Association of Opioid Agonist Treatment With Mortality in People With Opioid Dependence (2024)

Download PDF

• Error in Abstract

  JAMA Psychiatry

  Correction

  September 1, 2021

• Error in Abstract

  JAMA Psychiatry

  Correction

  May 1, 2022

• Errors in Data Reported in Meta-Analysis on Opioid Agonist Treatment and Mortality

  JAMA Psychiatry

  Comment & Response

  September 1, 2023

  ThomasSantoJr,MPH, PhD

• Errors in Results and Tables

  JAMA Psychiatry

  Correction

  September 1, 2023

Select Your Interests

Original Investigation

ThomasSantoJr,MPH¹; BrodieClark,BPsych¹; MattHickman,PhD²; et al JasonGrebely,PhD³; GabrielleCampbell,PhD^1,4; LuisSordo,MD, PhD^5,6; AileenChen,BPsych^1,7; Lucy ThiTran,BPsychSc¹; ChriannaBharat,BSc¹; PriankaPadmanathan,MRCPsych²; GrainneCousins,PhD⁸; JulieDupouy,MD, PhD^9,10; ErinKelty,PhD¹¹; RobertoMuga,MD, PhD¹²; BohdanNosyk,PhD^13,14; JeongMin,MSc¹³; RaimondoPavarin,MPH^15,16; MichaelFarrell,MD¹; LouisaDegenhardt,PhD¹

Author Affiliations Article Information

Key Points

Question Is opioid agonist treatment (OAT) associated with risk of overall and cause-specific mortality among people with opioid dependence?

Findings In this systematic review and meta-analysis, risk of all-cause, overdose, suicide, alcohol-related, cancer, and cardiovascular-related mortality was significantly lower for people with opioid dependence during OAT.

Meaning These findings suggest that increasing access to OAT and retention in treatment are critical for reducing rates of preventable mortality among people with opioid dependence.

Importance Mortality among people with opioid dependence is higher than that of the general population. Opioid agonist treatment (OAT) is an effective treatment for opioid dependence; however, there has not yet been a systematic review on the relationship between OAT and specific causes of mortality.

Objective To estimate the association of time receiving OAT with mortality.

Data Sources The Embase, MEDLINE, and PsycINFO databases were searched through February 18, 2020, including clinical trial registries and previous Cochrane reviews.

Study Selection All observational studies that collected data on all-cause or cause-specific mortality among people with opioid dependence while receiving and not receiving OAT were included. Randomized clinical trials (RCTs) were also included.

Data Extraction and Synthesis This systematic review and meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guidelines. Data on study, participant, and treatment characteristics were extracted; person-years, all-cause mortality, and cause-specific mortality were calculated. Crude mortality rates and rate ratios (RRs) were pooled using random-effects meta-analyses.

Main Outcomes and Measures Overall all-cause and cause-specific mortality both by setting and by participant characteristics. Methadone and buprenorphine OAT were evaluated specifically.

Results Fifteen RCTs including 3852 participants and 36 primary cohort studies including 749 634 participants were analyzed. Among the cohort studies, the rate of all-cause mortality during OAT was more than half of the rate seen during time out of OAT (RR, 0.47; 95% CI, 0.42-0.53). This association was consistent regardless of patient sex, age, geographic location, HIV status, and hepatitis C virus status and whether drugs were taken through injection. Associations were not different for methadone (RR, 0.47; 95% CI, 0.41-0.54) vs buprenorphine (RR, 0.34; 95% CI, 0.26-0.45). There was lower risk of suicide (RR, 0.48; 95% CI, 0.37-0.61), cancer (RR, 0.72; 95% CI, 0.54-0.98), drug-related (RR, 0.41; 95% CI, 0.33-0.52), alcohol-related (RR, 0.59; 95% CI, 0.49-0.72), and cardiovascular-related (RR, 0.69; 95% CI, 0.60-0.79) mortality during OAT. In the first 4 weeks of methadone treatment, rates of all-cause mortality and drug-related poisoning were more than double the rates during the remainder of OAT (RR, 2.81; 95% CI, 1.55-5.09) but not for buprenorphine (RR, 0.58; 95% CI, 0.18-1.85). All-cause mortality was 6 times higher in the 4 weeks after OAT cessation (RR, 6.01; 95% CI, 4.32-8.36), remaining double the rate for the remainder of time not receiving OAT (RR, 1.81; 95% CI, 1.50-2.18). Opioid agonist treatment was associated with a lower risk of mortality during incarceration (RR, 0.06; 95% CI, 0.01-0.46) and after release from incarceration (RR, 0.09; 95% CI, 0.02-0.56).

Conclusions and Relevance This systematic review and meta-analysis found that OAT was associated with lower rates of mortality. However, access to OAT remains limited, and coverage of OAT remains low. Work to improve access globally may have important population-level benefits.

Opioid dependence is increasing in many countries, particularly in North America, where there have been substantial increases in opioid-related health harms, specifically overdose.^1,2 In the US during the COVID-19 pandemic, opioid overdoses have increased in some states by up to 30% in 2020 compared with those in 2019.³ Population-level increases in “deaths of despair,” including suicides, injuries, and liver disease, have also been observed.⁴ People with opioid dependence are at an elevated risk of a range of causes of death beyond deaths of despair, including other acute and systemic causes such as unintentional opioid and suicide-related death, and all liver-related, alcohol-related, cancer-related, chronic respiratory–related, digestive-related, HIV-related, influenza- and pneumonia-related, and injection-related injuries.^2,5

Methadone and buprenorphine are classified by the World Health Organization as essential medicines for opioid agonist treatment (OAT) for opioid dependence.⁶ There is robust evidence from a recent systematic review that during OAT, overdose and all-cause mortality are reduced among people with opioid dependence.⁷ That review also found that people who cease OAT are at the highest risk of all-cause and overdose mortality in the first 4 weeks after treatment cessation⁷ and that risk of mortality is elevated in the first 4 weeks of OAT compared with the remainder of time receiving OAT.⁷

To our knowledge, there has not been a systematic examination of (1) the evidence on the potential association of OAT with other causes of death or (2) OAT provided in alternative settings, including during and immediately after incarceration. In this review, we aim to (1) examine and compare all-cause and cause-specific crude mortality rates (CMRs) during and out of OAT, for both randomized clinical trials (RCTs) and observational studies; (2) examine these rates according to specific periods during and after treatment; (3) examine and compare all-cause and cause-specific CMRs for OAT provided during incarceration, after release from incarceration while receiving OAT, and according to the amount of time receiving and not receiving OAT after release from incarceration; and (4) examine the association between risk of mortality during and out of OAT by participant and treatment characteristics.

This systematic review and meta-analysis was conducted from January 15 to February 18, 2020. As this was a review, this study was not approved by an institutional review board. We reported on published, peer-reviewed data. Each included study obtained approval from their respective jurisdictions. This study followed the Guidelines for Accurate and Transparent Health Estimates Reporting (GATHER),⁸ Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guideline,⁹ and Meta-analysis of Observational Studies in Epidemiology (MOOSE) guideline.¹⁰ The protocol was registered on PROSPERO.¹¹

We systematically searched 3 peer-reviewed databases (Medline, Embase, and PsycINFO) without language limitation; database searches were completed on February 18, 2020. Search terms included exploded MeSH terms and keywords for opioid dependence, OAT, and mortality (eAppendix in the Supplement). We consulted experts and investigators with ongoing studies of mortality among people with opioid dependence, authors of studies included in a previous review⁷ of OAT and mortality, systematic reviews of OAT and opioid dependence, and clinical registries.

Eligible studies had to report mortality data for people with opioid dependence during and out of OAT (prespecified exclusion criteria and excluded studies are in the eAppendix and eTable 1 in the Supplement). We included both observational studies and RCTs. Observational studies that reported any form of mortality and person-time data for time during OAT and out of OAT were included. Randomized clinical trials were required to report mortality data for participants allocated to OAT and those allocated to control or comparator interventions separately. Authors of studies that did not report mortality or person-time data during and after OAT separately were contacted for additional information.

Two reviewers independently reviewed the titles and abstracts identified in the search and retrieved articles to determine eligibility; full texts were also independently reviewed (T.S., B.C., L.T.T.). Conflicts were resolved by consensus with a third reviewer (L.D., J.G., G.C.). If additional data were required for the primary analyses in any setting or the secondary analyses (eg, medication type, time spent in or out of treatment, and participant characteristics), then the study authors were contacted.

We extracted information on participant and study characteristics, treatment modality characteristics, number of deaths, person-years at risk, and all-cause and cause-specific mortality rates during follow-up periods during and after treatment cessation (eTables 2, 3, and 4 in the Supplement). All extracted data were confirmed by a second person (T.S., B.C., or L.T.T.).

We assessed risk of bias using the Risk Of Bias In Nonrandomized Studies–of Interventions (ROBINS-I) tool for observational studies¹²; and Risk of Bias 2 (RoB-2)¹³ tool for RCTs. Two reviewers (T.S., B.C., or L.T.T.) assessed each study independently with conflicts resolved by a third party (L.D., J.G., or G.C.).

We reported on multiple causes of death using the primary cause of death assigned to each fatality. To standardize the definitions for cause-specific deaths included in each study, we contacted study authors asking them to provide data using specified International Classification of Diseases, Ninth Revision (ICD-9) and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10)¹⁴ codes (eTable 5 in the Supplement).

For RCTs, log-transformed rate ratios (RRs) were calculated by comparing binary death and participant totals using the Mantel-Haenszel fixed method. For observational cohorts, observed number of deaths were divided by person-years (PYs) to calculate crude mortality rates (CMRs) for all and specific causes during periods in and out of treatment. Unadjusted log-transformed RRs for each study were calculated by comparing CMRs for time during and out of treatment. Crude mortality rates for each period and RRs were pooled using random-effects meta-analyses with exact 95% CIs assuming a Poisson distribution. Heterogeneity was quantified using the I² statistic. Analyses were conducted in Stata, version 16.1,¹⁵ using the metan¹⁶ command for meta-analyses and the metareg¹⁷ command for meta-regressions (StataCorp). Significance was set at P < .05, and all P values were 2-sided.

Separate analyses of mortality risk were conducted for studies of patients who were incarcerated, observational studies of people released from prison with or without OAT, and studies that monitored time in or out of OAT after release from prison. Pooled analyses were conducted for all-cause and cause-specific mortality for the first 4 weeks during OAT and after OAT cessation and the remainder of time during and out of OAT. Crude mortality rates were calculated for each of the 4 periods, and RRs were calculated with the remainder of time in treatment as the reference period. We performed sensitivity analyses by excluding studies with participants who had no record of OAT during the study, studies at high risk of bias owing to missing data or loss to follow-up, and any studies of participants who were positive for HIV and released from prison.

Of the 7980 studies identified, 72 publications fulfilled inclusion criteria (eFigure 1 in the Supplement; a list of articles excluded during full-text review is in eTable 1 in the Supplement). A total of 15 RCTs including 3852 participants and 36 primary cohort studies including 749 634 participants were eligible for analysis. An additional 21 articles were not included analyses despite fulfilling the inclusion criteria due to participant overlap: 3 were included in subanalyses (C. Bharat, personal communication, 2021; eTable 6 in the Supplement for secondary publications).^15,18

Characteristics of RCTs are presented in eTables 4 and 6 to 13 in the Supplement. Eight of 15 RCTs (53%)^19-26 were conducted in North America and at single clinics (8 [53%]^19-26). Buprenorphine was studied in 7 of 15 of the RCTs (47%),^{19,22,23,27-30} and the most common comparator was detoxification alone (5 [33%]^{20,25,27,29,31}). Most RCTs (12 of 15 [80%]^{19-22,25-29,31-33}) commenced before 2010 and lasted 6 months or less (9 [60%]^{19,20,23-26,28-30}). Eight of the RCTs (53%)^{19,20,23,26,27,29-31} had greater than 20% of participants lost to follow-up.

In total, 45 deaths were reported across RCTs; 7 of 15 RCTs (47%) reported 0 deaths.^{19,20,23,25,28-30} There was no significant difference in all-cause mortality for patients allocated to OAT compared with comparison groups (RR, 0.86; 95% CI, 0.59-1.23) (eFigure 2 in the Supplement). Three of 15 RCTs (20%) evaluated the administration of OAT to patients who were incarcerated; no deaths were reported (eFigure 2 in the Supplement).^19,21,24

Characteristics of observational studies are presented in eTables 4 and 6 to13 in the Supplement. Cohorts ranged from 110 to 306 786 participants and included people from Europe (20 of 36 studies [58%]^34-53; 278 977 people), Australia (6 studies [17%]^54-59; 103 715 people), Asia (5 studies^60-64 [14%]; 311 658 people), and North America (5 studies^65-69 [14%]; 110 631 people). Seventeen studies (47%)^{34,35,39,41,42,45,47-50,57,58,61,64,66,68,69} concluded follow-up in 2010 or later. Sixteen studies (44%)^{34,36-39,41,42,45,50,51,54,57,58,66,68,69} used OAT registry data. Mortality data during and after OAT were stratified by methadone treatment for 28 of 36 studies (78%)^{37-40,42-46,48,49,52-58,60-67,69,70} and by buprenorphine treatment for 8 of 36 studies (22%).^{41,45,54,55,57-59,69} Type of OAT was unspecified in 6 of 36 studies (17%).^{35,36,47,50,51,68}

The pooled all-cause CMR while enrolled in any form of OAT was 11.00 deaths per 1000 PY (95% CI, 9.20-13.16) compared with 23.97 deaths per 1000 PY when not receiving OAT (95% CI, 19.92-28.83). Being in OAT was associated with more than half the risk of mortality (RR, 0.47; 95% CI, 0.42-0.53; 30 of 36 studies [83%]^{34-36,38,39,41-45,48-50,52-59,61-69}; 562 714 patients) (eFigure 3 in the Supplement). The association was not different for methadone (RR, 0.47; 95% CI, 0.41-0.54; 23 of 36 studies [64%]^{38-40,42-46,48,49,52-54,56,57,61-67,69}) or buprenorphine (RR, 0.34; 95% CI, 0.26-0.45; 8 of 36 studies [22%]^{41,45,54,55,57-59,69}) (Figure).

Table 1 shows mortality rates and RRs for time in and out of OAT according to participant characteristics, treatment administrators, study methodologies, and region. Although absolute mortality rates during and after OAT varied (Table 1), the RRs for time during and out of OAT were similar across subgroups of people with opioid dependence (eg, by sex, age group, HIV or hepatitis C virus [HCV] status, and taking drugs via injection). Study findings were consistent across study design, mortality ascertainment method, and region. Meta-regressions found no evidence for variation of RR by study year, sample size, follow-up time, age, proportion of women, proportion of people who inject drugs, or people with HIV or HCV (Table 1; eTable 14 in the Supplement).

Results of all-cause and cause-specific mortality analyses are presented in Table 2. While receiving OAT, people with opioid dependence were at lower risk of all injury and poisoning (pooled RR, 0.34; 95% CI, 0.27-0.42), suicide (pooled RR, 0.48; 95% CI, 0.37-0.61), cancer (pooled RR, 0.72; 95% CI, 0.54-0.98), and alcohol-related (pooled RR, 0.59; 95% CI, 0.49-0.72) and cardiovascular-related (pooled RR, 0.69; 95% CI, 0.60-0.79) mortality. The strongest association with lower mortality risk while receiving OAT was observed for deaths related to injury or poisoning, specifically, unintentional drug-related death (pooled RR, 0.41; 95% CI, 0.33-0.52) and suicide (pooled RR, 0.48; 95% CI, 0.37-0.61). Liver disease–related deaths in general were lower during OAT, but there was an association between OAT and viral hepatitis mortality (RR, 1.35; 95% CI, 1.15-1.60). Forest plots for each cause of death are presented in eFigures 4-24 in the Supplement.

Stratified OAT time interval and all-cause mortality data were available for 14^{38,39,41,42,45,49,50,54,58,61,62,65,69,70} cohort studies (39%; 175 213 people). The remainder of time spent in OAT (after the first 4 weeks of treatment) was the period of lowest mortality risk overall. In the first 4 weeks of OAT, all-cause mortality was almost double that in the remainder of time spent in treatment (RR, 1.92; 95% CI, 1.10-3.35). Compared with the remainder of time spent in OAT, all-cause mortality was 6 times higher in the first 4 weeks after OAT cessation (RR, 6.01; 95% CI, 4.32-8.36) and remained at double the rate for the remainder of time spent while not receiving OAT (RR, 1.81; 95% CI, 1.50-2.18) (Figure; eFigure 26 in the Supplement). A similar pattern emerged for drug-induced and unintentional opioid-related deaths.

Acute causes of mortality that may be associated with OAT are listed in eTable 15 and eFigure 25 in the Supplement; these causes were stratified by times in and out of OAT (forest plots are presented in eFigures 27-33 in the Supplement; data on other causes of death stratified by OAT time period are presented in eTable 16 in the Supplement; and analyses using first 2-week time periods are presented in eTable 17 in the Supplement). Cause-specific mortality rates were lowest after the first 4 weeks in OAT (ie, remainder in OAT; eTable 15 in the Supplement). For pooled RRs, rate of mortality during the remainder of time receiving OAT was used as the referent. Because of insufficient mortality data (eg, more than half the included studies reported 0 deaths for a specific cause), RRs were not reported for less common causes of death.

Table 3 presents results of analyses during and after incarceration for people receiving OAT compared to those not receiving OAT. Only 1 study¹⁸ that examined the OAT during incarceration was identified, with OAT associated with reduced all-cause (RR, 0.06; 95% CI, 0.01-0.46), drug-related, and suicide-related (RR, 0.10; 95% CI, 0.01-0.75) mortality during the first 4 weeks of treatment and the entire time spent in incarceration.

Three studies^47,60,71 reported mortality after release from prison for people who initiated OAT during incarceration, compared with those who had opioid dependence and were incarcerated but had not received OAT. All-cause (RR, 0.24; 95% CI, 0.15-0.37) and drug-related (RR, 0.19; 95% CI, 0.10-0.37) mortality was lower during the first 4 weeks after release from prison for people who left incarceration still receiving OAT compared with those who were released but were not receiving OAT.

Two studies^62,71 followed all-cause mortality rates while receiving and not receiving OAT after release from incarceration. Time spent in OAT after release was associated with lower all-cause (RR, 0.09; 95% CI, 0.09-0.56) and drug-related mortality (RR, 0.18; 95% CI, 0.10-0.31) in 1 study⁷¹ (eFigures 34-36 and eTable 18 in the Supplement).

Twelve RCTs (80%) were assessed as having high risk of bias largely due to missing outcome data and measurement of outcome (eFigures 44 and 45 in the Supplement).^{20,22,23,25-33} Assessments of observational studies using the ROBINS-I tool included in all-cause mortality analyses found that most studies were at moderate risk of bias (26 of 30 [87%]).^{34-45,49,51,54,56-59,61,63,66-70} Some studies were at serious risk of bias (4 [13%]),^50,55,64,65 and 1 study (3%) was at critical risk of bias owing to measurement of mortality data and participant loss to follow-up (eFigures 37 and 38 in the Supplement).⁴⁸ The sole study of OAT for patients who were incarcerated was at low risk of bias (eFigure 39 in the Supplement).¹⁸ The other studies with follow-up time during or out of OAT after incarceration were at moderate risk of bias (eFigures 40-43 in the Supplement).^47,60,62,71

Nineteen observational studies conducted multivariable analyses adjusting for a range of potential confounders on all-cause (and in some cases overdose-related) mortality (Table 4).^{18,34,38-42,45,47,51,54,60,62-64,66,68-70} There was minimal change in the RRs after controlling for confounding, regardless of the variables included (eTable 19 in the Supplement).

Sensitivity analyses are reported in eFigures 46-49 in the Supplement. None of the analyses yielded results that changed the direction or size of associations reported in the main findings.

To our knowledge, this systematic review and meta-analysis is the first to document the association of OAT across different settings with both all-cause and cause-specific mortality. We synthesized 36 observational cohort studies that assessed mortality risk during and out of OAT, which represented a 3-fold increase from a previous review of all-cause mortality⁷ that included 19 cohorts. Our findings suggest a potential public health benefit of OAT, which was associated with a greater than 50% lower risk of all-cause mortality, drug-related deaths, and suicide and was associated with significantly lower rates of mortality for other causes. The association was consistent across a range of participant and study characteristics.

Our results suggest that RCTs of OAT were underpowered to examine mortality risk. There was no significant association between OAT and mortality risk in the pooled community RCTs; in 7 studies,^{19,20,23,25,28-30} no deaths were reported. None of the 3 RCTs conducted on the effect of OAT during or after incarceration could be used to guide inferences.^19,21,24

Several examples of discrepant findings and conclusions on the benefits of treatment between RCTs and observational studies, such as hormone replacement treatment on cardiovascular outcomes⁷² and vitamin, mineral, or fish oil supplements on noncommunicable disease,⁷³ exist. Discrepancies between findings from RCTs and observational studies and mortality outcomes described in the current review are not the same because (1) associations between OAT and all-cause mortality (as well as overdose and suicide) were strong, and (2) there was little difference between adjusted and unadjusted estimates; therefore, it was unlikely that unmeasured confounding would alter the findings.

People with opioid dependence were at substantially lower risk of suicide, cancer, drug-related, alcohol-related, and cardiovascular-related mortality during OAT compared with time while not receiving OAT. The association between OAT and a lower risk of overdose was identified in previous reviews.⁷ The reductions in alcohol-related, cardiovascular-related, and cancer-related mortality may be due to reductions in alcohol use during OAT; similarly, reduced stimulant use may be associated with reduced cardiovascular mortality. The reductions in cardiovascular and cancer-related mortality may also reflect greater access to screening, early intervention, and treatment as a result of improved engagement with treatment administrators.

By contrast, viral hepatitis mortality was higher among those who received OAT. This may reflect an issue of competing risks⁷⁴ (with people retained in OAT living longer and therefore at higher risk of HCV infection). However, we believe this may have been due to misclassification. As we show in eFigure 49 in the Supplement, 8 cohorts^{45,49,50,54,57,58,69,70} reported both overall liver-related and viral hepatitis mortality during and out of OAT. There was an elevated risk of mortality due to viral hepatitis, suggesting that OAT status may have affected coronial decisions around attribution of the liver-related cause of death. Until very recently, there were no highly effective treatments for HCV, and coverage availability of HCV treatment has been historically low. There may be an opportunity to further reduce mortality through increasing coverage of HCV treatment as 1 strategy to reduce viral hepatitis mortality in this population.

Despite a hypothesized relationship between OAT and mortality risk due to injection-related injuries and diseases, such as bacterial infections, no such relationship was identified. This may reflect the way in which we operationalized exposure to OAT as overall exposure and according to specific periods of time, but not specifically measured as long-term and stable retention in OAT. The potential for OAT to have an influence on more chronic consequences of dependent opioid use would not be expected except insofar as OAT was being delivered at a certain level of quality and intensity. Although we extracted information on the context of OAT provision where reported, there was often very limited information on the manner of OAT delivery and the other services available to, and received by, people in the cohorts. It is likely that additional interventions will be required to reduce risk of mortality due to injection-related injuries and diseases.

There is a need for more detailed investigation and intervention development to minimize mortality risk during induction of OAT and to maintain patients on OAT who are in critical need of such therapy. A clinical decision support system, stratifying clients’ risk of dropout in real time, may facilitate the identification of those in need of service enhancements to increase engagement and prevent dropout. This will be challenging given limited evidence to support the influence of a number of factors on retention, including age, substance use, OAT dose, legal issues, and attitudes toward OAT.⁷⁵ There is evidence that people with multiple treatment episodes continue receiving OAT for progressively longer periods in later treatment episodes.^39,76 Findings from a Scottish cohort suggest that survival benefits increase with cumulative exposure to OAT.⁷⁷

Evidence from previous studies suggests a strong association between OAT and lower risk of mortality when incarcerated, when released from incarceration while receiving OAT, and when receiving OAT after release from incarceration, with particularly lower risk of suicide and overdose mortality. Opioid agonist treatment was also found to be cost-saving as an intervention to reduce the risk of mortality after release from incarceration.⁷⁸ Guidelines have been developed in the UK recommending and informing the use of OAT in carceral settings and after release⁷⁹; international guidelines are also needed.

Future RCTs that withhold OAT are no longer ethical, and it is unlikely that future trials of OAT will be large enough to detect differences in mortality outcomes. The evidence base can be transformed and improved in 2 ways: (1) investing in research that tests how alternative ways of delivering OAT (and different medications) and adjunct interventions can improve retention, minimize elevated risk of mortality at initiation and maximize cost-effectiveness and (2) investing in data linkage studies that can create better-emulated trials and natural experiments of OAT delivery and combinations of interventions on mortality risk.

This study had some limitations. The current evidence base in general was unrefined, lacking detail on clinical characteristics of patient history, intervention delivery, and consistent measures of confounders. For example, most cohorts did not specify whether treatment ceased because of dropout or completion. Those that did specify suggested that most patients dropped out of treatment as opposed to having a planned completion date.^76,80 Morbidity was strongly associated with mortality risk and may also have been associated with retention.

There have been divergent findings according to which comorbidities are considered. One study found retention was higher among people with greater comorbidity (measured as the number of chronic diseases)⁷⁶; another study suggested no association of HIV or HCV status with retention⁸¹; and an Australian study suggested that depression and other substance use disorders were associated with increased retention, whereas psychosis was associated with reduced retention (C. Bharat, personal communication, 2021). Moreover, cohort studies that have adjusted for comorbidity did not find changes in the estimated mortality risk by time during and out of OAT.

The evidence depends on observational studies that were subject to bias, confounding, and selection of participants. In most cohorts, there was little bias in measuring the outcome, but there were problems with missing data in some studies. However, we did find consistent data and little difference between adjusted and unadjusted estimates (Table 4). The evidence base could be substantially strengthened if future studies linked more detailed clinical information with mortality records so that more refined adjustments could be made for patterns and severity of opioid dependence, comorbidity, and environmental hazards on OAT retention and mortality risk.

Studies were largely limited to those who had received OAT for opioid dependence at some point. It is possible that mortality risk is higher for those who never receive OAT than it is for out-of-OAT periods among those who have some OAT experience, leading to an underestimate of the association between OAT and risk of mortality.

Despite the large numbers of people in the identified cohorts, for some causes of death, we had limited power, particularly when evaluating association during specific periods in and out of OAT. Similarly, there were insufficient studies to examine potential differences between buprenorphine and methadone by specific periods in and out of treatment.

Although associations were seen in this review, there have been too few studies of the influence of OAT during incarceration and after release. More studies using linked data that follow people through OAT during incarceration and after release, as conducted previously,^18,71 are needed.

The results of this systematic review meta-analysis suggest that OAT is an important intervention for people with opioid dependence, with the capacity to reduce multiple causes of death. Despite this positive association, access to OAT remains limited in many settings, and in the US and globally, coverage for this type of treatment is low. Future work to increase access could have important population-level benefits.

Accepted for Publication: April 1, 2021.

Published Online: June 2, 2021. doi:10.1001/jamapsychiatry.2021.0976

Correction: This article was corrected on June 30, 2021, to fix an error in the Abstract; on March 16, 2022, to fix an incorrect rate ratio reported in the Abstract; and on July 19, 2023, to fix errors in the Results and Tables.

Corresponding Author: Thomas Santo Jr, MPH, National Drug and Alcohol Research Centre, UNSW Sydney, 22-32 King St, Randwick, NSW 2031, Australia (t.santo@unsw.edu.au).

Author Contributions: Mr Santo had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Santo, Clark, Hickman, Grebely, Tran, Farrell, Degenhardt.

Acquisition, analysis, or interpretation of data: All authors.

Drafting of the manuscript: Santo, Clark, Hickman, Chen, Farrell, Degenhardt.

Critical revision of the manuscript for important intellectual content: Santo, Clark, Hickman, Grebely, Campbell, Sordo, Tran, Bharat, Padmanathan, Cousins, Dupouy, Kelty, Muga, Nosyk, Min, Pavarin, Farrell, Degenhardt.

Statistical analysis: Santo, Clark, Grebely, Bharat, Cousins, Dupouy, Kelty, Nosyk, Min.

Obtained funding: Degenhardt.

Administrative, technical, or material support: Clark, Hickman, Sordo, Tran, Bharat, Nosyk, Degenhardt.

Supervision: Santo, Hickman, Campbell, Farrell, Degenhardt.

Other material support: Pavarin, Farrell.

Conflict of Interest Disclosures: Mr Santo reported receiving the Australian Government Research Training Program Fee Offset scholarship and Australian Federal Government Department of Health Grants National Centre Core Funding during the conduct of the study. Dr Hickman reported receiving grants from National Institute for Health Research & Medical Research Council for analysis of the data set included in this review during the conduct of the study and speaker honoraria from Merck Sharp & Dohme and Gilead in the past 3 years outside the submitted work. Dr Grebely reported receiving grants from AbbVie, Cepheid, Gilead Sciences, Hologic, Indivior, and Merck, and personal fees from AbbVie, Cepheid, Gilead Sciences, and Merck outside the submitted work. Dr Campbell reported receiving grants from Indivior and National Health and Medical Research Council Early Career Fellowship during the conduct of the study. Dr Bharat reported receiving the National Drug and Alcohol Research Centre and University of New South Wales Scientia PhD Scholarships outside the submitted work. Dr Dupouy reported being a member of a working group for and writing a recommendation on the proper use of prescribed opioid analgesics for the French High Authority of Health. Dr Farrell reported receiving grants from the Australian Federal Government Department of Health National Centre Core Funding, an united grant from Indivior to evaluate new opioid medications in Australia, and grants from Seqirus United to evaluate new opioid medications in Australia outside the submitted work. Dr Degenhardt reported receiving grants from National Health and Medical Research Council Fellowship, project funding and grants from the National Institutes of Health Project funding, grants from Indivior United to evaluate new opioid medications in Australia, and grants from Seqirus United to evaluate new opioid medications in Australia outside the submitted work. No other disclosures were reported.

Funding/Support: This work was supported in part by an Australian National Drug and Alcohol Research Centre scholarship (Mr Santo) and a University of New South Wales Sydney PhD scholarship (Mr Santo). The National Drug and Alcohol Research Centre is supported by funding from the Australian Government Department of Health under the Drug and Alcohol Program. This work was supported in part by grant APP1135991 from National Health and Medical Research Council Senior Principal Research Fellowship and grant R01DA1104470 from a National Institutes of Health National Institute on Drug Abuse grant (Dr Degenhardt); grant RFA-AI-18-026 from National Institutes of Health NIAID and grant APP1150078 from the Advancing the Health of People Who Use Drugs: Hepatitis C and Drug Dependence Program (Dr Degenhardt); support from National Institutes of Health Project Health Protection Research Unit in Behavioural Science and Evaluation at University of Bristol (Dr Hickman); support from the National Drug and Alcohol Research Centre and University of New South Wales Scientia PhD Scholarships (Ms Bharat); support from an National Health and Medical Research Council Emerging Leadership grant (Dr Kelty); support from National Health and Medical Research Council Investigator Award and the Kirby Institute, and grant 1176131 from National Health and Medical Research Council Investigator Grant (Dr Grebely); from grant AES PI19/00982 from the Acción Estratégica en Salud del Gobierno de España AES PI19/00982 (Dr Sordo); from the Michael Smith Foundation for Health Research Scholar (Dr Nosyk); from the Health Canada Substance Use and Addictions Program (Ms Min); by grant 1119992 from the National Health and Medical Research Council Early Career Fellowship and grant HRAPHR-2015-1088 from the Health Research Board, Ireland (Dr Campbell); by grant MR/N00616X/1 from Medical Research Council Addiction Research Clinical Training Programme (Ms Padmanathan); and by grants RD17/0017/0003 and PI20/0883 from the Ministry of Science and Innovation, Institute of Health Carlos III, Spain (Dr Sordo).

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The views expressed in this publication do not necessarily represent the position of the Australian government.

Additional Contributions: We thank (Kevin) Kun-Chia Chang, PhD, Charles Huang, PhD, Anders Ledberg, PhD, and Olga Morozova, PhD, for providing additional, unpublished data to inform our review. We thank Steven Batki, MD, Dana Bernson, MPH, John Devlin, PharmD, Javier Esteban, PhD, Elizabeth Evans, PhD, Linn Gjersing, PhD, Valerie Gruber, PhD, Sharon Hutchinson, PhD, Simeon Kimmel, MD, Marina Klein, MD, Barrott Lambdin, PhD, Barbara Lovrecic, MD, David Marsh, PhD, Michelle McKenzie, MPH, Tim Millar, PhD, Timothy Nielsen, MPH, Matthias Pierce, PhD, Stefanie Rezansoff, PhD, Josiah Rich, MD, Elana Rosenthal, MD, Angela Russolillo, MSc, Davida Schiff, MD, Andrew Stone, MD, Eric Strain, MD, Kinna Thakarar, DO, Helge Waal, MD, and George Woody, MD, for providing clarification of the data available for their cohorts. We thank Annabeth Simpson, MPH, for assistance through the screening and extraction stages. We thank Hayley Jones, PhD, for assistance with providing additional data for our review. No financial compensation was received for these contributions. Finally, we would like to thank Annick Bórquez, BSc, MSc, PhD, from University of California San Diego for identifying the errors in Table 3, which have been corrected.