Depression history as a predictor of outcomes during buprenorphine-naloxone treatment of prescription opioid use disorder

Abstract

Background:

In the multi-site Prescription Opioid Addiction Treatment Study (POATS), the best predictor of successful opioid use outcome was lifetime diagnosis of major depressive disorder. The primary aim of this secondary analysis of data from POATS was to empirically assess two explanations for this counterintuitive finding.

Methods:

The POATS study was a national, 10-site randomized controlled trial (N=360 enrolled in the 12-week buprenorphine-naloxone maintenance treatment phase) sponsored by the NIDA Clinical Trials Network. We evaluated how the presence of a history of depression influences opioid use outcome (negative urine drug assays). Using adjusted logistic regression models, we tested the hypotheses that 1) a reduction in depressive symptoms and 2) greater motivation and engagement in treatment account for the association between depression history and good treatment outcome.

Results:

Although depressive symptoms decreased significantly throughout treatment (p <.001), this improvement was not associated with opioid outcomes (aOR=0.98, ns). Reporting a goal of opioid abstinence at treatment entry was also not associated with outcomes (aOR=1.39, ns); however, mutual-help group participation was associated with good treatment outcomes (aOR=1.67, p <.05). In each of these models, lifetime major depressive disorder remained associated with good outcomes (aORs=1.63–1.82, ps =.01–.055).

Conclusions:

Findings are consistent with the premise that greater engagement in treatment is associated with good opioid outcomes. Nevertheless, depression history continues to be associated with good opioid outcomes in adjusted models. More research is needed to understand how these factors could improve treatment outcomes for those with opioid use disorder.

1. Introduction

In response to the high prevalence of prescription opioid use disorder (Han et al., 2015), the past decade has seen an increase in research on the efficacy of treatments for this disorder. The National Drug Abuse Treatment Clinical Trials Network Prescription Opioid Addiction Treatment Study (POATS) was the first large, multi-site randomized trial of buprenorphine-naloxone for prescription opioid dependence (Weiss et al., 2010a; Weiss et al., 2011). In this trial, different durations of buprenorphine-naloxone treatment (a 4-week brief treatment followed for some participants with an extended, 12-week treatment) and different intensities of drug counseling were evaluated as treatments for patients with prescription opioid dependence. Results indicated that nearly half of participants met the threshold for successful opioid outcome at the end of the treatment in the trial (Weiss et al., 2011). This trial, along with others (e.g. Fiellin et al., 2014; Sigmon et al., 2013), support the efficacy of pharmacotherapy for prescription opioid use disorders. However, significantly less is known about predictors of successful treatment outcome in this population. Identifying predictors of outcome is particularly important to improving treatment for the 50% of patients who do not adequately respond to this intervention.

In a previous study, Dreifuss and colleagues (2013) analyzed baseline data from the POATS trial to determine which factors best predicted successful opioid use treatment outcome. Somewhat surprisingly, a diagnosis of lifetime major depressive disorder (MDD) was the best predictor of good outcome in this trial, above and beyond other factors such as age, route of administration, and prior opioid use disorder treatment (Dreifuss et al., 2013). This finding seems counterintuitive, as depression is generally linked to poor outcomes in substance use disorders (Greenfield et al., 1998; Najt et al., 2011), including opioids (Hasin et al., 2002; Nunes, et al., 2004; Rounsaville et al., 1986); moreover, depression predicts opioid use over time (Sullivan et al., 2006). However, this puzzling result is consistent with results of several other studies of opioid use disorder outcomes in which better outcomes for substance use are associated with depression symptoms, history of suicide attempts, and depression diagnoses (Gerra et al., 2004; Joe et al., 1994; Marcovitz et al., 2016; Marsch et al., 2005; Rao et al., 2004).

Two possible explanations for this effect were raised by Dreifuss and colleagues (2013). First, given the potential antidepressant properties of buprenorphine (cf. Peciña et al., 2019; Saxena and Bodkin, 2019), treatment with this medication during the study may have helped to treat symptoms of depression; and second, major depression may confer greater motivation to engage in treatment (e.g., Rounsaville and Kleber, 1985). The first of these potential explanations is based on the premise that a reduction in depressive symptoms during opioid treatment is a possible mechanism driving successful outcomes among those with a diagnosis of depression, whereas the second is based on evidence that a diagnosis of depression may lead to more engagement in substance use treatment, thus leading to better outcomes. In the present study, we completed a secondary analysis of additional data from the POATS trial to test if either of these two explanations could help to explain the good prognostic effects of depression history on opioid treatment outcome in patients receiving buprenorphine.

Next, we briefly review two potential mechanisms for beneficial effects of depression history on opioid treatment outcome. One explanation involves the possible association between MDD diagnosis and depressive symptom severity: if participants with a history of MDD entered the POATS trial with a higher level of current depression symptoms compared to those without, they would likely experience a greater reduction in depressive symptoms that could lead to better overall outcomes in treatment for opioids. Many studies show that elevations in current depression symptoms are common in individuals presenting for treatment of opioid use disorder, and that depressive symptom severity often declines during the course of medication-based treatment (Dean et al., 2004; Kosten et al., 1990; Krupitsky et al., 2016; Latif et al., 2019; Mysels et al., 2011; Romero-Gonzalez et al., 2017; Zaaijer et al., 2015). For many individuals, this reduction in depressive symptoms may be due to remission of substance-induced depression (that is, depressive symptoms resulting from opioid use that are not present during abstinence from opioids; see Nunes et al., 2015, for review). Research on alcohol use disorders and depression shows that depressive symptoms typically resolve quickly for those with substance-induced depression (Brown et al., 1995). However, a diagnosis of MDD in the POATS trial required that depressive symptoms were independent from opioid dependence. Thus, any explanations for why MDD diagnosis predicts good outcome in the POATS trial cannot be solely explained by remission of substance-induced depression.

In the POATS trial, a reduction in depressive symptoms among all participants is plausible given the studies described above, as well as the potential antidepressant effects of buprenorphine itself (Ragguett et al., 2018; Stanciu, et al., 2017). Several studies suggest that buprenorphine may be efficacious in reducing symptoms of depression (Bodkin et al.,1995; Emrich et al.,1982; Karp et al., 2014; Nyhuis et al., 2008) and suicidal ideation (Yovell et al., 2016); some recent clinical trials also suggest that buprenorphine paired with samidorphan may be efficacious for treatment-resistant depression (Ehrich et al., 2015; Fava et al., 2016). In contrast, the question of whether buprenorphine confers an antidepressant effect in opioid use disorder is far less studied, and to date results of these studies are mixed (Ahmadi and Jahromi, 2018; Dean et al., 2004; Kosten et al. 1990; Romero-Gonzalez et al., 2017). In summary, although few studies have tested antidepressant effects of buprenorphine in prescription opioid use disorder, individuals entering treatment for opioid use disorder often show a reduction in depressive symptoms during medication treatment, and buprenorphine may have antidepressant properties that could support such a reduction in the POATS study.

Another potential explanation for the good prognostic effect of lifetime MDD diagnosis is motivation to engage in treatment. Although deficits in motivation are a central component of MDD (Barch et al., 2016), findings suggest that depression diagnoses or symptoms are often related to higher levels of motivation to change substance use (Blume et al., 2001; Haukkala et al., 2000; Holt et al., 2009; Morris et al., 2018; Smith and Tran, 2007), including motivation to change opioid use (Morris et al., 2018). Given that motivation is related to treatment engagement among individuals with opioid use disorder (Simpson et al., 1995; Simpson et al., 1997), one goal of the present study was to test if motivation and treatment engagement could explain the effect of depression history on substance use outcome in the POATS trial. Although a motivation for change measure was not administered in this study, we took advantage of several aspect of the POATS study design that allowed us to test proxies for motivation and engagement in treatment. First, all participants completed a measure of abstinence goals. Endorsement of a goal of abstinence is considered one aspect of motivation and is robustly correlated with related constructs such as self-efficacy for sobriety, as well as engagement in treatment (Laudet and Stanick, 2010). Second, we evaluated engagement in treatment by examining mutual-help group attendance among POATS participants. Mutual-help group participation was encouraged at each study medical management session, but it was not required of study participants. Thus, given well-established links between motivation and treatment engagement (e.g., Simpson et al., 1995), frequency of mutual-help attendance may reflect a direct behavioral correlate of motivation.

The primary aim of this secondary analysis of data from the POATS trial was therefore to empirically assess two explanations for why a history of MDD diagnosis predicted good opioid use outcomes for prescription opioid-dependent patients treated with buprenorphine-naloxone in the POATS study. First, we tested the hypothesis that a reduction in depression symptoms would account for the association between a lifetime diagnosis of MDD and good treatment outcome. Based on evidence suggesting that depressive symptoms typically improve during medication treatment for opioid dependence, we evaluated whether change in depressive symptoms accounted for the effect of depression history on substance use outcomes. Second, we tested the hypothesis that greater motivation and engagement in treatment accounted for the association between MDD diagnosis and good treatment outcome, and that treatment motivation and engagement accounted for the association between MDD diagnosis and opioid use outcomes. Motivation to change was assessed via a question assessing abstinence goals, and engagement in treatment was assessed by participation in mutual-help groups.

In testing hypotheses about effects of depression diagnosis on opioid use disorder treatment, we considered the role of gender. Women are much more likely than men to be diagnosed with MDD (e.g., Hankin and Abramson, 2001), and according to epidemiological data, women are more likely than men to have MDD and simultaneously to use non-medical prescription opioids (Fink et al., 2015). Reports of attendance at mutual-help groups show mixed results, with some studies finding gender differences and others showing no difference in attendance by gender, although none of the samples in these studies was comprised primarily of opioid use disorder patients (Moos et al., 2006; Weiss et al., 2000; Witbrodt and Delucchi, 2011). Prior analyses of POATS data revealed no effect of gender on opioid use outcomes yet this study also indicated important gender differences, including differences in motivations for opioid use and higher rates of depression diagnosis and current depressive symptoms among women (McHugh et al., 2013). Thus, we included gender in our analyses of depression and opioid treatment outcome.

2. Method

This secondary analysis used data from the Prescription Opioid Addiction Treatment Study, sponsored by the National Drug Abuse Treatment Clinical Trials Network. This national, 10-site randomized controlled trial used a two-phase adaptive treatment research design to compare different durations of buprenorphine-naloxone treatment and different intensities of counseling to treat patients with prescription opioid dependence (for details, see Weiss et al., 2010b). The first phase of the study (N=653) consisted of a 4-week buprenorphine-naloxone taper plus an 8-week follow-up. Those who relapsed to opioid use during Phase 1 were then offered an extended, 12-week treatment phase. In both phases, patients were randomly assigned to standard medical management alone (Fiellen et al., 1999) or standard medical management plus additional individual opioid dependence counseling (Pantalon et al., 1999. This report focuses on the second phase of the trial (n=360), during which patients were maintained on buprenorphine-naloxone stabilization for 12 weeks.

Following approval by the Institutional Review Board at each study site, potential participants were evaluated. Participants were required to meet DSM-IV criteria for current opioid dependence and to be ≥18 years old. Key exclusion criteria included past-month heroin use on >4 days, a lifetime diagnosis of opioid dependence due to heroin alone, or a history of heroin injection (Weiss et al., 2010a). Those who needed to continue opioid use for pain management and those with current unstable psychiatric illness or ongoing formal SUD treatment were also excluded (for details, see Weiss et al., 2011). Written informed consent was obtained from all subjects.

2.1. Measures

During the 12-week extended treatment phase of the study, daily opioid use other than the study medication was assessed weekly with a self-report measure, the Substance Use Report, using a calendar technique to assist recall, similar to the Timeline Follow-back method (Sobell and Sobell, 1992). This measure was corroborated at each weekly visit by urine drug assays. Participants were considered to have a successful outcome if they were abstinent during week 12 of the extended treatment phase as well as abstinent during at least 2 out of the previous three weeks (for details, see Weiss et al., 2010b).

At baseline, additional assessments were administered to all participants. The Composite International Diagnostic Interview (Robins et al., 1988), a widely used and well-validated interview, was used to assess substance use disorders, major depressive disorder, and posttraumatic stress disorder. The Beck Depression Inventory-II (Beck et al., 1996), a 21-item self-report measure of severity of current depressive symptoms, was administered at baseline and throughout treatment. BDI scores are a well-established indicator of current depression severity; prior research indicates the validity of this measure in a variety of settings and populations (Dozois et al., 1998). Two measures developed for this study were also administered: the self-report Pain and Opioid Analgesic Use History and the Concomitant Treatments interview, a log of all non-study treatments including mutual-help groups. Demographic characteristics were collected by self-report.

2.2. Analysis Plan

First, a mixed effects model was used to assess change over time in depression scores, adjusted for a lifetime history of MDD. Next, we examined a series of multivariable logistic regression models adjusted for gender, treatment condition, and lifetime heroin use. Initially, we present the association previously reported between depression history and opioid outcome (Dreifuss et al., 2013). To test the hypothesis that a reduction in depressive symptoms accounts for the association between depression and good treatment outcome, we added depression scores at week 4 of treatment to the model, adjusted for baseline depression scores. To test the hypothesis that greater motivation and engagement in treatment account for the association between depression and good treatment outcome, we added opioid abstinence as a goal and mutual-help group attendance to the model. Adjusted odds ratios are presented. Data were analyzed using SPSS v.24.

3. Results

3.1. Sample description at baseline

Study participants at baseline (N=360) were 18 to 63 years old, with a mean age of 32.5 (sd=9.7). Most (90.6%) were white, with 7 to 22 years of education, mean (sd)=12.9 (2.2). Less than half (41.9%) were female, most (60.3%) were employed full-time, and 50.0% were never married.

About one-third (35.0%) reported any prior opioid dependence treatment, including mutual-help groups (21.1%); 26.1% had ever used heroin. Most participants (82.8%) had used opioids by a non-standard route of administration such as intranasal use or crushing. Approximately half (55.3%), reported that their first source of prescription opioids was a legitimate prescription, typically taken for physical pain (64.7%). The most commonly reported goal for treatment was lifetime abstinence from prescription opioids(60.3%). Although most had prescription opioid dependence only, 18.6% had an additional substance dependence diagnosis.

Beck Depression Inventory scores were high, on average (mean=23.1, sd=12.0); one-third of the participants had lifetime major depression disorder (34.2%) and one-fifth had current major depressive disorder (20.0%). Current posttraumatic stress disorder was diagnosed for 12.8% of the participants. Overall, 50.0% had a current psychiatric disorder other than substance use disorder. Current chronic pain was common (41.4%).

3.2. Depression scores during the treatment study

Beck Depression Inventory scores were examined at baseline and during the 12-week buprenorphine treatment. As shown in Figure 1, participants with lifetime major depression diagnoses scored significantly higher on the depression scale than those without major depression at each time, from baseline through the 12-week treatment, as expected F(1, 347.10)=20.95, p<.001). Figure 1 also shows that depressive symptom scores decreased over time both for participants with and without lifetime major depression (F(3, 322.31)=194.22, p<.001); the greatest reduction was early in treatment, regardless of MDD diagnosis.

Figure 1.

BDI scores by lifetime major depression from baseline through treatment (N=360)^a,b

^aChange over time: F(3, 322.31)=194.22, p<.001; Post hoc tests: baseline-week 4, p<.001; week 4–8, p=.005; week 8–12, p=.34^bMajor depression: F(1, 347.10)=20.95, p<.001 The horizontal bar is set at 13 to note the cutoff score for mild depression (Beck et al., 1996).

3.3. Multivariable models

The counterintuitive finding from the POATS trial that a major depression diagnosis predicted successful opioid use outcomes was the starting point of the current analysis (Table 1, Model 1): the likelihood of successful opioid outcomes at the end of the trial (weeks 9–12) was greater for participants with a history of MDD (aOR=1.82, p<.01) in an adjusted multivariable model (Dreifuss et al., 2013).

Table 1.

Models to predict successful opioid outcomes during a 12-week treatment study

Predictors of good opioid outcomes	Odds ratios, adjusted
	Model 1	Model 2	Model 3
	(N=360)	(N=317)	(N=317)
Gender	0.93	0.82	0.97
Treatment conditiona	0.78	0.71	0.76
Heroin ever	0.56*	0.57*	0.53*
Major depression lifetime	1.82**	1.63t	1.81*
Depression score at baseline		1.01
Depression score at week 4 of treatment		0.98
Abstinence goal			1.39
Mutual-help groups during treatment			1.67*

^tp<.055;

^*p<.05;

^**p<.01

^aStandard medical management alone or combined with opioid counseling

Note. All variables were assessed at baseline (N=360), with the exception of week 4 depression symptoms and mutual-help group attendance (N=317).

Hypothesis 1 stated that a reduction in depression symptoms accounts for the association between history of MDD and good treatment outcome. To test this hypothesis, the original multivariable model was expanded (Table 1, Model 2). Depression symptom scores at week 4 of treatment, adjusted for baseline scores, were not significantly associated with opioid outcome at the end of the trial (weeks 9–12). Hence a reduction in depressive symptoms did not account for the association between MDD and good treatment outcome. The effect of major depression was slightly attenuated when depression scores were added (Model 2; aOR=1.63, p=.055).

Hypothesis 2 stated that greater motivation and engagement in opioid use treatment account for the association between a history of depression and opioid use outcome. The following variables served as proxies for treatment motivation: 1) the self-reported treatment goal of opioid abstinence (measured at baseline), and 2) engagement: any attendance at mutual-help groups during treatment. To test Hypothesis 2, these two variables were added to Model 1 (Table 1, Model 3). Patients who attended mutual-help groups during the treatment trial were more likely to have successful opioid outcomes (aOR=1.67, p =.022), whereas opioid abstinence as a treatment goal was not associated with opioid outcome. Again, the effect of major depression remained statistically significant in this model (aOR=1.81, p =.012). Participants with any lifetime heroin use were half as likely to have successful opioid outcomes (aOR=0.53–0.57, p <.05), whereas gender and treatment condition were not associated with opioid outcomes in any of the models. The interaction between gender and lifetime MDD as a predictor of opioid outcome was also tested, and this effect was not significant. Since gender differences in mutual-help group participation have been reported, we examined the association between gender and mutual-help group participation in this sample. Gender was not associated with mutual-help attendance (58.4% of men vs. 64.9% of women had attended mutual-help groups during treatment).

4. Discussion

Given the general evidence for depression as a predictor of poor outcomes in substance use disorders, the finding from the POATS trial that depression history predicted successful opioid use outcomes is counterintuitive. This secondary analysis tested several pathways by which depression history may influence opioid use outcomes. Contrary to hypotheses, no evidence emerged to support either a reduction in depression symptoms or a baseline goal of abstinence as links between depression history and good treatment outcome. However, consistent with one hypothesized mechanism, we found that greater engagement in treatment, as reflected by attendance at mutual-help groups, predicted successful opioid use outcome. It is noteworthy that mutual-help group attendance contributed to a good opioid outcome without reducing the effect of MDD.

We did not find evidence to support the hypothesis that the influence of a history of depression on good opioid outcomes could be explained by a reduction in current depressive symptoms, perhaps by buprenorphine. Although depressive symptom levels were high at the beginning of treatment (with averages falling in the “severe” range for those with a history of depression and the “moderate” range even for those participants with no history of depression; Beck et al., 1996), patients with and without a history of major depression showed significant declines in symptoms during treatment; this change in symptoms did not predict opioid outcomes. These decreases in depressive symptoms are consistent with previous studies showing improvement in depressive symptoms during treatment for opioid use disorders (Dean et al., 2004; Kosten et al., 1990; Krupitsky et al., 2016; Latif et al., 2019; Mysels et al., 2011; Romero-Gonzalez et al., 2017). Moreover, prior studies report that patients show robust declines in BDI scores simply upon entering treatment for opioid use (Stice et al., 1991), and that depression symptoms are typically higher among patients who seek treatment for opioid use, as compared to those not seeking treatment (Rounsaville and Kleber, 1985). Results of the present study are thus consistent with relatively high levels of depression followed by general improvement in depressive symptoms upon treatment initiation, regardless of prior depression history.

Further, our results failed to support the hypothesis that one of our measures of motivation for treatment, the goal of opioid abstinence, was associated with good opioid outcomes. Based upon previous analyses of POATS data indicating gender differences in clinical characteristics of prescription opioid use (McHugh et al., 2013), and on evidence that depression may differentially influence substance use outcomes for men and women (for review, see Greenfield et al., 2007), gender was included in analyses of depression. However, gender was not associated with opioid use outcome in multivariate models.

Considered together, the present findings provide limited support for hypothesized mechanisms between depression history and good opioid treatment outcome. Although the original finding of depression history as a predictor of good opioid outcomes has been reported in other samples (e.g., Gerra et al., 2004; Rao et al., 2004), a significant number of other studies find that depression history has deleterious effects on the course of substance use disorders, raising the possibility that the link between depression history and opioid use outcome may be spurious. These inconsistent findings on depression history and substance use outcome may be related to the complicated temporal relationship between depressive episodes and substance use. For example, both the course of substance use disorders and the impact of depression treatment on substance use outcomes may vary based on whether a depressive episode was present prior to the substance use disorder (e.g., Hasin et al., 2002; Nunes et al., 1998). Although depressive episodes were verified as independent in the present study, it is unclear if their onset was prior to or following first use of opioids. Future studies should carefully assess the timelines of depression and substance use disorder to assess the impact of depressive episodes.

Findings of this study illustrate the role of treatment engagement as a significant contributor to opioid use outcomes. Although pre-treatment motivation (operationalized as a self-reported goal of abstinence) did not predict opioid outcomes, greater actual engagement in treatment (measured by attendance at mutual-help groups) significantly predicted opioid use outcomes, over and above the influence of depression. Others have argued that enhanced motivation due to depression could also motivate greater engagement in substance use disorder treatment (e.g., Morris et al., 2018); more research is needed to specifically test this possibility in opioid use disorder specifically.

These analyses are limited by our focus on depression, as other psychiatric symptoms such as anxiety disorders may also motivate treatment engagement. Given that much of the literature on psychiatric symptoms and motivation for treatment is based on depressive symptoms, expanding this question to other types of psychiatric illness may be an important goal for future research. Finally, although previous research on the efficacy of mutual-help groups for co-occurring depression and substance use disorder has been mixed (Glasner-Edwards et al., 2007; Kelly et al., 2003), results of the present study are encouraging with the regard to the benefits of mutual-help participation on opioid use outcomes, regardless of depression. Indeed, our 42-month follow-up study to the POATS trial found mutual-help group attendance (along with buprenorphine maintenance treatment) to be correlated with opioid abstinence over time (Weiss et al., 2019). An important next step for future studies will be to examine potential mediators of the associations among depression, mutual-help groups, and opioid use outcomes.

Several limitations are important to note. As noted above, the POATS trial was not designed to compare buprenorphine to another pharmacological treatment; thus, it is not possible to compare potential antidepressant effects of buprenorphine to another condition. Also as noted above, the study did not include a validated measure of motivation to engage in treatment, which limits our ability to understand the role of motivation in driving the outcomes of this study. Although depression independent of substance use was verified by interview at baseline, participants may have experienced substance-induced symptoms at the time of treatment, limiting our ability to specify how changes in depressive symptoms may have differed between the groups by depression history. Finally, although depressive episodes were verified as independent, this analysis did not assess the temporal course of depressive episodes and opioid use, which limits our ability to assess how the course of lifetime depression may have affected opioid treatment outcome.

5. Conclusions

In summary, this secondary analysis examined pathways from major depression to treatment outcomes for prescription opioid use disorders, and found limited evidence that hypothesized mechanisms of change in depression symptoms and motivation could explain the beneficial effect of depression history on treatment outcome. Findings also demonstrate the benefit of attendance at mutual-help groups. Future research is needed to understand how recovery from opioid use disorder may contribute to changes in depressive symptoms over time. A better understanding of these factors could eventually improve treatment outcomes for those with opioid dependence.

Highlights.

• A secondary analysis of the Prescription Opioid Addiction Treatment Study (POATS).

• Tested explanations for the finding that MDD history predicted good opioid outcome.

• Depression decreased during treatment, but was not associated with opioid outcomes.

• Mutual-help group participation was associated with good treatment outcomes.

• Lifetime Major Depressive Disorder remained associated with good outcomes.