intention to treat
Summary
Intention-to-treat (ITT) analysis analyzes all randomized participants in the groups to which they were originally assigned, regardless of adherence, crossover, or dropout. It preserves randomization, minimizes selection/attrition bias, and reflects real-world effectiveness, but it tends to bias results toward the null (underestimates efficacy).
Detail
DEFINITION: In ITT analysis, every patient is analyzed according to the group they were randomized to ("analyze as randomized"), even if they never received the treatment, stopped early, crossed over, or were lost to follow-up (outcomes still counted, often via imputation).
WHY IT MATTERS: - Preserves the benefit of randomization: confounders (known and unknown) remain balanced between groups. Excluding non-adherent patients reintroduces selection bias, because adherence is often related to prognosis (healthy adherer effect). - Avoids attrition bias. - Reflects real-world EFFECTIVENESS (includes the effect of non-compliance and side effects) rather than ideal-conditions EFFICACY. - Conservative: non-adherence and crossover dilute the difference between groups, so ITT biases toward the null hypothesis. This lowers the chance of a false-positive (type I error) but may increase the chance of a false-negative (type II error). It is the standard primary analysis for superiority trials.
COMPARISON WITH OTHER ANALYSES: - Per-protocol analysis: includes only patients who completed the protocol as assigned. It estimates efficacy under ideal conditions but is prone to bias, because the groups are no longer truly randomized. It is often preferred as a secondary analysis, and it is more appropriate in non-inferiority/equivalence trials, where ITT's bias toward the null could falsely suggest similarity. - As-treated analysis: analyzes patients by the treatment they actually received, regardless of assignment. It breaks randomization and is prone to confounding. - Modified ITT: excludes certain post-randomization patients (e.g., those who never received any dose or were found ineligible after randomization). It can reintroduce bias if applied unevenly.
CLINICAL/BOARD PEARLS: - Classic stem: "Patients who crossed over from the surgery arm to the medical arm were analyzed in the surgery group." The answer is intention-to-treat analysis. - Stem: "Which analysis best preserves randomization and avoids bias from dropouts?" The answer is ITT. - Stem: "Which analysis would be expected to underestimate the true treatment effect?" The answer is ITT, because crossover and non-compliance dilute the effect. - ITT applies to randomized controlled trials. Related concepts include allocation concealment, blinding, and lost to follow-up (handled by imputation, e.g., last observation carried forward, or worst-case scenarios). - Strengths: preserves randomization, minimizes bias, reflects practical effectiveness. Limitations: dilutes the treatment effect, may underestimate the efficacy and harm of the intervention, and requires handling of missing data.
Sources
- First Aid for the USMLE Step 1 – Biostatistics & Epidemiology (Clinical trials, Bias)
- UWorld Step 1/Step 2 CK – Biostatistics: Study design and clinical trials
- Kaplan USMLE Step 1 Lecture Notes: Behavioral Science and Social Sciences/Biostatistics
- Gupta SK. Intention-to-treat concept: A review. Perspect Clin Res. 2011;2(3):109-112
- CONSORT Statement (Consolidated Standards of Reporting Trials)
- Fletcher RH, Fletcher SW. Clinical Epidemiology: The Essentials
Reviewed by Medusa editorial — medical student review. Information here is for study reference only and is not medical advice. Spotted an error? Let us know.