Psychological Medicine

Original Articles

Assessing the heritability of anorexia nervosa symptoms using a marginal maximal likelihood approach

S. E. Mazzeoa1a2 c1, K. S. Mitchella1a3, C. M. Bulika4, T. Reichborn-Kjenneruda5a6a7, K. S. Kendlera3a8 and M. C. Nealea3a8a9

a1 Department of Psychology, Virginia Commonwealth University, Richmond, VA, USA

a2 Department of Pediatrics, Virginia Commonwealth University, Richmond, VA, USA

a3 Virginia Institute for Psychiatric and Behavioral Genetics, Virginia Commonwealth University, Richmond, VA, USA

a4 Departments of Psychiatry and Nutrition, University of North Carolina, Chapel Hill, NC, USA

a5 Division of Mental Health, Norwegian Institute of Public Health, Oslo, Norway

a6 Institute of Psychiatry, University of Oslo, Norway

a7 Department of Epidemiology, Columbia University, New York, NY, USA

a8 Department of Psychiatry, Virginia Commonwealth University, Richmond, VA, USA

a9 Department of Human Genetics, Virginia Commonwealth University, Richmond, VA, USA


Background Assessment of eating disorders at the symptom level can facilitate the refinement of phenotypes. We examined genetic and environmental contributions to liability to anorexia nervosa (AN) symptoms in a population-based twin sample using a genetic common pathway model.

Method Participants were from the Norwegian Institute of Public Health Twin Panel (NIPHTP) and included all female monozygotic (MZ; 448 complete pairs and four singletons) and dizygotic (DZ; 263 complete pairs and four singletons) twins who completed the Composite International Diagnostic Interview (CIDI) assessing DSM-IV Axis I and ICD-10 criteria. Responses to items assessing AN symptoms were included in a model fitted using the marginal maximum likelihood (MML) approach.

Results Heritability of the overall AN diagnosis was moderate [a2=0.22, 95% confidence interval (CI) 0.0–0.50] whereas heritabilities of the specific items varied. Heritability estimates for weight loss items were moderate (a2=0.31–0.34) and items assessing weight concern when at a low weight were smaller (0.18–0.29). Additive genetic factors contributed little to the variance of amenorrhea, which was most strongly influenced by unshared environment (a2=0.16, e2=0.71).

Conclusions AN symptoms are differentially heritable. Specific criteria such as those related to body weight and weight loss history represent more biologically driven potential endophenotypes or liability indices. The results regarding weight concern differ somewhat from those of previous studies, highlighting the importance of assessing genetic and environmental influences on variance of traits within specific subgroups of interest.

(Received September 24 2007)

(Revised February 20 2008)

(Accepted March 06 2008)

(Online publication May 19 2008)