In previously published studies of this sample, the TSF group had greater 12-step affiliation and greater reductions in depression compared to ICBT during treatment, while substance use outcomes were similar between groups . For the current study we hypothesized that the superior depression outcomes in the TSF group would be mediated by greater 12-step attendance and affiliation. Furthermore, we predicted that the effects of 12-step attendance and affiliation on future alcohol and drug use would be mediated by depressive symptoms. As depression often decreases with initial abstinence , we controlled for past alcohol and drug use in longitudinal models, to test whether the hypothesized relations were independent of prior substance use. The current study involves secondary analyses of 209 veterans who participated in a trial of outpatient group psychotherapy for comorbid SUD-MDD . Demographics of the sample are presented in Table 1. Participants met DSM-IV criteria for: lifetime dependence on alcohol, cannabis, or stimulants with use in the past 90 days, and major depressive disorder with ≥ 1 lifetime episode occurring independent of substance use, assessed via the Composite International Diagnostic Interview . Exclusion criteria included opiate dependence with intravenous administration, bipolar disorder or psychotic disorder, living more than 50 miles from the facility, or severe memory impairment interfering with assessment. The trial was approved by the University of California, San Diego and VA San Diego Healthcare System Institutional Review Boards. Research staff obtained referrals from the VASDHS dual diagnosis clinic, briefly screened individuals for eligibility, and met with eligible veterans to explain study procedures and obtain informed consent. Participants consented to group psychotherapy, video recording of sessions, monthly psychotropic medication visits,vertical grow system random toxicology screens, and research assessments at baseline and every 3 months. Veterans agreed to participate onlyin the assigned form of treatment for the duration of group psychotherapy.
Rates of prescription antidepressant utilization during the group psychotherapy phase were high for both treatment conditions . Group psychotherapy was initiated via a rolling admission procedure, with start dates scheduled every 2 weeks and alternating assignment of patients to the condition with the next start date. Group sessions occurred twice/week for the first 3 months and weekly for the next 3 months. Session attendance was not significantly different across groups. Both interventions were co-delivered by senior clinicians and doctoral students trained to criterion via manual review, direct observation, and weekly supervision. Therapists rotated across treatment conditions every 6-12 months and treatment adherence was assessed via client report of content and videotape review. The TSF and ICBT groups did not differ significantly on the demographic and clinical characteristics assessed at baseline . For the TSF condition we modified TSF from Project MATCH to allow focus on multiple substances and group delivery. The three core modules of the TSF protocol covered Steps 1-3, general Twelve-Step topics and literature, and Steps 4-5. Sessions involved discussions of 12-step readings and recovery tasks . Depression was only discussed in the context of 12-Step themes. For ICBT, material was adapted from two empirically-validated treatments: group cognitive-behavioral treatment of depression and cognitive-behavioral therapy from Project MATCH . The three core modules of ICBT were Thoughts , Activities , and Interpersonal . We employed hierarchical linear models to examine the relations between treatment condition, 12-step variables, depression, and alcohol/drug use in multilevel mediation analyses . The use of HLM allows inclusion of multiple time points nested within individuals and both static and time-varying covariates. By examining associations between sets of time-varying predictors and outcomes, these analyses were akin to running multiple models for multiple time points and averaging their effects.
We included all available data via maximum likelihood estimation, a preferred method of estimation when the data contain information that is assumed missing-at-random . Analyses revealed no differences on any study variables between individuals with complete data and those with any missing data, supporting this assumption. Separate HLMs were conducted in Stata 10.1 to examine individual paths in each mediation model . Covariates for each HLM included treatment group, time, baseline level of outcome, demographics and post-traumatic stress disorder , as these covariates have previously predicted substance use in this sample or others . Repeated variables were treated as time-varying covariates, with PDD and PDDRG as time-varying covariates in the final HLMs to determine if effects were independent of current alcohol and drug use. Finally, in Model 2 all time-varying covariates were lagged, with prior variables predicting future alcohol and drug use.Consistent with current conventions in mediation analysis , formal tests to estimate the magnitude, statistical significance, and effect sizes of mediated effects were computed using products-of coefficients with asymmetric 95% confidence limits, which has more accurate Type I error rates and greater power to detect mediated effects than alternate approaches . Estimates of mediated effects and confidence limits were obtained using the PRODCLIN program by importing coefficients and standard errors from the HLMs. Mediated effect size was reported with the proportion of the direct effect explained by the mediated effect . Separate HLMs examined the effects of 12-step attendance and affiliation on depression at Months 3 and 6, controlling for treatment condition, baseline depression, time, and other covariates . Greater 12-step attendance and greater 12-step affiliation both predicted lower depression, and the effect of treatment condition was no longer statistically significant.
When examining the 12-step variables simultaneously, only 12-step meeting attendance independently predicted depression , thus further analyses did not include 12-step affiliation.A final model included current PDD and PDDRG as time varying predictors of depression, to determine if the effects of 12-step meetings on depression were independent of current alcohol and drug use. Lower PDD and PDDRG both significantly predicted lower depression . Controlling for these effects, greater 12-step meeting attendance still predicted lower depression , indicating that relations between meeting attendance and depression were independent of current drinking and drug use. To assess the statistical and clinical significance of mediation we obtained an estimate of the mediated effect with asymmetric 95% confidence intervals. In Model 1 this refers to the specific portion of the group difference on depression that is mediated through 12-step meeting attendance. These results indicated that the effect of TSF on depression was mediated through 12-step meeting attendance, with the mediated effect explaining 24.3% of the direct effect of TSF on depression. Lagged measures of 12-step attendance at Months 3 and 6 were used to predict future frequency of drinking and drug use, controlling for covariates, none of which significantly predicted drinking or drug use in the subsequent HLMs. As shown in Table 4, greater 12-step meeting attendance significantly predicted lower future PDD but not future PDDRG . This indicated individuals with greater 12-step meeting attendance had lower future drinking frequency,growing cannabis outdoors but 12-step attendance was not significantly related to future drug use. Support for the predictor-to-mediator path in Model 2 was previously established in Model 1. To complete Model 2 analyses, we examined the mediator-to-outcome path by testing whether depression at Month 3 and 6 predicted future alcohol and drug use, controlling for the effects of the predictor . As shown in Table 4, lower lagged depression significantly predicted lower future PDD but not future PDDRG . The effects of 12-step attendance were no longer statistically significant . Given that effects of depression on future drinking could be confounded by current drinking, our final model controlled for the effects of lagged PDD. As shown in Table 4, prior PDD was significantly and strongly related to future PDD , but lagged depression was still significantly predictive of future PDD . In other words, lower month 3 and 6 depression uniquely predicted lower drinking at months 6 and 9, above and beyond the effects of prior drinking. In our original trial of ICBT and TSF for adults with SUD and MDD, the ICBT group was expected to have superior reductions in depression symptoms during treatment, but the results were contrary to these expectations . The current study utilized mediation analyses to examine whether greater engagement in 12-step resources explained lower within-treatment depression for the TSF condition. Greater attendance at community-based 12-step meetings was associated with lower depression and mediated the group difference in depression, providing preliminary evidence for 12-step attendance as a therapeutic mechanism for reductions in depressive symptoms. Others have noted that general therapeutic factors inherent to self-help groups may be especially helpful for SUD patients with psychiatric conditions , and patients in TSF likely gained greater exposure to such therapeutic factors through greater community 12- step meeting attendance. Greater meeting attendance could also signal greater behavioral activation, a component of effective psychotherapy for depression.
While interpersonal and behavioral coping skills were also targeted in ICBT, our findings suggest 12-step meetings could be an important and readily accessible vehicle for general symptom reduction in patients with SUD and MDD. While the ICBT group had less reduction in depression during treatment, our prior report revealed superior post-treatment depressive symptoms and substance use for ICBT , and future studies of this sample may provide insight into the unique vs. common mechanisms that explain differential patterns of symptom change in these alternative treatments.However, our study was unique in finding this effect was mediated by depression, even when controlling for current drinking. In a similar study of Project MATCH, depressive symptoms were not uniquely related to meeting attendance or future drinking when current drinking was controlled . This discrepancy suggests that changes in depressive symptoms linked to 12-step meeting attendance may be an especially critical therapeutic process for patients with comorbid MDD. Other studies suggest negative affect has a large role in the maintenance of substance use for these patients. Individual changes in depression and substance use were highly correlated in this sample , and SUD patients with MDD are more likely to experience depressed mood prior to relapse . The importance of negative affect was also confirmed in recent work with Project MATCH , in which the mediating effects of depression were significant and independent of other process variables in the aftercare sample, who may be more representative of patients enrolled in our study. In our sample 12-step meeting attendance no longer predicted future drinking when accounting for the mediating effects of depressive symptoms. This suggests that even when treatment processes focus explicitly on substance use, ancillary effects on mood may be an important mechanism of change for patients with comorbid MDD. Conversely, patients with depressive symptoms that persist despite continued abstinence and frequent 12-step meeting attendance may need additional interventions to control depression and prevent future relapse. Two aspects of our hypotheses were not confirmed and merit discussion. When examined concurrently, 12-step affiliation did not predict depression independent of 12- step meeting attendance. While 12-step meetings may be more instrumental in reducing depressive symptoms due to the positive effects of social interaction and behavioral activation, the lack of effects for affiliation here could also be due to statistical and methodological limitations. The two 12-step variables were highly correlated at each wave, and this multi-collinearity could have biased regression coefficients . Our measure of 12-step affiliation was brief and basic, and other measures may capture detailed components of affiliation more predictive of depression. Secondly, future drug use was not predicted by 12-step attendance or depression. We had less power to detect this effect, because fewer patients were drug-dependent or using drugs at intake. Still, the effects of 12-step meetings could be stronger for patients who primarily use alcohol, possibly due to greater availability of specific meetings . The effects of 12-step meetings have differed between alcohol and drug users in prior research , so the lack of effects on drug use in this study may be important and deserving of further exploration. This study has limitations, most notably in the sample’s restricted demographic characteristics which limit the generalizability of findings. These results may be applicable to a large proportion of SUD patients due to the high prevalence of comorbid MDD in clinical settings , but replication in other samples is needed. Importantly, factors other than 12-step meeting attendance could have accounted for better initial depression response in TSF, but the groups had similar attendance at group therapy sessions and similar rates of antidepressant medication use.