Longitudinal missing data strategies for substance use clinical trials using generalized estimating equations: an...

Area of Science:

• Clinical Trials Methodology
• Biostatistics
• Substance Use Research

Background:

• Longitudinal substance use clinical trials frequently encounter missing data.
• Sub-optimal methods for managing missing information are prevalent, potentially biasing results.
• Accurate analysis of treatment effects requires appropriate handling of missing data.

Purpose of the Study:

• To compare the effectiveness of different methods for handling missing data in substance use trials.
• To evaluate the impact of buprenorphine/naloxone tapering schedules on urine analysis positivity.
• To assess the utility of multiple imputation (MI) in analyzing longitudinal substance use data.

Main Methods:

• Utilized listwise deletion, positive urine analysis (UA) imputation, and multiple imputation (MI).
• Employed generalized estimating equations (GEE) to analyze the probability of a positive UA (UA+) over a 4-week period.
• Compared outcomes based on baseline substance use and a 7-day versus 28-day buprenorphine/naloxone tapering schedule.

Main Results:

• Listwise deletion and positive UA imputation models indicated that a 28-day taper group was less likely to have a positive opioid UA (OR=0.57 and OR=0.43, respectively).
• The multiple imputation (MI) model also showed a similar effect for the 28-day taper group (OR=0.57).
• Effect sizes across methods were comparable, suggesting MI aligns with traditional approaches.

Conclusions:

• Multiple imputation (MI) is a valuable analytic strategy for longitudinal substance use data when data are missing at random.
• Combining MI with generalized estimating equations (GEE) provides a robust analytical framework.
• Researchers should consider MI for improved accuracy in substance use clinical trial analyses.