Abstract:
|
In some real-world disease registries and longitudinal studies, some probands provided rich phenotypic data, but died before providing DNA. The phenotypic and genotypic data of their relatives are often available. Under the assumption of genotypes missing completely at random (MCAR), we derive a power gain formula with a dichotomous outcome, comparing the statistical power with ungenotyped relatives included and that with them excluded. With the theoretical power gain result, we further use simulations to mimic real world data and explore important factors in the power gain under different scenarios. The missingness mechanism, study design, phenotypic heritability, genetic variation frequency, and genetic variation specific heritability are important factors in the change in power for dichotomous outcomes in real world databases with relatives who contribute incomplete genetic information in addition to rich phenotypic data. Our study joins the efforts to leverage real world data with uncertain genetic information and may help uncover real world evidence on the effects of genetic variants on adverse events or diseases.
|