Trajectory Classes of Opioid Use Among Individuals in a Randomized Controlled Trial Comparing Extended-Release Naltrexone and Buprenorphine-Naloxone

1. Introduction

More than 2.1 million people in the U.S. have an opioid use disorder (OUD), a chronic condition with 65% to 84% of individuals relapsing within a month after completing inpatient or outpatient detoxification (Bailey et al., 2013; Day and Strang, 2011). Opioid overdose deaths have steadily increased over the last 15 years with 47,600 reported in 2017 (Center for Disease Control and Prevention, 2019). Synthetic opioids (other than methadone) such as the highly potent fentanyl, fentanyl analogs, and tramadol are responsible for more than half (29,000) of opioid-related overdose deaths (National Institute on Drug Abuse (NIDA), 2018); between 2002 and 2017 there was a 22-fold increase in the total number of overdose deaths attributable to synthetic opioids. Thus, there is a dire need to identify more effective and targeted methods of treating OUD.

Three pharmacologically distinct FDA-approved medications for OUD (MOUD) have been found to be effective, the full opioid agonist methadone, the partial agonist buprenorphine-naloxone [BUP-NX], and the antagonist naltrexone formulated as an extended-release injection [XR-NTX]. These are considered front-line treatments (Volkow, 2018) and traditionally combined with psychosocial interventions to provide comprehensive care. Methadone in the U.S. is only available in regulated specialty settings (i.e., Opioid Treatment Programs), while BUP-NX and XR-NTX can be prescribed across a range of healthcare settings, and have the potential to increase the accessibility of MOUD. BUP-NX and XR-NTX each have unique induction requirements portending distinct challenges in initiation and maintenance of treatment among patients. This is particularly true of XR-NTX which requires patients to be fully detoxified from opioids (e.g., no opioids for approximately 7–10 days; Jarvis et al., 2018); however, once induction hurdles are overcome, XR-NTX appear to be as effective as BUP-NX (Lee et al., 2018; Tanum et al., 2017). For example, a multi-site randomized controlled trial (RCT) comparing daily oral BUP-NX with flexible dosing (from 4 to 24 mg/day) to monthly intramuscular XR-NTX (380 mg) revealed no differences between the medications in reducing opioid and other illicit drug use in the short-term (i.e., 12 weeks; Tanum et al., 2017). A multi-site RCT in the United States, compared BUP-NX (administered sublingually via film) to XR-NTX (administered monthly via intramuscular injection), in a sample of individuals with OUD, over 24 weeks (Lee et al., 2018). Although significantly fewer participants successfully initiated XR-NTX, among those who were fully inducted on either XR-NTX or BUP-NX (per protocol treatment groups), there were no significant differences between the two treatments in weeks to relapse, opioid-negative urines, and self-reported opioid abstinent days. Thus, although XR-NTX may present unique induction challenges, studies support its efficacy as an additional treatment for OUD.

While the aforementioned trials provided important information on mean-level differences in rates of opioid use and relapse, there is scarce information on the variability in how individuals’ opioid use changes over time while receiving MOUD. Identifying differential use patterns and treatment response and the characteristics that predict them is critical in understanding how MOUD work among subgroups of individuals, and can inform the timing and targeting of intervention efforts (Grella and Lovinger, 2011; Hser et al., 2017; Teesson et al., 2017). The limited research to date has demonstrated heterogeneity in opioid use trajectories for people on MOUD. In a sample of 795 people who use opioids and participated in a RCT of methadone or BUP-NX, Hser et al. (2017) identified 4 opioid use trajectories over time (up to 55 months posttreatment): low use, high use, decreasing use, and increasing use. More than 40% of the participants were in the low use group (demonstrating consistently low opioid use over time) after MOUD treatment. Eastwood and colleagues (2018) examined opioid use trajectories in a sample of 7,719 individuals continuously enrolled in community based MOUD (methadone or buprenorphine) for at least five years and followed posttreatment for another two years. They found five heroin use trajectory classes: gradual decreasing, decreasing then increasing, continued low-level, rapid decreasing, and continued high-level (Eastwood et al., 2018). Variables that predicted non-response to treatment included current injection drug use, previous treatment, and higher social deprivation. These studies demonstrated variability in MOUD treatment response and highlight the benefit of adherence to treatment. Overall, studies show the strongest predictor of abstinence over follow-up was being on MOUD (Weiss et al., 2015). Findings underscore the need to identify treatment response early in the course of treatment to modify and optimize treatment interventions. Notably, no studies have examined patterns of treatment response during the first 24 weeks of treatment and in particular among patients randomized to XR-NTX.

The aim of this secondary analysis was to examine trajectories of opioid use for patients with OUD who were randomized to receive 24 weeks of treatment with either BUP-NX or XR-NTX, and to identify characteristics associated with each trajectory class. In particular, we were interested in exploring opioid use trajectories among individuals randomized to XR-NTX since this analysis had not previously been conducted.

2. Method

3. Results

4. Discussion

To advance our understanding of MOUD treatment response, the identification of distinct subgroups and factors associated with differential treatment response is critical to inform intervention efforts in terms of optimal targets and intervention leverage points. This secondary analysis identified four distinct opioid use trajectories for BUP-NX and three distinct opioid use trajectories for XR-NTX. Overall, the trajectory classes for XR-NTX and BUP-NX were quite similar (i.e., both had “near abstinent/no use”, “low use”, and “moderate use, increasing use over time” classes; BUP-NX had an additional class of “low use, increasing use over time”). We found the “near abstinent/no use” classes across both BUP-NX and XR-NTX treatments had the highest number of medical management weeks (mean of 20 weeks) relative to other classes. Medical management and adherence to treatment are deeply intertwined. Medical management focused on establishing and maintaining provider-patient rapport, reviewing medication adherence and side-effects and non-study opioid use, and promotion of use of psychosocial treatments. In contrast, those who were in the “moderate use, increasing use over time” class across both BUP-NX and XR-NTX attended only 7.8 and 8.1 weeks of medical management, respectively, suggesting interventions that increase attendance to medical management as well as adherence to MOUD, especially early in treatment, are needed. It is likely the medical management visits reflected general treatment adherence. Individuals who attended fewer medical management visits were less likely to receive or take their dose of BUP-NX or XR-NTX and thus, were more likely to relapse. Or conversely, those who relapsed were less likely to attend their medical management visits.

Hser and colleagues (2017) also identified four opioid use trajectories for BUP-NX, however, their trajectories were different from those identified in this analysis. They found trajectories of low use, high use, decreasing use, and increasing use after treatment with methadone or buprenorphine-naloxone. Unlike Hser, we did not find a decreasing use trajectory for BUP-NX. Eastwood et al (2018) identified five trajectories of heroin use (gradually decreasing; decreasing then increasing; continued low-level; rapid decreasing; continued high level) over five years of treatment with methadone or buprenorphine. Eastwood et al.’s continued low-level subgroup is consistent with our low use subgroups; likewise, their decreasing then increasing subgroup is similar to BUP-NX’s low-use then increasing use class. Differences between the current and previous trajectory analyses are likely a function of differences in study and patient characteristics. The prior studies were multi-year posttreatment follow-up class analyses and this current analysis focused on identifying within treatment trajectory classes.

History of previous successful treatment emerged as an important pre-treatment variable distinguishing the BUP-NX classes, with a greater proportion of those in the low use group reporting prior treatment success. Given the chronic relapsing condition of OUD and the need for multiple treatment attempts, it makes sense that history of successful treatment is predictive of more positive treatment outcomes (Dennis et al., 2005). The finding that, among those randomized to XR-NTX, more Hispanics were in the moderate use, increasing use over time trajectory class is also consistent with studies demonstrating poorer opioid use outcomes for those who are ethnic minorities (e.g., in Hser et al.’s (2017) sample, Hispanics were more likely to be in the high use group relative to the low use group). It is possible that the medical management services did not have the same benefit for Hispanics; or Hispanic identity may serve as a proxy for other variables such as social and environmental disadvantages that contribute to health disparities in general (Saloner and Cook, 2013; Windsor et al., 2015). Moreover, although all Hispanic patients enrolled in this trial were English-proficient, consideration of the ways in which language/cross-cultural barriers between patients and providers may be related to access and utilization of mental health and substance abuse treatment is imperative (Guerrero et al., 2013; Sentell et al., 2007). Finally, history of psychiatric treatment emerged as a significant pre-treatment variable distinguishing the XR-NTX classes, with a greater proportion of those in the low use group reporting receiving prior psychiatric treatment. It is unknown why this variable emerged as a significant distinguishing factor for the XR-NTX classes and not for those who received BUP-NX.

Several limitations must be noted. First, our opioid use outcome variable was based on self-reported drug use, which is subject to recall bias and social desirability effects. Nevertheless, self-reported substance use has been shown to correspond well with urine toxicology, when self-reports are elicited in research contexts where negative consequences will not ensue from acknowledging drug use (Clark et al., 2016; Wilcox et al., 2013). Second, our retained subsample included participants who were not successfully inducted onto an initial dose of study medication (4% of BUP-NX and 24% of XRNTX participants), thus overall findings should be interpreted with caution. Some of the participants in the continuous opioid use classes were likely in those subgroups because they never started the medications. Given non-compliant participants tend to have poorer outcomes, their exclusion tends to overestimate the quality of the treatment outcomes. Thus, the advantage of using the retained subsample was that we avoided overestimating the benefits of MOUD (Gupta, 2011). Third, given BUP-NX and XR-NTX were initiated after participants completed inpatient detoxification and with varying induction strategies, findings may not be generalizable to those who initiate treatment in an outpatient setting. Fourth, several of our trajectory classes had small sample sizes, thus it is possible we were underpowered to detect significant differences among certain subgroups. In addition, although we examined an extensive array of demographic, psychosocial, and treatment characteristics, we likely did not measure all potential factors that could have differentiated the subpopulations.

Future research is recommended to replicate study findings with a larger, outpatient sample size and a greater number of predictors of trajectory class membership. Despite limitations, this study is one of the few clinical trials that have examined trajectories of opioid use during a RCT of BUP-NX and the first study to examine trajectories of use after randomization to XR-NTX.

The study findings have several implications for clinical treatment and policy. The largest subgroups across those randomized to receive BUP-NX or XR-NTX were in the no use/near abstinent groups, which was associated with attending more medical management appointments, underscoring the need for novel engagement strategies that increase medical management visits and treatment adherence. Contingency management and community reinforcement approaches have shown promise in these areas (Carroll and Weiss, 2016). The differential patterns of opioid use trajectories in MOUD treatment highlight the importance of providers regularly monitoring their patient’s opioid use over time to anticipate treatment response and adjust treatment accordingly. Studies have shown that opioid use during the early weeks of treatment predicts poorer outcomes (Weiss and Rao, 2017). Thus, if a patient’s opioid use is increasing over time, providers may consider either an increase in the current medication dosage, change in type of medication received, or an increase in psychosocial treatment or support that enhance treatment adherence (Murphy et al., 2007). More research is needed on the best medical or psychosocial strategy to improve prognosis among those with OUD. The findings that demographic factors such as ethnicity and homelessness were associated with particular opioid use trajectory classes emphasize the need for interventions tailored to these specific subpopulations. Enhancing recovery capital in the form of safe housing and additional social/treatment supports may help promote sustained abstinence from opioids (Groshkova et al., 2013; Laudet and White, 2008).

In sum, our findings highlight the significant heterogeneity of opioid use during a RCT of MOUD treatment and several factors associated with longitudinal patterns of opioid use that can be effectively targeted in the context of active clinician monitoring and adaptive treatment strategies tailored to patient needs. Policies that increase access and adherence to MOUD treatment, alongside an understanding of the unique patterns of MOUD response and the factors that influence it, will be critical in reducing the current opioid crisis.

Supplementary Material