Intention-to-treat

The principle is once randomized, always analyzed. Excluding people who stopped treatment, switched arms, or were non-compliant breaks the balance that randomization created, because those departures are rarely random. Keeping everyone in their original group maintains comparable arms and yields an estimate that reflects real-world use, where people do not always adhere.

The alternative, per-protocol analysis, includes only those who followed the regimen and estimates the effect under perfect adherence. It answers a different question and is more vulnerable to bias, so ITT is the primary analysis for superiority trials while per-protocol serves as a sensitivity check. Non-inferiority trials complicate this, since ITT can bias toward declaring equivalence.

Reviewers should confirm that the primary analysis is ITT, that all randomized participants are accounted for, and that missing outcome data are handled with appropriate methods rather than simply dropped. A large gap between ITT and per-protocol results signals adherence problems worth scrutinizing.