Sexual Orientation and Substance Use Treatment Outcomes Across Five Clinical Trials of Contingency Management

Abstract

Objective:

Lesbian, gay, and bisexual (LGB) individuals have elevated rates of substance use disorders and present to treatment with more severe substance use problems. Despite this health disparity, recent reviews highlight the paucity of studies reporting sexual orientation in substance use research (e.g., Flentje, Bacca, & Cochran, 2015). Using data from five clinical trials of contingency management (CM), the current study investigated the impact of sexual orientation on three substance use outcomes: treatment retention, longest duration of abstinence, and percent negative samples submitted.

Method:

Participants (N = 912; mean age = 36.6; 51.1% female; 45% African American, 42.2% Caucasian) were randomized to standard care in community-based intensive outpatient treatment (IOP) or the same plus CM.

Results:

Patients identifying as LGB made up 10.6% of the sample. A significant proportion identified as bisexual (8.2% of the total sample). Regardless of sexual orientation, participants receiving CM achieved better treatment outcomes than those receiving IOP alone. There were no statistically significant differences between LGB and heterosexual participants in their response to IOP in general, and CM specifically, across all three treatment outcomes (ps>.05). However, equivalence testing revealed that outcomes were not statistically equivalent for LGB and heterosexual participants, with the exception of percentage of negative samples, which was equivalent within the CM group only.

Conclusions:

Differences in treatment response to CM and standard community-based IOP do not reach the level of statistical significance; however, in most cases, we cannot conclude that treatment response is equivalent for LGB and heterosexual individuals.

A substantial body of research indicates that lesbian, gay, and bisexual (LGB) persons evidence elevated rates of substance use disorders compared to their heterosexual counterparts (Green & Feinstein, 2012; McCabe, Hughes, Bostwick, West, & Boyd, 2009; Mereish & Bradford, 2014). LGB individuals are more likely to seek treatment for substance use disorders (Cochran & Mays, 2000; McCabe, West, Hughes, & Boyd, 2013) and enter substance use treatment with more severe substance use problems than heterosexual patients (Cochran & Cauce, 2006). Despite this known health disparity, clinical trials of substance use disorders treatments rarely include demographic data on sexual orientation. A recent review of published studies on substance use problems found that sexual orientation was reported in only an estimated 4.9% of PsychINFO and 6.5% of PubMed articles in 2012, and these rates did not show a significant upward trend compared to published articles in 2006 (Flentje, Bacca, & Cochran, 2015). A second systematic review found that this problem is not specific to the substance use field, with sexual orientation demographics rarely reported among randomized controlled trials of psychosocial interventions for anxiety and depression (Heck, Mirabito, LeMaire, Livingston, & Flentje, 2017).

In 2016, the National Institute on Minority Health and Health Disparities (NIMHD) designated sexual and gender minorities as a health disparity population for research purposes (NIMHD, 2016). The lack of published data on representation of LGB persons in clinical trials research for substance use disorders limits our understanding of the effectiveness of psychosocial treatments in this important minority group. A handful of studies have examined the efficacy of evidence-based treatments for substance use disorders in samples that solely consist of LGB persons, finding some promising effects for approaches such as cognitive behavioral therapy, contingency management, and motivational interviewing (e.g., Morgenstern et al., 2007, 2009; Shoptaw et al., 2005). However, these studies have focused almost exclusively on gay/bisexual men. Further, such studies do not allow for analysis of the moderating effects of sexual orientation on treatment outcome, which is key to informing treatment approaches with LGB patients and, if needed, the development of culturally specific treatment adaptations. Only a few studies have been published on this topic. For example, one study compared LGB to non-LGB adolescents on their response to Community Reinforcement Approach, finding that both groups showed reductions in substance use and mental health problems but that LGB youth showed more drastic reductions (Grafsky, Letcher, Slesnick, & Serovich, 2011). In a sample of individuals receiving mental health treatment, a population that likely overlaps with individuals in substance use treatment, findings were comparable for LGB and heterosexual individuals overall but bisexual individuals had worse outcomes compared to other sexual orientations (Beard, Kirakosian, Silverman, Winer, & Wadsworth, 2017). To remedy the overall lack of data in this area and to clarify these inconsistencies, Heck and colleagues (2017) called for the collection and pooling of sexual orientation data across randomized controlled trials in order to investigate whether sexual orientation moderates treatment outcomes.

Using data from five randomized trials comparing standard intensive outpatient treatment (IOP) to IOP plus contingency management (CM; Petry et al., 2004, 2006; Petry, Alessi, Marx, Austin, & Tardif, 2005; Petry, Barry, Alessi, Rounsaville, & Carroll, 2012; Petry, Weinstock, & Alessi, 2011), we sought to determine whether sexual orientation is associated with key substance use treatment outcomes (i.e., treatment retention, longest duration of abstinence, percent of urine/breath samples negative for illicit drugs/alcohol). CM is one of the most efficacious approaches for substance use disorders (Dutra et al., 2008; see Prendergast, Podus, Finney, Greenwell, & Roll, 2006 for a review), and IOP is common in clinical settings. Thus, these trials allow for secondary analyses related to sexual orientation in two important treatment modalities.

Method

Study Design

These secondary analyses involve five randomized clinical trials comparing CM interventions to standard care in IOP settings (Petry et al., 2004, 2005, 2006, 2011, 2012). In all of these trials, participants randomized to CM in addition to IOP achieved better treatment outcomes relative to those randomized to IOP alone. The studies assessed the same primary outcomes using the same measures, were conducted in community-based clinics, and targeted similar patient populations (i.e., treatment-seeking individuals with substance use disorders). All five studies involved CM+IOP and IOP only arms. The studies also utilized similar inclusion and exclusion criteria, assessment and outcome measures, follow-up schedules, and procedures. These similarities in methods and design provide a rationale for combining samples, thus providing greater statistical power to test for differences between LGB and heterosexual participants.

Participants

The sample consisted of 920 individuals with substance use disorders who were initiating intensive outpatient treatment at one of seven community clinics. These clinics are all non-profit community-based clinics serving primarily indigent and uninsured patients in the New England area. All participants were 18 years of age or older, English speaking, and met diagnostic criteria for past year substance abuse or dependence using DSM-IV criteria (American Psychiatric Association, 1994). The vast majority of participants were cocaine dependent and most had polysubstance use problems. Exclusion criteria were uncontrolled severe psychiatric problems, inability to understand study procedures, and being in recovery from pathological gambling. Those with severe psychiatric illnesses who were reasonably stable (e.g., not actively suicidal or manic) were not excluded if they met other study criteria. Of the 920 participants, six were missing data on sexual orientation and two were missing data other key measures, leaving a final sample of 912 (see Table 1). Study procedures were approved by the university Institutional Review Board, and all participants provided written informed consent.

Table 1.

Demographic and Clinical Characteristics of LGB and Heterosexual Participants

	LGB (n = 96)	Heterosexual (n = 816)	p-value
Study, % (n)
Petry et al. (2004)	9.4 (9)	13.6 (111)	.157
Petry et al. (2005)	16.7 (16)	15.4 (126)
Petry et al. (2006)	5.2 (5)	9.7 (79)
Petry et al. (2011)	21.9(21)	25.7 (210)
Petry et al. (2012)	46.9 (45)	35.5 (290)
CM condition, % (n)	59.4 (57)	58.8 (480)	.917
Sex (% female)	80.2 (77)	47.8 (390)	<.001
Race, % (n)
African American	40.6 (39)	45.7 (373)	.601
Caucasian	44.8 (43)	42.0 (343)
Other	14.6 (14)	12.3 (100)
Marital status, % («)
Never married	69.8 (67)	54.4 (444)	.032
Married/cohabitating	9.4 (9)	17.3 (141)
Separated/divorced	18.8 (18)	26.3 (215)
Widowed	2.1 (2)	2.0 (16)
Employment status, % (n)
Full time	22.9 (22)	43.4 (354)	.001
Part-time	28.1 (27)	23.0 (187)
Unemployed	40.6 (39)	23.4 (191)
Retired/disability	6.3 (6)	6.6 (54)
Student/controlled environment	2.0 (2)	3.7 (30)
Intake positive sample, % (n)	27.1 (26)	24.4 (199)	.567
Cocaine dependent, % (n)	91.7 (88)	84.4 (689)	.059
Alcohol dependent, % (n)	59.4 (57)	52.8 (431)	.223
Opiate dependent, % (n)	29.1 (25)	31.8 (219)	.609
Age, M(SD)	33.67 (8.15)	37.09 (9.19)	<.001
Years of education, M (SD)	11.42 (2.08)	11.99 (1.92)	.006
Years of cocaine use, M (SD)	9.42 (6.47)	9.87 (7.98)	.529
Years of alcohol use, M (SD)	8.67 (8.86)	10.12 (10.36)	.139
Years of heroin use (among users), M(SD)	7.43 (7.04)	7.46 (7.58)	.983
ASI Scores, M(SD)
Medical	.25 (.35)	.25 (.35)	.965
Employment	.81 (.24)	.72 (.28)	.001
Alcohol	.18 (.21)	.22 (.24)	.164
Drug	.16 (.10)	.16 (10)	.577
Legal	.15 (.21)	.13 (.21)	.290
Family/social	.20 (.23)	.18 (.21)	.512
Psychiatric	.33 (.24)	.27 (.24)	.021

Note. ASI=Addiction Severity Index. CM=contingency management. LGB=lesbian/gay/bisexual.

Procedures

At a baseline assessment, participants completed a battery of measures, which included a demographic survey, a diagnostic criteria checklist for substance use disorders based on the Structured Clinical Interview for DSM-IV (First, Spitzer, Gibbon, & Williams, 1996), and the Addiction Severity Index (ASI; McLellan et al., 1985). The ASI produces composite scores ranging from 0.00 to 1.00 on seven domains related to consequences of substance use: medical, employment, alcohol, drug, legal, family/social, and psychiatric. Higher scores indicate greater problem severity in each domain. A single item assessed sexual orientation (“do you consider yourself to be…”), offering the following choices: heterosexual (straight), gay, lesbian, or bisexual. Following the baseline assessment, eligible participants were randomized to treatment condition, as described below.

Treatments

All participants received either standard intensive outpatient (IOP) substance use treatment delivered by the community clinics or IOP plus CM.

Standard care.

Standard care was similar across the five studies and seven clinics, consisting of IOP therapy sessions, up to 5 hours/day, 3–5 times/week for up to 6 weeks. The frequency of therapy sessions then gradually tapered for up to 12 additional weeks. Aftercare involved one group session per week for up to 12 months. Therapy included 12-step treatment, HIV/AIDS education, relapse prevention, and life skills training. In addition to the treatment received at the clinics, participants submitted up to 24 breath and urine samples during the 12-week study intervention period. Research staff collected and screened samples and encouraged participants to discuss results with clinic staff if they tested positive. However, all data, including breath and urine sample results were confidential and not shared with non-study staff, as these clinics did not regularly screen for substance use. Thus, standard care for study participants was as similar as possible to standard care for non-study participants, with the exception of additional testing.

Contingency management (CM) + Intensive Outpatient.

Participants in the CM +IOP conditions received the same standard care described above, including collection of breath and urine samples. In addition, they earned reinforcement for verified target behaviors. The target behaviors varied across studies and included abstinence from certain illicit drugs and, in some studies, alcohol; attendance at IOP treatment sessions; and/or completion of treatment goal-related activities. Magnitude of available reinforcement (range = $80–$874), reinforcement delivery system (prizes versus vouchers), and format (group versus individual CM) also varied across studies and CM+IOP conditions. The primary aims of each study are described briefly below, and the original reports contain detailed descriptions of study design and treatment conditions. Petry et al. (2004) examined the efficacy of low magnitude CM reinforcement ($80) or a higher magnitude reinforcement ($240) plus standard IOP relative to IOP alone. Petry et al. (2005) varied the delivery system, comparing prize CM+IOP versus voucher CM+IOP versus IOP alone. Petry et al. (2006) compared three conditions: CM+IOP for abstinence, CM+IOP for goal-related activities, and IOP alone. Petry et al. (2011) compared CM+IOP delivered in a group session to IOP alone. The Petry et al. (2012) study had two subcomponents based on the participants’ initial urine toxicology results. Participants testing negative for cocaine at intake were randomized to CM+IOP for abstinence, CM+IOP for attendance, or IOP alone. Participants who were cocaine positive at intake were randomized to different magnitudes of CM reinforcement for abstinence or the standard care IOP condition.

Data Analysis

For the purposes of these analyses and to enhance uniformity among CM+IOP conditions, we combined all CM conditions targeting abstinence because these conditions improved outcomes relative to standard care (IOP alone) in all studies and did not differ from one another in terms of the target behavior. Participants in CM conditions not targeting abstinence (47 randomized to CM for goal-related activities condition in Petry et al. [2006] and 107 participants randomized to CM for attendance in Petry et al. [2012]) were excluded from all analyses because the activities condition did not significantly improve outcomes, and the attendance condition yielded mixed effects, inconsistent with the abstinence CM condition in that study. As noted above, 8 participants were missing data on key variables and not included in analyses. The remaining sample consisted of 912 participants randomized to either CM for abstinence plus IOP or IOP alone.

We compared participants who identified as LGB to those identifying as heterosexual on intake and demographic characteristics using chi-square tests of independence for nominal variables and independent t-tests for continuous variables. We present corrected t values and degrees of freedom when heterogeneous variances were present.

Primary outcomes were treatment retention (range = 0–12 weeks), longest duration of abstinence (LDA), and percent negative samples submitted. LDA represents the longest period of abstinence uninterrupted by positive samples or unexcused missed urine drug screens/alcohol breath screens (range = 0–12 weeks). The percent of negative samples submitted was calculated with the total number of samples submitted as the denominator and the number of negative samples submitted as the numerator (range = 0–100%). In this manner, the two drug use outcomes present the full range of possibilities regarding missing samples, i.e., considering missing samples as positive (for LDA) and as missing (in calculating % negative). In order to receive reinforcement in all of the CM conditions, participants’ samples were required to test negative for cocaine, alcohol, and opioids for the Petry (2004, 2005, 2006, 2012) studies. Samples in the Petry et al. (2011) study also had to be negative for methamphetamine (but these comprised <1% of samples). Samples that tested positive for one or more of the targeted substances interrupted the string of abstinence for LDA and were considered positive for the calculation of negative samples submitted.

Analyses examined the main effects of sexual orientation (LGB versus heterosexual) and treatment condition and their interaction. CM conditions were collapsed and represented by a dichotomous variable indicating randomization to IOP alone or any abstinence CM+IOP condition. We examined these effects in three separate models using analyses of covariance (ANCOVA), with the three primary outcomes (treatment retention, LDA, percent negative samples) entered as dependent variables and relevant variables included as covariates (see below). Data analyses were conducted using SPSS (version 24).

Finally, equivalence testing using the confidence interval approach (Rogers et al., 1993; Westlake, 1981) was conducted for all models that did not indicate statistically significant differences between the heterosexual and LGB participants. As indicated by Rogers et al. (1993), significance testing and equivalence testing are not mutually exclusive. The latter allows us to conclude whether heterosexual and LGB participants’ mean responses to treatment are close enough together to be considered equivalent to one another. These analyses were conducted separately for CM+IOP and IOP alone, retaining the same covariates as in the ANCOVA models. First, an equivalence interval was calculated for each outcome. Because there are no agreed upon standards for how close values must be in these study outcomes (treatment retention, LDA, and percentage of negative screens) to be considered equivalent, we chose ±20%, as this is the most commonly used equivalence interval in social sciences. Then, a 90% confidence interval was calculated for the mean difference between the heterosexual and LGB groups on each outcome for CM+IOP and IOP alone. If this 90% confidence interval falls within the equivalence interval, the two groups are considered to be equivalent on that outcome (Rogers et al., 1993).

Results

Of the 912 participants with complete data, 816 (89.5%) identified as heterosexual, 8 as gay men (0.9%), 13 (1.4%) as lesbian, and 75 (8.2%) as bisexual. Of the 75 participants who identified as bisexual, 11 (14.7%) were male and 64 (85.3%) were female. Given that individuals identifying as bisexual comprised a majority of the LGB group, additional post-hoc analyses were conducted specifically for bisexual participants.

Table 1 presents demographic and intake characteristics by sexual orientation. The sexual orientation distribution differed across sex, with significantly more female participants identifying as LGB. Compared to heterosexual participants, LGB participants were less likely to be currently or ever married, reported less full-time employment/more unemployment, were younger, had fewer years of education, and reported greater employment and psychiatric problems on the ASI. There were no significant differences in sexual orientation across studies or treatment conditions, in substance use variables, or on other subscales of the ASI. Gender, age, and scores on the ASI employment and psychiatric subscales were included as covariates in all subsequent models. Education was not included as a covariate due to its high correlation with employment problems, and the marriage variable was not included, as several of the studies were conducted prior to the legalization of same-sex marriage (though the marriage variable included both same-sex and heterosexual marriages).

Figure 1 presents ANCOVA results for all three substance use treatment outcomes. For treatment retention, after controlling for covariates and sexual orientation, the main effect of treatment condition was significant, F(1, 904) = 20.40, p < .001, partial ƞ² = .022. Participants assigned to CM+IOP stayed in treatment almost 2 weeks longer (M = 6.70 weeks, SE = 0.28) than those randomized to IOP alone (M = 4.75 weeks, SE = 0.34). Sexual orientation was not significantly associated with treatment retention, F(1, 904) = .34, p = .56, partial ƞ² < .001, nor was the interaction between sexual orientation and treatment condition, F(1, 904) = 2.97, p = .09, partial ƞ² = .003. Identification as bisexual was similarly not significantly related to treatment retention, F(1, 904) = .01, p = .94, partial ƞ² < .001, nor was the interaction between bisexual orientation and treatment condition, F(1, 904) = 1.45, p = .23, partial ƞ² = .002.

Figure 1.

Mean treatment retention (0–12 weeks), longest duration of abstinence (0–12 weeks), and percent of negative samples (0–100%) by treatment condition and sexual orientation. Values are adjusted means.

For LDA, a similar pattern of results emerged for the main and interaction effects. There was a significant main effect of treatment condition, F(1, 904) = 27.73, p < .001, partial ƞ² = .030. Participants randomized to CM+IOP achieved over 2 additional weeks of continuous abstinence (M = 5.55, SE = .29) compared to those in IOP alone (M = 3.22, SE = .34). Again, the main effect for sexual orientation, F(1, 904) = .38, p = .54, partial ƞ² < .001 and the sexual orientation by treatment condition interaction, F(1, 904) = 1.81, p = .18, partial ƞ² = .002, were not significant. Identification as bisexual was similarly not significantly related to LDA, F(1, 904) = .06, p = .81, partial ƞ² < .001, nor was the interaction between bisexual orientation and treatment condition, F(1, 904) = 2.56, p = .11, partial ƞ² = .003.

The main effect of treatment condition on percent negative samples was not significant, F(1, 904) = 3.01, p = .08, partial ƞ² = .003. LGB participants had a lower percentage of negative screens (M = 74.96%, SE = 3.26) than heterosexual participants (M = 81.00%, SE = 3.26) but this difference was not statistically significant, F(1, 904) = 3.07, p = .080, partial ƞ² = .003. Likewise, the sexual orientation by condition interaction effect was not significant, F(1, 907) = 1.47, p = .23, partial ƞ² = .002. Identification as bisexual was similarly not significantly related to percent negative samples, F(1, 904) = 1.96, p = .16, partial ƞ² = .002, nor was the interaction between bisexual orientation and treatment condition, F(1, 904) = .99, p = .32, partial ƞ² = .001.

Equivalence testing was conducted for all three outcomes, separately for the CM+IOP and IOP alone groups. See Table 2. For CM+IOP, heterosexual and LGB participants were found to be equivalent on percent negative screens but not on treatment retention or LDA. In both cases, the LGB means were higher than the heterosexual means (i.e., longer treatment retention, longer durations of abstinence). Heterosexual and LGB participants in IOP alone were not found to be equivalent on any of the three outcomes, with LGB participants having lower treatment retention, LDA, and % negative screens than heterosexual participants. Thus, although LGB and heterosexual participants did not have significantly different outcomes in response to CM+IOP or IOP alone, we cannot conclude that they are equivalent, except in the case of percent negative screens in response to CM+IOP.

Table 2.

Equivalence Tests

	Equivalence Interval	Mean Group Difference (Hetero minus LGB)	90% CI for Mean Group Difference	Equivalence
	CM Group
Treatment Retention	±1.29	−0.41	−1.36 – 0.53	No
LDA	±1.08	−0.22	−1.26 – 0.83	No
% Negative	±16.39	2.36	−4.59 – 9.32	Yes
	Standard Care Group
Treatment Retention	±1.05	0.94	−0.17 – 2.04	No
LDA	±0.73	0.68	−0.30 – 1.66	No
% Negative	±15.99	9.38	0.04 – 18.73	No

Note. CM = Contingency Management. CI = Confidence Interval. LDA = Longest Duration of Abstinence.

Discussion

This study is one of the first to examine whether substance use treatment outcomes differ based on sexual orientation in adult populations. LGB and heterosexual participants receiving CM+IOP did not demonstrate statistically significant differences on key substance use treatment outcomes (treatment retention, LDA, and percent negative screens). Similarly, no significant differences were found specifically for bisexual participants, who made up the majority of the LGB group. These findings are consistent with past studies of CM that failed to find statistically significant differences in treatment response for racial/ethnic minorities (Barry, Sullivan, & Petry, 2009; Montgomery & Carroll, 2017) and across sexes (Garcia-Fernandez et al., 2011; Rash & Petry, 2015; Wong, Badger, Sigmon, & Higgins, 2002) and income categories (Kinnaman, Slade, Bennett, & Bellack, 2007; Rash, Andrade, & Petry, 2013; Rash, Olmstead, & Petry, 2009; Secades-Villa et al., 2013). However, this study went a step further and examined whether LGB and heterosexual participants could be deemed equivalent on these outcomes. We found that they were indeed equivalent in their response to CM+IOP on one key outcome, percentage of negative urine drug screens submitted, but not on treatment retention or LDA. For these latter two outcomes, the LGB group means indicated a superior response to CM compared to the heterosexual group that was large enough to indicate a potentially clinically significant difference, though it did not reach the level of statistical significance. Thus, we cannot conclude that sexual orientation is not related to CM’s impact on these outcomes.

Previous studies have found that LGB patients have elevated rates of substance use disorders (McCabe et al., 2009; Mereish & Bradford 2014) and enter treatment with more severe substance use problems (Cochran & Cauce, 2006). In the current study, LGB individuals were overrepresented (10.6% of the sample) compared to the national prevalence rate of 3.5% identifying as LGB (Gates, 2011). Bisexual individuals were particularly overrepresented (8.2% of the current sample versus 1.8% of the general population; Gates, 2011). However, the percentage of bisexual participants was similar in studies of individuals seeking mental health treatment (8%; Beard et al., 2017) and substance abuse treatment (5.6%; Senreich, 2012). In our study, LGB patients did not differ significantly from heterosexual patients on many indices of substance use severity, including positive urine drug screen at intake, number of years of substance use, or endorsing DSM-IV dependence criteria, but were significantly elevated on other problems related to substance use, including employment and psychiatric difficulties. Standard IOP did not lead to statistically significant outcomes for LGB and heterosexual participants. However, equivalence testing showed that treatment outcomes for the IOP alone group could not be deemed to be equivalent between the LGB and heterosexual participants, with means indicating worse outcomes for LGB participants (i.e., opposite of the finding in the CM condition). The result of these equivalence tests suggests that, despite a lack of statistical significance, there is a meaningful difference in treatment outcomes between LGB and heterosexual clients that warrants further attention.

Some efforts have been made to develop culturally sensitive substance use disorders treatments for LGB populations, but evaluation of such adaptations have produced limited positive results (see Green & Feinstein, 2012 for a review). One such study (Shoptaw et al., 2005) randomized gay and bisexual men who were methamphetamine dependent to one of four treatment conditions: 1) cognitive behavioral therapy (CBT); 2) CM; 3) CBT and CM; or 4) culturally tailored gay-specific CBT. Individuals assigned to either of the two CM groups had better treatment outcomes (treatment duration, LDA) than those in a non-CM condition. The study did not find a benefit of culturally tailored compared to standard CBT on drug use outcomes, particularly at follow-up. The results of the current study also found that CM works well for LGB patients. In addition, building on Shoptaw et al.’s (2005) finding that a culturally tailored version of CBT was not superior to standard CBT, we found that treatment response to standard IOP services did not differ significantly by sexual orientation, even in the absence of LGB-specific services. Of note, such findings do not necessarily indicate that culturally adapted interventions should not be developed and tested to address the unique needs of LGB patients. As in this study, it is likely important to analyze not only statistically significant outcomes but also whether treatment outcomes are equivalent between LGB and heterosexual patients. In addition, treatment studies do not, by their very nature, include individuals who do not feel comfortable even seeking treatment in the first place due to their sexual orientation and their perceptions about the relevance of traditional substance use treatment services in meeting their needs. Additional research is needed on LGB individuals with substance use disorders who never seek treatment in the first place.

This study has several strengths, including the use of a large heterogeneous sample, broad inclusion criteria and limited exclusion criteria, and recruitment from community-based clinics, all of which support the generalizability of the findings. In addition, the use of objective indices of substance use as the primary treatment outcome limits the potential for biased substance use results. Finally, analysis of data across five clinical trials allowed for a substantial sample size, without which examination of sexual orientation as a predictor of outcomes in clinical trials would not be feasible due to the low baseline rates.

The findings should be considered in light of certain limitations. Although the use of multiple clinical trials allowed for this preliminary analysis of the effects of sexual orientation on treatment outcomes, a critical limitation involves the size of our LGB sample. The small size necessitated that we pool lesbian, gay, and bisexual participants together for the primary analyses, which is less than ideal given the variability in patterns of use and associated psychosocial stressors among LGB persons (see Green & Feinstein, 2012 for a review). In addition, a majority of the LGB participants in this sample were female (80.2%). Although gender was included as a covariate in the analysis, we were unable to explore gender differences within the LGB subgroup. We strongly encourage other researchers to examine their data to determine whether larger and more nuanced analyses can be conducted to further investigate if and how sexual orientation, including subgroups of LGB individuals, may impact substance use disorder treatment outcomes. Second, the measurement of sexual orientation was limited to gay, lesbian, bisexual, or heterosexual and did not capture alternative orientations (e.g., asexual). We also did not assess gender identity as part of these studies, nor did we have assessments of sexual behavior or attraction, which may not be fully consistent with identification of sexual orientation (e.g., men who have sex with men but identify as heterosexual). Future studies should use more detailed assessments of sexual orientation and also assess gender identity. In addition, it is possible that some LGB individuals did not identify themselves as such due to stigma and apprehension about reporting. The equivalence intervals used in this study were based on a rule of thumb, as standards for clinical significance for the outcomes examined in this study have not been firmly established. It is possible that we would arrive at different conclusions regarding equivalence if we used more or less conservative equivalence intervals. Finally, all studies were conducted in the northeast US and may not generalize to populations of LGB patients living in other areas of the US or in other countries.

Findings from this study provide initial evidence that sexual orientation does not lead to statistically significant differences in treatment outcomes for CM or standard IOP treatment. However, only one of these outcomes was found to be statistically equivalent for CM+IOP and none were equivalent for IOP alone. This is an intriguing finding and will hopefully compel other researchers to conduct follow-up equivalence analyses within their clinical trials. An important caveat is that our LGB sample contained a large proportion of bisexual women and, thus, results may be most relevant for this specific group. Follow-up analyses confirmed similar results for the bisexual group as the larger LGB group. Despite these findings, many questions remain unanswered. Compared to other evidence-based treatments for substance use disorders, individual patient characteristics may be less likely to influence treatment outcomes for CM, given its primarily behavioral components. Future studies of other psychosocial treatments (e.g., CBT, motivational interviewing) should likewise examine the impact of sexual orientation and gender identity on treatment efficacy, using combined datasets across multiple clinical trials when available. These studies can inform decision making about the necessity and focus of culturally tailored substance use treatments for sexual and gender minority groups. Further, as recommended by recent reviews (Flentje et al., 2015; Heck et al., 2017), all clinical trials for substance use disorders treatments should measure and report sexual orientation and gender identity and, when there is adequate power, examine these important demographic values as predictors of treatment outcome using tests of both statistical differences and equivalence.