Recruitment bias in twin research: The rule of two-thirds reconsidered

A review of recent studies of adult volunteer twins confirms an earlier report that, in such studies, male and dizygotic twin pairs are underre-presented. The Martin and Wilson [(1982). Behav. Genet.12:467–472] model of recruitment bias assumes that individuals who are twins respond to recruitment in the same way that singletons do. It is argued that twins recruited for twin studies may be influenced by additional factors, and evidence in support of this view is presented. The consequences of the augmented model. illustrated by Monte Carlo methods, suggests the following conclusions: (1) recruitment bias can have unpredictable and sometimes large effects in the form of under- or, more likely, overestimation of the true twin correlations; (2) these effects will tend to be larger among dizygotic, than among monozygotic, twins; and (3) differential recruitment bias is likely to be accompanied by monozygotic (MZ): dizygotic (DZ) differences in the total variance of the traits measured. Showing that the dependent variable has as large a variance in the twin sample as in the reference population, or that the means and variances of the MZ and DZ samples are similar, provides considerable reassurance that the sample statistics may be representative.This work was supported in part by Grant MH37860 from the National Institute of Mental Health.