A clinical trial investigating the effectiveness of a new tuberculosis regimen in Indian patients is halfway through enrollment. The data safety monitoring board (DSMB) has identified a significant imbalance in baseline characteristics between the intervention and control groups despite randomization. Which investigation/procedure is most appropriate to assess whether this imbalance could have occurred by chance alone?

Explanation

## Baseline Imbalance Assessment in RCTs **Key Point:** When baseline imbalances are observed in a randomized trial, the investigator must first *test* whether the imbalance is statistically significant or merely due to random variation. This is done using hypothesis tests on baseline characteristics. ### Statistical Testing of Baseline Imbalance **High-Yield:** Baseline imbalances are **expected in small samples** due to random allocation. The question is: Is this imbalance statistically significant? #### Appropriate Tests by Variable Type | Baseline Variable Type | Test | Purpose | |------------------------|------|----------| | Continuous (age, BP) | Independent samples t-test | Compare means between groups | | Categorical (sex, stage) | Chi-square test or Fisher's exact | Compare proportions between groups | | Multiple variables | ANOVA or logistic regression | Assess overall imbalance | **Clinical Pearl:** Small imbalances in baseline characteristics are *not* a threat to validity if randomization was truly performed. The DSMB should assess whether the imbalance is: 1. **Statistically significant** (unlikely to occur by chance) 2. **Clinically meaningful** (likely to bias the outcome) If neither condition is met, the trial proceeds without modification. ### Why This is the "Investigation of Choice" Statistical testing allows the DSMB to: - Quantify the probability that the imbalance occurred by random allocation alone - Decide whether adjustment (e.g., stratified analysis) is needed in the final analysis - Determine if the trial's validity is compromised **Mnemonic:** **TEST-THEN-ACT** — **T**est baseline balance → **E**valuate significance → **S**tratify if needed → **T**hen proceed → **A**djust analysis if necessary → **C**ontinue trial → **T**rust randomization.