Asymptotic powers for matched trend tests and robust matched trendtests in case-control genetic association studies

摘要

The matched trend test (MTT), developed using a conditional logistic regression, has beenproposed to test for association in matched case-control studies to control the bias ofknown confounding effects and reduce the potential impact of population stratification.The MTT requires a known genetic model. When the genetic model is unknown, a MonteCarlo robust test, MAX, has been proposed for the analysis of matched case-control studies.The MAX statistic takes the maximum of three MTTs optimal for three common geneticmodels. We derive the asymptotic power for MTTs and robust tests. m particular, we derivethe asymptotic p-value for MAX. Using these analytical results, we conduct simulationstudies to compare the performance of MAX and the two-degree-of-freedom Chi-squaretest for matched case-control studies, where the latter is implemented in most computingsoftware. Our simulation results show that MAX is always asymptotically more powerfulthan the two-degree-of-freedom Chi-square test under common genetic models. Ourresults provide guidelines for the analysis of genetic association using matched case-control data. An illustration of our results to a real matched pair case-control etiologic studyof sarcoidosis is given.