Estimating statistical power for open-enrollment group treatment trials

Area of Science:

• Clinical Psychology
• Psychiatric Epidemiology
• Biostatistics

Background:

• Substance abuse treatment trials often use closed-enrollment designs, which do not reflect real-world open-enrollment group therapy settings.
• Group membership turnover in open-enrollment settings presents a significant analytical challenge, creating a disconnect between clinical practice and research designs.
• Latent class pattern mixture models (LCPMMs) are advanced statistical tools suitable for analyzing longitudinal data with complex membership dynamics.

Purpose of the Study:

• To illustrate a method for conducting power analyses specifically for open-enrollment substance abuse treatment trial designs.
• To demonstrate the application of Latent Class Pattern Mixture Models (LCPMMs) in power analyses for trials with participant turnover.
• To address the methodological discrepancies between traditional power analysis assumptions and the proposed use of LCPMMs in open-enrollment trial settings.

Main Methods:

• Utilized Monte Carlo simulation to perform power analyses for open-enrollment designs.
• Employed Latent Class Pattern Mixture Models (LCPMMs) to account for group membership turnover.
• Derived model parameters from published data of a randomized controlled trial comparing Seeking Safety to Community Care for women with comorbid PTSD and SUD.

Main Results:

• The study provides a practical framework for power analysis in open-enrollment treatment trials.
• The proposed approach using LCPMMs can accommodate the complexities of participant turnover inherent in real-world treatment settings.
• The example highlights how LCPMM-based power analyses can better align with the analytical strategies intended for open-enrollment trial data.

Conclusions:

• LCPMMs are a viable and recommended approach for analyzing data from open-enrollment substance abuse treatment groups with turnover.
• Implementing LCPMM-based power analyses can enhance the design and interpretation of future substance abuse treatment trials.
• This methodological advancement supports research that more accurately reflects clinical practice in behavioral treatments.