Objective
Evidence linking childhood body mass index (BMI) with subsequent eating disorders is equivocal. Potential explanations include different study populations and size, and that anorexia nervosa (AN) and bulimia nervosa (BN) should be studied separately. We examined whether birthweight and childhood BMI were associated with subsequent risk of AN and BN in girls.

Method
We included 68,793 girls from the Copenhagen School Health Records Register born between 1960 and 1996 with information on birthweight and measured weights and heights obtained from school health examinations at ages 6–15 years. Diagnoses of AN and BN were retrieved from Danish nationwide patient registers. We used Cox proportional hazards regression to estimate hazard ratios (HRs) and 95% confidence intervals (CIs).

Results
We identified 355 cases of AN (median age: 19.0) and 273 cases of BN (median age: 21.8). Higher childhood BMI was linearly associated with decreasing risk of AN and increasing risk of BN at all childhood ages. At age 6, the HR for AN was 0.85 (95% CI: 0.74–0.97) per BMI z-score and the HR for BN was 1.78 (95% CI: 1.50–2.11) per BMI z-score. Birthweight >3.75 kg was associated with increased risk of BN compared to a birthweight of 3.26–3.75 kg.

Conclusion
Higher BMI in girls at ages 6–15 years was associated with decreasing risk of AN and increasing risk of BN. Premorbid BMI could be relevant for the etiology of AN and BN, and in identifying high risk individuals.

Public significance
Eating disorders are associated with elevated mortality, especially AN. Using a cohort of Copenhagen school children, we linked information on BMI at ages 6–15 years for 68,793 girls with nationwide patient registers. Low childhood BMI was associated with increased risk of AN, whereas high childhood BMI was associated with increased risk of BN. These findings may assist clinicians in identifying individuals at high-risk of these diseases.