Continuous covariate imbalance and conditional power for clinical trial interim analyses

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Statistical Inference

Background:

• Clinical trial analyses often adjust for influential covariates, even with baseline imbalance.
• Covariate adjustment may not always be feasible or desirable, impacting generalizability.
• Baseline covariate imbalance can significantly affect interim and final trial outcomes.

Purpose of the Study:

• To illustrate the impact of influential continuous baseline covariate imbalance on unadjusted conditional power (CP).
• To assess the effect of covariate imbalance on clinical trial decisions, particularly futility stopping bounds.
• To evaluate the robustness of this relationship across different covariate distributions.

Main Methods:

• Simulated clinical trial data with varying continuous baseline covariate distributions (normal, skewed, bimodal).
• Calculated unadjusted conditional power (CP) under different scenarios of covariate imbalance.
• Assessed the influence of imbalance on decisions related to futility stopping rules.

Main Results:

• Influential continuous baseline covariate imbalance substantially affects unadjusted conditional power (CP).
• The relationship between imbalance and CP holds for normal, skewed, and bimodal covariates.
• Unadjusted CP calculations are sensitive to covariate imbalance, influencing trial decision-making.

Conclusions:

• Unadjusted conditional power requires careful interpretation when influential covariate imbalance is present.
• Trialists must consider the potential impact of baseline imbalance on statistical analyses and decisions.
• The findings underscore the importance of evaluating covariate effects in clinical trial planning and analysis.