Abstract:
|
In randomized trials with time-to-event outcomes and their meta-analyses, the difference in restricted mean survival times (RMST) is a useful measure of the treatment effect. However, the difference in RMST depends upon the time horizon and the randomized trials included in a meta-analysis frequently use different follow-up durations. A fixed time horizon chosen for the meta-analysis may be larger than the follow-up duration in some trials, and smaller in others, leading to the exclusion of data. Here, we describe fixed-effect and random-effects methods for multivariate meta-analysis of difference in RMST at multiple follow-up durations. The models allow using all available data and some borrowing of strength across follow-up times. We show how to calculate the covariance of the differences in RMST at successive time points. We illustrate the models for the comparison of transcatheter aortic valve implantation against surgical aortic valve replacement in patients with aortic stenosis.
|