (±standard deviation) were 43±4 years for the women and 51±18 years for the men. All subjects had been previously scheduled for an abdominal or pelvic computed-tomography scan at the Department of Radiology, St Paul's Hospital, Vancouver, British Columbia, Canada, and partici-pated in the study on a volunteer basis. Main outcome measures A better agreement was found between the skinfold calipers and computed-tomography methods than between the ultrasound and computed-tomography method for the measurement of subcutaneous body fat. This was observed when the data were analyzed for both correlational agreement and for graphical interpretation. Statistical analyses performed The relationships among skinfold, ultrasound, and computed-tomography measure-ments were analyzed by determining Pearson correlation coefficients. A graphical method described by Bland and Altman was also used to assess agreement among the three methods. Results Significant correlation coefficients were observed between skinfold calipers and computed tomography at all three abdominal sites (site 1, r=.60, P=.003; site 2, r=.70, P=.0001; site 3, r-=.73, P=.0001). Ultrasound and computed-tomography methods only showed a significant correlation at site 3 (r=.54; P=.009). The graphical method revealed that the variation in the ultrasound measurements was much greater than that of the skinfold measurements when compared to computed-tomography values. Applications/conclusions The results of this study indicated that relative agreement in the measurement of subcutaneous body fat between skinfold and computed-tomography measure-ments was superior to that exhibited between ultrasound and computed-tomography measurements. This finding enhances the potential use of skinfold calipers in the clinical setting, particularly in view of the fact that measurement of subcutane-ous body fat at different body sites is becoming increasingly important for the characterization of risk of certain disease states. J Am Diet Assoc. 1994;94:855-858. L. McCargar is an assistant professor and C. Orphanidou is a master of science degree candidate in the School of Family and Nutritional Sciences, University of British Columbia, Vancouver, British Columbia, Canada. C. L. Birmingham is with the Department of Internal Medicine, J. Mathieson is with the Department of Radiology, and E. Goldner is with the Department of Psychiatry, St Paul's Hospital, Vancouver, British Columbia, Canada. Address correspondence to: L. McCargar, PhD, RDN School of Family and Nutritional Sciences, 2205 East Mall, University of British Columbia, Vancouver, British Columbia, Canada V6T1Z4. 105 M easurement of subcutaneous fat thickness at distinct body sites is an important component in the assessment of regional fat distribution. Investigators generally agree that disproportionate fat distribution plays a major role in the development of certain metabolic disorders and can ulti-mately influence morbidity and mortality (1,2). More specifically, android (upper body) fat distribution, most commonly observed in men, has been strongly associated with diabetes mellitus, hypertriglyceridemia, and hypertension (3-5). In contrast, these metabolic aberrations have shown little association with gynoid (lower body) fat distribution, which is more common in women (6). JOURNAL OF THE AMERICAN DIETETIC ASSOCIATION / 855 RESEARCH Table Characteristics of study subjects Subjects Age Weight Height BMI* WHR\" (y) (kg) (cm) « mean±SDc > Men(n = 13) 51 ±18 73.4 + 9.9 174.0 + 6.1 24.3±3.8 0.96 + 0.06 Women(n = 9) 43 + 14 59.9+10.2 164.2±5.7 22.1 +2.6 0.82 + 0.06 aBMI = body mass index (weight [kg]/height [m2]). °WHR = waist to hip ratio (cm/cm). Waist circumference was measured at the level of the umbilicus. Hip circumference was measured at the level of the maximum posterior extension of the buttocks. CSD = standard deviation. Traditionally, skinfold calipers have been the most widely used tool for evaluating body fat (7). They provide a fast, inexpensive, and convenient way of measuring subcutaneous fat in field situ-ations (8). Some of the problems associated with this technique include selecting an appropriate site, accurately defining the area to be measured, and having sufficient experience to obtain repro-ducible and reliable results (7). Furthermore, a number of as-sumptions have been questioned regarding the use of skinfolds, including the constancy of skinfold compressibility and skin thickness and adipose tissue fat content (9). Recently, there has been increased interest in the use of radiologic techniques for assessing body composition. Adaptation of contemporary instrumentation was deemed necessary to over-come some of the limitations of skinfold calipers. Computed tomography, which has the unique feature of differentiating between subcutaneous and visceral fat, is considered to be the reference method for quantifying body-fat depots, particularly in the abdominal area (10-12). However, the dose of radiation and the high cost limit the use of computed tomography in field surveys and routine clinical assessments (13,14). The usefulness of ultrasonography for measurement of body composition has also been investigated (8,12,14-19). Ultrasound overcomes some of the limitations of computed tomography in that it is portable, it does not emit radiation, and it is less expensive (14). The purpose of this investigation was to compare skinfold caliper and ultrasound measurements of subcutaneous body fat at three abdominal sites with computed-tomography measurements. We assumed that computed tomography represented the \"gold standard.\" METHODS Subjects Twenty-two subjects (13 men and 9 women), 24 to 81 years of age, volunteered to participate in the study. A i l subjects were recruited from a group of patients who had been scheduled for an abdominal or pelvic scan by computed tomography for diagnostic purposes in the Department of Radiology, St Paul's Hospital, Vancouver, British Columbia, Canada. The study was approved by the human ethics committees of St Paul's Hospital and the University of British Columbia. Informed consent was obtained from all subjects before their participation. Procedures Body composition and anthropometric measurements were taken for each subject in the following sequence: fat measurement by computed tomography, ultrasound, and skinfold calipers; weight; height; and waist to hip ratio. At the time of measurement, all subjects were dressed in hospital gowns. A l l measurements for each individual were done at the same time, on the same day, within a 1-hour period. The calibration of all equipment, was checked on a regular basis. The abdominal sire is difficult, to measure by ultrasound; nevertheless, the location was of specific interest. This was a pilot study to determine whether measurement with skinfold calipers and/or ultrasound was a reliable technique to use in a subsequent study to investigate body-fat deposition during weight gain in patients with anorexia nervosa. Measurement sites were marked with an indelible marker on the patient's skin, just before the computed-tomography scan was performed. A smaU, radiopaque grid marker was affixed to the skin to allow the sites to be identified. Al l measurement sites chosen were 5 cm to the left of midline, at 6 cm (site 1), 9 cm (site 2), and 12 cm (site 3) caudal to the inferior tip of the xiphoid process. All the computed-tomography scans were performed on a modern high-resolution scanner (spatial resolution range=0.1 mm) (GE Highlight Advantage, GE Medical Systems, Milwaukee, Wis) and were made directly from the operator's console by one of the authors, a radiologist (J.M.). Skinfold calipers seem to be the desirable method for measuring subcutaneous fat in clinical settings; the method is accurate, quick, easy to perform, noninvasive, and inexpensive The patients were then transferred to a separate room where ultrasound measurements were taken at the identical sites by one of two trained ultrasound technicians. A l l the measurements were made with a commercially available, high-resolution, real-time ultrasound unit (Aloka 650, Aloka Overseas Corporation, Tokyo, Japan) using a 5-MHz probe. The transducers were applied to the skin and then gradually withdrawn, and the ultrasound measure-ments were made with the least compression possible. Care was taken so that the transducer was perpendicular to the skin surface to avoid off-axis measurement errors. Both ultrasound and com-puted-tomography measurements were rounded off to the near-est millimeter. Skinfold measurements were obtained by means of Harpenden calipers (Creative Health Products, Plymouth, Mich). Three mea-surements were taken at each site while the subject was standing. Half of the double fold of skin and subcutaneous fat was recorded. The mean value of the two measurements with the smallest difference was then calculated. All measurements were made by the same experienced investigator (CO. ) . Weight, height, and waist to hip ratio were measured to further describe the population. The waist circumference was measured at the level of the umbilicus, and the hip circumference was measured at the level of the maximum posterior extension of the buttocks (20). Statistical Analysis A sample-size calculation revealed that a sample of 22 would have suitable statistical power. The relationships among skinfold, ultra-sound, and computed-tomography measurements were analyzed 856 / AUGUST 1994 VOLUME 94 NUMBER 8 106 by determining Pearson correlation coefficients (Systat, statistical software, version 4.1,1989, Systat Corporation, Evanston, 111). In addition, a graphical method described by Bland and Altman (21) was used to assess agreement between different methods. RESULTS Measurements of subcutaneous fat using the three different methods were completed on 22 subjects (13 men and 9 women). Subject characteristics are shown in the Table. Comparison of the methods by Pearson correlation coefficients showed significant correlations at site 1 for skinfold 1 vs ultrasound 1 (r=.59;P=.004) and skinfold 1 vs computed tomography 1 (r=.60; P=.003), but not for ultrasound 1 vs computed tomography 1. Significant correlations at site 2 were observed for skinfold 2 vs ultrasound 2 (r= . 66; P=. 001) and skinfold 2 vs computed tomography 2 (r=.70; P=.0001), but not for ultrasound 2 vs computed tomography 2. At site 3, significant correlations were observed for all measure-ments: skinfold 3 vs ultrasound 3 (r=.64; P=.001), skinfold 3 vs computed tomography 3 (r=.73; P=.0001), and ultrasound 3 vs computed tomography 3 (r=.54; P=.009). When the data for men and women were analyzed separately, of the nine possible comparisons (three measurements at three sites), eight of nine comparisons showed strong significant corre-lations for the men. In the women, only three of nine comparisons showed significant correlations (P<.05) (skinfold vs computed tomography at all three sites). The Figure shows the mean differences at site 2 between computed-tomography and skinfold measurements and between computed-tomography and ultrasound measurements plotted against the absolute values for computed tomography. The Figure shows ±2 standard deviations of the mean differences. The variation in the ultrasound measurements is much greater than that of the skinfold measurements compared with computed-tomography values. DISCUSSION In this study, for the measurement of subcutaneous body fat skinfold measurements showed better agreement with the results from computed tomography than did ultrasound: This was the case whether the data were analyzed for correlational agreement or graphical interpretation (21). Because of the limitations of using skinfold calipers (measure-ment error, compressibility) or computed tomography (radiation, expense), we anticipated that ultrasound might be a suitable alternative measurement tool to avoid these problems. Kuczmarski et al (14) found that body fat can be estimated with nearly the same degree of accuracy using either ultrasound or skinfold calipers. Other studies (13,18) also suggest that ultrasound is a reliable method for measurement of fat depots. This conclusion, however, is not supported by our study. Several sources of measurement error are possible with ultra-sound. The variability of ultrasound measurement of subcutane-ous fat is primarily attributable to direct compression of the subcutaneous fat by the pressure of the overlying transducer. Even when care is taken to minimize the amount of pressure on the abdominal wall, good contact between the skin and the transducer is required to form an image; the result is a minimum, but inevitable, amount of compression. Additionally, in some cases it is difficult to identify which of the reflecting surfaces represent the boundaries of subcutaneous fat. Furthermore, axial measurements in ultrasound depend on the speed of transmission of sound through tissue. Most ultrasound machines are calibrated to assume that the sound wave is passing through uniform soft tissue, but there are slight differences in the speed of sound wave propagation through fat compared with muscle or hepatic paren-chyma. Finally, if the transducer is not held perpendicular to the 107 CT2 vs SF2 O -2 - , H -12 -14 —1—I—1—I—I—I—r- ~T i i i I i i i i i i i r 2 4 6 8 10 12 14 16 18 20 22 24 VALUES FOR COMPUTED TOMOGRAPHY (CT2) (mm) CT2 VS US2 1° I ' => 4 j -'0 -i I : -12 '—I—I—I—I—I—I 12 14 16 ~ I I I 1 1—1—I—[~ 18 20 22 24 0 2 4 6 8 10 VALUES FOR COMPUTED TOMOGRAPHY (CT,) (mm) Mean differences in subcutaneous fat measurement at abdominal site 2. CT=computed tomography; SF=skinfold calipers; US=ultrasound; SD=standard deviation. JOURNAL OF THE AMERICAN DIETETIC ASSOCIATION / 857 RESEARCH skin surface, an off-axis artifact is introduced, thereby causing overestimation of the actual distance (18). The disagreement between studies may be due, in part, to differences in equipment used, technical procedures, and sites measured (8). Our study was conducted based on the assumption that com-puted tomography is the gold standard for fat measurement. Although studies have been done using cadavers (10), which provide a direct measure of body fat, in most cases this research option is not possible. Researchers generally agree that computed tomography provides the greatest precision of all indirect meth-ods currently available (10,11,13). In clinical settings, accuracy of the skinfold caliper method can be enhanced by setting standards for site selection and measurement techniques and by using one observer An unexpected result was the weak correlation observed be-tween methods when only the data from women were analyzed. This result could not be explained by differences in absolute measures between men and women. The data from men alone showed significant correlations in eight of nine possible relation-ships. Our findings suggest not only gender differences but also characteristics inherent to abdominal fat in women that make it difficult to.measure. The structure and function of the fat tissue may be highly variable because of differences in fat-cell size, number, membrane structure, or mobilizability. We acknowledge, however, that this finding may be an artifact of the small sample size. As a result of this pilot study, ultrasound was not used in a subsequent study to investigate body-fat gain in patients with anorexia nervosa. APPLICATIONS The results of our study indicate that relative agreement in measurement of subcutaneous body fat between skinfold calipers and computed tomography was superior to that exhibited be-tween ultrasound and computed tomography. This finding en-hances the potential use of skinfold calipers in the clinical setting. A practical comparison among the three methods shows that skinfold calipers, like ultrasound, are noninvasive. Computed tomography is invasive, as it uses x-rays. In terms of cost, com-puted tomography is three times as expensive as ultrasound; both methods require about the same amount of time. Measurements by skinfold calipers are the easiest and quickest to perform. The only cost is associated with the initial purchase of the calipers. Site selection and measurement technique account for sub-stantial differences between observers. With a single observer, however, skinfold measurements can yield extremely reproduc-ible results (22). Although training is required, a standardized measurement routine can'be readily established (22). • The authors thank the British Columbia Medical Services Foundation, part of the Vancouver Foundation, for providing the financial support for this research. Also, the excellent technical assistance of Beverley Keeley and Susan 108 / Spiro (ultrasound technicians) and Janice Brown (com-puted-tomography technician) from the Department of Radiology, St Paul's Hospital, Vancouver, British Colum-bia, Canada, is gratefully acknowledged. References 1. Kissebah AH, Vydelingum N, Murray R, Evans DJ, Hartz AJ, Kalkoff RK, Adams PW. Relation of body fat distribution to metabolic complica-tions of obesity. J Clin Endocrinol Metab. 1982; 54:254-260. 2. Krotkiewski M, Bjorntorp P, Sjostrom L, Smith U. Impact of obesity on metabolism in men and women: importance of regional adipose tissue distribution. J Clin Invest. 1983; 72:1150-1162. 3. Gillum RF. The association of body fat distribution with hyperten-sion, hypertensive heart disease, coronary heart disease, diabetes and cardiovascular risk factors in men and women. J Chronic Dis. 1987; 40:421-428. 4. Evans DJ, Hoffman RG, Kalhoff RK, Kissebah AH. Relationship of body fat topography to insulin sensitivity and metabolic profiles in premenopausal women. Metabolism. 1984; 33:68-75. 5. Anderson AJ, Sobocinski KA, Freedman DS, Barboriak JJ, Rimm AA, Gruchow HW. Body fat distribution, plasm lipids and lipoproteins. Arteriosclerosis. 1988; 9:88-94. 6. Bjorntorp P. Adipose tissue in obesity. In: Hirsch J, Van Itallie, T, eds. Recent Advances in Obesity Research. 4th ed. London, England: John Libbey; 1985:163-170. 7. Bray GA, Greenway FL, Molitch ME, Dahms WT, Atkinson RL, Hamilton RS. Use of anthropometric measures to assess weight loss. Am JClinNutr. 1978;31:769-773. 8. Ramirez ME. Measurement of subcutaneous adipose tissue using ultrasound images. Am JPhys Anthropol. 1992; 89:347-357. 9. Martin AD, Ross WD, Drinkwater DT, Clarys JP. Prediction of body fat by skinfold caliper: assumptions and cadaver evidence. Int J Obes. 1985; 9:SI:31-39. 10. AshwellM, Cole TJ, Dixon AK. Obesity: new insight into the anthropometric classification of fat distribution shown by computed tomography. BMJ. 1985; 290:1692-1694. 11. Krist H, Chowdhury B, Grangard U, Tylen U, Sjostrom L. Total and visceral adipose tissue volumes derived from measurement with com-puted tomography in adult men and women: predictive equations. Am JClinNutr. 1988;48:1351-1361. 12. van der Koy K, Seidell JC. Techniques for the measurement of visceral fat: a practical guide. Int J Obes. 1993; 17:187-196. 13. Ferland M, Despres JP, Tremblay A, Pinault S, Nadeau A, Moorjani S, Lupien PJ, Theriault G, Bouchard C. Assessment of adipose tissue distribution by computed tomography in obese women: association with body density and anthropometric measurements. Br J Nutr. 1989; 61:139-148. 14. Kuczmarski RJ, Fanelli MT, Koch GG. Ultrasonic assessment of body composition in obese adults: overcoming the limitations of the skinfold caliper. Am J Clin Nutr. 1987; 45:717-724. 15. Fanelli MT, Kuczmarski RJ. Ultrasound as an approach to assessing body composition. Am J Clin Nutr. 1984; 39:703-709. 16. Borkan GA, Hiuts DE, Cardarelli J, Burrows BA. Comparison of ultrasound and skinfold measurements in assessment of subcutaneous and total fatness. Am J Phys Anthropol. 1982; 58:307-313. 17. Volz PA, Ostrove SM. Evaluation of a portable ultrasonoscope in assessing the body composition of college-age women. Med Sci Sports Exerc. 1984; 16:97-102. 18. Booth RA, Goddard BA, Paton A. Measurement of fat thickness in man: a comparison of ultrasound, harpenden calipers and electrical conductivity. Br J Nutr. 1966; 20:719-725. 19. Black D, Vora J, Hayward M, Marks R. Measurement of subcutane-ous fat thickness with high frequency pulsed ultrasound: comparisons with a caliper and a radiographic technique; Clin Phys Physiol Meas. 1988; 9:57-64. 20. Lohman TG, Roche AF, Martorell R, eds. Anthropometric Stan-dardization Reference Manual. Champaign, 111: Human Kinetics Books; 1988. 21. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 1:307-310. 22. Jebb SA, Marinos E. Techniques for the measurement of body composition: a practical guide. Int J Obes. 1993; 17:611-621. 858 / AUGUST 1994 VOLUME 94 NUMBER 8 APPENDIX 2 ETHICS APPROVAL CERTIFICATES 109 APPENDIX 3 SUBJECT CRITERIA Based on: American Psychiatric Association. (1987). Diagnostic and Statistical Manual of Mental Disorders (3rd. ed. revised). Washington, DC: Author. 112 Diagnostic Criteria for Anorexia Nervosa A. Refusal to maintain body weight over a minimal normal weight for age and height, eg. weight loss leading to maintenance of body weight 15% below that expected; or failure to make expected weight gain during period of growth, leading to body weight 15% below that expected. B. Intense fear of gaining weight or becoming fat, even though underweight. C. Disturbance in the way in which one's body weight, size, or shape is experienced, eg. the person claims to \"feel fat\" even when emaciated, believes that one area of the body is \"too fat\" even when obviously underweight. D. Absence of at least three consecutive menstrual cycles when otherwise expected to occur (primary of secondary amenorrhea). 113 APPENDIX 4 SUBJECT RECRUITMENT ADVERTISEMENT 114 APPENDIX 5 INFORMED WRITTEN CONSENT 116 3. Bioelectrical Impedance Assessment (BIA) This method will be done to measure body composition, It is a non-invasive procedure that can be done while you are lying down. Small electrodes are attached to one wrist and one ankle, and measurement of resistance of a small electrical current is used for the calculation of body composition. 4. Dual Energy X-ray Absorptiometry (DEXA) This is also a non-invasive method for measuring body composition. It involves laying on a bed for approximately 10 minutes. There is a low level of radiation exposure which is less than 1/10th that of a standard chest x-ray. This is a low level of radiation and it is not associated with any known side effects. 5. Dietary Intake Records You will be asked to record everything that you eat and drink for four days. 6. Physical Activity Assessment You will be asked to complete a questionnaire to estimate the frequency, duration and intensity of any physical activity that you have participated in, for four days. ** Each appointment will take 1 hour, so your entire participation in the study will require no more than 2 hours. Risks and Significant Side Effects There are no risks associated with skinfold, bioelectrical impedance, or dual energy x-ray absorptiometry measurements. Potential Benefits Although the question as to what areas of the body gain fat during weight gain has been addressed, it has not been previously investigated. The results of this study will shed some light into the fat redistribution patterns that occur during weight gain. It will also aid in the nutritional assessment and intervention for patients with eating disorders. Monetary Compensation Unfortunately, we will be unable to provide monetary compensation for your participation in the study. Confidentiality Only the investigators listed on page one, and the study co-ordinator Ms. Orphanidou will have access to confidential data which identifies you by name or initials. A l l possible precautions will be taken to assure your anonymity and the confidentiality of the data gathered in the study. Al l of the data obtained will be coded by number, and no names will ever be used in any reports of the study. Only combined group data will be presented. 118 If you have any questions or concerns at any time during the study, you may contact Dr. Birmingham, Dr. McCargar, or Ms. Orphanidou at the phone numbers listed above. You will be informed of any significant new information pertaining to your safety. $ $ $ $ $ $ $ $ I have read the above information and I have had the opportunity to ask questions to help me understand what my participation would involve. I freely consent to participate in the study and acknowledge receipt of a copy of the consent form. Signature of Participant Date Signature of Guardian (if applicable) Signature of Witness 119 APPENDIX 6 ANTHROPOMETRIC MEASUREMENT STANDARDS SKINFOLDS Adopted from: 1. Gibson, R. S. (1991). Principles of Nutritional Assessment, New York: Oxford University Press. 2. T. G. Lohman, A. F. Roche, R. Martorell, (Eds.), Anthropometric Standardization Reference Manual. Illinois: Human Kinetics Books. 120 Triceps Skinfold The triceps SKF is taken on the back of the right arm at the mid-point between the tip of the acromion and the tip of the olecranon, with the elbow flexed to 90°. The triceps SKF is picked up with the left thumb and index finger approximately 1 cm proximal to the marked level, and the tips of the calipers are applied to the SKF at the marked level. C h e s t S k i n f o l d Chest SKF thickness is measured using a skinfold with its long axis directed to the nipple. The SKF is picked up as high as possible. The thickness is measured 1 cm inferior to this. 121 Subscapula Skinfold The participant stands with the shoulders relaxed and the arms by the sides. The SKF is raised so it can be measured on a diagonal line coming from the vertebral border of the scapula to a point 1 cm beneath the inferior angle. The SKF runs \\ Left arm behind back downward and outward at an angle approximately 4 5 ° to the spine. Axilla Skinfold Measured at the level of the xiphi-sternal junct ion, in the mid-axi l lary l ine, with the S K F hor izonta l . 122 Abdomen Skinfolds Abdomen S K F measurements were performed at 2 sites: Abdomen a: Vertical S K F raised 3 cm lateral to the mid-point of the umbilicus and 1 cm inferior to it. Abdomen b: Vertical S K F raised 5 cm lateral to and at the level of the umbilicus. Iliac Skinfold Measured in the mid-axillary line immediately superior to the iliac crest. An oblique S K F is grasped just posterior to the mid-axillary l ine f o l l ow ing the natura l cleavage lines of the skin. It is aligned infero-medially at 45° to the horizontal. The caliper jaws are applied about 1 cm from the fingers holding the S K F . t—Mid-Axillary Line 123 T h i g h S k i n f o l d A vertical fold is measured at the midline of the anterior aspect of the thigh, midway between the inguinal crease and the proximal border of the patella. I ngu ina l Crease M i d - T h i g h S k i n f o l d P r o x i m a l • B o r d e r of P a t e l l a C a l f S k i n f o l d The knee and hip are flexed to approximately 9 0 ° . The level of the maximum calf circumference is marked on the medial aspect of the calf. From a position in front of the subject, a S K F is raised parallel to the long axis of the calf on its medial aspect. M e d i a l A s p e c t 124 APPENDIX 7 ANTHROPOMETRIC MEASUREMENT STANDARDS CIRCUMFERENCES Adopted from: T. G. Lohman, A. F. Roche, R. Martorell, (Eds.), Anthropometric Standardization Reference Manual. Illinois: Human Kinetics Books. 125 Arm Circumference Measured at the mid-point between the tip of the acromion and the tip of the olecranon. Forearm Circumference Measured with the arm hanging downward and palms facing anteriorly. The tape is placed loosely around the proximal part of the forearm, perpendicular to its long axis, and moved up and down until the level of the maximum CIRC is located. 126 Chest Circumference Measured at the level of the fourth costo-sternal joints. The tape is held in the right hand while the free end of the tape is passed in front of the subject and retrieved by the measurer's left hand as it passes around the subject's back. The measurement is made in a horizontal plane at the end of a normal expiration. W a i s t C i r c u m f e r e n c e s Waist C IRC measurements were performed at 2 sites: Waist a: measured at the level of the natural waist which is the narrowest part of the torso, as seen from the anterior aspect. Waist b: measured at the level of the umbilicus. 127 H i p C i r c u m f e r e n c e The measurer stands at the site of the subject so that the level of max imum ex tens ion of the buttocks can be seen. The tape is placed around the buttocks in a horizontal plane at this level. T h i g h C i r c u m f e r e n c e s Thigh CIRC measurements were performed at 2 sites: Thigh a (mid-thigh): The tape is p laced horizontal ly around the thigh at the level of the thigh S K F measurement, ie. midway between the mid-point of the inguinal crease and the proximal border of the patella. Thigh b (proximal thigh): The tape is passed horizontally around the thigh immediately distal to the gluteal furrow. Inguinal 128 C a l f C i r c u m f e r e n c e The tape measure is positioned horizontally around the calf and moved up and down to locate the maximum C I R C in a plane perpendicular to the long axis of the calf. 129 APPENDIX 8a DEXA WHOLE BODY IMAGE COMPUTER PRINT-OUT 130 S t . P a u l ' 8 H o s p i t a l N u c l e a r M e d i c i n e D e p t . 1081 B u r r a r d S t . V a n c o u v e r , B . C . P h . 6 3 1 - 5 0 0 8 Name Ethnic CAUCASIAN ID J 5 3 9 Height 170 Age 29 Sex Fema le Weight kl .1 H Body 0 8 / 2 0 / 9 3 Sequence 1 Bone image not f o r diagnosis | MD T o t a l BMD ( g / c m ) 0 . 8 2 1 T o t a l BMC (g) 2033 T o t a l L e a n M a s s ( g ) 34811 T o t a l F a t Mass (g) 11041 T o t a l F a t % ! 2 3 . 1 S i r i UWE F a t % : 1 7 . 7 B r o z e k OWE F a t % i 17 .6 S o f t T i s s u e F a t % ! 2 4 . 1 % TBMC/FFM : 5 . 5 STD CV f o r T o t a l BMD: 1.0 See Cuide f o r other CVs. 6.5x13.0 ran, lSOran/s, 61.75 cm Rev. 2.5.0 / 1.3.0 C a l i b . 08/20/93 COMMENTS 131 St. Paul's H o s p i t a l Nuclear Medicine Dept. 1081 Burrard St. Vancouver, B.C. Ph. 631-5008 Name Address Telephone History Treatment Medications days eves ID J539 Ethnic CAUCASIAN Age 29 Menoage Sex Female Height 170 freight 47.7 Armspan Comments SCAN INFORMATION Type Scan Date-Seq Analysis Date Calibration Date Technician Physician Body 08/20/93 1 08/20/93 08/20/93 RF DR. BIRMINGHAM Resolution Speed Width Host/Scanner Analysis Revision 6.5 x 13.0 mm 180 mm/s 61.75 cm 2.5.0 / 1.3.0 2.5.0 DETAILED RESULTS BMD BMC AREA LENGTH WIDTH LEAN MASS FAT MASS g/cm2 g cm2 cm cm g g Head 1.466 436.4 297.7 3394 747.8 Trunk 0.727 529.6 728.0 16911 3833 Abdomen 0.705 234.7 333.1 9961 1566 Arms 0.562 322.1 573.3 2841 1640 Legs 0.850 744.7 876.3 11665 4821 Total 0.821 2033 2475 34811 11041 QUANTITIES MEASURED Whole body calculations are performed for the following quantities: AREA = Cursor region area, in cm2 BMC = Bone mineral content, in g BMD = Bone mineral density, in g/cm2 STM = Soft tissue mass, in g TBM = Total body mass, in g TBMC Total BMC TSTM Total STM 132 APPENDIX 8b PRE-DETERMINED DEXA REGIONS 133 134 APPENDIX 8c DEXA REGIONS USED IN STUDY 135 BI&ZA JRJB(BJtd2>NS 136 APPENDIX 9 BIA ELECTRODE PLACEMENT 137 138 APPENDIX 10 FOOD INTAKE RECORD FORM 139 GUIDELINES FOR KEEPING A FOOD RECORD A FOOD RECORD IS A DETAILED DESCRIPTION OF EACH FOOD OR BEVERAGE ITEM TAKEN OVER 24 HRS DURING A DAY. AN ACCURATELY COMPLETED FOOD RECORD CAN PROVIDE VALUABLE INFORMATION ABOUT THE NUTRITIONAL CONTENT OF AN INDIVIDUAL'S USUAL DIET. TO ASSESS YOUR DIET RECORD CORRECTLY, WE MUST BE ABLE TO CLEARLY PICTURE THE FOODS AND BEVERAGES THAT YOU HAVE RECORDED. THE GUIDELINES BELOW DESCRIBE THE INFORMATION THAT IS IMPORTANT FOR YOU TO RECORD. PLEASE READ THESE GUIDELINES BEFORE YOU START FILLING IN YOUR FOOD RECORDS. PLEASE KEEP A RECORD OF EVERYTHING THAT YOU EAT OR DRINK ON THE ATTACHED FORMS, FOR A PERIOD OF 4 DAYS (3 CONSECUTIVE WEEKDAYS AND 1 WEEKEND DA.Y). 1. THE PORTION SIZE (QUANTITY) NEEDS TO BE ACCURATELY RECORDED You can discribe portion sizes in as many ways as you like. For example, you might record: Volume 1 cup or 8 oz or 250 ml of 2%milk tablespoon or 15 ml of peanut butter teaspoon or 5 ml of sugar 1 1 Size 1 medium egg, poached small apple medium blueberry muffin 1 1 Weight 2 oz or 60 grams of hamburger meat or chicken or fish 140 2. DETAILED DESCRIPTION OF FOOD ITEMS IS ESSENTIAL Tell us as much as you can about the foods that you eat. BE SPECIFIC ABOUT THE TYPE OF FOOD. BRAND NAME IF APPLICABLE, AND THE CONTENT OF MIXED DISHES. For example: If you eat cookies, please tell us what type (eg. chocolate chip), what brand name (eg. Dare's or homemade), as well as the size (eg. 2\"). If you have milk, tell us if it is canned or fresh, and whether it is whole, 2% or skim, as well as the amount (eg. 1 cup). DESCRIBE MIXED FOODS AS IF YOU WERE WRITING A RECIPE Everyone has their own way of making everyday foods - please tell us how you do it! For example: If you made a cheese sandwich: What type of bread and cheese did you use? Did you add margarine of butter? Did you use mayonnaise or miracle whip? Did you add lettuce or tomato slices? How much of each item did you use? If you did not make the food yourself, describe the contents as best as you can. For example: If you had 1 cup of tuna casserole, let us know that it was about 1/2 macaroni, 1/4 tuna, and 1/4 peas and celery. 3. RECORD IMMEDIATELY AFTER EACH MEAL AND SNACK Take your food record with you if you go out to eat. Please keep track throughout the day - it is easy to forget exactly what you have eaten! 141 H E R E IS A N E X A M P L E FOR Y O U TO USE AS A GUIDE: Last Name: ^/Tl'th First Name: Diet Record Day #: J=_ Day nf Week: ^JLMCUW^ Date: ^1CU-Ol 3 '? 3 TIME & PLACE DESCRIPTION OF FOOD & BEVERAGE ITEMS QUANTITY CONSUMED Please specify amounts eg. oz, g, tsp, cups etc. La-rich 13 ; 3 0 hlMLOosoni and Ch