In independent studies lower self-efficacy was correlated with neurocognitive impairment and mediated the effects of impairment on drinking . Impairment also moderated the effects of self-efficacy and Alcoholics Anonymous affiliation, such that these variables were less predictive of substance use for impaired patients. Other studies support a moderating role of impairment , as the protective influence of frequent, positive social support at baseline was greater for patients with executive impairment . Each these studies employed “threshold” models of impairment as opposed to continuous measures, which may more sensitive to the level of cognitive disturbance required to significantly impact the process of psychotherapy. Our previous studies found that 12-step affiliation and self-efficacy predicted drinking in patients with substance dependence and MDD, but it is not known whether neurocognitive impairment relates to individual differences in these well-established therapeutic process variables, or moderates their effects on substance use for patients with psychiatric comorbidity. Previous research also indicates negative affect is an important correlate of post treatment substance use , especially for patients with comorbid MDD . If neurocognitive deficits impact the course of depressive symptoms in patients with substance dependence and MDD, this is another pathway through which neurocognitive impairment may impact substance use. Previous research supports this pathway, as depression severity was linked to executive functioning ,cannabis grow tray and fluid reasoning ability predicted reductions in depression among depressed, hazardous drinkers . Also of interest is whether impairment moderates the effects of depressive symptoms on substance use.
In studies of non-clinical samples the effect of depression on drinking was moderated by impulsivity and distress tolerance , other neurobehavioral constructs that could relate to neurocognitive impairment in more severe, treatment seeking patients. We previously found that greater depressive symptoms predicted greater future drinking in the current sample . If impaired patients have worse acquisition of self-efficacy and coping skills as suggested in prior studies , they may have exaggerated risk for drinking and drug use following periods of increased negative affect. The overall goal of this study was to explore the impact of neurocognitive impairment on substance use treatment outcomes in patients with comorbid substance dependence and MDD. These aims are guided, in part, by recent models suggesting that substance use is a complex interplay of distal risk, proximal risks, intrapersonal processes, and contextual factors , and that thorough investigations of therapeutic processes should examine intricate inter-relationships between these factors . Based on the literature reviewed thus far, we hypothesized that the effects of neurocognitive impairment on future drinking and drug use would be mediated by lower self-efficacy, lower 12-step affiliation and greater depressive symptoms. We also tested interactions between these proximal factors and neurocognitive impairment, expecting that impairment would moderate prospective relations between these variables and future substance use, and that moderating effects would be strongest during periods of increased negative affect. This study included 197 veterans who completed neurocognitive testing prior to participation in a controlled trial of outpatient group psychotherapy for comorbid substance dependence and MDD . Inclusion criteria included a diagnosis of lifetime dependence on alcohol , cannabis , or stimulants with use in the past 90 days, and DSM-IV diagnosis of MDD, with at least one depressive episode occurring independently of substance use.
Participants were excluded if they were using opiates with intravenously, had bipolar or other psychotic disorder, lived excessively far from the SDVAHS, or had severe memory impairments prohibiting accurate recall in assessments. The current sample averaged 49.3 years of age and was mostly male and Caucasian . At intake very few participants were working or married . In the three months preceding the study 85% of the sample had used alcohol, with an average of 15.07 drinks per drinking day, and 47% had used drugs. The average intake Hamilton Depression score was 28.35 , indicative of severe depressive symptoms. The procedures for this study were approved by the University of California, San Diego and VASDHS Institutional Review Boards and have been described in greater detail in prior publications and are explained briefly here. Initial referrals from the VASDHS dual diagnosis clinic were first reviewed for eligibility. Research staff contacted veterans to conduct brief screenings before meeting with eligible veterans to explain procedures and obtain informed consent. Veterans consented to random allocation to group psychotherapy, assessment visits every three months during 6 months of treatment and 12 months of follow-up, recording of group sessions, random toxicology screens, psychotropic medication consultation, and review of electronic medical records. After completing the baseline assessment veterans were sequentially allocated to TSF or ICBT on an alternating, two-week rotation. For the duration of the group intervention provided in the study, veterans agreed to receive no other formal treatment for substance use or depression, while participation in other formal interventions was allowed during follow up. Because treatment group differences in long-term outcomes and process variables are not a focus of this study and have previously been published , the treatments are described only briefly here.
The manual for TSF was modified to allow group delivery and targeting both drugs and alcohol, while Integrated Cognitive-Behavioral Therapy was developed by integrating cognitive-behavioral relapse prevention and group cognitive-behavioral therapy for depression . Group sessions occurred twice/week for three months and weekly thereafter . Co-therapists were trained via manual review, direct observation, and weekly supervision, and rotated across conditions every 6-12 months. All participants scheduled pharmacotherapy appointments with VA physicians, and nearly all utilized medication management. Our index of neurocognitive impairment was modeled after previous studies examining similar relationships in substance-dependent samples . Analyses were structured to address our two overarching research questions: whether effects of neurocognitive impairment on substance use were mediated by self efficacy, 12-step affiliation, or depressive symptoms, and whether impairment moderated relationships between these process variables and substance use outcomes. Hierarchical linear models are ideally suited for examining these questions in longitudinal data. By allowing random effects multiple time points nested within individuals can be analyzed in one analysis. Repeated measures of process variables can be included as time-varying predictors, and cross-level interactions can be specified to examine cross-level interactions between predictors at the between-individual and within-individual levels. All available data was included via maximum likelihood estimation, a preferred method for handling missing data when assumed missing-at-random . Although in practice this assumption cannot be fully confirmed, we found no significant differences on study variables between participants with complete data vs. those with any missing data, supporting this assumption. With the exception of group and time effects, all predictor variables were grand-mean centered prior to inclusion in analyses. All statistical analyses were performed in Stata 10.1 . Prior to mediation and moderation analyses, we examined static predictors of PDD and PDDRG in preliminary models to identify any statistically significant covariate effects,vertical grow systems for sale including demographics, therapy attendance, baseline severity of PDD/PDDRG, group, time , and the group by time interaction. To examine the effects of neurocognitive impairment on substance use outcomes and process variables , each variable was analyzed separately in HLM with impairment as an individual-level predictor, controlling for effects of time, group, and the group by time interaction. For each process variable predicted by neurocognitive impairment, we then examined the variable’s effect on future substance use by including it as a time-varying, lagged predictor of PDD/PDDRG. While these sequential steps detect likely presence of mediation, formal statistical tests are better indicators of mediated effects. Consistent with current conventions in mediation , we examined the statistical significance of mediated effects using the products-of-coefficients with asymmetric 95% confidence limits . This method has more accurate Type I error rates and greater power to detect mediated effects than alternate approaches , and is appropriate for the analysis of nested data structures .
In the final models we tested whether neurocognitive impairment moderated the strength of association between process variables and future drinking/drug use. After estimating the covariate model, a main effects model was estimated that simultaneously tested the effects of all process variables and neurocognitive impairment on PDD/PDDRG. To test the moderating effects of impairment, three cross-level interaction terms were created, one for each process variable. To reduce nonessential multi-collinearity, all predictor variables were grand-mean centered prior to creation of interaction terms . In separate HLMs each interaction term was added to the main effects model to test the statistical significance of moderation. For any significant two-way interaction terms, we also tested three-way interactions with other process variables, to determine whether these interactions were further moderated by other treatment processes. In the next series of analyses we tested the effects of baseline neurocognitive impairment on substance use outcomes and mediating variables. In mediation terminology these analyses would test for the presence of significant relations between the predictor and outcome and for significant effects of the predictor on mediators, namely self-efficacy, 12-step affiliation, and depression. Neurocognitive impairment did not significantly predict PDD , or PDDRG . However, as displayed in Table 12, impairment did significantly predict self-efficacy, 12-step affiliation, and depressive symptoms, in the hypothesized directions. Thus, while individuals with greater impairment, on average, did not drink or use drugs more frequently, they did have lower self-efficacy, lower 12-step affiliation, and greater depressive symptoms during the course of the study. With the first presumptive paths for mediation supported, the next analyses examined whether self-efficacy, 12-step affiliation, and depression predicted future drinking and drug use. Each process variable was tested as time-varying predictor in separate HLMs for PDD and PDDRG, which controlled for previously mentioned covariates and neurocognitive impairment. Future PDD was significantly predicted by self-efficacy, 12-step affiliation, and depressive symptoms, and future PDDRG was also predicted by each of these process variables, supporting the second presumptive path for mediation. These results indicated that individuals with greater self-efficacy, greater 12- step affiliation, and lower depressive symptoms had greater future drinking and drug use frequency. Further analyses were conducted to determine whether neurocognitive impairment moderated relations between these process variables and future substance use. In contrast to analyses conducted thus far, the next set of HLMs specified and tested cross-level interactions between neurocognitive impairment and process variables in the prediction of PDD and PDDRG. To proceed in a parsimonious fashion given the large number of potential interactions, separate models tested each impairment by process variable interaction separately, before building significant terms into a full model with potential three-way interactions. Results for PDD indicated that effects of lagged self-efficacy and lagged depressive symptoms were not moderated by impairment, but that impairment significantly moderated effects of lagged 12-step affiliation on PDD .As displayed in Table 4, the interaction of impairment and 12-step affiliation was statistically significant in the full model controlling for all process variable main effects and other interactions. Moreover, this interaction was further qualified by a significant 3-way interaction between impairment, affiliation, and depression, . As depicted in Figure 2, the effect of 12-step affiliation on PDD was strongest for impaired patients when levels of depressive symptoms were high. Overall, these results suggest that in our sample the benefits of 12- step affiliation on future drinking were greatest for patients with greater levels of neurocognitive impairment, especially when they were severely depressed. Separated analyses examined identical relationships in the prediction of drug use. Impairment did not moderate the effects of lagged self-efficacy or lagged 12-step affiliation , but impairment did moderate the effects of depressive symptoms . As displayed in Figure 1, depressive symptoms predicted PDDRG to a greater extent in patients with lower impairment. However, this interaction was no longer statistically significant when controlling for lagged self-efficacy and 12-step affiliation. Results indicated that, contrary to predictions, more severe depressive symptoms predict greater future drug use to a lesser extent among patients with greater cognitive impairment, but these effects were not independent of prior levels of self-efficacy and 12-step affiliation. Despite strong clinical rationale that neurocognitive deficits will negatively impact the outcome of psychotherapy for treatment of substance dependence, a number of studies have not found worse outcomes for patients with neurocognitive functioning . More recent investigations suggest that neurocognitive impairment impacts substance use outcomes through more indirect mechanisms: by influencing treatment process variables like self efficacy and AA affiliation, or by moderating the strength of relations between these variables on substance use .