Introduction

Globally, the use of opioids, including the use of heroin, opium and the non-medical use of pharmaceutical opioids, has stabilized at high levels, affecting 0.7% of the world’s adult population aged 15–64 [1]. With an estimated 33 million affected individuals in the world in 2014, the use of opiates and prescription opioids continues to produce severe health and social problems [2]. Opioid use disorder (OUD) is associated with high risk of premature death. Worldwide, illicit opioid dependence was the largest contributor to the direct burden of DALYs (disability-adjusted life years, 9.2 million; 95% confidence interval (CI): 7.1–11.4), accounting for 45.8% among the 20 million DALYs caused by illicit drugs [3]. In addition, the epidemic of prescription opioids overdose in recent years has underscored the importance of addressing OUD as a worldwide public health challenge [4]. The estimated all-cause mortality rate among those with regular or dependent heroin and other opioids use was 2.09 per 100 person-years (PY), which was almost 15 times that of the general population [5]. Specifically death from drug overdose is the primary cause of drug-related deaths worldwide and opioids are the main drug type implicated in these deaths [1]. In Europe, opioids were found in 82% of fatal overdoses, of which heroin or its metabolites was detected in the majority of overdose deaths; other opioids including methadone, buprenorphine, fentanyl, and tramadol were also regularly found in toxicological reports [6].

Medication-assisted treatment (MAT) including the agonists, methadone and buprenorphine, and opioid antagonist, naltrexone is the primary evidence-based treatment for OUD [7]. Opioid agonist therapy with methadone and buprenorphine, mainly known as opioid substitution treatment (OST), is the main form of care for OUD, which has showed effectiveness in reducing both illicit opioids and prescription opioids use [8, 9], and risk of HIV infection [10], as well as improving physical and mental health [11]. Recently, extended-release naltrexone also represented a promising treatment to maintain abstinence, reduce relapse, and retain patients in treatment, which was crucial for reducing the risk of death from overdose [12, 13]. Studies have shown that MAT is effective in reducing mortality in patients with OUD [14, 15]. Research on mortality estimates has been carried out in those with OUDs [5, 16], those who inject drugs [17], and alcohol use disorder patients [18].

Mortality risk varies before entering, during and after MAT, and also varies with different medicine types among individual with OUDS [19, 20]. A recent meta-analysis estimated mortality rate of periods during and after OST, and found that receiving methadone and buprenorphine treatment was associated with substantial reductions in the mortality of those with OUDs [21]. The risk of death of individuals with OUDs reduced largely after entry into MAT [22], but the pooled level of net mortality reduction due to MAT was still uncertain compared to untreated individuals with OUDs. Another previous meta-analysis also failed to determine the risk of death in untreated individuals with OUDs, as it combined those discharged from treatment with untreated individuals as a comparison group [5]. A clear distinction between the discharged group and the untreated group could help for assessing the effect of MAT on mortality more objectively and exactly among those with OUDs. Previous meta-analysis mainly focused on the opioid agonist treatment with methadone and buprenorphine, and no estimate of the overall mortality rate related to naltrexone antagonist treatment has been made to date.

Other characteristics such as retention time, medication dose, sex, and geographical region are also important factors that influence the estimates of mortality rate and risk. In previous systematic reviews, there were no comprehensive assessments of the effects of different MAT characteristics on mortality rates and risks among those with OUDs [5, 16, 23]. In recent years, more large-scale longitudinal cohort studies and cumulative evidence examined the association of MAT with mortality by different characteristics of treatment (including different treatment status, medication type, retention duration, periods of treatment transition, and drug dose) [14, 15, 19, 20, 24, 25]. This made it possible to quantitatively evaluate the specific mortality rate and risk in those with OUDs based on different MAT-related characteristics. These data are essential for understanding the scope of problems related to OUDs as well as the impact of MAT as a strategy to combat opioid-related problems. Hence, the present study reviewed and quantitatively integrated existing data on mortality associated with MAT characteristics from longitudinal cohort studies by meta-analysis.

Methods

Data sources and search strategy

We performed a systematic review and meta-analysis in accordance with Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [26]. We searched PubMed and Embase (through the Ovid platform) from inception through 15 August 2017 using both MESH terms and text words with a limitation of English language. Briefly, our search strings included terms relating to substitution, maintenance, methadone, buprenorphine, naltrexone, mortality, and cohort studies. The search was restricted to human studies. In addition, full texts of systematic reviews on the topic of MAT for OUDs under study during the titles and abstracts screening stage were read to identify all potentially relevant studies. All studies included in the earlier reviews [5, 16, 21, 23] were assessed for inclusion in the present study. Additionally, reference lists of included studies were screened to detect complementary articles which met the inclusion criteria. The detailed search strategy for PubMed was shown in eAppendix 1 in the Supplement.

Eligibility criteria

All studies were eligible in this meta-analysis for the following inclusion criteria: (a) study design: longitudinal cohort study; (b) participants: patients with OUD; (c) treatment: MAT involving either methadone, LAAM (levo-alpha-acetylmethadol), buprenorphine, naltrexone, or heroin; (d) comparison group: those with OUDs either discharged from MAT or not treated in MAT; (e) outcome measures: all-cause or overdose mortality rate (crude mortality rate [CMR] and standardized mortality ratio [SMR]), mortality rate ratio (RR), or hazard ratio (HR); (f) duration of follow-up: at least one year. We included studies that directly examined the effects of MAT on mortality among those with OUDs. Studies that assessed the mortality risk of different medication types or dosage levels were also included.

We excluded cross-sectional or serial cross-sectional studies and comparative analysis of autopsy records or medical records to identify death relating to MAT, as well as studies in prison and recently-release settings. Letters, reviews and comments were also excluded. We excluded studies which did not specify the drug of use and studies focusing on only HIV-positive individuals [27, 28]. Studies regarding other types of treatment but not MAT (e.g., detoxification treatment, therapeutic community, or psychological treatment alone) or unknown treatment modality were not included. We also excluded studies if mortality rate was not calculated by dividing the number of deaths by the total person-time in the corresponding treatment status. Studies with no comparison group or reporting mortality not separated by treatment status were also excluded. For duplicate studies from the same cohort, studies with detailed mortality data by treatment status or longer follow-up duration were included.

Study selection and data extraction

After removal of duplicates, titles and abstracts of retrieved articles were screened and clearly irrelevant articles were excluded. Full texts of the remaining studies were read to determine whether they met eligibility criteria for inclusion in the meta-analysis. Data were extracted from the eligible articles by two independent authors. The extracted data included study characteristics (country, study design, follow-up period); sample characteristics (sample source, sample size, age, and percent male); comparison modalities (treatment status, different periods in or out of treatment, medication type, treatment duration, or different dosage levels); and mortality estimates (such as all-cause CMRs and SMRs, overdose CMRs, RRs and HRs). Discrepancies were resolved by discussion or through consultation with a third author. For studies providing separate mortality rate and risk estimates by different medicines without a summary estimate [20, 29, 30], data were extracted separately and divided into different medication groups.

Quality assessment

The methodological quality of the studies was independently assessed using the nine-star Newcastle–Ottawa scale (NOS) [31] by two authors. Each study was evaluated based on eight items with a total score of 9, containing some perspectives about sample selection, comparability of cohorts, and outcome assessment for cohort studies. Studies with a score of 7 or greater were defined as high quality [32].

Statistical analysis

Meta-analyses were performed to generate pooled estimates of CMRs for corresponding treatment status across studies, and SMRs, if data permitted. The CMR was expressed as the number of deaths per 100 PY. In some cases, mortality rates were not reported but could be estimated using reported raw data with the method described by Degenhardt [5]. CMRs were both pooled for all-cause and overdose death among opioids users by three treatment-related status (i.e. while receiving MAT, after medication treatment cessation, and not treated in MAT), as well as medication type (agonist [methadone or buprenorphine], antagonist [naltrexone], and unspecified).

We also performed meta-analyses to compare the relative risks (RRs) of mortality based on treatment status and medication type. Within studies, RRs were calculated by the following comparative groups: untreated versus in MAT, discharged from MAT versus in MAT. When the number of deaths within a subgroup was zero, we dealt with this problem by setting the number of deaths at 0.5 to allow inclusion in comparative analyses [33]. All statistical analyses were performed by using STATA, version 12.1. Heterogeneity between and within studies was estimated using the Cochran Q-statistic and the I2 statistic. Random effects models were used for all analyses in current study due to the expected large between-study heterogeneity.

To explore heterogeneity, cohorts were divided into subgroups according to sex (male or female), geographical region (Europe, North America, Australasia, or Asia), follow-up duration (<8 years or ≥8 years), study design (data linkage or follow-up investigation), cohort size (<2500 or ≥2500), and year of publication (before 2010 or in the year of 2010 and later). Differences by these characteristics were estimated using meta-regression analysis. We also performed meta-analyses based on different periods on or off MAT (first two weeks following treatment initiation, maintenance period during treatment, first 2 weeks after terminating MAT, or remaining period after terminating MAT), treatment duration (≤1 year or >1 year) and methadone dose (<60 mg/day or ≥60 mg/day). Finally, the potential for publication bias was examined using funnel plot and Begg’s test [34]. Sensitivity analyses were conducted to examine the influence of individual study on pooled estimates, as well as the impact of quality of the included studies.

Results

Studies included

The flow chart summarizing the process of study selection was shown in Fig. 1. Among 2993 articles, 30 studies involving 370,611 participants (1,378,815 PY) were eligible for the meta-analysis. Twenty-four studies (165,732 participants, 883,821.6 PY) reported all-cause mortality rate or risk ratio pertaining to different treatment status [14, 15, 19, 20, 29, 30, 35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52], of which three studies only provided mortality risk ratio [45, 47, 52]. Sixteen studies involving 272,210 participants (1,183,985 PY) reported overdose mortality rate or risk ratio [14, 20, 35,36,37,38, 41,42,43,44, 46, 50, 51, 53,54,55].

Fig. 1
figure 1

Flow chart of identification of eligible studies

Full size image

Table 1 showed the characteristics of the included studies. Twenty-one studies provided all-cause mortality rate during in treatment [14, 15, 19, 20, 29, 30, 35,36,37,38,39,40,41,42,43,44, 46, 48,49,50,51]; eighteen studies provided all-cause mortality rate after terminating MAT [14, 15, 19, 20, 29, 30, 35,36,37, 39,40,41, 43, 44, 48,49,50,51]; and eight studies provided all-cause mortality rate for an untreated period [36, 38, 40,41,42, 46, 48, 50]. Twenty-one studies compared all-cause mortality risk during in treatment with that after terminating MAT [14, 15, 19, 20, 29, 30, 35,36,37, 39,40,41, 43,44,45, 47,48,49,50,51,52], and nine compared all-cause mortality risk while receiving MAT with untreated period [36, 38, 40,41,42, 46, 48, 50, 52]. For patients treated in MAT, methadone was the main drug of maintenance treatment (n = 21), with other medicines including buprenorphine (n = 3) [20, 30, 47], naltrexone (n = 2) [29, 30], and unspecified drugs (n = 7) [19, 36, 37, 42, 46, 51, 55]. Of the seven studies with unspecified drug types, six studies combined patients treated with methadone or buprenorphine as a whole [19, 36, 37, 42, 46, 51], and one study did not report the type of medicine prescribed for MAT [55]. Studies were mainly from Europe (n = 15) and Australasia (n = 6), mostly using linked registers (n = 22) to identify treatment-related information and mortality data. The population of the 30 studies consisted of opioid-dependent individuals (n = 19) [14, 15, 19, 20, 24, 25, 29, 30, 36, 37, 39, 43, 44, 46, 49, 51, 53, 55, 56], heroin-dependent individuals (n = 10) [35, 38, 40,41,42, 45, 48, 50, 52, 54], and outpatients with opioid use disorder (n = 1) [47]. Of the 30 studies, 4 studies reported that the participants were diagnosed according to the DSM-IV criteria for dependence [25, 42, 43, 45]. Follow-up duration ranged from 2 to 20 years, with a mean of 8.7 years. Newcastle–Ottawa scale for quality assessment showed that the overall score ranged from 4 to 9 and there were 19 studies with a score of 7 or more.

Table 1 Characteristics of included studies
Full size table

Estimates of mortality rate and risk in different treatment status

Table 2 showed the all-cause and overdose mortality rates. Random effects models suggested pooled all-cause CMRs of individuals with OUDs while receiving MAT, after terminating MAT and for an untreated period were 0.92 per 100 PY (95% CI: 0.79–1.04; Fig. 2a), 1.69 (1.47–1.91; Fig. 2b), and 4.89 (3.54–6.23; Fig. 2c), respectively. The pooled overdose mortality rates were 0.24 (0.20–0.28) during in-treatment period, 0.68 (0.55–0.80) after treatment discharge, and 2.43 (1.72–3.15) for an untreated period (Table 2; Supplementary Figure S1). Individuals with OUDs had higher risks of all-cause death (2.33, 95% CI: 2.02–2.67; Fig. 3a1) and overdose death (3.09, 2.37–4.01; Fig. 3a2) after terminating MAT than that while receiving MAT. Meanwhile, compared with patients receiving MAT, both higher all-cause mortality risk (RR 2.56, 1.72–3.80; Fig. 3b1) and overdose mortality risk (8.10, 4.48–14.66; Fig. 3b2) were presented in untreated individuals with OUDs.

Table 2 Pooled estimates and subgroup analyses of crude all-cause mortality rates (CMRs)
Full size table
Fig. 2
figure 2

Pooled estimates of crude all-cause mortality rates and risks by different medicines. (a) Medicine-specific and overall all-cause CMR of individuals with OUD while receiving MAT. (b) Medicine-specific and overall all-cause CMR of individuals with OUD after exiting from MAT. (c) Overall all-cause CMR of individuals with OUD for an untreated period. (d) Risk of all-cause death after exiting from MAT of individuals with OUD treated with buprenorphine compared with those treated with methadone

Full size image
Fig. 3
figure 3

Pooled estimates of all-cause and overdose mortality rate ratios (RRs) by treatment status. (a1) Risk of all-cause death after exiting from MAT compared with that while receiving MAT in individuals with OUD. (a2) Risk of overdose death after exiting from MAT compared with that while receiving MAT in individuals with OUD. (b1) Risk of all-cause death in untreated individuals with OUD compared with patients treated in MAT. (b2) Risk of overdose death in untreated individuals with OUD compared with patients treated in MAT

Full size image

Estimates of mortality rate and risk by different medicines

As can be seen in Table 2, the all-cause CMRs during and after agonist therapy were 0.93 (0.76–1.10), and 1.79 (1.47–2.10), respectively. The all-cause CMRs while in active antagonist treatment and after discontinuation of antagonist treatment were 0.26 (0–0.59), and 1.97 (0–5.18), respectively. Figure 2 showed the all-cause mortality rate stratified by different medicines. During the periods of receiving MAT, all-cause mortality rate for naltrexone treatment was the lowest (0.26, 0–0.59), followed by buprenorphine treatment (0.38, 0.31–0.46), and methadone treatment (1.05, 0.86–1.25; Fig. 2a). After terminating MAT, all-cause mortality rates for patients exiting from methadone treatment and naltrexone treatment were 2.03 (1.67–2.39) and 1.97 (0–5.18), respectively. Compared to these two medicines, patients exiting from buprenorphine treatment had a lower all-cause mortality rate (0.80, 0.38–1.22; Fig. 2b). Further analysis showed that buprenorphine groups had lower mortality risk than methadone groups after treatment discharge (RR 0.81, 0.70–0.93; Fig. 2d).

Effects of retention time, treatment transition, and drug dose on mortality

Longer retention duration was associated with a lower all-cause mortality rate (>1 year vs. ≤1 year: 1.62 per 100 PY, 1.31–1.93 vs. 5.31, −0.09–10.71; Fig. 4a). The all-cause mortality rate also varied during different stages on or off treatment. During periods of MAT initiation, the all-cause mortality rate decreased from 4.89 (3.54–6.23; Fig. 2c) in untreated periods to 1.44 (0.20–2.69; Fig. 4b) in the first two weeks following treatment onset. Then the all-cause mortality rate decreased continuously and the lowest all-cause mortality rate was observed in MAT maintenance period (0.57, 0.52–0.62). Subsequently, the mortality rate increased sharply in the first two weeks after discharging from medication treatment (4.71, 1.69–7.73), and then gradually decreased at the remaining period after exiting from MAT (1.28, 1.03–1.53; Fig. 4b).

Fig. 4
figure 4

Pooled estimates of all-cause mortality rate by retention-related characteristics. (a) All-cause CMR of individuals with OUD treated in MAT ≤ 1 year and those treated in MAT > 1 year, respectively. (b) All-cause CMR of individuals with OUD in the first 2 weeks following MAT initiation, during MAT maintenance period, in the first 2 weeks and subsequent period after exiting from MAT, respectively

Full size image

The pooled all-cause CMR of low-methadone dosage group (1.66, 0.78–2.53) was higher than that of high-dosage group (1.14, 0.72–1.57; Supplementary Figure S2A); however, the mortality risk was not statistically significant between different methadone dose groups (RR 1.02, 0.71–1.46; Supplementary Figure S2B).

Pooled estimates of mortality rates stratified by other characteristics

Subgroup analyses and meta-regression were conducted to explore differences in study-level characteristics. Males had higher CMRs than females while receiving MAT (1.46 [1.25–1.67] vs. 1.28 [1.00–1.57]), after exiting from MAT (3.92 [3.40–4.44] vs. 3.23 [1.52–4.94]), and for an untreated period (7.55 [1.39–13.72] vs. 4.20 [1.37–7.03]; Table 2). The univariable meta-regression also showed that all-cause mortality rates were significantly different by geographic region, cohort size and year of publication during in medication treatment and after medication treatment cessation. The pooled all-cause CMRs both during and after treatment were lower in studies from Australasia, studies published after 2010 and studies with larger cohort size than the corresponding comparison groups. We found no evidence that the mortality varied by follow-up duration both for treated and discharged groups (Table 2). Full meta-regression models were not conducted because of the limited studies.

Sensitivity analyses and publication bias

From the results of the leave-one-out sensitivity analysis, all the results above in our main analyses were not materially altered. We found no evidence that the results were statistically different stratified by study quality. The funnel plot suggested potential publication bias for mortality estimates. No evidence of publication bias for the analyses of risk estimates was observed based on funnel plots and Begg’s tests (p > 0.1; Supplementary Figure S3).

Discussion

The present meta-analysis investigated the mortality rates pertaining to three treatment statuses (0.92 per 100 PY during in MAT, 1.69 after cessation and 4.89, for an untreated period) in those with OUDs. In this analysis, we had extended our estimates of mortality rate and risk to untreated individuals with OUDs. The risk of all-cause death and overdose death among untreated patients with OUDs were 2.56 times and 8.10 times that of patients receiving MAT, respectively. The findings verified that longer retention in MAT was associated with a reduction in mortality rate and that discontinuation of MAT increased the risk of death, especially in the first 2 weeks after medication treatment discharge. In addition, this meta-analysis was the first study to assess the mortality rate and risk of three medicines delivered in MAT, agonist (methadone or buprenorphine), and antagonist (naltrexone). The mortality rate was lower while receiving naltrexone treatment than that during buprenorphine and methadone treatment. However, there is still a substantial need for research on the mortality of naltrexone treatment given the limited number of studies, particularly naltrexone delivered in the form of implants and extended-release.

The protective effect of MAT on mortality

These analyses provided pooled evidence that those with OUDs not entering into MAT had the highest level of mortality, especially the risk of overdose death, which was 8.10 times of patients in medication treatment. The present study indicated a markedly net mortality reduction associated with MAT compared to untreated individuals with OUDs. The lowest all-cause and overdose mortality rates were found while receiving MAT, indicating a protective effect of MAT on mortality. Researches showed that MAT was associated with decrease in drug injection [57] and HIV infection [10], which may contribute to the reduction of mortality. Medication treatment cessation was a major risk factor for both all-cause and overdose mortality, and the risk of all-cause death and fatal overdose after terminating MAT were 2.33 times and 3.09 times higher than that during in MAT. In contrast, longer retention time was associated with a reduction of mortality rate. These analyses were the first to provide integrated pooled data demonstrating that increased retention of treatment was beneficial for survival during MAT, and we found a lower mortality rate in patients who retained in medication treatment more than 1 year. Studies also showed that the hazard of death fell 5–13% for each additional year of treatment [56, 58].

Mortality risk was higher during the first two weeks following treatment onset than that during the remaining maintenance period. The reason for this may be the use of inappropriate starting dosage prescribed during induction into treatment, without taking into account the patient’s tolerance to opioids [59]. A prior study found that patients who enrolled in OST had higher risk of fatal overdose in the initial weeks of treatment [53]. Similarly, the mortality rate was especially high during the first two weeks after discontinuation of MAT, which may contributed to the high level of mortality rate following medication treatment cessation. The findings suggested that maintaining retention in treatment and extending treatment duration could be effective in reducing deaths. In fact, a considerable proportion of patients in this meta-analysis comprised multiple episodes of medication treatment initiation and cessation [14, 19, 37]. This indicated that MAT clients frequently transitioned into and out of treatment, which increased exposure to the known risk periods of drug overdose [60]. It is essential to stringently assess a patient’s tolerance prior to treatment initiation and improve retention over the long term. Moreover, a post-treatment follow-up should be considered after medication treatment cessation, particularly in the first few weeks following treatment discharge.

As a supplement of our findings, evidence showed that OST was also an effective strategy to reduce deaths both in prison and the immediate post-release period. Cohort studies found that prison-based OST was associated with a 74% reduction of hazard of death during incarceration, and continuation of OST after release was associated with a substantial reduction in all-cause and overdose mortality in the first month after release [61, 62]. Receiving OST in prison contributed to greater community treatment engagement in the first month of release [63]. This continuation of OST after release could further reduce the risk of death from overdose in the immediate post-release period [64]. These evidence supported the scale-up of MAT provision in both community and prison settings to reduce mortality risk.

The impact of different medicines on mortality estimates and risks

Methadone, buprenorphine, and naltrexone are currently three medications approved for treating OUDs [7]. As a full opioid agonist, methadone is the most frequently used medication in OST for the management of OUDs [65]; however, buprenorphine, a partial agonist, is thought to have a superior safety profile to methadone as it is less likely to induce respiratory depression, posing a lower mortality risk from overdose, especially during treatment induction [20]. This is consistent with the results of our study which showed a lower all-cause mortality rate during buprenorphine treatment than during methadone treatment. Moreover, we found a significantly lower mortality risk with buprenorphine than methadone after medication treatment cessation.

Naltrexone is an opioid antagonist that blocks the pleasurable and euphoric effects of heroin and other opiates [7]. Compared to methadone and buprenorphine, the evidence for the efficacy of naltrexone antagonist therapy in mortality remains comparatively weak. The pooled result in our study showed that mortality rate during naltrexone treatment was lower than that during methadone and buprenorphine treatment. This was the first study to integrate data about the effects of naltrexone on mortality rate and risk, which verified naltrexone as an effective medication treatment on prevention of premature mortality, especially extended-release or implant naltrexone treatment. Implant formulations represent a preferable option to deliver naltrexone treatment, since it has been associated with increasing overall time in treatment [12]. Prior research has shown that patients treated with implant naltrexone have significantly lower mortality than patients treated with oral naltrexone [66], and naltrexone implants had a favorable mortality profile both during and after treatment than buprenorphine [30]. A randomized clinical trial also indicated that extended-release naltrexone was as safe and effective as buprenorphine–naloxone in preventing relapse for patients who successfully initiated medication [67]. Additionally, more benefits could be gained by retaining patients in naltrexone and buprenorphine treatment, as the magnitude of RRs for all-cause mortality after treatment versus during treatment were larger for naltrexone and buprenorphine than for methadone. Taken altogether, these results underscore the potential benefits of naltrexone as a treatment of opioid use disorder. More large-scale studies directly comparing the mortality risk of naltrexone, particularly in the form of implants and extended-release, with other medications are warranted in the future.

The impact of other characteristics on mortality estimates

There was significant heterogeneity among the included studies, such as large geographic variation and sex difference. Australasia had lower all-cause CMR estimates both during in-treatment period and after exiting from MAT compared with other regions. This may be explained by the low background mortality among opioids users in Australasia [5]. Moreover, treatment settings (primary care-based or specialized treatment clinics), mode (low-threshold or high-threshold) and quality (supervised or take-home medication dispensation) of treatment delivered in different countries were varied, being the possible reasons of different mortality risk [68, 69]. Notably, no studies from regions such as Africa were identified, and further evidence of the effectiveness of MAT is required in these settings. We found that female opioids users had lower risk of mortality, which may be related to the fact that females seem to have a greater chance of surviving an overdose [53]. Additionally, studies published after 2010 tend to report lower mortality rates either in or off MAT. This may be related to the increasing proportion of opioids users entering buprenorphine therapy over time [70], which is associated with reduction in overdose deaths [71].

Differences were also found in mortality rates according to cohort size and study design. These factors may explain the large heterogeneity among included studies. Related to the dosage of medicine, no significant difference in mortality risk was found between low- and high-methadone dosage groups; however, few studies were available for analysis. Prior evidence has shown that higher methadone dosages were superior in retaining patients [43]. Further research is needed to examine the role that dosage might play in reducing mortality across the available medications.

Strengths and limitations

This meta-analysis was the first study to provide pooled net reduction of all-cause and overdose mortality rate in individuals with OUDs who receiving the MAT, compared with the individuals not getting medication treatment, which was different from the recent meta-analysis mainly focusing on the mortality rate during and after OST [21]. We identified several additional studies that were not included in the above meta-analysis [24, 25, 29, 38, 39, 42, 45,46,47, 49, 51, 52, 55, 56], and added the estimate of mortality rates and risks among different medication types, treatment retention, and drug doses. In addition to OST including methadone and buprenorphine, we firstly pooled the mortality estimate of antagonist treatment with naltrexone among individual with OUDs. Thus, our findings have extended and added to the current knowledge on the comprehensive effects of MAT on mortality risk among patients with OUDs receiving medicines treatment. In addition, we provided the pooled evidence that longer retention time of MAT reduced the risk of mortality.

Our review has several limitations. First, most included studies were registry-based studies which were vulnerable to selection bias. Differences between individuals receiving different medications could account for the differing effects that were associated with the medications. Misclassification may have occurred in death and treatment group in studies based on data linkage. Besides, the number of studies regarding to naltrexone treatment was limited. Second, the studies included were still predominantly from Europe and Australasia. Thus, the results of the present study may not reflect the effectiveness of MAT in worldwide. Studies in low and middle income countries are especially needed. Third, information about ancillary services such as counseling, or psychosocial services was limited. The provision of additional interventions such as psychosocial interventions might additionally reduce the risk of death [54]. Finally, most of individuals included in this meta-analysis were either from drug-free programs or underwent a short duration of detoxification when waiting for entry into MAT [41]. These individuals may differ from the individuals who never received any drug treatment. Notwithstanding these limitations, this analysis has added quantitative evidence on the effectiveness of MAT in reducing mortality among patients with OUDs in community settings.

Implications

This study provided strong quantitative evidence that links mortality rate and mortality risk to different characteristics of MAT, further highlighting the importance of MAT in the prevention of premature death among individuals with OUDs. Opioids-related deaths remain a large burden to public health [4]. Increased efforts and interventions should be focused on lengthening retention time and target patients in their early stage of medication treatment initiation and cessation, and post-treatment follow-up should be emphasized after medication treatment cessation. Despite the protective effect of MAT on mortality, barriers still exist in the access to and utilization of MAT [7]. Facing the situation of limited coverage of OST and other harm reduction interventions in worldwide [72], increased coverage and expanding access to MAT are crucial components of the effort to reduce opioids-related deaths and alleviate disease burden. Naltrexone represents an alternative option to deliver medication treatment for those with OUDs, especially in the form of implants or extended-release, given that it was associated with a lower mortality rate than methadone and buprenorphine treatment. More researches are still needed regarding to the effects of long-acting naltrexone treatment on mortality and other aspects among individuals with OUDs.