Some Biases That May Affect Kin-Cohort Studies for Estimating the Risks from Identified Disease Genes

Abstract

The kin-cohort design can be used to estimate the absolute risk of disease (penetrance) associated with an identified mutation. In the kincohort design a volunteer (proband) is genotyped and provides information on the disease histories (phenotypes) of his or her first degree relatives. We review some of the strengths and weaknesses of this design before focusing on two types of bias. One bias can arise if the joint distribution of phenotypes of family members, conditional on their genotypes, is mis-specified. In particular, the assumption of conditional independence of phenotypes given genotypes can lead to overestimates of penetrance. If probands are sampled completely at random, a composite likelihood approach can be used that is robust to residual familial correlations, given genotypes. If the sample is enriched in probands with disease, however, one is forced into making some assumptions on the joint distribution of the phenotypes, given genotypes. For phenotypes characterized by age at disease onset, biases can result if no allowance is made for an influence of genotype on competing causes of mortality or on mortality rates following disease onset.