Per-protocol

Per-protocol analysis includes only subjects who completed treatment without protocol deviations. May overestimate efficacy. Used as secondary or sensitivity analysis, never primary in confirmatory trials.