Missing data imputation in two phase III trials treating HIV1 infection.

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• HIV/AIDS Research

Background:

• Longitudinal clinical trials often face patient dropouts, leading to missing data.
• Accurate imputation of missing data is essential for intent-to-treat analyses.
• Existing imputation methods may yield varying results, impacting primary endpoint reliability.

Purpose of the Study:

• To compare standard imputation methods with novel approaches for HIV1-RNA data.
• To assess the robustness of primary endpoint results using different imputation strategies.
• To evaluate imputation method performance on both real and simulated clinical trial data.

Main Methods:

• Compared last-observation-carried-forward, baseline carried forward, and multiple imputation.
• Evaluated a novel nearest-neighbour hot-deck method for HIV1-RNA data imputation.
• Assessed a censored regression analysis of last-observation-carried-forward heuristic.
• Utilized data from two clinical trials of the HIV1 fusion inhibitor enfuvirtide.
• Supplemented analysis with simulated datasets covering diverse missing data patterns.

Main Results:

• Different imputation methods yielded varying results for HIV1-RNA data analysis.
• Sensitivity analyses are critical to confirm the robustness of primary endpoints.
• Novel imputation methods showed potential for specific applications in HIV trials.

Conclusions:

• The choice of imputation method significantly influences clinical trial outcomes.
• Robustness checks using multiple imputation techniques are vital for reliable results.
• Further research into tailored imputation methods for specific data types like HIV1-RNA is warranted.