Toward a Generalized Framework and Flexible Software Environment for Power Analysis of Alcohol Treatment Randomized Controlled Trials

点击翻译按钮获取中文摘要

PROJECT SUMMARY Pharmacological and behavioral treatments of alcohol use disorder are only useful to the extent that their effects are robust and reliable. A priori determination of statistical power is one of the most significant steps to maximizing scientific reliability; however, it has historically been largely overlooked in the health sciences. Alcohol researchers are increasingly utilizing complex statistical models (e.g., growth curves, mixture models, social networks); yet, accessible methods for sample size determination for all but the simplest statistical tests have not been broadly disseminated. Moreover, even for models in which software programs are readily available, scientists consistently express uncertainty regarding expected parameter estimates, prohibiting the utility of power analysis itself. As a result, many randomized controlled trials (RCTs) are likely not optimally designed to allow for conclusive statistical inference. The overall objective of this proposal is to close this gap by developing a generalized method for power analysis using a combination of established and innovative approaches that can be used to generate and test a vast array of models used in treatment research. This objective will be achieved by pursuing two specific aims: develop a flexible, user-friendly software interface for statistical power and sensitivity analysis (Aim 1); and develop a generalized framework for determining model parameter estimates (Aim 2). The proposal is innovative because the software interface will be grounded in state-of-the-art advancements in computational technology and human-computer interaction. In addition, the framework will be fortified by advances in generalizability-theory-based meta-analyses and guided walk and machine-learning-based simulation studies investigating the effects of uncertainty in individual model parameters on the likelihood of observing hypothesized effects. The software will include automated sensitivity analyses to identify parameters with the largest impact on power and will recommend starting parameters. Development and testing will be grounded in the naltrexone and cognitive behavioral therapy literatures. Their widespread use and diversity of methodological approaches make them ideal testbeds from which to generalize to other treatment domains. This research is synergistic with each of the goals within the NIAAA Strategic Plan (2017-2021). Perhaps most significantly, methodology and guidelines will be developed for designing RCTs that have sufficient precision to distinguish trajectories, mechanisms, heterogeneity, and specificity of treatment efficacy. These efforts will provide researchers with the required a priori information to execute studies that ensure the robustness of their results. The development of these tools for RCT design will have implications for all areas of alcohol treatment research, including in-the-moment risk behaviors, person- level (comorbid) characteristics, targeting underserved and vulnerable populations, and large-scale initiatives. As a result, this proposal will allow for more definitive identification of efficacious interventions and treatments of alcohol use disorder.