Psychosocial Treatment Effect with Adult Substance Users: A Meta-Analysis of Cognitive-Behavioral Treatments

Cognitive-behavioral (CB) interventions for adult substance use disorders have been of ongoing interest to evaluation researchers, but the corresponding body of literature has rarely been subjected to quantitative review. Based primarily on Marlatt and Gordon's (1985) model of relapse prevention, these treatments target cognitive, affective, or situational triggers for substance use and skills training related to coping alternatives. Given this range of possible foci, CB treatments may be particularly useful with secondary areas of client functioning. In a meta-analytic review of relapse prevention, psychosocial effect sizes were nearly twice that found for alcohol or other drug use (Irvin et al., 1999), but this outcome remains poorly understood. This study reports secondary psychosocial outcomes from a recent meta-analytic review of CB treatment with substance use disorders (Magill & Ray, under review). The purpose of the current analyses was: (1) to examine the magnitude of psychosocial treatment effect and (2) to test whether this effect was associated with time or reactivity of outcome measurement. The meta-analysis examined 21 qualifying randomized controlled trials. Effect estimates were calculated using Hedges' g and were inverse variance weighted prior to pooling (see Hedges & Olkin, 1985). Analyses tested the possible presence of between study heterogeneity and publication bias, and effect moderation was assessed in fixed effects sub-groups. CB treatments produced a small-moderate effect (see Cohen, 1977) over comparison conditions (g = .310, p < .005), and this effect was larger than that found previously for substance use outcomes (g = .163, p < .005; Magill & Ray, under review). The Q-test was significant, and the model of inference was therefore considered conditionally random(random effects value g = .356, p < .005). The findings support previous research that suggests greater treatment response for measures of secondary functioning compared to alcohol or illicit drug use. However, effect magnitude varied by the two putative moderators considered. Effect sizes were greatest at six to nine months (g = .634, n = 5, Q >.05), followed by posttreatment to four months (g = .236, n = 14, Q < .005), and twelve months outcomes showed a small negative treatment effect (g = -.011, n = 2, Q > .05). Non-reactive measures (e.g., related problems, social support) were smaller (g = .237, p < .05, n = 13, Q < .05) than those considered potentially behaviorally-reactive (i.e., self-efficacy, coping skills; g = .502, p < .005, n = 8, Q < .05). Diagnostic tests did not suggest the presence of publication bias. CB treatments are a viable option for frontline substance use treatment providers and may be particularly effective with secondary areas of client functioning. However, overlap among effect size moderators should be considered when interpreting the present findings and further research should examine whether secondary outcomes are mechanisms of main substance use effects.