. . .an analysis of the external validity of a trial [generalizability of results] is only possible if all eligible patients are followed regardless of their randomization status.