Group therapy for women with substance use disorders: In-session affiliation predicts women’s substance use treatment outcomes

Abstract

In-session affiliation among members is a hypothesized mechanism of action of group therapy for women with substance use disorders (SUDs). We evaluate group affiliation as an independent predictor of SUD treatment outcome in women (n=100), 18 years or older diagnosed with substance dependence, who were randomized to the single-gender Women’s Recovery Group (WRG) or mixed-gender group therapy (Group Drug Counseling; GDC). Affiliative statements made by members in both groups were measured for 39 women in each treatment arm. We studied the relationship between frequency of affiliative statements categorized in quintiles and the trajectory of days of any drug use during 3 months treatment and 6 months post-treatment using a Poisson regression model with estimation via generalized estimating equations. Furthermore, we investigated whether the effect of affiliation on substance use was moderated by group therapy type. The relationship between amount of affiliation and substance use reduction was non-linear. At the end of the treatment phase (3 months), women who experienced the highest level of affiliation (>65 affiliative statements on average) were found to reduce substance use by about 1.75 days more (p-value=0.02) than women who experienced the lowest level of affiliation (<26). The effects of affiliation persisted 6 months post-treatment and were moderated by therapy group, whereby women enrolled in the single-gender WRG appeared to benefit more from affiliation post-treatment. Training therapists to facilitate verbal affiliation may provide added therapeutic benefit to group therapy for women with SUDs.

1. Introduction

Group therapy continues to be the most widely used modality of substance use disorder (SUD) treatment (Center for Substance Abuse Treatment, 2005) with demonstrated effectiveness in reducing substance use (Sobell, Sobell, & Agrawal, 2009; Weiss, Jaffee, Menil de, & Cogley, 2004). Nevertheless, there is a paucity of research that elucidates the mechanisms that contribute to the effectiveness of SUD group therapy (Weiss et al., 2004). Effectiveness of group therapies has been ascribed to several potential processes (Sugarman et al., 2016) including in-session factors (Joyce, Piper, & Ogrodniczuk, 2007) such as group cohesion, an in-session mechanism that is defined as the feeling of belonging to the group (Yalom, 1995). Group cohesion is considered a key therapeutic factor (Restek-Petrović et al., 2014) that can increase the opportunity of connection and support among participants (Brook, Galanter, & Kleber, 2008; Joyce et al., 2007; Marmarosh, 2015; Yalom, 1995).

Group cohesion may be particularly important for women. Literature shows that when there are higher percentages of women in a group, there are higher ratings of group cohesion (Curşeu, Pluut, Boroş, & Meslec, 2015). In single-gender groups, group cohesion is rated higher when the composition is all-female compared to all-male (Martin & Good, 2015). Moreover, there is evidence that cohesion in all-female groups is associated with greater attendance (Crino & Djokvucic, 2010; Smith-Ray, Mama, Reese-Smith, Estabrooks, & Lee, 2012) and enhanced psychiatric outcomes (Gallagher, Tasca, Ritchie, Balfour, & Bissada, 2014).

Although group cohesion has been positively associated with therapeutic outcome in a variety of clinical populations (Burlingame, McClendon, & Alonso, 2011; Lecomte, Leclerc, Wykes, Nicole, & Abdel Baki, 2015; Taube-Schiff, Suvak, Antony, Bieling, & McCabe, 2007; Tschuschke & Dies, 1994) there is a lack of consensus in the field regarding how best to measure cohesion (Burlingame et al., 2011; Hornsey, Dwyer, & Oei, 2007). Group cohesion has typically been assessed using factors such as attendance (Piper, 1984), self-report questionnaires (Lecomte et al., 2015), and group size (Lo Coco, Gullo, Lo Verso, & Kivlighan, 2013). Although these measures may assess self-report of belonging to a group, and attendance could represent members’ sense of belonging to a group, these measures do not assess observable behavior or in-session processes such as verbal interactions among group members including statements of support, empathy, understanding, feeling supported, and belonging to the group. Accordingly, researchers have advocated for more specific, observable measures (Greenfield, Kuper, Cummings, Robbins, & Gallop, 2013; Hornsey et al., 2007).

Using data from the Women’s Recovery Group Study [NCT01318538], our group developed a procedure to quantify one observable characteristic of in-session group cohesion by measuring the frequency of in-session verbal affiliative statements (Greenfield, Cummings, Kuper, Wigderson, & Koro-Ljungberg, 2013; Sugarman et al., 2016). We defined affiliative statements as “supportive, positive, or empathic” comments among members of the group. Observational studies have found that women show higher rates of verbal affiliation in social interaction compared to men (Luxen, 2005); thus, for women, verbal affiliation may be of particular importance for connecting with others in group settings. We have previously demonstrated that verbal affiliative statements can be reliably coded from video and audiotapes of group therapy sessions (Greenfield, Cummings, et al., 2013; Sugarman et al., 2016).

The single-gender Women’s Recovery Group (WRG) is an evidence-based, manualized, relapse-prevention group therapy that uses a cognitive behavioral approach and includes gender-specific content (Greenfield, 2016; Greenfield et al., 2014; Greenfield, Trucco, McHugh, Lincoln, & Gallop, 2007). As previously described (Sugarman et al., 2016), the single-gender group composition of the WRG is hypothesized to enhance group cohesion, comfort, and support for participants, while the women-focused content provides women with education regarding the gender-specific antecedents and consequences of substance use. It is hypothesized that these two components (group composition and content) synergize to enhance outcomes compared with a standard, mixed-gender recovery group. The WRG was tested in two trials in which women with SUDs were randomized to 12 weekly sessions of the WRG or to the evidence-based active comparison condition, mixed-gender Group Drug Counseling (GDC) (Greenfield et al., 2014; Greenfield, Trucco, et al., 2007). The WRG was compared to mixed-gender GDC in a Stage I treatment development randomized controlled trial using a semi-open group format (Greenfield, Trucco, et al., 2007), and in a larger Stage II trial (Greenfield et al., 2014) using an open, rolling group format. In both trials, women in the WRG and GDC demonstrated clinically relevant reductions in substance use after 12 weeks of group treatment. In the Stage I trial, these reductions persisted 6 months post-treatment for women in WRG (Greenfield, Trucco, et al., 2007) but not GDC. Results from the Stage II trial showed that GDC is no more effective than WRG (Greenfield et al., 2014) and that the WRG can be feasibly implemented in community practice in a rolling group format.

Women with SUDs often report preferences for women-only treatment because they perceive it as more comfortable, open (Kauffman, Dore, & Nelson-Zlupko, 1995), honest, safe, and intimate (Greenfield, Brooks, et al., 2007; Greenfield, Cummings, et al., 2013). Recent studies have investigated the additional benefits of single-gender group therapies, which can provide a more effective treatment environment for women with SUDs (Bride, 2001; Claus et al., 2007; Greenfield, 2016; Greenfield et al., 2014; Greenfield, Trucco, et al., 2007; Grella, 2008; Orwin, Francisco, & Bernichon, 2001). We, therefore, hypothesized that enhanced affiliation may be a critical component of creating a cohesive and supportive group therapy environment for women with SUDs in general, and for the WRG specifically.

To investigate this hypothesis, we developed a coding manual in order to quantify in-session group affiliation by measuring the frequency of affiliative statements made by group members in each treatment condition (i.e., WRG and GDC) for the Stage I trial (Greenfield, Kuper, et al., 2013), and then refined this coding manual for the larger Stage II trial (Sugarman et al., 2016). Using Noldus XT software (Zimmerman, Bolhuis, Willemsen, Meyer, & Noldus, 2009), members of our group analyzed the videotaped sessions of the WRG trials, extracting and coding affiliative statements. We found that the single-gender group therapy for women with SUDs (i.e., WRG) was associated with enhanced group affiliation (measured as number of sentences of encouragement, support, and affiliation during therapy sessions) relative to mixed-gender group therapy (i.e., GDC), and that the number of affiliative statements was 66% higher in WRG than in the mixed-gender GDC (Sugarman et al., 2016). Based on these results, we hypothesized that greater frequency of affiliative statements among group members could be one mechanism of enhanced support and efficacy of the WRG compared with mixed-gender group therapy for SUDs. However, it is unknown whether affiliation, quantified by using these in-session observational data, is associated with substance use outcomes following group therapy.

We are not aware of any previous studies that have quantified the effect of affiliation on women’s substance use treatment outcomes in the context of either single- or mixed- gender group therapies for SUDs. The aims of the present study were to investigate the relationship of affiliation with women’s substance use treatment outcomes. In particular, we quantified the relationship of group affiliation measured via frequency of in-session affiliative statements to days of any substance use during the 12-week treatment phase and 6 months post-treatment. Furthermore, we investigated whether the effect of affiliation on women’s substance use treatment outcomes is moderated by single- versus mixed-gender group therapy.

2. Material and Methods

A detailed description of the Women’s Recovery Group (WRG) treatment and the Stage I and Stage II clinical trials can be found in previous publications (Greenfield, 2016; Greenfield, Kuper, et al., 2013; Greenfield et al., 2014; Greenfield, Trucco, et al., 2007). Details of methods pertinent to the present analyses of the Stage II trial are presented below.

2.1. Group Treatments

Groups for GDC and WRG comprise 12 weekly 90-minute sessions (Greenfield, 2016). GDC (Crits-Christoph et al., 1999; Daley, Mercer, & Carpenter, 2002) is an evidence-based, mixed-gender group therapy that promotes abstinence, provides education on addiction and recovery, encourages participation in mutual help groups, and teaches coping skills. The session topics and education provided in the GDC are not gender-specific. Eight female therapists were assigned to either WRG or GDC. Groups were audio- and videotaped to examine group dynamics and therapist adherence to the treatment manual. Therapists were supervised weekly to assure fidelity to each group treatment manual. Participants signed consent for audio- and videotaping of sessions. In order to ensure confidentiality of participants, only therapists were visible in videotaped recordings.

2.2. Participants

The protocol was listed in clinicaltrials.gov (Identifier: NCT01318538), approved by the McLean Hospital Institutional Review Board, and written informed consent was obtained from all participants. Participants were eligible if they were 18 years or older, met DSM-IV criteria for substance dependence for at least one substance (in addition to nicotine if present), and used substances in the 60 days prior to enrollment. One hundred and fifty-eight participants were enrolled (100 women, 58 men). The focus of this study is on women’s SUD treatment outcomes in single and mixed-gender groups, therefore, women were randomized to WRG (n=52) or GDC (n=48) and men were assigned to GDC. Women were randomized using simple randomization with stratification on the basis of whether or not the subject was recruited from either a hospital, residential, or partial hospital versus outpatient setting, as we recognized that recent engagement in these higher levels of care may affect outcome or interact with the study treatments. In order to minimize selection bias for women specifically interested in women’s treatment, recruitment materials indicated that the study was for investigational group therapy for substance use disorders, and did not mention the WRG. The study took place at two community outpatient treatment sites: McLean Hospital’s Alcohol and Drug Abuse Treatment Program in Belmont, MA and Stanley Street Treatment and Resources (SSTAR) in Fall River, MA. Site details and baseline characteristics of women participants were previously reported (Greenfield et al., 2014). The most common SUD diagnoses were alcohol (89%), cocaine (18%), and opioids (17%). Participants were predominately white (94%) and non-Hispanic (99%), with a mean age of 47 years (SD=12.1; range of 23–79 years).

2.3. Outcome

In the present study, substance use was assessed as days of any substance use (including alcohol and drugs) using the Timeline Follow-Back (TLFB) (Sobell & Sobell, 1992). The TLFB uses a calendar method to calculate days of any drug use including alcohol. To validate self-reported substance use, urine toxicology screens were obtained at group sessions and during monthly assessments. Participants completed assessments at baseline. Follow-up assessments were conducted monthly during treatment (months 1–3), monthly in the three months post-treatment (months 4–6), and again at six months post-treatment (month 9).

2.4. Affiliative statements

Affiliative statements were defined as supportive, positive, or empathic comments among members of the group and assigned to eight categories (Sugarman et al., 2016): (i) agreement statements (e.g. “That’s exactly what I feel”); (ii) supportive statements (e.g. “That’s true, it isn’t easy”); (iii) positive statements about the group (e.g. “I’ve found this group very helpful.”); (iv) therapeutic statements (e.g. “How do you plan to stay sober?”); (v) completing another member’s thought (e.g. Participant A: “I’m getting old, my body…” Participant B: “…doesn’t tolerate it anymore.”); (vi) strategy (e.g. “You can try taking your walks somewhere else”); (vii) engaging question (e.g. “Who is going to be at the dinner?”); and (viii) shared experience (e.g. “I relate to that”). Two independent raters, to ensure reliability, were trained to code the group therapy videos using Noldus XT (Zimmerman et al., 2009). There was a total of 446 group sessions throughout the study; the number of participants per group ranged from 1–7. Group videos were excluded from analyses if they only included one participant, and GDC groups where only men or only women were present were also excluded. This left 362 eligible videos, and one in five (20%; n=74) of these were randomly selected for coding. Figure 1 describes the process of the selection of group sessions for the analytic sample. In order to obtain a representative set of videos, videos were randomly ordered based on therapist, and approximately the same number of videos from each of the eight therapists was randomly selected and coded. Twenty-five percent of the 74 videos (n=19) were coded by a second rater. The average kappa coefficient across the 19 dually coded videos was .80, suggesting moderate to high inter-rater reliability. The detailed coding procedure is described in Sugarman et al. 2016 (Sugarman et al., 2016).

Figure 1.

Group sessions included in analytic sample

Information on group affiliation experienced by the participants, obtained as the sum of the affiliative statements falling into one of the eight categories (agreement statements, supportive statements, positive statements about the group, therapeutic statements, completing another member’s thought, strategy, engaging question, shared experience), was available for 120 participants (78 women and 42 men) out of the 158 enrolled in the study. For the purposes of these analyses, only women were included in the analytic sample. Figure 2 provides an illustration of how the analytic sample was constructed. For female participants who attended more than one group session with a coded tape, an overall measure of group affiliation experience was obtained averaging the number of affiliative statements across the eight categories over the number of group sessions that were analyzed, in which they were present. Out of the 78 women, 31 had affiliation data from a single session, 23 had data from two sessions, 16 had data from three sessions, 5 had data from four sessions, 2 had data from five sessions, and 1 had data from six sessions. In our analyses, the overall measure of affiliative statements was categorized in quintiles.

Figure 2.

Participant inclusion in analytic sample

2.5. Statistical Analysis

We evaluated differences in baseline characteristics between the analytic sample compared to the overall sample using chi-square tests for categorical variables and independent t-tests for continuous variables. We then conducted regression analyses of substance use accounting for the longitudinal and clustered nature of the data. We studied the relationship between affiliative statements categorized in quintiles and the trajectory of days of any substance use during treatment phase (months 1–3) and post-treatment phase (months 4–9) using a Poisson regression model with estimation via generalized estimating equations (GEE) (Fitzmaurice, Laird, & Ware, 2011). We fitted two separate models for the in-treatment and post-treatment phase. Each model included the effects of time in months (3 time points in the treatment phase and 6 time points in the post-treatment phase), affiliative statements (5 levels), and the affiliative statements by time interaction. Comparisons of levels of affiliation in terms of changes in outcomes were based on the test of interaction. We included all data available from randomized participants with information on experience of group affiliation under an intent-to-treat paradigm. All models adjusted for potential confounders of the relationship between affiliative statements and number of days of substance use based on prior studies such as group therapy status (WRG/GDC), baseline number of days of substance use (drug and alcohol), and self-efficacy. Self-efficacy at baseline is defined as one’s belief in one’s ability to succeed in remaining abstinent from substances during a high risk situation and was coded as a binary variable based on the clinical cutoff score of 80 on the Drug-Taking Confidence Questionnaire, which is considered high self-efficacy level (Cummings, Greenfield, & Gallop, 2010). In a sensitivity analysis we furthermore adjusted for an individual level measure of attendance, defined as the number of sessions attended.

To evaluate potential effect modification by treatment group we added a time by treatment by affiliative statement interaction, as well as treatment by affiliative statement and time by treatment interactions. A test for significance in the three-way interaction was used to assess whether the effect of affiliation on changes in outcomes is modified by treatment group status.

3. Results

Table 1 presents the baseline characteristics of this analytic sample (n=78) and demonstrates no significant differences in baseline characteristics between the analytic sample and the sample of all women (n=100) who enrolled in the Stage II WRG trial.

Table 1.

Descriptive statistics in the Stage II trial sample of women excluded and included in the analytic sample of the current study

Sample	Excluded from Analytic Sample(n = 22)		Included in Analytic Sample (n = 78)
Variable	n	%	n	%
Marital status
Married	3	13%	29	37%
Education *
More than high school	17	77%	62	79%
Graduated high school	1	4%	13	17%
Less than high school	4	18%	3	4%
Household income
$0 – $20,000	9	41%	21	27%
$20,001 – $50,000	5	23%	17	21%
$50,001 – $100,000	4	18%	18	23%
More than $100,000	4	18%	22	28%
Substance use diagnoses
Alcohol	16	72%	60	77%
Cocaine	2	9%	7	9%
Cannabis	0	0%	2	2%
Opioids	1	5%	4	5%
Other	3	13%	5	7%
Self efficacy
>= 80	7	32%	33	42%
	Mean	SD	Mean	SD
Days of any substance use (including alcohol) in the past 30 days	15.59	9.54	17.56	9.65

^*Significant difference between included and excluded women, p < 0.05.

3.1. Frequency of Statements

Confirming prior findings of the group (Sugarman et al., 2016), there was a greater number of total affiliative statements in WRG compared to GDC (66% higher, p=0.0038) (Figure 3). We also observed differences in frequency of affiliative statements within treatment group across therapists. The p-values of an ANOVA test for the association between therapist and affiliative statements were <0.001 for both treatment groups.

Figure 3.

Boxplot of frequency of affiliative statements by treatment group.

For regression analyses we categorized the overall measure of affiliative statements into 5 categories corresponding to the quintiles of the distribution with cut-offs at 26, 41, 50, and 65 for number of affiliative statements. The minimum number of affiliative statements recorded per participant experience was 5 and the maximum 122.

3.2. Affiliative Statements and Substance Use

We plotted trajectories of days of any substance use during treatment and post-treatment phase by affiliative statements quintile and by treatment group (Figure 4). On average, women exposed to a high level of affiliation (on average >65 statements) experienced greater substance use reduction, which persisted over time. In the WRG, women who experienced the lowest affiliation (on average <26 statements) were the ones who had lowest reduction in substance use.

Figure 4.

Trajectories plotting the mean of days of any substance use disorder use daysfor each time point during (months 0–3) and post treatment (months 4–9) phase by treatment (left panel GDC – mixed gender group and right panel WRG – single gender group) and quintiles of affiliative statements (in each panel different colors represent the 5 groups of affiliative statements experienced by the patient: <26, 27–41,42–50, 51–65, > 65 statements).

In the longitudinal regression analyses, enhanced affiliation was significantly associated with decrease in substance use during the 12-week treatment phase. The effect was non-linear, whereby a large frequency of affiliation is required to yield benefits. Participants who experienced the highest level of affiliation (>65) when compared to those who experienced the lowest level (<26) were found to reduce days of any substance use by approximately half a day more at each month (coefficient of time x affiliation interaction: beta=−0.58, p-value=0.02, Table 2, Model 1, treatment phase). At the end of the treatment phase (after 3 months) women who experienced the highest level of affiliation were found to reduce substance use by about 1.75 days more than women who experienced the lowest level of affiliation.

Table 2.

Poisson regression model with estimation via generalized estimating equations (GEE) to investigate the relationship between frequency of affiliative statements categorized in quintiles and the trajectory of days of any substance use during treatment (months 1–3) and post-treatment (months 4–9).

	Treatment Phase		Post-treatment Phase
	Model 1	Model 2	Model 1	Model 2
	Estimate (S.E.)	Estimate (S.E.)	Estimate (S.E.)	Estimate (S.E.)
Intercept	0.334 (0.637)	0.573 (0.605)	1.404** (0.431)	1.202* (0.497)
time	0.323* (0.150)	0.186 (0.137)	0.005 (0.024)	0.015 (0.037)
Affiliation (26,41]	0.526 (0.721)	0.934 (0.744)	−0.589 (0.596)	−0.481 (0.571)
Affiliation (41,50]	0.639 (0.738)	1.735 (1.074)	−0.049 (1.007)	0.787 (1.251)
Affiliation (50,65]	0.015 (0.786)	0.344 (1.042)	−0.502 (0.941)	0.301 (1.481)
Affiliation (65,122]	0.669 (0.667)	1.645* (0.755)	−1.524* (0.736)	−4.548** (0.959)
WRG	0.293 (0.398)	−0.456 (0.841)	0.328 (0.424)	1.333 (0.706)
High Self-Efficacy	0.021 (0.019)	−0.178 (0.383)	0.021 (0.022)	−0.768 (0.420)
Baseline days of substance use	−0.275 (0.382)	0.019 (0.019)	−0.719 (0.416)	0.013 (0.017)
time x Affiliation (26,41]	−0.226 (0.198)	−0.269 (0.181)	0.033 (0.085)	0.077 (0.104)
time x Affiliation (41,50]	−0.297 (0.310)	−1.168** (0.420)	−0.021 (0.109)	−0.010 (0.132)
time x Affiliation (50,65]	−0.101 (0.206)	−0.212 (0.143)	0.082 (0.103)	0.023 (0.163)
time x Affiliation (65,122]	−0.583* (0.246)	−1.023** (0.247)	0.069 (0.073)	0.429** (0.063)
time x WRG		0.468** (0.138)		−0.022 (0.046)
WRG x Affiliation (26,41]		−1.392 (1.173)		−0.511 (0.983)
WRG x Affiliation (41,50]		−1.046 (1.342)		−1.897 (1.917)
WRG x Affiliation (50,65]		−0.120 (1.442)		−1.677 (1.763)
WRG x Affiliation (65,122]		−1.029 (1.098)		2.853* (1.219)
time x WRG x Affiliation (26,41]		0.202 (0.345)		−0.077 (0.151)
time x WRG x Affiliation (41,50]		0.844 (0.466)		−0.038 (0.241)
time x WRG x Affiliation (50,65]		−0.087 (0.227)		0.105 (0.203)
time x WRG x Affiliation (65,122]		0.307 (0.309)		−0.398** (0.113)

^*p-value<0.05;

^**p-value<0.01.

Model 1 Main effect model; Model 2 interaction model between affiliative statements and WRG. Both models adjust for baseline days of any substance use and self efficacy.

During the 12-week treatment, both mixed-gender and single-gender groups significantly benefitted from enhanced affiliation and there were no significant differences in the effect of affiliation between the two groups, as we found no evidence of interaction between affiliative statements and specific group therapy modality (Table 2, Model 2, treatment phase). At the beginning of the post-treatment phase, women in the GDC who experienced the highest level of affiliation had a reduction of 4.5 days of use relative to women in the GDC who experienced low affiliation (coefficient of affiliation: beta=−4.5, p-value<0.01, Table 2, Model 2, post-treatment phase). However, this beneficial effect of group affiliation in GDC reduced over time. This can be seen from a positive time x affiliative statements interaction (beta=0.43, p-value<0.01, Table 2, Model 2, post-treatment phase). On the other hand, a significant negative 3-way interaction indicated that women enrolled in WRG had persistent reduction over time in days of any substance use (coefficient of time x WRG x affiliation interaction: beta=−0.40, p-value<0.001, Table 2, Model 2, post-treatment phase). The significant 3-way interaction indicates that for those in GDC the benefits of group affiliation decreased during the post-treatment phase, whereas for those in WRG the benefits persisted and remained stable over the same period of time.

4. Discussion

Group cohesion is considered an essential therapeutic factor in group therapy (Restek-Petrović et al., 2014) and can enhance the opportunity in the group setting for connection as well as support among participants (Brook et al., 2008). Moreover, there is some evidence that cohesion is associated with enhanced psychiatric treatment outcomes in all-female groups (Gallagher et al., 2014). However, the role of in-session processes that underlie cohesion among group participants in explaining group therapy’s short and long-term effects on substance use is limited. To our knowledge, our group is the first to examine group cohesion by measuring the observable phenomena of verbal affiliations within group sessions in either single-gender or mixed-gender SUD treatment groups. Furthermore, this is the first study to examine verbal affiliation as an independent predictor of women’s substance use disorder clinical treatment outcome.

In the present study, we found that women who experienced a high level of affiliation during group treatment sessions, measured as the 5^th quintile of number of affiliative statements (>65 statements on average), displayed a 90% reduction in days of substance use during the 12-week group treatment. By contrast, women who experienced a low level of affiliation, measured as the 1^st quintile of number of affiliative statements (<26 statements on average) displayed a 70% reduction in days of substance use during the 12-week group treatment. Importantly, these overall reductions in days of substance use were sustained during the 6-month post-treatment follow-up period in the single-gender group (i.e., WRG), while the beneficial effects of affiliation appear less persistent in the mixed gender group (i.e., GDC). This finding is consistent with the Stage I trial results that showed 6-month post-treatment persistence of positive treatment effects in women randomized to WRG but not to GDC (Greenfield, Trucco, et al., 2007). The findings also lend support to previous literature that has shown women’s preference for single-gender group therapy based on their perception of the enhanced safety, support, and comfort of these groups (Bride, 2001; Claus et al., 2007; Greenfield, 2016; Greenfield, Cummings, et al., 2013; Greenfield et al., 2014; Greenfield, Trucco, et al., 2007; Grella, 2008; Orwin et al., 2001). The effect of affiliation is non-linear, whereby the effects are found at high levels of experiencing affiliative statements. There is weaker evidence that experiencing moderate levels of group affiliation has beneficial effects on reductions in days of substance use that persist over time.

Although there were no statistically significant in-treatment differences in the effects of exposure to enhanced affiliative statements, it is clinically significant that the effects of group affiliation exposure degraded over the post-treatment period for women in the GDC, whereas it remained stable among women in the WRG. One possibility for this finding is that women in the WRG who experienced more affiliative statements may have formed stronger bonds or social ties with other women compared to those in the GDC, which may have translated into supportive networks following treatment completion.

In order to examine the effect of therapist on substance use, we conducted post hoc longitudinal analyses. Differences in reduction in substance use within treatment group across therapists were found. The p-values of an ANOVA test for the effect of therapist on reduction in days of substance use were <0.001 for both treatment groups during and post-treatment phases combined. Adjustment for affiliative statements in the models reduced the therapist effect on the outcome, indicating that ability to encourage group affiliation might partly explain the success of the therapists in clinical outcomes of group participants in reduction of days of substance use. This finding indicates that therapists vary with respect to their elicitation of group participants’ verbal statements of affiliation within sessions. Therapist promotion of affiliative statements may therefore be a modifiable factor in enhancing the effectiveness of SUD group therapies for women – in both single- and mixed-gender groups. We consider the analyses involving therapists as exploratory due to the low sample size within levels of therapist and affiliation categories. In the future it would be important to conduct a more thorough evaluation of therapists’ effect on affiliative statements and substance use reduction in the context of single and mixed gender therapy groups in an adequately powered study.

Measuring verbal affiliation is an observable and quantifiable method of examining in-session group cohesion. This study demonstrates that women’s experience of enhanced verbal affiliation within substance use disorder group sessions may be a mechanism of action for both single-gender and mixed-gender group therapies. Consistent with previous literature showing that group cohesion is an important therapeutic factor of group treatments (Joyce et al., 2007; Restek-Petrović et al., 2014; Yalom, 1995), higher levels of verbal affiliation may increase connection and support among participants, which is an important part of the recovery process (Bond, Kaskutas, & Weisner, 2003; Longabaugh, Wirtz, Zywiak, & O’Malley, 2010).

Study Limitations:

This study has a number of limitations. The sample size was insufficient for a more formal assessment of affiliative statements as mediators of the group therapy and therapist effect via mediation analyses (Valeri & VanderWeele, 2013). However, our quantitative analyses provide important preliminary findings to motivate larger studies to assess the mediating mechanisms of group therapy involving affiliative statements. In particular, our findings suggest that verbal affiliation as a mechanism of action might be modified by gender composition of the group. Another limitation concerns the measurement error in our measure of affiliation, which might lead to conservative estimates of the effect of affiliative statements on substance use. Unmeasured confounding of the affiliative statement-substance use relationship might bias our findings as well; therefore, affiliative statements, while predicting reduction of days of substance use might be a proxy of other factors. In sensitivity analyses, findings were robust to further adjustment for number of sessions attended. We did not consider number of sessions attended in our primary outcome models as this variable could act as either a confounder of the affiliation-substance use relationship or as an intermediate variable on the pathway between affiliation and substance use.

In addition, our study might not be generalizable to other treatment-seeking populations with SUDs that may be different with respect to the demographic and clinical characteristics of our sample (Substance Abuse and Mental Health Services Administration, 2012). In spite of these limitations, this study is significant in that it is the first empirical study of the experience of participant affiliation, as measured by the observable phenomena of in-session affiliative statements by group participants, and its effect on outcomes in women with SUDs during the 12-week treatment phase and 6-months post-treatment.

5. Conclusion

This study is the first to measure group chesion by the observable, quantifiable phenomenon of participant affiliative statements and to provide evidence that such verbal affiliation among participants may be one mechanism of action of group therapy for women with SUDs. Our work has important implications. This analysis of data from the Stage II two-site randomized controlled trial of single-gender WRG versus mixed-gender GDC demonstrates that group affiliation, measured as the number of in-session affiliative statements experienced by participants, predicts reduction of days of substance use during the 12-week treatment period, and that this positive effect persists 6 months post treatment. Moreover, there is additional substance use reduction over time for women in groups with high affiliation who were enrolled in the single-gender WRG, indicating that the effect of verbal affiliation on clinical outcome might be further modified by gender composition of the group. This study also finds that there is variation among therapists with regard to group affiliation. Regardless of group type (WRG or GDC), therapists whose groups were characterized by greater verbal affiliative statements by participants saw greater reduction in days of substance use among women participants. These findings suggest that promotion of group affiliation, and in particular participant statements of empathy, encouragement, and support, a key component of the WRG, can enhance SUD outcomes among women in SUD group therapy. Verbal affiliation among members in SUD group therapy may be especially significant for women’s substance use treatment outcomes. Training therapists to facilitate such verbal affiliation may provide added therapeutic benefit to group therapy for women with SUDs.