Opioid use disorder in primary care

Data sources

The evidence team created a search strategy with guidance from an experienced librarian for each of the clinical questions created. Two authors (D.P., J.T.) performed the search for systematic reviews and randomized controlled trials (RCTs) for each clinical question with no language restrictions. The search was restricted to nonanimal studies. The databases and resources used to search for relevant systematic reviews included MEDLINE, Cochrane Library, Google, published guidelines on OUD, and reference lists of the included systematic reviews. The search included any articles up to June 2018, but was generally limited to the past 5 to 10 years. Key words opioid or opiate were used for all searches. Specifics for each question and the corresponding key words, timelines, and search strategies used can be found in the appendix (CFPlus).^* After the search for systematic reviews was complete, an additional search of MEDLINE was undertaken to find RCTs published since the most recent systematic review for each clinical question. Reference lists of included articles were hand searched to identify potentially missed articles.

Study selection

Beyond systematic reviews and newer RCTs, inclusion criteria were studies of adult patients with OUD reporting on at least 1 of the following outcomes: morbidity and mortality, social outcomes, quality of life and symptoms, or opioid use outcomes (these are defined in Table 1). Systematic reviews of observational studies were included; however, observational data were only considered when RCTs did not exist. Individual observational studies were not used to inform recommendations. Exclusion criteria were studies on detoxification from opioids; studies in pediatric, pregnant, or cancer patients; and studies completed within a prison setting. Any exceptions made were recorded in the appendix (CFPlus).^*

Dual title, abstract, and full-text review were completed for all systematic review and RCT searches to determine study eligibility. Single review was completed for guidelines and their references, with dual assessment if full-text review was required. Disagreements over inclusion were resolved by consensus.

Data extraction

Dual data extraction was completed using templates created by 2 authors (C.R.F., J.T.), one specifically for systematic reviews and one for RCTs. For systematic reviews, data extracted included author, year, title, study design, general characteristics, setting, sex, mean age, mean duration, duration range, outcomes reported (along with number of studies, RCTs, and patients for each outcome), values associated with the outcomes, the intervention, and the control. If no usable data were found in a given systematic review, authors attempted to obtain that data from the included trials.

Following extraction, data tables of systematic reviews and RCTs were created with headings for total studies, age, population, relevant studies, duration of studies, intervention, outcomes, and risk-of-bias quality assessment. The data tables created can be found in the appendix (CFPlus).^*

Risk-of-bias assessment

Risk of bias was assessed using a modified AMSTAR (A Measurement Tool to Assess Systematic Reviews) rubric for systematic reviews, focusing on the 6 most relevant questions^7,8:

• Was study selection and data extraction performed by dual reviewers?

• Was the literature search comprehensive?

• Were the included study characteristics described?

• Was the quality of the included studies assessed and reported?

• Were the methods used to combine results appropriate?

• Were conflicts of interest reported?

For systematic reviews, each question was scored as 1 (completed) or 0 (not completed). These individual scores were then summated, with a higher total score suggesting a lower risk of bias. For RCTs, the Jadad 5-point scoring rubric was used.⁹ The risk-of-bias assessment for each article was completed by at least 2 independent authors, and disagreement was resolved by consensus or a third author.

Analysis

Following data extraction, we used study outcomes and meta-analyses to answer each clinical question. We reported study characteristics and outcomes descriptively using means and other statistical results as per original papers. We prioritized systematic reviews of RCTs and individual RCT results over systematic reviews of observational data. Where outcomes were measured in various ways, we preferentially reported on the more objective outcomes. For example, for the outcome of continued opioid use in studies of pharmacotherapy, we report on the results of urine drug tests over self-report outcomes.

Performing new meta-analyses

If no relevant meta-analyses existed or if relevant RCTs had been published since the most recent systematic review, a new meta-analysis was completed using the RevMan 5 software. We used a Mantel-Haenszel statistical method and focused on reporting risk ratios (RRs) when appropriate. Not wanting to overweigh smaller studies, we chose a fixed-effects analysis if there was no reason to speculate that the effect of the intervention would deviate meaningfully between studies. We assessed heterogeneity using the I² statistic. Values greater than 50% were indicative of “high heterogeneity” and suggested a sensitivity analysis be completed to determine the cause of the heterogeneity. Additionally, we performed an exploratory meta-analysis of the effects of buprenorphine, methadone, and naltrexone on mortality. Owing to the low event rate, mortality events from the 3 treatments were combined and meta-analysis was completed using the exact method with odds ratios. ¹⁰

No RCT data available

Overall, 9 of the 17 systematic reviews we completed had either no RCT data available for the specified outcomes or the data were considered inconclusive (Box 1). No systematic review or RCT had data to support all outcomes, and no individual systematic review or RCT provided adequate data on morbidity and mortality (Table 2).

Management of OUD in primary care

No previous meta-analysis was available; however, 4 RCTs compared the management of OUD in primary care with that in specialty care (numbers of participants ranged from 46 to 221). Three of these looked at patient satisfaction rates and found statistically significantly higher rates (ie, more satisfaction) with primary care (eg, 77% vs 38%). We performed a meta-analysis of the effect of treatment setting on retention and found program retention was 86% in primary care versus 67% in a specialty clinic (RR = 1.25, 95% CI 1.07 to 1.47, I² = 18%) (Figure 1).¹¹ Street opioid abstinence was also higher in primary care settings (53% vs 35%, RR = 1.50, 95% CI 1.12 to 2.01, I² = 74%); however, heterogeneity was high and this included both self-reported and urine-confirmed data (Figure 2).¹¹

• Download figure
• Open in new tab

Figure 1.

Treatment retention in primary care versus specialty care

^*Mantel-Haenszel fixed method.

Meta-analysis of studies from Lagisetty et al.¹¹

• Download figure
• Open in new tab

Figure 2.

Street opioid abstinence in primary care versus specialty care

^*Mantel-Haenszel fixed method.

Meta-analysis of studies from Lagisetty et al.¹¹

Diagnosis

Fourteen systematic reviews on identifying patients with OUD were found, but none assessed diagnostic criteria. Two case-finding tools were compared with the Diagnostic and Statistical Manual of Mental Disorders (4th or 5th edition) criteria: the COMM (Current Opioid Misuse Measure), a 17-question scale, and the POMI (Prescription Opioid Misuse Index), a 6-question checklist. Both have been assessed in only 1 cohort study each, reporting positive likelihood ratios of 3.35 and 10.3, respectively (CFPlus).^*

Treatment

Pharmacotherapy

Buprenorphine: We found 2 systematic reviews and an additional 6 RCTs (as 9 publications) of buprenorphine alone or combined with naloxone. Compared with placebo, buprenorphine significantly retained more patients in treatment (64% vs 39% for placebo, number needed to treat [NNT] of 4 at 30 days to 52 weeks; RR = 1.66, 95% CI 1.52 to 1.82, I² = 86%) (CFPlus).^*

Methadone: One systematic review and 1 RCT of methadone were found. Retention in treatment was higher with methadone compared with no methadone (73% vs 22% for controls, NNT = 2 at 45 days to 2 years; RR = 3.37, 95% CI 2.83 to 4.02, I² = 73%) (CFPlus).^*

Our meta-analysis of 24 RCTs directly comparing buprenorphine with methadone revealed higher retention rates with methadone (45% vs 60% with methadone, NNT=7; RR = 0.75, 95% CI 0.71 to 0.80) (Figure 3).^12–15 However, substantial heterogeneity was present (I² = 72%). This also differed from a published systematic review that found no difference in retention rates between buprenorphine and methadone.¹⁶ Neilsen and colleagues’ systematic review meta-analyzed subgroups of patients from 3 of the above studies who used prescription opioids rather than heroin.¹⁶

• Download figure
• Open in new tab

Figure 3.

Retention in treatment: Buprenorphine versus methadone.

^*Mantel-Haenszel fixed method.

Meta-analysis of Neumann et al,¹² Piralishvili et al,¹³ Potter et al,¹⁴ and studies from Mattick et al (all other studies).¹⁵

Overall, opioid abstinence appears higher with methadone than with buprenorphine (Figure 4).^12,14,15 However, there was a statistically significant difference between subgroups of studies that measured abstinence objectively and those that relied on self-report (P < .001). If only studies that used objective measures are included, there is no difference in abstinence between buprenorphine and methadone (RR = 0.99, 95% CI 0.78 to 1.24, I² = 0%).

• Download figure
• Open in new tab

Figure 4.

Abstinence: Buprenorphine versus methadone.

^*Mantel-Haenszel fixed method.

Meta-analysis of Neumann et al,¹² Potter et al,¹⁴ and studies from Mattick et al (all other studies).¹⁵

Adverse effects were poorly reported in both the buprenorphine and the methadone literature. Two RCTs found no difference between the drugs, except for more sedation with methadone (58% vs 26% with buprenorphine) in 1 RCT. Two RCTs found fewer adverse effects with buprenorphine than in controls (CFPlus).^*

Naltrexone: Two systematic reviews and 5 RCTs were found on the opioid antagonist naltrexone. Indirect comparison reveals lower rates of retention than with OAT, but naltrexone is still better than placebo or usual care (33% vs 25% for controls, RR = 1.35, 95% CI 1.11 to 1.64, I² = 0%) (CFPlus).^* Although results of subgroup analysis of oral naltrexone were not statistically significant (RR = 1.32, 95% CI 0.97 to 1.79), they were numerically similar to the results for injectable naltrexone, and results of the test for subgroup differences between oral and injectable forms were not significant (P = .86). Naltrexone also increased abstinence from opioids (39% vs 27% for controls, RR = 1.48, 95% CI 1.11 to 1.98, I² = 63%) (CFPlus).^* Based on 4 small RCTs, naltrexone decreases re-incarceration (24% vs 33% for controls, RR = 0.69, 95% CI 0.51 to 0.94, I² = 0%) (CFPlus).^*

Buprenorphine, methadone, and naltrexone combined event rates: As mortality rates were very low across buprenorphine, methadone, and naltrexone studies, we performed an exploratory meta-analysis combining event rates for all 3 drugs and found a statistically significant reduction in overall mortality with the use of pharmacotherapy in patients with OUD (odds ratio of 0.34, 95% CI 0.10 to 0.95, 7 RCTs).

Psychosocial supports.

Eight systematic reviews were identified on psychosocial supports. There was substantial variation with regard to inclusion criteria and analysis; thus, we prioritized 5 key interventions and assessed individual RCTs identified from the systematic reviews.

The addition of standard counseling to OAT is more effective in retaining people in treatment than no or minimal counseling (74% vs 62% for controls, NNT = 8; RR = 1.20, 95% CI 1.06 to 1.36, 3 RCTs); however, the heterogeneity was high (I² = 74%) (Figure 5).^17–19 No difference was noted between extended counseling sessions (45 to 60 minutes) compared with “standard” sessions of 15 to 20 minutes (RR = 1.19, 95% CI 0.88 to 1.62) (CFPlus).^*

• Download figure
• Open in new tab

Figure 5.

Retention in treatment: Counseling versus minimal to no counseling.

^*Mantel-Haenszel fixed method.

Meta-analysis of Liu et al¹⁷ and studies from Amato et al (Chawarski, 2011)¹⁸ and Dugosh et al (Gu, 2013).¹⁹

The use of contingency management, defined as either “rewards” for desired behaviour (eg, vouchers or prizes) or loss of privileges for undesired behaviour (eg, loss of medication carries for positive urine drug screening results), increases retention in treatment (RR = 1.11, 95% CI 1.06 to 1.17) (Figure 6).^18,20,21 Subgroup analysis suggests the benefits are primarily from positive contingencies (RR = 1.15, 95% CI 1.09 to 1.21), with negative or medication-related contingencies worsening retention (68% vs 77% for no contingency, RR = 0.86, 95% CI 0.76 to 0.99) (test for subgroup difference P < .001). Methods of reporting opioid use were too heterogeneous to be meta-analyzed.

• Download figure
• Open in new tab

Figure 6.

Retention in treatment: Contingency management versus no contingency management.

^*Mantel-Haenszel fixed method.

Meta-analysis of studies from Amato et al (Bickel, 2008; Chopra, 2009; Gross, 2006; Jiang, 2012; Kosten, 2003; Oliveto, 2005; Petry, 2005; Silverman, 2004; Stitzer, 1992),¹⁸ Ainscough et al (Chutuape, 1999; Chutuape, 2001; Kidorf, 1996; Ling, 2013; Peirce, 2006; Petry, 2002; Petry, 2007; Preston, 2000),²⁰ and Davis et al (Chen, 2013; Defulio, 2012; Dunn, 2013; Everly, 2011; Hser, 2011; Jiang, 2012; Kidorf, 2013).²¹

Management of comorbidities in patients with OUD.

There was inadequate RCT evidence in all searched areas (CFPlus).^*

Limitations

Limitations of this review included a lack of consistent terminology regarding OUD (eg, heroin abuse, opioid use, addiction, opioid dependency), which might have affected our ability to identify all relevant studies. Treatment studies were generally open label, suffered from high dropout rates, and included primarily patients using heroin as opposed to prescription opioids. Most studies were not designed to determine effects on morbidity and mortality, but instead focused on drug use outcomes and retention, which were inconsistently assessed and reported across trials.