Comparative efficacy and acceptability of psychosocial interventions for individuals with cocaine and amphetamine addiction: A systematic review and network meta-analysis

1Drug use disorders are the 15th leading cause of disability-adjusted life years in high-income countries. Cocaine and amphetamines are the most commonly abused stimulants in people aged 15–64 years, with an annual prevalence of misuse of 0.38% and 1.20%, respectively. Patients addicted to stimulants experience a range of psychological and physical sequelae including psychosis and other mental illnesses, neurological disorders and cognitive deficits, cardiovascular dysfunctions, sexually transmitted diseases, and blood-borne viral infections such as HIV and hepatitis B and C, and are at increased risk of all-cause mortality. Moreover, the social burden of stimulant abuse is worsened by its association with crime, violence, and sexual abuse.

2Currently, international clinical guidelines recommend the use of psychosocial interventions for cocaine and/or amphetamine addiction as first-line treatment, and there is little evidence supporting pharmacotherapy or brain stimulation treatments. In the absence of approved pharmacotherapies, several structured psychosocial and self-help approaches are available, such as contingency management (CM) (a behavioural approach that consists in providing stimulant users with rewards upon drug-free urine samples), community reinforcement approach (a multi-layered intervention involving functional analysis, coping-skills training, and social, familial, recreational, and vocational reinforcements), and 12-step programme (a set of guiding principles outlining a course of action for self-help recovery from addiction). International guidelines are unclear on whether any specific intervention should be considered first; for example, the National Institute for Health and Care Excellence (NICE) recommends CM alone, cognitive behavioural therapy (CBT) alone, or self-help groups based on 12-step programme alone for the treatment of individuals with stimulant use disorders. However, a recent systematic review showed that CM and CBT were well accepted and moderately efficacious at the end of treatment, but not at follow-up after treatment completion.

3Previous pairwise meta-analyses relied on a limited number of studies with direct comparisons between different interventions. From a clinical perspective, it is important to assess whether psychosocial interventions are effective and acceptable in both the short and long term, and also whether the combination of 2 approaches can produce a significant benefit. We therefore performed a network meta-analysis to compare and rank the efficacy and acceptability of individual and combined psychosocial interventions for the treatment of cocaine and/or amphetamine addiction at different time-points.

4This network meta-analysis was conducted following a pre-established protocol registered on PROSPERO (CRD 42017042900), and is reported according to the PRISMA for Network Meta-Analyses (PRISMA-NMA) guidelines. We searched the Cochrane Drugs and Alcohol Group Specialised Register, PubMed, Embase, CINAHL, ISI Web of Science, and PsycINFO from the date of database inception to 8 April 2018. We also screened international registers, hand-searched the reference lists of retrieved articles, and looked at key conference proceedings (for the full search strategy, see S1 Text). When needed, we contacted the investigators and relevant trial authors to obtain information about unpublished or incomplete trials. All searches included non-English language literature.

5We included randomised controlled trials (RCTs) comparing any structured psychosocial intervention against a control intervention—another psychosocial intervention or treatment as usual (TAU)—for the treatment of individuals with cocaine and/or amphetamine addiction, according to the Diagnostic and Statistical Manual of Mental Disorders (DSM) III, III-R, IV, IV-TR, or V or the International Classification of Diseases–10th revision (ICD-10) criteria. CBT, CM, community reinforcement approach, meditation-based therapies, non-contingent rewards, supportive-expressive psychodynamic therapy, 12-step programme, and their combinations were all identified as structured psychosocial interventions. We excluded studies on occasional users not actively seeking treatment and RCTs with study duration less than 4 weeks. We did not exclude studies on individuals with a comorbid substance use disorder (including opioid, alcohol, or cannabis use) or with a comorbid psychiatric disorder.

6Four authors (FDC, GLD, MC, RDG) independently screened the references retrieved by the search, selected the studies, and extracted the data, using a predefined data-extraction sheet. The same reviewers discussed any uncertainty regarding study eligibility and data extraction until consensus was reached; conflicts of opinion were resolved with another member of the review team (AC). 2 authors (GLD, MC) independently assessed the risk of bias of the included studies with the Cochrane tool, Three authors (FDC, CDG, AC) used the Grading of Recommendations Assessment, Development and Evaluation (GRADE) approach, through the Confidence in Network Meta-Analysis Software (CINeMA), to evaluate the strength of evidence for results at the end of treatment from the network meta-analysis (S2 Text).

7We considered as primary outcomes the efficacy and the acceptability of the interventions at the end of treatment. Efficacy was measured as the proportion of individuals abstinent (assessed by urinalysis), and acceptability as the proportion of individuals who dropped out from the study due to any cause. As secondary outcomes, we also measured efficacy and acceptability at 12 weeks from the start of treatment and at the longest follow-up (with follow-up starting at the end of treatment, independent of the duration of the intervention). If 12-week data were not available, we used data ranging between 4 and 20 weeks (giving preference to the time-point closest to 12 weeks). Other secondary outcomes were the longest duration of abstinence measured both at 12 weeks and at the end of treatment.

8We performed first pairwise meta-analyses using a random-effects model to estimate pooled odds ratios (ORs) for dichotomous outcomes and standardised mean differences (SMDs) for continuous outcomes with their 95% confidence intervals (CIs) using STATA. We assessed statistical heterogeneity in each pairwise comparison with τ2, I2 statistic, and P value. If at least 10 studies were available, we used the funnel plot and Egger’s test to detect publication bias.

9We incorporated indirect comparisons with direct comparisons using random-effects network meta-analyses within a frequentist framework using STATA (network package), and results are presented with the network graphs package. We report the results of network meta-analyses for both primary and secondary outcomes in league tables with effect sizes (OR or SMD) and their 95% CIs. When dichotomous outcome data were missing, we assumed that patients who dropped out after randomisation had a negative outcome. Missing continuous outcome data were analysed on an endpoint basis, including only participants with a final assessment, as reported by the original study authors. We also calculated the number needed to treat (NNT), which is the number of patients that need to be treated in order for 1 to benefit from the intervention compared with TAU.

10We assessed incoherence between direct and indirect sources of evidence using local and global approaches. Coherence (or consistency) is an important assumption to check in network meta-analyses because it is the manifestation of transitivity in the data from a network of interventions: coherence exists when treatment effects from direct and indirect evidence are in agreement (subject to the usual variation due to heterogeneity in the direct evidence). Local incoherence was measured by using a loop-specific approach (which identified inconsistent loops of evidence) and a side-splitting approach (which separated evidence on a particular comparison into direct and indirect evidence). Global incoherence was measured with the between-studies standard deviation (SD) (heterogeneity parameter) by using both a coherence and incoherence model and by measuring the chi-squared incoherence, with its P value. We estimated the presence of publication bias by plotting comparison-adjusted funnel plots for the network meta-analyses with a linear regression line. We also estimated the ranking probabilities for all treatments, i.e., their probability of being at each possible rank for each intervention. We report the treatment hierarchy as the surface under the cumulative ranking curve (SUCRA) and as the mean rank (S2 Text).

11To determine whether the results were affected by study characteristics, we performed subgroup network meta-analyses for abstinence and dropout at the end of treatment according to the following variables: year of publication, sex ratio, mean age group, intensity of the treatment, type of stimulant, risk of bias, opioid therapy, sample size, and comorbid alcohol misuse. Additionally, we performed sensitivity network meta-analyses for the primary outcomes by considering (a) only trials on individuals addicted to cocaine and no other stimulant and (b) only trials on individuals addicted to stimulants and on opioid substitution therapy.

12From the initially identified 7,261 citations, we retrieved 160 potentially eligible articles in full text (Fig 1). We excluded 88 reports, but then included 4 additional studies (3 from trial registers and 1 from screening the references), resulting in 76 publications (S3 Text) describing 50 RCTs (6,942 participants), published between 1993 and 2016 (Fig 2; Table 1), comparing 12 psychosocial interventions or TAU (listed and defined in S4 Text). Overall, 5,158 participants were randomly assigned to psychosocial treatments, and 1,784 to TAU. Full clinical and demographic characteristics are reported in Table 1. The mean study sample size was 139 participants, ranging between 19 and 487 participants. The median duration of treatment was 12 weeks (range 6–36). Dropout rates varied between 15.1% (CM + CBT) and 60.2% (meditation-based therapies) (S1 Table). A total of 37 studies were followed up after study completion, for a mean duration of 41.4 weeks (range 16–96). A total of 42 (84%) trials recruited patients from North America, 6 from Europe, 1 from Latin America, and 1 from Oceania. About a third of the population was women (35.9%), and the mean age was 36.8 years. A total of 76% of trials (38 of 50) enrolled participants with cocaine addiction, 8% of trials (4 of 50) with amphetamine addiction, and 16% (8 of 50) with both. About one-third of trials (18 of 50) enrolled participants on methadone maintenance. The mean addiction severity was moderate/high (S2 Table). In terms of risk of bias, 22 (44%) trials were rated low risk, 13 (26%) moderate, and 15 (30%) high (S3 Table; S1 Fig).

13We present all the networks for specific outcomes in S2 Fig. Eight psychosocial interventions had at least 1 trial versus TAU, and all of them were directly compared with at least another psychosocial intervention. We obtained unpublished or supplementary information for 5 of the included studies.

14The pairwise meta-analyses are presented in S4 Table, while data on heterogeneity are presented in S5 Table. The pairwise meta-analyses showed some statistically significant results in terms of abstinence and dropout. CM plus CBT, CM, and 12-step programme were superior to TAU in terms of abstinence at the end of treatment, while CBT, CM, and the combination of CM plus community reinforcement approach were superior to TAU in terms of dropout at the end of treatment (S4 Table).

15The results of the network meta-analysis are presented in Fig 3. In terms of abstinence at the end of treatment, the combination of CM plus community reinforcement approach, the combination of CM plus CBT, and CM alone were superior to non-contingent rewards (OR ranging between 2.59 [95% CI 1.70–3.93], P < 0.001, and 3.31 [95% CI 1.32–8.28], P = 0.010) and to TAU (OR ranging between 2.22 [95% CI 1.59–3.10], P < 0.001, and 2.84 [95% CI 1.24–6.51], P = 0.013). Moreover, the combination of CM plus community reinforcement approach was also superior to the combination 12-step programme plus non-contingent rewards and to CBT (OR 4.07 [95% CI 1.13–14.69], P = 0.031, and 2.43 [95% CI 1.02–5.88], P = 0.045, respectively), while CM alone and the combination of CM plus CBT were superior to CBT (OR 1.88 [95% CI 1.52–2.85], P = 0.003, and 2.08 [95% CI 1.28–3.33], P = 0.002, respectively). In terms of dropouts at the end of treatment, the combination of CM plus community reinforcement approach, community reinforcement approach alone, non-contingent rewards, CM alone, and CBT were better accepted than TAU (OR ranging between 1.41 [95% CI 1.10–1.82], P = 0.007, and 3.63 [95% CI 2.01–6.55], P < 0.001). Moreover, the combination of CM plus community reinforcement approach was better accepted than CBT, CM alone, CM plus CBT, community reinforcement approach plus non-contingent rewards, meditation-based therapies, non-contingent rewards, supportive-expressive psychodynamic therapy, 12-step programme alone, and 12-step programme plus non-contingent rewards (OR ranging between 2.06 [95% CI 1.04–4.08], P = 0.037, and 4.61 [95% CI 1.92–11.06], P < 0.001).

16Fewer studies reported results for abstinence measured at 12 weeks of treatment (S3 Fig) and at the longest follow-up after treatment completion (S4 Fig), but findings were in line with the outcome data at the end of treatment. Comparative abstinence and dropout at different time-points for each psychosocial intervention versus TAU are presented in Fig 4.

17In terms of the longest duration of abstinence measured at 12 weeks (S5 Fig), we found that the combination of CM plus CBT and CM alone were superior to TAU (SMD 0.75 [95% CI 0.31–1.19] and 0.62 [95% CI 0.43–0.80], respectively). Likewise, for the longest duration of abstinence measured at the end of treatment (S6 Fig), we found that the combination of CM plus CBT and CM alone were superior to TAU (SMD 0.74 [95% CI 0.43–1.06] and 0.60 [95% CI 0.43–0.76], respectively).

18The common heterogeneity SD for the coherence model was 0.46 and 0.21 for abstinence and dropout at the end of treatment, respectively (it was 0.47 and 0.19 for abstinence and dropout at 12 weeks, respectively). The global incoherence was not significant for all the outcomes considered (S6 Table). Tests of local incoherence did not show any inconsistent loops for abstinence and dropout at the end of treatment, although in some cases the ratio of the odds ratios (RoR) from direct and indirect evidence was large (i.e., RoR > 2), and we cannot definitely exclude the presence of incoherence. We found only 1 inconsistent loop for abstinence measured at 12 weeks and no other inconsistent loops for the other outcomes considered at 12 weeks (S7 Table; S7 Fig). The test of incoherence from the side-splitting model did not show significant differences for abstinence at the end of treatment but found some differences between some comparisons for dropout at the end of treatment (S8 Table). The comparison-adjusted funnel plots of the network meta-analysis for abstinence and dropout at the end of treatment were not suggestive for significant publication bias (S8 Fig).

19The ranking of psychosocial interventions based on cumulative probability plots and SUCRAs is presented in S9 Table and S9 Fig. We also performed subgroup analyses for abstinence and dropout at the end of treatment to study the effect of several potential moderator variables, the findings of which did not substantially differ from those of the primary analysis for most of the comparisons (S10 Table). Pre-planned sensitivity analysis on individuals addicted to cocaine only did not affect the main results (S10 Fig), while pre-planned sensitivity analysis on individuals on opioid substitution therapy showed a superiority of CM alone and CM plus CBT over TAU and non-contingent rewards, and a superiority of CM alone over CBT (S11 Fig). Predictivity intervals of mixed estimates are presented in S11 Table, while the overall limitations per comparison are presented in S12 and S13 Figs.

20We found that 4 patients needed to be treated with CM plus community reinforcement approach to have 1 additional patient abstinent at the end of treatment compared toTAU (NNT 4.07, 95% CI 2.29–21.95), with consistent results at longest follow-up after treatment completion (NNT 3.68, 95% CI 2.36–14.24). Similarly, 3 patients needed to be treated with CM plus community reinforcement approach to have 1 fewer patient dropping out at the end of treatment compared to TAU (NNT 3.25, 95% CI 2.42–5.79). For abstinence at the end of treatment, the NNT for the combination of CM plus CBT and for CM alone versus TAU was 4.80 (95% CI 2.99–12.12) and 5.44 (95% CI 3.74–9.75), respectively. For dropout at the end of treatment, the NNT ranged from 4.02 (95% CI 2.58–12.62) for CRA to 7.15 (95% CI 4.15–27.66) for non-contingent rewards, 10.52 (95% CI 5.83–53.65) for CBT, and 11.82 (95% CI 6.74–43.26) for CM.

21The overall strength of evidence according to GRADE is summarised in S12 Table for abstinence and for dropout at the end of treatment. For CM plus community reinforcement approach versus TAU, the strength of evidence was rated as “moderate” for abstinence and as “high” for dropout due to any cause. For CM plus CBT and CM alone versus TAU, the strength of evidence for abstinence was rated as “moderate” for both, while the strength of evidence for dropout due to any cause was rated as “very low” and “moderate”, respectively.

22This network meta-analysis is based on 50 studies including 6,942 individuals randomly assigned to 12 different psychosocial interventions or TAU. To our knowledge, it is the most comprehensive synthesis of data for all available psychosocial interventions in individuals with cocaine and/or amphetamine addiction.

23We found that CM alone or in combination with either community reinforcement approach or CBT had superior efficacy and acceptability compared to TAU at 12 weeks and at the end of treatment. This effect was not significantly influenced by clinical modifiers in the subgroup analyses and remained significant in the sensitivity analyses. Moreover, CM in combination with community reinforcement approach and community reinforcement approach alone were more effective than TAU at the longest follow-up after treatment completion. The clinical relevance of this finding is key, because achieving long-term abstinence is the main treatment goal for individuals with cocaine and/or amphetamine addiction.

24Several major guidelines recommend the use of either CBT or CM alone for the treatment of cocaine and/or amphetamine addiction. Self-help groups following the 12-step programme are also recommended. Our results do not support these recommendations. We found that CBT alone was more acceptable than TAU (NNT 10.5, 95% CI 5.8–53.6), but it was not superior for abstinence on any dichotomous or continuous outcome measured and was less effective than CM alone. CM alone showed greater efficacy (NNT 5.2, 95% CI 3.6–9.3) and acceptability (NNT 12.5, 95% CI 7.0–48.8) than TAU at 12 weeks of treatment and at the end of treatment (NNT 5.4, 95% CI 3.8–9.8, and 11.9, 95% CI 6.8–43.3, respectively), but the effect was not sustained at the longest follow-up after treatment completion. Both CBT and CM were inferior to CM in combination with community reinforcement approach at the longest follow-up after treatment completion. Community reinforcement approach alone was not different from TAU for abstinence at 12 weeks of treatment or at the end of treatment, but showed increased abstinence at the longest follow-up after treatment completion (NNT 4.1, 95% CI 2.4–36.2). CM in combination with community reinforcement approach was superior to TAU for abstinence at 12 weeks of treatment (NNT 2.1, 95% CI 1.6–6.2), at the end of treatment (NNT 4.1, 95% CI 2.3–21.9), and at the longest follow-up after treatment completion (NNT 3.7, 95% CI 2.4–14.2), as well as for acceptability at 12 weeks of treatment (NNT 3.1, 95% CI 2.2–6.1) and at the end of treatment (NNT 3.3, 95% CI 2.3–6.3), as shown in Fig 4. CM plus community reinforcement approach was also superior to 12-step programme for abstinence and dropout at 12 weeks of treatment, for dropout at the end of treatment, and for abstinence at the longest follow-up after study completion.

25Behavioural interventions have proved effective for the treatment of other forms of addiction. There is growing evidence that reducing punishment—such as incarceration—and adopting positive reinforcement for people with substance use improves their access to services, their reintegration into society, and, ultimately, public safety. Recent experimental data emphasise the potential of interventions that focus on improving goal-directed behaviour and positive reinforcement rather than punishment in people with cocaine addiction. The efficacy of a purely behavioural intervention—such as CM alone—shows that financial rewards can compete with biological rewards mediated by cocaine and amphetamine cues and incentives. This seems to be true only if rewards are contingent upon the provision of drug-free urine samples, as non-contingent rewards were not shown to be effective (Fig 4). Indeed, CM strategies help individuals to overcome apathy or resistance to the recovery process. In this study we found that, although CM was efficacious at the end of treatment, the effect of CM alone was not sustained at longest follow-up after treatment completion (Fig 4). Cocaine and amphetamine addiction is conceptualised as a chronic and recurrent brain disease, which entails behavioural and psychological abnormalities (primarily reward-processing deficits) following the learned or conditioned pairing of situational and social cues with the reinforcing effects of drug use. It seems unlikely that a behavioural strategy alone could address in the long term the whole complexity of biological, psychological, and behavioural factors that underlie addiction. The addition of community reinforcement approach to CM potentiates an otherwise purely behavioural intervention with psychological and social components that may enhance its effect. Notably, community reinforcement approach alone performs no differently from TAU in the short term, but its effect is more sustained at longest follow-up. The combined intervention, CM plus community reinforcement approach, overall achieves the best outcomes.

26Cocaine and amphetamine addiction is highly prevalent in the world and is incredibly costly economically. In 2015, illicit drugs cost tens of millions of disability-adjusted life years, with Europeans proportionately experiencing more, but with the greatest mortality rate in low- and middle-income countries. We did not do a formal cost-effectiveness analysis. Indeed, recent cost-effectiveness analyses on psychosocial interventions for substance use are encouraging, but without a full economic model our recommendation cannot be made unequivocally.

27This study has some limitations. Some comparisons were appraised as having low or very low quality, potentially restricting the validity of those results. All RCTs of psychosocial interventions for cocaine and/or amphetamine addiction are not blinded, which increases the risk of performance bias for self-reported outcomes. For this reason, we only reported data based on objective outcomes (abstinence on urinalysis and data on attrition), which are less likely to be influenced by the lack of blinding. The risk of selective study reporting was minimised as we contacted study authors to retrieve unpublished data, but we cannot exclude that some unpublished studies remain missing or that published reports overestimated the efficacy of treatments. Finally, some interventions are designed to last more than 12 weeks, namely, CM in combination with community reinforcement approach (24 weeks), community reinforcement approach alone (time unlimited, but in the studies included the intervention lasted always 24 weeks), and supportive-expressive psychodynamic therapy (which could be either time-limited or unlimited and lasted 36 weeks in the only study included). For these interventions, the evaluation at 12 weeks was extracted before the end of treatment, which was a disadvantage over other interventions requiring shorter duration.

28The results of this network meta-analysis support the use of combined CM plus community reinforcement approach as the most effective and acceptable intervention for both short- and long-term treatment of individuals with cocaine and/or amphetamine addiction. The provision of evidence-based psychosocial treatments for stimulant use disorders is all the more important because of the lack of validated pharmacological or brain-stimulation-based treatment for cocaine and/or amphetamine addiction. These findings may influence clinical guidelines; however, further studies are warranted to confirm these results and evaluate cost-effectiveness.