[Skip to Navigation]
Sign In
Table 1. 
Prevalence by Sex, Cohort, and Diagnostic Definition*
Prevalence by Sex, Cohort, and Diagnostic Definition*
Table 2. 
Sample Description of Female Participants Controlling for Age*
Sample Description of Female Participants Controlling for Age*
Table 3. 
1972-1973–Measured Variables Predicting All-Source AN in Female Participants of All Ages, Covarying Age at 1973
1972-1973–Measured Variables Predicting All-Source AN in Female Participants of All Ages, Covarying Age at 1973
Table 4. 
Descriptive Twin Data for Univariate Analyses of AN Definitions*
Descriptive Twin Data for Univariate Analyses of AN Definitions*
Table 5. 
Parameter Estimates of Univariate Analyses of Anorexia Definitions in Female Participants*
Parameter Estimates of Univariate Analyses of Anorexia Definitions in Female Participants*
1.
Sullivan  PF Mortality in anorexia nervosa.  Am J Psychiatry 1995;1521073- 1074PubMedGoogle Scholar
2.
Harris  ECBarraclough  B Excess mortality of mental disorder.  Br J Psychiatry 1998;17311- 53PubMedGoogle ScholarCrossref
3.
Hoek  Hvan Hoeken  D Review of the prevalence and incidence of eating disorders.  Int J Eat Disord 2003;34383- 396PubMedGoogle ScholarCrossref
4.
McKnight Investigators, Risk factors for the onset of eating disorders in adolescent girls: results of the McKnight Longitudinal Risk Factor Study.  Am J Psychiatry 2003;160248- 254PubMedGoogle ScholarCrossref
5.
Wittchen  HUNelson  CBLachner  G Prevalence of mental disorders and psychosocial impairments in adolescents and young adults.  Psychol Med 1998;28109- 126PubMedGoogle ScholarCrossref
6.
Eagles  JJohnston  MHunter  DLobban  MMillar  H Increasing incidence of anorexia nervosa in the female population of northeast Scotland.  Am J Psychiatry 1995;1521266- 1271PubMedGoogle Scholar
7.
Jones  DFox  MBabigan  HHutton  H Epidemiology of anorexia nervosa in Monroe County, New York: 1960-76.  Psychosom Med 1980;42551- 558PubMedGoogle ScholarCrossref
8.
Lucas  ARBeard  CMO’Fallon  WMKurland  LT Anorexia nervosa in Rochester, Minnesota: a 45-year study.  Mayo Clin Proc 1988;63433- 442PubMedGoogle ScholarCrossref
9.
Møller-Madsen  SNystrup  J Incidence of anorexia nervosa in Denmark.  Acta Psychiatr Scand 1992;86197- 200PubMedGoogle ScholarCrossref
10.
Szmukler  G The epidemiology of anorexia nervosa and bulimia.  J Psychiatr Res 1985;19143- 153PubMedGoogle ScholarCrossref
11.
Willi  JGrossman  S Epidemiology of anorexia nervosa in a defined region of Switzerland.  Am J Psychiatry 1983;140564- 567PubMedGoogle Scholar
12.
Milos  GSpindler  ASchnyder  UMartz  JHoek  HWilli  J Incidence of severe anorexia nervosa in Switzerland: 40 years of development.  Int J Eat Disord 2004;36118- 119PubMedGoogle ScholarCrossref
13.
Hall  AHay  P Eating disorder patient referrals from a population region 1977-1986.  Psychol Med 1991;21697- 701PubMedGoogle ScholarCrossref
14.
Hoek  HBartelds  ABosveld  Jvan der Graaf  YLimpens  VEMaiwald  MSpaaij  CJ Impact of urbanization on detection rates of eating disorders.  Am J Psychiatry 1995;1521272- 1278PubMedGoogle Scholar
15.
Joergensen  J The epidemiology of eating disorders in Fyn County, Denmark, 1977-1986.  Acta Psychiatr Scand 1992;8530- 34PubMedGoogle ScholarCrossref
16.
Nielsen  S The epidemiology of anorexia nervosa in Denmark from 1973-1987: a nationwide register study of psychiatric admission.  Acta Psychiatr Scand 1990;81507- 514PubMedGoogle ScholarCrossref
17.
Willi  JGiacometti  GLimacher  B Update on the epidemiology of anorexia nervosa in a defined region of Switzerland.  Am J Psychiatry 1990;1471514- 1517PubMedGoogle Scholar
18.
Lucas  ARBeard  CMO’Fallon  WMKurland  LT 50-Year trends in the incidence of anorexia nervosa in Rochester, Minn: a population-based study.  Am J Psychiatry 1991;148917- 922PubMedGoogle Scholar
19.
Lilenfeld  LKaye  WGreeno  CMerikangas  KRPlotnicov  KPollice  CRao  RStrober  MBulik  CMNagy  L A controlled family study of restricting anorexia and bulimia nervosa: comorbidity in probands and disorders in first-degree relatives.  Arch Gen Psychiatry 1998;55603- 610PubMedGoogle ScholarCrossref
20.
Strober  MFreeman  RLampert  CDiamond  JKaye  W Controlled family study of anorexia nervosa and bulimia nervosa: evidence of shared liability and transmission of partial syndromes.  Am J Psychiatry 2000;157393- 401PubMedGoogle ScholarCrossref
21.
Strober  MFreeman  RLampert  CDiamond  JKaye  W Males with anorexia nervosa: a controlled study of eating disorders in first-degree relatives.  Int J Eat Disord 2001;29263- 269PubMedGoogle ScholarCrossref
22.
Wade  TDBulik  CMNeale  MKendler  KS Anorexia nervosa and major depression: shared genetic and environmental risk factors.  Am J Psychiatry 2000;157469- 471PubMedGoogle ScholarCrossref
23.
Kortegaard  LSHoerder  KJoergensen  JGillberg  CKyvik  KO A preliminary population-based twin study of self-reported eating disorder.  Psychol Med 2001;31361- 365PubMedGoogle ScholarCrossref
24.
Klump  KLMiller  KBKeel  PKMcGue  MIacono  WG Genetic and environmental influences on anorexia nervosa syndromes in a population-based twin sample.  Psychol Med 2001;31737- 740PubMedGoogle ScholarCrossref
25.
Stice  EPresnell  KBearman  S Relation of early menarche to depression, eating disorders, substance abuse, and comorbid psychopathology among adolescent girls.  Dev Psychol 2001;37608- 619PubMedGoogle ScholarCrossref
26.
Fairburn  CGCooper  ZDoll  HWelch  SL Risk factors for anorexia nervosa.  Arch Gen Psychiatry 1999;56468- 476PubMedGoogle ScholarCrossref
27.
Jacobi  CHayward  Cde Zwaan  MKraemer  HAgras  W Coming to terms with risk factors for eating disorders: application of risk terminology and suggestions for a general taxonomy.  Psychol Bull 2004;13019- 65PubMedGoogle ScholarCrossref
28.
Walsh  BBulik  CFairburn  CHalmi  KHerzog  DGolden  NEA Defining eating disorders. In:Evans  DFoa  EGur  R  et al. eds. Treating and Preventing Adolescent Mental Health Disorders: What We Know and What We Don't Know New York, NY Oxford University Press, Annenberg Foundation Trust at Sunnylands, and Annenberg Public Policy Center of the University of Pennsylvania2005;Google Scholar
29.
Cnattingius  SHultman  CDahl  MSparen  P Very preterm birth, birth trauma, and the risk of anorexia nervosa among girls.  Arch Gen Psychiatry 1999;56634- 638PubMedGoogle ScholarCrossref
30.
Lichtenstein  PDe Faire  UFloderus  BSvartengren  MSvedberg  PPedersen  N The Swedish Twin Registry: a unique resource for clinical, epidemiological and genetic studies.  J Intern Med 2002;252184- 205PubMedGoogle ScholarCrossref
31.
Pedersen  NLichtenstein  PSvedberg  P The Swedish Twin Registry in the third millennium.  Twin Res 2002;5427- 432PubMedGoogle ScholarCrossref
32.
First  MSpitzer  RGibbon  MWilliams  J Structured Clinical Interview for DSM-IV Axis I Disorders, Research Version, Patient Edition.  New York Biometrics Research, New York State Psychiatric Institute1997;
33.
World Health Organization, International Classification of Diseases, Eighth Revision (ICD-8).  Geneva, Switzerland World Health Organization1967;
34.
World Health Organization, International Classification of Diseases, Ninth Revision (ICD-9).  Geneva, Switzerland World Health Organization1978;
35.
World Health Organization, International Statistical Classification of Diseases, 10th Revision (ICD-10).  Geneva, Switzerland World Health Organization1992;
36.
National Board of Health and Welfare, The cause of death register.  July1 2004;Available at:http://www.sos.se/epcAccessed July 2, 2004
37.
Cachelin  FMMaher  B Is amenorrhea a critical criterion for anorexia nervosa?  J Psychosom Res 1998;44435- 440PubMedGoogle ScholarCrossref
38.
Garfinkel  PLin  EGoering  PSpegg  CGoldbloom  DKennedy  SKaplan  ASWoodside  DB Should amenorrhoea be necessary for the diagnosis of anorexia nervosa? evidence from a Canadian community sample.  Br J Psychiatry 1996;168500- 506PubMedGoogle ScholarCrossref
39.
Bulik  CSullivan  PKendler  K An empirical study of the classification of eating disorders.  Am J Psychiatry 2000;157886- 895PubMedGoogle ScholarCrossref
40.
Walters  EEKendler  KS Anorexia nervosa and anorexic-like syndromes in a population-based female twin sample.  Am J Psychiatry 1995;15264- 71PubMedGoogle Scholar
41.
American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition.  Washington, DC American Psychiatric Association1987;
42.
World Health Organization, Composite International Diagnostic Interview (CIDI).  Geneva, Switzerland World Health Organization1990;
43.
Eysenck  HJEysenck  SBG Manual of the Eysenck Personality Questionnaire.  London, England Hodder and Stoughton1975;
44.
Viken  RJRose  RJKaprio  JKoskenvuo  M A developmental genetic analysis of adult personality: extraversion and neuroticism from 18 to 59 years of age.  J Pers Soc Psychol 1994;66722- 730PubMedGoogle ScholarCrossref
45.
Pedersen  NPlomin  RMcClearn  GFriberg  L Neuroticism, extraversion, and related traits in adult twins reared apart and reared together.  J Pers Soc Psychol 1988;55950- 957PubMedGoogle ScholarCrossref
46.
Floderus-Myrhed  BPedersen  NRasmuson  I Assessment of heritability for personality, based on a short-form of the Eysenck Personality Inventory: a study of 12 898 twin pairs.  Behav Genet 1980;10153- 162PubMedGoogle ScholarCrossref
47.
SAS Institute Inc, SAS/STAT Software: Version 9.  Cary, NC SAS Institute Inc2004;
48.
Zeger  SLLiang  KYAlbert  PS Models for longitudinal data: a generalized estimating equation approach.  Biometrics 1988;441049- 1060PubMedGoogle ScholarCrossref
49.
Zeger  SLiang  K Longitudinal data analysis for discrete and continuous outcomes.  Biometrics 1986;42121- 130PubMedGoogle ScholarCrossref
50.
Zeger  SLiang  K An overview of methods for the analysis of longitudinal data.  Stat Med 1992;111825- 1839PubMedGoogle ScholarCrossref
51.
Diggle  PLiang  KZeger  S Analysis of Longitudinal Data.  Oxford, England Oxford Science1994;
52.
Falconer  DS The inheritance of liability to certain diseases, estimated from the incidence among relatives.  Ann Human Genet (Lond) 1965;2951- 76Google ScholarCrossref
53.
Pearson  K Mathematical contributions to the theory of evolution, VIII: on the correlation of characters not quantitatively measurable.  Proc R Soc 1901;66241- 244Google ScholarCrossref
54.
Sullivan  PEaves  L Evaluation of analyses of univariate discrete twin data.  Behav Genet 2002;32221- 227PubMedGoogle ScholarCrossref
55.
Neale  MBoker  SXie  GMaes  H Mx: Statistical Modeling. 5th Richmond Medical College of Virginia, Dept of Psychiatry2001;
56.
Neale  MCMiller  MB The use of likelihood-based confidence intervals in genetic models.  Behav Genet 1997;27113- 120PubMedGoogle ScholarCrossref
57.
Cohen  J A coefficient of agreement for nominal scales.  Educ Psychol Meas 1960;2037- 46Google ScholarCrossref
58.
Landis  JKoch  G The measurement of observer agreement for categorical data.  Biometrics 1977;33159- 174PubMedGoogle ScholarCrossref
59.
Sullivan  PFBulik  CMFear  JLPickering  A Outcome of anorexia nervosa.  Am J Psychiatry 1998;155939- 946PubMedGoogle Scholar
60.
 JuhaeriStevens  JJones  DArnett  D Associations of aging and birth cohort with body mass index in a biethnic cohort.  Obes Res 2003;11426- 433PubMedGoogle ScholarCrossref
61.
Michels  KEkbom  A Caloric restriction and incidence of breast cancer.  JAMA 2004;2911226- 1230PubMedGoogle ScholarCrossref
62.
Mellemkjaer  LEmborg  CGridley  GMunk-Jorgensen  PJohansen  CTjonneland  AKjaer  SKOlsen  JH Anorexia nervosa and cancer risk.  Cancer Causes Control 2001;12173- 177PubMedGoogle ScholarCrossref
63.
Korndorfer  SLucas  ASuman  VCrowson  CKrahn  LMelton  L  III Long-term survival of patients with anorexia nervosa: a population-based study in Rochester, Minn.  Mayo Clin Proc 2003;78278- 284PubMedGoogle ScholarCrossref
64.
Sullivan  P Discrepant results regarding long-term survival of patients with anorexia nervosa?  Mayo Clin Proc 2003;78273- 274PubMedGoogle ScholarCrossref
65.
Albanes  D Caloric intake, body weight, and cancer: a review.  Nutr Cancer 1987;9199- 217PubMedGoogle ScholarCrossref
66.
Kushi  LGiovannucci  E Dietary fat and cancer.  Am J Med 2002;113 ((suppl 9B)) 63S- 70SPubMedGoogle ScholarCrossref
67.
Kaye  WBulik  CThornton  LBarbarich  BSMasters  KPrice Foundation Collaborative Group, Comorbidity of anxiety disorders with anorexia and bulimia nervosa.  Am J Psychiatry 2004;1612215- 2221PubMedGoogle ScholarCrossref
68.
Bulik  CSullivan  PFear  JJoyce  P Eating disorders and antecedent anxiety disorders: a controlled study.  Acta Psychiatr Scand 1997;96101- 107PubMedGoogle ScholarCrossref
69.
Eaves  LJEysenck  HJMartin  NG Genes, Culture and Personality: An Empirical Approach.  New York, NY Oxford University Press Inc1989;
70.
Costa  PTMcCrae  RR Major contributions to the psychology of personality. In:Modgil  SModgil  Ceds. Hans Eysenck: Consensus and Controversy Philadelphia, Pa Falmer Press1986;63- 72Google Scholar
71.
Eysenck  HJ The Maudsley Personality Inventory.  San Diego, Calif EdITS1962;
72.
Cattell  RBEber  HWTatsuoka  MM The Handbook for the Sixteen Personality Factor Questionnaire.  Champaign, Ill Institute for Personality and Ability Testing1970;
73.
Cloninger  CR A unified biosocial theory of personality and its role in the development of anxiety states.  Psychiatr Dev 1986;4167- 226PubMedGoogle Scholar
74.
Bech  PJørgensen  BJeppesen  KPoulson  DLVanggaard  T Personality in depression: concordance between clinical assessment and questionnaires.  Acta Psychiatr Scand 1986;74263- 268PubMedGoogle ScholarCrossref
75.
Kendell  REDiScipio  WJ EPI scores of patients with depressive illnesses.  Br J Psychiatry 1968;114767- 770PubMedGoogle ScholarCrossref
76.
Strober  M Pathologic fear conditioning and anorexia nervosa: on the search for novel paradigms.  Int J Eat Disord 2004;35504- 508PubMedGoogle ScholarCrossref
77.
Grice  DEHalmi  KAFichter  MMStrober  MWoodside  DBTreasure  JTKaplan  ASMagistretti  PJGoldman  DBulik  CMKaye  WHBerrittini  WH Evidence for a susceptibility gene for anorexia nervosa on chromosome 1.  Am J Hum Genet 2002;70787- 792PubMedGoogle ScholarCrossref
Original Article
March 2006

Prevalence, Heritability, and Prospective Risk Factors for Anorexia Nervosa

Author Affiliations

Author Affiliations: Departments of Psychiatry (Drs Bulik, Sullivan, and Tozzi) and Genetics (Drs Sullivan and Furberg) and Department of Nutrition, School of Public Health (Dr Bulik), University of North Carolina at Chapel Hill; and Department of Medical Epidemiology and Biostatistics, Karolinska Institute, Stockholm, Sweden (Drs Lichtenstein and Pedersen).

Arch Gen Psychiatry. 2006;63(3):305-312. doi:10.1001/archpsyc.63.3.305
Abstract

Context  Anorexia nervosa (AN) is a serious mental illness with marked morbidity and mortality.

Objective  To explore the prevalence, heritability, and prospectively assessed risk factors for AN in a large population-based cohort of Swedish twins.

Design  During a 4-year period ending in 2002, all living, contactable, interviewable, and consenting twins in the Swedish Twin Registry (N = 31 406) born between January 1, 1935, and December 31, 1958, underwent screening for a range of disorders, including AN. Information collected systematically in 1972 to 1973, before the onset of AN, was used to examine prospective risk factors for AN.

Setting  Population-based sample of twins in Sweden.

Participants  Cases of AN were identified as those individuals who met full DSM-IV criteria by means of clinical interview of the Swedish Twin Registry, who had a hospital discharge diagnosis of AN, or who had a cause-of-death certificate including an AN diagnosis.

Results  The overall prevalence of AN was 1.20% and 0.29% for female and male participants, respectively. The prevalence of AN in both sexes was greater among those born after 1945. Individuals with lifetime AN reported lower body mass index, greater physical activity, and better health satisfaction than those without lifetime AN. Anorexia nervosa was inversely associated with the development of overweight (odds ratio, 0.29; 95% confidence interval [CI], 0.16-0.54 [P<.001]). The heritability of narrowly defined DSM-IV AN (additive genetic effects) was estimated to be a2 = 0.56 (95% CI, 0.00-0.87), with the remaining variance attributable to shared environment (c2 = 0.05; 95% CI, 0.00-0.64) and unique environment (e2 = 0.38; 95% CI, 0.13-0.84). Neuroticism measured about 3 decades before the diagnostic assessment was significantly associated with the development of later AN (odds ratio, 1.62; 95% CI, 1.27-2.05 [P<.001]).

Conclusions  The prevalence of AN was higher in both male and female participants born after 1945. Individuals who survive AN and who no longer have body mass indexes in the AN range appear to be at lower risk for the development of overweight. Prospectively assessed neuroticism was associated with the subsequent development of AN, the liability to which is under considerable genetic influence.

Anorexia nervosa (AN) is a perplexing psychiatric illness that primarily strikes female patients and is associated with the highest mortality rate of any mental disorder.1,2 Hoek and van Hoeken3 report the average prevalence of AN to be 0.3%, with the prevalence of subthreshold variants ranging from 0.37% to 1.3%.4,5 The data on temporal incidence trends are conflicting, with some studies suggesting that the incidence is increasing6-12 and others reporting stable rates.13-17 Hoek and van Hoeken3 indicate that the incidence of AN increased during the past century, reaching a plateau in the 1970s. The peak age of onset is 15 to 19 years.18

Anorexia nervosa is familial, with significantly elevated relative risks for AN, bulimia nervosa, and eating disorders not otherwise specified in family members.19,20 The relative risk of AN in family members of males with AN is even higher (>20).21 Given the relative rarity of the condition, twin studies have used various strategies to enhance statistical power, including reporting the heritability of AN in the context of a bivariate twin analysis with AN and major depression (a2 = 58%; 95% confidence interval [CI], 0.33-0.84)22; basing analysis on a simple question of “Have you ever had anorexia nervosa?” (a2 = 48%; 95% CI, 0.27-0.65)23; and using a broad definition of AN syndrome (a2 = 76%; 95% CI, 0.35-0.95).24 In each case, shared environment was negligible (c2), and the remaining variance was primarily attributable to unique environmental factors (e2). Although important contributions, each of these approaches is suboptimal, and no heritability estimates of narrowly defined threshold DSM-IV AN currently exist.

Despite the seriousness of the disorder, we know little about risk factors for the development of AN. Risk factor studies for AN have been challenged by the relative rarity of AN in that longitudinal observational studies often identify only a small number of cases.4,25 Case-control comparisons have highlighted perfectionism and negative self-evaluation as associated with AN and bulimia nervosa.26 Comprehensive reviews of risk factors for AN exist.27,28 Jacobi et al27 reported several general risk factors for eating disorders, including sex, race/ethnicity, early childhood eating and gastrointestinal tract problems, elevated shape and weight concerns, negative self-evaluation, sexual abuse and other adverse experiences, and general psychiatric morbidity, although none was specific for AN. In terms of factors specifically predictive of AN, only pregnancy- and birth-related complications have been reported from a Swedish population-based study.29 Many of the other identified predictors could simply be prodromal symptoms of the disorder.

The present study had the following goals: (1) to report the prevalence of AN in a large cohort of female and male Swedish twins born between January 1, 1935, and December 31, 1958; (2) using a prospective longitudinal design, to identify variables measured in 1972 to 1973 that were associated with the subsequent onset of AN; and (3) to provide estimates of the genetic and environmental contributions to liability to AN.

Methods
Swedish twin registry

The Swedish Twin Registry (STR) (http://www.mep.ki.se/twinreg/index_en.html) is the largest population-based registry of twin births in the world.30,31 For this study, we focused on the STR subset of twin births in Sweden between January 1, 1935, and December 31, 1958, with both twins alive and who were participants in the Screening Across the Lifespan of Twins cohort (SALT). The SALT interviews included questions on eating habits, smoking habits, and medical illnesses. For the entire STR, data on current vital status, marital status, address, and place of birth are available for more than 99% of the 160 000 registered individuals.

Interview procedures and data sources

During a 4-year period ending in 2002, all living, contactable, interviewable, and consenting twins in the STR underwent screening for a range of disorders, including AN. The interview procedure had the following steps. Each month, we matched the master twin register to national records to update vital status and current address information; we randomly selected approximately 1000 pairs for interviews and sent introductory letters describing the study; and we then attempted to complete a telephone interview about 2 weeks later. This process was repeated during the 4 years of the study. In the SALT interviews, lifetime history of AN was assessed via an interview based on the Structured Clinical Interview for DSM-IV Axis I Disorders.32 All interviews were conducted by trained interviewers with adequate medical background using a computer-based data collection system. Detailed procedural information about the SALT interviews can be found in Lichtenstein et al30 and Pedersen et al.31

Hospital discharge registry identification of cases

The Swedish Hospital Discharge Register covers all public, inpatient hospitalizations in Sweden and is maintained by the Centre for Epidemiology at the National Board of Health and Welfare (http://www.sos.se/epc/english/ParEng.htm). For psychiatric diagnoses, the register is complete from 1973. Each record contains the admission and discharge dates, the main discharge diagnosis, and as many as 8 secondary diagnosis codes from the International Classification of Diseases, Eighth Revision (ICD-8),33International Classification of Diseases, Ninth Revision (ICD-9),34 or International Statistical Classification of Diseases, 10th Revision (ICD-10).35 These diagnoses are based on the clinical diagnosis made by the attending physician. The psychiatric discharge registry contains data from the 3.8 million hospitalizations of 960 000 people. We searched this registry for discharge diagnoses corresponding to AN (codes 306.50 [ICD-8], 307B [ICD-9], and F50.0 [ICD-10]).

Cause-of-death register

The national cause-of-death registry contains information on all deaths in Sweden since 1961.36 The diagnoses and causes of death are coded according to the International Classification of Disease revisions.33-35 To define cases of AN, we used codes 306.50, 307B, and F50.0. Exploration of other codes that might overlap with AN did not reveal additional cases consistent with the presence of AN.

Definitions of an

Given that we had multiple sources of information (the SALT interview, hospital discharge registry, and cause-of-death registry) that could inform us of eating disorder status, we studied 3 definitions of AN. The first definition, all-source AN, includes individuals who met full DSM-IV criteria by means of the SALT interview (n = 121), any registry discharge diagnosis of AN (n = 143), or a death certificate that included an AN diagnosis (n = 2). The second definition, narrow DSM-IV AN, includes individuals who met full DSM-IV criteria by means of the SALT interview (excluding amenorrhea for male participants). The third definition, broad DSM-IV AN, includes individuals who met all DSM-IV criteria for AN with the exception of amenorrhea, as amenorrhea has been shown to be an unreliable diagnostic criterion for AN.37-40

Assessment of additional conditions

In the SALT interview, lifetime DSM-III-R major depression41 and generalized anxiety disorder were assessed during the SALT telephone interview via an adaptation of the Composite International Diagnostic Interview.42 We also report on 4 dichotomously scored general health variables. Affirmative responses to the general health item reflect those individuals who rated their current health as excellent or good. Positive responses to the change in health item reflect those individuals who state that their current health is better than or the same as their health 5 years earlier. Health limitation variables include the “health limits activities” item with a positive score being no limitation and “days health limited activities” item with a positive score being limitations of fewer than 7 days in the past year.

Prospective risk factor questionnaire

A questionnaire was sent in 1972 and 1973 to twins of like-sexed pairs and assessed demographics (height and weight), physical illnesses (including gastric problems), physical activity level, personality, stress, and work exposures. Neuroticism and extraversion were measured using a short form of the Eysenck Personality Inventory43 that has been widely used in previous Scandinavian twin studies.44-46 Each scale score is based on the sum of no (0) or yes (1) responses to 9 items. We used mean imputation if 1 item for a particular scale was missing. Individuals having more than 1 missing item for a particular scale were excluded. Raw scale scores were standardized by using regression to adjust for the effects of age. Stress score was based on the question “Do you experience your daily existence as being very ‘stress-filled’?”

Zygosity determination

Zygosity information for like-sexed pairs was obtained at the time of registry compilation on the basis of questions about childhood resemblance and updated during the most recent screening for those with previously uncertain diagnoses. Four separate validation studies using serology and/or genotyping have shown that these items correctly classify the zygosity of 98% of twin pairs.30

Statistical analyses

Descriptive statistics and basic analyses were conducted using SAS software, version 9.47 Asymmetric, exact CIs were obtained using PROC FREQ.47 We used linear and logistic regression with generalized estimating equations48-51 to investigate the associations with demographic and prospective risk factors.

Most biometrical twin analyses of complex disorders are based on the liability-threshold model.52,53 This model postulates the existence of an unobserved (or latent) liability to an illness that results from numerous components. These components act additively and may be genetic and/or environmental in origin. The existence of a liability threshold is postulated that defines the observed affection status: individuals whose liability is above this threshold are affected; those whose liabililty is below are unaffected. We assume that the observed similarity of members of a twin pair results from variance in 3 types of causes. Additive genetic effects (a2, the cumulative impact of multiple genes individually of small effect) contribute twice as much to the monozygotic as to the dizygotic twin correlation under the assumptions that monozygotic twins are genetically identical, whereas dizygotic twins share half their genes identical by descent (this assumption is clearly not precisely correct but may serve as a first approximation). Shared environmental effects (c2, environmental effects to which both members of a twin pair are exposed) contribute equally to the correlation in monozygotic and dizygotic twins. Individual-specific environmental effects (e2) reflect environmental experiences that are not shared by the members of a twin pair and tend to make them different in liability to illness. The sum of a2, c2, and e2 is constrained to unity.

The goal of univariate twin analysis is to decompose the variance in liability to the disorder in the population studied into that due to a2, c2, and e2. We were able to fit models that estimated these parameters in female participants only, as the prevalence of AN was too low in males to allow biometrical modeling. Thresholds were estimated from the data. By tradition, twin modelers have come to use a parsimony index to determine the best-fitting submodel determined by eliminating effects due to a2 and/or c2. However, we have shown that this is ill-advised, because simulations indicated that this practice frequently yielded the incorrect model and sharply and misleadingly biased parameter estimates.54 Therefore, we present the full ACE model (a2, c2, and e2), including parameter estimates and CIs, together with the χ2 goodness-of-fit statistic. We used Mx for all analyses55 with a script from the GenomEUtwin library (http://www.psy.vu.nl/mxbib) to analyze contingency tables and with CI calculation.56

Results
Sample description

Of all eligible twins (N = 41 499), 31 406 individual twins responded, giving an individual response rate of 75.68%. Data were obtained from both members of 12 407 pairs and from 1 member of 6592 pairs. Of the complete pairs, 3269 pairs were monozygotic, 9010 pairs were dizygotic, and 128 pairs were of unknown zygosity.

The median age of the sample was 54.6 years (interquartile range, 49.0-57.6 years). The sex distribution had a slight female predominance (50.6%). Participants had completed a median of 11.0 years of education (interquartile range, 9-14 years). The 2 cohorts consisted of twins born between 1935 and 1944 and those born between 1945 and 1958.

Reliability and predictors of participation

To establish test-retest reliability, 105 individual twins who completed the telephone screening were selected at random and reinterviewed within 2 weeks. We assessed short-term reliability for 2 key variables associated with eating disorders—low weight and dieting behavior. The Cohen κ57 was 0.63 for AN criterion A (“Did you ever weigh less than other people thought you should weigh?”) and 0.48 for the broader variable querying dieting history. These κ values suggest moderate to substantial test-retest reliability.58

We had access to data on all eligible twins independent of participation and were thus able to examine the predictors of participation. Significant predictors of participation in a multivariate logistic regression model included being a monozygotic twin (odds ratio [OR], 1.68 [P<.001]), having a co-twin who was also eligible for the present study (OR, 1.44 [P<.001]), female sex (OR, 1.39 [P<.001]), fewer total hospitalizations (dichotomized at the 75th percentile of 4 hospitalizations; OR, 1.18 [P<.001]), and age (OR, 1.01 [P<.001]), but not registration in the Swedish Cancer Registry (OR, 1.01 [P = .85]). Having a hospital discharge diagnosis of AN was not a significant predictor of participation in the SALT interview (OR, 0.79; 95% CI, 0.56-1.11). A hospital discharge diagnosis of AN was associated with increased mortality according to the death registry (OR, 2.18; 95% CI, 1.33-3.58).

Zygosity bias

There was no association between zygosity and any anorexia-related variable in same-sex female twins (P≥.19 for all comparisons). The absence of statistically significant association between zygosity and the traits under study suggests that zygosity bias did not have an impact on the results.

Prevalence

Table 1 presents the prevalence of the 3 diagnostic definitions of AN by sex and cohort. The prevalence of AN for all 3 definitions of illness was significantly greater in the younger cohort of twins (all-source AN, χ2 = 5.73 [P<.001]; narrow DSM-IV AN, χ2 = 4.91 [P<.001]; broad DSM-IV AN, χ2 = 6.19 [P<.001]). There were no cases of broad or narrow DSM-IV AN identified in male participants in the older cohort of SALT twins.

Sample characteristics

Table 2 presents the demographic and clinical characteristics obtained from SALT interviews of individuals with narrow DSM-IV AN and broad DSM-IV AN compared with those without AN. Individuals with narrow DSM-IV AN were significantly younger and had significantly more education than those without AN. In addition, after controlling for age, they had significantly lower current body mass index (BMI) (calculated as weight in kilograms divided by the square of height in meters), but there were no differences in lifetime maximum BMI. Their age at maximum lifetime BMI was significantly lower than those without AN. As expected, individuals with narrow DSM-IV AN reported significantly greater behavioral features of eating disorders, including dieting, purging, laxative use, vomiting, and excessive exercise, than those without AN. They also reported significantly higher prevalences of lifetime major depressive disorder and generalized anxiety disorder. Notably, a significantly greater proportion of individuals with AN reported themselves to be in good or excellent health. The pattern of results for broad DSM-IV AN was identical, with the exception of there being no significant difference in self-rated current general health.

Impact of an on later overweight

To determine the extent to which lifetime AN is protective against the development of later overweight (BMI, ≥25) (http://www.cdc.gov/nccdphp/dnpa/obesity/defining.htm), we used logistic regression with generalized estimating equation corrections, with overweight status at the time of the SALT interview as the outcome variable. We focused on individuals with lifetime narrow DSM-IV AN who did not have a BMI at the time of the SALT interview of 17.5 or lower (the BMI cutoff for ICD-10 AN, which was used as a proxy for current illness to exclude individuals from the analysis who may still have had symptomatic AN at the time of the SALT interview). Results indicated that 36.5% of individuals without AN and 12.5% of individuals with a history of AN were overweight (OR, 0.29; 95% CI, 0.16-0.54 [P<.001]), suggesting that a lifetime history of AN is protective against the later development of overweight.

Prospective risk factors

Table 3 presents the results of logistic regressions using generalized estimating equation corrections predicting all-source AN status from prospective data collected in 1973 for female participants only. We chose to use the all-source AN variable as the outcome to minimize the possibility of including individuals who were cases of AN in the control sample. We restricted the analyses to female participants because there were few males with AN. Using this definition, we excluded 48 individuals who had all-source AN but who reported their onset of AN to be before 1973 (using a 5-year grace period to account for the duration of the 1973 assessment and inaccuracies of recall). Seventy-six individuals met criteria for all-source AN and reported their onset of AN to be after 1973. These 76 women constitute the sample for this analysis in comparison with 12 999 women who did not meet the criteria for AN. Based on the literature and available data, we chose the following 7 self-reported potential predictors assessed in 1973: BMI, gastric problems, excessive exercise (defined as much or very much physical activity vs little or almost no physical activity), perceived excessive life stress, neuroticism, and extraversion. Of these variables, only higher neuroticism in 1973 was strongly predictive of all-source AN (P<.001).

Tetrachoric correlations

For twin modeling, we focused on twin data from the SALT interview and therefore present results for narrow and broad DSM-IV AN. Table 4 presents descriptive data for the twins used in this report. The tetrachoric correlations and contingency table data for narrow and broad DSM-IV AN are shown for the different types of twin pairs (monozygotic male, dizygotic male-male, monozygotic female, dizygotic female-female, and dizygotic male-female and female-male twin pairs).

First, as can be seen in the rightmost columns of Table 4, the rarity of AN in male participants and the infrequency of concordant pairs rendered it impossible to calculate tetrachoric correlations for male twins for either definition of illness. Second, for both definitions of illness in female participants, the tetrachoric correlations for monozygotic twin pairs were substantially larger than those for dizygotic pairs. Third, for opposite-sex dizygotic twin pairs, tetrachoric correlations were negative, reflecting the absence of concordant male-female pairs for either definition of illness and the disproportionate female-male ratio of AN.

Twin modeling

We conducted twin analyses only for female-female twin pairs, given the rarity of AN in male participants. Table 5 depicts the results of structural equation modeling for the narrow and broad DSM-IV definitions of AN from SALT. For both definitions of illness, the structural equation models provided an adequate fit to the data. We suggest a predominant focus on the 95% CIs rather than the point estimates because these contain information not only about the estimate but also about the precision with which the estimate is known.54 The heritability of narrow DSM-IV AN was estimated to be a2 = 0.56 (95% CI, 0.00-0.87) with the remaining variance attributable to shared environment (c2 = 0.05; 95% CI, 0.00-0.64) and unique environment (e2 = 0.38; 95% CI, 0.13-0.84). Estimates for broad DSM-IV AN are presented in Table 5.

Comment

This study represents, to our knowledge, the largest twin study conducted to date of individuals with rigorously diagnosed AN. Our results confirm and extend the findings of previous studies on prevalence, risk factors, and heritability.

Consistent with several studies, the lifetime prevalence of AN identified by all sources was 1.20% in female participants and 0.29% in male participants, reflecting the typically observed disproportionate sex ratio. Similarly, our data show a clear increase in prevalence of DSM-IV AN (broadly and narrowly defined) with historical time in Swedish twins. The increase was apparent for both sexes. Hoek and van Hoeken3 also reported a consistent increase in prevalence, with a leveling out of the trajectory around the 1970s. Future studies in younger STR participants will allow verification of this observation.

Several observed differences between individuals with and without AN were expected, ie, more frequent endorsement of symptoms of eating disorders. Other differences are noteworthy. Consistent with previous observations, individuals with lifetime AN reported lower BMIs at the time of interview than did individuals with no history of AN. Although this could be partially accounted for by the presence of currently symptomatic individuals in the sample, our results remained unchanged when we excluded individuals likely to have current AN (ie, current BMI, ≤17.5). Previous studies have shown that, even after recovery, individuals with a history of AN have a low BMI.59 Although perhaps obvious, a history of AN appears to offer protection against becoming overweight. The protective effect also holds for obesity (BMI, ≥30), although there were too few individuals in the sample with histories of AN who had become obese for meaningful analyses. Despite the obvious nature of this observation, the mechanism whereby protection against overweight is afforded is not immediately clear. Those with a history of AN reported greater current exercise and a perception of being in better physical health. One possible interpretation of this pattern of findings is that individuals with a history of AN continue to display subthreshold symptoms of AN (ie, excessive exercise and caloric restriction) that contribute to their low BMIs. Alternatively, symptoms that were pathologic during acute phases of AN, such as excessive exercise and decreased caloric intake, may resolve over time into healthy behaviors, such as consistent exercise patterns and a healthful diet, that result in better weight control and self-rated health.

Regardless of which of these hypotheses is true, another intriguing difference is that individuals with lifetime AN report a lower age at highest BMI, although the magnitude of the highest lifetime BMI does not differ in those with and without a history of AN. Those with AN report their highest lifetime BMIs early in their fourth decade of life on average, whereas those without AN report their highest BMIs in the middle of their fifth decade of life (close to the age at interview). On a population level, adults tend to gain on average 2.25 kg (5 lb) per decade until reaching their eighth decade of life.60 Although more detailed data are necessary to make definitive statements about different weight trajectories, our results suggest not only that individuals with AN may maintain low BMIs but also that they may not follow the typical adult weight gain trajectories. These data are particularly intriguing in light of recent reports of AN being associated with reduced risk of certain cancers61,62 and protective against mortality due to diseases of the circulatory system.63,64 Energy intake is closely related to fat intake and obesity, both of which have also been related to cancer development65,66 and both of which are reduced in AN. Further detailed studies of the weight trajectories and health of individuals with histories of AN are required to explicate the nature and magnitude of these intriguing findings.

Of the variables assessed in 1972 to 1973, neuroticism emerged as the only significant prospective predictor of AN. This is notable because there have been few truly prospective risk factor studies of AN. This finding is intriguing and entirely consistent with studies reporting increased antecedent anxiety disorders in individuals in whom AN later developed.67,68 Individuals with high neuroticism scores are characterized by emotional instability, low self-esteem, and feelings of anxiety, depression, and guilt.69 The construct of neuroticism is extraordinarily robust70: neuroticism or a very similar construct can be found in essentially every major theory of personality.71-73 Our results suggest that the early existence of this trait may predispose to AN. However, neuroticism is known to covary with depression,74,75 and in the absence of data on depression in the 1972-1973 assessment, we were unable to determine the extent to which high neuroticism reflected current mood state. Nonetheless, these findings, coupled with the elevated rates of generalized anxiety disorder and major depression in individuals with AN observed in this sample, are in line with an increasingly large body of clinical literature suggesting that anxiety proneness may represent a risk pathway to AN.76 What remains unknown is whether neuroticism is a nonspecific predictor of the development of psychopathology in general or whether it is specifically predictive of the emergence of AN. Finally and somewhat surprisingly, BMI and exercise levels did not predict AN onset.

Although the strengths of this study include the sample size and prospective risk factor data, several limitations must be considered. First, our sample represents individuals born between 1935 and 1958, before widespread recognition of eating disorders. Although our diagnostic interviews were lifetime retrospective interviews, many individuals in this cohort may never have been detected or labeled as having had AN. Moreover, the likelihood of detection may have increased in the latter birth years, meaning that the observed increase in prevalence, rather than being a true secular trend, may merely be a reflection of increased detection. Second, as is often the case in large population-based investigations, one is forced to make trade-offs between the size of a population and the comprehensiveness of an evaluation. Thus, we are hampered by somewhat cursory evaluations of some of our variables such as exercise, health, and personality.

Owing to the relative rarity of the condition, we have not had reliable estimates of the contribution of genes and environment to liability to AN. To our knowledge, ours represents the first twin study of both narrowly and broadly defined AN characterized by structured diagnostic interviews. As such, these results suggest a substantial contribution of genetic factors to liability to AN, with heritability estimates of 0.56 and 0.31 for narrow and broad DSM-IV AN, respectively. Despite the size of the study, our estimates remain imprecise, as reflected in the broad CIs. Comparing the biometrical models for narrow and broad DSM-IV AN, the narrow DSM-IV AN model fit the data better and yielded a higher heritability estimate. The more narrowly defined disorder may represent a more homogeneous phenotype that yields higher heritability estimates. Indeed, molecular genetic research has found that restricting the clinical definition of AN to the narrow phenotype of restricting AN enabled the detection of linkage, whereas broader phenotypes failed to yield significant linkage signals.77

In conclusion, the prevalence of AN increased in both sexes between 1935 and 1958, while consistently afflicting females disproportionately. Individuals with a history of AN appear to be protected from the development of overweight later in life. Anorexia nervosa is a moderately heritable psychiatric disorder that may be predicted by the presence of early neuroticism.

Correspondence: Cynthia M. Bulik, PhD, Department of Psychiatry, University of North Carolina at Chapel Hill, CB 7160, 101 Manning Dr, Chapel Hill, NC 27599-7160 (cbulik@med.unc.edu).

Submitted for Publication: January 25, 2005; final revision received July 1, 2005; accepted July 28, 2005.

Funding/Support: This study was supported by grants NS-041483 (Dr Pedersen), CA-085739 (Dr Sullivan), and AI-056014 (Dr Sullivan) from the National Institutes of Health, Bethesda, Md; and by a grant from the Foundation of Hope, Raleigh, NC (Dr Bulik). The Swedish Twin Registry is supported by grants from the Swedish Department of Higher Education, Stockholm, the Swedish Scientific Council, Stockholm, and an unrestricted grant from AstraZeneca, Lund, Sweden.

References
1.
Sullivan  PF Mortality in anorexia nervosa.  Am J Psychiatry 1995;1521073- 1074PubMedGoogle Scholar
2.
Harris  ECBarraclough  B Excess mortality of mental disorder.  Br J Psychiatry 1998;17311- 53PubMedGoogle ScholarCrossref
3.
Hoek  Hvan Hoeken  D Review of the prevalence and incidence of eating disorders.  Int J Eat Disord 2003;34383- 396PubMedGoogle ScholarCrossref
4.
McKnight Investigators, Risk factors for the onset of eating disorders in adolescent girls: results of the McKnight Longitudinal Risk Factor Study.  Am J Psychiatry 2003;160248- 254PubMedGoogle ScholarCrossref
5.
Wittchen  HUNelson  CBLachner  G Prevalence of mental disorders and psychosocial impairments in adolescents and young adults.  Psychol Med 1998;28109- 126PubMedGoogle ScholarCrossref
6.
Eagles  JJohnston  MHunter  DLobban  MMillar  H Increasing incidence of anorexia nervosa in the female population of northeast Scotland.  Am J Psychiatry 1995;1521266- 1271PubMedGoogle Scholar
7.
Jones  DFox  MBabigan  HHutton  H Epidemiology of anorexia nervosa in Monroe County, New York: 1960-76.  Psychosom Med 1980;42551- 558PubMedGoogle ScholarCrossref
8.
Lucas  ARBeard  CMO’Fallon  WMKurland  LT Anorexia nervosa in Rochester, Minnesota: a 45-year study.  Mayo Clin Proc 1988;63433- 442PubMedGoogle ScholarCrossref
9.
Møller-Madsen  SNystrup  J Incidence of anorexia nervosa in Denmark.  Acta Psychiatr Scand 1992;86197- 200PubMedGoogle ScholarCrossref
10.
Szmukler  G The epidemiology of anorexia nervosa and bulimia.  J Psychiatr Res 1985;19143- 153PubMedGoogle ScholarCrossref
11.
Willi  JGrossman  S Epidemiology of anorexia nervosa in a defined region of Switzerland.  Am J Psychiatry 1983;140564- 567PubMedGoogle Scholar
12.
Milos  GSpindler  ASchnyder  UMartz  JHoek  HWilli  J Incidence of severe anorexia nervosa in Switzerland: 40 years of development.  Int J Eat Disord 2004;36118- 119PubMedGoogle ScholarCrossref
13.
Hall  AHay  P Eating disorder patient referrals from a population region 1977-1986.  Psychol Med 1991;21697- 701PubMedGoogle ScholarCrossref
14.
Hoek  HBartelds  ABosveld  Jvan der Graaf  YLimpens  VEMaiwald  MSpaaij  CJ Impact of urbanization on detection rates of eating disorders.  Am J Psychiatry 1995;1521272- 1278PubMedGoogle Scholar
15.
Joergensen  J The epidemiology of eating disorders in Fyn County, Denmark, 1977-1986.  Acta Psychiatr Scand 1992;8530- 34PubMedGoogle ScholarCrossref
16.
Nielsen  S The epidemiology of anorexia nervosa in Denmark from 1973-1987: a nationwide register study of psychiatric admission.  Acta Psychiatr Scand 1990;81507- 514PubMedGoogle ScholarCrossref
17.
Willi  JGiacometti  GLimacher  B Update on the epidemiology of anorexia nervosa in a defined region of Switzerland.  Am J Psychiatry 1990;1471514- 1517PubMedGoogle Scholar
18.
Lucas  ARBeard  CMO’Fallon  WMKurland  LT 50-Year trends in the incidence of anorexia nervosa in Rochester, Minn: a population-based study.  Am J Psychiatry 1991;148917- 922PubMedGoogle Scholar
19.
Lilenfeld  LKaye  WGreeno  CMerikangas  KRPlotnicov  KPollice  CRao  RStrober  MBulik  CMNagy  L A controlled family study of restricting anorexia and bulimia nervosa: comorbidity in probands and disorders in first-degree relatives.  Arch Gen Psychiatry 1998;55603- 610PubMedGoogle ScholarCrossref
20.
Strober  MFreeman  RLampert  CDiamond  JKaye  W Controlled family study of anorexia nervosa and bulimia nervosa: evidence of shared liability and transmission of partial syndromes.  Am J Psychiatry 2000;157393- 401PubMedGoogle ScholarCrossref
21.
Strober  MFreeman  RLampert  CDiamond  JKaye  W Males with anorexia nervosa: a controlled study of eating disorders in first-degree relatives.  Int J Eat Disord 2001;29263- 269PubMedGoogle ScholarCrossref
22.
Wade  TDBulik  CMNeale  MKendler  KS Anorexia nervosa and major depression: shared genetic and environmental risk factors.  Am J Psychiatry 2000;157469- 471PubMedGoogle ScholarCrossref
23.
Kortegaard  LSHoerder  KJoergensen  JGillberg  CKyvik  KO A preliminary population-based twin study of self-reported eating disorder.  Psychol Med 2001;31361- 365PubMedGoogle ScholarCrossref
24.
Klump  KLMiller  KBKeel  PKMcGue  MIacono  WG Genetic and environmental influences on anorexia nervosa syndromes in a population-based twin sample.  Psychol Med 2001;31737- 740PubMedGoogle ScholarCrossref
25.
Stice  EPresnell  KBearman  S Relation of early menarche to depression, eating disorders, substance abuse, and comorbid psychopathology among adolescent girls.  Dev Psychol 2001;37608- 619PubMedGoogle ScholarCrossref
26.
Fairburn  CGCooper  ZDoll  HWelch  SL Risk factors for anorexia nervosa.  Arch Gen Psychiatry 1999;56468- 476PubMedGoogle ScholarCrossref
27.
Jacobi  CHayward  Cde Zwaan  MKraemer  HAgras  W Coming to terms with risk factors for eating disorders: application of risk terminology and suggestions for a general taxonomy.  Psychol Bull 2004;13019- 65PubMedGoogle ScholarCrossref
28.
Walsh  BBulik  CFairburn  CHalmi  KHerzog  DGolden  NEA Defining eating disorders. In:Evans  DFoa  EGur  R  et al. eds. Treating and Preventing Adolescent Mental Health Disorders: What We Know and What We Don't Know New York, NY Oxford University Press, Annenberg Foundation Trust at Sunnylands, and Annenberg Public Policy Center of the University of Pennsylvania2005;Google Scholar
29.
Cnattingius  SHultman  CDahl  MSparen  P Very preterm birth, birth trauma, and the risk of anorexia nervosa among girls.  Arch Gen Psychiatry 1999;56634- 638PubMedGoogle ScholarCrossref
30.
Lichtenstein  PDe Faire  UFloderus  BSvartengren  MSvedberg  PPedersen  N The Swedish Twin Registry: a unique resource for clinical, epidemiological and genetic studies.  J Intern Med 2002;252184- 205PubMedGoogle ScholarCrossref
31.
Pedersen  NLichtenstein  PSvedberg  P The Swedish Twin Registry in the third millennium.  Twin Res 2002;5427- 432PubMedGoogle ScholarCrossref
32.
First  MSpitzer  RGibbon  MWilliams  J Structured Clinical Interview for DSM-IV Axis I Disorders, Research Version, Patient Edition.  New York Biometrics Research, New York State Psychiatric Institute1997;
33.
World Health Organization, International Classification of Diseases, Eighth Revision (ICD-8).  Geneva, Switzerland World Health Organization1967;
34.
World Health Organization, International Classification of Diseases, Ninth Revision (ICD-9).  Geneva, Switzerland World Health Organization1978;
35.
World Health Organization, International Statistical Classification of Diseases, 10th Revision (ICD-10).  Geneva, Switzerland World Health Organization1992;
36.
National Board of Health and Welfare, The cause of death register.  July1 2004;Available at:http://www.sos.se/epcAccessed July 2, 2004
37.
Cachelin  FMMaher  B Is amenorrhea a critical criterion for anorexia nervosa?  J Psychosom Res 1998;44435- 440PubMedGoogle ScholarCrossref
38.
Garfinkel  PLin  EGoering  PSpegg  CGoldbloom  DKennedy  SKaplan  ASWoodside  DB Should amenorrhoea be necessary for the diagnosis of anorexia nervosa? evidence from a Canadian community sample.  Br J Psychiatry 1996;168500- 506PubMedGoogle ScholarCrossref
39.
Bulik  CSullivan  PKendler  K An empirical study of the classification of eating disorders.  Am J Psychiatry 2000;157886- 895PubMedGoogle ScholarCrossref
40.
Walters  EEKendler  KS Anorexia nervosa and anorexic-like syndromes in a population-based female twin sample.  Am J Psychiatry 1995;15264- 71PubMedGoogle Scholar
41.
American Psychiatric Association, Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition.  Washington, DC American Psychiatric Association1987;
42.
World Health Organization, Composite International Diagnostic Interview (CIDI).  Geneva, Switzerland World Health Organization1990;
43.
Eysenck  HJEysenck  SBG Manual of the Eysenck Personality Questionnaire.  London, England Hodder and Stoughton1975;
44.
Viken  RJRose  RJKaprio  JKoskenvuo  M A developmental genetic analysis of adult personality: extraversion and neuroticism from 18 to 59 years of age.  J Pers Soc Psychol 1994;66722- 730PubMedGoogle ScholarCrossref
45.
Pedersen  NPlomin  RMcClearn  GFriberg  L Neuroticism, extraversion, and related traits in adult twins reared apart and reared together.  J Pers Soc Psychol 1988;55950- 957PubMedGoogle ScholarCrossref
46.
Floderus-Myrhed  BPedersen  NRasmuson  I Assessment of heritability for personality, based on a short-form of the Eysenck Personality Inventory: a study of 12 898 twin pairs.  Behav Genet 1980;10153- 162PubMedGoogle ScholarCrossref
47.
SAS Institute Inc, SAS/STAT Software: Version 9.  Cary, NC SAS Institute Inc2004;
48.
Zeger  SLLiang  KYAlbert  PS Models for longitudinal data: a generalized estimating equation approach.  Biometrics 1988;441049- 1060PubMedGoogle ScholarCrossref
49.
Zeger  SLiang  K Longitudinal data analysis for discrete and continuous outcomes.  Biometrics 1986;42121- 130PubMedGoogle ScholarCrossref
50.
Zeger  SLiang  K An overview of methods for the analysis of longitudinal data.  Stat Med 1992;111825- 1839PubMedGoogle ScholarCrossref
51.
Diggle  PLiang  KZeger  S Analysis of Longitudinal Data.  Oxford, England Oxford Science1994;
52.
Falconer  DS The inheritance of liability to certain diseases, estimated from the incidence among relatives.  Ann Human Genet (Lond) 1965;2951- 76Google ScholarCrossref
53.
Pearson  K Mathematical contributions to the theory of evolution, VIII: on the correlation of characters not quantitatively measurable.  Proc R Soc 1901;66241- 244Google ScholarCrossref
54.
Sullivan  PEaves  L Evaluation of analyses of univariate discrete twin data.  Behav Genet 2002;32221- 227PubMedGoogle ScholarCrossref
55.
Neale  MBoker  SXie  GMaes  H Mx: Statistical Modeling. 5th Richmond Medical College of Virginia, Dept of Psychiatry2001;
56.
Neale  MCMiller  MB The use of likelihood-based confidence intervals in genetic models.  Behav Genet 1997;27113- 120PubMedGoogle ScholarCrossref
57.
Cohen  J A coefficient of agreement for nominal scales.  Educ Psychol Meas 1960;2037- 46Google ScholarCrossref
58.
Landis  JKoch  G The measurement of observer agreement for categorical data.  Biometrics 1977;33159- 174PubMedGoogle ScholarCrossref
59.
Sullivan  PFBulik  CMFear  JLPickering  A Outcome of anorexia nervosa.  Am J Psychiatry 1998;155939- 946PubMedGoogle Scholar
60.
 JuhaeriStevens  JJones  DArnett  D Associations of aging and birth cohort with body mass index in a biethnic cohort.  Obes Res 2003;11426- 433PubMedGoogle ScholarCrossref
61.
Michels  KEkbom  A Caloric restriction and incidence of breast cancer.  JAMA 2004;2911226- 1230PubMedGoogle ScholarCrossref
62.
Mellemkjaer  LEmborg  CGridley  GMunk-Jorgensen  PJohansen  CTjonneland  AKjaer  SKOlsen  JH Anorexia nervosa and cancer risk.  Cancer Causes Control 2001;12173- 177PubMedGoogle ScholarCrossref
63.
Korndorfer  SLucas  ASuman  VCrowson  CKrahn  LMelton  L  III Long-term survival of patients with anorexia nervosa: a population-based study in Rochester, Minn.  Mayo Clin Proc 2003;78278- 284PubMedGoogle ScholarCrossref
64.
Sullivan  P Discrepant results regarding long-term survival of patients with anorexia nervosa?  Mayo Clin Proc 2003;78273- 274PubMedGoogle ScholarCrossref
65.
Albanes  D Caloric intake, body weight, and cancer: a review.  Nutr Cancer 1987;9199- 217PubMedGoogle ScholarCrossref
66.
Kushi  LGiovannucci  E Dietary fat and cancer.  Am J Med 2002;113 ((suppl 9B)) 63S- 70SPubMedGoogle ScholarCrossref
67.
Kaye  WBulik  CThornton  LBarbarich  BSMasters  KPrice Foundation Collaborative Group, Comorbidity of anxiety disorders with anorexia and bulimia nervosa.  Am J Psychiatry 2004;1612215- 2221PubMedGoogle ScholarCrossref
68.
Bulik  CSullivan  PFear  JJoyce  P Eating disorders and antecedent anxiety disorders: a controlled study.  Acta Psychiatr Scand 1997;96101- 107PubMedGoogle ScholarCrossref
69.
Eaves  LJEysenck  HJMartin  NG Genes, Culture and Personality: An Empirical Approach.  New York, NY Oxford University Press Inc1989;
70.
Costa  PTMcCrae  RR Major contributions to the psychology of personality. In:Modgil  SModgil  Ceds. Hans Eysenck: Consensus and Controversy Philadelphia, Pa Falmer Press1986;63- 72Google Scholar
71.
Eysenck  HJ The Maudsley Personality Inventory.  San Diego, Calif EdITS1962;
72.
Cattell  RBEber  HWTatsuoka  MM The Handbook for the Sixteen Personality Factor Questionnaire.  Champaign, Ill Institute for Personality and Ability Testing1970;
73.
Cloninger  CR A unified biosocial theory of personality and its role in the development of anxiety states.  Psychiatr Dev 1986;4167- 226PubMedGoogle Scholar
74.
Bech  PJørgensen  BJeppesen  KPoulson  DLVanggaard  T Personality in depression: concordance between clinical assessment and questionnaires.  Acta Psychiatr Scand 1986;74263- 268PubMedGoogle ScholarCrossref
75.
Kendell  REDiScipio  WJ EPI scores of patients with depressive illnesses.  Br J Psychiatry 1968;114767- 770PubMedGoogle ScholarCrossref
76.
Strober  M Pathologic fear conditioning and anorexia nervosa: on the search for novel paradigms.  Int J Eat Disord 2004;35504- 508PubMedGoogle ScholarCrossref
77.
Grice  DEHalmi  KAFichter  MMStrober  MWoodside  DBTreasure  JTKaplan  ASMagistretti  PJGoldman  DBulik  CMKaye  WHBerrittini  WH Evidence for a susceptibility gene for anorexia nervosa on chromosome 1.  Am J Hum Genet 2002;70787- 792PubMedGoogle ScholarCrossref
×