To understand ITT analysis more deeply, let’s compare it with other common methods of analysis: per-protocol (PP) and as-treated (AT).

1. Intention-to-Treat (ITT) Analysis

Definition:

Includes all participants as originally randomized, regardless of adherence to the protocol.

Key Characteristics:

• Preserves the benefits of randomization.
• Reflects the real-world effectiveness of the intervention.
• Handles deviations such as noncompliance or dropout by including all participants in their assigned groups.

Strengths:

• Minimizes bias and maintains the initial comparability of groups.
• Provides conservative estimates of treatment effects.

Weaknesses:

• Treatment effects may be diluted by noncompliance or protocol deviations.
• Requires robust methods to handle missing data (e.g., imputation).

Example:

In a diabetes medication trial, 100 participants are randomized to the treatment group, and 100 to a placebo group. If 10 in each group fail to complete the study, ITT analysis still includes all 200 participants based on their original assignment.

2. Per-Protocol (PP) Analysis

Definition:

Includes only participants who fully adhere to the intervention protocol.

Key Characteristics:

• Excludes noncompliant participants, crossovers, and those with missing data.
• Reflects the efficacy of the treatment under ideal conditions.

Strengths:

• Provides a clear picture of the treatment’s potential efficacy.
• Useful for secondary or exploratory analyses.

Weaknesses:

• Loses the balance achieved by randomization.
• Introduces selection bias, as excluded participants may differ systematically from those included.

Example:

In the same diabetes trial, only the 90 participants in each group who adhered to their assigned regimen are analyzed, ignoring those who dropped out or did not follow the protocol.

3. As-Treated (AT) Analysis

Definition:

Analyzes participants based on the treatment they actually received, regardless of their randomization.

Key Characteristics:

• Focuses on the actual exposure to the treatment.
• Groups may differ due to self-selection or other biases.

Strengths:

• Can help assess treatment effects based on actual behavior.
• Useful when there is significant crossover between groups.

Weaknesses:

• Completely breaks randomization.
• Highly susceptible to confounding and bias.

Example:

If participants in the diabetes trial switched from the treatment to the placebo group (or vice versa), AT analysis evaluates outcomes based on what they actually received rather than their assigned group.

Comparing the Approaches

Feature	ITT	PP	AT
Randomization Preserved	Yes	No	No
Reflects Real-World Use	Yes	No	Sometimes
Accounts for Adherence	No	Yes	Yes
Bias Potential	Low	Moderate	High
Estimates Effectiveness or Efficacy	Effectiveness	Efficacy	Neither (often unclear)

When to Use ITT vs. Other Approaches

1. Use ITT Analysis When:
   • The goal is to estimate real-world effectiveness.
   • Maintaining the randomization benefits is crucial.
   • The primary outcome needs to reflect all assigned participants, even with deviations.
2. Use PP Analysis When:
   • Assessing the efficacy of an intervention under perfect adherence.
   • Performing secondary or exploratory analyses.
3. Use AT Analysis When:
   • Investigating treatment effects based on what participants actually received.
   • Understanding the practical implications of treatment switches or deviations.

Illustrative Example: COVID-19 Vaccine Trial

Scenario:

A trial enrolls 10,000 participants, randomizing them equally to a vaccine or placebo group. Over time:

• 500 vaccine group participants do not return for follow-up.
• 300 placebo group participants receive the vaccine mid-trial.

Analysis Approaches:

1. ITT Analysis:
   • Analyzes all participants in their original groups (vaccine vs. placebo).
   • Accounts for real-world issues like noncompliance and crossover.
2. PP Analysis:
   • Includes only participants who followed the protocol (e.g., excludes the 500 vaccine dropouts and the 300 placebo crossovers).
   • Estimates the vaccine’s efficacy under ideal adherence.
3. AT Analysis:
   • Groups participants by what they actually received (e.g., includes the 300 crossovers in the vaccine group).
   • Risks bias because those who switched treatments may differ systematically from those who did not.

Conclusion

• ITT Analysis: Preferred for primary outcomes, ensuring unbiased, generalizable results.
• PP and AT Analyses: Useful as supplementary methods to understand treatment efficacy and practical impacts but should not replace ITT as the main analysis.