The Clinical Efficacy of Mindfulness-Based Treatments for Alcohol and Drugs Use Disorders: A Meta-Analytic Review of Randomized and Nonrandomized Controlled Trials

Alcohol and Drugs Use Disorders (respectively AUD and DUD) are one of the most prevalent mental disorders worldwide [1-4]. Specially, a recent epidemiological study revealed that 12-month and lifetime prevalence respectively ranged from 13.9 to 29.1% for AUD and from 3.9 to 9.9% for DUD [1, 5]. They also add to global morbidity and mortality [6-8] as well as to severe deficiencies in productivity, interpersonal, social, and psychological functioning [9-10].

Currently, it is well known that there are several evidence-based interventions that demonstrated promising results for both AUD and DUDs. For instance, the U.S. Food and Drug Administration has permitted 3 medications to treat alcohol dependence – disulfiram, oral and injectable naltrexone, and acamprosate – and nalmefene were permitted by the European Medicines Agency [11-12]. Numerous studies also confirmed how pharmacotherapy might be beneficial in DUDs treatment (for overviews see: [13-15]). Furthermore, a variety of psychological and behavioral therapies also demonstrated to be effective (e.g., cognitive-behavioral therapy [CBT], motivational enhancement therapy, 12-step facilitation therapy [16-19]). Nevertheless, across substance use disorders (SUDs), relapse in dysfunctional substance use is considered the core clinical features of such population [20, 21]. Accordingly, in the last 2 decades, an increasing interest for alternative interventions that might yield better outcomes in treating SUDs has been observed. In this situation, some authors hypothesized how mindfulness-based interventions (MBIs) might positively integrate CBT to promote effective programs for addictive behaviors [22, 23].

The first western definition of mindfulness was given by Kabat-Zinn [24] who described it as “paying attention in a particular way, on purpose, in the present moment, and non-judgmentally” [24]. Consecutively, Bishop et al. [25] operationalized mindfulness as particular focus of attention categorized by 2 distinct features that are largely related to the self-regulation of attention toward the immediate present moment and an attitude marked by inquisitiveness, openness, and acceptance. Finally, Shapiro et al. [26] added a third component that underlines the intention or the personal motivation in engaging with mindfulness practice.

Numerous studies demonstrated how individuals learn their purposeful control of attention through training methods such as mindfulness meditation (e.g., [27-29]); also, it was postulated that the development of an observing and acceptance attitude toward present-moment experiences might permit “the individual to more consciously choose thoughts, emotions, and sensations they will identify with, rather than habitually reacting to them” ([30]; p. 569). Additionally, the development of this mental position can facilitate a skillful response to a given situation [22, 26] that contrast everyday habitual mental functioning or being on “autopilot.”

Given the former evidences and assumptions, some authors proposed the rationale that sustains the integration of mindfulness practice into traditional treatments for SUDs individuals. For example, Groves and Farmer [31] confirmed “In the context of addictions, mindfulness might mean becoming aware of triggers for craving…and choosing to do something else which might ameliorate or prevent craving, so weakening the habitual response” (p. 189). Further, Witkiewitz et al. [32] continued how mindfulness meditation might disrupt the craving response system, which is considered by an association between environmental cues and rigid cognitive responding, by providing heightened awareness and acceptance of the initial craving response, without judging, analyzing, or reacting [32].

On the basis of previous deliberations, a manualized psychological treatment for addiction called “mindfulness based relapse prevention” (MBRP; [32-35]) has been developed. Likewise, during the last decade, other MBIs have been adapted for SUDs such as acceptance and commitment therapy (ACT; e.g., [36-38]), spiritual self-schema therapy (e.g., [39-41]), dialectical behavior therapy (e.g., [42-44]), mindfulness-based stress reduction (MBSR; e.g., [45]), and Vipassana Meditation (e.g., [46-48]).

The increasing body of empirical research on mindfulness-based programs use to treat SUDs led Zgierska et al. [49] and Chiesa and Serretti [50] to conduct 2 systematic reviews on this topic in order to draw some decisions about their efficacy in this clinical population. Even though the authors concluded in favor of hopeful results in using MBIs for addiction treatment, they underlined substantial methodological restrictions in most studies published and unclear evidences about which persons with addictive disorders might benefit most from these programs. Furthermore, Li et al. [51] recently published a quantitative meta-analytic review on the same field of research, concluding that MBIs could be beneficial in reducing substance use, craving, and perceived stress as well as in improving mindfulness skills.

However, this work has some relevant limitations regarding its informative value on incremental effectiveness of MBIs, compared to other standard programs, in treating AUD and DUDs. First of all, they included several studies carried out among nonclinical populations. Second, they separately considered results from randomized and nonrandomized trials (respectively RCTs and NRCTs), an aspect that might significantly influence studies outcomes [52], and they also included studies that compared MBIs with no active control conditions. Third, they did not present results related to an essential treatment outcome in SUDs that refer to the attrition rate (AR) [53]; also, they did not take into consideration several other secondary outcomes that are robustly associated with relapse in substance use among clinical populations (i.e., negative affectivity, depressive/anxiety and post-traumatic symptoms and avoidance coping strategies; e.g., [20, 54-57]). Furthermore, they did not evaluate the effect of clinical setting (i.e., group, group + individual, individual) that is considered a core aspect in SUDs treatment efficacy [58]. Finally, they did not compare pooled effect sizes associated with different treatment outcomes so as to clarify if MBIs could have specific or generalized therapeutic effects; also, they did not report quantitative robustness indexes of their conclusions.

Consequently, our aim is to accomplish a meta-analytic review of the literature in order to demonstrate if and to what extent MBIs could promote an incremental effectiveness compared to other active programs usually provided in clinical practice for AUD and DUDs treatment. Consistent with this possibility, we considered RCTs and NRCTs that compared MBIs with other active programs, examining primary and secondary outcomes, which comprehensively operationalize treatment efficacy. In line with previous considerations regarding Li et al. [51] work, we chose to take into consideration AR, abstinence (e.g., any substance use vs. no substance use; duration of abstinence) and levels of perceived craving as primary outcomes of treatments. Levels of perceived stress, negative affectivity, overall mental health and specific (i.e., depressive, anxiety, post-traumatic) psychiatric symptomatology, and the use of avoidance coping strategies were evaluated as secondary outcomes. Furthermore, we investigated the role of core clinical features (i.e., research design, length of intervention/follow-up, short and long-term effects, types of MBIs, types of control conditions, clinical settings, sample characteristics, intervention developed for treating the co-occurrence of SUDs with other psychiatric conditions) that might explain the heterogeneity of findings, reporting quantitative indexes for the robustness of our conclusions. Eventually, we explored if MBIs could produce similar effects on such outcomes, or their therapeutic effects might be related to specific clinical domains relevant for SUDs treatment.

In order to be included in this work, studies had to be published in scientific peer-reviewed journals. Consistently with Zeng et al. [169] suggestions, we referred to the Cochrane Collaboration’s tool for assessing the risk of bias [170] in order to evaluate the quality of RCTs. Conversely, the quality of NRCTs was assessed by Methodological index for nonrandomized studies [171].

PsycINFO, PubMed, ISI Web of Knowledge, and Scopus were the primary sources of information. We conducted the on-line research up to August 31st 2017. The main search terms were “mindfulness”, “mindfulness meditation,” “MBI,” “mindfulness training”, “MBSR,” “mindfulness-based cognitive therapy,” “MBRP,” “dialectical behavior therapy,” “ACT,” “spiritual self-schema therapy,” “Vipassana meditation,” and “Zen meditation” in combination with the name of each substance (i.e., substances, drugs, alcohol, marijuana, cocaine, opioid, heroin, methamphetamine). The references of reviews and meta-analyses were referred to as additional sources of information [49-51, 61-63].

Moreover, so as to assess the incremental effectiveness of MBIs in AUD and DUDs treatment, the studies included in this meta-analytic review had to feature an assessment of MBIs with other active interventions based on RCTs and NRCTs (e.g., [59, 60]). Particularly, studies had to compare MBIs with other psychological, psychoeducational, and/or pharmacological treatments usually provided in clinical practice. We considered studies that reported a clear description of the characteristics of interventions (e.g., protocol, setting, length of program/follow-up), especially referring to therapeutic strategies used to address AUD and DUDs clinical targets.

Finally, all studies had to refer to valid and reliable criteria for AUD and DUD diagnoses (Diagnostic and Statistical Manual of Mental Disorders) and all studies had to use valid and reliable instruments to assess treatment outcomes.

Figure 1 shows a detailed description of data extraction procedures and studies that met criteria for inclusion eligibility. Tables 2 and 3 summarize results of assessment procedures used to evaluate the quality of studies included in the current meta-analysis.

Cohen’s d [64] was calculated as a measure of effect size. The index was primarily calculated using descriptive statistics reported in the Results section of each study. In addition, t and χ² tests were used to evaluate Cohen’s d when descriptive statistics was not available [66].

To consider primary and secondary outcomes pooled effect sizes comparable, we decided to estimate Cohen’s d even in the case of binary data (i.e., AR; any substance use vs. no substance use). Accordingly, we computed OR and we converted them to Cohen’s d [65].

Furthermore, we used adequate procedures proposed by Morris [172] to estimate Cohen’s d when pre-post changes in outcomes measures were considered. Specifically, the Cohen’s d computation was based on pre-post scores differences, the pooled pre- and posttest standard deviation and the application of a bias correction factor. In the case of multiple comparisons over time performed by the original authors, we computed d for each contrast and obtained a single pooled coefficient, consistently with procedures clarified by Borenstein et al. [65].

Values of Cohen’s d less than or equal to 20, 50, and 80 were inferred as small, moderate, and large effect sizes respectively [64].

Overall the pooled effect size (dw) of each outcome measure was estimated using the weighted mean of the d value for each study [67]. The 95% CI was computed, as was its significance according to the ratio of dw to the standard error [67].

Heterogeneity in effect sizes was computed using the Q statistic [67] and I² index [68, 69]. Despite the small number of studies included in the current work, we used Begg and Mazumdar’s rank correlation test (r_B-M) [70] and Egger’s regression [71] to detect the publication bias. Given the small number of available studies, a bootstrap methodology (bias corrected and accelerated; [72]) was applied in calculating the significance of the previous parameters. A total of 1,000 bootstrap independent samples were used with p < 0.05 (2-tailed). Spearman’s rho was used to evaluate the significance of the correlation between effect size, sample size, year of publication, and length of intervention/follow-up. Subgroup analyses (i.e., RCTs vs. NRCTs; short- vs. long-term effects; MBIs + control condition vs. manualized MBIs; MBIs vs. CBTs/TAU; group vs. group + individual/individual settings; several SUDs vs. specific SUDs; SUDs-other psychiatric disorders vs. SUDs) were conducted using methodologies described by Borenstein et al. [65] based on the Z-test. Z-test was also used to compare pooled effect sizes within primary and secondary outcomes. Adequate Bonferroni correction was applied when we performed multiple comparisons.

Orwin’s fail-safe procedure [73] was assessed in order to measure the number of studies with null results needed to overturn our conclusions. For Orwin’s fail-safe N, the critical level was set at 20. Moreover, using the procedures proposed by Rosenthal [74], we computed the critical value (5k + 10; k = number of studies) of Orwin’s fail-safe N to evaluate the power of our conclusions.

Thirty-seven studies [34, 36, 38, 43, 44, 46, 47, 75-104] were eligible for a total of 3,531 AUD and DUDs patients admitted to the therapeutic programs.

Table 1 shows a comprehensive description of characteristics of each study. Seven studies implemented mindfulness practices into usually provided programs, 6 studies combined manualized MBIs (i.e., ACT, MBRP, MBSR) with other standard interventions. Twenty-one studies compared manualized MBIs adapted for SUDs with other active interventions.

Control treatments were represented by 12-steps focused programs (7 studies), CBTs (7 studies), individual counseling, or psychoeducational treatments or supportive groups (4 studies), mixed interventions which combined the previous treatment approaches (11 studies), and therapeutic community (3 studies). Figure 1 shows a detailed description of the distribution of clinical approaches.

As previously mentioned, we aggregated results for AR (28 studies), abstinence (21 studies), levels of perceived craving (7 studies) and stress (5 studies), negative affectivity (3 studies), overall mental health (6 studies), the severity of depressive (9 studies), anxiety (5 studies) and post-traumatic (3 studies) symptoms, and several forms of avoidance coping strategies (8 studies).

Tables 4-13 discretely summarize the effect sizes and show the forest plots for each outcome previously presented. Tables 4-13 also show detailed results related to Spearman’s correlations, subgroup analyses, publication bias, and Orwin’s fail-safe number of each outcome.

In the followings sections, we report main findings for primary and secondary outcomes. Figure 2 and 3 exhibit results related to multiple comparisons within the previous categories of treatment outcomes.

We observed a null pooled effect size in AR when MBIs were compared with other treatments. However, we detected a significant and high heterogeneity in effect sizes. Accordingly, we explored some possible source of heterogeneity. First of all, we excluded from analysis Marcus et al. [92] research because it represents a distribution outlier. Results displayed a decrease in extent of heterogeneity (I² = 55.85%), albeit it remained moderate and significant (Q_[26] = 58.90, p < 0.001), and null pooled effect size (d_w = –0.06 [–0.02 to 0.14]; ns). The research design (RCTs: d_w = 0.00 [–0.10 to 0.10]; Q_[19] = 29.80, ns; NRCTs: d_w = –0.18 [–0.33 to –0.05], p < 0.05; Q_[6] = 23.50, p < 0.001, I² = 74.47%; Z = –2.16; p < 0.05), the type of MBIs (manualized MBIs: d_w = 0.01 [–0.09 to 0.10], ns; Q_[18] = 29.58, p < 0.05, I² = 39.16%; MBIs + control condition: d_w = –0.22 [–0.37 to –0.07], p < 0.01; Q_[7] = 28.57, p < 0.01, I² = 68.98%; Z = 2.59; p < 0.05) and sample characteristics (several SUDs: d_w = –0.15 (–0.25 to –0.06); p < 0.01; Q_[20] = 41.72, p < 0.01; I² = 52.06%; specific SUDs: d_w = 0.18 [0.03–0.33], p < 0.05 Q_[6] = 3.52, ns; Z = –2.16; p < 0.05) moderated the extent of pooled effect sizes and partially explained results variability. Publication bias was not detected.

Table 1 provides a detailed description of abstinence assessment procedures administered within each study. Even though we observed a large variability in methods of evaluation (e.g., self-report vs. objective measures; maximum duration of assessed abstinence), abstinence was generally operationalized both as a binary (e.g., any substance use vs. no substance use) or continuous (e.g., days of any substance use) outcome.

A significant small pooled effect size was found in abstinence, in association with significant and moderate heterogeneity across the results. Specifically, MBIs seemed to promote abstinence maintenance better than other conditions. Excluding the study of Wupperman et al. [103], which represented an outlier, we confirmed the previous MBIs advantages (d_w = 0.37 [0.30–0.45], p < 0.001) and this effect was consistent across studies (Q_[19] = 29.59, ns; I² = 35.80%). We did not reveal bias of publication. As indicated by the value of Orwin’s fail-safe number, the beneficial effect previously reported is so far considered to be conclusive.

A large pooled effect size was found in relation to the levels of perceived craving during the interventions. Specifically, MBIs seemed to significantly decrease the levels of craving if they were compared with other approaches. We revealed a large heterogeneity across the results. However, MBIs seemed to be more effective than other active programs in reducing levels of craving when they were specifically carried out to treat the co-occurrence of SUDs and other psychiatric disorders (d_w = –2.36 [–2.61 to 2.11], p < 0.001; Q_[2] = 6.37, p < 0.05; I² = 68.62; only SUDs: d_w = –0.10 (–0.28 to 0.08), ns; Q_[3] = 5.43, ns; Z = –14.47, p < 0.001). The Orwin’s fail-safe number (n = 32.4; critical value = 25) associated to this finding was robust enough to draw definitive conclusion in favor of MBIs, specifically in the case of the co-occurrence between SUDs and other psychiatric disorders.

The improvement in levels of perceived craving was significantly different to abstinence (Z = 6.03, p < 0.001) and AR (Z = 10.03, p < 0.001). Additionally, the abstinence pooled effect size was significantly larger than AR (Z = 6.02, p < 0.001; Bonferroni correction: α = 0.0167).

Considering levels of perceived stress during the programs, we found a consistent and small-to-moderate pooled effect size. MBIs seemed to reduce the levels of perceived stress if they were compared with other treatments. Nonetheless, we found a significant -relationship between d and the sample size. Specifically, clinical trials characterized by larger samples showed smaller difference in this outcome. Bias of -publication was not revealed. As indicated by Orwin’s fail-safe number, the beneficial effect of MBIs on -reduction of perceived stress was not robust enough in -order to definitely conclude in favor of such treatments.

We observed a large pooled effect size for negative affectivity. When compared to other programs, MBIs seemed to significantly reduce negative emotional experiences. However, large heterogeneity was detected across the results. Specifically, only one study [84] demonstrated large differences between MBI and control condition in reducing negative affectivity. Conversely, the remaining 2 studies [91, 97] showed no differences between treatment approaches.

The paucity of studies did not permit in exploring possible sources of heterogeneity. Additionally, for the same reason, it was not possible to compute bias of publication.

The Orwin’s fail-safe number showed that the beneficial effects of MBIs on this outcome are not conclusive.

The null pooled effect size was observed in overall mental health. Results were consistent across studies. We did not detect bias of publication.

A moderate-to-large pooled effect size was observed, even though it was associated with high heterogeneity. MBIs seemed to support greater decrease of depressive symptomatology than other programs. Comparing pooled effect sizes, we found a significant difference when MBIs were specifically carried out to treat the co-occurrence of SUDs and other psychiatric disorders (d_w = –0.91 [–1.07 to –0.74], p < 0.001; Q_[5] = 48.03, p < 0.001; I² = 89.59%; only SUDs: d_w = –0.04 [–0.27 to 0.19], ns; Q_[2] = 3.64, ns; I² = 45.11%; Z = –6.03, p < 0.001); also, when we took into consideration the sample characteristics (several SUDs: d_w = –0.93 [–1.09 to –0.76], p < 0.001; Q_[5] = 43.50, p < 0.001; I² = 88.50%; specific SUD: d_w = 0.04 [–0.19 to 0.27], ns; Q_[2] = .38, ns; I² = 0.00%; Z = –6.64, p < 0.001). Further significant difference in pooled effect sizes was observed when it was considered the effect of clinical setting (Group: d_w = –1.09 [–1.28 to –0.90], p < 0.001; Q_[4] = 39.21, p < 0.001; I² = 89.80%; Group + Individual: d_w = –0.03 [–0.26 to 0.20], ns; Q_[2] = 5.12, ns; I² = 60.93%).

We did not detect bias of publication. However, the Orwin’s fail-safe number demonstrated that the improvement in depressive symptomatology associated to MBIs is not robust enough.

We found a moderate-to-large pooled effect size and large heterogeneity across the results. In detail, MBIs seemed to support greater decrease of anxious symptomatology than other approaches. We revealed significant differences between short- (d_w = –0.37 [–0.60 to –0.13], p < 0.001 [n = 2]) and long-term (d_w = –1.50 [–1.90 to –1.10], p < 0.001; Q_[3] = 13.71, p < 0.001; I² = 85.42%; Z = –4.85, p < 0.001) effects of MBIs. Additionally, MBIs demonstrated significant different outcomes when we separately considered specific control conditions (CBTs: d_w = –0.57 [–0.81 to –0.35], p < 0.001 [n = 2]; TAU: d_w = –1.00 [–1.26 to –0.73], p < 0.001; Q_[2] = 53.19, p < 0.001; I² = 96.24%). Nevertheless, the previous clinical aspects did not -explain the large variability observed in studies -results. On the other hand, we found a robust positive -relationship between effect sizes and the length of treatment.

Publication bias was not revealed. The Orwin’s fail-safe number did not permit to definitely support an advantage of MBIs on anxious symptomatology.

We found a large pooled effect size in decrease of post-traumatic symptomatology when MBIs were compared to other approaches, although the variability across results was significant. The paucity of studies did not permit to investigate possible sources of heterogeneity and to compute bias of publication. Nevertheless, this finding was robust enough in order to conclude in favor of MBIs in reducing these specific symptoms.

A small pooled effect size was observed. Findings were consistent across studies. MBIs seemed to reduce the use of avoidance coping strategies compared to other clinical approaches. We did not find bias of publication. The Orwin’s fail-safe number revealed that the pooled effect size was not robust so as to conclude in favor of a therapeutic effect specifically related to MBIs.

The improvement in post-traumatic symptomatology was significantly larger than the other secondary outcomes considered in the current meta-analytic review (4.55≤ Z ≤13.37; p < 0.001; Bonferroni -correction: α = 0.002). The decreases of negative affectivity and anxious symptoms were significantly greater than outcomes related to overall mental health and the use of avoidance coping strategies (3.72≤ Z ≤6.99; p < 0.001), but they did not significantly -differ from each other and from other depressive -symptoms.

The current meta-analytic review sought to demonstrate the incremental effectiveness of MBIs in AUD and DUDs treatment. In line with this objective, we included RCTs and NRCTs in order to draw conclusions about if and to what extent MBIs could promote additional benefits compared to other treatments usually provided in clinical practice. Additionally, we operationalized the efficacy in relation to primary (i.e., AR, abstinence maintenance, levels of perceived craving) and secondary outcomes (i.e., levels of perceived stress, negative affectivity, overall mental health, the severity of depressive, anxious and post-traumatic symptomatology and the use of avoidance coping strategies) for which there is a large consensus to consider them good indexes of therapeutic success. Eventually, we proposed multiple comparisons among primary and secondary outcomes so as to clarify if MBIs have a generalized effect on the previous dimensions, or they could be more effective for specific domains that are relevant for SUDs treatment.

We observed no difference between treatment conditions when we considered AR an outcome measure, especially when we tested the effect of RCTs. These findings might be in line with the literature that has shown how the dropout factor is the norm rather than the exception in treating SUDs (e.g. [105-108]). This evidence might be ascribed to some patient characteristics such as age and cognitive deficits that represented robust risk factors across several clinical trials with different orientations (for a meta-analytic review see: [53]). Interestingly, we found a significant difference in pooled effect sizes when RCTs and NRCTs were compared. Specifically, when MBIs were carried out as an NRCT, they seemed to exhibit a benefit, albeit modest, in reducing the dropout phenomenon. This result may reflect the lack of control of a crucial variable, which has been related to treatment retention in SUDs and it has also been involved in engaging in mindfulness practices that refer to motivation. To explain in detail, lower motivation was related to a higher dropout rate [109-111]; it was also demonstrated how the personal intention in meditation practice is a core aspect in order to identify one of the potential mechanisms to explain how mindfulness affects positive change [112-115]. Further, Mascaro et al. [116] showed how preexisting brain functioning predicts the consequent practice of mindfulness during a Cognitively-Based Compassion Training. Taken the previous considerations together, we might conclude that patients assigned to MBIs in NRCTs could be characterized by preexisting conditions, both neural and psychological, that facilitate the learning and practice of mindfulness abilities and sustain their engagement in treatment retention.

Additional variables that significantly influenced the AR referred to sample characteristics and type of MBIs. Particularly, MBIs seemed to show a slightly better treatment retention than other programs when they were carried out to treat mixed SUDs samples. Conversely, other approaches demonstrated less attrition, although nonsignificant, than MBIs in treating homogeneous SUDs samples. Although we might conclude that MBIs seemed to show preliminary advantages in sustain treatment retention when they were carried out to simultaneously treat several SUDs, future studies are needed in order to clarify which therapeutic strategies related to mindfulness approaches might be implicated in explaining this result. We might also extend the previous considerations regarding the relevance of studying therapeutic strategies in treatment retention to the difference observed between types of MBIs. To explain in detail, it seems that when MBIs are combined with standard programs show slightly less AR than manualized MBIs. Taking into consideration this finding, we might postulate that the combination of standard therapeutic strategies (e.g., relapse prevention skills, motivational enhancement interventions) with mindfulness principles (e.g., acceptance attitude) could reinforce the motivation to stay in treatment and reduce relevant interference factors (e.g., Abstinent Violation Effect) to treatment retention. Nevertheless, empirical process-outcome studies are necessary to demonstrate the previous clinical consideration.

As a whole, even though treatment retention seemed to be independent of clinical orientations and settings, considering the small benefit of MBIs in reducing AR in relation to specific conditions, we support a detailed pre-treatment assessment of SUDs co-diagnoses, motivational processes, and cognitive functioning in order to recognize subjects who would be the best candidates for these types of intervention [49].

MBIs seemed to consistently promote a slightly better abstinence. It is well established that several neurocognitive aspects related to impulsivity explain the ability to successfully achieve and maintain abstinence during and following addiction treatments [117]. Several empirical studies also demonstrated the existence of a large relationship between dispositional mindfulness and traits related to impulsivity (e.g., [118-120]). An overlap between some mindfulness abilities and impulsivity in explaining the levels of alcohol consumption [121] and alcohol use motivations was also found [122, 123]. Furthermore, it was observed how MBIs have an effect in reducing levels of impulsivity in several clinical and nonclinical samples (e.g., [124-127]). Consequently, we might hypothesize how the slight advantages of MBIs in supporting abstinence in SUDs treatment could be ascribed to the reduction of several forms of impulsivity that are related to lapse and relapse in addictive behaviors. Even though this conclusion is in line with mindfulness theoretical assumptions in addiction treatment (e.g.. [31-32]), these considerations were not robust enough as indicated by the Orwin’s fail-safe number. Consequently, future research in MBIs efficacy should focus on process-outcome studies that are needed to prove this hypothesis.

MBIs seemed to show large therapeutic effects in reducing levels of perceived craving in comparison with other approaches. During MBIs, patients are encouraged to bring consciousness to experience of craving and to learn to observe it without judgment and without expression any reaction [78], thereby reducing the activation of neural correlates of craving [128]. However, the heterogeneity of results was large. This variability was partially explained considering whether clinical trials were carried out to specifically treat the co-occurrence of SUDs and psychiatric disorders. Referring to the Orwin’s fail-safe number, we can conclude that MBIs are effective programs in reducing levels of craving when they were provided to treat SUDs in comorbidity with other psychiatric disorders. Conversely, no significant differences between treatment orientations were found in craving changes when interventions aimed to exclusively treat SUDs. Given the well-documented therapeutic effects of mindfulness in reducing specific psychiatric symptoms (i.e., depressive, anxious, post-traumatic experiences) [156, 161, 162], these results could be ascribed to secondary effects of MBIs on such symptoms, which were related to craving episodes [129-134], especially in individuals with co-occurring psychiatric disorders and SUDs [135, 136]. This assumption is also corroborated by our results concerning the comparison between MBIs and other treatment approaches in reducing depressive symptoms among SUDs individuals. Particularly, we exclusively found large advantages in favor of MBIs when they were specifically carried out to treat the co-occurrence of SUDs and other psychiatric disorders. Although MBI’s therapeutic effects on the decrease of levels of perceived craving seem to be limited to individuals who are affected by SUDs and other psychiatric disorders, such clinical target represents the best primary outcome when MBIs were compared with other active programs.

In light of all the previous considerations, we might preliminarily conclude that formal mindfulness practices, which represent the core feature that differentiate MBIs from treatments usually provided in clinical practice, could be considered effective craving-related coping skills, especially when craving episodes are functionally associated with other psychiatric symptoms. However, future research on MBIs in SUDs should systematically investigate the levels of perceived craving as an outcome measure so as to empirically clarify the role of mindfulness practice, and clarify in detail which mindfulness abilities are implicated, in reducing and/or managing this aspect that it is considered with a large consensus as one of the strongest predictors of relapse in addiction treatment (e.g., [137-139]).

Consistent with a large amount of literature (e.g., for a meta-analysis see: [140]), MBIs seemed to show consistent and small-to-moderate advantages in reducing levels of perceived stress. It was postulated how the ability to observe situations and thoughts nonjudgmentally without reacting to them impulsively, helps people to develop a more reflexive awareness of inner and outer experiences, and it could represent an efficacious tool for the reduction of stress [24, 141]. Even though this dimension might represent a promising treatment outcome, particularly in relation to the well-established role of stress in inducing craving and relapse (e.g., [142-144]), the difference between clinical conditions in reducing the level of perceived stress is not robust enough to definitively conclude in favor of MBIs. One possible explanation of the small difference between MBIs and other treatments might be related to a primary effect of detoxification itself (e.g., reduction of withdrawal symptoms) rather than a specific psychotherapeutic effect. As a consequence, future research should investigate which psychological processes could be exclusively improved by MBIs in response to discomfort and stress and how they are implicated in craving management and relapse prevention.

We observed great advantages in reducing the negative affectivity associated with MBIs, even though only one study [84] explained the pooled effect size. Generally speaking, this finding is principally consistent with the results from the empirical literature that demonstrated how mindfulness programs produced their benefits in clinical and nonclinical samples by decreasing the negative affect and by improving the positive affect (e.g., [26, 145]), as well as by enhancing emotion regulation [146-148]. It is well known, how one of the most prominent risk factors for craving and relapse in SUDs is a negative factor (e.g., [149-151]). Consistently, it is possible to assume how MBIs could work on several processes, both psychological and neural [146-148], that might prevent relapse reducing levels of negative emotionality. Nonetheless, the large variability of results, the paucity of studies that specifically investigate the previous aspect, and the extent of pooled effect size did not permit to draw definitive conclusions. Consequently, given the central role of negative affect in relapse prevention, future clinical research in SUDs treatment should systematically include the evaluation of this dimension as a secondary outcome, especially assessing its temporal stability after the end of interventions.

We observed no significant differences between MBIs and other approaches in improving the overall mental health. We might assume how this finding could reflect the consequence of abstinence. It is well established how a wide range of psychiatric symptoms presented by SUDs individuals at admission to treatment are associated with substance intoxication and show a quick remission during and after the treatment (e.g., [152-154]). Therefore, we may conclude that it is needed to demonstrate if mindfulness practice could really represent a long-term protective factor for relapse in psychopathology among SUDs subjects, as showed in other clinical populations (e.g., [155]) that represented an antecedent of relapse in substance use (e.g., [136]).

However, when we considered specific psychopathological symptoms, we found significant dissimilar findings. Considering depressive symptomatology, we observed moderate-to-large improvements when MBIs were compared to other treatments. However, we revealed the existence of a large variability across studies, several sources of heterogeneity, and pooled effect size was not large enough to consider it as definitely robust. In detail, MBIs showed significant larger benefits in reducing depressive symptoms when it was considered specific clinical features: (a) MBIs seemed to be more effective when they were carried out to treat the co-occurrence of SUDs and other psychiatric disorders; (b) MBIs supported larger improvements in mixed SUDs samples than in homogeneous populations; (c) MBIs sustained a larger decrease in depressive symptomatology when they were provided in a group setting than in a combined setting (i.e., group + individual).

The efficacy of MBIs is well supported by several RCTs in treating depression [158]. Taking into consideration such evidences and our results, we might preliminarily conclude that MBIs are also effective in reducing depressive symptoms among AUD and DUDs populations, especially when they co-occur with other psychiatric disorders. Additionally, given the controversial results regarding the effects of clinical settings in treating depression [157] and the well-established efficacy of peer-based groups in SUDs [158, 159], future research should explore possible therapeutic factors that sustain the efficacy of MBIs in reducing depressive symptoms when they are exclusively carried out in a group setting, and which interference processes could be associated to the patient-therapist relationship [160].

Consistent with data regarding the efficacy of MBIs in treating anxiety disorders [161], we might partially extend similar considerations to the decrease of anxious symptoms among SUDs individuals. Particularly, even though we observed large variability in studies results and not conclusive findings, we observed large differences between MBIs and TAU control conditions, as well as when long-term effects of MBIs were evaluated. Specifically, as demonstrated in other clinical populations [163], MBIs seemed to promote better long-term therapeutic effects than short-term ones in reducing anxious symptoms. Moreover, the significant difference observed in pooled effect sizes when we separately considered TAU and CBT control conditions might reflect the demonstrated efficacy of CBT interventions in treating anxiety disorders, even when they co-occur with SUDs [168].

Eventually, robust findings in favor of MBIs were found when we took into consideration the severity of post-traumatic symptoms. In line with the therapeutic efficacy of MBIs in treating post-traumatic stress disorder (PTSD) [162] and values of effect sizes found in the current meta-analytic review, we can conclude that such interventions might be considered an effective alternative in treating the co-occurrence of SUDs and PTSD.

In relation to the decrease of avoidance coping strategies, we found a consistent and small benefit associated with MBIs compared to other treatment approaches. Even though it was demonstrated that MBIs were more effective programs than other interventions in reducing avoidance coping strategies among several clinical populations [163-166], we cannot support the same conclusion among SUDs. This finding might be in line with Arch and Craske [167] who argued that both cognitive restructuring (CBT process) and cognitive defusion (e.g., ACT process) aim to decrease avoidance and enhance exposure to previously avoided and suppressed internal experiences. Consequently, we might consider the reduction of avoidance coping strategies as common therapeutic dimension that different therapeutic orientations seem to equally address with dissimilar techniques.

Eventually, taking into consideration results from multiple comparisons within secondary outcomes, we conclude that the most robust therapeutic effect of MBIs refers to the decrease of post-traumatic symptoms. Considering comparable pooled effect sizes of depressive/anxious symptoms and negative affectivity, we might also assume that MBIs improve several aspects related to the emotional well-being of patients.

In conclusion, MBIs seemed to show clinically significant, albeit preliminary, advantages compared to other clinical approaches when it was considered specific relapse factors that refer to anxious and depressive symptoms, especially when SUDs co-occur with other psychiatric disorders. Furthermore, we might assume that MBIs are valid and effective therapeutic alternatives when SUDs individuals are affected by PTSD. Formal mindfulness practices could be considered additional craving-related coping strategies, particularly for dual-diagnosis individuals. Modest, although significant, benefits in favor of MBIs were detected in relation to abstinence. Treatment retention was independent of the therapeutic approach. Eventually, taking into consideration null differences regarding AR between MBIs and other active programs, we also conclude that the benefits described above should be exclusively applied to patients who go through the complete course of treatment.

We wish to acknowledge the work of an expert English mother tongue translator on this document.

The authors report no relevant financial conflicts. This meta-analytic review did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors. The authors alone are responsible for the content and writing of this paper.