Exploring the performance of during-treatment substance use outcome measures in predicting longer-term psychosocial functioning and post-treatment abstinence.

Abstract

The selection of appropriate efficacy endpoints in clinical trials has been a long-standing challenge for the substance use disorder field. Using data from a large, multi-site National Drug Abuse Treatment Clinical Trials Network trial (CTN-0044; n=474), this secondary data analysis aimed to explore whether specific proximal (during-treatment) substance use outcome measures predict longer-term improvements in psychosocial functioning and post-treatment abstinence, and whether predictions vary depending on the specific substance (cannabis, cocaine/stimulants, opioids, and alcohol). Generalized linear mixed models examined associations between six during-treatment substance use outcome measures and social functioning impairment (Social Adjustment Scale Self-Report) and severity of psychiatric symptoms (Brief Symptom Inventory-18) at end-of-treatment, and 3- and 6-months after treatment as well as post-treatment abstinence. Maximum days of consecutive abstinence, proportion of days abstinent, ≥3 weeks of continuous abstinence, and the proportion of urine specimens negative for the primary substance were associated with post-treatment psychiatric and social functioning improvement and abstinence. However, only the effects of abstinence during the last 4 weeks of the treatment period on all three post-treatment outcomes was stable over time and did not differ between primary substance groups. In contrast, complete abstinence during the 12-week treatment period was not consistently associated with functioning improvements. Substance use outcome measures capturing the duration of primary substance abstinence during treatment are suitable predictors of post-treatment abstinence and longer-term psychosocial functioning improvement. Binary outcomes, such as end-of-treatment abstinence, may be particularly stable predictors and attractive given their ease of computation and straightforward clinical interpretability.