Individual gender and group gender composition as predictors of differential benefit from mindfulness-based relapse prevention for substance use disorders

Abstract

Objectives:

Mindfulness-based relapse prevention (MBRP) is an effective group-based aftercare treatment for substance use disorders (SUDs), yet few studies have examined moderators of MBRP efficacy. This secondary data analysis evaluated individual gender and group gender composition (e.g., proportion of women relative to men in each therapy group) as treatment moderators of MBRP.

Methods:

The analysis sample included 186 individuals with SUDs randomized to MBRP or relapse prevention (RP) as an aftercare treatment. Outcomes included number of heavy drinking days and drug use days at the 12-month follow-up.

Results:

There were no treatment moderation effects for models with heavy drinking days as the outcome (all ps > .05). Group gender composition, but not individual gender, moderated the effect of treatment condition on drug use days (p < .01). Individuals who received MBRP had significantly fewer drug use days at 12-months than those who received RP, but only among individuals in therapy groups comprising one-third or more women (p <0 .0001). Specifically, all women and men who received MBRP in groups with one-third or more women were abstinent from drugs at month 12, whereas those in RP groups with one-third or more women had an average of about eight drug use days at month 12 (corresponding to a large between-treatment condition effect size).

Conclusions:

Group-based MBRP may be more efficacious than group-based RP, particularly when women compose at least one-third of the therapy group. Further research is warranted on gender and group gender composition as moderators of MBRP.

Sex and gender are important to consider when studying the development and treatment of substance use disorders (SUDs; Becker, McClellan, & Reed 2017; Greenfield et al. 2007). For example, substance use tends to escalate more quickly for women as compared to men, with more rapid progression from first use to habitual use to seeking treatment (Bobzean, DeNobrega, & Perrotti 2014; Hernandez-Avila, Rounsaville, & Kranzler 2004). Women with SUDs have higher prevalence rates of co-occurring psychiatric disorders including mood, anxiety, eating and post-traumatic stress disorders (Khan et al. 2013; Najavits, Weiss, & Shaw 1997; Zilberman, Tavares, & Blume 2003). Furthermore, women face gender-specific barriers to entering treatment, such as pregnancy, fear of losing custody of their children, and childcare responsibilities (Greenfield, Brooks, et al. 2007), and women with SUDs tend to be more stigmatized than men (Greenfield, Brooks, et al. 2007; Khan et al 2013). Although relapse rates are similar across genders, relapse among women is more often related to negative affect, relationship stress, and trauma (Greenfield, Brooks, et al. 2007; Hyman et al. 2008; Walitzer & Dearing 2006).

The majority of addiction treatment is delivered in the format of group therapy, which has generally been shown to be effective (Weiss, Jaffee, de Menil, & Cogley 2004). In group therapy, both an individual’s gender and the group gender composition may affect treatment experience and outcome. Single-gender and mixed-gender group therapy for SUDs are both associated with reductions in substance use following treatment (Greenfield et al. 2014), with some limited evidence that substance use reductions may be better maintained for women following single-gender group therapy (Greenfield, Trucco, McHugh, Lincoln, & Gallop 2007). In studies comparing a single-gender women’s recovery group to a mixed-gender drug counseling group, women with higher psychiatric symptom severity had fewer substance use days if they received single-gender rather than mixed-gender treatment (Greenfield et al. 2008). Furthermore, in the mixed-gender group, women with greater self-efficacy demonstrated better substance use outcomes than women with lower self-efficacy (Cummings, Gallop, & Greenfield 2010). Taken together, these findings suggest a relationship between group gender composition, psychological symptom profiles, and substance use outcomes.

Group gender composition may be important to consider in group therapy research because it may impact group-level processes, such as group cohesion. Group cohesion has been associated with improved therapeutic outcomes in general group therapy research (Burlingame, McClendon, & Alonso 2011). Greenfield, Sugarman, and colleagues found that the number of affiliative statements (e.g., supportive statements, acknowledgement of shared experience, and strategy suggestions) verbalized between group members was greater in women’s single-gender groups than in mixed-gender groups (Greenfield, Kuper, Cummings, Robbins, & Gallop 2013; Sugarman et al. 2016). Furthermore, the same authors demonstrated that within a mixed-gender group, women made more affiliative statements than men, and the most frequent direction of affiliative statements was from women to men. Given the seemingly greater contribution from women than men to group affiliative processes, the percentage of women in groups may affect group cohesion. Indeed, the broader group process literature suggests that higher percentages of women in groups is associated with greater group emotional intelligence, decreased relational conflict, affective similarity (Curşeu, Pluut, Boroş, & Meslec 2015), as well as group cognitive complexity (Curşeu & Sari 2015) and collective intelligence (Woolley, Chabris, Pentland, Hashmi, & Malone 2010).

Qualitative experiences within group treatment for SUDs also vary as a function of group gender composition. Compared to women in mixed-gender groups, women in single-gender groups endorsed greater feelings of safety, intimacy, empathy, and honesty, plus the ability to focus on gender-relevant recovery topics (Greenfield, Cummings, Kuper, Wigderson, & Koro-Ljungberg 2013). Moreover, women in mixed-gender groups reported that women were more active contributors than men, and that women tended to prompt greater communication among men (Greenfield et al. 2013). In adolescent alcohol prevention groups, a greater proportion of girls in the alcohol prevention group was associated with more engagement and group satisfaction among both girls and boys (Garcia, Bacio, Tomlinson, Ladd, & Anderson 2015).

One group-based therapy for SUDs is mindfulness-based relapse prevention (MBRP), a protocolled aftercare program which integrates mindfulness meditation training with components of cognitive behavioral relapse prevention (Bowen, Chawla, & Marlatt, 2011). Multiple randomized controlled trials of MBRP support its effectiveness as an aftercare treatment for SUDs (Bowen et al. 2009, 2014; Li et al. 2017; Witkiewitz et al. 2014). The role of gender in predicting differential response to mindfulness-based interventions (MBIs) for SUDs, however, is still not clear. In a systematic review of gender differences in MBIs, Katz and Toner (2013) found that most existing randomized trials of MBIs for SUDs have not evaluated individual gender as a treatment moderator, and in the single randomized trial that did (Bowen et al. 2009), there was no significant moderation effect. However, Katz and Toner (2013) also found that quasi-experimental studies and case series demonstrate that women are more attracted to MBIs than men. Hence, they recommend further research on the role of gender in MBIs for SUDs in larger randomized trials. Despite preliminary research on individual gender as a moderator of SUD treatment outcomes, it appears that no studies to date have examined group gender composition as a moderator of MBIs for SUD.

Accordingly, the current study evaluated both individual gender and group gender composition (e.g., the proportion of women relative to men in each therapy group) as potential treatment moderators for drug and alcohol use outcomes in the largest randomized controlled trial of MBRP to date. We did not put forth a hypothesis for individual gender as a treatment moderator. Prior research suggests individual gender does not impact the efficacy of MBRP (Bowen et al., 2009) and because MBRP content was not designed for any particular gender, MBRP content may be equally relatable for men and women. Research suggesting that women may be more attracted to MBIs (Katz & Toner, 2013) raises the possibility that women might be more engaged in MBRP and thus benefit more. We also did not put forth a hypothesis for group gender composition as a treatment moderator. There is no prior research indicating that group gender composition influences the efficacy MBIs and MBRP therapists are trained to engage participants and facilitate group cohesion, regardless of group composition. The broader psychotherapy literature indicates that groups with more women have greater group cohesion, a factor which may influence the efficacy of MBRP, given that MBRP involves considerable disclosure of personal experiences and observations related to personal mindfulness practice.

Method

Participants

The current study was a secondary data analysis drawing from a recent trial (Bowen et al. 2014) in which 286 individuals with SUDs were randomly assigned to one of three treatment conditions: MBRP, relapse prevention (RP), or treatment as usual. Inclusion criteria were: age 18 or older, fluency in English, medical clearance, ability to attend treatment sessions, agreement to random assignment, and prior completion of either intensive outpatient or inpatient care for SUD. Exclusion criteria were: current psychotic disorder, dementia, suicidality, imminent danger to others, or participation in prior MBRP trials. The current study only included participants assigned to MBRP or RP because there were no data available on group gender composition for the treatment as usual groups. Five participants assigned to MBRP or RP were missing data on gender and were not included in these analyses, for a final sample of N = 186.

Procedures

Participants were recruited from two sites within the same community SUD treatment agency. Research staff collected all data during private sessions. One set of therapists (n = 10) delivered MBRP and a separate set of therapists (n = 9) delivered RP. MBRP (Bowen et al., 2011) was delivered in a closed-cohort group format over the course of 8 weeks (2-hour group sessions once per week). Two therapists facilitated each group session. Four MBRP therapists were clinical psychologists, one was in a doctoral training program, and five had Master’s degrees. Key features of MBRP include: a) guided in-session mindfulness meditations (e.g., body scan) and imaginal exposure practices (e.g., urge surfing), b) discussion about participants’ direct experiences during and following meditation practices, c) discussion about how mindfulness principles relate to substance use relapse and recovery, d) incorporation of cognitive-behavioral relapse principles into discussion and home assignments (e.g., monitoring of urges), and e) assignment of formal and informal (i.e., “on-the-go”) mindfulness home practices. The RP intervention was based on a cognitive-behavioral relapse prevention program (Monti et al. 2002), adapted to match MBRP in time, format, size, location, and amount of assigned homework. RP focuses on teaching participants coping and problem-solving skills to manage key triggers and life stressors. Six RP therapists were clinical psychologists, one was in a doctoral training program, and two had Master’s degrees.

Measures

Demographics.

Gender, age, race, educational level, and employment status were assessed with a self-report questionnaire. The questionnaire asked about gender (male, female, or “other” options) and allowed a free text response for “other.” One participant selected “other” and four individuals did not provide an answer for the gender question. None of the participants with “other” or missing data for the gender question provided a free text response, thus these five individuals were excluded from subsequent analyses, and gender was treated as binary in analyses. The questionnaire did not inquire about sex assigned at birth or gender identity. Given that participants were explicitly asked about gender, we use the term “gender” throughout this paper.

Substance use disorder severity.

The Severity of Dependence Scale, a 5-item self-report measure (SDS; Gossop et al. 1995), was used to assess SUD severity at baseline. The SDS demonstrated good internal consistency reliability in the current sample (Cronbach’s α = 0.85).

Substance use.

The Timeline Followback interview (Sobell & Sobell 1992) was used to assess substance use, including use of alcohol and drugs. The Timeline Followback is calendar-based method for assessing substance use. Primary outcomes were number of drug use days and number of heavy drinking days (defined as 4+/5+ standard drinks for women/men) over the 90-day period prior to the 12-month post-treatment follow-up assessment.

Statistical Analyses

Mplus Version 8 (Muthén & Muthén 2012) was used for primary study analyses. Primary substance use outcomes were drug use days (DUD) and heavy drinking days (HDD) (coded as 4+/5+ standard drinks for women/men) during the 90-days prior to the 12-month post-treatment assessment. We conducted negative binomial regression analyses to examine the main effects of treatment, gender, and gender composition on substance use outcomes, and to examine gender and group gender composition as moderators of the effect of treatment on substance use outcomes. We chose negative binomial regression models to account for a high frequency of zero values for substance use outcomes. For all regression models, we used all available data and parameters were estimated with full information maximum likelihood (Witkiewitz et al. 2014). Additionally, we used the sandwich estimator (White 1980) in Mplus to account for dependency in the data due to clustering by therapy group. We chose a cut-off of one-third women in a given therapy group to reflect the total proportion (54 out of 186; 29%) of women in the sample, which was approximately one-third.

For all models, we controlled for age, baseline dependence severity, number of prior treatment episodes, treatment hours completed, and treatment site because these covariates were used in the main outcome paper of the parent trial (Bowen et al. 2014). We also controlled for race, given recent findings of racial differences in treatment outcomes following MBRP (Greenfield et al., 2018; Witkiewitz, Greenfield, & Bowen 2013). In models in which individual gender was the moderator variable, we controlled for group gender composition, and vice versa. To probe significant interactions, we tested the effect of treatment at each level of the moderator using separate negative binomial regression models (i.e., simple slopes analysis). For the interaction models and follow-up simple slope analyses by group, we set statistical significance at p < .01.

For analyses that examined treatment effects, we quantified the magnitude of these effects by reporting effect sizes in terms of incidence rate ratios (IRRs). IRRs can be interpreted as the rate of increase (when the IRR is above 1.0) or the rate of decrease (when the IRR below 1.0) in heavy drinking or drug use days for a 1-unit increase in the predictor (with other predictors in the model held constant). An IRR of approximately 1.44 (for IRRs above 1), or 0.69 (for IRRs below 1), corresponds to a Cohen’s d effect size of 0.2 (small effect); an IRR of approximately 2.48 or 0.40 corresponds to a Cohen’s d effect size of 0.5 (medium effect); and an IRR of approximately 4.27 or 0.23 corresponds to a Cohen’s d effect size of 0.8 (large effect; Borenstein, Hedges, Higgins, & Rothstein 2009).

Results

Table 1 presents the descriptive statistics both in the full available sample (N = 186) and by individual gender and group gender composition. There were no significant differences on these baseline and demographic variables by individual gender, except for substance dependence severity. Compared to men, women had significantly higher substance dependence severity scores at baseline. Compared to groups with more than two-thirds men, groups with one-third or more women had a higher proportion of individuals with at least some college education and a lower proportion of racial/ethnic minorities.

Table 1.

Descriptive Statistics for Study Variables, n (%) or Mean (standard deviation)

	Men (n = 132)	Women (n = 54)	Individuals in groups with one-third or more women (n = 90)	Individuals in groups with more than two-thirds men (n = 96)	Full Available Sample (MBRP and RP conditions only; n = 186)
Self-reported Race/Ethnicity
Non-Hispanic White	71 (53.8%)	29 (53.7%)	41 (46.1%)*	57 (60.6%)*	100 (53.7%)
Racial/Ethnic Minority	61 (46.2%)	25 (46.3%)	48 (53.9%)*	37 (39.4%)*	86 (46.2%)
Age	38.34 (11.05)	40.73 (10.41)	11.15 (10.98)	11.97 (12.51)	39.03 (10.89)
Educational Level
High School Degree/GED or Less	71 (54.2%)	25 (46.3%)	39 (43.8%)*	57 (59.4%)*	96 (51.9%)
Completed at Least some College	60 (45.8%)	29 (53.7%)	50 (56.2%)*	39 (40.6%)*	89 (48.1%)
Unemployed	87 (66%)	31 (57.4%)	34 (37.8%)	34 (35.4%)	118 (63.4%)
Abstinence from Drug Use at Baseline	120 (95%)	49 (90.7%)	85 (98.8%)	84 (90.3%)	174 (94.6%)
Abstinence from Drinking at Baseline	120 (95)	49 (90.7%)	81 (94.2%)	88 (94.6%)	174 (94.6%)
Substance Dependence Severity	9.34 (4.12)*	11.61 (4.88)*	10.15 (3.84)	9.56 (4.16)	9.85 (4.01)
Prior Treatment Episodes	1.68 (1.59)	1.47 (1.50)	1.52 (1.30)	1.70 (1.77)	1.62 (1.56)
Treatment Hours Completed During the Study	12.50 (7.51)	11.61 (4.88)	12.51 (5.40_	11.97 (7.99)	12.23 (6.85)

Note. MBRP = Mindfulness-Based Relapse Prevention. RP = Relapse Prevention.

^*= significant difference (p <.05) between men and women based on t-tests or chi-square tests

Among the 186 participants, there were 22 different therapy groups, 11 (50%) groups comprising one-third or more women, and the other 11(50%) comprising more than two-thirds men. In the sample, 90 (48.4%) participants were in groups with one-third or more women and 96 (51.6%) were in groups with more than two-thirds men. In the MBRP treatment condition, five out of 12 therapy groups had one-third or more women. In the RP condition, six out of 10 therapy groups had one-third or more women. Across treatment conditions, the average group size was 8.76 individuals (SD = 1.73). The average MBRP group size was 8.8 (SD = 2.2), and the average RP group size was 8.72 (SD = 1.27).

Main Effects of Treatment

In the current sample, treatment condition (coded 0 = RP, 1 = MBRP) was not a significant predictor of number of heavy drinking days (HDD) (B (SE) = −0.62 (0.85), p = 0.465, IRR = 0.54; 95% CI: 0.10, 2.89) at the 12-month follow up. However, treatment condition significantly predicted number of drug use days (DUD) (B (SE) = −1.85 (0.57), IRR = 0.16 (95% CI: 0.05, 0.48), p <0.001), such that MBRP participants reported 84% fewer DUD than RP participants.

Main Effects of Individual Gender and Group Gender Composition

Across treatment conditions, individual gender (coded 0 = men, 1 = women) was not a significant predictor of HDD (B (SE) = −0.55 (0.66), p = 0.41) or DUD (B (SE) = 1.34 (0.73), p = 0.07). Group gender composition (coded 0 = group with more than two-thirds men, 1 = group with one-third or more women) was also not a significant predictor of HDD (B (SE) = 0.15 (1.10), p = 0.89) or DUD (B (SE) = −2.02 (1.33), p = 0.13).

Interaction Effects

Moderation of Treatment by Gender.

As seen in Table 2, there was no significant interaction between individual gender and treatment in the prediction of HUD or DUD.

Table 2.

Summary of Results Across Interaction Models

	12-Month Heavy Drinking Days	12-month Drug Use Days
Predictors	B (SE)	B (SE)
Treatment X Gender	−2.11 (1.50)	−1.33 (1.33)
Treatment X Group Gender Composition	−1.41 (1.56)	−16.20 (1.49)**

Note. ** = p < .01. B =unstandardized regression coefficient; SE = standard error. For all interaction models, we controlled for: age, site, treatment hours completed, prior treatment episodes, baseline substance dependence severity, and race.

Moderation of Treatment by Group Gender Composition.

As seen in Table 2, there was not a significant interaction between group gender composition and treatment in the prediction of HDD. However, there was a significant interaction between group gender composition and treatment in the prediction of DUD. Follow-up simple slope analyses to probe the interaction revealed the following: a) among individuals in groups with one-third or more women, there was a significant effect of treatment (coded 0 = RP, 1 = MBRP) (B = −16.65 (0.786), IRR = 0.00 (95% CI:,0.00, 0.00), p <0.0001), such that MBRP participants reported 100% fewer drug use days than RP participants, and b) among individuals in groups with more than two-thirds men, treatment was not a significant predictor of DUD (B = 0.87 (0.76), IRR = 2.39 (95% CI:,0.54, 10.57), p = 0.25)

Table 3 presents the mean scores on HDD and DUD by treatment condition among defined subgroups based on individual gender and group gender composition. As shown in Table 3, among individuals in groups with one-third or more women, all MBRP participants (including both men and women) reported zero DUD, whereas RP participants reported an average of 8.80 DUD (SD = 23.10).

Table 3.

Mean Scores on Outcomes by Treatment Condition among Full Sample and among Subgroups

	Mean (SD) Heavy Drinking Days	Mean (SD) Drug Use Days
Full available sample (n = 186)	MBRP: 1.44 (7.66)	MBRP: 3.06 (15.08)
	RP: 3.88 (12.17)	RP: 6.10 (19.05)
Women (n = 54)	MBRP: 0.95 (4.24)	MBRP: 4.65 (20.10)
	RP: 3.14 (9.41)	RP: 11.49 (27.19)
Men (n = 132)	MBRP: 1.61 (8.53)	MBRP: 2.73 (13.70)
	RP: 4.41 (13.90)	RP: 2.46 (9.07)
Individuals in groups with one-third or more women (n = 90)	MBRP: 0.79 (3.58)	MBRP: 0.00 (0.00)
	RP: 4.07 (11.32)	RP: 8.80 (23.10)
Individuals in groups with more than two-thirds men (n = 96)	MBRP: 1.80 (9.17)	MBRP: 4.70 (18.55)
	RP: 3.54 (13.93)	RP: 1.12 (4.28)
Women in groups with one-third or more women (n = 39)	MBRP: 1.58 (5.48)	MBRP: 0.00 (0.00)
	RP: 3.91 (10.47)	RP: 14.43 (29.93)
Women in groups with more than two-thirds men (n = 15)	MBRP: 0.00 (0.00)	MBRP: 11.63 (31.68)
	RP: 0.17 (0.41)	RP: 0.00 (0.00)
Men in groups with one-third or more women (n = 51)	MBRP: 0.19 (0.40)	MBRP: 0.00 (0.00)
	RP: 4.22 (12.34)	RP: 3.17 (11.36)
Men in groups with two-thirds or more men (n = 81)	MBRP: 2.14 (9.97)	MBRP: 3.74 (15.98)
	RP: 2.14 (9.97)	RP: 1.56 (5.00)

Note. MBRP = Mindfulness-Based Relapse Prevention; RP = Relapse Prevention.

Sensitivity Analyses.

We conducted several sensitivity analyses. First, when conducting the interaction models with a continuous moderator variable, (i.e., percent of women in each therapy group), rather than a binary variable (i.e., groups with one-third or more women versus groups with two-thirds or more men), the results remained substantively unchanged. Second, given the significant difference between the gender composition groups in the proportion of those with at least some college education, we conducted sensitivity analyses in which we conducted the interaction models with education included as a covariate. The results remained substantively unchanged, even when controlling for education. Third, we conducted additional analyses and controlled for clustering of individual therapists. The results remained substantively unchanged. Finally, when conducting the interaction models with substance of choice included as a covariate, the results remained substantively unchanged.

Discussion

We evaluated group gender composition as a moderator of a group-based mindfulness-based intervention (MBI). Our analyses demonstrate that individuals who received mindfulness-based relapse prevention (MBRP) during aftercare substance use disorder (SUD) treatment had fewer drug use days 12 months following treatment than those who received RP, and this effect was most pronounced when the therapy group comprised one-third or more women. Both men and women benefited from receiving MBRP in therapy groups with one-third or more women, with 100% of these individuals abstinent from drugs at month 12 (corresponding to a large between-treatment condition effect size). Hence, the key finding from this study is that group-based MBRP may have a large advantage over group-based RP in reducing long-term drug use, particularly when therapy groups comprise at least one-third women.

Our study sheds light on contextual conditions within therapy groups that may optimize the longer-term effect of MBRP for SUDs. Yet findings do not elucidate why MBRP groups with one-third or more women may be more effective than RP groups with similar composition. First, it is important to note that the group therapy literature suggests that groups with more women exhibit greater group cohesion, which has been defined as the feeling of belonging or fitting in with a group (Greenfield et al., 2013; Sugarman et al., 2016). There may be specific factors that contribute to a greater sense of group cohesion in groups with more women, including more affiliative or supportive statements generated by women (Greenfield et al., 2013; Sugarman et al., 2016), increased emotional intelligence, affective similarity, decreased interpersonal conflict (Curşeu et al., 2015), cognitive complexity (Curşeu & Sari, 2015) and collective intelligence in groups with greater proportions of women members (Woolley et al., 2010). It is possible that relative to RP, therapeutic effects of MBRP may be more dependent on these group contextual conditions. For example, a key treatment component of MBRP is group inquiry, which involves the therapist eliciting and reinforcing client statements about direct sensory and emotional experiences during and following mindfulness practices, especially challenging experiences (e.g., intrusive thoughts, physical pain, difficult emotions; Bowen et al, 2011). Hence, it is possible that groups with more women foster greater group cohesion, which in turn may assist the MBRP therapist in effectively facilitating the group inquiry process. Effects may also be due to an increased sense of safety or cohesion with other members of often oppressed or underrepresented communities. A recent study (Greenfield et al., 2018) suggested better outcomes following MBRP as compared to RP when individual race/ethnic status was reflected in the group race/ethnicity (e.g., ethnic/racial minority individuals in groups comprising more than half minorities). Importantly, however, neither that nor the current study assessed group cohesion or other therapy group process variables. Future research to better understand group-level moderators and mediators of MBRP is warranted, including qualitative approaches like coding within session group processes or directly interviewing individuals in the group to understand experiences and perceptions of the group sessions.

We also examined individual gender as a treatment moderator. Consistent with prior findings (Bowen et al., 2009), this study found no significant effect of individual gender on treatment response to MBRP. Of note, because both men and women benefited from receiving MBRP in groups with one-third or more women, findings suggest that individual gender did not interact with group gender composition in predicting treatment response (i.e., it was not the case that only women, but not men, benefited from being in MBRP groups with one-third or more women). Our study suggests that gender is a factor that influences MBRP efficacy, but potentially via within-group gender dynamics, rather than other mechanisms, such as individual gender influencing one’s interest in and ability to relate to MBRP. MBRP was designed to be equally relevant for both men and women. However, it is important to note that individual gender might function differently as a moderator variable when MBRP is delivered individually instead of in groups. There is preliminary research that women may be more interested in MBIs than men (Katz & Toner, 2013) and interest with MBRP was not measured in the current study.

It is also important to note that while the interaction of treatment and group gender composition predicted drug use days, it did not predict heavy drinking at month 12. It is not clear why this was the case. Future research on MBRP and group-level moderators and mechanisms may need to further consider type of substance (e.g., alcohol vs. illicit drugs) because group dynamics may be influenced by the type of substance being discussed and number of individuals within a group who share the same drug of choice.

Alcohol and drug relapse rates following treatment are generally consistent across genders, yet there is evidence that for women, relationship stress, trauma history, and negative affect are stronger predictors of alcohol and drug relapse than they are for men (Greenfield, Brooks, et al., 2007; Hyman et al., 2008; Walitzer & Dearing, 2006). This study’s findings that both men and women reported fewer drug use days following MBRP if they were in groups comprised of more than one-third women suggest that these groups adequately addressed factors related to relapse in SUD. Previous research demonstrated that MBRP attenuated the relationship between post-treatment depressive symptoms and craving and subsequent substance use (Witkiewitz & Bowen, 2010). Therefore, future research may be needed to extend these findings and clarify the specific ways in which groups with higher proportions of women may have decoupled the connection between relationship stress, trauma, and negative affect from return to substance use in women.

Limitations and Future Research

This study had limitations, including the rather narrow assessment of gender. Participants self-reported gender using three options: female, male, or other. No data were collected about sex assigned at birth or more nuanced aspects of gender identity or gender roles that may affect SUD presentation or treatment. Furthermore, the original study did not systematically randomize gender composition of the treatment groups. Another limitation of the current study was that we selected the proportion of one-third women because that was equivalent to the amount of women in the full sample. The sample size in the current study was not large enough to examine the effect of different proportions of women (e.g., 50% women in a therapy group). Future research could examine whether different proportions of women in a treatment group may have more or less of an impact on treatment outcomes. The number of therapy groups with one-third or more women was not completely balanced across the two treatment conditions (six out of 10 groups in RP and five out of 12 groups in MBRP), which could have impacted the results. Future studies focused on the impact of therapy-group-level factors could seek to balance the group-level variables across treatment conditions. Lastly, we did not have a measure of group-level processes (e.g., group cohesion). Examining why men and women benefit more from MBRP when there are more women present in the group is an important future direction for clarifying mechanisms of change in MBRP. While individual gender did not moderate the efficacy of MBRP, there is limited evidence from quasi-experimental studies in the literature that women may be more attracted to MBIs than men (Katz & Toner 2013). Future research may investigate gender differences in self-selection into MBI treatment as well as attrition rates.

Altogether, this study suggests that delivering MBRP in therapy groups of at least one-third women optimizes effects on long-term drug use outcomes. Accordingly, future research is warranted to understand group gender composition as a moderator of the effect of MBRP.