When a case is not a case: effects of phenotype misclassification on power and sample size requirements for the transmission disequilibrium test with affected child trios.

Phenotype misclassification in genetic studies can decrease the power to detect association between a disease locus and a marker locus. To date, studies of misclassification have focused primarily on case-control designs. The purpose of this work is to quantify the effects of phenotype misclassification on the transmission disequilibrium test (TDT) applied to affected child trios, where both parents are genotyped. We compute the non-centrality parameter of the distribution corresponding to the TDT statistic when there is linkage and association of a marker locus with a disease locus and there is phenotype misclassification. We verify our analytic calculations with simulations and provide an example sample size calculation. In our simulation studies, the maximum absolute difference between statistical power computed by simulation and analytic methods is 0.018. In an example sample size calculation, we observe that to maintain equivalent power, the required sample size increases when the disease prevalence decreases or when the misclassification rate increases. A 39-fold sample size increase is required when the misclassification rate is 5% and the disease prevalence is 1%. Given the potentially substantial power loss for the TDT in the presence of misclassification, we recommend that researchers incorporate phenotype misclassification into their study design for genetic association using trio data. We have developed freely available software that computes power loss for a fixed sample size or sample size for a fixed power in the presence of phenotype misclassification.