Background: Published heritability estimates (h2) for body mass index (BMI) range from as low as 0.05 to as high as 0.90. The purpose of this paper is to introduce new data to help narrow the range of plausible estimates.
Subjects: Subjects were 53 pairs (23 M; 30 F) of monozygotic twins reared apart (MZAs), whose mean BMI was 24.2 (SD = 4.7). BMI's were transformed to approximate normality via the Box-Cox transformation. Twin paris came from the Finnish Twin Cohort (17 pairs), a data base of Japanese twins (10 pairs) and published case histories of primarily American twins (26 pairs).
Results: The h2 for MZAs is given by the correlation among the twin pairs. For the transformed data, the zero-order correlation of twins' BMIs was 0.79 for all twins, 0.63 for the Finnish twins, 0.73 for the Japanese twins and 0.85 for the 'archival' twins. When modeled with regression to control for relevant covariates, the estimate of h2 is either 0.50 or 0.70, depending on one's definition. The semipartial r was 0.50, suggesting that 50% of the total variance in BMI appears to the genetic in origin after controlling the covariates. The partial r was 0.70, suggesting that 70% of the variance in BMI that is not accounted for by the covariates can be attributed to genetic variation. Separation age had a small positive correlation with absolute intra-pair difference in BMI, suggesting that these estimates of h2 are not biased upwards due to early shared environment.
Conclusions: Findings are consistent with past studies of MZAs and suggest that h2 estimates between 0.50 and 0.70 are reasonable. Implications of this finding are discussed.