Modeling longitudinal drinking data in clinical trials: an application to the COMBINE study.

Area of Science:

• Clinical Psychology
• Psychopharmacology
• Biostatistics

Background:

• Current clinical trial methods for assessing drinking behavior lack consensus and fail to capture its complexity.
• Traditional statistical models analyze summary drinking measures separately, limiting a comprehensive understanding of treatment effects.
• A novel statistical approach is needed to analyze high-resolution drinking data effectively.

Purpose of the Study:

• To illustrate a statistical approach for examining treatment effects on high-resolution drinking data using the COMBINE study.
• To evaluate the effectiveness of naltrexone and combined behavioral intervention (CBI) on drinking outcomes.
• To assess the impact of treatments on abstinence and risky drinking patterns.

Main Methods:

• Secondary data analysis of the COMBINE study.
• Participants were randomly assigned to naltrexone, acamprosate, medical management, and/or CBI.
• A Poisson hurdle model was used to simultaneously analyze abstinence and the number of drinks, with a focus on "risky drinking" thresholds.

Main Results:

• Naltrexone significantly increased the odds of abstinence compared to placebo, particularly when not combined with CBI.
• Naltrexone reduced the risk of "risky drinking" and increased the odds of maintaining low-risk drinking.
• These positive effects of naltrexone were most pronounced when only medical management was provided, without CBI.

Conclusions:

• The hurdle model is a suitable statistical tool for analyzing the two-part drinking process (abstinence and number of drinks).
• Findings support the use of naltrexone for enhancing abstinence and reducing risky drinking.
• The interplay between naltrexone, CBI, and medical management in influencing drinking behavior requires careful consideration in treatment strategies.