Toward Identifying Clinically Useful Definitions of Treatment Success
Treatments for alcohol use disorder (AUD) have been historically evaluated by consumption-based outcomes, such as the percentage of days abstinent from alcohol (PDA). There are many advantages of alcohol consumption as a primary outcome, given that it is easily quantified (e.g., via standard drinks) and because alcohol consumption is inherently associated with AUD. However, researchers, clients, and treatment providers have all agreed that the process of recovering from AUD involves more than consumption alone (e.g., Donovan et al., 2005; Kaskutas et al., 2014; Neale et al., 2015). Recent efforts have called for a shift away from defining consumption as the marker of AUD treatment success (Witkiewitz, 2013a), given that defining treatment success primarily by amount of consumption often fails to adequately portray the recovery process (e.g., Donovan et al., 2012; Substance Abuse and Mental Health Services Administration, 2012; Tiffany et al., 2012). Yet, efforts to move toward including non-consumption outcome measures have been stymied for two primary reasons. First, it is assumed that consumption measures are more sensitive and, therefore, consumption must be used as a "surrogate" measure for more clinically meaningful non-consumption measures (FDA, 2015, p. 2). Second, research is needed to direct researchers toward the best "gold standard" measures to use among AUD populations and studies of measurement invariance over time of these measures are largely non-existent. The aims of the proposed study will address these two concerns via secondary data analysis of two of the largest AUD treatment research studies conducted in the United States: the COMBINE Study and Project MATCH (Anton et al., 2006; Project MATCH Research Group, 1997). The present study will examine the sensitivity and specificity (Aim 1), as well as the reliability, validity, and measurement invariance (Aim 2), of non-consumption measures available in the COMBINE study and Project MATCH. The non-consumption outcome variables to be examined are consistent with the variables identified as important to researchers, clients and their loved ones, and treatment providers: 1) alcohol-related problems, 2) mental health, 3) craving/temptation, 4) psychosocial functioning/quality of life, 5) physical health, 6) financial status, and 7) self-efficacy. Sensitivity and specificity will be evaluated using Receiver Operating Characteristic (ROC) curve analyses (Hanley & McNeil, 1982) predicting outcome data from immediately post- and 12-month post-treatment for: 1) abstinence, 2) no heavy drinking days (Falk et al., 2010), 3) the World Health Organization risky drinking levels (European Medicines Agency, 2010), and 4) a composite clinical outcome defined by consumption and alcohol-related problems (Cisler & Zweben, 1999). Reliability, validity, and measurement invariance will be assessed using latent variable modeling approaches. The two aims of this study will address longstanding and untested assumptions that have prevented researchers from shifting toward more clinically meaningful definitions of treatment success that include non-consumption outcomes.