Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Analysis of variable expressivity in neurofibromatosis 1 Szudek, Jacek 2001

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

Item Metadata

Download

Media
831-ubc_2001-715426.pdf [ 9.87MB ]
Metadata
JSON: 831-1.0090779.json
JSON-LD: 831-1.0090779-ld.json
RDF/XML (Pretty): 831-1.0090779-rdf.xml
RDF/JSON: 831-1.0090779-rdf.json
Turtle: 831-1.0090779-turtle.txt
N-Triples: 831-1.0090779-rdf-ntriples.txt
Original Record: 831-1.0090779-source.json
Full Text
831-1.0090779-fulltext.txt
Citation
831-1.0090779.ris

Full Text

ANALYSIS OF VARIABLE EXPRESSIVITY IN NEUROFIBROMATOSIS 1 by JACEK SZUDEK B.Sc, McGil University, 1995 A THESIS SUBMITTED IN PARTIAL FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Department of Medical Genetics) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA August 2001 © Jacek Szudek, 2001 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 of WHOd GfiLlrftlCZ The University of British Columbia Vancouver, Canada DE-6 (2/88) ABSTRACT Neurofibromatosis 1 (NF1) exhibits extreme clinical variability. This variabilty greatly increases the burden for afected familes. The relationship of genetic factors to variable expressivity in NF1 is poorly understood. To improve understanding of NF1,1 studied relationships between several disease features in individuals and among afected relatives. My studies used clinical information on 4731 NF1 patients from three independent databases: the National NF Foundation International Database, the NF Instiute Database and a population-based registry of NF1 patients in north-west England. My initial studies found associations between several pairs of features in affected probands and between the occurence of individual features in affected parents and children. This establishes that some patients are more likely than others to develop particular NF1 features. Furthermore, the results of my logistic regressive models are consistent with grouping 9 of the features into three sets of associated features: 1) cafe-au-lait spots, intertriginous freckling and Lisch nodules; 2 ) cutaneous, subcutaneous and plexiform neurofibromas; and 3) macrocephaly, optic glioma and other neoplasms. Also, the occurence of Unidentifed Bright Objects on magnetic resonance imaging in young (<21 years) NF1 patients was associated with other expressed diagnostic features. Clinical features within a group may share pathogenic mechanisms that differ, at least in part, from those underlying features in other groups. I found no local associations between the presence of cutaneous neurofibromas, plexiform neurofibromas, and cafe-au-lait spots in each of ten divisions of the body surface in NF1 patients. However, the corelation among relatives in the number of body segments afected with one or more lesions was positve and significant for al three ii features. The developments of cutaneous neurofibromas, plexiform neurofibromas, and cafe-au-lait spots in NF1 patients are each spatialy independent but influenced by familal factors. Familal aggregation paterns of NF1 features among various classes of afected relatives were used to examine familal aggregation in greater detail. Using multivariate analyses, statisticaly significant associations among diferent classes of relatives were found for several features. Three distinct paterns were observed among the associations for familial features: 1) Lisch nodules and cafe-au-lait spots had greater associations between 1st degree relatives than between 2 n d degree relatives; 2) Subcutaneous neurofibromas, plexiform neurofibromas, cafe-au-lait spots, and intertriginous freckling had greater associations between sibs than between parents and children; and 3) Head circumference and stature had similar associations for al afected relatives. These familal paterns suggest that unlinked modifying genes, the normal NF1 alele, and the mutant NF1 alele may al be involved in the development of particular clinical features of NF1, but that the relative contributions vary for diferent features. The results presented in this thesis suggest that genetic factors are involved in phenotypic variabilty in NFL These findings also provide specifc clues to pathogenesis of NF1 features that can be tested in molecular studies. The methods of biostatistical analysis developed as part of this thesis can be applied to the study of other complex disorders. T A B L E O F C O N T E N T S A B S T R A C T ii T A B L E O F C O N T E N T S iv LIST O F T A B L E S viii LIST O F F I G U R E S ix A C K N O W L E D G E M E N T S xi 1. I N T R O D U C T I O N 1 1.1 History 2 1.2 Prevalence 3 1.3 Diagnosis 3 1.4 Prevalences of clinical features 5 1.5 Pathogenesis of selected clinical features 6 1.6 Mortality 14 1.7 Gene location and structure 15 1.8 NF1 gene expression 16 1.9 Neurofibromin function 17 1.10 NF1 gene in other species 20 1.11 Animal models 20 1.12 Variable expressivity 22 1.13 Possible sources of variable expressivity 25 1.14 Hypotheses 32 1.15 Objectives 32 2. A S S O C I A T I O N S O F C L I N I C A L F E A T U R E S IN N E U R O F I B R O M A T O S I S 1 .. 33 2.1 Hypotheses 34 2.2 Objectives 34 2.3 Introduction 34 2.4 Subjects and methods 35 2.4.1 Patients and data description 35 2.4.2 Clinical features 36 2.4.3 Statistical analysis in individual probands 37 2.4.4 Statistical analysis in affected parents and children 39 2.5 Results 40 - 2.5.1 Associations in individuals 40 2.5.2 Parent-child associations 41 2.6 Discussion 41 2.6.1 Associations in individuals 41 2.6.2 Parent-child associations 46 2.7 Conclusion 49 3. G R O W T H IN N O R T H A M E R I C A N W H I T E C H I L D R E N W I T H N E U R O F I B R O M A T O S I S 1 50 3.1 Hypothesis 51 3.2 Objective 51 3.3 Introduction 51 3.4 Subjects and methods 52 iv 3.4.1 Subjects 52 3.4.2 Reference populations 53 3.4.3 Distribution analysis 53 3.4.4 Growth curves 54 3.5 Results : 55 3.5.1 Standardised stature and occipitofrontal circumference 55 3.5.2 Centile curves 60 3.6 Discussion 60 3.6.1 Population norms 60 3.6.2 Standardisation 69 3.6.3 Assessment of heterogeneity 69 3.6.4 Assessment of standardised distributions 69 3.6.5 Assessment of centile curves 71 3.6.6 Pathogenesis 71 3.7 Conclusion 73 4. LOGISTIC REGRESSIVE MODELS OF CLINICAL FEATURES IN NEUROFIBROMATOSIS 1 74 4.1 Hypothesis 75 4.2 Objective 75 4.3 Introduction 75 4.4 Subjects and methods 76 4.4.1 Subjects 77 4.4.2 Clinical features 77 4.4.3 Statistical models 78 4.4.4 Model validation 81 4.4.5 Interpretation 81 4.5 Results 82 4.5.1 Prevalence by age of NF1 clinical features 82 4.5.2 Model development and validation 109 4.5.3 Parameter estimate comparisons 109 4.5.4 Consistent parameters from validated models 110 4.6 Discussion 116 4.6.1 Ascertainment bias 116 4.6.2 Consideration of binary treatment of variables 117 4.6.3 Statistical signifcance 117 4.6.4 Previous reports of associations 118 4.6.5 Pathogenetic interpretation of associations 119 4.6.6 Cross-sectional nature of data 121 4.7 Conclusion 122 5. UNIDENTIFIED BRIGHT OBJECTS ON MRI ASSOCIATED WITH DIAGNOSTIC FEATURES OF NEUROFIBROMATOSIS 1 123 5.1 Hypothesis 124 5.2 Objective 124 5.3 Introduction 124 5.4 Subjects and methods 125 5.4.1 Subjects 125 v 5.4.2 Statistical analysis 125 5.5 Results 126 5.5.1 Frequency of clinical features analysed 126 5.5.2 Unidentifed bright objects by number of diagnostic features 126 5.5.3 Logistic regressive model of unidentifed bright objects versus diagnostic features 131 5.6 Discussion 131 5.6.1 Unidentifed bright object associations 131 5.6.2 Bias 133 5.6.3 Cross-sectional nature of data 134 5.6.4 Pathogenesis 135 5.7 Conclusion 135 6. ANALYSIS OF LOCAL AND FAMILIAL FACTORS IN NEUROFIBROMATOSIS 1 LESIONS 136 6.1 Hypothesis 137 6.2 Objective 137 6.3 Introduction 137 6.4 Subjects and methods 138 6.4.1 Subjects 138 6.4.2 Analysis of local efect 138 6.4.3 Skin surface area 139 6.4.4 Total number of neurofibromas 140 6.4.5 Analysis of familial factors 140 6.5 Results 141 6.5.1 Lesion frequency by body segment 141 6.5.2 Occurence of various lesions in body segments is independent 144 6.5.3 Relationship between segment size and number of neurofibromas 144 6.5.4 Familal corelations 147 6.6 Discussion 147 6.6.1 Lesion severity 147 6.6.2 Ascertainment issues 150 6.6.3 Local associations of lesions 150 6.6.4 Local factors in pathogenesis 151 6.6.5 Familal associations 152 6.6.6 Familal factors in pathogenesis 153 6.7 Conclusion 154 7. ANALYSIS OF INTRA-FAMILIAL PHENOTYPIC VARIATION IN NEUROFIBROMATOSIS 1 155 7.1 Hypothesis 156 7.2 Objective 156 7.3 Introduction 156 7.4 Subjects and methods 158 7.4.1 Subjects 158 7.4.2 Features 158 7.4.3 Confounding factors 159 7.4.4 Multivariate probit and normal models 159 vi 7.4.5 Assessment of intrafamilal corelations 161 7.5 Results 162 7.5.1 Feature prevalences 162 7.5.2 Individual covariates 162 7.5.3 Intrafamilal associations 166 7.6 Discussion 178 7.6.1 Variable expressivity in other disorders 178 7.6.2 Statistical signifcance 179 7.6.3 Representativeness of the sample 179 7.6.4 Comparison to other studies 180 7.6.5 Sample size considerations 181 7.6.6 Minimising the confounding efect of age 182 7.6.7 Statistical signifcance 182 7.6.8 Interpretation of intrafamilal associations 183 7.7 Conclusion 188 8. GENERAL DISCUSSION 189 8.1 Summary 190 8.1.1 Pair-wise analyses in individual NF1 patients." 190 8.1.2 Analyses of several diferent features at once 192 8.1.3 Familal analyses 193 8.2 How this study fits into curent NF1 research 195 8.3 Future directions 196 REFERENCES 200 APPENDIX - ANALYSIS OF SIMULATED DATA WITH MPROBIT 220 vii L I S T O F T A B L E S Table 1.1: National Instiutes of Health diagnostic criteria for neurofibromatosis 1 3 Table 1,2: Studies of the prevalence of neurofibromatosis 1 4 Table 1.3: Comparison of symptom prevalence among 6 studies of NF1 cases 6 Table 2.1: Prevalence of clinical features of NF1 in probands and afected relatives 38 Table 2.2: Associations of NF1 features in probands 42 Table 2.3: Associations of features between parents and children 43 Table 4.1: Logistic regressive models of NF1 features 111 Table 4.2: Parameter estimates for logistic regressive models of NF1 features 113 Table 4.3: Logistic regressive models summarised as odds ratios 115 Table 5.1: Features associated with Unidentifed Bright Objects 132 Table 6.1: Distribution of lesions by body segment 145 Table 6.2: Associations between lesions by body segment 146 Table 7.1: Comparison of NF1 feature frequencies from two studies 163 Table 7.2: Summary of multivariate models of NF1 clinical features 164 Table 7.3: Number of relatives used in multivariate models of NF1 features 167 L I S T O F F I G U R E S Figure 1.1: Pictures of selected NF1 clinical features 7 Figure 3.1: Distribution of sex- and age-standardised stature 56 Figure 3.2: Distribution of sex- and age-standardised occipitofrontal circumference 58 Figure 3.3(a): Stature centiles in males 2-18 years 61 Figure 3.3(b): Occipitofrontal circumference centiles in males 2-18 years 63 Figure 3.4(a): Stature centiles in females 2-18 years 65 Figure 3.4(b): Occipitofrontal circumference centiles in females 2-18 years 67 Figure 4.1: Prevalence by age of 6 or more cafe-au-lait spots : 83 Figure 4.2: Prevalence by age of intertriginous freckling 85 Figure 4.3: Prevalence by age of 2 or more subcutaneous neurofibromas 87 Figure 4.4: Prevalence by age of 2 or more cutaneous neurofibromas 89 Figure 4.5: Prevalence by age of plexiform neurofibromas 91 Figure 4.6: Prevalence by age of Lisch nodules 93 Figure 4.7: Prevalence by age of optic glioma 95 Figure 4.8: Prevalence by age of seizures 97 Figure 4.9: Prevalence by age of pseudarthrosis 99 Figure 4.10: Prevalence by age of scoliosis 101 Figure 4.11: Prevalence by age of macrocephaly 103 Figure 4.12: Prevalence by age of short stature 105 Figure 4.13: Prevalence by age of neoplasms 107 Figure 5.1: Prevalence by age of Unidentifed Bright Objects 127 Figure 5.2: Unidentifed Bright Objects by the number of diagnostic features 129 ix Figure 6.1: Body segment scheme used for recording location of lesions in the Neurofibromatosis Instiute Database 142 Figure 6.2: Number of afected segments versus number of neurofibromas 148 Figure 7.1: Feature associations among relatives with NF1 168 Figure 7.2: Feature associations among 1st and 2nd degree relatives with NF1 171 Figure 7.3: Feature associations among sib and parent-child pairs with NF1 174 Figure 7.4: Feature associations among mother- and father-child pairs with NF1 176 Figure A.l: Prevalence by age of a simulated clinical feature 226 Figure A.2: Associations of simulated features among al afected relatives 228 Figure A.3: Associations of simulated features among 1st and 2nd degree relatives 230 Figure A.4: Associations of simulated features among sib and parent-child pairs 232 Figure A.5: Associations of simulated features among mother- and father-child pairs. 234 x A C K N O W L E D G E M E N T S I would like to thank my research supervisor, Jan Friedman, my friend and co-worker, Patricia Birch, and my statistics guru, Hary Joe, for their guidance and support. I could have never completed this thesis without them. Thank you to my supervisory commitee, Barb McGilivray, Diana Jurilof  and Dessa Sadovnick, for their encouragement and feedback. To the members of the British Columbia Neurofibromatosis Foundation, thank you for teaching me that medical research can and should be important. Thank you to my family, Carmen Larsen and Ewa Szudek, for their companionship and faith. xi I N T R O D U C T I O N Neurofibromatosis 1 (NF1) is a generaly progressive human disorder that afects the skin, central nervous system, skeleton and other organs. It is inherited in an autosomal dominant fashion and afects around 1/3,500 people (Crowe et al. 1956; Poyhonen et al. 1997b), making it one of the most frequent autosomal dominant diseases in humans. The defining feature of NF1 is the neurofibroma, but patients often also develop other features, such as cafe-au-lait spots, Lisch nodules, pseudarthrosis, scoliosis, optic glioma and other neoplasms. Some of the features can be serious or life-threatening. Malignancy and vasculopathy are significant sources of morbidity (Zoler et al. 1995; Rasmussen et al. 2001). Owing to the progressive nature of NF1, the abilty to diagnose NF1 in a patient improves with age as more features develop or existing features become more numerous or prominent (DeBela et al. 2000b). 1.1 History Neurofibromatosis was first clinicaly recognised by von Recklinghausen in 1882 (Crump 1981). Since then, there have been several seminal contributions. The distinction between NF1 and NF2 was made in 1981 (Riccardi 1981). An NIH conference in 1988 (NIH 1988) developed consensus criteria for making the diagnosis of NF1. The NF1 gene was localised to the long arm of chromosome 17 (Goldgar et al. 1989) and cloned in 1991 (Marchuk et al. 1991). 2 1.2 Prevalence The prevalence of NF1 has been measured in several diferent populations. Table 1.2 summarises the methods and results. Prevalence estimates range from 1/960 to 1/7,800, but most are between 1/2,500 and 1/4,500. Ascertainment diferences probably contribute to the variation among the studies, but racial diferences have not been ruled out. Nimura (1990) surveyed 1,042 NF1 patients in Japan, but did not report prevalence 1.3 Diagnosis A person is diagnosed with NF1 if he or she satisfies two or more of the seven criteria listed in the folowing table (NIH 1988; Gutmann et al. 1997): Table 1.1: National Institutes of Health diagnostic criteria for neurofibromatosis 1. Cardinal Clinical Features (Any two or more are required for diagnosis) • Six or more cafe au lait spots over 5mm in greatest diameter in prepubertal individuals or over 15 mm in greatest diameter in postpubertal individuals. • Two or more neurofibromas of any type, or one plexiform neurofibroma • Freckling in the axilary or inguinal regions • Optic glioma • Two or more Lisch nodules (iris hamartomas) • A distinctive osseous lesion such as sphenoid dysplasia or thinning of the long bone cortex with or without pseudarthrosis. • A first degree relative (parent, sibling, or ofspring) with NF1 by the above criteria These criteria are very efective for making the diagnosis in adults or older children, but they cannot diagnose 5% of 8-year-olds with NF1 and almost half of 1-year olds (DeBela et al. 2000b). For this reason, other clinical features such as unidentifed bright objects on magnetic resonance imaging have been proposed as additonal diagnostic 3 GO © es E © © ^ 3 O 4» O S CN CM W ° s % s C J U CU CS ' M >• h a- § S <u <*i on O fi a 03 Pi a o <s <+H -H ^ " cu QJ H - » — OH « T3 H < 3 cu cu o c CU cd > CU P H GO CU • PN T3 3 1/3 T3 fi -*-» CO cd fi" <U B fi cd GO fi O cn a T3 cd 13 ' B , cn O -fi 13 f -CU c cu <+H o fi GO ON o f Ol • i-H o o o cn o o o f CO •5 a . . O cd S-H -t-n cn ^3 fi a ^ cu <+H o fi o cd — x fi cn cd o CU cn <-t—i -t-» o '3 * i OH f cn 'tu cd <tj _) cd s ° 1^ S -fi fi . 2 • —H a * cd <u X cn cu O — . f i cd [—i _o ^ °^  C P H T3 O • t-H cn >> -fi OH cn cn cu fi 0 0 . 2 5 fi «a> <+h cd ° cn O HI cn CU > CU > 13 kH <H-H o fi-cu a GO cn cu cn cn cd cu cr j d cn 13 -T3 ( » -kn <+H o o a fi 1 - 1 CU O cn ifi cu <U In ^ od cd o o CO ON cd "GO GO fi Ol OO o " '* fi CU CU C/3 O cn OH <+H o cn cu fi cr cn <U cn cd cu T3 CU H—» o C+H cd <H-H O cn (U > • i-H 13 kH <H-H o CU a cn cn cu cn cn cd o cn £ OH O cn CU • kH CU fi cr GO -T3 kH O o cd ' B H cn O -fi cn cu o cu cn cn o CU fi CU a o cu i2 CU -CJ cu >-l <l> *1 cu 13 t3 13 •2 i § o o oo" o o o rn o~ o f -CJ GO cd •5 u s. « fi lg f. — <x> > ^ N a o o oo o o o O ir> ON VO r-H^ 3T 5 t o f Ol ON Ol o-_ ; ON cd —1 > cu cS GO cn cu C * C+H o fi o _fi 13 > cu o t; cd OH cn cd fi o c 1 X cu "cd o cu •> cu kH o o cu cd o 4h ^ -fi fi U O S Ph T3 S O O-CU cd kH cn O O o f cn t--fi cd "fi o vo ON vo cn O m kH ON cu U H M T3 ON oo 1994 fi 00 ON . ON I — H cd 1994 13 +-* CU O cn 1 - 1 cd fi +-> Id CU • I-H 13 fi rels o « cn , - H cn 5 cd ret H - » fi to +H CU CU fi o x> Axel Huso ON cu a o -fi o-Axel Huso oo Full CU ON >^ ON Axel Huso ON Full O ON o O ON P H ^ H criteria (Curless et al. 1998), although they must be beter characterised before their use as diagnostic criteria is justified (DeBela et al. 2000a). 1.4 Prevalences of clinical features Riter and Riccardi studied 111 3-generation familes with NF and found no instance of skipped generation, suggesting that penetrance of NF is complete (Riter and Riccardi 1985). Although penetrance appears to be complete, NF1 can involve many diferent clinical features. Table 1.3 summarises the prevalences of NF1 clinical features from several clinical studies. Sample sizes are given under the authors' names. The prevalence of every feature was not reported in al of the studies, and a blank cel means that the study did not examine that feature. Nevertheless, each feature was reported in at least two studies, and comparisons can be made. Riccardi ascertained his subjects through the specialised Baylor NF Program (Riccardi 1992). Friedman and Birch ascertained their subjects through the National Neurofibromatosis Foundation International Database (NFDB), which colects standardised information from specialised clinics around the world (Friedman et al. 1993). Subjects in both studies were drawn mostly from pediatric centres, an ascertainment which contributes to the low mean ages of the samples. The ascertainment of subjects by Crowe et al., Huson at al. and Samuelsson and Axelsson is summarised in Table 1.2. The studies of Huson and Samuelsson are population based and provide useful standards for comparison, despite being smaler than the others. Diferences among the studies are partly due to inconsistent definitons of feature "presence" or "absence" -5 Friedman and Birch used the NIH criteria (Table 1.1), whereas the other studies largely predate the formal criteria. There is no known overlap between the databases. Table 1.3: Comparison of symptom prevalence among 6 studies of NF1 cases. Feature Friedman and Riccardi Crowe et Huson et Samuelsson and Birch 1997b 1992 al. 1956 al. 1989a Axelsson 1981 (n=1728) (n=953) (n=203) (n=135) (n=91) Cafe-au-lait spots 89% 100% 78% 84% 82% Dermal discrete 54% 93% neurofibromas Plexiform 23% 40% 16% 32% neurofibromas Xanthogranuloma 2% 2% 1% Lisch nodules 59% 84% 85% Pseudarthrosis 2% 3% 4% Scoliosis 24% 25% 16% 10% >10% Optic glioma 4% 10% 0% (symptomatic) Malignancy 5% 4% 5% 10% Seizures 6% 6% 7% 3-9% Mean age (years) 17.7 17.8 26.1 1.5 Pathogenesis of selected clinical features Cafe-au-lait spots (Figure 1.1(a) are present within the first year of life in almost al NF1 patients and tend to increase in number and size during the first decade of life (see Figure 4.1) (Riccardi 1982). This makes cafe-au-lait spots particularly useful in making the diagnosis of NF1 (DeBela et al. 2000b). They are typicaly 10 to 30 mm in diameter in adults and can occur anywhere on the skin except the scalp, eyebrows, palms and soles (Riccardi 1992). The basal layer of the epidermis contains melanocytes derived from the neural crest. In a cafe-au-lait spot, these cels characteristicaly have giant melanosomes 6 Figure 1.1: Pictures of selected NF1 clinical features. (a) Front view of an NF1 patient with 2 large cafe-au-lait spots, (b) Front view of a patient with hundreds of dermal discrete neurofibromas, (c) Leg hypertophy resulting from a difuse plexiform neurofibroma in a patient with NF1. (d) The iris of a patient with NF1, showing multiple Lisch nodules, (e) Rear view of an NF1 patient with scoliosis, (f) Radiograph of the right arm of a patient afected with NF1, showing long bone bowing. (g)Magnetic resonance image of a right optic pathway glioma in a patient afected with NF1. Most of these pictures have been published previously (Friedman et al. 1999). 7 8 - specialised intracelular organeles for melanin synthesis (Bolande 1974; Fitzpatrick 1981). Cafe-au-lait spots typicaly cause only cosmetic problems (Gutmann et al. 1997). Intertriginous (skin fold) freckles in NF1 patients are similar in colour to cafe-au-lait spots but are usualy 1 to 3 mm in diameter. Intertriginous freckles derive from a physiological pathway that has nothing to do with light exposure, but once present they darken with light exposure (Fitzpatrick 1981). Their development may be afected by increased temperature, moisture or salinity found in skin folds (Friedman and Riccardi 1999). Neurofibromas have the same histology whether they occur as part of NF1 or distinct from NF1. The neurofibroma is a complex benign tumour arising in the fascicles of peripheral nerves - an endoneurium surounds each nerve fibre and a bundle of nerve fibres is surounded by a perineurium to form a fascicle. Histologicaly, a local increase in endoneurial matrix within the fascicle is accompanied by a thickened perineurium, Schwann cels that increase in size and number (Harkin and Reed 1969), and an increased number of mast cels and fibroblasts (Giorno et al. 1989). Neurofibromas can be classified into two categories - dermal discrete (Figure 1.1(b) and plexiform (Figure 1.1(c). Dermal discrete neurofibromas include cutaneous and subcutaneous neurofibromas. Cutaneous neurofibromas move with the skin when it is stretched, while subcutaneous neurofibromas lie below the skin and do not move with it (Riccardi 1992). Plexiform neurofibromas include nodular and difuse plexiforms. Dermal discrete neurofibromas are confined to a single fascicle within a nerve, while difuse and nodular plexiform neurofibromas involve tortuous changes in the nerve fibres of one ore more fascicles (Burger and Scheithauer 1994). Discrete neurofibromas occur in most NF1 9 patients and tend to increase in number with age (see Figures 4.3-4.5). Patients who present with plexiform neurofibromas usualy do so during childhood. Some difuse and most deep nodular plexiform neurofibromas are asymptomatic and not apparent on surface exani (Riccardi 1992). Neurofibromas include nerve, Schwann, fibroblast and mast cels. Proportions vary both among and within tumours (Erlandson 1991), but 50-80% of neurofibroma cels react with Schwann cel-specific antibodies (Peltonen et al. 1988). Although NF1'1' fibroblasts have a greater proliferative capacity than normal or heterozygous fibroblasts in vitro (Atit et al. 1999), loss of heterozygosity at the NF1 locus in neurofibromas occurs primarily in a subpopulation of Schwann cels, and not in fibroblasts or other cels (Kluwe et al. 1999; Rutkowski et al. 2000; Sera et al. 2000). This suggests that Schwann cels are the pathogeneticaly culpable cels in neurofibromas. Lisch nodules (Figure 1.1(d) are almost exclusively found in NF1 patients (NM 1988). They are uncommon in young NF1 patients but occur in most patients over 10 years of age (see Figure 4.6). Histologicaly, Lisch nodules consist of aggregations of oval to round cels that form dome-shaped papules (smal, elevated areas) on the anterior layer of the iris (Lukacs et al. 1997). Electron microscope observations reveal that Lisch nodules are hamartomas (self-limiting, benign growths) composed of melanocytic cels, most of which contain immature melanosomes (Pery and Font 1982). Lisch nodules are not associated with visual problems or other ocular manifestations (Friedman and Riccardi 1999). Loss of heterozygosity at the NF1 locus has not been studied in cels from Lisch nodules. 10 Scoliosis (Figure 1.1(e) is frequent inNFl patients, but the degree of spinal curvature is variable. Dystrophic scoliosis involves changes in rib or vertebral shape, or vertebral spacing or rotation and can result in a severe, angular curve in the spine. Dystrophic scoliosis is considered characteristic of NF1 and is distinct from non-dystrophic scoliosis in NF1 patients, which is similar to idiopathic scoliosis (Winter et al. 1979; Calvert et al. 1989; Cox and Southern 1995) and may merely be coincidental to NF1 (Riccardi 1992). Dystrophic scoliosis inNFl is thought to be a developmental anomaly with an early onset (Funasaki et al. 1994), but dystrophic features can accumulate over time (Durani et al. 2000). Its pathogenesis may be congenital, but it is not radiologicaly evident at birth. Several factors have been implicated in dystrophic scoliosis including: bone deterioration (Wilde et al. 1994), intraspinal or paraspinal tumours (Crawford and Bagamery 1986) and dural ectasia (Stone et al. 1987). One possible mechanism is functional failure of the pedicles, which normaly join and transfer compressive forces between the lamina and vertebral bodies (Purkiss et al. 2001). Tibial or other long bone pseudarthrosis consists of a broken bone that does not heal properly. In fact, the etymology of the word "pseudarthrosis" is the Greek for "false joint". The basis is thought to be a congenital defect in bone formation (Riccardi 1999), and patients with this conditon usualy present at a very young age with a characteristic bowing (Figure 1.1 (f)) of a long bone before fracture occurs (Crawford and Bagamery 1986; Rudicel 1987). Very little is known about the natural history or pathogenesis of pseudarthrosis in NF1, although Stevenson et al. (1999) found a male predominance and no parent of origin efect. 11 Most optic pathway gliomas (astrocytomas) (Figure 1.1(g) inNFl involve the optic chiasm or the optic nerves running between the chiasm and the retina (Listernick and Gutmann 1999). Lesions that involve the optic chiasm can interfere with the hypothalamus and cause precocious puberty (Habiby et al. 1995). Unlike sporadic optic gliomas, those in NF1 rarely involve the optic tracts between the chiasm and the visual cortex, and their growth is usualy limited (Listernick et al. 1995). NF1 expression appears to be tightly regulated in astrocytes and may normaly inhibit growth at key points in development. NF1 expression is nearly undetectable in resting A^F7+/+astrocytes (Gutmann et al. 1996) but can be increased in response to certain stimuli (Hewet et al. 1995). Furthermore, Nfl+I~ mice have increased numbers of cerebral astrocytes and increased astrocyte proliferation compared to wild-type litermates (Nordlund et al. 1995; Gutmann et al. 1999). These activities may involve interacting proteins such as Ras (see Sectionl .9). One of the functions of neurofibromin is to downregulate p21-ra.s activity, and increased activity of p21-ras has been observed in sporadic (non-NFl) astrocytomas (Gutmann et al. 1996). The majority of tumours in NF1 patients are benign (Riccardi 1992), but patients are at increased risk for malignancy compared to unafected individuals. Although the lifetime risk for any malignancy is only marginaly higher, patients (particularly younger ones) are at much higher risk than usual for connective and other soft tissue and brain malignancies (Rasmussen et al. 2001). The most common malignancies in NF1 are astrocytomas, sarcomas, leukemias, pheochromocytomas and rhabdomyosarcomas (Sorensen et al. 1986; Huson et al. 1988; Riccardi 1992; North 1993). Astrocytomas in NF1 patients are usualy optic pathway gliomas (described above) but are found at other 12 CNS sites in about 1% of patients (Gutmann 1999b). The most common sarcomas are malignant peripheral nerve sheath tumours (MPNST), which usualy arise from plexiform neurofibromas (Korf 1999a). Leukemias include juvenile chronic myeloid leukemia (JCML) and myelodysplasia syndromes. Although leukemia is more common in NF1 than in the general population, it is stil  quite rare, and the excess of leukemia in NF 1 is limited to childhood (Gutmann 1999b). Defects in NF1 function can lead to greater ras activity and increased celular proliferation. NF1 codes for a tumour suppressor gene (see discussion below), and the development of tumours in NF1 patients may be the result of somatic mutations in the one remaining normal alele. This certainly seems true of JCML, in which loss of heterozygosity (LOH) has been shown for NF1 and activated ras has been linked to myeloid cel proliferation (Bolag et al. 1996). A similar mechanism may be involved in other NF1 tumours. LOH for NF1 has also been found in astrocytomas (Gutmann et al. 2000) but not in normal white mater (Lau et al. 2000). LOH for NF1 has been found in pheochromocytomas (Gutmann et al. 1994), in MPNSTs (Skuse et al. 1989; Skuse et al. 1991), and in some cels in dermal discrete and plexiform neurofibromas (Colman et al. 1995) but it is not clear if it is a cause or a result of tumour growth (Korf 1999a). The pathogenesis of non-tumour features in NF1, such as short stature and macrocephaly, is even less wel understood. Scoliosis and early or delayed puberty occasionaly influence stature. However, short stature associated withNFl usualy afects the whole skeleton proportionately, and no specifc cause is apparent in most cases (Riccardi 1992; Huson 1994). Disease features such as hydrocephalus and plexiform neurofibromas occasionaly afect occipitofrontal circumference (OFC). 13 However, increased OFC among NF1 patients usualy has no obvious cause and appears to result from prenatal or neonatal overgrowth of the brain (Riccardi 1992; Huson 1994). Seizures occur in less than 10% of NF1 patients and are no diferent than those in the general population (Gutmann 1999a) - which have a prevalence of less than 1% (Wright et al. 2000). They can occur at any age and may include grand mal tonic-clonic, absence, complex partial or petit mal seizures or hypsarhythmia. Seizures may be a consequence of haploinsuficiency or abnormal function of the NF1 gene, but relationship between seizures and NF1 function has not been investigated, and the pathogenesis of this feature in NF1 is unclear. 1.6 Mortality Many disease features are more prevalent or prominent among older patients (DeBela et al. 2000b), but few longitudinal studies of NF1 have been done. Zoler et al. conducted a 12-year folow-up study of 70 adult NF1 patients in the city of Goteborg, Sweden (Zoler et al. 1995). Mean life expectancy was 61.6 years, significantly lower than the expectancy of 75 years in the general Swedish population. Twenty-two of the patients died during the study, significantly higher than the 5.1 expected based on the general Swedish population. Malignancy occured in 12 (55%) of the deaths. 10 of the 12 patients with hypertension died during the period of study. No studies have compared the prevalence of hypertension in NF1 patients and the general population. A study of death certificates identified 3770 persons with NF1 and found a median age at death of 59 years compared with 74 years in the general population (Rasmussen et al. 2001). In addition, people afected with NF1 were significantly more 14 likely to have soft tissue or brain malignancy or vascular disease listed on their death certificates. This was especialy true among younger subjects. Valvular pulmonic stenosis appears to be more common in NF1, but it is unclear whether other vascular disease is also associated with NF1 (Friedman 1999). 1.7 Gene location and structure The NF1 gene has been mapped to 17ql 1.2; Goldgar et al. (1989) summarised the results ofthe international consortium for NF1 linkage. The gene was identified in 1990 (Viskochil et al. 1990; Walace et al. 1990). NF1 has an open reading frame of 8454 nucleotides (Marchuk et al. 1991) and spans 335 kb of genomic DNA (Li et al. 1995). Sixty exons have been identified, and the genomic sequence is available from GenBank (accession number AC004526). NF1 has one of the highest mutation rates of any gene in the human genome, estimated between lxlO"4 and 3xl0"5 (Sergeyev 1975; Huson et al. 1989a; Clementi et al. 1990). This is comparable to other genes of similar size, such as Duchenne muscular dystrophy (Vogel and Motulsky 1997), and may be partly due to the large number of nucleotides in these genes. Most reported mutations are unique, but some are observed repeatedly (Korf 1999b; Messiaen et al. 2000). NF1 -homologous loci have been detected by fluorescence in situ hybridisation analysis and genomic sequence homology at 2q21, 2q33, 14ql.2, 15ql 1.2, 18pl 1.2, 21ql 1.2 and 22ql 1.2 (Regnier et al. 1997). These NF1 -related loci arose through repeated (partial) gene duplications of the NF1 locus, rather than through a process involving reverse transcription (Bernards 1998). Most human NF1 -related loci are 15 closely associated with centromeres, consistent with emergence by pericentromeric interchromosomal transpositon (Regnier et al. 1997). Characterisation of the NF1 region itself shows that three other genes are embedded in intron 27b. EVI2A encodes a 232 amino acid peptide expressed in the brain and bone marow which is short-lived and appears to have a transmembrane domain (Cawthon et al. 1991). EVI2B encodes a 448 amino acid peptide also expressed in the bone marow but of unknown function (Cawthon et al. 1990). OMGP encodes a 416 amino acid peptide expressed on the surface of oligodendrocytes that appears to have an intracelular adhesion function (Mikol et al. 1990). Large deletions of the NF1 gene inevitably involve loss of embedded genes, but the consequences of homozygous inactivation of any of these 3 genes have not been demonstrated in vivo or in vitro (Viskochil 1999). NF1 is sometimes caused by deletions of the entire NF1 locus. These large deletions cluster around common breakpoints and can afect expressed regions flanking the NF1 gene, but it remains to be seen if these influence expression of the disease (Dorschner et al. 2000). 1.8 NF1 gene expression The NF1 mRNA transcript is 12-14 kb in size (Buchberg et al. 1990). RT-PCR has detected NF1 mRNA in almost al tissues, but the highest levels are in neurons, Schwann cels, oligodendrocytes, keratinocytes, astrocytes, adrenal medula and white blood cels (Daston et al. 1992). The protein product of the NF1 gene is neurofibromin (SWISS-PROT accession number P21359), a peptide over 220 kD in size (DeClue et al. 1991). Neurofibromin is ubiquitously expressed during embryogenesis in mice, with adult-specific tissue expression levels established one week after birth (Gutmann et al. 1995). Three major alternatively spliced and diferentialy expressed isoforms of NF1 mRNA have been identified in humans. The most common isoform includes an alternatively spliced exon within the GTPase-Activating Protein (GAP) related functional domain (Marchuk et al. 1991) (see section 1.9). The inclusion of this exon results in a decrease of GAP function (Andersen et al. 1993). This isoform is found in many animal species and is associated with diferentiated cels (Viskochil 1999). Another isoform involving a diferent alternatively spliced exon is preferentialy expressed in muscle (Gutman et al. 1993). A third isoform is found exclusively in the central nervous system during embryogenesis (Danglot et al. 1995). In additon to alternative splicing , NF1 is subject to RNA editing. Nucleotide 2914 can be deaminated from a cytosine to a uracil, leading to a premature end to translation (Skuse et al. 1996). This may be a form of inactivation of the normal NF1 alele, since varying levels of edited NF1 mRNA have been found in dermal discrete and plexiform neurofibromas and astrocytomas. A higher range of editng was observed in more malignant tumours compared to benign tumours (Cappione et al. 1997). 1.9 Neurofibromin function Sequence analysis of NF1 shows a region of marked homology to the catalytic domain of mammalin GTPase-Activating Protein (GAP) and yeast proteins that can downregulate p21-Ras activity (Xu et al. 1990). Functional studies have shown that both 17 ful-length neurofibromin and only the GAP-related domain (GRD) inactivate p21-Ras by stimulating its intrinsic GTPase activity (Bolag and McCormick 1991). The role of activated Ras is complex and varies among diferent cel types. Anchored on the cytoplasmic side of the mammalin plasma membrane, activated p21-Ras can stimulate proliferation through the Raf-MAK (mitogen activated kinase) pathway and can also inhibit apoptosis through the phosphoinositol 3' kinase (PI3 kinase) pathway. NF1 mutations generaly result in diferent tumours than Ras mutations (Mulvihil 1994), suggesting that other properties of neurofibromin are also involved in tumourgenesis. Neurofibromin has other biochemical properties that involve its GRD. Neurofibromin GRD interacts with other Ras-like proteins, such as R-Ras (Rey et al. 1994) and with tubulin (Bolag et al. 1993), which contains a domain similar to Ras (Pai et al. 1990; Nogales et al. 1998). This could represent an additonal activity of neurofibromin or a means by which its interaction with Ras is regulated (Viskochil 1999). In addition, neurofibromin GRD activity is diferentialy inhibited by several diferent lipids (Golubic et al. 1991) and prostaglandins (Han et al. 1991). The NF1 homologue in Drosophila melanogaster acts as an activator of the cAMP pathway as wel as a negative regulator of Ras (Guo et al. 1997). NFF1' mutant flies can be rescued not only by an NF1 transgene but also by expression of activated cAMP-dependent protein kinase A (PKA) (The et al. 1997). This suggests that PKA functions downstream of or paralel to neurofibromin in Drosophila. Neurofibromin may also interact with the cAMP pathway in humans. Analysis of NF1 mutations in NF1 patients suggests a second functional domain upstream of the GRD (Fahsold et al. 2000). This cysteine/serine-rich domain contains PKA binding sites (Marchuk et al. 1991). This 18 domain is subject to phosphorylation (Izawa et al. 1996) and may influence neurofibromin's interaction with microtubules (Gregory et al. 1993) and cAMP-dependent signaling (Guo et al. 1997; The et al. 1997). The finding of somatic mutations, or second "hits", in neurofibroma Schwann cels (Kluwe et al. 1999; Rutkowski et al. 2000; Sera et al. 2000), optic gliomas (Gutmann et al. 2000), pheochromocytomas (Xu et al. 1992), JCML (Bolag et al. 1996), and malignant peripheral nerve sheath tumours (Skuse et al. 1989; Skuse et al. 1991) in NF1 patients supports the role of neurofibromin as a tumour suppressor. However, loss of heterozygosity has not been extensively studied in other NF1 features, such as cafe-au-lait spots, freckles or Lisch nodules. Mutation of both NF1 aleles may be involved in the development of some NF1 features, but not for others. It is not known exactly how abnormal or insuficient neurofibromin can trigger the clinical features typicaly found in NF1 patients, but Ras and cAMP pathways may be involved. Haploinsuficiency of neurofibromin increases growth of mast cels (Ingram et al. 2000), which may be involved in neurofibroma formation (Riccardi 1990). Altered growth has also been demonstrated for keratinocytes resulting in hyperpigmentation and skin cancer among Nfl+/~ mice exposed to chemical carcinogens (Atit et al. 2000). A higher ratio of active to inactive Ras has been found in neurofibromas and malignant peripheral nerve sheath tumours (Guha et al. 1996). Activated PKA is known to stimulate proliferation in some cel types and may normaly contribute to body growth (Miyazaki et al. 1992; Kim et al. 1997). This may account for abnormalites in NF 1 such as a smaler skeleton in patients with short stature, and abnormal bone formation in pseudarthrosis and scoliosis. Reduced NF1 expression 19 results in an increase in glial cel proliferation (Gutmann et al. 1999). Normal stimulation of the PKA pathway accelerates diferentiation and inhibits proliferation of glial cels (Raible and McMoris 1990; Raible and McMoris 1993). Deficiencies in this pathway in glial cels may be involved in optic pathway gliomas and brain overgrowth in macrocephaly. 1.10 NF1 gene in other species The homologue of the NF1 gene has been cloned and characterised in several organisms. The coding sequence of the Nfl locus in mouse has 98% identiy to human NF1 over 2838 amino acids (Buchberg et al. 1990; Bernards et al. 1993; Mantani et al. 1994). The Drosophila melanogaster NF1 coding sequence has 60% identiy to human NF1 over 2802 amino acids (The et al. 1997). Caenorhabditis elegans has 27% identiy over 368 amino acids (GenBank Z46266). Saccharomyces cerevisiae has 23% identiy over 1490 amino acids (GenBank Z36009). The NF1 gene has been partialy characterised in several other organisms as wel. Identification of the NF1 gene in other species has resulted in the development of several animal models. 1.11 Animal models Animal models of NF1 have added valuable insights into the pathogenesis of NF1 lesions in humans. Although curent models have demonstrated only a few of NF1 disease features observed in humans, future models may be even more useful. 20 Some of these features have been modeled in animals without directly afecting the Nfl locus. Transgenic mice expressing the human T-lymphotropic virus type 1 tax (HTLV-tax) gene develop dermal discrete neurofibromas and occasionaly Lisch nodules (Hinrichs et al. 1987; Nerenberg et al. 1987; Green et al. 1992). These neurofibromas appear to contain the same cel types as human lesions and increase in size and number during pregnancy, just as human lesions do. The tax trans-regulator represses Nfl gene expression through a cw-acting element upstream of its transcriptional start site (Feigenbaum et al. 1996). Although repression by tax occurs in the absence of mutation at the Nfl locus, the mechanism by which it regulates transcription at the Nfl locus may shed light on NF1 expression and disease pathogenesis in humans. Damselfish exhibit a naturaly occuring disorder with tumours resembling human plexiform neurofibromas and regions of hyperpigmentation resembling cafe-au-lait spots (Schmale et al. 1983; Schmale et al. 1986). The disease is infectious and may be caused by a retrovirus that downregulates function of the damselfish NF1 gene (Schmale et al. 1996). Viruses have not been implicated in the NF1 gene or disease in humans. A mouse model has been engineered with a premature truncation in its Nfl gene (Jacks et al. 1994). Mice heterozygous for this mutation (Nfl+l')do not develop any of the lesions characteristic of human NF1, while homozygous mutants (Nfl'1') have double outlet right ventricle and die in utero (Jacks et al. 1994). Chimeric mice, composed of Nfl+I~ and Nf1' cels, develop multiple plexiform neurofibromas (Cichowski et al. 1999). Compound heterozygotes for both Nfl and Trp53 mutations on the same chromosome (Nfl+/~ Trp53+I' cis) frequently develop astrocytomas and malignant peripheral nerve sheath tumours, suggesting that there is biochemical interaction or co-operation between 21 the products of these two loci (Gichowski et al. 1999; Vogel et al. 1999). Interestingly, the frequency of astrocytomas in these Nfl+/~ Trp53+/~ cis compound heterozygotes varies depending on genetic background (Reily et al. 2000), suggesting that other genetic factors are also involved. Drosophila melanogaster NF1 mutant heterozygotes are phenotypicaly normal, but Drosophila homozygous for either of two particular NF1 nul mutants have a 20-25% smaler body size ( see section 1.9) (The et al. 1997). A subset of Holstein catle develops multiple cutaneous neurofibromas, and the trait was found to segregate with a polymorphism near the bovine NF1 locus (Sartin et al. 1994). However, no specific mutation of the bovine NF1 gene has yet been reported. 1.12 Variable expressivity One of the most remarkable aspects of NF1 in humans is the variabilty in its phenotype. This variabilty afects cardinal clinical features as wel as disease complications. Although most people with NF1 are not severely afected, some develop a wide variety of serious features and complications. "NF1 can be manifest in so many diferent ways and in so many tissues and organs from one family to another, from one person to another within a given family, and from one body part to another and from one time to another in a given person." (Riccardi 1992). This variabilty confounds clinical management and limits the predictive abilty of genetic counseling for afected familes. Determining the causes of the variable expressivity in NF1 is a logical step toward elucidating its pathogenesis and improving treatment. 22 Present knowledge about the natural history of NF1 comes mostly from hospital records (Crowe et al. 1956), specialised clinics (Riccardi 1992), smal population samples (Carey et al. 1979; Samuelsson and Axelsson 1981; Huson et al. 1989a), and regional (McGaughran et al. 1999) and international (DeBela et al. 2000b; Szudek et al. 2000b; Szudek et al. 2000a; Szudek et al. 2000c) databases. Some features of NF1 can vary over an afected individual's lifetime. Cutaneous neurofibromas usualy accumulate with age, while cafe-au-lait spots develop at an early age and then become lighter in colour in older individuals. Optic gliomas often undergo a short period of rapid growth but can then remain dormant and asymptomatic (Listernick and Gutmann 1999). The disease phenotype is also afected by systemic developmental changes. Adolescence is a time when neurofibromas often become apparent for the first time and/or those already present grow in size (Riccardi 1992). In a study of 105 women with NF1, sixty-four reported growth of new neurofibromas during pregnancy, and fifty-five noted enlargement of existing neurofibromas (Dugof and Sujansky 1996). A few studies looked for but did not find significant associations between common clinical features in individuals. In a longitudinal study, Zoler et al. (1995) found that rapid progression at one point in life does not necessarily predict severity later. Easton et al. (1993) found no associations among clinical features within 175 individual NF1 patients. Specificaly, no statisticaly significant relationship was observed among the folowing clinical features: number of cafe-au-lait spots, number of dermal discrete neurofibromas, and presence or absence of plexiform neurofibromas, optic gliomas, scoliosis, seizures and remedial education. The study minimised the confounding efects 23 of age but examined only a smal number of NF1 patients, many of whom were related. Furthermore, the study considered only pair-wise associations between features and not associations among several diferent features at once. The disease is also inconsistent among afected relatives - more so than can be accounted for by individual covariates such as age. Carey et al. (1979) examined 104 NF1 patients from 30 familes with two or more afected relatives and found that three-quarters of familes showed significant diferences in severity of NF1 between afected individuals. Riccardi et al. (1979) examined 221 NF1 patients from 127 familes and found that afected children of individuals with minimal or mild NF1 had a 25% chance of having a moderate or severe phenotype. A population-based study in south-east Wales examined 135 NF1 patients in 69 familes (Huson et al. 1989b) and, although intra-familial variabilty was not assessed quantiatively, the authors found that the severity of NF1 is "extremely variable" within familes. Huson (1994) compiled the clinical features of 19 pairs of monozygotic twins with NF1 from the literature. Twins are exactly the same age, so a high degree of concordance might be expected by chance alone. Nevertheless, a significant portion of the monozygotic pairs were discordant for individual features, particularly for plexiform neurofibromas, optic gliomas, scoliosis, seizures, and learning dificulties. Most clinical studies of familal associations have based their findings on a relatively subjective assessment without appropriate statistical analysis. The most statisticaly sophisticated study to date is by Easton et al. (1993), who looked for familal clustering of 8 diferent NF1 features in 175 afected individuals from 48 familes. The . results suggest that there may indeed be genetic influences on phenotypic variabilty in 24 NF1, but also that intra-familal variabilty is the rule rather than the exception. Most of the features examined, particularly plexiform neurofibromas, optic gliomas and scoliosis, were more often discordant than concordant between afected relatives. 1.13 Possible sources of variable expressivity The folowing are possible sources of phenotypic variabilty in NF1: I) Allelic heterogeneity is the presence of diferent mutant aleles at the same locus, each capable of producing an abnormal phenotype. An example is the fibroblast growth factor gene, FGFR2. Apert syndrome, characterised by craniosynostosis and syndactyly, results from specific mutations in the FGFR2 gene (Wilkie et al. 1995). Crouzon syndrome, characterised by craniosynostosis but normal limbs, results from mutations in a diferent portion of the same gene (Reardon et al. 1994). Jackson-Weiss syndrome, in which craniosynostosis is associated with broad haluces (big toes), tarsal-metatarsal fusions and syndactyly, results from a diferent mutation in the same exon of FGFR2 as Crouzon syndrome (Jabs et al. 1994). Also, mutation of the whole dystrophin gene, or at least one of two specific regions, results in a severe disease - Duchenne muscular dystrophy. If the mutation is . limited to a diferent region of the dystrophin gene, the result is a relatively mild disease - Becker muscular dystrophy (Thompson et al. 1991). Muler et al. (Muler et al. 1997) and Mehler (Mehler 2000) reviewed the molecular genetics of these syndromes. More than 400 diferent constiutional NF1 mutations have been reported (Korf 1999b; Fahsold et al. 2000; Messiaen et al. 2000). A more or less consistent phenotype (facial anomalies, learning disabilty or mental retardation, and large numbers of dermal 25 discrete neurofibromas) occurs in association with deletions involving the entire NF1 gene (Tonsgard et al. 1997; Dorschner et al. 2000), but little other evidence has been found of alele-phenotype corelations in NF1. Similar clinical features have been observed among afected members of a few familes with the NF1 variants Watson syndrome (Alanson et al. 1991), familial cafe-au-lait spots (Abeliovich et al. 1995) or familal spinal neurofibromas (Pulst et al. 1991; Poyhonen et al. 1997a; Ars et al. 1998). These observations are consistent with an alele-phenotype correlation, but no particular kind of NF1 mutation has been found in familes with these or other phenotypic variants. Likewise, diferent phenotypes can result from the same mutation. Upadhyaya et al. (Upadhyaya et al. 1996) reported two patients with very diferent NF1 phenotypes who both have the same mutation in exon 37 of the NF1 gene. Although alelic heterogeneity may be a factor in the variable expressivity in NF1 (diferences in the rest of the NF1 gene were not ruled out), other factors may also be involved. 2) Polymorphic variations in the normal allele could also modify phenotypic efects of the mutant alele. One example of such an efect is the Rh blood factor. Rh-positve blood specimens usualy react with anti-Rh serum. Rh-negative specimens usualy do not. In some cases, blood specimens give an intermediate, rather than a strong positve or strong negative response to anti-Rh D serum. In several familes, this intermediate response was observed only in members having a particular homologous alele at the Rh complex (Vogel and Motulsky 1997). Although cystic fibrosis is recessively inherited and NF1 is dominantly inherited, the cystic fibrosis transmembrane conductance regulator (CFTR) gene ilustrates how variations in the level of normal transcripts can afect disease severity. A significant 26 corelation exists between the level of the normal CFTR transcripts and the severity of lung disease. A splicing variant of the CFTR gene results in varying levels of normal transcript among diferent individuals and between diferent organs of the same individual (Rave-Harel et al. 1997). Since alternate splicing has also been demonstrated for NF1 mRNA (Metheny et al. 1995), such a variant could contribute to the observed variable expressivity in NF1 patients who cary it as their normal NF1 alele. 3) The action of modifying genes has been used to explain variable phenotypes in the haemoglobin system, retinitis pigmentosa, cystic fibrosis, and many other genetic diseases. Sickle-cel anemia, caused by homozygosity for a mutation in the P-globin gene, is less severe when accompanied by hereditary persistence of fetal haemoglobin (HPFH), a group of genetic disorders at other loci that produce deregulation of the y-globin gene. HPFH increases the amount of fetal haemoglobin (HbF) in afected red cels (Odenheimer et al. 1987; Piat et al. 1991), which reduces the aggregation of a-globin triggered by the imbalance resulting from the p-globin mutation. As a result, patients who are homozygous for the sickle cel mutation but have large amounts of HbF have few or no symptoms of sickle-cel anemia. Similarly, sickle-cel disease is associated with less anemia and improved survival in patients with alpha-thalassemia, usualy caused by deletions at the Hb a locus (Weatheral 2001). Peripherin-2 and rom-1 proteins assemble into oligomers involved in photoreceptor development. Individuals who inherit particular peripherin-2 as wel as rom-1 mutations are aflicted with retinitis pigmentosa, whereas individuals who inherit only one defective gene at either locus are normal (Loewen et al. 2001). 27 As mentioned above, variabilty of splicing the CFTR gene product has been corelated with the extent of lung disease. Inter-individual and organ diferences in CFTR splicing (Rave-Harel et al. 1997) may be modulated by overexpression of celular alternative splicing factors (Nissim-Rafinia et al. 2000). In another example, calcitonin causes calcium and phosphate absorption from the blood and depositon in bone. Calcitonin mRNA is processed diferently in the hypothalamus than in the thyroid, resulting in diferent amounts of two protein variants in the two tissues (Amara et al. 1982). This process is also influenced by splicing recogniton factors (Lou et al. 1998). Diferentialy spliced forms of the NF1 transcript have been observed in diferent tissues and stages of development. Since splicing can be influenced by genetic modifers (Nissim-Rafinia et al. 2000), an individual's splicing machinery may be suficiently variable to contribute to the phenotypic variabilty of NF1 (Metheny et al. 1995). Another argument for the likely importance of genetic modifers can be made on the basis of neurofibromin's function. Since neurofibromin downregulates p21ras (Xu et al. 1990), other known regulators of the Ras signal transduction cascade or variants of Ras itself may modify neurofibromin's influence and thus the disease phenotype. In addition, pseudogenes with partial homology to NF1 have been found on several chromosomes (see section 1.7). Expression of nitric oxide synthase is suppressed by RNA transcribed from a pseudogene in the snail (Korneev et al. 1999). If a homologous NF1 gene were found to be expressed, it is possible that its product influences NF1 expression. By studying related NF1 patients, Easton et al. (1993) corelated genetic relatedness with the likelihood of concordance of NF1 features such as cafe-au-lait spots, 28 neurofibromas, scoliosis and optic gliomas. Phenotypic concordance between relatives increased with increased genetic similarity, supporting a modifying efect. The higher frequency of tumours observed among women with NF1 than among afected men (Samuelsson and Axelsson 1981; Airewele et al. 2001) and the higher frequency of pseudarthrosis observed among males (Stevenson et al. 1999) might also be due to genetic modifers that are diferentialy expressed in men and women. 4) Epigenetic effects are factors that can afect the phenotype without change in the genotype. They include imprinting, which is the diferential expression of genetic material depending on whether the material has been inherited from the mother or the father. An imprinted gene cluster at 1 lpl5.5 has been implicated in Beckwith-Wiedemann syndrome (BWS), a congenital overgrowth disorder (Reik and Maher 1997). The genetics of BWS are complex, and there are 3 major sub-groups of patients -familial, sporadic and those with cytogenetic abnormalites (Maher and Reik 2000). In each of these subgroups, BWS pathogenesis depends on parent-of-origin efects. Maternal transmission is associated with greater penetrance in familal cases. Some sporadic cases are associated with paternal uniparental disomy. Among those with cytogenetic abnormalites, duplications are derived from the father, whereas inversions and translocations are derived from the mother. Albright hereditary osteodystrophy (AHO) also can be divided into two subgroups: those with features of AHO alone and those with AHO and associated resistance to multiple hormones (Weinstein and Yu 1999). Patients from both groups ,have a similar -50% deficiency of the Gs protein and are often found within the same 29 family (Levine et al. 1986). This variable expressivity is due to tissue specifc-imprinting of the Gs gene (GNAS1) (Weinstein and Yu 1999). Maternal transmission of the disease leads to hormone resistance while paternal transmission does not (Davies and Hughes 1993). In NF1, Crowe et al. (1956) and Miler and Hal (1978) suggested that maternal inheritance of the mutation results in a more severe overal phenotype than paternal inheritance. Miler and Hal studied 62 patients from one centre, and the statistical signifcance of their observation relies on an excess of very severely afected ofspring born to afected mothers. There is no evidence that this group was representative of the NF1 population as a whole. Other investigators, using much larger data sets, did not find a parent of origin efect (Carey et al. 1979; Riccardi and Wald 1987; Huson et al. 1989a). Shannon et al. (1992) examined parent of origin of the NF1 mutation in 21 children with NF1 and juvenile chronic myelogenous leukemia (JCML). JCML developed in 12 boys and four girls who inherited NF1 from their mothers, and in five boys who inherited the disease from their fathers. Father-to-daughter transmission was not observed. This finding could be explained by imprinting of NF1, but requires further investigation. 5) Non-genetic factors, such as environmental exposures and trauma, may also be responsible for some of the variabilty in NFL For instance, tissue injury seems to play a role in the development of at least some neurofibromas (Riccardi 1992). Systemic changes during pregnancy (Dugof and Sujansky 1996) and adolescence (Riccardi 1992) lead to changes in the number and/or size of neurofibromas. Infectious agents can influence NF1 expression in mice and animals (see section 1.11), but this phenomenon 30 has not been observed in humans. By far the most important non-genetic factor afecting the NF1 phenotype is age. Riccardi (1992) has noted that, for example, the number of peripheral neurofibromas increases with age. Chance seems to account for some ofthe phenotypic variabilty of NF1 (Riccardi 1992; Easton et al. 1993). Evidence of loss of heterozygosity at the NF1 locus in neurofibrosarcomas, peripheral neurofibromas and several other tumours (see sections 1.5 and 1.9) suggests that random "second hit" mutations in somatic tissues may be important in development of the tumours characteristic of NF 1. Scoliosis can be caused by plexiform neurofibromas exerting pressure on the spine. The location of a plexiform neurofibroma under the skin is probably afected by chance and, therefore, chance is likely to be involved in the development of scoliosis in many cases (Riccardi 1992). For haemoglobinopathies, cystic fibrosis and craniofacial syndromes, there is evidence that more than one ofthe above five possibilties plays a role in the observed variable expressivity (Muler et al. 1997; Mickle and Cuting 2000; Serjeant 2001). These diseases are very diferent from NF1, but they demonstrate the diferent genetic mechanisms that can contribute to variable expressivity in humans. The work on NF1 is not as developed as for these diseases, but it is likely that several ofthe factors from this list and interactions between them contribute to its variability. 31 1.14 Hypotheses The hypotheses tested in this thesis are: 1) that subgroups of NF1 patients can be identified who are more likely to develop particular disease features, 2) that some NF1 features cluster in certain familes, and 3) that familial aggregation of features, and thus expressivity of NF1, is influenced by genetic factors, such as alelic diferences or modifying genes. 1.15 Objectives My objectives were: 1) to estimate risks of various disease features in clinicaly-defined subsets of NF1 patients, 2) to test for and estimate associations between the occurence of features in individual NF 1 patients and afected family members, and 3) to examine genetic efects on the occurence of disease features among relatives with NF1, in order to gain insight into the sources of the variable expressivity. 32 2. A S S O C I A T I O N S O F C L I N I C A L F E A T U R E S IN N E U R O F I B R O M A T O S I S 1 33 2.1 Hypotheses Clinical features of neurofibromatosis 1 (NF1) do not occur independently in: (1) afected individuals or (2) between afected relatives. 2.2 Objectives To test for and estimate: (1) associations between pairs of clinical features in NF1 probands, and (2) associations of individual features between afected parents and children. 2.3 Introduction Substantial variabilty is wel documented in NF1 among afected individuals and familes (see section 1.13). Nevertheless, there is evidence of intra-familal corelations for several NF1 clinical features (Easton et al. 1993) and a few rare NF1 phenotypes appear to "breed true" in familes (Alanson et al. 1991; Pulst et al. 1991; Abeliovich et al. 1995; Poyhonen et al. 1997a; Ars et al. 1998). Recogniton of clinical heterogeneity within a disease may provide important pathogenetic insights. For example, understanding that neurofibromatosis 1 and 2 are diferent diseases was a seminal contribution (Riccardi 1982). In order to determine if clinical heterogeneity exists within NF1 itself, I tested three large clinical data sets for associations between pairs of clinical features in probands. I also tested for familal determinants of clinical variabilty by looking for associations of individual clinical features between parents and children with NF 1. 34 2.4 Subjects and methods 2.4.1 Patients and data description All patients included in this analysis were diagnosed with NF1 according to established clinical criteria (Table 1.1) (NIH 1988; Gutmann et al. 1997). The study was performed using clinical data from three independent sets of NF1 patients. At the time of this analysis, the National Neurofibromatosis Foundation International Database (NFDB) (Friedman and Birch 1997b) contained descriptions of 2509 NF1 probands, 211 afected parents and 289 of their afected children. 83% of the NF1 cases are Caucasian, 7% Asian, and 4% African-American. The remaining 6% are mostly combinations of these three ethnic groups. The Neurofibromatosis Instiute Database (NFID) (Riccardi 1992) includes standard clinical information on 774 NF1 probands, 132 afected parents and 189 of their afected children. 72% of the cases are Caucasian, 14% Hispanic, 13% African-American and 1% Asian. The Manchester NF1 database (MANF1) is a population-based registry of north-west England and includes clinical information on 270 probands, 94 afected parents and 140 of their afected children (McGaughran et al. 1999). 92% of the cases are Caucasian, 4% East Indian, 2% Black, 1% Bangladeshi and 1% Pakistani. These three databases contain information on many of the same NF1 features. There is no known overlap among the patients included in these three databases. Specifc NF1 mutations have been identified by molecular analysis in fewer than 1% of these patients. 35 2.4.2 Clinical features Twelve of the most common or important clinical features of NF1 were selected for inclusion in this study: intertriginous freckling, discrete cutaneous or subcutaneous neurofibromas (refered to as "discrete neurofibromas"), difuse or nodular plexiform neurofibromas (refered to as "plexiform neurofibromas"), learning disabilty or mental retardation, Lisch nodules, scoliosis, tibial or other long bone bowing or pseudarthrosis, optic glioma, macrocephaly, short stature, seizures and neoplasms (other than neurofibromas or optic glioma). Cafe-au-lait spots were not included in this study because they were coded as "present" in al subjects in the NFID, which is incompatible with pair-wise analysis (see below). Table 2.1 summarises the prevalence of these 12 features in the three databases. Most of the features were identified by physical examination. Discrete neurofibromas were coded as "present" if the subject had two or more cutaneous or subcutaneous neurofibromas. Short stature was coded as "present" if the subject's height was two or more standard deviations below the age- and gender-matched population mean. Subjects with pseudarthrosis, early or delayed puberty, scoliosis, vertebral dysplasia, or spinal compression were excluded from analyses involving height. Macrocephaly was coded as "present" if the subject's head circumference was two or more standard deviations above the age- and sex-matched population mean. Subjects with plexiform neurofibroma of the head, early or delayed puberty, or hydrocephalus were excluded from analyses involving head circumference. Lisch nodules were diagnosed or excluded by a slit lamp examination. The presence or absence of optic glioma was determined by cranial MRI or CT examination; individuals who did not have 36 cranial imaging were coded as "unknown". Only patients who had definite presence or absence of a feature were considered in comparisons involving that feature. 2.4.3 Statistical analysis in individual probands Pair-wise combinations ofthe presence or absence of each feature were analysed in probands by 2-by-2 tables using SAS (1996). The prevalences of many features of NF1 increase with age (Riccardi 1992; Cnossen et al. 1998). Two features that both increase with age may show a strong association because older patients are likely to have both features and younger patients are likely to have neither. Therefore, patients from each database were stratified into 5-year age groups to reduce confounding by age. Patients were also stratified by gender, but not by race because the number of non-Caucasians is heterogeneous and too smal for useful comparison. The method of Mantel and Haenszel (1959) was used to estimate a summary odds ratio with 95% confidence intervals over the age and gender strata. This method weighs the proportion in the table for each stratum by its respective sample size, and, therefore, is robust to large diferences in sample size among the strata (Everit 1992). However, it is not likely to be informative in cases where the relationship between the two variables being compared is very diferent among the strata. Therefore, the Breslow-Day method (SAS 1996) was used to test each triad of odds ratios for homogeneity between the three databases. The test compares the diference between the Mantel and Haenszel summary odds ratio and the odds ratios for the individual strata. Triads with p-values greater than 0.05 were considered homogenous. I performed the analyses in three independent data sets (NFDB, NFID and MANF1) because I expected to observe many associations that 37 s J> CD 00 cu «g C*M r^ 5- cu c es -Q o s-CM *g cu $ £ «2 — 53 cu T3 fi CS o i-of) "O cu <u > *- .-. 53 v < & so -O fi CS -Q O i-P. N ( N ^ ^ — cn cn t— vo Sf .-C r-1 VO CU S •M CS CU Is w ^ O N O O N O i n ^ - v f - -O ^ O T t i n r - i o m OO ON OO i n CN ^H i n CN CN ,-1 VO »-i O N O • ^ d O N cn t-~ ^H rn ^ ON VO cn i >—i m oo i n o o c N O ^ o r - - r - ~ m CN CN (N i rn CS >H (N vo vo -si- o 'sj- oo cn cn CN 1 ' CN O ON VO ON o i n ON © cn N C N r n vo cn cn ^ ^ cn cn vo oo oo "sf VO oo r--oo cn i n r-- i n i n CN ON 00 i n r- oo <—i cn CN vq od ON 00 CN vo i n cn CN r--cn i n CN CN O cn cn VO r-H vo r-OO 00 r - i n cn c--l n ON cn cn oo O CN o cn cn ^ ON CN i n oo i n vo i n o IT) VO i n CN oo i n ON O >n vo oo © cn CN O c-- i n O ON CN ON it 'st <N CN CN i n o ON CN cn vo cn oo oo vo ^ O O N o o o i n —' CN CN ON ON oo ON cn ON CN O VO r-l d - - i CN ON i n 00 ON o CN CN CN i n i n i n oo CN OO m CN i n CN vd i n fi M o to kH oo +-» CD k-O on <D O fi o oo Q P H h-l ca a o '-fi oo fi 'S OH o Q 00 O M fi CD oo P H oo o ca OH CD o o kH o ca CD 3 o 43 00 oo o CN ON r~-d CN ON i n o i n CN cn cn o i n vo vo cn CN r-- i n VO vo o i n cn i—i cn cn^ cn s ' s — " oo cn r r oo m cn ON ON i n o O i n ON m CN, T-H CS vq vo i n d CN vo CN CN vd 00 a 00 _ca "S, o CD z, reached nominal statistical signifcance by chance as a result of multiple comparisons. Odds ratios with 95% confidence intervals that excluded 1.0 in at least two of the three databases were considered unlikely to be due to chance alone. The chance of observing statistical signifcance for a given pair-wise comparison in two of the three databases and not the third is 3x0.05x0.05x0.95=0.007. The likelihood of signifcance in al three databases is relatively negligible. Since 66 diferent pair-wise comparisons were made, 66x0.007=0.47 comparisons are expected to reach statistical signifcance by chance alone. 2.4.4 Statistical analysis in afected parents and children The second analysis included afected relatives. A feature that increases with age may show a strong intra-familal association because the ages of sibs within a family are usualy similar. Therefore, I limited my analysis to parents and children, who usualy difer in age by at least 20 years. For each of the 12 features, a 2-by-2 table was used to compare the frequency of a given feature in children with NF1 of parents with NF1 who had the feature to the frequency in children of parents who lacked the feature. Each individual was counted only once. Twelve contingency tables were generated separately in each database. Odds ratios with 95% confidence intervals were calculated for contingency tables without blank cels. The Breslow-Day method (SAS 1996) was used to test for homogeneity, and the Mantel-Haenszel method (Mantel and Haenszel 1959) was used to estimate summary odds ratios. 39 2.5 Results 2.5.1 Associations in individuals Pair-wise associations between each of the 12 clinical features were tested in 2509 NF1 probands from the NFDB, 774 NF1 probands from the NFID and 270 NF1 probands from the MANF1 database (Table 2.2). Most of the associations are moderate in strength - positve associations generaly have odds ratios in the range of 2.0-3.0 and negative associations have odds ratios in the range of 0.3-0.5. In the NFDB, an odds ratio of 1.0 was excluded from the 95% confidence limits for 26 of 66 associations tested. There were 23 nominaly significant positve associations and 3 nominaly significant inverse associations. In the NFID, which contains fewer than 1/3 as many cases as the NFDB, an odds ratio of 1.0 was excluded from the 95%) confidence limits for 13 of 66 associations tested. Ten of these nominaly significant associations were positve and 3 were negative. In the MANF1, which is about 1/9 as large as the NFDB, an odds ratio of 1.0 was excluded from the 95% confidence limits for 5 of 55 associations tested. All of these were positive. Odds ratios could not be calculated in the remaining 11 associations due to blank cels in the contingency tables. Overal, 6 of 66 tested associations between pairs of features are statisticaly significant and in the same direction in at least two of the databases (Table 2.2). Four of these 6 associations are statisticaly homogenous (p>0.05) between the three databases. One statisticaly significant inverse association was observed in at least two independent databases. The associations are shown in Table 2.2 as odds ratios for each database and as summary odds ratios for al three databases together. 40 2.5.2 Parent-child associations Table 2.3 summarises the associations for occurence of the 12 features between: 211 NF1 parents and 289 of their NF1 children from the NFDB; 132 NF1 parents and 189 of their NF1 children from the NFID; and 94 NF1 parents and 140 of their NF1 children from the MANF1. The associations are expressed as odds ratios for each database and as summary odds ratios for al three databases together. Odds ratios could not be calculated for one association in the NFBD, three associations in the NFID and two associations in the MANF1, due to blank cels in contingency tables. A summary odds ratio of 1 was excluded from the 95% confidence limits for Lisch nodules, optic glioma, learning disabilty or mental retardation, macrocephaly and short stature, but not for intertriginous freckling, discrete neurofibromas, plexiform neurofibromas, seizure, pseudarthrosis, scoliosis or neoplasms. Three of the five statisticaly significant associations are homogenous between the three databases. No significant negative associations were observed. 2.6 Discussion 2.6.1 Associations in individuals The large number of cases in these three databases enabled me to find significant associations between several common features of NF1 (Table 2.2). The concordance between the findings in the three independent databases is remarkable. About three (p=0.05 multiplied by 66) nominaly statisticaly significant associations were expected by chance in each database, and one would expect chance associations to difer in the 41 pH T3 X> O T3 '"5 x co T3 a i a> bo c3 bo o fi CO <D 1-o ^ S a .2 x .2 o o o o. co **" co C « •"• b oo .O 11 «JH —1 rv <-> °^ e O O ° as in "3 ON 3^ CO ii 1 -S co a -a OS T3 H O O • 1—I <u O a o 3 C/3 <D S-H to o (Zl »—1 <N OO 00 >—1 O O r- <—1 0 0 O O O O ^ >/"> 0 -+ r-- t ' -t r 1 r'l — ft—< —• — — - ' fN — — ro r i r i t?" 3 r i r i >/-! O — r'l *o —I , ! , ! —- •— " —^ —-r ] f. oc r-i -—s ^ s 0 oc ri <N f. r'l ir, — r i — — : oc >r, r 1 — r i — fN CO <u T3 O o CO bO n o CO X co CL) O o CO co CO O M -a o CO CO CO CO PL, z ii -t-> -4-» <u <U U i-l O O O CO CO CO CO o co PL Q Q PH Q 1—I co ii S-H 00 I c o CN <L> c3 3 CL. CO - H c So o ,0 T3 ° U 60 .s .g o Q. <U - fci XI o 3 o « o =3 B & nj E ru C T3 O T3 ca O c u 'S 60 to ca -a 60 ±3 ro rt o § g £ •c 60 3 .. A c G. S •o O « > S U Q PH .fi e CD -fi o c CD c3 PH fi CD CD • CD C X 2 >» -e 2 -3 fi -H C3 O CI O a >» .S 53 CJ C O 00 CS oo cu cd on fi 5^ kH ). -fi 3 <D A CI £ s ° a a O e ° N ? 15 ^ 53 i n *3 ON 00 +3 00 TH < * 00 ^1 CN cd i-H CU 1 -S 00 e3 T 3 H O IS o I ^ CO -4—> '53 a fcfj o I ^ 0f> c ^ r n r-i 4.3) 00 s' c^ <N N -~ 1 <N NO Or? rn 00 C N »A r'-i •~^  •A <A C N C N ?c£ C ^ 1— •—. 1^ C ^ <s c £ >o vo O N N O 1 C S °°. vo rn <N-i PQ Q !-CN cn i n t-» 00 0 vo m f i ON cn 0 ON d d d d d d d d •7.4) CN •5.3) •61.2 in ON f i r i vo oc O^ 0 , d d —— '— m d 10.5 rn r 1 cn •16.1) CN <A 0 , d OO vo d f i d ON^ ON" f i 00 O N •4 d d d O N 00 00 —1 d DO .fi o CD kH PH 00 PH CD H-» CD kH O 00 00 PH eg • »—I X CD 0 0 C N 1-—1 1—H 00 • 1 m V D — O O v — y •—' 00 0 ^ H in C N r-i oc 00 CD T3 O z, -fi O 00 ca a 3 o O H O 00 CD <D G O PH Q r-1 C N " * 1 vo vo C N i n • 00 d o C N 00 o fi CD 00 PH 00 'o o la -fi O H CD O o kH o ca I H-» in o -fi 00 o o •160) •6.6) •12.4) 9.2) 485) oc m in vo f , d d , d 3 00 co vq C N f i C N 25.5 cn ^ v f l CN O N 00 r-i r+ vd O N CN f i r i • (N 00 in r i ^ \ f . O d d d d ^ r-i r i CN f i r-H r-; r i O r - 1 in 1 - 1 oc •n in 00 ja o CD .a c -S s $ -o 5 2 <rj o 6 « 5 H o .5 k. Xfi T 3 O .a c 6 "3 ft c a. CO T 3 0) -D O C o o 2£ •rj OJ T3 ca C3 cd 0 s o m (fi o '2 £ .SP cj C3 'fi o n) •fi V C3 -73 0 0 O c o -a S3 c u E kH O 3^ C3 CL> ccj T3 ca o n k. '-H e G O CD 0) kH — 0 0 3 • ; § £ a 43 NFDB, NFID and MANF1. The reproducibilty of my results suggests that these associations are probably not due to chance alone. Two of the associations in Table 2.2 were not homogenous among the three databases. The summary odds ratio between discrete neurofibromas and Lisch nodules ranged from 1.6 to 2.6. The summary odds ratio between discrete neurofibromas and pseudarthrosis ranged from 0.3 to 0.5 but could not be calculated in the MANF1 because the contingency table contained an empty cel. Although these ranges are not wide, the summary odds ratios should be viewed with caution. Seizure disorders can lead to cognitve deficit (Goldstein and Reynolds 1999), but a causal relationship has not been demonstrated in NF1 (Hughes 1994). The positve associations observed may reflect shared pathogenetic mechanisms underlying the associated features. For example, NF1 probands with seizures may be more likely also to have learning disabilties or mental retardation than patients without seizures (Table 2.2) because the efect of the NF1 mutation on brain development is greater in patients who have seizures. The association observed between the occurence of plexiform and discrete neurofibromas (Table 2.2) is consistent with the histopathological similarity between these lesions (Harkin and Reed 1969; Burger and Scheithauer 1994). In addition, both kinds of neurofibromas are associated with acquired somatic loss or mutation of the normal NF1 alele in at least some cases (Sawada et al. 1996; Sera et al. 1997). NF1 patients who develop plexiform neurofibromas usualy do so during childhood (Riccardi 1992). In contrast, discrete neurofibromas are uncommon in young children but are almost universaly present among adults with NF1. The association I observed is much 44 stronger in younger than in older NF1 patients. The odds ratio was 6.9 among patients under five years old, 3.1 among those 5-9, but only 1.3 among those over 40. This raises the interesting possibilty that NF1 patients with plexiform neurofibromas develop discrete neurofibromas earlier than patients without plexiform lesions - a hypothesis that can only be tested with longitudinal data. Previous studies found both discrete neurofibromas and intertriginous freckling to be more common in NF1 patients with Lisch nodules than in NF1 patients without Lisch nodules (Pietruschka 1961; Zehavi et al. 1986), but the responsible mechanism is unknown. Lisch nodules (Pery and Font 1982) and freckles (Fitzpatrick 1981) are derived from cels of melanocytic origin, and al three lesions involve cels derived from the embryonic neural crest (Weston 1981). This is consistent with the suggestion that NF1 is a neurocristopathy (Huson and Hughes 1994) but does not explain why other neural crest-derived tissues, in which the NF1 gene is expressed, such as the sympathetic ganglia, thyroid C-cels, and parathyroids, are rarely involved in NF1. Moreover, the involvement of neural-crest derived tissues has not been reported for many NF1 features, such as learning disabilities, dysplastic scoliosis, and tibial pseudarthrosis. Although plexiform neurofibromas growing near the spine can cause abnormal curvature and result in scoliosis, the association I observed between plexiform -neurofibromas and scoliosis does not lose signifcance when patients known to have plexiform neurofibromas of the trunk are excluded. Furthermore, two diferent forms of scoliosis may occur in NF1 patients - a dystrophic form that occurs within the first decade of life and is often severe and rapidly progressive, and a milder form that occurs later and resembles idiopathic scoliosis (see section 1.5). The association I observed 45 involves only early onset scoliosis. Many cases of NF1 come to atention because of pseudarthrosis or scoliosis, but it is not clear why these patients are more likely to also have discrete neurofibromas than patients who lack pseudarthrosis or scoliosis. The pathogenetic basis for these associations is obscure. Most of the associations observed among probands in this study are moderate in strength - positve associations generaly have odds ratios in the range of 2.0-3.0 and negative associations have odds ratios in the range of 0.3-0.5 (Table 2.2). Such pair-wise associations are not strong enough to be useful clinicaly for predictive classification of patients. Furthermore, the diagnosis of NF1 requires the presence of two or more clinical criteria (Table 1.1) and features included in the criteria are not completely independent from one another. This confounding factor wil be addressed in analyses of afected non-probands (Chapter 4), who require the presence of only one clinical feature for diagnosis. Nevertheless, my observation of similar associations in three independent databases strongly suggests that common disease features do not occur entirely at random in NF1 and that some patients are more likely than others to develop particular features. This interpretation contrasts with the view that any NF1 patient may develop any manifestation of the disease (Bernhart and Halperin 1990; Riccardi 1992). Most of the associations I observed have never been noted before. 2.6.2 Parent-child associations My observations in probands suggest that shared pathogenetic mechanisms underlie several common features of NF1. If genetic factors influence these pathogenetic mechanisms, one would expect familal aggregation of such features to occur. Therefore, 46 I tested for associations between the occurence of the 12 features among afected relatives. The ages of sibs within a family are usualy similar, and an association may be noted because older sib pairs are more likely to both have a feature and younger sib pairs to both lack a feature that increases in prevalence with age. Parents and children usualy difer in age by at least 20 years, so significant associations between the occurence of a feature in a parent and child are unlikely to be inflated by age confounding. Due to this age diference, I expect the odds-ratios from parent-child comparisons to yield conservative estimates of intra-familal associations. Consequently, I limited my analysis to afected parents and children. Afected parents with more than one afected child were counted more than once in each 2-by-2 table. This means that the calculated standard erors for odds ratios are underestimates of the true standard errors. The 95% confidence limits in Table 2.3 are slightly narower than the true 95% confidence limits. Several strong associations (odds ratios from 2.0 to 8.7) were found by comparing the presence or absence of the 12 features between afected parents and children (Table 2.3). The summary odds ratios for Lisch nodules, optic glioma, macrocephaly and short stature were significant and homogenous among the three databases, although the odds ratio for optic glioma could not be calculated in the NFDB. The NFDB draws its data mainly from pediatric centres and contains no family in which parent and child both have optic glioma. Nevertheless, these associations are probably not due to ascertainment bias. All subjects were assessed in specialised NF clinics, and the family was excluded from a particular analysis if the presence or absence of the feature in question was not 47 known in both the afected parent and child. The summary odds ratios for learning disabilty or mental retardation was significant but was not homogenous among the three databases. The summary odds ratio for neoplasms was not significant and not homogenous among the three databases. This may be due to diferences among centres in how the feature is diagnosed. No negative associations were found between afected parents and children. This is consistent with my hypothesis that afected relatives have a more similar NF1 phenotype than unrelated patients. The absence of negative associations also supports the statistical validity of my observations. One would expect to observe negative, as wel as positive, associations by chance, if there were realy no intrafamilal associations. I observed familial associations for the occurence of Lisch nodules, macrocephaly, short stature, and learning disabilty or mental retardation. In a previous study, Easton et al. (1993) examined 175 individuals with NF1 and found evidence of intra-familal corelations in the number of cafe-au-lait macules and neurofibromas and in the presence or absence of optic gliomas, scoliosis, seizures and referral for remedial education. Easton et al. observed no corelations for head circumference or plexiform neurofibromas. Furthermore, phenotypic similarity of these NF1 features was found to decrease with decreasing genetic similarity - a trend examined in Chapter 7. Unlike my study, the results of Easton et al. rely heavily on data from six pairs of monozygotic twins. Although the studies difer in design and found diferent associations both are consistent with the hypothesis that genetic factors influence the phenotypic expression of NF1 mutations in patients with NF 1. 48 The statisticaly significant phenotypic similarity among relatives may be evidence of an NF1 alele-phenotype correlation. Although generaly not striking inNFl, phenotypic modifcation by the nature of the mutant alele has been demonstrated in complete deletions of the NF1 gene, which tend to result in a severe phenotype (Tonsgard et al. 1997). Other genetic factors that might influence the phenotype in NF1 patients include modifying genes at other loci. First-degree relatives share half of their DNA sequences at other loci. Similarities at these other loci may contribute to the phenotypic similarities observed in familes with NF1. My findings complement those of Easton et al. (1993) and are consistent with their hypothesis that modifying genes influence the NF1 phenotype. The NF1 protein, neurofibromin, is known to interact with many other proteins, including tubulin (Bolag et al. 1993), kinases (Marchuk et al. 1991) and Ras (Buchberg et al. 1990; Xu et al. 1990). Variations in these proteins (Mot et al. 1997; Pepperkok et al. 2000; Saragoni et al. 2000) might also influence the NF1 phenotype . 2.7 Conclusion Although the NF1 phenotype is highly variable, some patients are more likely than others to develop certain disease features. Genetic factors may influence the particular phenotypic features that develop in many cases. Further clinical, epidemiological, and molecular studies are necessary to elucidate the pathogenesis of this complex disease fuly, but my investigations provide hope that some serious complications of NF1 can be predicted or prevented. 49 3. G R O W T H I N N O R T H A M E R I C A N W H I T E C H I L D R E N W I T H N E U R O F I B R O M A T O S I S 1 50 3.1 Hypothesis Changes in growth afect only a subset of patients with NF1. 3.2 Objective To analyse the distributions of and generate growth charts for stature and occipitofrontal circumference (OFC) inNFl patients. 3.3 Introduction Short stature (>2 standard deviations below the population mean) and macrocephaly (>2 standard deviations above the population mean) are more common in people afected with NF1 than in the general population (Weichert et al. 1973; Carey el al. 1979; Huson et al. 1988; Riccardi 1992). It has been suggested that short stature and macrocephaly are "al-or-none" phenomena that afect only a subset of NF1 patients (Riccardi 1992). According to this hypothesis, NF1 patients would be expected to fall into two distinct groups: (1) those whose stature is in the same normal distribution as unafected people of the same age and (2) those whose stature is decreased. NF1 patients would also be expected to fal into two distinct groups with respect to macrocephaly: (1) those whose OFC is in the same normal distribution as unafected people of the same age and (2) those whose OFC is increased. I examined the distributions of these measurements to determine whether changes in growth afect al or only a subset of patients with NF 1. I also generated centile curves for stature and OFC by age and gender. 51 3.4 Subjects and methods 3.4.1 Subjects All patients included in this study meet the NIH Diagnostic Criteria for NF1 (NIH 1988; Gutmann et al. 1997). Measurements of patient stature and occipitofrontal circumference (OFC) were obtained from the National Neurofibromatosis Foundation International Database (NFDB) (Friedman et al. 1993). At the time of this analysis, the NFDB included extensive demographic and cross-sectional clinical and anthropometric data on 569 Caucasian NF1 patients examined during 1980-98 at 14 participating centres in North America. Information was colected and recorded on each patient using a standard procedure. Patient stature was measured without shoes using a stadiometer. OFC was measured at the largest diameter over the occiput and forehead using a tape measure. The data were subjected to automated range checking and routinely screened for quality and consistency by the database administrator. Only measurements from each patient's first visit to a participating clinic were included in the analysis. Patients who were known to have one or more of the folowing features on any clinical visit were excluded from analyses of stature: pseudarthrosis (n=22, 3%), early (under 10 years) (n=13, 2%) or delayed (over 15 years) (n=51, 1%) puberty, optic glioma (n=66, 9%), scoliosis (n=98, 14%), vertebral dysplasia (n=19, 3%). The ultimate sample for analyses of stature consisted of 183 males and 202 females. Patients with one or more of the folowing features were excluded from analyses of OFC: plexiform neurofibroma of the head (n=46, 6%), early or delayed puberty, optic glioma, or hydrocephalus (n=23, 3%). The ultimate sample for analyses of OFC consisted of 216 males and 220 females. 52 3.4.2 Reference populations Standard population norms for stature by age were obtained from the National Center for Health Statistics (NCHS) studies during 1963-74 (Hamil et al. 1977). The NCHS standards are based on a sample consisting of 83 percent White or Hispanic subjects and 17 percent Black subjects living in the United States. Standard population norms for OFC by age were obtained from the Fels Instiute study conducted during 1929-75 (Hamil et al. 1977). The Fels Instiute sample is slightly less heterogeneous than the NCHS sample. 3.4.3 Distribution analysis Stature and OFC measurements were standardised using z-scores to control for both age and gender: z = (measurement of patient) - (mean of the sex- and age- matched control group) standard deviation of the sex- and gender-matched control group Patients with stature and OFC measurements coresponding to a z-score with an absolute value greater than 7 were excluded to minimise data entry errors. Four (1%) stature and 2 (0.5%) OFC measurements coresponded to z-scores below -7. One (0.3%) stature and 2 (0.5%) OFC measurements coresponded to z-scores above 7. Single data entry, as used in this study, has an eror rate around 2% (Reynolds-Haertle and McBride 1992; Gibson et al. 1994; Horbar and Leahy 1995). After these exclusions, I expect that about 1% of the remaining measurements contain errors. I tested the standardised data by analysis of variance to determine if significant 53 diferences exist among the measurements made by the major contributing centres. Distributions of z-scores for stature and OFC were ploted in histograms using SAS (1996). Each histogram is based on the z-scores compiled from males and females of al ages. In addition, the deviation from unimodality of each distribution was quantifed by computing its dip statistic (Hartigan 1985). Dip approaches zero for unimodal distributions. The signifcance of a given dip value is determined by comparing it to the distribution of values from a known unimodal distribution. 3.4.4 Growth curves Centiles were generated directly from the data for NF 1 patients of various ages and compared to the coresponding centiles from reference populations. NF1 patients were divided into gender and age groups matching those of the curves available for population norms. A typical series of age-groups had medians of 2, 2.5, 3... 18 years. Age-group limits were determined by spliting the diference between a given median and the next lowest and highest medians. Patients with ages equidistant from two medians were assigned to the older age group. For example, the 2.5 year-old group included patients ages 2.250 to 2.749 years old. The 5th, 25th, 50th, 75th and 95th centiles for stature and OFC were determined for both genders in each age group of NF1 patients and ploted along side the centiles from the coresponding population standards. The data were ploted and smoothed using SAS (1996). Smoothing was done by producing a cubic spline that minimises deviation from fit (Reinsch 1967; SAS 1996). Smoothed curves were inspected to ensure that the final results reasonably represent the data. 54 Splining was used and described in detail in a study that generated standard curves for the NCHS (Hamiletal. 1977). 3.5 Results 3.5.1 Standardised stature and occipitofrontal circumference 183 males and 202 females were included in analyses of stature. 216 males and 220 females were included in analyses of OFC. Analysis of variance for heterogeneity among the 14 North American contributing centres revealed no centre bias for age- and sex-standardised stature (p=0.72) or OFC (p=0.10). Figures 3.1 and 3.2 show the distributions of standardised measurements of stature and OFC among NF1 patients and population norms. Mean standardised stature among NF1 patients is lower than the mean in the reference population. Thirteen percent of the NF1 patients lie two or more standard deviations below the reference population mean, compared to 2% of norms. Mean standardised OFC among NF1 patients is greater than the mean in the reference population. Twenty-four percent of NF1 patients lie 2 or more standard deviations above the reference population mean. The histograms for stature and OFC appear unimodal (Figures 3.1 and 3.2) - their dip statistics, which measure departures from unimodality, are 0.014 and 0.012, respectively. These corespond to the 10th centiles in distributions of dip based on known unimodal distributions. In other words, the deviations from normality are likely to have resulted from chance alone. The standardised stature distribution has a skewness of 0.32, and a kurtosis of 0.19 - cases are clustered to the left of the mean, and the distribution 55 Figure 3.1: Distribution of sex- and age-standardised stature. NF1 patient measurements (histogram) are from the National NF Foundation Database. Unafected norms (smooth curve) are from the National Center for Health Statistics and the Fels Institute. 56 50 Standardized Stature Figure 3.2: Distribution of sex- and age-standardised occipitofrontal circumference. NF1 patient measurements (histogram) are from the National NF Foundation Database. Unafected norms (smooth curve) are from the National Center for Health Statistics and the Fels Institute. 5 9 peaks more abruptly than a normal distribution. The standardised OFC distribution has a skewness of-0.16 and a kurtosis of 0.87 - cases are clustered to the right of the mean, and the distribution peaks more abruptly than normal. 3.5.2 Centile curves Stature and OFC centiles by age and gender are shown in Figures 3.3(a)-3.4(b). Median stature is as much as 7 cm lower and OFC 2 cm greater in NFDB NF1 patients than in the standard pediatric growth charts, depending on age and gender. 3.6 Discussion 3.6.1 Population norms The NCHS and Fels standards were used for comparison to stature and OFC of NF1 patients because these studies cover a wide range of ages and are commonly used clinicaly to diagnose short stature and macrocephaly. These normal population studies were longitudinal and therefore more accurately represent growth than the cross-sectional studies I used. More recent NCHS standards for boys and girls are available for stature but not for occipitofrontal circumference (htp:/www.cdc.gov/growthcharts). These new stature standards for boys and girls are remarkably similar to the ones I used (Hamil et al. 1977; Kuczmarski et al. 2000). 60 Figure 3.3(a): Stature centiles in males 2-18 years. N F 1 patient measurements are from the National NF Foundation Database and are denoted by solid lines. Unafected norms are from the National Center for Health Statistics and the Fels Instiute and are denoted by dashed lines. 62 Figure 3.3(b): Occipitofrontal circumference centiles in males 2-18 years. NF1 patient measurements are from the National NF Foundation Database and are denoted by solid lines. Unafected norms are from the National Center for Health Statistics and the Fels Instiute and are denoted by dashed lines. 63-62-59 58 57 56 55 54 53 52 51 50 49 48 47 46 J AUCAS A \LES (n=216) -^ ^ / / / i NF1:5th, 2Sth, 50th, 75th and 95th Centiles | - - Norms: 5th, 50th and 95th Centiles 1 1 1 I 1 ! 1 1 ' | l | l | l | l | i | l | l | , 1 1 1 1 i i i i 63 62 59 54 53 h 52 51 - 50 - 49 48 47 46 2 3 4 5 6 7 9 10 11 12 13 14 15 16 17 Age (years) 64 Figure 3.4(a): Stature centiles in females 2-18 years. NF1 patient measurements are from the National NF Foundation Database and are denoted by solid lines. Unafected norms are from the National Center for Health Statistics and the Fels Instiute and are denoted by dashed lines. 170 150 140 120 110 90 70 - --- H A I IPAQ AM-FFMA .ES (n=202 i] / / ^ -/ / / / / / / / / / y / / / y / / / / / y ' / s / . / / / / / / / / . // // // / / / / // /•/ / / / . / / /y y // / / / s / / / / // / / / / // y / / / / / / / / / // // / / / / . / / / / / / / / / / / y ' / / // / / / / / / . / / / / ' / / // A / / '/ / / / / / / ' / . / / '  // / / / / / / / ' / / ' / / / // J /  y // / / , / / / / . / / ' / / / / '// / "// —x~ / / / / / // / / / 1 /// // > ' / / / // (//_ _ NF1:5th, 25th, 50th, 75th and 95th Centiles | — Norms: 5th, 50th and 95th Centiles ! ! 1 i 1 i i 1 1 1 1 . | . | . | . | i | i | i | i | i | i 170 160 150 140 120 110 90 70 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Age (years) 66 Figure 3.4(b): Occipitofrontal circumference centiles in females 2-18 years. NF 1 patient measurements are from the National NF Foundation Database and are denoted by solid lines. Unafected norms are from the National Center for Health Statistics and the Fels Instiute and are denoted by dashed lines. AUG m MAL 1=Z y y y y y y y y y y y y y y y y y y Sy y y y y y y y / / / / // / / y y y y y *• y y ' / y y " // / y y / y / // / V '/ // 1/ / / y y / / / / I 1 NF1:5th, 25th, 50th, 75th c ind 95 th Centiles 1 1 1 i i i i 1 - - Norms: 5th, 50th and 95th Centiles r I ! I I I I I I I 1 I 1 1 1 1 1 I 1 I 1 I 1 I 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 Age (years) 3.6.2 Standardisation Standardisation for age and gender by z-scores is a transformation that alows pooling of measurements across groups that difer in age and gender. This transformation can be applied to measurements for which standard population distributions are approximately normal. The distributions of stature and OFC satisfy this criterion (Hamil et al. 1977). Thus, a subject's z-score closely coresponds to his or her centile rank. 3.6.3 Assessment of heterogeneity I was concerned that diferences among examiners and equipment at diferent centres would increase the variabilty of my sample and diminish my abilty to analyse the standardised distributions. However, analysis of variance detected no statisticaly significant diferences for stature or OFC among the contributing centres. 3.6.4 Assessment of standardised distributions The shifts in the standardised distributions of stature and OFC (Figures 3.1 and 3.2) confirm that, on average, the NF1 patients in this study are shorter and have larger heads than standard populations. The shifts are very similar to those in a recent longitudinal study by Carmi et al. (1999). The population norms were not taken during the same years as my sample, and some were taken 20 or more years earlier. Secular trends suggest that stature and OFC may have increased in the normal population over 69 this time (Roche 1979; Byard and Roche 1982; Ounsted et al. 1985; SAS 1996). If year-specifc standards were used in my study and the one by Carmi et al., the shift in stature may be slightly larger than indicated here and the shift in OFC may be slightly smaler. Ascertainment bias may also afect the distribution shifts. Although short stature and macrocephaly are not among the diagnostic criteria for NF1, these features may have contributed to the patients' referral to the contributing NF clinics (NIH 1988; Gutmann et al. 1997). Therefore, the group in this study may be shorter and have bigger heads than a population-based sample of children with NF1. The distributions of standardised stature and OFC have positve kurtosis values -they have fewer data points within one standard deviation of the mean than does a normal distribution. This variabilty could result from several factors: 1) The NFDB patient measurements are cross-sectional and, therefore, more variable than longitudinal data; 2) The NFDB patient group is geographicaly heterogeneous; 3) A smal proportion of cases may have data-entry errors; 4) Ascertainment bias may increase the frequency of outliers; 5) Such distributions might represent composites of more than one normaly distributed group with the same mean but diferent variances (Zar 1999); 6) The estimates of the standard deviation used in the standardisation may have been diferent than the standard deviation among patients. Riccardi has suggested that short stature and macrocephaly in NF1 are "al-or-none" phenomena, i.e., that two diferent groups of NF1 patients exist: those with short stature (or macrocephaly) and those without (Riccardi 1992). Under this hypothesis, the distributions should be bimodal. My findings are not consistent with this suggestion. 70 The distributions (Figures 3.1 and 3.2) indicate that stature is reduced to some degree and OFC enlarged to some extent in al NF1 patients. 3.6.5 Assessment of centile curves The centile curves for stature and OFC (Figures 3.3(a)-3.4(b) are comparable to those from a recent study of Italian NF1 patients (Clementi et al. 1999). Minor diferences may be partly due to line-smoothing techniques and geographic variation. Deviation from these NF1 standards may indicate the efect of a specifc disease feature such as optic glioma or hydrocephalus. On the other hand, NF1-specifc charts may provide reassurance that an afected child's growth, although outside the "normal" range on standard pediatric growth charts, is actualy normal for a child with NF1. I also created charts for body mass index and the ratio OFC/stature by age and gender in Caucasian NF1 patients. These charts are available from htp:/www.medgen.ubc.ca/friedmanlab. 3.6.6 Pathogenesis Patients with known hydrocephalus and plexiform neurofibromas of the head were excluded from my analyses of OFC, so enlargement of the head in the remaining patients must be due to enlargement of the scalp, skul or brain. In NF1, enlargement of the brain is the likely cause (Riccardi 1992; Huson 1994). Glial cel proliferation is increased in vitro by sera from NF1 patients, compared to sera from unafected individuals (Caronti et al. 1998). Optic or other CNS gliomas are another manifestation 71 of glial cel proliferation. They were observed by MRI in 10% of NF1 patients in the NFDB. Other studies have observed optic gliomas in 1.5% of 135 and 15% of 217 NF I patients (Lewis et al. 1984; Huson et al. 1988). Glial overgrowth is an important part of NF1 and it may be responsible for macrocephaly in NF1 patients. Patients with puberty disturbance or bone abnormalites were excluded from my analyses of stature. The cause of the stature reduction in the remaining NF1 patients is unknown, but is thought to afect the skeleton proportionately (Riccardi 1992). Data reviewed by Howel et al. (1998) indicate that growth hormone replacement therapy resulted in a moderate increase in stature for NF1 patients with biochemical evidence of growth hormone deficiency. Although growth hormone levels were not measured routinely in the NFDB patients, less than 1% are known to have ever had documented growth hormone deficiency. Growth hormone deficiency was found in only 3 (2.5%) of 122 children with NF1 in another study (Cnossen et al. 1997). Stature appears to be reduced in many more NF1 patients than can be atributed to such deficiency. The findings in this study are consistent with known molecular function of the NF1 gene and protein in humans and Drosophila (see section 1.9). Mutations in neurofibromin in GAP-related domain produce hyperactivity of p21ras which may contribute to increased glial (astrocyte) cel proliferation and to enlargement of the brain in NF1 patients (Nordlund et al. 1993; Rizvi et al. 1999). The NF1 homologue in Drosophila also acts as an activator of the cAMP pathway (Guo et al. 1997; The et al. 1997). Drosophila homozygous for either of two particular NF1 mutants that lack expression of NF1 protein are 20 to 25% smaler than flies of the parental strain (The et al. 1997) - a phenotype rescued by activating the cAMP pathway through expression of 72 activated protein kinase A. Human neurofibromin also has cAMP dependent protein kinase A (PKA) binding sites (Marchuk et al. 1991; Fahsold et al. 2000). Deficiencies in this pathway may contribute to a smaler phenotype in humans as wel. Activated PKA. is known to stimulate proliferation in some cel types and may normaly contribute to body growth (Miyazaki et al. 1992; Kim et al. 1997). Normal stimulation ofthe PKA pathway also accelerates diferentiation and inhibits proliferation of glial (oligodendrocyte) cels (Raible and McMoris 1990; Raible and McMoris 1993). Neurofibromin involvement in or between the PKA and p21ras pathways may contribute to the larger heads observed in people diagnosed with NF1 (Izawa et al. 1996). However, patients with the smalest stature did not also have the largest heads. 3.7 Conclusion Short stature and macrocephaly are wel-recognised clinical features of NF1. This study suggests that these changes in growth afect al NF1 patients and are not limited to particular subgroups. Therefore, the growth curves presented here can be used for al children with NF1. The mechanisms by which mutations of the NF1 gene produce these phenotypic efects are unknown, but understanding how they do so may provide an important clue to the pathogenesis of other NF1 features. 73 4. L O G I S T I C R E G R E S S I V E M O D E L S O F C L I N I C A L F E A T U R E S I N N E U R O F I B R O M A T O S I S 1 74 4.1 Hypothesis Clinical features of neurofibromatosis 1 (NF1) are not randomly distributed among afected individuals. 4.2 Objective To identify and estimate moderate to strong associations in individual NF1 probands between several diferent clinical features and to validate these associations in afected relatives and in a population based sample. 4.3 Introduction NF1 expressivity is tremendously variable, but subtle phenotypic paterns may exist within subgroups of afected patients. The existence of such subgroups is supported by the observation of a relatively consistent phenotype among patients with deletions of the entire NF1 and in familes with rare NF1 variants (see section 1.13). In Chapter 2,1 demonstrated several associations between pair-wise combinations of clinical features among age-stratified probands with NF1. These analyses support the existence of phenotypic subgroups but were limited in two ways: only two features could be examined at once and some of the comparisons may have been confounded by age. Although I analysed age in 5-year strata, virtualy every feature of NF1 exhibits a diferent relationship with age, and there stil  may have been considerable age-related variability, especialy among the youngest patients. In another study, numerical taxonomy was used on NFDB data to identify five 75 distinct clinical groups of NF1 patients (Friedman et al. 1995). However, these clusters were also subject to confounding by age and other variables. Log-linear models have been used previously to test for associations among several diferent congenital malformations and determine which features tend to occur together (Beaty et al. 1991). However, log-linear models, like my previous study based on contingency tables, can only treat age as a categorical variable. Since virtualy every feature of NF1 exhibits a diferent relationship with age, I chose to control age as precisely as possible - as a continuous variable. Unlike log-linear models, logistic regressive models alow the analyst to use continuous variables as wel as binary and categorical ones. In this study, I have extended my analysis of associations among clinical features in NF 1 patients by using logistic regression to consider joint and interactive efects of several clinical features at once and to control for age as a continuous variable. My findings clarify and refine the associations among clinical features in NF1 patients and provide further clues to the pathogenesis of these features. 4.4 Subjects and methods This study involved analysis of four separate clinical samples of patients with NF1 - developmental, validation, and relative samples from the National Neurofibromatosis Foundation International Database (NFDB) and a population-based sample from the Manchester NF1 database (MANF1), as described below. Logistic regressive models were built from an initial series of one-covariate models, by progressively adding covariates and interaction terms, in the developmental sample. The 76 best fitting of these models were then tested in each of the other samples, using both the parameters from the developmental sample and by refiting the parameters in each of the other samples. 4.4.1 Subjects Subjects were obtained from two large clinical databases: the NFDB and the MANF1. All patients included in this analysis were diagnosed with NF1 according to established clinical criteria (NIH 1988; Gutmann et al. 1997). At the time of this analysis, the NFDB included extensive demographic and cross-sectional clinical information on 2797 NF1 probands and 511 of their afected relatives examined since 1980 at 25 participating centres in North America, Europe and Australia. 83% of the cases are Caucasian, 7% Asian, 4% African-American, 6% other or mixed race. Al information was colected and recorded on each patient using a standard procedure (Friedman and Birch 1997b). The data were audited for quality and consistency by the NFDB administrator. The Manchester NF1 Database (MANF1) sample is described in section 2.4.1. There is no overlap among the patients included in the NFDB and MANF1 databases. 4.4.2 Clinical features I selected 13 important or frequent clinical features of NF1 for this study: cafe-au-lait spots, intertriginous freckling, discrete cutaneous neurofibromas, discrete subcutaneous neurofibromas, difuse or nodular plexiform neurofibromas (refered to as 77 "plexiform neurofibromas"), Lisch nodules, scoliosis, tibial or other long bone bowing or pseudarthrosis ("pseudarthrosis"), optic glioma, macrocephaly, short stature, seizures and neoplasms (other than neurofibromas or optic glioma). Each of these features was coded as either "present", "absent" or "unknown". Age, coded to the nearest 0.01 year, and gender were considered as covariates. Cafe-au-lait spots were considered "present" in patients with 6 or more spots of suficient size (see Table 1.1). Most of the other features were identified according to the criteria described in section 2.4.2. Patients coded as "unknown" for a particular feature were not considered in models involving that feature. 4.4.3 Statistical models Thirteen separate logistic regression models were built, with the logit of each of the 13 NF1 features analysed set as the binary response variable (Y) in a diferent model. Whereas linear regression atempts to compute the mathematical relationship between the covariates and the response variable directly, logistic regression atempts to compute the relationship between the covariates and the log of the odds of the response variable being "present". The frequencies of many features change with age, but this efect is not uniform among the features (Friedman et al. 1999). Therefore, age was controled as precisely as possible, as a continuous explanatory variable. First, a one-covariate model was constructed using age as the only covariate: f 78 where: p(l\x) is the probabilty that the feature is "present", given the covariates, x; the logit is the log function on the left side of the equation involving ,p(l|x); a is the y-intercept; /? is the slope; and AGE is the age of the subject at the time the feature was assessed. Maximum likelihood techniques were used to generate parameter estimates (SAS 1996). Linearity in the logit was examined in each model, and age was transformed when necessary to meet the requirement of linearity in the logit. log = a + j3]AGETRF where AGETRF = exp(-cxAGE) At AGE zero, the value of this function is a + /?/. For negative values of Pi, the value of the function approaches a as AGE gets larger. This function was used to approximate the frequency-by-age curves of the NF1 features considered in this study. It was necessary to use this transformation of AGE to maintain linearity of the logit for most outcome variables in this study. A series of two-covariate analyses was then performed using the equation, 79 log r = a+ B] AGETRF + B2x V 7 in which each of the 13 features was set in turn as the response variable, and AGETRF and one of the 12 remaining features (x) were used as explanatory variables to screen for potential main efects. Variables with parameters (JTs) with p<0.2, a standard cut-of value that alows more variables to influence the outcome (Hosmer and Lemeshow 1989), were included as explanatory variables (x,'s) in multi-covariate analyses. AGETRF and gender were included as covariates in al models. Folowing maximum likelihood estimation of the parameters in the multivariable model, the importance of each explanatory variable was reassessed. Explanatory variables with parameters greater than zero with p<0.2 were used to refit the model and interaction terms (ffs) among the explanatory variables were considered by forward selection. For example, log /?(l|x) a + Bi AGETRF + B2MALE + B3x3 + /?4x4 + B5x log f = a+BxA GETRF + B2 MALE + B.x, + B,x4 + «5, x3 x4 V 7 80 4.4.4 Model validation Fited logistic regressive models always perform favourably on the sample used to generate them (Hosmer and Lemeshow 1989). Therefore, a random subsample consisting of 1,384 of the 2,797 NF1 probands from the NFDB was excluded, and models were developed on data from the remaining 1,413 NFDB probands (the "developmental sample"). These models were tested on data from the 1,384 NFDB probands who were originaly excluded, the "validation sample". The models were also tested on data from 511 afected relatives of the 2797 NFDB probands and on population-based data from the MANF1, which includes both probands and afected family members. The Hosmer and Lemeshow (1989) goodness-of-fit test was used to assess how wel the parameter estimates from the developmental sample fit the validation, afected relative, and MANF1 samples. This statistic compares the likelihood of a feature predicted by model parameters to its actual occurence in the data. Models with a goodness-of-fit p>0.05 were considered adequate. In addition, parameters for covariates and significant explanatory variables from the best-fiting models derived in the developmental sample were re-estimated by maximum likelihood in the validation, and afected relative samples, and in the independent MANF1 sample, to alow more detailed comparison. 4.4.5 Interpretation Logistic regressive models have a straightforward interpretation in terms of odds-ratios. The strength of association between the response variable (Y) and an explanatoiy variable (x/) in a one-covariate model is measured by /?/. Subjects with variable x\ coded as "present" are exp(/?/) times more likely to also have feature Y than are subjects with 81 feature x/ absent. The strength of interaction between Y and explanatory variables (x/ and xi) in a two-covariate model is measured by /?/, B2, and 5j. Subjects with variables X] and X2 present are exp(y#/+/?2+<5/) times more likely to also have the response feature. Subjects with variable xi present and x2 absent are exp(/?/) times more likely to also have the response feature. Confidence intervals were calculated as ±1.96xSE, where SE is the standard eror for /?. For odds ratios involving more than one P>, confidence intervals were calculated using individual standard erors and their covariances: 2 2 ±1.96xsqrt(SEi +SE2 -2xcovariancei2). Odds ratios with 95% confidence intervals that excluded 1.0 were considered unlikely to be due to chance alone. 4.5 Results 4.5.1 Prevalence by age of NF1 clinical features Whereas Table 1.3 shows feature prevalences for subjects of al ages, Figures 4.1 - 4.13 show the prevalence by age of cafe-au-lait spots, intertriginous freckling, subcutaneous neurofibromas, cutaneous neurofibromas, plexiform neurofibromas, Lisch nodules, optic glioma, seizures, pseudarthrosis, scoliosis, macrocephaly, short stature and neoplasms (other than optic glioma and neurofibromas). These curves were generated from cross-sectional clinical information on 3308 NF1 patients recorded in the NFDB. The total number of patients included for each clinical feature varies because I excluded cases in which the presence or absence of the particular feature could not be determined unequivocaly from available data. 82 Figure 4.1: Prevalence by age of 6 or more cafe-au-lait spots. The curve is based on 3244 NF1 patients from the National Neurofibromatosis Foundation Database. "Presence" required 6 or more cafe-au-lait spots over 5mm greatest diameter in prepubertal individuals or over 15 mm in greatest diameter in postpubertal individuals Figure 4.2: Prevalence by age of intertriginous freckling. The curve is based on 3198 NF1 patients from the National Neurofibromatosis Foundation Database. Figure 4.3: Prevalence by age of 2 or more subcutaneous neurofibromas. The curve is based on 3255 NF1 patients from the National Neurofibromatosis Foundation Database. Figure 4.4: Prevalence by age of 2 or more cutaneous neurofibromas. The curve is based on 3283 NF1 patients from the National Neurofibromatosis Foundation Database. Figure 4.5: Prevalence by age of plexiform neurofibromas. The curve is based on 3283 NF1 patients from the National Neurofibromatosis Foundation Database. 92 Figure 4.6: Prevalence by age of Lisch nodules. The curve is based on 2432 NF1 patients from the National Neurofibromatosis Foundation Database. 94 Figure 4.7: Prevalence by age of optic glioma. The curve is based on 1195 NF1 patients from the National Neurofibromatosis Foundation Database. Figure 4.8: Prevalence by age of seizures. The curve is based on among 3308 NF1 patients from the National Neurofibromatosis Foundation Database. 97 o m -i i 1 1 ! 1 1 1 1 !- O C I > 0 0 I ^ C D L 0 ^ f C 0 C N | - r - O 0 0 0 0 0 0 0 0 0 80U8|BA8Jd 98 Figure 4.9: Prevalence by age of pseudarthrosis. The curve is based on 3287 NF1 patients from the National Neurofibromatosis Foundation Database. o -i 1 1 i 1 1 1 1 1 !- o 0 ) 0 0 h - - C D l O ' ^ - C 0 C M i - O O O O O O O O O O aouaiBAajd 100 Figure 4.10: Prevalence by age of scoliosis. The curve is based on 3057 NF1 patients from the National Neurofibromatosis Foundation Database. o in o m co o co CM O CN CD < d oo d d co d d d co d aoua|BA9Jd 102 Figure 4.11: Prevalence by age of macrocephaly. The curve is based on 2129 NF1 patients from the National Neurofibromatosis Foundation Database. Figure 4.12: Prevalence by age of short stature. The curve is based on 1918 NF1 patients from the National Neurofibromatosis Foundation Database. o LO I 1 1 1 1 ! 1 1 1 !- o < J > o q r ^ c q i n ^ - c o c N j - r - o o o o o o o o o o aOU9|BA9Jd 106 Figure 4.13: Prevalence by age of neoplasms. The neoplasms do not include optic glioma or neurofibromas. The curve is based 3129 NF1 patients from the National Neurofibromatosis Foundation Database. 4.5.2 Model development and validation A multi-covariate logistic regressive model was generated for each of 13 diferent NF1 clinical features, using age (transformed using a log function) and gender as covariates, and each of the 12 other features as possible explanatory variables. Maximum likelihood parameter estimates were used to determine the best fitting model for each of the clinical features in a developmental sample of NF1 probands from the NFDB, and goodness of fit of each model was then evaluated in three other independent samples - an independent "validation" sample of probands from the NFDB, non-proband afected relatives from the NFDB, and the population-based MANF1 sample that includes both probands and non-probands. The best-fiting models in the developmental sample for the folowing outcome features had parameter estimates that also had an adequate fit (Hosmer and Lemeshow goodness-of-fit p>0.05) in the validation, afected relative and MANF1 samples: intertriginous freckling, subcutaneous neurofibromas, plexiform neurofibromas, optic glioma, pseudarthrosis, macrocephaly, and other neoplasms (Table 4.1). Models for cafe-au-lait spots, cutaneous neurofibromas, Lisch nodules, seizures, scoliosis and short stature had an inadequate fit (Hosmer and Lemeshow goodness-of-fit p<0.05) in at least one of the samples. 4.5.3 Parameter estimate comparisons Parameter estimates were independently generated in each of the four samples for the folowing features: cafe-au-lait spots, intertriginous freckling, cutaneous 109 neurofibromas, subcutaneous neurofibromas, plexiform neurofibromas, Lisch nodules, scoliosis and short stature (Table 4.2). Parameter estimates for optic glioma, seizures, pseudarthrosis, macrocephaly and other neoplasms could not be generated in al four samples, due to sparseness of data in at least one of the samples. The coresponding cels in Table 4.2 are blank. Standard erors are shown for al the parameter estimates in the development subsample (Table 4.1) but not for parameter estimates in the other three subsamples (Table 4.2). Whereas the development subsample was used to identify statisticaly significant parameter estimates, the later three subsamples were used to determine which variables had more or less consistent parameter point estimates. 4.5.4 Consistent parameters from validated models Some of the parameters from models that had an adequate fit (Hosmer and Lemeshow goodness-of-fit p>0.05) in al four samples were not consistent when generated independently in each sample. In the plexiform neurofibroma model, the parameter estimates for scoliosis and other neoplasms difered greatly among the four samples. In the pseudarthrosis model, the estimate for freckling was inconsistent. Models from the ill-fit samples difered dramaticaly often by the estimate of only one parameter. The 10 models that had an adequate fit in at least three of the four samples were recalculated including only variables with consistent parameters. Recalculated parameters for the developmental sample are nearly identical to the initial parameters in Table 4.1 and are summarised as odds ratios with 95% confidence intervals in Table 4.3. 110 oo . - f i cu to Z &o cu c « 'a "3 > co. es a "a cu s-3 C3 3 a 3 IO VO OH fi O V I  OH O 00 00 in oo o ON OH a oo ON OH fi Tl- oo oo Tl" — ' TT m m oo CN CN vo cn .—< d o' d —< PH 00 7 CU r^ H -fi OO . 3 * § ; § - f i ^ PH O O ^ oo U w j £ 00 fi o )H PH GO i—I u o OH 00 03 O oo O r-i I  I  OH fi O cn I  I  OH C i n oo O ON I  I  OH fi 0 0 t--O ON I  I  OH fi O O ON T f CN T f vo vo ON ^H ^ TI- CN cn d d d - H 00 CD -a o fi n f i o fi o CU kH PH ^ oo O CN V I  OH C 2 ™ . ON O T f I  I  OH C CN CN m . cn o —< I  I  OH fi ON « N O r-H OH fi m vo ON ^H ^H rn r- ^ H vo vo o o o PH PH Z Z Vi PH O M -a fi <u oo PH PH Z u oo PH Z 00 fi O CU fi U O CN I  I  OH fi O T f I  I  OH fi VO o cn cn O - H OH fi . m O - H I  I  OH fi 0O O —H ON —i r-i CN CN o ON c— m rn vq oo vo d o d d PH Z PH * PH Z u ,00 HH o u £ PH Z 00 PH 00 fi O <U fi C53 cu fi 00 ^ o in O CN I  I  OH fi 2 vo d TT I  I  OH fi 2 M . CN O r-c I  I  OH fi £2 d - H I  I  OH fi CN CN 00 —( O CN CN '—1 cn cn in m CN ON '—< oo r-n vo rn d ^< d d d *Z 00 f H Z Z & O oo oo PH 00 PH CN t--O *~H I  I  OH fi o cn I  I  OH fi O ON I  I  OH fi vo O ON I  I  OH fi o CN cn TT TT Tf Tt- r- TJ-TJ- vo r- oo m o o ^ H o oo f-< "o. * o oo <U i_l PH Z - o £ u oo CU 3 -d o fi , f i o oo 111 S a a o o PL, 5 c o , 3 > fl CD e > fe CZi CO. o I " P , CD u Ife PH f l 8 p O - H I  I  PH o ^ r-o 00 I  I  P * fl 00 CN ft C 0 1 0 O O 0 II m 1  0 11 II PH c PH 11 fl m m vo 00 r~ vo ON o o r ~ ; ca .£5 a CD O o PH O JS "a, o CD ca C50 O 2 o . in O CN I  II II II ft C O V I  OH A 8 0 0 in O CN I  I  OH fl 0 0 1—H n VO o j^-I  I  OH fl . CN O - H I  I  OH fl I  I  OH A 00 CN m ON in cn ro ^ t--o ON o m >o '—< © CN ON O © CN © © CN l in CD ^ • 7 4 % A* C/3 ca CD 03 ^ CN o m I  I  OH C O V I  OH fl ^ CN O I  I  OH fl H CN O I  I  OH fl —H in vo <n m o TJ- ON 00 in r-- in 0 0 0 0 DO .fl ID PL, CD |7 PH U CD fl CD OH CD CD =2 CO ' c o O CD CO OH in r- r-o © O OH CO O o 00 CN ^ ON o r-I  I  OH fl in o I  I  OH fl O - H I  I  OH C O I  I  OH fl o in ON r- • <n <n m >—1 q t_ - 1 © < < CN <u 2 CD 3 S fa T ) O S fl O 3 « S C 3 ) t o ca o " t ! & j O 53 £ CO OH CD O o I-CD ca r-O V I  OH fl VO CN OH fl H CN o vo I  I  OH fl o CN vo o I  I  OH C vo o r-CN cn cn •5f CN O ^ CS OH CD CD O ) H CD a CD ) H fl o n ^ r O I  I  PH fl ON ^_ O I  I  PH fl o j^-I  I  PH fl ON ^ © I  I  OH C 00 in cn vo I m vo ON ON r--o —5 T -J ca CO s s .2 c o O PH l - l CD CO in " P , O CD 112 fl CD 3 ID C T3 CD >, CD fl" J2 CD CD C O CD e OH fl ^ -=1 .5 H CD co CD C CD oo on CD > c o fl VA sc ca i- ^ -t* ca £ :§ z E J2 o « CO © S S i o ca CD CD «. O i CO - CO CD « bb CD bo o £ ,o CO oj co - tu 05 ca e CU CD S" S - , CD CD 4> CD s s 1 a « W OH fe C^_ .. O <s >> 5 s 03 3 fl CD co CD l H CD CD CO ! H a co ca •n « +-» 3 52 co >, fl .O CD T3 £ CD OH ta ° fa w ca CO JD SH £ I 2H i-i ca O ,CD ca fl xi —< ca OH CD X |_ W £ c o "3 tt 2 * a> -o .s • CO 2 a cS - ° <tl CD T3 -5 z ca s _© 03 > c E CH o >• Q 03 CO & S-C3 3 "& « X <y fe fe s-a •** 03 V fe -<-< s a s O < - H vo o r-in ^- cn M O >—< O CN co r - 1 ^3> o >n in oo o o o ^ CN VO O co CN CN vq d o d o in co oo cN CN vo cn * d d d ^ co *7 CD —^ o c bo CD CD fe bfi C fe O Z -2 00 fe CD CD S H fe OO i - l U o OH CO H-» ja fl ca ca O O m oo r-~ vo o> o - J »-<' ^ J ON VO in vo c— o CN cn d d d ^ • ON cn r~ ^ H vo oo d d d d vo vo ON —i *—' ^vt" CN cn d d d - J CO CD 3 O C fe o oo _ -2 J b0 .fl CD CD l H fe VO 00 -vf ^ cn ^ —< d CN o o o ON vo VO ON o d d oo r- i—i ^- vo vo o d d o «3 o fl CD fl CO fl O ^ CD tu, c • ca o -a fl CD co fe r- vo co o CN O •—< d CN d d vo o o O — c o cn CN oo CN ON r-. m d d d d O ON C co cn vq oo vo d o d o fe z fe * 00 fe fe fe O O fe o ca a o > H « O fl CD fl CO fl O CD "5 fe " z fl & oo w cn cn in oo c • •n- in in TJ- CN d d d ^ d CN —i ON r-~ -^i ON ^ H - 3 - cn o d —; d - J —; • i ON O CN CN t CN CO CN d —^  ' ' m m o o d • in cn cN ON < oo ^ H vq co t-^  d ^ J d d d i • CO ca o o fe \'r"\ U O0 00 £ O0 fe fe o « o 3 CD fl £ fe ">< ? CD fe fe ^ co ON I—i o vq ^ o CN d - J d d o o m CN co ON O "\T d - H d ^ O <N in in - H vo oo d - J o o • T J - r-^ vo d d ~ oo CO I 0 0 fe O U a l l CO f -H ca u "OH * o co £ U CO CD O C -fl CD CO 113 CU s a a o u H to CU a 3 "es c cu £ a % a © 93 Vi s X CU to to cu u s • M ft CU to s a S O m oo r--cn oo rn * d CN o oo ^ 1^ 1 T l " ON d d vo oo r--VO T f ON O O r - H ccS K i CD to crj a o H-» CM O vo ON r-- vo cn >—i O rn vo ci ai <d> ci <zi o O N o m m '—J O C N O N O d IN d d N fi ccj J2 ofi OH o CD CD * M to I £ < GO OH o CD CD GO cn O N cn CN O N O 00 © m d r J d d d cn vo m oo cn vo cn cn m vo CN d d d d d d O Tf- ON Tt" oo in r- m d d © d • i .0 o CD (D to §r-£ J 43 u z S &0 • i-H &0 O T3 fi CD czi OH O N cn m in d d ^ H rn r- vo d d to O OH &0 O o GO ^ H T f o cn in oo TT O O r H in in m ^ H T t O Tj" d ^ ^ ri i <o 2 w 3 a b fi bfi to ccj .2 OH 8 U CCS -fi OH 0) o o kH o CCS in o ^ H VO CN t— o d d ON CN Tj-ON o ON o d d - H T-H i - H rn ON ON oo OO o ON m CN oo Tj" vq CN d d i - H d d O - H TJ- CN O O — i GO to *c3 .fi OH CD O o kH o cd o H f i GO TI- m vo ON ON t--d H CD cd '&o fi a s .2 -2 fi "OB "ci n f i O "3 O \ f i 3 OT OH S kJ O OH CD &0 "BH o CD 60 a '•B c o n, to cu c o o <u JS H CO Ck CD E o &0 114 Table 4.3: Logistic regressive models summarised as odds ratios. Summary of associations from validated logistic regressive models of NF1 clinical features. These parameters are based on a recalculation of the development subset. Only models with adequate fit and consistent parameter estimates in at least three of the four samples are shown. F e a t u r e A s s o c i a t e d F e a t u r e s O d d s - R a t i o ( 9 5 % C . I . ) Cafe-au-lait spots (CLS) Freckling 1.4 (0.8-2.0) SNF 0.5 (0.2-1.3) Both 2.3 (1.2-3.7) Freckling CLS 1.2 (0.7-1.9) SNF 0.6 (0.3-1.4) Lisch nodules 1.3 (0.9-2.0) All three 3.7 (1.8-7.4) Cutaneous neurofibromas SNF 1.6 (1.2-2.2) (CNF) Plexiform 2.0 (1.4-2.7) Both 3.2 (2.1-4.7) Subcutaneous CLS 1.4 (1.0-1.9) neurofibromas (SNF) CNF 2.0 (1.4-2.8) PNF 2.4 (1.6-3.6) All three 3.4 (2.1-5.7) Plexiform neurofibromas CNF 2.8 (2.0-4.8) (PNF) SNF 2.5 (1.5-3.6) Both 3.6 (2.3-5.5) Lisch Nodules CLS 1.6 (1.0-2.4) CNF 1.7 (0.9-4.3) Both 2.2 (0.4-11.0) Optic glioma PNF 1.9 (0.9-4.0) Macrocephaly 1.6 (0.9-2.9) Neoplasms 7.1 (2.8-18.1) All three 22.4 (5.8-86.6) Pseudarthrosis CNF 0.6 (0.3-1.2) Neoplasm 1.8 (0.9-5.4) Male gender 1.6 (1.1-2.9) All three 1.7 (0.7-7.5) Macrocephaly Lisch nodules 1.6 (0.7-3.5) Optic glioma 2.9 (1.2-6.9) Short stature 0.2 (0.1-1.1) Neoplasm 0.1 (0.1-1.1) All four 0.1 (0.1-1.1) Neoplasm (other) Lisch nodules 2.6 (1.1-6.1) Optic glioma 6.9 (3.3-14.5) Pseudarthrosis 5.8 (1.6-20.9) All three 102 (17.1-616) 115 For example, intertriginous freckling was found to be 20% more common (odds ratio = 1.2) in subjects with cafe-au-lait spots, 40% less common (odds ratio = 0.6) in those with subcutaneous neurofibromas, and 30% more common (odds ratio = 1.3) in those with Lisch nodules. Although only Lisch nodules were significantly associated on their own, freckling was found to be 3.7 times more common in subjects with al three features. 4.6 Discussion 4.6.1 Ascertainment bias The models I have developed include several associations confirmed in two independent samples of NFDB probands, in their afected relatives and in NF1 patients from the population-based MANF1 sample. The NFDB is comprised of patients seen at specialised clinics, so the development and validation samples of probands are probably more severely afected than the NF1 population in general. The afected relative sample was drawn from the same specialised clinics, but their severity is not as biased as that of the probands (Friedman and Birch 1997b). Nevertheless, since half of NF1 cases represent new mutations, and the NFDB only contains data on 511 afected relatives of 2979 probands, it is likely that many afected relatives of these probands are not included in the NFDB. I expect that afected relatives who are included in the NFDB may be more severely afected than those who were not. In contrast, the MANF1 was colected through genetic registries in north-west England by a limited number of physicians. Both parents were routinely examined. Its ascertainment is nearly 70% and is thought to be representative of the regional NF1 population (McGaughran et al. 1999). Model 116 parameters that have been confirmed in al four samples are unlikely to reflect database or specialised clinic biases. Instead these models probably reflect trends that exist in the NF1 population at large. 4.6.2 Consideration of binary treatment of variables Features such as optic glioma, seizures and pseudarthrosis naturaly fal into a binary ("present" or "absent") coding scheme, while it might be more informative to treat cafe-au-lait spots, cutaneous and subcutaneous neurofibromas, scoliosis, macrocephaly, short stature and others as ordinal or continuous variables. Although the NFDB contains ordinal data on many variables, the MANF1 contains mostly binary data. All 13 of the features in this study were treated as binary variables to permit comparison of NFDB models in the MANF1. 4.6.3 Statistical signifcance The fit is beter for most models (Table 4.1) in the development sample than in any other sample because the parameter estimates were generated in the developmental sample. Models based on afected relatives from the NFDB and MANF1 samples contain multiple data from familes. Standard erors for parameters in these models slightly underestimate true standard errors. Many of the associations in Table 4.3 do not have 95% confidence intervals that exclude 1.0. However, several of these models include three-way interactions (Table 4.2), and the first order parameters must be included to 117 adhere to the principle of a hierarchicaly wel formulated model (Kleinbaum 1992). Also, a variable can contribute to model fit without being significant itself at p<0.05, so the criterion for inclusion in a logistic regressive model was extended to p<0.2 (Hosmer and Lemeshow 1989). 4.6.4 Previous reports of associations Associations involving freckling, Lisch nodules, and plexiform, cutaneous and subcutaneous neurofibromas, are reported in Chapter 2 as pair-wise associations of weak magnitude. For example, freckling and Lisch nodules were shown to have a pair-wise age-stratified odds ratio of 1.8 (95% C.I=1.3-2.4). This chapter shows that most of these associations not only persist when controling for age and other common NF1 features, but increase slightly in strength when the presence of multiple features is considered. The presence of cafe-au-lait spots and subcutaneous neurofibromas as wel as Lisch nodules make freckling 3.7 (95% C.I =1.8-7.4) times more likely. Furthermore, this chapter shows that these pair-wise associations exist simultaneously. For example, cutaneous and subcutaneous neurofibromas are both significantly associated with plexiform neurofibromas (Table 4.3). The pair-wise association between optic glioma and neoplasms has also been previously reported with an odds ratio of 5.8 (Friedman and Birch 1997a) but gains even more strength when other features are taken into consideration. Optic glioma is 22.4 (95% C.I.= 5.8-86.6) times more common when plexiform neurofibromas and macrocephaly, as wel as neoplasms, are present and age is taken into consideration. 118 4.6.5 Pathogenetic interpretation of associations Many of the associations I observed were non-reciprocal - only one of a pair of features appears in the other's model. This suggests that the two features were not of primary importance in accounting for each others' status. I also observed several reciprocal associations. As a conceptual framework to understand the interdependencies of clinical features of NF1, it is not easy to pool the results of 13 separate models. Numerical taxonomy and log-linear models are easier to interpret, but they are more susceptible to confounding by age and do not provide initial parameters for the final study in this thesis, multivariate analyses (see Chapter 7). Still, a few features appeared to be closely associated (see below). In general, features were considered to be closely associated if each feature appeared as an explanatory variable with a positve parameter estimate in each ofthe other group members' models. Such a relationship means that these features must be taken into account to accurately describe the occurence of other features belonging to the same group. Fundamental pathogenetic diferences may exist between subjects who have one or more of a group's features and those who do not, and the mechanisms shared by associated features may be diferent for each group of features. However, these NF1 features are not mutualy exclusive, and many patients belong to more than one group. 1) Cafe-au-lait spots, intertriginous freckles and Lisch nodules are al derived from cels of melanocytic origin (Weston 1981; Pery and Font 1982). Cafe-au-lait spots contain melanosomes with giant pigment particles. Intertriginous freckles develop through a process that does not require light exposure, but they too involve pigment and darken with sun exposure (Fitzpatrick 1981). Histologicaly, Lisch nodules are 119 melanocytic hamartomas. Associations between Lisch nodules and pigmentary features have been previously reported (Pietruschka 1961; Zehavi et al. 1986), but the responsible mechanism is unknown. 2) The associations observed between the occurence of plexiform, cutaneous and subcutaneous neurofibromas are consistent with the histopathological similarity between these lesions (Harkin and Reed 1969; Burger and Scheithauer 1994). In addition, each type of neurofibroma is associated with acquired loss or mutation of the normal NF1 alele in at least some cases (Sawada et al. 1996; Sera et al. 1997). The negative 3-way interaction terms in two of the three models suggest that associations involving neurofibromas are not additive. The association between subcutaneous neurofibromas and cafe-au-lait spots is negative in the cafe-au-lait spot model, but positve in the subcutaneous neurofibroma model (Table 4.1). This is because the coeficient for subcutaneous neurofibromas in the cafe-au-lait spot model changed from positve to negative after adding the interaction term. Similarly, the coeficient for subcutaneous neurofibromas in the intertriginous freckling model changed from positve to negative after adding the interaction term between cafe-au-lait spots and subcutaneous neurofibromas, indicating a positve three-way interaction. Melanocytes in cafe-au-lait spots and intertriginous freckles and Schwann cels in subcutaneous neurofibromas are derived from the embryonic neural crest (Weston 1981). This is consistent with the suggestion that NF1 is a neurocristopathy (Huson and Hughes 1994) but does not explain why these and not other neural crest-derived tissues are involved in NF1 and why many features of NF1 do not appear to be abnormalites of neural-crest derived tissues (see section 2.6.1). 120 3) The common thread between optic glioma, other neoplasms and macrocephaly could be glial hyperplasia resulting from haploinsuficiency of neurofibromin. Most of the other neoplasms in patients in this study involve the central nervous system and most of these are gliomas (Friedman and Birch 1997a). Patients with hydrocephalus and plexiform neurofibromas on the head were excluded from the analyses of head circumference, so enlargement of the head in the remaining patients must be due to enlargement of the scalp, skul or brain. In NF1, enlargement of the brain is the likely cause (Riccardi 1992; Huson 1994). Gutmann et al. (1999) have directly demonstrated an efect of NF1 haploinsuficiency on glial cel proliferation. 4.6.6 Cross-sectional nature of data While these models are accurate descriptors of feature occurence, they cannot be used to predict which patients wil get what features. The NFDB data are largely cross-sectional, with 74% of the subjects seen only once. The MANF1 is exclusively cross-sectional. A fited logistic regressive model can be used to predict the risk for an individual developing a particular feature in folow-up studies, but not in cross-sectional studies such as this one (Kleinbaum 1992). Curently available longitudinal clinical data in NF1 are too limited in number of subjects and duration of study for this purpose; large-scale longitudinal studies of the natural history of NF1 would be necessary to develop predictive models. 121 4.7 Conclusion The occurence of NF1 clinical features can be described to some extent by taking other disease features and age into account. These associations have been demonstrated in probands, their afected relatives and in a completely independent population-based sample. This suggests that the associations are true of NF1 patients in specialised clinics and of NF1 patients at large. Phenotypic studies of afected relatives can determine the importance of familial and genetic factors in the development of these common NF1 features. The models developed here identify and quantify the covariates that must be taken into account in familial studies in Chapter 7 . 122 5. U N I D E N T I F I E D B R I G H T O B J E C T S O N M R I A S S O C I A T E D W I T H D I A G N O S T I C F E A T U R E S O F N E U R O F I B R O M A T O S I S 1 123 5.1 Hypothesis Unidentifed bright objects (UBOs) do not occur independently of other neurofibromatosis 1 (NF1) features. 5.2 Objective Use logistic regressive models to identify individual features whose occurence is associated with the occurence of UBOs in NF1 patients. 5.3 Introduction "Unidentifed bright objects" (UBOs) have been observed on MPJ in 43-93% of children with neurofibromatosis 1 (NF1) (Bognanno et al. 1988; Grifiths et al. 1999). UBOs are areas of increased image brightness that can be visualised only under particular scanning configurations (Truhan and Filpek 1993). They show no mass efect or contrast enhancement (Sevick et al. 1992). Pathologicaly, they corespond to areas of vacuolar or spongiotic change in the brain substance and may represent increased fluid within the myelin associated with hyperplastic or dysplastic glial proliferation (DiPaolo et al. 1995). UBOs evolve over time and are almost never seen in patients over the age of 20 years (Sevick et al. 1992; DiMario and Ramsby 1998). A corelation between the location of these lesions and IQ score has been suggested, but no consistent relationship has been found between the learning disabilties that frequently occur in NF1 patients and the presence of UBOs (North 1999). I show here that UBOs may also be associated with the occurence of other clinical features in young (<21 years) NF1 patients. 124 5.4 Subjects and methods 5.4.1 Subjects 523 patients between the ages of 2 and 20 years who met the NIH Diagnostic Criteria for NF1 and had cranial MRI examinations were selected from the National NF Foundation International Database (NFDB) (Friedman et al. 1993). The presence or absence of the folowing NIH Diagnostic features was known in each patient: ^ 6 cafe-au-lait spots of suficient size for age, intertriginous freckling, ^ 2 Lisch nodules, ^ 2 neurofibromas or one plexiform neurofibroma, a typical bony lesion and optic glioma. Although an afected first-degree relative is also one of the NIH Diagnostic Criteria and accounts for the fact that some of the patients included had only one diagnostic feature of NF1, family history was not considered in this analysis. 5.4.2 Statistical analysis The presence or absence of UBOs on cranial MRI examination was determined for each patient (DeBela et al. 2000a). The presence or absence of each of the six diagnostic features and of CNS neoplasms (other than optic glioma) was determined during the same clinical visit (Gutmann et al. 1997). The %2 and Cochran-Armitage trend statistics were used to test whether the presence of UBOs was independent of the number of diagnostic features. The Cochran-Armitage test partions the x 2 into component parts -2 2 % for linear trend and % f°r departure from linear trend (Zar 1999). Logistic regression was used to quantify the associations between UBOs and each of the diagnostic features simultaneously, while controling for age and gender (SAS 1996). Features with p>0.10 125 in the first model were excluded from subsequent models. Parameters for the remaining features were re-estimated and second order interactions were evaluated. 5.5 Results 5.5.1 Frequency of clinical features analysed 259 (50%) of the 523 NF1 patients from the National Neurofibromatosis Foundation International Database (NFDB) under 21 years of age who had undergone head MRI had UBOs. 492 (94%) had >6 cafe-au-lait spots, 426 (81%) intertriginous or axilary freckling, 253 (48%) Lisch nodules, 128 (24%) >2 subcutaneous neurofibromas, 85 (16%) >2 cutaneous neurofibromas, 100 (19%) >1 plexiform neurofibromas, 73 (14%) a typical bony lesion, 70 (13%) had optic glioma - 23 (4%) symptomatic and 47 (9%) asymptomatic. 11 (2%) patients had other CNS neoplasms - 7 non-optic gliomas, 1 ependymoma, 1 ganglioneuroma, 1 hamartoma of the pons and midbrain, and 1 hypothalamic neoplasm. Figure 5.1 shows the frequency of UBOs by age. 5.5.2 Unidentifed bright objects by number of diagnostic features Figure 5.2 shows the frequency of UBOs among 523 NF1 patients from the NFDB 2-20 years of age according to the number of diagnostic features present in each patient. The frequency of UBOs in my sample does not increase with age in this range (Figure 5.1) but increases from 29% in patients who have only one diagnostic feature to 100% in patients who have six features (x =28.0, p<.0001; Cochran-Armitage Trend = -5.201, p<.001). The mean age of patients with 1 or 2 diagnostic features was 8 years 126 Figure 5.1: Prevalence by age of Unidentified Bright Objects. The curve is based on 523 neurofibromatosis 1 (NF1) patients from the National Neurofibromatosis Foundation International Database who had cranial MRI. Doted lines indicate 95% confidence intervals. 127 o CM 00 CD 00 CO CM 12 re I o O) < s)U8j)ed jo luoojod 128 Figure 5.2: Unidentified Bright Objects by the number of diagnostic features. Percentage of Unidentifed Bright Objects on MRI by the number of diagnostic features in 523 NF1 patients between 2 and 20 years of age from the National Neurofibromatosis Foundation International Database. The label above each column indicates the number of patients with the respective number of diagnostic features. 129 and the mean age of those with 4, 5 or 6 diagnostic features was 12 years. 5.5.3 Logistic regressive model of unidentifed bright objects versus diagnostic features Table 5.1 summarises the associations between UBOs and other clinical features estimated by multi-covariate logistic regression. Model 1 suggests that cafe-au-lait spots, freckling, cutaneous and plexiform neurofibromas and characteristic bony lesions are not associated with UBOs. These variables were excluded and the remaining associations recalculated in Model 2. Model 2 suggests that significant associations exist among NF1 patients between UBOs and the presence of Lisch nodules, subcutaneous neurofibromas, optic gliomas and neoplasms, but not age or gender. None of the second order interactions was significant. 5.6 Discussion 5.6.1 Unidentifed bright object associations I found UBOs to be associated with the number of diagnostic features in young NF1 patients (Figure 5.2), but most significantly with optic gliomas, other CNS neoplasms, subcutaneous neurofibromas and Lisch nodules (Table 5.1). The percentage of patients with Lisch nodules and neurofibromas increases with age (DeBela et al. 2000b). However, my approach adjusts for age (and gender), so the observed associations are not confounded by age and are not an artefact of age-specifc sampling. The frequencies of these features in NFDB patients are similar to those reported for other patients in this age group (Grifiths et al. 1999; McGaughran et al. 1999). 131 X CD it fl ^ o CD "fl 60 O cd <n CD M C H °0 nn-fl SH fl H 7* . 3 CD ° -fl " O +-* O fe 52 M O OJ  o H-l cn rfl fl CD £ a "2 ° s ° 8 CD 00 ccj OH Z cn C N m o CD - O 1 O '52 PQ 42 CD CD I f l X! A O CO fl C o oo § o _fl c fl CD CD CD fe CD 3 nfl os H o ^ cd ^ ^ .s CD -T3 - 1 CD • 2 2 j a ° 60 fe C+H o T 3 C H - CD C >-< S D -"O CD a £ P -fl 8 g co O u ^ O N S fl "g 5 — J2 " CD "O fe 03 CD •3 * O CD to "T3 52 co 03 o <» S. 0> 60 I- cd 3 ^ 03 ti—i cU O fe co fl fe* .2 • H—» <u 'o 03 co H < CD • fl X ) O - - H w V co 3 OH g .3 £j3 CD fe 0 1 / 5 ^ 1 2 & u « S cd CD fl. ^ cu cd CD 2H fl T 3 O - C -~ s co £ fl <d xi <N o o C3 > v. ca ca 3 ON ^ 3 3 1—* CD 3 u fe a 03 r o O in O N O N O N m o C N in m o cn o 00 00 —< ,—| CD 60 JH CD T 3 fl CD O CD "cd O OH C O +-» 'cd fe) k I ."CD fe cd C N m C N 1 o O N in C N o C N cn o 60 . f l CD CD CO J D 3 O z x 0 CO O fe HJ co cd X o CD z 00 o o o o C N C N O o >n O N o 00 O ^ H CO fl o CD fl u co fl o CD fl CD - O fl C O C N V O cn 1 o C N 00 o O N O O O N *™H m o d d fl . 0 'co CD fe fl O O N O N VO O N O N C N 00 o •xl" o cd OH O CO a CO JS "E, O CD z C O z U | H CD -fl 132 5.6.2 Bias The patients in this study were selected from the NFDB because they had cranial MRI. Most of the centres that contribute data to the NFDB do not routinely do MRI examinations on al young NF1 patients, in accordance with curent recommendations (Gutmann et al. 1997). I did not know why the patients in this study underwent MRI. Therefore, I was concerned that the patients who did have MRIs were more likely to have symptoms of intracranial pathology. All patients in this study underwent MRI examinations, but the frequency of such intracranial lesions as optic and other CNS gliomas was not unusualy high compared to a prospective (Grifiths et al. 1999) and a population-based (McGaughran et al. 1999) studies. Nevertheless, an inclusion bias has to be considered. MRIs on patients in my study were interpreted by a diferent radiologist at each centre. Most of the patients in my study were seen at pediatric clinics (Friedman et al. 1993) and agreement between diferent pediatric radiologists with respect to the presence or absence of UBOs on MRIs of NF1 patients has been shown to be around 85% (DeBela et al. 2000a). Therefore, I expect the criteria for defining UBOs and diferentiating them from initial low-grade gliomas or other brain lesions to be less consistent than if al patients had been examined by the same radiologist. Patient information was not kept from the radiologists who interpreted the MRI scans, and they may have been more likely to recognise UBOs if a diagnosis of NF1 had already been established or was obvious on clinical or radiological examination. This is unlikely to explain the association observed between the occurence of UBOs and such non-prominent features as Lisch nodules. Moreover, I found no statisticaly signifcant 133 association between UBOs and two of the most obvious diagnostic features - cafe-au-lait spots and cutaneous neurofibromas. 5.6.3 Cross-sectional nature of data A recent longitudinal MRI study by Grifiths et al. (1999) found brain tumours in several children with NF1 that developed at the site of a previously-recognised UBOs. In addition, the children who developed brain tumours tended to have more UBOs than other children their age with NF1. I did not analyse the precise location or number of UBOs in these patients, and my data are cross-sectional, not longitudinal. Also, folow-up may help define some of the lesions labeled as UBOs. Nevertheless, I did observe a strong association between the occurence of UBOs and CNS gliomas. The cross-sectional nature of my data preclude conclusions about whether UBOs develop before or after the other associated features in individual NF1 patients. It would be important to assess this prospectively because UBOs are more prevalent at a younger age than Lisch nodules, subcutaneous neurofibromas and neoplasms (Friedman and Riccardi 1999; DeBela et al. 2000a; DeBela et al. 2000b) and might predict the future development of other features in young NF1 patients. For this reason they have been proposed as an additonal diagnostic criterion in children (Curless et al. 1998). UBOs may turn out to be a reliable diagnostic criterion, but not before they are beter defined in terms of location, size and number, and their sensitivity and specificity are properly studied (DeBela et al. 2000a). 134 5.6.4 Pathogenesis My observations support a pathogenetic relationship between UBOs and certain other features in young NF1 patients. The common thread between the associated features (optic gliomas, other neoplasms, Lisch nodules, and subcutaneous neurofibromas) may be dysregulated celular proliferation resulting from haploinsuficiency of neurofibromin (Gutmann et al. 1999; Ingram et al. 2000), but it is hard to understand what this has to do with UBOs if they are simply areas in which the myelin is immature and contains increased fluid. The key may be underlying hyperplastic or dysplastic glial cel proliferation (DiPaolo et al. 1995), which leads to formation of the altered myelin in UBOs. If widespread dysregulation of glial cel proliferation in the brains of young NF1 patients is involved in the pathogenesis of UBOs, the mechanism could be similar to that underlying the formation of optic and other CNS gliomas (Friedman and Birch 1997a), and, by analogy, to the formation of Lisch nodules and subcutaneous neurofibromas in these patients. 5.7 Conclusion UBOs and other NF1 clinical features do not occur independently and may be pathogeneticaly related. 135 A N A L Y S I S O F L O C A L A N D F A M I L I A L F A C T O R S I N N E U R O F I B R O M A T O S I S 1 L E S I O N S 6.1 Hypothesis The development of cafe-au-lait spots, plexiform neurofibromas and cutaneous neurofibromas is influenced by both local and familal factors. 6.2 Objective To test for and estimate associations within body segments and within familes for the occurence of cafe-au-lait spots, plexiform neurofibromas and cutaneous neurofibromas. 6.3 Introduction The defining feature of neurofibromatosis is the neurofibroma: a complex benign tumour arising in the fascicles of peripheral nerves (see section 1.5). Cafe-au-lait spots are another cardinal pathologic feature of NF1. They are present soon after birth in almost al NF1 patients, and their number and size tend to increase during the first decade of life (Riccardi 1982). It has been hypothesised that histamine or other products secreted by mast cels may influence the growth of neurofibromas (Giomo et al. 1989; Riccardi 1992). A neurofibroma that contains an excess of mast cels could stimulate not only its own growth, but that of other neurofibromas nearby. Secreted factor may also be triggered by local trauma. Freckles and other areas of hyperpigmentation tend to occur in skin folds, presumably due to local environmental factors (Fitzpatrick 1981; Riccardi 1992). Cutaneous neurofibromas may arise in an area that has been injured (Riccardi 1990). 137 I have shown in Chapters 2 and 4 that individuals with difuse plexiform neurofibromas are more likely also to have cutaneous neurofibromas and those with cafe-au-lait spots are more likely to have subcutaneous neurofibromas. Also, the occurence of several NF1 clinical features was found to be associated in afected parents and children: Lisch nodules, optic glioma, learning disabilty or mental retardation, macrocephaly and short stature. Here I test the hypotheses that the development of these lesions may be influenced by local or familal factors. 6.4 Subjects and methods 6.4.1 Subjects 547 NF1 patients, including 117 afected individuals in 52 familes, were selected from the NF Instiute database (Riccardi 1992). All of these patients were evaluated between 1979 and 1995 by Dr. Vincent Riccardi, and al meet the NIH diagnostic criteria for NF1 (NIH 1988; Gutmann et al. 1997). For each patient, the presence of 1 or more cafe au lait spots, 1 or more cutaneous neurofibromas, and 1 or more difuse plexiform neurofibromas was recorded for each of the ten divisions of the body surface shown in Figure 6.1. 6.4.2 Analysis of local efect I used two-layered Mantel-Haenszel tests (see section 2.4.3) to look for local associations between the presence of difuse plexiform neurofibromas and cutaneous neurofibromas in individual body segments of each NF1 patient (SPSS 1998). I stratified 138 simultaneously by the body segment being considered and by the number of other body segments with 1 or more cutaneous neurofibromas (a categorical variable with range 0 to 9). This stratification was used to adjust for the fact that an NF1 patient who has a greater total number of body segments with 1 or more neurofibromas is more likely to have at least one neurofibroma in any particular segment than an NF1 patient who has fewer total body segments afected. Confidence intervals for the summary odds-ratios were obtained using a jack-knife based on 20 diferent subgroups -suficient to get stabilty in the estimate. The jack-knife involved calculating the standard eror an additonal 20 times, each time excluding a diferent subgroup. The mean and variance of these 20 additonal standard erors were then used to estimate the true standard eror (Miler 1974). Odds-ratio homogeneity was assessed using the Breslow-Day test (SPSS 1998). Local associations between cafe-au-lait spots and cutaneous neurofibromas and between cafe-au-lait spots and plexiform neurofibromas were also analysed in this manner. 6.4.3 Skin surface area The body divisions used in this study cover varying amounts of skin surface area, so I checked for an association between segment surface area and the presence of >1 cutaneous neurofibroma. Using logistic regression, I set the segment area as the independent variable and the presence or absence of cutaneous neurofibromas as the dependent variable. I also checked for an association between surface area and prevalence of difuse plexiform neurofibromas. Since the median age of my patients was 12 years, I approximated the surface area percentages of the body divisions by taking the 139 mean of the values cited for children and for adults (Palehorse 1997). The proportions of total surface area assigned to each body segment were: head=9%, neck=3%, right upper torso=9%, left upper torso=9%, right lower torso=9%, left lower torso=9%, right arm=9%, left arm=9%, right leg=17%, and left leg=17%. 6.4.4 Total number of neurofibromas In additon to the data on whether each body segment was afected by 1 or more cutaneous neurofibromas, complete counts of cutaneous neurofibromas were available for 44 of the patients. Counts of neurofibromas ranged from none to several hundred and appeared to increase logarithmicaly with the number of afected segments: I log-transformed the counts of neurofibromas and ploted them against number of divisions with 1 or more cutaneous neurofibromas to test whether the number of body segments with 1 or more cutaneous neurofibromas provided a good representation of the total number of cutaneous neurofibromas in an individual. I used linear regression to quantify this relationship (SPSS 1998). Counts of total number of cafe-au-lait spots were not made, and few subjects had more than one plexiform neurofibroma, so these variables were not analysed in this manner. 6.4.5 Analysis of familial factors For the familial analysis, I stratified subjects into 5-year age intervals, calculated the deciles for the total number of segments afected with cutaneous neurofibromas in each stratum, and then ranked each subject by decile for the stratum in which he or she 140 lay. I then used random efects models to obtain maximum likelihood estimates and confidence intervals for intrafamilal corelations for decile (Spjotvol 1967; Dormer et al. 1989). Cafe-au-lait spots and plexiform neurofibromas were also analysed in the same manner. 6.5 Results I studied the recorded distributions of cafe-au-lait spots, cutaneous neurofibromas, and difuse plexiform neurofibromas in 10 segments ofthe body surface of each of 547 patients with NFL Figure 6.1 shows a front view of the 10 segments. Each segment also extends to the dorsal side of the body. Fifty-three percent (n=385) of the subjects were female, and 47% (n=344) male; 77% (n=426) were White, 12% (n=) were Hispanic, 8% (n=44) were Black and 1 % (n=8) were of other or mixed origin. Mean age was 17 years, and median age was 12 years. 6.5.1 Lesion frequency by body segment 210 patients had no cutaneous neurofibromas in any segment. 4 patients had no cafe-au-lait spots in any segment. 331 patients had no plexiform neurofibromas in any segment. Table 6.1 shows the frequency of these lesions in each of the 10 body segments among the NF1 patients included in this study. Cutaneous and plexiform neurofibromas occured with similar frequencies in al 10 body segments. Cafe-au-lait spots occured in almost al patients in al body segments except the head and neck, where they were less frequent. 141 Figure 6.1: Body segment scheme used for recording location of lesions in the Neurofibromatosis Institute Database Each of the 10 segments also extends to the dorsal side of the body. 142 143 6.5.2 Occurence of various lesions in body segments is independent Table 6.2 shows the 10 body segments examined and the pair-wise odds-ratio between neurofibromatosis lesions for each segment. No association was observed between the occurence of cutaneous and difuse plexiform neurofibromas in the same body segment. The summary odds-ratio was 1.20 (95% 0=0.81, 1.79). There was no evidence for heterogeneity across body divisions (p=0.37). Similarly, there was no association between the presence of cafe-au-lait spots and either cutaneous or difuse plexiform neurofibromas within a single body segment. The summary odds-ratios were 1.36 (95% 0=0.91, 2.03) for cafe-au-lait spots and cutaneous neurofibromas and 1.25 (95%) 0=0.74, 2.12) for cafe-au-lait spots and plexiform neurofibromas. There was no evidence of heterogeneity across body divisions for the occurence of plexiform neurofibromas with cafe-au-lait spots (p=0.52), but there was significant (p=0.03) heterogeneity in the occurence of cutaneous neurofibromas and cafe-au-lait spots, with a positve association seen in the neck (odds ratio=2.59; 95% CI=1.23, 5.47). 6.5.3 Relationship between segment size and number of neurofibromas I observed no association between the relative size of the body surface area in a segment and the presence of one or more cutaneous neurofibromas (p=0.74) or of a difuse plexiform neurofibroma (p=0.17). The number of body segments afected with one or more cutaneous neurofibromas was strongly corelated with the total number of cutaneous neurofibromas in 44 NF1 patients in whom both total counts and data on the number of afected body segments were available (r=0.95, p<0.001). The relationship is 144 • CD -*-» ^ - M • - H Cd -T3 C H PH S ' O S-i o kH 3 c r -Tf in a S S3 fa of o tu « CU So C . >, cd j-r» ^ 3 " o -o-a £ oo S « o 00 --H -<_> — cd cd « M ft~r _ o cd © CU *cU • - M « cd cd cu O fa — kH Cd oo cu Q ft cd " * I vo is -° S I 1 3 .2 3 C*H VJ « s a •a I PH 3 cu Z Sj oo fi 3 s c « 'S •s * 3 « rj 3 w cu 3 cu a W) cu 0 s Cj=" NF C<=" o x C<=" ^ ? C^ oo CO r - oo r - Tf CO vo vo ON H t~» ON CN ON ON ° 1 w w w r-» CN ,_, co Tf ,_ t~- in o ON co co co CO I—1 •—1 CN CN > H^ CO in in in in in m «n N ° ^ ^ - N \ ° /-v o x xP vO x o O x vO O 0 s 0 s O 0 s ooinvoTf^Ht-~-Tfc<-i^HOO r - O N C N ^ m ^ H ^ H O N T f i n Tf CN co CN in Tf CN 1 in Tf Co C? Co" C° o x Co" C°" o x Co" oN CO —-H r— t~- CN CN OO OO O O CO CO Tf Tf, in in CO CO Tf Tf vo_ "—' " ' "—' ON 00 ON oo in vo oo ON o VO in in oo 00 o o r—H CN "o •—1 1 — 1 CN CN CN CN CN CN CN CN i CN co Tf m vo r— o 00 ON rH >x O o PH co J5 I CO cd a o X> o =3 CD C eg § J H a ^ 00 "d T3 o „ .a oo cd -° s - O o CD c co fl O CD ii X) CO a .2 '3 o CO CO fl o fl CD CD C CD '-5 PH 52 ^ .2 -S ^° . S fl «U ' O CD — O ^ 00 S3 oo H < a CD oo O a C3 U CA fe z . a CO U > «2 XI CD CO fe z CO s O CD B C5 3 U •o c. CO s 83 i <** 93 U CO fe z CO s © a S3 S U 00 fe z . a CO i. CN r~- co VD (N Id fe e a B V 50 r-~ —^^  —^s CN IT) CO ON t-~ I—I VO CN CO ON ON CO •"ct CN 00 oo d >—I CN •xf o ,_! CN fe CN in VO t-x o o O o d, d d^ © , d VO CN CN -xl-ON CN — i C N ~ H vo C N IT) o o i C N CO d d d C N . "xt , ~ h rn tx •xt -xf °. °\ r- oo vo r-- ^ ^ vo m m C N ^ ~ O © fe -xf fe iri H H 0\ oo CN fe d o © © © © © © "xt Xt O C N o o o •xt ON vo C N 00 CO o o CO >n C N <—1 ON >—1 ON ON C N C N d d d 1 1 T ~ H d d o" CO 00 s OO C N C N ON •xt VO •xt o © © vo fe © "xt" C N CO "xt vq in CO C N CO C N C N C N fe in >—1 <—1 4 ,—1 CO VO C N in ON 00 C N ON CO IT) C N o CO CO o © ON d d d d d d d d fe d m O N co O N m >—. t-- fe fe o O N C O 0 0 C O 0 0 O N f e © 0 N V O o H d d d d H r t ( » i r o ' 1-3 k> -H -H >—1 i> -fl; CD -fl; CD ^ « fe c2 fe >—i C N co in vo t~x o 00 ON - H log linear (Figure 6.2); the regression equation is log(total number of neurofibromas +1) = 0.23*(number of segments afected) + 0.014. 6.5.4 Familal corelations I estimated intrafamilal corelations in the age-adjusted number of body segments (age-specifc decile for number of afected segments) that included one or more cutaneous neurofibromas, one or more cafe-au-lait spots, or one or more plexiform neurofibromas in 117 afected members of 52 familes. I found significant intrafamilal corelations in the number of body segments afected by each of these clinical features. The corelation among relatives in the age-adjusted number of body segments with 1 or more cutaneous neurofibromas was 0.37 (95% CI=0.15,0.55). The corelation among relatives in the age-adjusted number of body segments with 1 or more plexiform neurofibromas was 0.35 (95% CI=0.15,0.57). The corelation among relatives in the age-adjusted number of body segments with 1 or more cafe-au-lait spots was 0.45 (95% CI=0.18,0771). 6.6 Discussion 6.6.1 Lesion severity The number of body segments afected by 1 or more cutaneous neurofibromas appears to provide a good measure of how severely each of these NF1 patients was afected by this disease feature. I found a very high corelation with the number of body segments afected in 44 patients for whom counts of the total number of cutaneous 147 Figure 6.2: Number of affected segments versus number of neurofibromas. Corelation between the number of body segments afected with one or more cutaneous neurofibromas and the total number of cutaneous neurofibromas in 44 NF 1 patients (r=0.95, p<0.001). Total neurofibroma counts and data on the number of afected body segments were available in al patients. The relationship is log linear; the regression equation is Log(total number of neurofibromas +1) = 0.23*(number of segments afected) + 0.014. 148 CN • 00 • • • • • • • • • • • • • CD P ^ • CM h o J l CM O CN o CN LO LO o o L O (l + sewojqyojrtsu snoeuemo jo J8qwnu)6o| neurofibromas were available (Figure 6.2). It seems likely that a corelation also exists between the number of body segments afected with cafe-au-lait spots or plexiform neurofibromas and the severity of each of these disease features, but count data were not available in this study to demonstrate this. 6.6.2 Ascertainment issues The subjects were refered to a specialised clinic, so I was concerned that they are more severely afected than the NF1 patient population at large. The frequencies of cutaneous neurofibromas, plexiform neurofibromas and cafe-au-lait spots are comparable to those from another large database (Friedman and Birch 1997b) and population based studies (Samuelsson and Axelsson 1981; Huson et al. 1989a; McGaughran et al. 1999). Al of the subjects in this study were examined by the same clinician, ruling out the bias inherent in using multiple examiners. Body segment data were colected over a single visit for each patient, but the colection period lasted 16 years. I expect this to make the data less consistent than if they were gathered over a shorter period of time. 6.6.3 Local associations of lesions I have shown in Chapter 2 that individuals with difuse plexiform neurofibromas are more likely also to have cutaneous neurofibromas, but this association did not take into account the location or number of these lesions. The curent study is the first to examine this association within body divisions. Since virtualy al difuse plexiform neurofibromas are of congenital origin (Riccardi 1992), I wanted to find out if they 150 influence the subsequent development of cutaneous neurofibromas. These findings suggest that the occurence of cutaneous neurofibromas in NF1 patients is not influenced by the local presence of a difuse plexiform neurofibroma. In fact, I found that al three of the lesions studied (cafe-au-lait spots, cutaneous neurofibromas, and plexiform neurofibromas) occured independently of each another in almost al of the body segments analysed (Table 6.2). This is consistent with an independent triggering mechanism for each lesion. 6.6.4 Local factors in pathogenesis Mast cels have been implicated in the pathogenesis of plexiform neurofibromas (Giorno et al. 1989). It has been hypothesised that histamine or other products secreted by these cels may influence the growth of neurofibromas, either the plexiform neurofibroma itself or other cutaneous or subcutaneous lesions (Riccardi 1992). This efect would presumably be strongest near the original plexiform neurofibroma. My results argue against the hypothesis that factors secreted by a difuse plexiform neurofibroma stimulate the development of nearby cutaneous neurofibromas. Local trauma has also been implicated in the pathogenesis of neurofibromatosis lesions. Cutaneous neurofibromas may arise in an area that has been injured (Riccardi 1990). Freckles and other areas of hyperpigmentation tend to occur in skin folds, presumably due to local environmental factors (Fitzpatrick 1981; Riccardi 1992). I found a significant association between cafe-au-lait spots and cutaneous neurofibromas only in the neck segment. Cafe-au-lait spots were found on al segments except the head and neck in almost al patients. Most body segments may lack the variation needed to 151 observe an association, although a strong (but not statisticaly significant) association was observed between cafe-au-lait spots and plexiform neurofibromas on the left leg (Table 6.2). The result could also be due to chance alone, since many diferent associations were calculated. A pathological reason the neck might be afected by both lesions is recurent minor trauma to the skin associated with flexion, extension, and rotation of the head (Riccardi 1990). It would be interesting to study local associations in more detail by checking specific lesions, such as cutaneous neurofibromas, tend to aggregate in the same body segment. Clearly, however, other factors are also involved in the pathogenesis of cafe-au-lait spots and neurofibromas, as indicated by the familal corelations I observed for the age-adjusted number of body segments afected by each ofthe three lesions studied. 6.6.5 Familal associations The significant intrafamilal corelations I found are consistent with other evidence that familial factors contribute to the development of cutaneous neurofibromas and cafe-au-lait spots in patients with NF1 (Easton et al. 1993) (see Chapters 2 and 7). The number of familial patients and the prevalences of cutaneous and plexiform neurofibromas and cafe-au-lait spots in patients were similar among the three study groups (Riccardi 1992; Easton et al. 1993; Friedman and Birch 1997b). The present study found a similar corelation for cafe-au-lait spots but higher corelations for cutaneous neurofibromas than Easton et al.. I also found a significant corelation for plexiform neurofibromas, whereas Easton only analysed this feature as a binary trait and found no familial association. My use of 10 body segments alows for more precise 152 stratification than Easton's 5 ordinal categories and may account partly for the diferences. 6.6.6 Familal factors in pathogenesis These familial associations suggest a genetic influence on the severity of NF1 lesions. Contributing factors may include efects of the mutant NF1 alele itself, efects of the normal NF1 alele, or modifying efects of other loci. The moderate magnitudes of the intrafamilal corelation coeficients show that family history alone is insuficient to predict the degree to which a patient wil be afected with these lesions. Other elements must also be involved. Previous studies support the hypothesis that some patients are predisposed to lesions such as cafe-au-lait spots, cutaneous neurofibromas and plexiform neurofibromas. The results of this chapter are consistent with the possibilty that diferent pathogenetic mechanisms are responsible for the three lesions studied. Cutaneous neurofibromas are composed of Schwann cels, fibroblasts, mast cels and nerve cel axons (Korf 1999a). Schwann cels are thought to play a major role in tumourgenesis since they have been found to undergo loss of heterozygosity at the NF1 locus, while other neurofibroma cels have not (Kluwe et al. 1999; Rutkowski et al. 2000). Although the mechanism by which loss of NF1 function leads to neurofibromas is unknown, Schwann cels migrate during embryonic development from the neural crest (Johnston et al. 1981) and give rise to cutaneous neurofibromas arise through neoplasia (excess celular division), at least in many cases (Stark et al. 1995). Although plexiform neurofibromas also contain Schwann cels, fibroblasts, mast 153 cels and axons, they are congenital and have more characteristics of dysplasia (irregular tissue arangement) than do cutaneous neurofibromas (Riccardi 1992). Plexiform neurofibromas usualy have extensive vascularisation, involve many diferent tissues and can spread to distort adjacent tissues (Korf 1999a). Chimeric mice composed in part of Nfl"1" cels develop plexiform neurofibromas but not cutaneous neurofibromas (Cichowski et al. 1999; Vogel et al. 1999). On the other hand, transgenic mice expressing the human T-lymphotropic virus type 1 tax gene develop cutaneous neurofibromas (Hinrichs et al. 1987; Nerenberg et al. 1987; Green et al. 1992; Feigenbaum et al. 1996). Although repression of Nfl expression by tax occurs in the absence of mutation at the Nfl locus, these observations suggests that plexiform and discrete neurofibromas can arise by pathways that are independent, at least in mice. Like neurofibromas, cafe-au-lait spots contain neural crest-derived cels, but these are melanocytes with abnormaly large pigment particles (Fitzpatrick 1981) rather than Schwann cels, as in neurofibromas. Some familes with NF1 mutations develop cafe-au-lait spots but no tumours, consistent with diferent pathogenetic factors being involved in the development of neurofibromas (Abeliovich et al. 1995). 6.7 Conclusion These findings are consistent with multiple factors being involved in the pathogenesis of both plexiform and cutaneous neurofibromas as wel as of cafe-au-lait spots. Some of these factors appear to be genetic, but others do not. Although some of the pathogenetic factors may be shared among these three lesions, others appear to difer. 154 A N A L Y S I S O F I N T R A - F A M I L I A L P H E N O T Y P I C V A R I A T I O N I N N E U R O F I B R O M A T O S I S 1 155 7.1 Hypothesis Genetic sources of variable expressivity are generaly important in NF 1 and vary for diferent clinical features. 7.2 Objective To measure associations of NF1 features among afected sibs, children and their mothers and fathers, and 2 n d degree relatives using methods that take other clinical features and age into account and adjust for the non-independence of afected relatives. 7.3 Introduction Many clinical features of NF1 are progressive, but the rate of progression and the occurence of serious manifestations vary greatly from one patient to another (Friedman and Riccardi 1999). This variabilty and the confounding efect of age have hindered eforts to characterise the relationship of genetic factors at the NF1 locus or other loci to disease variability. The large size of the NF1 gene makes mutational analysis dificult and there is not much evidence for alele-phenotype corelations in most NF1 patients (see section 1.13). Variation in the mutant NF1 alele itself is insuficient to account for the variabilty of most disease features. Mouse models provide evidence that genetic factors at other loci can afect the phenotype associated with Nfl mutations but are limited to only a few of NF1 disease features observed in humans (see section 1.11). Easton et al. (1993) found evidence for genetic factors influencing the number or presence of several features in 156 NF1 patients. These results rely heavily on near-perfect concordance among 6 pairs of monozygotic twins and some were not adjusted for the non-independence of multiple relative-pairs from the same family. I have shown in Chapters 2 and 4 that several statisticaly significant associations exist between the occurence of individual clinical features in unrelated probands with NF1. The results in Chapter 6 suggest that cafe-au-lait spots, cutaneous neurofibromas and plexiform neurofibromas are influenced by familal factors. I also found significant associations in the occurence of Lisch nodules, optic glioma, learning disability, macrocephaly and short stature in afected parent-child pairs (see Chapter 2) but made no atempt to adjust for the non-independence of multiple relative-pairs from the same family or for associations among clinical features in individuals in that preliminary study. The statistical methodology for testing for such non-independence is wel developed for continuous variables and is widely available as part of software packages such as S.A.G.E. (1997). Many NF1 clinical features are discrete by nature, and the methodology for this kind of analysis is not wel developed for discrete variables. Prior to this study, the only available programme was Generalized Estimating Equations 2 (GEE2), but it did not alow for complete intra-familal relationship classifications (Liang and Beaty 1991). In this study, I used newly revised programmes for multivariate probit analysis (MPROBIT) and multivariate normal analysis (MVNFAM) to extend my previous analyses (Joe 2000). I measured associations of NF1 features among afected sibs, children and their mothers and fathers, and 2nd degree relatives using methods that take other clinical features and age into account and adjust for the non-independence of afected relatives. By comparing the associations for relatives of various types, I provide 157 evidence that genetic sources of variable expressivity are generaly important in NF1 and vary for diferent clinical features. 7.4 Subjects and methods 7.4.1 Subjects All patients in this study meet the NIH diagnostic criteria for NF1 (NIH 1988; Gutmann et al. 1997). Data were obtained from the National NF Foundation International Database (NFDB) (Friedman et al. 1993) on 913 individuals from 373 familes with 2 or more afected members, including 268 sib-sib, 373 parent-child and 74 2nd degree relative pairs. 7.4.2 Features For analysis of familiality, I selected 12 clinical features of NF1: cafe-au-lait spots, intertriginous freckling, Lisch nodules, cutaneous neurofibromas, subcutaneous neurofibromas, plexiform neurofibromas, head circumference, stature, seizures, scoliosis, optic glioma and neoplasms other than neurofibromas and optic gliomas ("other neoplasms"). Most of the features were identified by physical examination and treated as binary variables. The NIH Criteria (NIH 1988) (Table 1.1) were used whenever applicable. Cafe-au-lait spots were coded as "present" if the subject had 6 or more spots. Cutaneous or subcutaneous neurofibromas were coded as "present" if the subject had two or more lesions of the same type. Plexiform neurofibroma was coded as "present" if the subject had one or more lesions. Stature and head circumference were treated as 158 continuous variables and standardised prior to analysis using population norms (see section 3.4.3). Lisch nodules were diagnosed or excluded by a slit lamp examination. The presence or absence of optic glioma was determined by cranial MRI or CT examination. Only patients with definite presence or absence of a feature were considered in models involving that feature. The complete data set used in this study is available from ftp:/ftp.stat.ubc.ca/pub/hjoe. 7.4.3 Confounding factors I included age as a covariate in al analyses. Age is one of the most important factors influencing the NF1 phenotype (Zoler et al. 1995). Many NF1 features, including Lisch nodules, subcutaneous neurofibromas, cutaneous neurofibromas, other neoplasms, intertriginous freckling, seizures and scoliosis have a higher prevalence in older patients. This is ilustrated by DeBela et al. (2000b) and in Figures 4.1-4.13. Furthermore, clinical features do not occur independently inNFl patients, even after adjusting for the efect of age (see Chapters 2 and 4). Therefore, I also controled for the presence or absence of other associated features to minimise confounding in the present study. 7.4.4 Multivariate probit and normal models My general approach was to treat the features of NF1 as if each were a disorder occuring in a population afected with NF1. Familal aggregation of each binary feature among various classes of relatives was estimated using multivariate probit regression 159 models (Joe 1997), which assume that each of the binary dependent features reflects an underlying latent quantiative variable (Thompson et al. 1991). Whereas linear regression atempts to compute the mathematical relationship between the covariates and the response variable directly, probit regression atempts to compute the relationship between the covariates and the probabilty of the response variable being "present" -expressed in terms of the normal distribution. Familal aggregation of continuous features (head circumference and stature) among various classes of relatives was estimated using multivariate normal regression models. This is similar to probit regression except that the continuous variable is modeled directly, rather than as a probabilty. The program MPROBIT was used for binary features and MVNFAM for continuous features (Joe 2000). Multivariate probit regression consists of two sets of parameters: regression parameters for covariates such as related features, interactions between related features (each represented by a distinct variable equal to the product of the two interacting features), age and gender; latent corelation parameters for binary features between specifc classes of relatives. For head circumference and stature, the second set of parameters are corelations (not latent corelations) of the continuous feature between specifc classes of relatives. MPROBIT and MVNFAM provide maximum likelihood estimates of the regression coeficients together with standard errors. The programmes estimate the corelation or latent corelation coeficients and standard erors for the intra-familial relationships specifed and a covariance matrix of al parameter estimates. I used the results of my study of individuals with NF1 (see Chapter 4) to obtain appropriate functions for age (e.g. e'ageU) and initial regression parameter estimates for 160 covariates representing related features, interactions between related features and gender. Familal aggregation was assessed among sibs, parent-child pairs (including mother-child and father-child pairs separately) and 2nd degree relatives. 7.4.5 Assessment of intrafamilal corelations Parameters and coeficients with 95% confidence intervals that excluded zero were deemed statisticaly significant. Standard erors and covariance matrices were used to test for diferences between intra-familal corelation coeficients for diferent comparisons. For example, to test for a diference between sib-sib corelation and parent-child corelation I used the folowing formulas: Z = where s = y (SEr*) + (SErpc) - 2 cov(r«, rpc) s Z-scores were converted into p-values according to the standard normal distribution. I used one-tailed tests to compare corelations between 1st degree and 2nd degree relatives and between sib pairs and parent-child pairs because I had a prior expectation that corelations between 1st degree relatives would be at least as strong as those between 2nd degree relatives (Easton et al. 1993) and that corelations between sibs would be at least as strong as those between parents and children (see Chapter 2). I used two-tailed tests to compare mother-child corelations to father-child corelations. 161 7.5 Results 7.5.1 Feature prevalences I studied 913 individuals with NF1 from 373 familes with two or more afected members. 91% of the individuals studied were White, 2% Asian, 1% Black, 1% Latin, and the remainder either of "other" or "unknown" origin. Table 7.1 shows the prevalences of each of the 12 NF1 clinical features in afected fathers, mothers and their afected children in this NFDB study sample and compares them to the prevalences in the sample used by Easton et al. (1993). 7.5.2 Individual covariates Familal aggregation among various classes of relatives was estimated using multivariate probit and multivariate normal regression models. Table 7.2 shows the regression parameters and standard erors for the covariates that were included in each model. The importance of these covariates was first identified in Chapter 4. Unlike the parameter estimates in Table 4.2, the estimates in Table 7.2 have been calculated for each feature taking into account intra-familal associations (see next section). The strength of association between the modeled feature and a covariate is measured by /?. The regression coeficient in probit regression is approximately half that for logistic regression. A unit increase in the value of the covariate means the modeled feature is exp(2/7) times more likely to be present. For example, subjects with intertriginous freckling were exp(2x.51)=2.8 times more likely also to have cafe-au-lait spots than 162 oo cd <u to yi <D cd on fi O > O N ON cd oo 00 cd 13 CU O N a O N o 00 C3 to o to » 00 5 " CU OO CU w > ^ -s = a! < oo M © ja •<-< a « Q to >- — 2 -S 2 tS C J .<u cu •** cu .4> _ o on CU -fi to tu M s • M tu to N ; 00 oo ON CN N ? VO —> ON N ? N ? CN CN O in vo r-ON O Tf o Tf N " Nr m m O oo o x in in in ,—i ON IT) m Tf ON ON NO m CN NO ^ Co" o x Co o x Tf c- in cn oo r~- CN CN ,_, r- Tf 00 oo NO m ON CN r-i r n O NO CN 00 NO in r--^ N ? 00 NO — i NO Tf —' Tf N ? N O CN CN ^H Tf NO m ON m Tf O N r n CN m vo cn ON o-- NO r--fi TT m NO - M Co" o x Co" Co" Co" o x m Tf oo CN f-H CN c--r- oo CN cn cn o x o oo fi —< o cn r-» m CN 0 s ON NO ON 0 s - ON T J - in CN ON ON o x 00 00 Tf .o 00 00 NO fi cn NO NO 00 <D Co" ox Co" Co" ox Co" O cn cn 00 NO 00oo Tf O ON NO OO in < oo m CN ^ - V N ° O N O r- CN CN m >—i m 0 \ T f r - CN r - Tf CN N ; CN o--CN m ON Tf CN N ° N ° O X O X NO NO 00 o x o x 00 Tf r- Tf fi O fi .fi o oo 00 cd 00 cd a o cd C2 oo fi a o 00 1-1 s - fi O 00 fib O CD fib ule n fib fib ule tane< o "oh o % tane< CD O • i-H bcu CD c fi fi a fi fi U O 0 0 o S * s i . fi 8 M 0 0 OH 0 0 O Table 7 . 2 : Summary of multivariate normal and probit models of NF1 clinical features. Summary of regressions in multivariate models for 12 clinical NF1 features. The 1st column lists the 12 modeled features. The 2nd-4th columns show the covariates and their regression parameter estimates (P) with standard erors (SE) used in each model. Po is the intercept in the model equation. Each regression accounts for covariates such as related features, interactions between related features, age and gender. Interactions are depicted by features separated by an "*" and their values equal the product of the two interacting features. Modelled Feature Intercept and Covariates P SE Lisch nodules Po .65 (.08) Age -3.55 (.32) Male gender -.01 (.08) CLS .23 (.15) CNF .44 (.20) CLS * CNF -.09 (.22) Cafe-au-lait spots Po .28 (.14) (CLS) Age -.66 (.25) Male gender .03 (.09) Freckling .51 (.12) SNF -.41 (.26) Freckling * SNF .61 (.28) Head circumference Po -.99 (.10) (OFC) Age .62 (.21) Male gender -.09 (.31) Lisch nodules -.06 (.36) Optic glioma .56 (.44) Stature .34 (.04) Neoplasms .10 (.75) Cutaneous neurofibromas Po -1.62 (.11) (CNF) Age -5.56 (.36) Male gender .01 (.10) SNF .62 (.11) PNF .36 (.12) Stature Po -.62 (.09) Age -.82 (.31) Male gender -.03 (.09) OFC .04 (.01) Optic glioma Po -1.02 (.13) Age .72 (.57) Male gender .06 (.17) PNF .01 (.37) OFC .19 (.07) Neoplasms .55 (.49) Table 7.2 continued: Modeled Feature Intercept and Covariates P SE Subcutaneous neurofibromas Po -1.72 (•12) (SNF) Age -3.78 (.35) Male gender -.04 (.08) CLS .43 (.11) CNF .73 (.13) PNF .52 (.17) Freckling * PNF -.24 (.23) Intertriginous freckling Po .49 (.15) (Freckling) Age -1.58 (.30) Male gender -.23 (.12) CLS .52 (.14) SNF -.18 (.27) Lisch nodules .55 (.14) CLS * SNF .62 (.33) Seizures Po -1.43 (.11) Age -.88 (.65) Male gender -.04 (.15) Plexiform neurofibromas (PNF) Po -1.11 (.11) Age -.88 (.38) Male gender .07 (.09) SNF .46 (.16) CNF .37 (.14) SNF * CNF -.21 (.22) Scoliosis Po -1.11 (.09) Age -.57 (.34) Male gender -.02 (.11) Other neoplasms Po -.95 (.23) Age -4.07 (2.11) Male gender -.06 (.21) Lisch nodules -.55 (.25) Optic glioma .32 (.31) subjects of the same age and gender without intertriginous freckling. Also, subjects with intertriginous freckling and subcutaneous neurofibromas were exp [2x(.51-.41+.61)]=4.1 times more likely to also have cafe-au-lait spots. The parameter estimates for age were highly significant (p<0.001) for Lisch nodules, subcutaneous neurofibromas, cutaneous neurofibromas, and intertriginous freckling; significant (p<0.05) for cafe-au-lait spots, optic gliomas, plexiform neurofibromas, and the continuous variables head circumference and stature; and not significant (p>0.05) for other neoplasms, seizures and scoliosis. The parameter estimate for gender was not significant in any of the models. However, parameters that are not statisticaly significant on their own can stil  contribute to model interpretation and signifcance when other related features are also considered. 7.5.3 Intrafamilal associations Table 7.3 shows the number of sib, parent-child (including mother-child and father child) and 2nd degree relative pairs used in each model. Subjects were included in a model only if the status ("presence" or "absence") of the modeled feature and al covariates was known. The measure of association for multivariate probit analysis of binary variables is the latent correlation. The measure of association for multivariate normal analysis of continuous variables is the correlation. Figure 7.1 shows the adjusted intrafamilal corelation or latent corelation coeficients and their 95% confidence intervals for each 166 CD I 3 £ o 5 o •5 <N o s CO fe "3 .g O CO , C D <U fe C H C3 fe c3 O > H O CU o •2 -=> co G «D ri u c 3 H CD S3 Xj r ^ CD 3 fl . C D CD Z ^ o S f l CD | 3 g FO O C  © U X! S .C CU i O .3 CN 3 « CD w X3 - X! 3 • CD S CD .fl n fe o s i s cs "5 •O C H > « _ J H co -O O 3 CD fe 8 3 CO 03 -*-< r-i ^ CD « CN . H u fl .. . ca % CD fe ° XI 2 CJ CO J ) _ 0 H CD * CO H S XS 60 s c -§ z .. nTcN ^ 8 ~" r- „ fl aj •» cs «S X , fl H h " § CO J-S3 OH H CD • J3 ' CD OS CO ' « fe C3 fe co U *3 fe s-X i H .9 CO '3 fe CU i. 3 es CD fe <u "3 T3 o ^ O ^- * ONir iTj -ONro r O V O C N V O c N ^ t V O r o r ^ v o i n r o ON fe fe O fe fe O N r N C ^ r n r x v o r o i o - ^ - r o r o o f-~.fer<-)fer<->cNfet~xfefefecN O N O Tf O o o m ' x t - f e lOfeTtrNxtr^cNTi-mcNONro feCNVOfNNOrOCNfefNcNfern ( N o o m - x t m m a oo oo ON-v)-ovoomir>rx-vovo(NC~~-feCNfeCNfe>^fNfe(NCNCN^|-fe co O 167 Figure 7.1: Feature associations among relatives with NF1. Adjusted intrafamilal corelation or latent corelation coeficients and their 95% confidence intervals are shown for each of the 12 features among al 913 relatives with NF1 from the 373 familes studied. In these estimates, al relatives are treated the same regardless of relationship. 168 o in © c o « 0) o o o CN (0 a> -a o c o to a) o c a> (0 re E o E U T3 (0 <D I 3 0) C w 3 O a> c ra 4-1 3 U •Q 3 CO CO ra E o 3 a> c CO 3 o a> c 3 o 3 i2 o Q. CO C IS o a> 3 ra iff ra O co 3 O c t re E o O) o a O CO E CO n Q. O a> c >_ <D x: CO CU N '53 CO CO 'to o o o CO CO re E o 3 O c E £ x 0) 169 of the 12 features among al 913 relatives withNFl from the 373 familes studied. In these estimates, al relatives are treated the same regardless of relationship. Statisticaly significant positve intrafamilal corelations or latent corelations were observed for Lisch nodules, head circumference, subcutaneous neurofibromas, cutaneous neurofibromas, stature, cafe-au-lait spots and intertriginous freckling. Latent corelations for optic glioma, other neoplasms, seizures, scoliosis and plexiform neurofibromas, although positive, were not statisticaly diferent from zero. However, the number of individuals who had the later features, especialy optic glioma, other neoplasms, or seizures, was smal, and the confidence intervals are very wide. Figure 7.2 shows the adjusted intrafamilal corelation or latent corelation coeficients and 95% confidence intervals for 8 clinical features among 746 afected 1st degree relatives and among 148 afected 2nd degree relatives. MPROBIT failed to converge on latent corelation coeficients between 2nd degree relatives for optic glioma, other neoplasms, seizures or scoliosis because of the low frequency of these features and insuficient sample size. I did obtain latent corelation coeficients between 1st degree relatives for these features, but none was significantly diferent from zero. Statisticaly significant positve corelations or latent corelations between 1st degree relatives were found for 7 ofthe 8 other features listed in Figure 7.1. Signifcant positve corelations or latent corelations between 2nd degree relatives were also found for 4 of these 8 features. Signifcant negative corelations were not observed for any of the features. Latent corelations were significantly greater among 1st degree relatives than among 2nd degree relatives for Lisch nodules (p=0.0001) and cafe-au-lait spots (p=0.0004). Corelations or 170 Figure 7.2: Feature associations among 1 s t and 2 n d degree relatives with NF1. Adjusted intrafamilal corelation or latent corelation coeficients and 95% confidence intervals are shown for 8 clinical features among 746 afected 1st degree relatives and among 148 afected 2nd degree relatives. A star indicates a significant diference between the coeficient among 1st degree relatives and the coeficient among 2nd degree relatives. MPROBIT failed to converge on latent corelation coeficients among 2nd degree relatives for optic glioma, other neoplasms, seizures or scoliosis because of the low frequency of these features and insuficient sample size. 171 CO o 3 o c o I/) 0) u c CD E 3 U "o T3 (0 CD X (A (A 3 re (0 3 re neo E E neo o i_ neo o i_ re 4-1 -Q neo .a • H re •«--ubci O +J o ubci eur Cu eur V) c c <D 3 re (0 (0 o a <A re 3 re re O CD C a CD (A 3 O c TO »Z CD +J C (A re i i o »-x -1= CD O E 3 CD C latent corelations among 1st degree relatives were not statisticaly diferent from corelations among 2nd degree relatives for head circumference (p=0.15), subcutaneous neurofibromas (p=0.06), cutaneous neurofibromas (p=0.49), stature (p=0.30), intertriginous freckling (p=0.07) or plexiform neurofibromas (p=0.11). Figure 7.3 shows the adjusted intrafamilal corelation or latent corelation coeficients and 95% confidence intervals for 8 features among 268 afected sib pairs and among 373 afected parent-child pairs. Again, MPROBIT failed to converge on latent corelation coeficients between sib or parent-child pairs for optic glioma, other neoplasms, seizures or scoliosis. Statisticaly significant positve corelations or latent corelations between sibs were found for al 8 features in Figure 7.3. Signifcant positve corelations or latent corelations between parents and children were found for 6 of the 8 features. Significant negative corelations were not observed for any of the features. Latent corelations were significantly greater between sibs than between parents and children for subcutaneous neurofibromas (p=0.04), cafe-au-lait spots (p=0.001), intertriginous freckling (p=0.03) and plexiform neurofibromas (p=0.02). Corelations or latent corelations between sibs were not statisticaly diferent from the corelations between parents and children for Lisch nodules (p=0.40), head circumference (p=0.45), cutaneous neurofibromas (p=0.29), or stature (p=0.20). Figure 7.4 shows the adjusted intrafamilal corelation or latent corelation coeficients and 95% confidence intervals for 8 features between 233 afected mother-child pairs and between 140 afected father-child pairs. Statisticaly significant positve corelations or latent corelations between mothers and children were found for 5 of the 8 features. Significant positve corelations or latent corelations between fathers and 173 Figure 7.3: Feature associations among sib and parent-child pairs with N F 1 . Adjusted intrafamilal corelation or latent corelation coeficients and 95% confidence intervals are shown for 8 features among 268 afected sib pairs and among 373 afected parent-child pairs. A star indicates a significant diference between the coeficient among sibs and the coeficient among parents and children. MPROBIT failed to converge on latent corelation coeficients among sibs or parent-child pairs for optic glioma, other neoplasms, seizures or scoliosis. 174 Figure 7.4: Feature associations among mother- and father-child pairs with NF1. Adjusted intrafamilal corelation or latent corelation coeficients and 95% confidence intervals are shown for 8 features between 233 afected mother-child pairs and between 140 afected father-child pairs. A star indicates a significant diference between the coeficient between mother-child pairs and the coeficient between father-child pairs. 176 "D CO O) S O "E ^ <5 177 children were found for 6 of the 8 features. Signifcant negative corelations were not observed for any of the features in either relationship. Latent corelations between fathers and children are significantly greater than latent corelations between mothers and children for Lisch nodules (p=0.001), subcutaneous neurofibromas (p=0.0001) and cutaneous neurofibromas (p=0.02). Corelations or latent corelations do not difer significantly between father-child pairs and mother-child pairs for head circumference (p=0.85), stature (p=0.40), cafe-au-lait spots (p=0.62), intertriginous freckling (p=0.71) and plexiform neurofibromas (p=0.17). 7.6 Discussion 7.6.1 Variable expressivity in other disorders Variable expressivity is a characteristic of many dominantly-inherited human genetic diseases and may have genetic or non-genetic causes. Possible genetic causes of variable expressivity include the efects of diferences in the mutant alele, efects of the normal alele, and the efects of modifying genes. For example, analysis of phenotypic variation in von-Hippel-Lindau (VHL) disease has implicated unlinked modifying genes in the pathogenesis of ocular tumours (Webster et al. 1998). In another example the risk of ovarian cancer in BRCA1 mutation cariers is modifed by alelic variation at the unlinked H-RAS locus (Phelan et al. 1996). My study was designed to evaluate the relative importance of various genetic mechanisms in the interfamilal and intrafamilal variabilty of NF1. 178 7.6.2 Statistical signifcance I analysed familial latent corelations for 10 NF1 clinical features treated as discrete variables and corelations for 2 NF1 clinical features treated as continuous variables, while adjusting for other related features, age and gender through statistical modeling. I found 7 of the features to have significant overal intra-familal associations (corelations or latent corelations) (Figure 7.1). I was also able to test for diferences between associations among various classes of relatives for 8 of the 12 features studied. Diferences between various classes of relatives were found for 6 of the 7 features with significant overal intra-familal associations (Figures 7.2-7.4). Several features had significantly positve associations among 2nd degree relatives, but none were significantly greater than the associations for the same feature among 1st degree relatives (Figure 7.2). Similarly, several features had significantly positve associations between parents and children, but none were greater than associations for the same feature between sibs (Figure 7.3). I observed no significant negative associations, or positve associations that were greater in 2nd degree than 1st degree relatives or greater in parent-child pairs than among sibs. I would expect to observe such associations if chance were a major factor. This supports the statistical validity of my approach. 7.6.3 Representativeness of the sample The NFDB draws its information from specialised clinics, so I was concerned about the representativeness of my sample. Furthermore, patients with unknown status of a feature were excluded from models involving that feature. Nevertheless, frequencies of 179 most features found among the familial cases used in this study (Table 7.1) are comparable to those found in another family study of variable NF1 expressivity (Easton et al. 1993). Cutaneous neurofibromas are twice as prevalent in the NFDB sample than in the sample used by Easton et al., due to a larger number of older patients in the later study. Feature prevalences are also comparable to prevalences from two available population-based studies of NF1 patients (Samuelsson and Axelsson 1981; Huson et al. 1989a) (Table 1.3). 7.6.4 Comparison to other studies Easton et al. (1993) studied 175 individuals with NF1 from 48 familes, including 6 pairs of monozygotic twins, 76 pairs of sibs, 60 parent-ofspring pairs, 54 2nd degree relative pairs and 43 3rd degree relative pairs. They examined 8 NF1 clinical features and found significant intrafamilal corelations for 3 quantiative variables: number of cafe-au-lait spots, number of cutaneous neurofibromas and head circumference. They also analysed 5 traits as binary variables, but these comparisons did not include adjustments for age. Furthermore, none of their analyses adjusted for the non-independence of multiple relative-pairs from the same family or of various clinical features. My sample size is 5 times larger, and I examined 12 clinical features, 6 of whicli are the same as Easton's. Also, I included associations between features as covariates in the familial analyses. Unlike Easton et al., I did not have counts of cafe-au-lait spots and dermal discrete neurofibromas, but Easton's quantiative investigations of these features complement my binary analyses nicely. Both studies found are consistent with modifying gene influence on cafe-au-lait spots, but not on dermal discrete neurofibromas. 180 In all, 10 of my 12 features were treated as binary variables -1 had quantiative data only on stature and head circumference. Many of the clinical features of NF1 (and other diseases) are by nature binary, and ours is the first study to examine associations for binary traits among diferent familial relationships while accounting for continuous covariates such as age. Similar methods have been used to study lens opacites (Framingham 1994) and liver cancer (Liang and Beaty 1991) in individuals who do not have NF1, but I may be the first to study an autosomal dominant disease in this manner. 7.6.5 Sample size considerations Although this is by far the largest group of NF1 familes ever studied, I only had 74 pairs of 2nd degree relatives. Models for most features used even fewer 2nd degree relatives because the data were incomplete. Subjects were included in a model only if the status of the modeled feature and al covariates was known (Table 7.3). These relatively smal sample sizes are reflected in the wide 95% confidence intervals for the corelation and latent corelation coeficients among 2nd degree relatives (Figure 7.2). Furthermore, statistical techniques are less reliable for smaler sample sizes, so I must atach an additonal note of caution to the point estimates for the corelation and latent corelation coeficients between 2nd degree relatives, particularly for Lisch nodules, head circumference, stature and intertriginous freckling, in which the analysis included 35 or fewer pairs of 2nd degree relatives (Table 7.3). 181 7.6.6 Minimising the confounding efect of age The most important confounding factor in familal analyses of NF1 is age. Many disease features are more prevalent in older NF1 patients (Cnossen et al. 1998), and, if not appropriately controled, age might produce an association between afected relatives of similar age (e.g., sibs) or obscure an association between relatives of very diferent ages (e.g., parents and children). My multivariate models minimise the confounding efect of age, but they may not eliminate this efect completely. The covariate representing age was significant in models for most features, but it is possible that a residual age efect is contributing to the observed diferences between sib and parent-child pairs for features such as subcutaneous neurofibromas and intertriginous freckling that become more prevalent with age (Figure 7.3). Age is less likely to influence the intrafamilal associations for cafe-au-lait spots or plexiform neurofibromas, which, when considered as discrete variables, occur with a relatively stable frequency with age (Figures 4.1 and 4.5) (Riccardi 1992; DeBela et al. 2000b). 7.6.7 Stati stical si gnifcance I used one-tailed tests for 1st degree vs. 2nd degree and sib-sib vs. parent-child comparisons. Several of the results just reach a level of nominal statistical signifcance using one-tailed z-tests, and several others fal only a little short of doing so. Clearly these results require independent confirmation in future studies. 182 7.6.8 Interpretation of intrafamilal associations Lisch nodules had significantly higher latent corelations among 1st degree relatives than among 2nd degree relatives (Figure 7.2). Higher latent corelations for 1st than 2nd degree relatives are consistent with efects produced by modifying genes at unlinked loci but might also result from environmental factors that are more likely to be shared among closer relatives. However, it is hard to imagine how specific environmental factors (such as viruses) could contribute to the development of Lisch nodules. Furthermore, sib pair and parent-child associations were similar for Lisch nodules (Figure 7.3), suggesting that environmental factors (at least those shared by sibs but not by parents and children) are not responsible for this association. No family studies have previously been done on Lisch nodules, and factors contributing to their development are unknown. My observations are consistent with the efect of a modifying gene on the pathogenesis of Lisch nodules. Cafe-au-lait spots also had significantly higher latent corelations among 1st degree relatives than among 2nd degree relatives (Figure 7.2) - consistent with the influence of modifying genes. In addition, cafe-au-lait spots were more strongly associated between sibs than parents and children (Figure 7.3). Both sib pairs and parent-child pairs are 1st degree relatives who would be expected to share a similar proportion of non-alelic modifying genes. The diferences I observed in latent corelations between sib pairs and parent-child pairs are unlikely to result from efects of modifying genes, unless dominant alelic variants are common at these modifying loci. If dominant alelic variants are common at these loci, then the sib-sib association is expected to be stronger than the parent-child association. Afected sibs would be expected to share the same 183 normal NF] alele by descent half of the time, but parent-child pairs rarely would. Efects of functional polymorphisms ofthe normal NF1 alele might explain a higher latent corelation of these features among sib pairs than among parent-child pairs, but no direct evidence is available on this possibilty, and the frequency of functional polymorphisms of the NF1 locus is unknown. Interestingly, sharing of the normal NF1 alele among sibs may also contribute to the diference between 1st and 2nd degree relatives observed for cafe-au-lait spots (Figure 7.2), since 1st degree relatives include sibs. The diference in association between sibs and parent-child pairs (Figure 7.3) is also consistent with the involvement of environmental factors shared by sibs. However, cafe-au-lait spots have a very early onset suggesting a limited role for environmental factors. Furthermore, no environmental trigger has been proposed for these lesions. Although I minimised the efect of age it could stil  contribute to stronger associations among sibs, who tend to be of similar age, than between parents and children, who difer greatly in age. However, the simulations shown in the appendix suggest that age was adequately controled. Easton et al. (1993) found a higher corelation for cafe-au-lait spots between MZ twins than between sibs, suggesting the efect of a genetic locus or loci in additon to NF]. My findings of a strong latent corelation for cafe-au-lait spots in 1st degree relatives but no latent corelation among 2nd degree relatives are consistent with this interpretation. Lisch nodules are melanocytic hamartomas that arise in iris tissue (Pery and Font 1982). Cafe-au-lait spots are pigmented macules composed of melanocytes with abnormaly large pigment particles (Fitzpatrick 1981). Lisch nodules and cafe-au-lait spots share an origin from neural crest-derived tissue, but this is also true of some other 1 8 4 lesions characteristic of NF1, including neurofibromas of al types and intertriginous freckling (Bolande 1981). I previously reported an association between the occurence of Lisch nodules and cafe-au-lait spots in individual NF1 patients (see Chapter 2), but intertriginous freckling was also associated - a feature that shows no indication of a stronger familial latent corelation among 1st degree than 2nd degree relatives (Figure 7.2). If the development of Lisch nodules and cafe-au-lait spots is influenced by modifying genes, it is unclear what the nature of these modifying factors is or whether they are the same or diferent for these two features. Associations between these features (Chapter 4) suggest that they share steps in pathogenesis. This hypothesis could be examined further by determining if the two features are transmited together more often than expect by chance alone. Intertriginous freckling, subcutaneous neurofibromas, plexiform neurofibromas had similar latent corelations in 1st and 2nd degree relatives (Figure 7.2) but higher latent corelations between sibs than between parents and children (Figure 7.3). As stated above, both sib pairs and parent-child pairs share a similar proportion of non-alelic modifying genes, so the diferences I observed in these latent corelations are unlikely to result from efects of modifying genes in the absence of dominance. Easton et al. (1993) found that concordance for dermal discrete neurofibromas (which include subcutaneous neurofibromas) between monozygotic twins was much higher than between sibs, an observation that suggests the involvement of a genetic factor. As stated above, afected sibs would be expected to share the same normal NF1 alele half of the time, but parent-child pairs rarely would. Efects of functional polymorphisms of the normal NF1 alele might explain a higher latent corelation of these features among sib pairs than among 185 parent-child pairs, but no direct evidence is available on this possibilty, and the frequency of functional polymorphisms ofthe NF1 locus is unknown. Another possible explanation is diferences in environmental factors that are more likely to be shared among sibs than between a parent and child. Intertriginous freckling occurs in skin folds, and local environmental factors may play a role in the development of such freckling (Riccardi 1992). There is anecdotal evidence that subcutaneous neurofibromas may also develop as a result of trauma (Riccardi 1990). This hypothesis has not been tested formaly, and it seems unlikely to account for the development of congenital difuse plexiform neurofibromas. In any case, it is unclear why factors like cutaneous trauma would be more similar in sibs than in parent-child pairs. Intertriginous freckling, subcutaneous neurofibromas, plexiform neurofibromas and cafe-au-lait spots al share an origin from neural crest-derived cels. I found that cafe-au-lait spots and intertriginous freckling tended to occur together in individual NF 1 patients, and so did cutaneous, subcutaneous, and plexiform neurofibromas, but associations were not seen between the features in these two groups (see Chapter 4). In the present study, I did not find a stronger latent corelation for cutaneous neurofibromas in sibs than in parent-child pairs, as I did for subcutaneous and plexiform neurofibromas (Figure 7.3). Lisch nodules, subcutaneous neurofibromas, and cutaneous neurofibromas had higher latent corelations between afected fathers and children than between afected mothers and children (Figure 7.4). My sample included twice as many mother-child pairs as father-child pairs, so I was concerned about ascertainment bias. Perhaps mothers are more likely to bring children to the atention of an NF1 clinic that contributed data to the 186 NNFF International Database, whereas only severely afected father-child pairs tend to be seen in the NF clinics. However, the frequencies of al features studied were similar in afected fathers as in afected mothers and in their afected children (Table 7.1). Shared environment is unlikely to be the sole cause of associations between parents and children, due to large diferences in age. It is also unlikely that shared environment is responsible for the diference in latent corelations between mother-child and father-child pairs. Likewise, a multifactorial influence with a more extreme threshold for males than for females cannot explain the observations for these features. Gender is not a significant predictive factor in any of my models (Table 7.2), and feature frequencies among afected daughters of afected fathers are similar to those among afected sons of afected mothers. Parent-of-origin efects on severity of NF1 have been suggested (Miler and Hal 1978; Hal 1981), but most studies do not support this possibilty (Riccardi and Wald 1987; Huson et al. 1989a). One study found a male predominance among NF1 patients with pseudarthrosis but no significant parent-of-origin efect (Stevenson et al. 1999). Male-to-male inheritance is unlikely since gender is not a significant factor in any of my models (Table 7.2). Father-son concordances for Lisch nodules, subcutaneous neurofibromas, cutaneous neurofibromas are similar to father-daughter concordances, which argues against a Y-linked factor. My findings are consistent with a parent-of-origin efect on the strength of the parent-child association rather than with a more severe phenotype in afected ofspring of parents of one gender when compared to afected ofspring of parents of the other gender. Similar parent-child aggregation paterns have been reported for body mass index (Friedlander et al. 1988) and blood pressure (Hurwich et al. 1982) in the general population, but they are unprecedented in NF 1. Such a patern 187 could be caused by a triplet repeat that can expand when transmited through mothers but not fathers. I do not know of any simple genetic mechanism that can explain this phenomenon in NF 1 or for body mass index or blood pressure in the general population. Head circumference and stature had similar corelations for al relationships. This suggests that the mutant NF1 alele itself is most important in determining these corelations. Easton et al. (1993) also found evidence of the importance of the mutant alele in head circumference. The distributions of head circumference and stature in NF1 patients are unimodal (see Chapter 3), but both NF1 distributions are shifted relative to unafected norms, suggesting that head circumference and stature are afected to a degree in al NF1 patients. Taken together with the results of the present study, it appears that the magnitude of this efect depends, at least partly, on the mutant NF1 alele. The mutant NF1 genotype also has a very strong efect on the phenotypic manifestations in patients with Watson syndrome (Alanson et al. 1991) or deletions of the whole NF1 locus (Tonsgard et al. 1997; Dorschner et al. 2000). 7.7 Conclusion The paterns of familial associations shown here suggest that genetic factors may be involved in determining the occurence of various clinical features of NF1 and these factors may vary depending on the feature. In some instances, the mutant NF1 alele may be most important. In other instances, the efects of the normal NF1 alele or of unlinked modifying genes may predominate. More than one genetic factor may be involved, and the relative importance of various genetic and non-genetic efects may vary for diferent features. 188 G E N E R A L D I S C U S S I O N 8.1 Summary NF1 is a complex human disease with highly variable expressivity. NF1 is a large gene whose functions are only beginning to be understood. It interacts in complex biochemical pathways involving Ras and protein kinase A (PKA), but the phenotypic consequences of NFJ mutations exceed what can be atributed to Ras or PKA. The study of other complex human diseases has benefited tremendously from model systems or from in vitro biochemical experiments. However, humans are the only animal known to have a disease caused by heterozygosity for mutation at the NF1 locus. Suitable animal models have been developed for a smal number of NF1 disease features using chimeric double knockouts or double heterozygotes with specifc mutations at other loci. These models are adequate for studying the occurence of a few disease features, but there is curently no animal model that coresponds to most disease features found in humans. In order to improve our understanding of the causes of variable expressivity in NF1,1 used statistical methods to analyse the largest colections of NF1 clinical information in the world. In an efort to simplify the NF1 phenotype, I treated each of the features as if it were a trait segregating in NF1 patients. I have described how these "individual traits" relate to each other and, taking these relationships into account, how they aggregate among relatives with NF1. 8.1.1 Pair-wise analyses in individual NF1 patients Replicable associations among common NF1 disease features in three independent databases (Chapters 2) strongly suggest that these features do not occur 190 entirely at random in NF1 and that some patients are more likely than others to develop particular features. This interpretation contrasts with the generaly held view that any NF1 patient may develop any manifestation of the disease (Bernhart and Halperin 1990; Riccardi 1992). Most of the associations I observed have never been reported before, and their identification encouraged me to look for familal paterns of phenotypic aggregation. A few of the individual clinical features were investigated further. Short stature and macrocephaly are wel-recognised clinical features of NFL The results in Chapter 3 suggest that these changes in growth afect al NF1 patients and are not limited to particular subgroups. These results suggest that stature and head circumference should be treated as continuous variables in future analyses. The mechanisms by which mutations ofthe NF1 gene produce these phenotypic efects are unknown. Understanding such mechanisms may provide an important clue to the pathogenesis of more serious manifestations of NF1. NF1 mutations in Drosophila result in a smaler phenotype through the protein kinase A pathway (Guo et al. 1997; The et al. 1997) and human neurofibromin also has protein kinase A binding sites (Marchuk et al. 1991; Fahsold et al. 2000), but their role, if any, in growth is unknown. My findings in Chapter 6 suggest that the occurence of cutaneous neurofibromas in NF1 patients is not influenced by the local presence of difuse plexiform neurofibromas. In fact, al three of the lesions studied in this chapter (cafe-au-lait spots, cutaneous neurofibromas, and plexiform neurofibromas) occured independently of each another in almost al of the body segments analysed (Table 6.2). 191 8.1.2 Analyses of several diferent features at once The analyses in Chapters 2 and 6 considered only two features at a time in individuals. I examined the relationship among several diferent features at the same time and summarised the results in Chapters 4 and 5. In these analyses, I observed several associations among clinical features. It is not easy to pool the results of the 13 separate models I generated into a single conceptual framework of interdependencies of clinical features in NF1. Nevertheless, the associations I observed are consistent with the existence of three groups of features among the 13 features studied (Figure 4.14) — pigmentary changes, neurofibromas and CNS findings. Al features belonging to a group must be taken into account to best describe the occurence of other features belonging to the same group. Subjects who have one or more of a group's features may be fundamentaly diferent from those who do not have any of a group's features. Each group may share particular pathogenic elements that difer between groups. For instance, patients who tend to have pigmentary changes may have melanocytes with a particular neurofibromin activity profile - a diferent profile than those who tend not to have pigmentary changes. The melanocytes in the afected group might be more susceptible to limited proliferation in response to heat or friction. Alternatively, the efect could be on splicing of a particular isoform of neurofibromin or on the level of neurofibromin expression. Analogous diferences may exist in the Schwann cels of patients who tend to have neurofibromas or in the glial cels of those with CNS features. Each of these efects could be due to several diferent factors at the NF1 locus or at unlinked loci. Subsequent analyses atempted to diferentiate between various genetic causes and are discussed below. 192 The mechanisms shared by associated features may be diferent and independent for each group of features. This is supported by the observation that no feature, except Lisch nodules, was found to belong to more than one group. In terms of pathogenesis, this means that a single factor or set of factors could account for each of the three groups of features, but one factor is unlikely to account for al three groups of features. These NF1 features are not mutualy exclusive, and many patients belong to more than one group. Logistic regressive models in individuals were also used to investigate an NF1 clinical feature of special interest - unidentifed bright objects on MRI (UBOs). My observations in Chapter 5 suggest that UBOs and other NF1 clinical features do not occur independently and may be pathogeneticaly related. The common thread between the associated features (optic gliomas, other neoplasms, Lisch nodules, and subcutaneous neurofibromas) may be dysregulated celular proliferation resulting from haploinsuficiency of neurofibromin (Gutmann et al. 1999; Ingram et al. 2000). Analyses of individual patients (Chapters 2,4-6) suggest that shared pathogenetic mechanisms underlie several common features of NF1. If genetic factors influence these pathogenetic mechanisms, one would expect familal aggregation of such features to occur. 8.1.3 Familal analyses Preliminary familial analyses in Chapter 2 suggest that some clinical features tend to aggregate in afected parents and children. The age-adjusted number of afected body segments for al three of the lesions studied in Chapter 6 (cafe-au-lait spots, cutaneous 193 neurofibromas, and plexiform neurofibromas) was corelated among relatives with NFL These significant intrafamilal associations are consistent with other evidence that familal factors contribute to the development of cutaneous neurofibromas and cafe-au-lait spots in patients with NF1 (Easton et al. 1993). Taking local and familial analyses into consideration, multiple factors appear to be involved in the pathogenesis of both plexiform and cutaneous neurofibromas as wel as of cafe-au-lait spots. Some of these factors may to be genetic, but others are not. The intrafamilal corelations I observed were statisticaly significant, but they were also moderate in magnitude. Even the twins reported by Easton et al. (1993) were not perfectly concordant for plexiform neurofibromas and cafe-au-lait spots. Non-familal and non-genetic factors must also influence the occurence of these lesions. The analyses reported in Chapters 2 and 4 suggest that individual features should not be analysed independently, but that other features, as wel as age, must be taken into account. With this in mind, familal analyses were extended to several diferent types of familal relationship, and the results were used to evaluate the diferent familial mechanisms that could be contributing to NF1 variable expressivity. The paterns of familial associations summarised in Chapter 7 suggest that genetic factors involved in determining the occurence of diferent clinical features of NF1 vary. This conclusion is based on the genetic similarities and diferences between diferent classes of afected relatives. Head circumference and stature had similar corelations among al classes of afected relatives and may be influenced by the type of mutation at the NF1 locus. Lisch nodules and cafe-au-lait spots had higher latent corelations among 1st degree relatives than among 2nd degree relatives and so may be influenced by other 194 unlinked loci. Intertriginous freckling, subcutaneous neurofibromas, plexiform neurofibromas and cafe-au-lait spots had higher latent corelations among sibs than parents and children and so may be influenced by: variations in the normal NF1 alele, dominant aleles at modifying loci or shared environment. These interpretations are coroborated by the results from an equivalent analysis of simulated data (Appendix). More than one genetic factor may be involved, and the relative importance of various genetic and non-genetic efects may vary for diferent features. 8.2 How this study Jits into current NF1 research NF1 research today is proceeding in two opposite but complementary directions. On the one hand, researchers are trying to investigate the natural history and pathogenesis of individual features. The aim of such studies is to develop treatment strategies for individual features and their complications. On the other hand, molecular biologists are trying to understand how the NF1 gene and its product, neurofibromin, work. Presumably neurofibromin interacts with Ras via its GRD and with PKA via its CSRD. a newly discovered functional domain. However, both of these pathways are complex and dynamic and the celular consequences of NF1 mutations are largely a mystery. Clinical and molecular approaches ultimately hope to address the same question - how do defects in NF1 gene result in the myriad phenotypes of neurofibromatosis 1 ? Indeed the NF1 phenotype may be too complex to examine al at once. Previous atempts to assign a single severity score to a patient have not borne much fruit (Carey et al. 1979; Riccardi 1992). Such scales are inevitably subjective and may depend too much on stochastic factors such as where a tumour is located on the body. My study suggests 195 that some clinical features tend to be found together in the same patients. In addition, and perhaps most importantly, age must be taken into account. Due to their progressive nature, many NF1 features are more prevalent in older age groups. A severity scale relative to other patients of similar age would be much more meaningful than one that ignores age. This study describes how various clinical features of NF1 are related and what genetic mechanisms could be involved in their development. This may provide clues to researchers on both clinical and molecular fronts. Clinical researchers can get a more quantiative idea of which patients are at higher risks for the development of specifc disease features. Molecular biologists can get a beter idea of what genetic mechanisms are likely to play a part in the pathogenesis of particular features. For instance, I found that plexiform neurofibromas are more common in patients with subcutaneous neurofibromas. Although these results should not be used to make predictions in individual patients, they may stimulate interest in the natural history of neurofibromas. They may also spur researchers to compare patients who have both plexiform and subcutaneous neurofibromas and contrast them to those who lack these features. 8.3 Future directions The results of this study provide clues to the pathogenesis of several common clinical features of NF1. In addition, the methods used in this study can be applied to other complex disorders, particularly those with notable intra-familal variability. One of the initial goals of this study was to generate age-specifc risks for the development of several NF1 clinical features, taking into account covariates such as gender and the 196 presence or absence of other associated clinical features. The cross-sectional nature of the available data precludes reliable risk estimates. Risk estimates can be gleaned only from longitudinal studies, but a long and costly folow-up study of a large cohort of NF1 patients has been difficult to justify until now. My observations of several associations among NF 1 features in individuals and among relatives go a long way in helping to justify long-term longitudinal studies of NF1. NF1 alelic heterogeneity is one hypothesis to explain the results presented here. Many NF1 patients have been genotyped at the NF1 locus. However, phenotypic information has not been colected in a consistent manner. A standard phenotyping protocol would alow the data to be pooled and enable testing for genotype-phenotype corelations. It would be important to emphasie primary clinical features rather than their consequences. Riccardi has long espoused the need for this distinction (Riccardi 1992). For example, the pathogenesis of scoliosis depends on whether or not it is caused by spinal neurofibromas. Such distinctions are not made in the curent databases, and the heterogeneity of this feature may help to explain why I found no associations involving scoliosis in individuals or familes. Clearly, features such as neoplasms, pseudarthrosis and scoliosis are of great concern to patients and their physicians, but other, often less threatening features, like Lisch nodules or cafe-au-lait spots, may also provide valuable insights into NF1 function and disease pathogenesis. They should not be neglected during phenotyping. The results presented here suggest that the majority of the features studied (and most of the cardinal clinical features) are familal and may have a genetic component. The paterns of variable expressivity are subtle, so data would be required on a large 197 number of patients. An objective measurement such as a count of these lesions (e.g. number of neurofibromas) would enable a more detailed analysis of familal segregation paterns and would require a smaler number of patients than binary variables of the type used in most of the studies reported here. Given the progressive nature of many NF1 disease features and the potentialy confounding efects of age on analysis, it is also essential that the data be representative of al age groups. These results show that much of the phenotypic variabilty in NF1 may be due to genetic factors unlinked to the NF1 locus. Neurofibromin interacts with Ras and possibly PKA. Segregation analysis is typicaly used to identify the number of other genes involved and their modes of inheritance. Segregation analysis for possible modifying genes was previously performed in a smal sample of NF1 patients by Easton et al. (1993), but the results were inconclusive. Segregation analysis with the data used in the present study may be problematic for several reasons. The data are binary and segregation analysis methods are not as robust for binary variables as they are for continuous or polytomous variables (Khoury et al. 1993). MostNFl clinical features are highly dependant on age and do not occur independently of one another. I have atempted to minimise these efects, but they may be harder to control in more complex models. With the completion of the Human Genome Project, genetic markers throughout the genome are available - one study reports 1.42 milion single nucleotide polymorphisms (SNPs), providing one SNP for every 1.9 kilobases. (Sachidanandam et al. 2001). Over 80% are polymorphic at frequencies above 10%). Although a genome scan for modifying loci is theoreticaly possible, it is curently impractical to genotype 198 NF1 patients at hundreds of thousands of markers distributed throughout the genome and test if the markers are associated with specific clinical features or groups of features. Mouse and Drosophila models have been helpful in the understanding of neurofibromin biochemical interactions and pathogenesis of features such as malignant peripheral nerve sheath tumours (MPNSTs) and astrocytomas. Astrocytomas occur in 15% of NF1 patients, and MPNSTs afect up to 10% (see section 1.5). Far more common are problems related to dermal discrete neurofibromas or bony abnormalites. Animal models of these and other common features would enable studies into their pathogeneses. NF1 is disorder wel known for its highly variable expressivity. The results of my study refine this perception: it is a disorder which also has subtle but definite associations among its features both in individuals and among afected relatives. An ardent search for putative genetic factors involved in variable expressivity of the NF1 phenotype assumes that such factors exist, but the evidence supporting this assumption has hitherto been scant. My study suggests that genetic factors are involved in phenotypic variabilty in NF1. Multivariate probit analyses has been used for perhaps the first time in a study of this type. I have applied them to both real and simulated data. These methods can also be applied to other human disorders. NF1 may afect the human species exclusively, but it is not the only complex human disorder that lacks understanding. It may be very useful to perform genetic epidemiological studies of complex human disorders in additon to molecular investigation. I hope my findings encourage other investigations of NF1 and other complex human disorders. Ultimately, the answers wil rely on molecular genetic experiments, and I am eager to see whether they wil confirm my conclusions. 1 9 9 R E F E R E N C E S Abeliovich D, Gelman-Kohan Z, Silverstein S, Lerer I, Chemke J, Merlin S, Zlotogora J (1995) Familal cafe au lait spots: a variant of neurofibromatosis type 1. J Med Genet 32:985-986 Airewele GE, Sigurdson AJ, Wiley KJ, Frieden BE, Caldarera LW, Riccardi VM, Lewis RA, Chintagumpala MM, Ater JL, Plon SE, Bondy ML (2001) Neoplasms in neurofibromatosis 1 are related to gender but not to family history of cancer. Genet Epidemiol 20:75-86. Alanson J, Upadhaya M, Watson G, Partington M, MacKenzie A, Lahey D, MacLeod H, Sarfarazi M, Broadhead W, Harper P, Huson S (1991) Watson syndrome: Is it a subtype of type 1 neurofibromatosis? J Med Genet 28:752-756 Amara SG, Jonas V, Rosenfeld MG, Ong ES, Evans RM (1982) Alternative RNA processing in calcitonin gene expression generates mRNAs encoding diferent polypeptide products. Nature 298:240-4. Andersen L, Fountain J, Gutmann D, Tarle S, Glover T, Dracopoli N, Houseman D, Colins F (1993) Mutations in the neurofibrmatosis 1 gene in sporadic malignant melanoma cel lines. Nature Genet 3:1993 Ars E, Kruyer H, Gaona A, Casquero P, Rosel J, Volpini V, Sera E, Lazaro C, Estavil X (1998) A clinical variant of neurofibromatosis type 1: Familal spinal neurofibromatosis with a frameshift mutation in the NF1 gene. Am J Hum Genet 62:834-841 Atit RP, Crowe MJ, Greenhalgh DG, Wenstrup RJ, Ratner N (1999) The Nfl tumor suppressor regulates mouse skin wound healing, fibroblast proliferation, and colagen deposited by fibroblasts. J Invest Dermatol 112:835-42. Atit RP, Mitchel K, Nguyen L, Warshawsky D, Ratner N (2000) The neurofibromatosis type 1 (Nfl) tumor suppressor is a modifer of carcinogen-induced pigmentation and papiloma formation in C57BL/6 mice. J Invest Dermatol 114:1093-100. Beaty TH, Yang P, Khoury MJ, Haris EL, Liang KY (1991) Using log-linear models to test for associations among congenital malformations. Am J Med Genet 39:299-306. Bernards A (1998) Evolutionary comparisons. In: Upadhyaya M, Cooper D (eds) Neurofibromatosis 1: from genotype to phenotype. Bios, Washington, DC, pp 175-190 Bernards A, Snijders AJ, Hannigan GE, Murthy AE, Gusela JF (1993) Mouse neurofibromatosis type 1 cDNA sequence reveals high degree of conservation of both coding and non-coding mRNA segments. Hum Mol Genet 2:645-50. 200 Bernhart B, Halperin J (1990) Genetic counseling for Neurofibromatosis. In: Rubenstein A, Korf B (eds) Neurofibromatosis : a handbook for patients, familes, and health-care professionals. Thieme Medical Publishers, New York, pp 201-8 Bognanno J, Edwards M, Lee T, Dunn D, Roos K, Klate E (1988) Cranial MR imaging in neurofibromatosis. AJR 151:381-388 Bolande R (1974) The neurocristopathies: A unifying concept of disease arising in neural crest development. Hum Pathol 5:409-429 Bolande R (1981) Neurofibromatosis - the quintessential neurocristopathy: Pathogenic concepts and relationships. Adv Neurol 29:67-75 Bolag G, Clapp DW, Shih S, Adler F, Zhang YY, Thompson P, Lange BJ, Freedman MH, McCormick F, Jacks T, Shannon K (1996) Loss of NF1 results in activation of the Ras signaling pathway and leads to aberant growth in haematopoietic cels. Nat Genet 12:144-8 Bolag G, McCormick F (1991) Diferential regulation of rasGAP and neurofibromatosis gene product activities. Nature 351:576-9 Bolag G, McCormick F, Clark R (1993) Characterization of ful-length neurofibromin: tubulin inhibits Ras GAP activity. Embo J 12:1923-7 Buchberg AM, Cleveland LS, Jenkins NA, Copeland NG (1990) Sequence homology shared by neurofibromatosis type-1 gene and IRA-1 and IRA-2 negative regulators of the RAS cyclic AMP pathway [see comments]. Nature 347:291-4 Burger P, Scheithauer B (1994) Tumors of the central nervous system Atlas of tumor pathology. Vol. Fascicle 10. Armed Forces Instiute of Pathology, Washington, DC Byard P, Roche A (1982) Secular trend for recumbent length and stature in the Fels longitudinal growth study. In: Borms J, Hauspie R, Sand A, Susanne C, Hebbelinck M (eds) Human growth and development. Plenum Press, New York, pp 209-214 Calvert PT, Edgar MA, Webb PJ (1989) Scoliosis in neurofibromatosis. The natural history with and without operation. J Bone Joint Surg Br 71:246-51 Cappione AJ, French BL, Skuse GR (1997) A potential role for NF1 mRNA editng in the pathogenesis of NF1 tumors. Am J Hum Genet 60:305-12 Carey J, Laub J, Hal B (1979) Penetrance and variabilty in neurofibromatosis: A genetic study of 60 familes. Birth Defects 15(5B):271-281 201 Carmi D, Shohat M, Metzker A, Dickerman Z (1999) Growth, puberty, and endocrine functions in patients with sporadic or familial neurofibromatosis type 1: a longitudinal study. Pediatrics 103:1257-62 Caronti B, Butareli FR, Giustini S, Calderaro C, Calandrielo L, Calvieri S, Paladini G (1998) Serum mitogenic activity on in vitro glial cels in Neurofibromatosis type 1. Brain Res 793:21-8 Cawthon R, Weiss R, Xu G Viskochil D, Culver M, Stevens J, Robertson M, Dunn D, Gesteland R, O'Connel P, White R (1990) A major segment of the neurofibromatosis type 1 gene: cDNA sequence, genomic structure, and point mutations. Cel 62:193-201 Cawthon RM, Andersen LB, Buchberg AM, Xu GF, O'Connel P, Viskochil D, Weiss RB, Walace MR, Marchuk DA, Culver M, et al. (1991) cDNA sequence and genomic structure of EV12B, a gene lying within an intron of the neurofibromatosis type 1 gene. Genomics 9:446-60 Cichowski K, Shih TS, Schmit E, Santiago S, Reily K, McLaughlin ME, Bronson RT, Jacks T (1999) Mouse models of tumor development in neurofibromatosis type 1. Science 286:2172-6 Clementi M, Barbujani G, Turola L, Tenconi R (1990) Neurofibromatosis-1: A maximum likelihood estimation of mutation rate. Hum Genet 84:116-118 Clementi M, Milani S, Mammi I, Boni S, Moncioti C, Tenconi R (1999) Neurofibromatosis type 1 growth charts [see comments]. Am J Med Genet 87:317-23 Cnossen M, de Goede-Bolder A, van den Broek K, Waasdorp C, Oranje A, Stroink H, Simonsz H, van den Ouweland A, Haley D, Niermeijer M (1998) A prospective 10 year folow up study of patients with neurofibromatosis type 1. Arch Dis Child 78:408-412 Cnossen M, Stam E, Cooiman L, Simonsz H, Stroink H, Oranje A, Haley D, de Goede-Bolder A, Niermeijer M, de Muinck Keizer-Schrama S (1997) Endocrinologic disorders and optic pathway gliomas in children with neurofibromatosis type 1. Pediatrics 100:667-670 Colman S, Wilams C, Walace M (1995) Benign neurofibromas in type 1 neurofibromatosis (NF1) show somatic deletion of the NF1 gene. Nat Genet 11:90-92 Cox JS, Southern MR (1995) Orthopaedic complications of childhood neurofibromatosis. Orthop Nurs 14:25-30; quiz 31 Crawford AH, Bagamery N (1986) Osseous manifestations of neurofibromatosis in childhood. J Pediatr Orthop 6:72-88 202 Crowe F, Schul W, Neel J (1956) A Clinical, Pathological, and Genetic Study of Multiple Neurofibromatosis. Charles C. Thomas, Springfield, Ilinois Crump T (1981) Translation of case reports in Ueber die multiplen Fibrome der Haut und ihre Beziehung zu den multiplen Neuromen by F. v. Recklinghausen. Adv Neurol 29:259-75 Curless R, Siatkowski M, Glaser J, Shatz N (1998) MRI diagnosis of NF-1 in children without cafe-au-lait skin lesions. Pediatr Neurol 18:269-271 Danglot G, Regnier V, Fauvet D, Vassal G, Kujas M, Bernheim A (1995) Neurofibromatosis 1 (NF1) mRNAs expressed in the central nervous system are diferentialy spliced in the 5' part of the gene. Hum Moi Genet 4:915-20 Daston MM, Scrable H, Nordlund M, Sturbaum AK, Nissen LM, Ratner N (1992) The protein product of the neurofibromatosis type 1 gene is expressed at highest abundance in neurons, Schwann cels, and oligodendrocytes. Neuron 8:415-28 Davies SJ, Hughes HE (1993) Imprinting in Albright's hereditary osteodystrophy. J Med Genet 30:101-3 DeBela K, Poskit K, Szudek J, Friedman JM (2000a) Use of unidentifed bright objects on MRI for diagnosis of neurofibromatosis 1 in children. Neurology 54:1646-51 DeBela K, Szudek J, Friedman JM (2000b) Use ofthe national instiutes of health criteria for diagnosis of neurofibromatosis 1 in children. Pediatrics 105:608-14 DeClue JE, Cohen BD, Lowy DR (1991) Identification and characterization of the neurofibromatosis type 1 protein product. Proc Natl Acad Sci U S A 88:9914-8 DiMario, FJ, Ramsby G (1998) Magnetic resonance imaging lesion analysis in neurofibromatosis type 1. Arch Neurol 55:500-505 DiPaolo DP, Zimmerman RA, Rorke LB, Zackai EH, Bilaniuk LT, Yachnis AT (1995) Neurofibromatosis type 1: pathologic substrate of high-signal-intensity foci in the brain. Radiology 195:721-4 Dormer A, Wels G, Eliasziw M (1989) On two approximations to the F-distribution: application to testing for intraclass corelation in family studies. Canadian Journal of Statistics 17:209-215 Dorschner MO, Sybert VP, Weaver M, Pletcher BA, Stephens K (2000) NF1 microdeletion breakpoints are clustered at flanking repetiive sequences. Hum Moi Genet 9:35-46 203 Dugof L, Sujansky E (1996) Neurofibromatosis type 1 and pregnancy. Am J Med Genet 66:7-10 Durani AA, Crawford AH, Chouhdry SN, Saifuddin A, Morley TR (2000) Modulation of spinal deformites in patients with neurofibromatosis type 1. Spine 25:69-75 Easton D, Ponder M, Huson S, Ponder B (1993) An analysis of variation in expression of neurofibromatosis (NF) type I (NFI): Evidence for modifying genes. Am J Hum Genet 53:305-313 Erlandson RA (1991) The enigmatic perineurial cel and its participation in tumors and in tumorlike entities. Ultrastruct Pathol 15:335-51 Everit B (1992) The analysis of contingency tables Monographs on statistics and applied probabilty ; 45. Chapman & Hal, London ; New York, pp vii, 164 Fahsold R, Hofmeyer S, Mischung C, Gile C, Ehlers C, Kucukceylan N, Abdel-Nour M, Gewies A, Peters H, Kaufmann D, Buske A, Tinschert S, Nurnberg P (2000) Minor lesion mutational spectrum of the entire NFI gene does not explain its high mutabilty but points to a functional domain upstream of the GAP-related domain. Am J Hum Genet 66:790-818 Feigenbaum L, Fujita K, Colins FS, Jay G (1996) Repression of the NFI gene by Tax may expain the development of neurofibromas in human T-lymphotropic virus type 1 transgenic mice. J Virol 70:3280-5 Fitzpatrick TB (1981) Melanin synthesis pathways in the pathogenesis of neurofibromatosis. Adv Neurol 29:209-11 Framingham (1994) Familal aggregation of lens opacites: the Framingham Eye Study and the Framingham Ofspring Eye Study. Am J Epidemiol 140:555-64 Friedlander Y, Kark JD, Kaufmann NA, Bery EM, Stein Y (1988) Familal aggregation of body mass index in ethnicaly diverse familes in Jerusalem. The Jerusalem Lipid Research Clinic. Int J Obes 12:237-47 Friedman J (1999) Vascular and endocrine abnormalites. In: Friedman J, Gutmann D, MacColin M, Riccardi V (eds) Neurofibromatosis : phenotype, natural history, and pathogenesis. Johns Hopkins University Press, Baltimore, pp 274-276 Friedman J, Birch P (1997a) An association between optic glioma and other tumours of the central nervous system in neurofibromatosis type 1. Neuropediatrics 28:131-132 Friedman J, Birch P (1997b) Type 1 neurofibromatosis: A descriptive analysis of the disorder in 1,728 patients. Am J Med Genet 70:138-143 204 Friedman J, Greene C, Birch P, and the NNFF International Database P (1993) National Neurofibromatosis Foundation International Database. Am J Med Genet 45:88-91 Friedman J, Gutmann D, MacColin M, Riccardi V (1999) Neurofibromatosis : phenotype, natural history, and pathogenesis. Johns Hopkins University Press, Baltimore, pp xiv, 381 Friedman J, Riccardi V (1999) Clinical and epidemiological features. In: Friedman J, Gutmann D, MacColin M, Riccardi V (eds) Neurofibromatosis : phenotype, natural history, and pathogenesis. Johns Hopkins University Press, Baltimore, pp 29-86 Friedman JM, Seywerd M, Birch P (1995) Analysis of clinical sub-groups inNFl. Am J Hum Genet 57:A89 Fuler L, Cox B, Gardner R (1989) Prevalence of von Recklinghausen neurofibromatosis in Dunedin, New Zealand. Neurofibromatosis 2:278-283 Funasaki H, Winter RB, Lonstein JB, Denis F (1994) Pathophysiology of spinal deformites in neurofibromatosis. An analysis of seventy-one patients who had curves associated with dystrophic changes. J Bone Joint Surg Am 76:692-700 Garty BZ, Laor A, Danon YL (1994) Neurofibromatosis type 1 in Israel: survey of young adults. J Med Genet 31:853-7 Gibson D, Harvey AJ, Everet V, Parmar MK (1994) Is double data entry necessary? The CHART trials. CHART Steering Commitee. Continuous, Hyperfractionated, Accelerated Radiotherapy [see comments]. Control Clin Trials 15:482-8 Giorno R, Lieber J, Claman HN (1989) Ultrastructural evidence for mast cel activation in a case of neurofibromatosis. Neurofibromatosis 2:35-41 Goldgar DE, Green P, Pary DM, Mulvihil JJ (1989) Multipoint linkage analysis in neurofibromatosis type I: an international colaboration. Am J Hum Genet 44:6-12 Goldstein S, Reynolds CR (1999) Handbook of neurodevelopmental and genetic disorders in children. Guilford Press, New York, pp xvi, 602 Golubic M, Tanaka K, Dobrowolski S, Wood D, Tsai MH, Marshal M, Tamanoi F, Stacey DW (1991) The GTPase stimulatory activities of the neurofibromatosis type 1 and the yeast IRA2 proteins are inhibited by arachidonic acid. Embo J 10:2897-903. Green JE, Baird AM, Hinrichs SH, Klintworth GK, Jay G (1992) Adrenal medulary tumors and iris proliferation in a transgenic mouse model of neurofibromatosis. Am J Pathol 140:1401-10 205 Gregory PE, Gutmann DH, Mitchel A, Park S, Boguski M, Jacks T, Wood DL, Jove R, Colins FS (1993) Neurofibromatosis type 1 gene product (neurofibromin) associates with microtubules. Somat Cel Moi Genet 19:265-74 Grifiths PD, Blaser S, Mukonoweshuro W, Armstrong D, Milo-Mason G, Cheung S (1999) Neurofibromatosis bright objects in children with neurofibromatosis type 1: a proliferative potential? Pediatrics 104:e49 Guha A, Lau N, Huvar I, Gutmann D, Provias J, Pawson T, Boss G (1996) Ras-GTP levels are elevated in human NFI peripheral nerve tumors. Oncogene 12:507-13 Guo HF, The I, Hannan F, Bernards A, Zhong Y (1997) Requirement of Drosophila NFI for activation of adenylyl cyclase by PACAP38-like neuropeptides. Science 276:795-8 Gutman DH, Andersen LB, Cole JL, Swaroop M, Colins FS (1993) An alternatively-spliced mRNA in the carboxy terminus of the neurofibromatosis type 1 (NFI) gene is expressed in muscle. Hum Moi Genet 2:989-92 Gutmann D (1999a) Abnormalites ofthe nervous system. In: Friedman J, Gutmann D, MacColin M, Riccardi V (eds) Neurofibromatosis: phenotype, natural history, and pathogenesis. Johns Hopkins University Press, Baltimore, pp 190-202 Gutmann D (1999b) Other malignancies. In: Friedman J, Gutmann D, MacColin M, Riccardi V (eds) Neurofibromatosis: phenotype, natural history, and pathogenesis. Johns Hopkins University Press, Baltimore, pp 231-249 Gutmann D, Aylsworth A, Carey J, Korf B, Marks J, Pyeritz R, Rubenstein A, Viskochil D (1997) The diagnostic evaluation and multidisciplinary management of neurofibromatosis 1 and neurofibromatosis 2. JAMA 278:51-57 Gutmann DH, Cole JL, Colins FS (1995) Expression of the neurofibromatosis type 1 (NFI) gene during mouse embryonic development. Prog Brain Res 105:327-35 Gutmann DH, Cole JL, Stone WJ, Ponder BA, Colins FS (1994) Loss of neurofibromin in adrenal gland tumors from patients with neurofibromatosis type I. Genes Chromosomes Cancer 10:55-8 Gutmann DH, Donahoe J, Brown T, James CD, Pery A (2000) Loss of neurofibromatosis 1 (NFI) gene expression in NFI-associated pilocytic astrocytomas. Neuropathol Appl Neurobiol 26:361-7 Gutmann DH, Giordano MJ, Mahadeo DK, Lau N, Silbergeld D, Guha A (1996) Increased neurofibromatosis 1 gene expression in astrocytic tumors: positve regulation byp21-ras. Oncogene 12:2121-7 206 Gutmann DH, Loehr A, Zhang Y, Kim J, Henkemeyer M, Cashen A (1999) Haploinsuficiency for the neurofibromatosis 1 (NFI) tumor suppressor results in increased astrocyte proliferation. Oncogene 18:4450-9 Habiby R, Silverman B, Listernick R, Charow J (1995) Precocious puberty in children with neurofibromatosis type 1. J Pediatr 126:364-367 Hal J (1981) Possible maternal and hormonal factors in neurofibromatosis. Adv Neurol 29:125-131 Hamil P, Drizd T, Johnson C, Reed R, Roche A (1977) NCHS growth curves for children birth-18 years, U. S. 1967-73 Vital and health statistics. Vol. Series 11, No. 165. U.S. Government Printing Ofice, Washington, pp 78-1650 Han JW, McCormick F, Macara IG (1991) Regulation of Ras-GAP and the neurofibromatosis-1 gene product by eicosanoids. Science 252:576-9 Harkin J, Reed R (1969) Tumors of the peripheral nervous system Atlas of Tumor Pathology. Armed Forces Instiute of Pathology, Washington, DC, pp 67-100 Hartigan P (1985) Computation of the dip statistic to test for unimodality. Applied Statistics 34:320-325 Hewet SJ, Choi DW, Gutmann DH (1995) Expression of the neurofibromatosis 1 (NFI) gene in reactive astrocytes in vitro. Neuroreport 6:1565-8 Hinrichs SH, Nerenberg M, Reynolds RK, Khoury G, Jay G (1987) A transgenic mouse model for human neurofibromatosis. Science 237:1340-3 Horbar JD, Leahy KA (1995) An assessment of data quality in the Vermont-Oxford Trials Network database. Control Clin Trials 16:51-61 Hosmer DW, Lemeshow S (1989) Applied logistic regression Wiley series in probabilty and mathematical statistics. Applied probabilty and statistics section. Wiley, New York, pp xiii, 307 Howel SJ, Wilton P, Lindberg A, Shalet SM (1998) Growth hormone replacement and the risk of malignancy in children with neurofibromatosis [see comments]. J Pediatr 133:201-5 Hughes R (1994) Neurological complication^  of neurofibromatosis 1. In: Huson S, Hughes R (eds) The Neurofibromatoses: A Pathogenic and Clinical Overview. Chapman & Hal, London, pp 204-232 207 Hurwich BJ, Rosner B, Nubani N, Kass EH, Lewiter Fl (1982) Familal aggregation of blood pressure in a highly inbred community, Abu Ghosh, Israel. Am J Epidemiol 115:646-56 Huson S (1994) Neurofibromatosis 1: A clinical and genetic overview. In: Huson S, Hughes R (eds) The Neurofibromatoses: A Pathogenic and Clinical Overview. Chapman & Hal, London, pp 160-203 Huson S, Compston D, Clark P, Harper P (1989a) A genetic study of von Recklinghausen neurofibromatosis in south east Wales: I. Prevalence, fitness, mutation rate, and efect of parental transmission on severity. J Med Genet 26:704-711 Huson S, Compston D, Harper P (1989b) A genetic study of von Recklinghausen neurofibromatosis in south east Wales: II. Guidelines for genetic counseling. J Med Genet 26:712-721 Huson S, Harper P, Compston D (1988) Von Recklinghausen neurofibromatosis: A clinical and population study in South-East Wales. Brain 111:1355-1381 Huson S, Hughes R (1994) The Neurofibromatoses: a pathogenetic and clinical overview. Chapman & Hal, New York Ingram DA, Yang FC, Travers JB, Wenning MJ, Hiat K, New S, Hood A, Shannon K, Wilams DA, Clapp DW (2000) Genetic and biochemical evidence that haploinsuficiency of the Nfl tumor suppressor gene modulates melanocyte and mast cel fates in vivo. J Exp Med 191:181-8 Izawa I, Tamaki N, Saya H (1996) Phosphorylation of neurofibromatosis type 1 gene product (neurofibromin) by cAMP-dependent protein kinase. FEBS Let 382:53-9 Jabs EW, Li X, Scot AF, Meyers G, Chen W, Eccles M, Mao JI, Charnas LR, Jackson CE, Jaye M (1994) Jackson-Weiss and Crouzon syndromes are alelic with mutations in fibroblast growth factor receptor 2. Nat Genet 8:275-9 Jacks T, Shih T, Schmit E, Bronson R, Bernards A, Weinberg R (1994) Tumour predispositon in mice heterozygous for targeted mutation of NF1. Nature Genet 7:353-361 Joe H (1997) Multivariate models and dependence concepts Monographs on statistics and applied probabilty ; 73. Chapman & Hal, London ; New York, pp xviii, 399 Joe H (2000) Programs for multivariate binary (logit/probit) models. (Available from ftp: /ftp. stat .ubc. ca/pub/hj oe/mbin/) Johnston MC, Vig KW, Ambrose LJ (1981) Neurocristopathy as a unifying concept: clinical corelations. Adv Neurol 29:97-104 208 Khoury MJ, Beaty TH, Cohen BH (1993) Fundamentals of genetic epidemiology Monographs in epidemiology and biostatistics ; v. 19. Oxford University Press, New York, pp vi, 383 Kim HA, DeClue JE, Ratner N (1997) cAMP-dependent protein kinase A is required for Schwann cel growth: interactions between the cAMP and neuregulin/tyrosine kinase pathways. J Neurosci Res 49:236-47 Kleinbaum D (1992) Logistic regression: a self-learning text. Spinger-Verlag, New York Kluwe L, Friedrich R, Mautner VF (1999) Loss of NF1 alele in Schwann cels but not in fibroblasts derived from an NF1-associated neurofibroma. Genes Chromosomes Cancer 24:283-5 Korf B (1999a) Neurofibromas and malignant tumors of the peripheral nerve sheath. In: Friedman J, Gutmann D, MacColin M, Riccardi V (eds) Neurofibromatosis : phenotype, natural history, and pathogenesis. Johns Hopkins University Press, Baltimore, pp 142-161 Korf B (1999b) NNFF International NF1 Genetic Analysis Consortium Mutation Summary Data. National Neurofibromatosis Foundation Korneev SA, Park JH, O'Shea M (1999) Neuronal expression of neural nitric oxide synthase (nNOS) protein is suppressed by an antisense RNA transcribed from an NOS pseudogene. J Neurosci 19:7711-20 Kuczmarski R, Ogden C, Grummer-Strawn L (2000) CDC growth charts: United States. htp:/www.cdc.gov/nchs/data/ad314.pdf Lau N, Feldkamp MM, Roncari L, Loehr AH, Shannon P, Gutmann DH, Guha A (2000) Loss of neurofibromin is associated with activation of RAS/MAPK and PI3-K/AKT signaling in a neurofibromatosis 1 astrocytoma. J Neuropathol Exp Neurol 59:759-67 Levine MA, Jap TS, Mauseth RS, Downs RW, Spiegel AM (1986) Activity of the stimulatory guanine nucleotide-binding protein is reduced in erythrocytes from patients with pseudohypoparathyroidism and pseudopseudohypoparathyroidism: biochemical, endocrine, and genetic analysis of Albright's hereditary osteodystrophy in six kindreds. J Clin Endocrinol Metab 62:497-502 Lewis RA, Gerson LP, Axelson KA, Riccardi VM, Whitford RP (1984) von Recklinghausen neurofibromatosis. II. Incidence of optic gliomata. Ophthalmology 91:929-35 Li Y, O'Connel P, Breidenbach HH, Cawthon R, Stevens J, Xu G, Neil S, Robertson M, White R, Viskochil D (1995) Genomic organization of the neurofibromatosis 1 gene (NF1). Genomics 25:9-18 209 Liang KY, Beaty TH (1991) Measuring familal aggregation by using odds-ratio regression models. Genet Epidemiol 8:361-70 Listernick R, Darling C, Greenwald M, Strauss L, Charow J (1995) Optic pathway tumors in children: The efect of neurofibromatosis type 1 on clinical manifestations and natural history. J Pediatr 127:718-722 Listernick R, Gutmann D (1999) Tumours of the optic pathway. In: Friedman J, Gutmann D, MacColin M, Riccardi V (eds) Neurofibromatosis : phenotype, natural history, and pathogenesis. Johns Hopkins University Press, Baltimore, pp 203-230 Loewen CJ, Moritz OL, Molday RS (2001) Molecular characterization of peripherin-2 and Rom-1 mutants responsible for digenic Retinitis pigmentosa. J Biol Chem 10:10 Lou H, Neugebauer KM, Gagel RF, Berget SM (1998) Regulation of alternative polyadenylation by Ul snRNPs and SRp20. Moi Cel Biol 18:4977-85 Lukacs A, Junk AK, Stefani FH, Kampik A, Schiren CG, Plewig G (1997) [Lisch nodules. Markers of neurofibromatosis 1 and immunohistochemical references for neuroectodermal diferentiation]. Hautarzt 48:38-41 Maher ER, Reik W (2000) Beckwith-Wiedemann syndrome: imprinting in clusters revisited. J Clin Invest 105:247-52 Mantani A, Wakasugi S, Yokota Y, Abe K, Ushio Y, Yamamura K (1994) A novel isoform of the neurofibromatosis type-1 mRNA and a switch of isoforms during murine cel diferentiation and proliferation. Gene 148:245-51 Mantel N, Haenszel W (1959) Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 22:719-748 Marchuk DA, Saulino AM, Tavakkol R, Swaroop M, Walace MR, Andersen LB, Mitchel AL, Gutmann DH, Boguski M, Colins FS (1991) cDNA cloning ofthe type 1 neurofibromatosis gene: complete sequence of the NFI gene product. Genomics 11:931-40 McGaughran JM, Haris Dl, Donnai D, Teare D, MacLeod R, Westerbeek R, Kingston H, Super M, Haris R, Evans DG (1999) A clinical study of type 1 neurofibromatosis in north west England. J Med Genet 36:197-203 Mehler MF (2000) Brain dystrophin, neurogenetics and mental retardation. Brain Res Brain Res Rev 32:277-307 Messiaen LM, Calens T, Mortier G, Beysen D, Vandenbroucke I, Van Roy N, Spelman F, Paepe AD (2000) Exhaustive mutation analysis of the NFI gene alows identification 210 of 95% of mutations and reveals a high frequency of unusual splicing defects. Hum Mutat 15:541-55 Metheny L, Cappione A, Skuse G (1995) Genetic and epigenetic mechanisms in the pathogenesis of neurofibromatosis type 1. J Neuropath Exp Neurol 54:753-760 Mickle JE, Cuting GR (2000) Genotype-phenotype relationships in cystic fibrosis. Med Clin North Am 84:597-607 Mikol DD, Gulcher JR, Stefansson K (1990) The oligodendrocyte-myelin glycoprotein belongs to a distinct family of proteins and contains the HNK-1 carbohydrate. J Cel Biol 110:471-9 Miler M, Hal J (1978) Possible maternal efect on severity of neurofibromatosis. Lancet 2:1071-1073 Miler R (1974) The jackknife - areview. Biometrika 61:1-15 Miyazaki M, Wahid S, Bai L, Namba M (1992) Efects of intracelular cyclic AMP and cyclic GMP levels on DNA synthesis of young-adult rat hepatocytes in primary culture. Exp Cel Res 200:404-9 Mot HR, Carpenter JW, Campbel SL (1997) Structural and functional analysis of a mutant Ras protein that is insensitve to nitric oxide activation. Biochemistry 36:3640-4 Muler U, Steinberger D, Kunze S (1997) Molecular genetics of craniosynostotic syndromes. Graefes Arch Clin Exp Ophthalmol 235:545-50 Mulvihil J (1994) Malignancy: Epidemiological^  associated cancers. In: Huson S, Hughes R (eds) The Neurofibromatoses: A Pathogenic and Clinical Overview. Chapman & Hal, London, pp 305-315 Nerenberg M, Hinrichs SH, Reynolds RK, Khoury G, Jay G (1987) The tat gene of human T-lymphotropic virus type 1 induces mesenchymal tumors in transgenic mice. Science 237:1324-9 NIH (1988) Neurofibromatosis: Conference statement. National Instiutes of Health Consensus Development Conference. Arch Neurol 45:575-8 Nimura M (1990) Neurofibromatosis in Japan. In: Ishibashi Y, Hori Y (eds) Tuberous Sclerosis and Neurofibromatosis: Epidemiology, Pathophysiology, Biology, and Management. Elsevier, Amsterdam, pp 23-31 Nissim-Rafinia M, Chiba-Falek O, Sharon G, Boss A, Kerem B (2000) Celular and viral splicing factors can modify the splicing patern of CFTR transcripts carying splicing mutations. Hum Moi Genet 9:1771-8. 211 Nogales E, Wolf SG, Downing KH (1998) Structure of the alpha beta tubulin dimer by electron crystalography. Nature 391:199-203 Nordlund M, Gu X, Shipley MT, Ratner N (1993) Neurofibromin is enriched in the endoplasmic reticulum of CNS neurons. J Neurosci 13:1588-600 Nordlund ML, Rizvi TA, Brannan CI, Ratner N (1995) Neurofibromin expression and astrogliosis in neurofibromatosis (type 1) brains. J Neuropathol Exp Neurol 54:588-600 North K (1993) Neurofibromatosis type 1: Review of the first 200 patients in an Australian clinic. J Child Neurol 8:395-402 North K (1999) Cognitve function and academic performance. In: Friedman J, Gutmann D, MacColin M, Riccardi V (eds) Neurofibromatosis: phenotype, natural history, and pathogenesis. Johns Hopkins University Press, Baltimore, pp 162-189 Odenheimer DJ, Sarnaik SA, Whiten CF, Rucknagel DL, Sing CF (1987) The relationship between fetal hemoglobin and disease severity in children with sickle cel anemia. Am J Med Genet 27:525-35 Ounsted M, Moar VA, Scot A (1985) Head circumference charts updated. Arch Dis Child 60:936-9 Pai EF, Krengel U, Petsko GA, Goody RS, Kabsch W, Witinghofer A (1990) Refined crystal structure of the triphosphate conformation of H-ras p21 at 1.35 A resolution: implications for the mechanism of GTP hydrolysis. Embo J 9:2351-9 Palehorse (1997) Burns: Identification, Types and Treatment. JRH Enterprises Peltonen J, Jaakkola S, Lebwohl M, Renval S, Risteli L, Virtanen I, Uito J (1988) Celular diferentiation and expression of matrix genes in type 1 neurofibromatosis. Lab Invest 59:760-71 Pepperkok R, Hotz-Wagenblat A, Konig N, Girod A, Bossemeyer D, Kinzel V (2000) Intracelular distribution of mammalin protein kinase A catalytic subunit altered by conserved Asn2 deamidation. J Cel Biol 148:715-26 Pery H, Font R (1982) Iris nodules in von Recklinghausen's neurofibromatosis: Electron microscopic confirmation of their melanocytic origin. Arch Ophthalmol 100:1635-1640 Phelan CM, Rebbeck TR, Weber BL, Devilee P, Rutledge MH, Lynch HT, Lenoir GM, Straton MR, Easton DF, Ponder BA, Cannon-Albright L, Larsson C, Goldgar DE, Narod SA (1996) Ovarian cancer risk in BRCA1 cariers is modifed by the HRAS1 variable number of tandem repeat (VNTR) locus. Nat Genet 12:309-11 212 Pietruschka G (1961) Zur Symptomatic der Neurofibromatosis multiplex nach von Recklinghausen im Bereich des Sehorgans. Med Bild 4:8-11 Piat OS, Thorington BD, Brambila DJ, Milner PF, Rosse WF, Vichinsky E, Kinney TR (1991) Pain in sickle cel disease. Rates and risk factors. N Engl J Med 325:11-6 Poyhonen M, Leisti E-L, Kytola S, Leisti J (1997a) Hereditary spinal neurofibromatosis: A rare form of NF1? J Med Genet 34:184-187 Poyhonen M, Niemela S, Herva R (1997b) Risk of malignancy and death in neurofibromatosis. Arch Pathol Lab Med 121:139-143 Pulst S, Riccardi V, Fain P, Korenberg J (1991) Familal spinal neurofibromatosis: Clinical and DNA linkage analysis. Neurology 41:1923-1927 Purkiss SB, Tredwel SJ, Sawatzky BJ (2001) The role of the pedicle in dystrophic scoliosis: clinical cases versus theory. (Unpublished results) Raible DW, McMoris FA (1990) Induction of oligodendrocyte diferentiation by activators of adenylate cyclase. J Neurosci Res 27:43-6 Raible DW, McMoris FA (1993) Oligodendrocyte diferentiation and progenitor cel proliferation are independently regulated by cyclic AMP. J Neurosci Res 34:287-94 Rasmussen SA, Friedman JM (2000) NF 1 gene and neurofibromatosis 1. Am J Epidemiol 151:33-40 Rasmussen SA, Yang Q, Friedman JM (2001) Mortality in Neurofibromatosis 1: An Analysis Using U.S. Death Certificates. Am J Hum Genet 68:5 Rave-Harel N, Kerem E, Nissim-Rafinia M, Madjar I, Goshen R, Augarten A, Rahat A, Hurwitz A, Darvasi A, Kerem B (1997) The molecular basis of partial penetrance of splicing mutations in cystic fibrosis. Am J Hum Genet 60:87-94 Reardon W, Winter RM, Rutland P, Puleyn LJ, Jones BM, Malcolm S (1994) Mutations in the fibroblast growth factor receptor 2 gene cause Crouzon syndrome. Nat Genet 8:98-103 Regnier V, Meddeb M, Lecointre G, Richard F, Duverger A, Nguyen VC, Dutrilaux B, Bernheim A, Danglot G (1997) Emergence and scatering of multiple neurofibromatosis (NFl)-related sequences during hominoid evolution suggest a process of pericentromeric interchromosomal transpositon. Hum Mol Genet 6:9-16 Reik W, Maher ER (1997) Imprinting in clusters: lessons from Beckwith-Wiedemann syndrome. Trends Genet 13:330-4 213 Reily KM, Loisel DA, Bronson RT, McLaughlin ME, Jacks T (2000) Nfl ;Trp53 mutant mice develop glioblastoma with evidence of strain-specific efects. Nat Genet 26:109-13 Reinsch C (1967) Smoothing by spline functions. Numerische Mathematik 10:177-183 Rey I, Taylor-Harris P, van Erp H, Hal A (1994) R-ras interacts with rasGAP, neurofibromin and c-raf but does not regulate cel growth or diferentiation. Oncogene 9:685-92 Reynolds-Haertle RA, McBride R (1992) Single vs. double data entry in CAST. Control Clin Trials 13:487-94 Riccardi V (1981) Von Recklinghausen neurofibromatosis. N Engl J Med 305:1617-1627 Riccardi V (1982a) The multiple forms of neurofibromatosis. Pediatr Rev 3:292-298 Riccardi V (1982b) Neurofibromatosis: Clinical heterogeneity. Cur Probl Cancer 7(2): 1-34 Riccardi V (1990) The potential role of trauma and mast cels in the pathogenesis of neurofibromas. In: Ishibashi Y, Hori Y (eds) Tuberous sclerosis and neurofibromatosis : epidemiology, pathophysiology, biology, and management. Elsevier, Amsterdam, pp 167-190 Riccardi V (1992) Neurofibromatosis: Phenotype, natural history, and pathogenesis. The Johns Hopkins University Press, Baltimore Riccardi V (1999) Skeletal system. In: Friedman J, Gutmann D, MacColin M, Riccardi V (eds) Neurofibromatosis : phenotype, natural history, and pathogenesis. Johns Hopkins University Press, Baltimore, pp 250-273 Riccardi V, Wald J (1987) Discounting an adverse maternal efect on severity of neurofibromatosis. Pediatrics 79:386-393 Riccardi VM, Kleiner B, Lubs ML (1979) Neurofibromatosis: variable expression is not intrinsic to the mutant gene. Birth Defects Orig Artie Ser 15:283-9 Riter J, Riccardi V (1985) Von Recklinghausen neurofibromatosis (NF-I): an argument for very high penetrance and a comparison of sporadic and inherited cases. Am J Hum Genet 37:A135 Rizvi TA, Akunuru S, de Courten-Myers G, Switzer RC, 3rd, Nordlund ML, Ratner N (1999) Region-specifc astrogliosis in brains of mice heterozygous for mutations in the neurofibromatosis type 1 (Nfl) tumor suppressor. Brain Res 816:111-23 214 Roche AF (1979) Secular trends in human growth, maturation, and development. Monogr Soc Res Child Dev 44:1-120 Rudicel S (1987) The orthopaedic manifestations of neurofibromatosis. Conn Med 51:221-2 Rutkowski JL, Wu K, Gutmann DH, Boyer PJ, Legius E (2000) Genetic and celular defects contributing to benign tumor formation in neurofibromatosis type 1. Hum Moi Genet 9:1059-66 S.A.G.E. (1997) Statistical analysis for genetic epidemiology, Release 3.0. (Available from htp:/darwin.cwru.edu/pub/sage.html) Sachidanandam R, Weissman D, Schmidt SC, Kakol JM, Stein LD, Marth G, Shery S, et al (2001) A map of human genome sequence variation containing 1.42 milion single nucleotide polymorphisms. Nature 409:928-33 Samuelsson B, Axelsson R (1981) Neurofibromatosis: A clinical and genetic study of 96 cases in Gothenburg, Sweden. Acta Dermatovenereolog Suppl 95:67-71 Saragoni L, Hernandez P, Maccioni RB (2000) Diferential association of tau with subsets of microtubules containing postranslationaly-modifed tubulin variants in neuroblastoma cels. Neurochem Res 25:59-70 Sartin EA, Doran SE, Riddel MG, Herera GA, Tennyson GS, D'Andrea G, Whitley RD, Colins FS (1994) Characterization of naturaly occuring cutaneous neurofibromatosis in Holstein catle. A disorder resembling neurofibromatosis type 1 in humans. Am J Pathol 145:1168-74 SAS (1996) SAS, Cary, NC. Version 6.12 Sawada S, Florel S, Purandare S, Ota M, Stephens K, Viskochil D (1996) Identification of NFI aleles of a dermal neurofibroma. Nat Genet 14:110-112 Schmale MC, Aman MR, Gil KA (1996) A retrovirus isolated from cel lines derived from neurofibromas in bicolor damselfish (Pomacentrus partitus). J Gen Virol 77:1181-7. Schmale MC, Hensley G, Udey LR (1983) Neurofibromatosis, von Recklinghausen's disease, multiple schwannomas, malignant schwannomas. Multiple schwannomas in the bicolor damselfish, Pomacentrus partius (pisces, pomacentridae). Am J Pathol 112:238-41 Schmale MC, Hensley GT, Udey LR (1986) Neurofibromatosis in the bicolor damselfish (Pomacentrus partius) as a model of von Recklinghausen neurofibromatosis. Ann N Y Acad Sci 486:386-402 215 Sergeyev A (1975) On the mutation rate of neurofibromatosis. Hum Gene 28:129-138 Serjeant GR (2001) The emerging understanding of sickle cel disease. Br J Haematol 112:3-18 Sera E, Puig S, Otero D, Gaona A, Kruyer H, Ars E, Estivil X, Lazaro C (1997) Confirmation of a double-hit model for the NFI gene in benign neurofibromas. Am J Hum Genet 61:512-9 Sera E, Rosenbaum T, Winner U, Aledo R, Ars E, Estivil X, Lenard HG, Lazaro C (2000) Schwann cels harbor the somatic NFI mutation in neurofibromas: evidence of two diferent Schwann cel subpopulations. Hum Moi Genet 9:3055-64 Sevick RJ, Barkovich AJ, Edwards MS, Koch T, Berg B, Lempert T (1992) Evolution of white mater lesions in neurofibromatosis type 1: MR findings. AJR Am J Roentgenol 159:171-5 Shannon KM, Waterson J, Johnson P, O'Connel P, Lange B, Shah N, Steinherz P, Kan YW, Priest JR (1992) Monosomy 7 myeloproliferative disease in children with neurofibromatosis, type 1: epidemiology and molecular analysis. Blood 79:1311-8 Skuse GR, Cappione AJ (1997) RNA processing and clinical variabilty in neurofibromatosis type I (NFI). Hum Moi Genet 6:1707-12 Skuse GR, Kosciolek BA, Rowley PT (1989) Molecular genetic analysis of tumors in von Recklinghausen neurofibromatosis: loss of heterozygosity for chromosome 17. Genes Chromosomes Cancer 1:36-41 Skuse GR, Kosciolek BA, Rowley PT (1991) The neurofibroma in von Recklinghausen neurofibromatosis has a unicelular origin. Am J Hum Genet 49:600-7 Sorensen S, Mulvihil J, Nielsen A (1986) Long-term folow-up of von Recklinghausen neurofibromatosis: Survival and malignant neoplasms. N Engl J Med 314:1010-1015 Spjotvol E (1967) Optimum invariant tests in unbalanced variance component models. Ann Math Statist 38: 422-428 SPSS (1998) SPSS for Windows, Chicago. Version 9.0.0 Stark M, Assum G, Krone W (1995) Single-cel PCR performed with neurofibroma Schwann cels reveals the presence of both aleles of the neurofibromatosis type 1 (NFI) gene. Hum Genet 96:619-23 Stevenson DA, Birch PH, Friedman JM, Viskochil DH, Balestrazzi P, Boni S, Buske A, Korf BR, Nimura M, Pivnick EK, Schory EK, Short MP, Tenconi R, Tonsgard JH, 216 Carey JC (1999) Descriptive analysis of tibial pseudarthrosis in patients with neurofibromatosis 1. Am J Med Genet 84:413-9 Stone JW, Bridwel KH, Shackelford GD, Abramson CL (1987) Dural ectasia associated with spontaneous dislocation of the upper part of the thoracic spine in neurofibromatosis. A case report and review of the literature. J Bone Joint Surg Am 69:1079-83 Szudek J, Birch P, Friedman JM (2000a) Growth charts for young children with neurofibromatosis 1 (NFI) [leter; comment]. Am J Med Genet 92:224-8 Szudek J, Birch P, Friedman JM (2000b) Growth in North American white children with neurofibromatosis 1 (NFI). J Med Genet 37:933-8 Szudek J, Birch P, Riccardi VM, Evans DG, Friedman JM (2000c) Associations of clinical features in neurofibromatosis 1 (NFI). Genet Epidemiol 19:429-39 The I, Hannigan GE, Cowley GS, Reginald S, Zhong Y, Gusela JF, Hariharan IK, Bernards A (1997) Rescue of a Drosophila NFI mutant phenotype by protein kinase A. Science 276:791-4 Thompson MW, Mclnnes RR, Wilard HF, Thompson JS (1991) Thompson & Thompson Genetics in medicine. W.B. Saunders Co., Philadelphia, pp xi, 500 Tonsgard J, Yalavarthi K, Cushner S, Short M, Lindgren V (1997) Do NFI gene deletions result in a characteristic phenotype? Am J Med Genet 73:80-86 Truhan A, Filpek P (1993) Magnetic resonance imaging: Its role in the neuroradiologic evaluation of neurofibromatosis, tuberous sclerosis, and Sturge-Weber syndrome. Arch Dermatol 129:219-226 Upadhyaya M, Osborn M, Maynard J, Harper P (1996) Characterization of six mutations in exon 37 of neurofibromatosis type 1 gene. Am J Med Genet 67:421-3. Viskochil D (1999) The structure and function of the NFI Gene: molecular pathophysiology. In: Friedman J, Gutmann D, MacColin M, Riccardi V (eds) Neurofibromatosis : phenotype, natural history, and pathogenesis. Johns Hopkins University Press, Baltimore, pp 119-141 Viskochil D, Buchberg A, Xu G, Cawthon R, Stevens J, Wolf R, Culver M, Carey J, Copeland N, Jenkins N, White R, O'Connel P (1990) Deletions and a translocation interupt a cloned gene at the neurofibromatosis type 1 locus. Cel 62:187-192 Vogel F, Motulsky AG (1997) Human genetics : problems and approaches. Springer, Berlin ; New York, pp xxxvi, 851 217 Vogel KS, Klesse LJ, Velasco-Miguel S, Meyers K, Rushing EJ, Parada LF (1999) Mouse tumor model for neurofibromatosis type 1. Science 286:2176-9 Walace M, Marchuk D, Andersen LB Letcher R, Odeh H, Saulino A, Fountain J, Brereton A, Nicholson J, Mitchel A, Brownstein B, Colins F (1990) Type 1 neurofibromatosis gene: Identification of a large transcript disrupted in three NF1 patients. Science 249:181-186 Webster AR, Richards FM, MacRonald FE, Moore AT, Maher ER (1998) An analysis of phenotypic variation in the familal cancer syndrome von Hippel-Lindau disease: evidence for modifer efects. Am J Hum Genet 63:1025-35 Weichert KA, Dine MS, Benton C, Silverman FN (1973) Macrocranium and neurofibromatosis. Radiology 107:163-6 Weinstein LS, Yu S (1999) The Role of Genomic Imprinting of G-alpha in the Pathogenesis of Albright Hereditary Osteodystrophy. Trends Endocrinol Metab 10:81-85. Weston JA (1981) The regulation of normal and abnormal neural crest cel development. Adv Neurol 29:77-95 Wilde PH, Upadhyay SS, Leong JC (1994) Deterioration of operative corection in dystrophic spinal neurofibromatosis. Spine 19:1264-70 Wilkie AO, Moriss-Kay GM, Jones EY, Heath JK (1995) Functions of fibroblast growth factors and their receptors. Cur Biol 5:500-7 Wilson AF, Sorant AJ (2000) Equivalence of single- and multiocus markers: power to detect linkage with composite markers derived from bialelic loci. Am J Hum Genet 66:1610-5 Winter RB, Moe JH, Bradford DS, Lonstein JE, Pedras CV, Weber AH (1979) Spine deformity in neurofibromatosis. A review of one hundred and two patients. J Bone Joint Surg Am 61:677-94 Wright J, Pickard N, Whitfield A, Hakin N (2000) A population-based study of the prevalence, clinical characteristics and efect of ethnicity in epilepsy. Seizure 9:309-13 Xu GF, O'Connel P, Viskochil D, Cawthon R, Robertson M, Culver M, Dunn D, Stevens J, Gesteland R, White R, et al. (1990) The neurofibromatosis type 1 gene encodes a protein related to GAP. Cel 62:599-608 Xu W, Muligan LM, Ponder MA, Liu L, Smith BA, Mathew CG, Ponder BA (1992) Loss of NF1 aleles in phaeochromocytomas from patients with type I neurofibromatosis. Genes Chromosomes Cancer 4:337-42 218 Zar JH (1999) Biostatistical analysis. Prentice Hal, Upper Saddle River, N.J., pp 1 v. (various pagings) Zehavi C, Romano A, Goodman R (1986) Iris (Lisch) nodules in neurofibromatosis. Clin Genet 29:51-55 Zoler M, Rembeck B, Akesson H, Angerval L (1995) Life expectancy, mortality and prognostic factors in neurofibromatosis type 1: A twelve-year folow-up of an epidemiological study in Goteborg, Sweden. Acta Derm Venerol (Stockh) 75:136-140 219 A P P E N D I X - A N A L Y S I S O F S I M U L A T E D D A T A W I T H M P R O B I T 220 Introduction The results of multivariate probit (MPROBIT) (Joe 1997; Joe 2000) analysis have been used in this study to support the involvement of genetic factors in the phenotypic variabilty of NF1. However, MPROBIT is newly revised for biostatistical analysis in this study; it has never before been used to analyse pedigree data. I wanted to assess the validity of this method. To that end, I have simulated a monogenic disease with five diferent features, each with diferent genetic component. I then used MPROBIT to analyse each feature and checked if the intra-familal latent corelation patern found for each feature was consistent with the genetic components used to simulate them. Subjects and Methods The Genometric Analysis Simulation Program (GASP) is a software tool that can generate samples of family data based on user specifed genetic models (Wilson and Sorant 2000). I used GASP to simulate a sample of pedigrees similar to the ones described in the National Neurofibromatosis Foundation International Database (NFDB). The simulated disease is caused by one 12-alele locus - 10 diferent dominant disease-causing aleles, each with a frequency of 0.01, and two diferent normal aleles, each with a frequency of 0.45. This is similar to NF1, where there are many rare mutant aleles and only a few common normal aleles are known. Five diferent features were generated simultaneously with the disease. (1) Random: a random feature was generated by finding the inverse of the standard normal cumulative distribution of random numbers between 0 and 1 inclusive. (2) Mutant: 60% of the variation in this feature is due to the mutant aleles at the disease locus, and 40% of the variation is due to random efects. (3) 221 Polygenic: 90% of the variation in this feature is due to a polygenic component, and 10% of the variation is due to random efects. The polygenic component for parents is generated using a random deviate from a normal distribution. The polygenic component for children is based on mid-parental value and a random deviate from a normal distribution. (4) Unlinked: 90% of the variation in this feature is due to a single locus unlinked to the disease locus, and 10% of the variation is due to random efects. The unlinked locus has two aleles with equal frequencies but opposite efects. One homozygous genotype increases susceptibilty, the other homozygous genotype decreases susceptibilty and the heterozygous genotype has no efect. (5) Wild: 90% of the variation in this feature is due to the normal aleles at the disease locus, with 10% of the variation due to random efects. Prevalence of most NFI features is typicaly greater than zero at birth, increases with age and is less than 100% late in life. Although these prevalence by age curves are typicaly non-linear, they can usualy be represented as linear if age is transformed using a logarithmic function. Each of the features was simulated as a continuous liability that was then compared to an age-specifc threshold (Figure A.l). The same threshold was used for al  five features. For convenience, this threshold for each age was the z-score coresponding to a probabilty of 1- e ' ! 0 / ( a g e + 4 ) . The final derived feature was defined as "present" if its normaly distributed genetic liability exceeded the age-specifc threshold. No parent of origin efect was included in these models. One sample of 4995 individuals in 555 familes was generated. Each simulated family has two parents, three ofspring, two aunts or uncles and two grandparents. NFI. is inherited as a dominant disease and mutant homozygotes are thought not to be viable. 222 Therefore only the 820 subjects in 244 familes with exactly one mutant alele were used in the subsequent analysis. These include 277 afected sib pairs, 255 afected mother-child pairs, 271 afected father-child pairs and 244 pairs of afected 2nd degree relatives. MPROBIT was used to estimate intra-familal latent corelation coeficients for each of the five features, while minimising the confounding efects of age (see section 7.4.4). Results Figure A.2 shows the adjusted intrafamilal latent corelation coeficients and their 95% confidence intervals for each of the five features among al 820 afected relatives with NF1 from the 244 familes studied. In these estimates, al relatives are treated the same regardless of relationship. Statisticaly significant positve intrafamilal latent corelations were observed for al of the features except the random feature. Figure A.3 shows the adjusted intrafamilal latent corelation coeficients and 95% confidence intervals for the five clinical features among afected 1st and 2nd degree relatives. Statisticaly significant positve latent corelations between 1st degree relatives were found for al of the features except the random feature. A significant positve latent corelation between 2nd degree relatives was also found for the feature influenced by the mutant alele. Significant negative latent corelations were not observed for any of the features. Latent corelations were significantly greater among 1st degree relatives than among 2nd degree relatives for features influenced by the normal alele (p=0.0008), a polygenic component (p=0.03), and a single unlinked locus (p=0.006). Latent corelations among 1st degree relatives were not statisticaly diferent from latent 223 corelations among 2na degree relatives for the random feature (p=0.47) or the feature influenced by the mutant alele (p=0.18). Figure A.4 shows the adjusted intrafamilal latent corelation coeficients and 95% confidence intervals for the five features among afected sib pairs and afected parent-child pairs. Statisticaly significant positve latent corelations between sibs were found for al but the random feature. Signifcant positve latent corelations between parents and children were found for three features. Signifcant negative latent corelations were not observed for any of the features. Latent corelations were significantly greater between sibs than between parents and children for the feature influenced by the normal alele (p<0.0001) Latent corelations between sibs were not statisticaly diferent from the latent corelations between parents and children for features influenced by an unlinked locus (p=0.22), a polygenic component (p=0.38), the mutant alele (p=0.50) or the random feature (p=0.44). Figure A.5 shows the adjusted intrafamilal latent corelation coeficients and 95%) confidence intervals for the five features between afected mother-child pairs and between afected father-child pairs. Statisticaly significant positve latent corelations between mothers and children were found for three of the features. Signifcant positve latent corelations between fathers and children were also found for the three features. Signifcant negative latent corelations were not observed for any of the features in either relationship. Latent corelations do not difer significantly between father-child pairs and mother-child pairs for any of the features: the feature influenced by the normal alele (p=0.17), an unlinked locus (p=0.94), a polygenic component (p=0.29), the mutant alele (p=0.92) or the random feature (p=0.37). 224 Conclusion Statisticaly significant intra-familal latent corelations were observed for the four features with genetic components, but not for the random feature. Higher latent corelations for 1st than 2nd degree relatives were observed for the features whose major genetic component was one or more modifying genes at unlinked loci, but not for any other features. A similar latent corelation for al relationships was observed for the feature with a normal alele component. This suggests that genetic components can result in significant intra-familal latent corelation coeficient estimates using MPROBIT. Furthermore, these latent corelation coeficient estimates are consistent with the simulated genetic components underlying the disease features. 225 Figure A . l : Prevalence by age of a simulated clinical feature. This feature was used to generate an age-specifc liability threshold. Prevalence of most NFI features is typicaly greater than zero at birth, increases with age in a non-linear manner and is less than 100% late in life. Each of the five features was initialy simulated as a continuous liability. Each of their continuous liabilities was compared to an age-specifc threshold. The final derived feature was defined as "present" if its normaly distributed genetic liability exceeded the age-specifc threshold. 226 o o co o o TT CD < O CO o CN I I I 1 1 1 1 [ — c n o q r ^ c n i o T j - c o c N d d d d d c i d c i 227 Figure A.2: Associations of simulated features among all affected relatives. Adjusted intrafamilal latent corelation coeficients and their 95% confidence intervals are shown for each of the five features among al 820 relatives with NF1 from the 244 familes simulated. In these estimates, al relatives are treated the same regardless of relationship. 228 iq o c o • o ° ir o o IT) CM O •a a> c "E D 10 d o 'E CO o Q. c CO E o T J C CO 01 229 Figure A.3: Associations of simulated features among 1st and 2 n d degree relatives. Adjusted intrafamilal latent corelation coeficients and 95% confidence intervals are shown for five clinical features among afected 1st degree and 2nd degree relatives. A star indicates a significant diference between the coeficient among 1st degree relatives and the coeficient among 2nd degree relatives. 230 1 231 Figure A.4: Associations of simulated features among sib and parent-child pairs. Adjusted intrafamilal latent corelation coeficients and 95% confidence intervals are shown for five features among afected sib pairs and among afected parent-child pairs. A star indicates a significant diference between the coeficient among sibs and the coeficient among parents and children. 232 233 Figure A.5: Associations of simulated features among mother- and father-child pairs. Adjusted intrafamilal latent corelation coeficients and 95% confidence intervals are shown for five features between afected mother-child pairs and between afected father-child pairs. 234 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0090779/manifest

Comment

Related Items