All Times EDT
Keywords: Estimand, Intercurrent event, Missing data, Sample size
Missing data are not intercurrent events; missing data are consequences of intercurrent events. When planning clinical trials with potential missing data, a conventional approach is two-stage: (1) calculate a required sample size N without considering missing data; (2) adjust the sample size using N*=N/(1-r), where r is the expected missing data rate. However, this approach is not estimand oriented. Clinical trial design should be aligned with estimand. Sample size calculation is a key step in clinical trial design, so methods for sample size calculation should be aligned with estimand. ICH E9(R1) summarizes five strategies for dealing with intercurrent events. We consider five basic approaches for sample size calculation when planning clinical trials with intercurrent events, with each approach aligned with one of these five strategies. We extend the approaches to scenarios where some combination of multiple strategies is applied to deal with intercurrent events. Being aligned with estimands and strategies for dealing with intercurrent events, these methods can be used for sample size calculations when planning clinical trials with intercurrent events.