USMLE Step 2 CK practice question: Intention-to-treat analysis preserves randomization in clinical trials
USMLE Step 2 CKBiostatisticsEpidemiology
Which of the following analytic approaches best preserves the comparability of the groups created by randomization?
- Analyze only participants who took the assigned treatment as directed
- Analyze participants according to the treatment they actually received
- Exclude participants who stopped treatment or changed therapy
- Analyze all participants in the groups to which they were randomized
- Compare adherent drug users with a matched cohort of nonusers
Show answer and explanation
Correct answer: D. Analyze all participants in the groups to which they were randomized
Randomization balances known and unknown confounders between groups, but only if the groups are analyzed as randomized. In an intention-to-treat (ITT) analysis, every participant is analyzed in the group to which they were assigned, regardless of adherence, crossover, or withdrawal. Because people who stop a drug often differ systematically from those who continue (here they were older and sicker), removing or reassigning them reintroduces selection bias and confounding. ITT also mirrors real-world effectiveness, in which some patients do not take or tolerate their medication.
ITT is conservative: nonadherence and crossover dilute the difference between groups and bias results toward the null. In superiority trials this protects against falsely claiming a benefit, but in noninferiority trials a bias toward the null can falsely suggest equivalence, so per-protocol results are reported alongside ITT. Per-protocol analysis includes only participants who adhered to their assigned treatment, and as-treated analysis groups participants by the treatment actually received; both estimate efficacy under ideal adherence but lose the protection of randomization.
Why the other options are wrong
- A. Analyze only participants who took the assigned treatment as directed — A per-protocol analysis estimates efficacy under ideal adherence but loses the balance of randomization, because adherers differ from nonadherers.
- B. Analyze participants according to the treatment they actually received — An as-treated analysis regroups participants by the treatment received, breaking randomization and reintroducing confounding.
- C. Exclude participants who stopped treatment or changed therapy — Excluding nonadherent participants creates selection bias, here removing older, sicker people mainly from one group.
- E. Compare adherent drug users with a matched cohort of nonusers — Comparing self-selected adherent users with matched nonusers is an observational design that remains subject to confounding.
Key point
Intention-to-treat analysis, which analyzes all participants in their randomized groups regardless of adherence, preserves the benefits of randomization.
Memory hook
Once randomized, always analyzed as randomized. ITT keeps the coin toss honest; per-protocol trades it for ideal-world efficacy.
15 free bank questions a day, exam-length timed blocks, spaced repetition and 3D anatomy — in English and Russian.
Play free