What is Intention-to-Treat Analysis in Statistics?
Intention-to-treat (ITT) analysis is a statistical method in randomized controlled trials where all participants are analyzed according to their original group assignment at randomization, regardless of whether they actually received the assigned treatment, dropped out, violated the protocol, or had any other deviation from the study plan. 1
Core Principle
The fundamental concept is straightforward: analyze participants in the groups to which they were randomly assigned, not based on the treatment they actually received. 2, 3 This approach preserves the critical benefit of randomization—the avoidance of bias when allocating interventions to trial participants. 1
Why ITT Analysis is Preferred
ITT analysis is generally favored because it avoids bias associated with non-random loss of participants. 1 The method maintains the prognostic balance generated from the original random treatment allocation and provides an unbiased estimate of treatment effectiveness at the level of adherence observed in the study. 3, 4
Key Advantages:
- Preserves randomization benefits that eliminate selection bias 3
- Provides pragmatic evaluation of treatment benefit in real-world conditions, not just ideal circumstances 2
- Prevents bias from excluding participants who don't adhere to protocol 1
Contrast with Per-Protocol Analysis
The alternative approach—"per-protocol" or "on-treatment" analysis—restricts analysis only to participants who fulfilled the protocol in terms of eligibility, interventions, and outcome assessment. 1 Excluding participants from analysis can lead to erroneous conclusions. 1 For example, in a trial comparing medical versus surgical therapy for carotid stenosis, per-protocol analysis showed surgery reduced risk of stroke and death, but ITT analysis based on all participants as originally assigned did not show superior effect of surgery. 1
Implementation Challenges
Missing Data Problem
Strict ITT analysis is often difficult to achieve because patients may drop out or be lost to follow-up. 1, 4 When outcome data are missing:
- For mortality outcomes, register information may allow inclusion 1
- For other outcomes, imputation techniques may be needed 1
- "Complete case" analysis (including only those with known outcomes) loses statistical power and may introduce bias if dropout is related to treatment response 1
Common Misuse
The term "intention-to-treat analysis" is often inappropriately used. 1 For example, excluding patients who did not receive the first dose of trial drug violates ITT principles. 1 In a review of 403 RCTs in major medical journals, 62% reported using ITT analysis, but only 39% actually analyzed all participants as randomized, with more than 60% having missing data in their primary analysis. 1
Critical Caveats
- ITT analysis may underestimate potential treatment benefit when there is significant non-compliance with assigned therapy 1
- The estimate of treatment effect in ITT analysis is generally conservative 4
- Simple imputation methods like "last observation carried forward" are widely criticized and may introduce bias without accounting for uncertainty 1
- Authors frequently mislabel their analyses as ITT when they have excluded randomized participants, making it difficult to identify true ITT analyses for meta-analyses 5, 6
Reporting Standards
According to CONSORT guidelines, authors should clearly state which and how many participants are included in each analysis, regardless of whether they use the term "intention-to-treat." 1 The number of participants per group should be given for all analyses, and results should be expressed as fractions to help readers assess whether randomized participants were excluded. 1