Effects of Occult Heterogeneity of Binary Traits on Estimates of Genetic Variation

Diagnostic criteria may not distinguish between two or more subtypes of a disorder, and these subtypes may have entirely different etiologies.  Such 'occult' heterogeneity could systematically affect familial... [ view full abstract ]

Diagnostic criteria may not distinguish between two or more subtypes of a disorder, and these subtypes may have entirely different etiologies.  Such 'occult' heterogeneity could systematically affect familial resemblance for a disorder, and lead to biased estimates of genetic or other sources of variation.  We explored this possibility using simulation studies, in which two or more independent, equally frequent subtypes exist, any one of which is sufficient to cause the disorder.  In the classical twin study, both MZ and DZ correlations are attenuated, but DZ resemblance decreases by a larger proportion, which has the effect of converting additive genetic to non-additive genetic variance.  These effects are more pronounced for rarer (e.g., schizophrenia, 1%) than more common (e.g., depression, 10%) disorders.  We consider different numbers of subtypes, various genetic and environmental correlations between them, and disorder prevalence to understand how hidden subtypes may bias estimates of heritability from studies of twins and genomic assays of individuals with no ostensible family relationship.