Intention to treat and per protocol analyses: differences and similarities.

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Epidemiology

Background:

• Randomized trials employ explanatory or pragmatic approaches.
• Explanatory trials assess efficacy under optimal conditions, often using per-protocol (PP) analysis.
• Pragmatic trials assess effectiveness in real-world settings, prioritizing intention-to-treat (ITT) analysis.

Purpose of the Study:

• To elucidate the distinct objectives, strengths, and limitations of intention-to-treat (ITT) and per-protocol (PP) analyses in randomized trials.
• To highlight how adherence impacts the validity and interpretation of both ITT and PP analyses.
• To introduce advanced statistical methods for addressing biases in trial analysis.

Main Methods:

• Discussion of the conceptual underpinnings of ITT and PP analyses.
• Examination of how adherence and non-adherence influence trial outcomes and interpretation.
• Introduction of generalized methods (g-methods) like inverse probability weighting.

Main Results:

• PP analysis risks losing randomization benefits and introducing post-randomization confounding due to non-adherence.
• ITT analysis may dilute treatment effects and face applicability issues with varying adherence.
• Both ITT and PP analyses are susceptible to selection bias from missing data.

Conclusions:

• ITT and PP analyses are complementary but have inherent limitations influenced by adherence.
• Post-randomization confounding and selection bias can impact trial validity.
• Advanced statistical methods like g-methods can mitigate biases in both ITT and PP analyses.