Examining the stability of genetic risk effect as evidence accumulates in the context of meta-analysis

The recursive cumulative meta-analysis (RCM) of genetic association studies explores the relative change of the cumulative risk effect (e.g. OR) in time, indicating the stability of risk effect as evidence accumulates. However, the stability in risk effect is currently evaluated empirically with a graphical approach. A Monte Carlo permutation test for examining the instability in RCM is proposed. The statistic used is a function of the difference between the observed change in risk effect and the expected change, and is expressed (stepwise) cumulatively from the last published GAS to the first one. The permutation method is based on the individual studies and the number of studies in each time step. The test was demonstrated using data from two large scale meta-analyses of GAS. The performance of the test was also explored by simulating data from meta-analyses with different settings in terms of heterogeneity and significance. Significance instability was detected when wide oscillations in risk effect were presented and vice versa. The proposed test for assessing stability may provide the framework for claiming or denying the existence of an association as evidence accumulates.