That's an excellent and crucial starting point for analyzing RCT data according to CONSORT! Intention-to-Treat (ITT) analysis is the gold standard for preserving the benefits of randomization and providing an unbiased estimate of a treatment's effectiveness in a real-world setting.

Let's break down how to conduct ITT analysis, focusing on "including all randomized participants."

1) Conducting Intention-to-Treat (ITT) Analysis Including All Randomized Participants

The core principle of ITT is "Once randomized, always analyzed." This means every participant's data is analyzed according to the group they were originally assigned to at randomization, regardless of what actually happened during the study (e.g., whether they received the assigned intervention, switched treatments, dropped out, or adhered to the protocol).

Why ITT is Crucial (and CONSORT-Compliant):

• Preserves Randomization: It maintains the balance of confounding variables between groups established by randomization.


• Minimizes Bias: By analyzing participants as randomized, it prevents selection bias that can occur if participants who don't complete treatment or switch groups are excluded.


• Reflects Real-World Effectiveness: It accounts for non-adherence, crossovers, and withdrawals, which are common in clinical practice, thereby providing a more pragmatic estimate of the treatment's effect.


• More Conservative Estimate: ITT analysis often yields a more conservative (smaller) estimate of the treatment effect compared to a per-protocol analysis, which is generally preferred to avoid overstating efficacy.

How to Conduct ITT Analysis - Step-by-Step

The biggest challenge in ITT analysis, and what "including all randomized participants" truly implies, is dealing with missing data. If a participant drops out, refuses treatment, or misses follow-up appointments, you won't have complete outcome data for them. Simply excluding them violates the ITT principle and introduces bias.

Step 1: Define the Analysis Population

• All randomized participants: Your ITT population consists of every single participant who was randomized into the trial. This is your denominator.

Step 2: Assign Participants to Their Randomized Group

• Each participant's data is categorized based on the treatment arm they were originally assigned to at randomization. This is irrespective of:


• Whether they actually received the treatment.


• Whether they received the full dose or duration.


• Whether they crossed over to the other treatment arm.


• Whether they withdrew consent or dropped out.

Step 3: Crucially - Address Missing Outcome Data

This is the most critical and often complex step to ensure "all randomized participants" are truly included. You cannot simply exclude participants with missing outcome data, as this converts your analysis into a "complete case analysis," which is a form of per-protocol analysis and can introduce significant bias.

Here are common strategies for handling missing data in ITT, ordered by general preference:

• Multiple Imputation (MI):


• Description: This is generally considered the gold standard for handling missing data in ITIT analysis. MI involves creating multiple complete datasets by imputing (estimating) the missing values several times, using statistical models based on the observed data and covariates (including baseline characteristics, the assigned treatment group, and other predictors of outcome). Each imputed dataset is then analyzed separately, and the results are combined using specific rules (Rubin's rules) to produce a single, robust estimate and standard error that accounts for the uncertainty of imputation.


• Advantages: Provides valid inferences under the Missing At Random (MAR) assumption, incorporates uncertainty, uses all available information.


• Disadvantages: Can be computationally intensive, requires statistical software proficiency.


• Recommendation: Strongly preferred if sufficient data is available for imputation.


• Maximum Likelihood (ML) / Expectation-Maximization (EM) Algorithm:


• Description: These methods are used directly within some statistical models (e.g., mixed models, structural equation modeling). They estimate parameters by directly maximizing the likelihood function based on the observed data, implicitly accounting for missing values without explicitly filling them in.


• Advantages: Also provides valid inferences under MAR, can be more straightforward than MI for certain models.


• Disadvantages: Less flexible than MI across different types of analyses, can be complex to set up.


• Recommendation: Excellent for longitudinal data with repeated measures (e.g., using Linear Mixed Models).


• Last Observation Carried Forward (LOCF):


• Description: For participants with missing post-baseline data, their last observed value is "carried forward" and used as their final outcome.


• Advantages: Simple to implement.


• Disadvantages: Highly problematic and generally discouraged. It assumes that a participant's condition remains unchanged after dropout, which is often clinically implausible and can lead to biased estimates (usually towards no effect or even an exaggerated effect, depending on the outcome and when dropouts occur). It tends to underestimate the variance.


• Recommendation: Only use for sensitivity analysis to see how robust your primary results (e.g., from MI) are under a very conservative assumption, but never as the primary ITT method.


• Baseline Observation Carried Forward (BOCF):


• Description: For participants with missing post-baseline data, their baseline value is carried forward and used as their final outcome.


• Advantages: Simple to implement.


• Disadvantages: Even more problematic than LOCF. It assumes no change from baseline, which is usually a very strong and unrealistic assumption, potentially biasing results towards the null.


• Recommendation: Similar to LOCF, only for very specific sensitivity analyses (e.g., if you assume people who drop out have no improvement or even worsen to baseline levels), but not as a primary method.


• Inverse Probability Weighting (IPW):


• Description: This method weights observed data points by the inverse of their probability of being observed, effectively giving more weight to participants who are similar to those who dropped out.


• Advantages: Can provide unbiased estimates under MAR, even when the proportion of missing data is high.


• Disadvantages: Can be complex, requires careful modeling of the probability of missingness.


• Complete Case Analysis (CCA) / Listwise Deletion:


• Description: This method simply excludes any participant who has missing data for any variable included in the analysis.


• Advantages: Simplest to perform.


• Disadvantages: Strongly violates the ITT principle. Introduces bias unless data are Missing Completely At Random (MCAR), which is rarely the case in RCTs. Reduces statistical power.


• Recommendation: Avoid as a primary ITT analysis. Only use for comparison in sensitivity analyses, or if the proportion of missing data is extremely small (<1-2%) and demonstrably MCAR.

Step 4: Perform the Statistical Analysis

Once you have handled missing data appropriately (e.g., using MI to create complete datasets, or using ML directly in your model), you apply your chosen statistical test or model to compare outcomes between the randomized groups.

• For Continuous Outcomes:


• t-test or ANOVA: For simple comparisons of means.


• Linear Regression or ANCOVA (Analysis of Covariance): ANCOVA is often preferred as it can adjust for baseline values of the outcome and other important covariates, which increases statistical power and precision, even when randomization was successful.


• Linear Mixed Models: Excellent for longitudinal data with repeated measures, inherently handles missing data under MAR assumptions.


• For Binary/Categorical Outcomes:


• Chi-squared test or Fisher's Exact Test: For comparing proportions.


• Logistic Regression: To estimate odds ratios or risk ratios, adjusting for covariates.


• For Time-to-Event Outcomes:


• Kaplan-Meier survival curves with Log-rank test: For comparing survival distributions.


• Cox Proportional Hazards Regression: To estimate hazard ratios, adjusting for covariates.

Step 5: Interpretation of Results

• The results of an ITT analysis provide an estimate of the effectiveness of the intervention under real-world conditions (i.e., when some participants may not fully adhere).


• Be prepared that the treatment effect observed in ITT analysis might be smaller than what a per-protocol analysis would show, precisely because it accounts for non-adherence and dropouts. This is generally considered a more realistic and conservative estimate.

CONSORT Reporting Requirements for ITT

CONSORT guidelines strongly advocate for ITT analysis and require clear reporting of how it was conducted:

• CONSORT Flow Diagram:


• Crucially, the flow diagram must show the number of participants randomized to each group, the number receiving allocated intervention, the number completing intervention, the number lost to follow-up, and the number included in the primary analysis (which should ideally be all randomized for ITT).


• It clearly illustrates the "all randomized" principle.


• Methods Section:


• Explicitly state that an Intention-to-Treat (ITT) analysis was performed as the primary analysis.


• Clearly describe how missing data were handled. Specify the method used (e.g., "Missing data were handled using multiple imputation, generating 50 imputed datasets based on [list variables used for imputation, e.g., baseline characteristics, treatment arm, observed outcome data].").


• Mention any sensitivity analyses performed to assess the robustness of the ITT results to different missing data assumptions (e.g., "Sensitivity analyses were conducted using complete case analysis and Last Observation Carried Forward imputation to assess the robustness of findings.").


• Results Section:


• Present the results of the ITT analysis as the primary findings.


• Report the number of participants included in the primary ITT analysis (ideally, all randomized participants with imputed data).


• If sensitivity analyses were performed, summarize their results.

In summary, to conduct a robust ITT analysis "including all randomized participants":

• **Define your analysis set as all randomized participants.**


• **Assign each participant to their randomized group.**


• Address missing outcome data using appropriate methods, with Multiple Imputation (MI) or Maximum Likelihood (ML) being the preferred choices. Avoid simply excluding participants.


• Perform your chosen statistical analysis on this complete (or imputed) dataset.


• Clearly report your ITT methodology and how missing data were handled in accordance with CONSORT guidelines.