Missing data in substance abuse treatment research: current methods and modern approaches

Area of Science:

• Clinical Trials
• Biostatistics
• Addiction Medicine

Background:

• Missing data in clinical trials can introduce significant bias in treatment effect interpretation.
• Common methods like listwise deletion and single imputation (e.g., positive urine analysis imputation) are frequently used but may be problematic.
• Multiple imputation (MI) is a statistically advanced method for handling missing data.

Purpose of the Study:

• To compare the impact of listwise deletion, positive urine analysis (UA) imputation, and multiple imputation (MI) on treatment effect interpretation.
• To evaluate these missing data procedures using data from a clinical trial on opioid addiction treatment.

Main Methods:

• Utilized publicly available data from a clinical trial (Clinical Trial Network 0003) comparing 7-day and 28-day buprenorphine/naloxone tapering schedules.
• Applied listwise deletion, positive UA imputation (scoring missing UA as positive), and MI to handle missing data.
• Examined the effect of tapering schedules on the likelihood of a positive urine analysis (UA).

Main Results:

• Listwise deletion yielded a nonsignificant treatment effect for the buprenorphine/naloxone taper.
• Positive UA imputation replicated original findings, showing a significant treatment effect.
• MI also indicated a significant effect, but with a smaller effect size and larger standard errors compared to positive UA imputation.

Conclusions:

• The method chosen for handling missing data critically influences study outcomes and interpretation of treatment effects.
• Listwise deletion and single imputation (positive UA) are discouraged due to potential biases.
• Multiple imputation (MI) is recommended for its advantages in internal and external validity, provided the 'missing at random' assumption is met.