Risk of Experiencing an Overdose Event for Patients Undergoing Treatment With Medication for Opioid Use Disorder
Publication: American Journal of Psychiatry
Abstract
Objective:
Overdose risk during a course of treatment with medication for opioid use disorder (MOUD) has not been clearly delineated. The authors sought to address this gap by leveraging a new data set from three large pragmatic clinical trials of MOUD.
Methods:
Adverse event logs, including overdose events, from the three trials (N=2,199) were harmonized, and the overall risk of having an overdose event in the 24 weeks after randomization was compared for each study arm (one methadone, one naltrexone, and three buprenorphine groups), using survival analysis with time-dependent Cox proportional hazard models.
Results:
By week 24, 39 participants had ≥1 overdose event. The observed frequency of having an overdose event was 15 (5.30%) among 283 patients assigned to naltrexone, eight (1.51%) among 529 patients assigned to methadone, and 16 (1.15%) among 1,387 patients assigned to buprenorphine. Notably, 27.9% of patients assigned to extended-release naltrexone never initiated the medication, and their overdose rate was 8.9% (7/79), compared with 3.9% (8/204) among those who initiated naltrexone. Controlling for sociodemographic and time-varying medication adherence variables and baseline substance use, a proportional hazard model did not show a significant effect of naltrexone assignment. Significantly higher probabilities of experiencing an overdose event were observed among patients with baseline benzodiazepine use (hazard ratio=3.36, 95% CI=1.76, 6.42) and those who either were never inducted on their assigned study medication (hazard ratio=6.64, 95% CI=2.12, 19.54) or stopped their medication after initial induction (hazard ratio=4.04, 95% CI=1.54, 10.65).
Conclusions:
Among patients with opioid use disorder seeking medication treatment, the risk of overdose events over the next 24 weeks is elevated among those who fail to initiate or discontinue medication and those who report benzodiazepine use at baseline.
In the United States, opioid overdose deaths have increased significantly over the past two decades (1, 2). This trend has been attributed to several factors, including an increased rate of prescription of opioid painkillers (3), subsequent cross-conversion to heroin use (4), and a dramatic rise in the availability of fentanyl and related high-potency synthetic opioids (5). A substantial body of evidence has shown that treatment with medication for opioid use disorder (MOUD) can decrease both overdose risk and all-cause mortality among people with opioid use disorder (OUD) (6). Therefore, MOUD treatment is one of the critical pathways for overdose prevention (7, 8) in addition to other effective public health–oriented and systems-level interventions (9–11). However, although substantially lower compared with periods out of care, an excess risk of drug-related death seems to remain during pharmacological treatment (12–15), with a particularly elevated risk of mortality during the induction phase to MOUD and the time immediately after leaving treatment (16, 17).
It is important to understand what factors place an individual at increased risk for opioid overdose, and this has been a target of predictive modeling. For example, based on a retrospective cohort, a model including five variables (age, age-squared, medications dispensed, tobacco use, and mental health and substance use diagnoses) predicted 66%–82% of overdoses that occurred among individuals on chronic opioid treatment over the next 2 years (c-statistic=0.75) (18). Another retrospective cohort study, using Prescription Drug Monitoring Program data, showed that patterns of prescription fills (including long- and short-acting opioids, buprenorphine, benzodiazepines, and muscle relaxants) can help identify those at risk of fatal overdose events (19). The focus of these efforts is on using information gathered by primary care, pain management, and other physicians at the time of initial engagement with patients to make a decision to either reduce or eliminate opioid prescriptions (20). Risk prediction models can also make a specific but not sensitive estimate of opioid overdose risk within the general population (21), and given that drug overdose is a statistically rare event in the general population, it is difficult to apply these models in clinical contexts.
On an individual level, the risk of an overdose event during a course of treatment with MOUD has not been clearly delineated. Therefore, a complementary goal for risk scoring modeling in this field is to focus on the patient population that is more relevant to addiction treatment professionals, rather than the overall population or patients who are receiving opioid prescriptions for pain control: Once a patient is diagnosed with opioid use disorder (OUD) and engaged in MOUD treatment, what is the risk of this individual having an overdose event? Another relevant question is whether the medication used influences the risk of overdose in the period immediately after treatment initiation. Retrospective cohort studies have suggested that buprenorphine treatment is associated with lower overdose risk compared with methadone (22) and extended-release naltrexone (23), while crude mortality rates of patients treated with methadone, buprenorphine, or naltrexone seem to be comparable (24). Another retrospective comparative effectiveness study (25) indicated that treatment with buprenorphine or methadone, but not naltrexone, was associated with a reduced risk of overdose compared with no treatment during the 12-month follow-up period. However, given the lack of randomization, these studies are limited by differences between cohorts in baseline characteristics and confounding by indication.
Our aim in this study was to address these research gaps by leveraging a new data set, harmonizing three large pragmatic clinical trials of MOUD treatment conducted within the National Institute on Drug Abuse’s National Drug Abuse Treatment Clinical Trials Network (CTN), where the treatment group was randomly assigned and harmonized safety monitoring data were used to track fatal and nonfatal overdose events. Using a panel of baseline predictors, we constructed predictive models that estimate the risk of overdose during MOUD treatment and compared the overall risk for each of the treatment groups (methadone, buprenorphine, and extended-release injection naltrexone). We hypothesized that several common predictor variables, such as baseline substance use, particularly use of benzodiazepines and opioids (19, 26), and medication nonadherence (16, 17) would be associated with an increased overdose risk during MOUD treatment.
Methods
Clinical Trial Data Sources
Data from three large randomized controlled MOUD clinical trials from the CTN were included in our analysis: CTN0027 (START) (27), CTN0030 (POATS) (28), and CTN0051 (X:BOT) (29). Full protocols for the individual studies have been published. Briefly, all three trials enrolled adult participants who met DSM-IV-TR criteria for opioid dependence or DSM-5 criteria for OUD. The studies employed broad pragmatic inclusion criteria and few exclusion criteria other than major medical and unstable psychiatric comorbidities (and individuals with heroin use only [CTN0030] and current methadone treatment [CTN0051]). Given the open-label design of these studies, participants who failed to initiate the study medication or stopped their medication during the trial were not automatically excluded from the study. Rather, induction failure (29) or treatment exposure (i.e., percentage of dispensed medication doses) (27) were assessed as secondary outcomes.
CTN0027, which had a 24-week treatment period, randomized participants to open-label outpatient methadone or buprenorphine treatment after 12–24 hours of abstention from opioids and presenting with mild withdrawal. The maximum maintenance dosage for buprenorphine was 32 mg/day (mean maximum daily dose, 22.1 mg [SD=8.2]), and no maximum methadone maintenance dosage was specified (mean maximum daily dose, 93.2 mg [SD=42.2]).
CTN0030 had two phases. In phase 1, participants who were primary prescription opioid users seeking outpatient treatment were inducted on buprenorphine, with a maintenance dosage of 8–32 mg/day. Participants were randomized to standard medical management (a manual-based brief intervention delivered by the study physician) or opioid dependence counseling (longer sessions by trained nonphysician psychotherapy providers). Buprenorphine was tapered at weeks 3–4, and participants were followed for another 4–8 weeks. About 6.8% of participants succeeded by week 12 in maintaining abstinence, and the rest were invited to enter phase 2, which had a 16-week treatment period: a 12-week buprenorphine stabilization treatment followed by a 4-week taper; the study also included an 8-week posttreatment follow-up. For the present analysis, we considered baseline data collected at the start of phase 1 and clinical data from phase 2, given that this portion of the study matched the condition of the other two trials. There were no recorded overdose events in phase 1.
In CTN0051, which had a 24-week treatment period, participants were initially detoxified (using a variety of pragmatic protocols) in an inpatient setting and were randomized to outpatient treatment with buprenorphine or extended-release naltrexone. Naltrexone was given every 28 days, and initial dosing required passing a naloxone challenge, with rechallenge in case of a missed dose before reinitiation of naltrexone. The maximum maintenance dosage of buprenorphine was 24 mg/day.
A helical team approach for harmonization was used, where data integrity was verified by two teams of statisticians for the baseline covariates. For the present analyses, we included the following harmonized baseline characteristics: age, sex, race/ethnicity; self-reported use of heroin, prescription opioids, benzodiazepines, cannabis, and stimulants (using the timeline followback method [30]); alcohol intoxication (defined as ≥3 drinks/day, assessed with the Addiction Severity Index–Lite [31] for CTN0030 and the timeline followback for CTN0027 and CTN0051) in the 28 days before baseline; current nicotine dependence (using the Fagerström Test for Nicotine Dependence [32]); and lifetime diagnoses of panic disorder, anxiety disorder, and major depression. We also harmonized the adverse event logs with case report forms from all three trials. Adverse event and serious adverse event forms were completed weekly until week 24 for CTN0027 and CTN0051, and biweekly until week 16 for CTN0030. Nonfatal self-reported overdose events and deaths were queried using standard adverse event and serious adverse event forms and were reviewed by an independent medical monitor.
Identification of Overdose Events in the Data
We first compiled a list of events that contained the term “overdose” in any row in available safety data from both the adverse event form and the serious adverse event narratives of the randomized sample (N=2,199). The following information was abstracted for each event: any information reported on the adverse event form, the Medical Dictionary for Regulatory Activities (MedDRA) preferred term for the event, event onset date, name of serious adverse event, description of serious adverse event, treatment information (including type of treatment, dosage, starting date, and most recent date of treatment administration), relevant history, and study. Three independent raters (L.B., F.C., S.X.L.) reviewed this list to identify likely overdose events using clinical judgment. Events were coded 1, 2, or 3, indicating likely opioid overdose, likely non-opioid overdose, and not overdose, respectively. The raters achieved substantial agreement (Fleiss’ kappa=0.74) (33).
Subsequently, any disagreements in the ratings were discussed in consultation with a fourth author (E.V.N.). The final categorization of overdose events followed a classification that had been used previously to explore CTN0051 overdose data (34). Events were categorized as a likely overdose if “overdose” or “opioid overdose” was the MedDRA preferred term for the event. Alternatively, “overdose” or “opioid overdose” was not the MedDRA preferred term for the event, but the word “overdose” appeared in the listing on the adverse event form or in the serious adverse event narrative.
Type of overdose (opioid, other substance, substance unknown) was coded as a separate variable. Any likely overdose event that met at least one of the following criteria was classified as a likely opioid overdose: the adverse event or serious adverse event name or MedDRA preferred term indicated heroin, opioid, or opiate overdose or toxicity; the adverse event or serious adverse event was treated with naloxone; the death certificate indicated heroin, opioid, or opiate overdose; the patient was treated for the adverse event or serious adverse event in the hospital and the narrative indicated heroin or opioids.
Data Analysis
We included only the first overdose for participants with multiple events and censored cases without an overdose event by study week 24. A log-rank test was used to test differences in the probability of an overdose event by the time of last follow-up or week 24 between groups with different demographic and baseline characteristics. Binary indicators of baseline substance use, rather than continuous indicators (number of days of use in the 28 days before baseline), were used because there was no linear relationship between days of use and overdose risk. In addition, binary indicators may be more useful in clinical practice given their ease of assessment and computation and their straightforward clinical interpretability. We report the observed percentage of having an overdose event in the 24 weeks after randomization by assigned treatment (buprenorphine, methadone, and extended-release naltrexone) overall and depending on induction status (inducted onto medication, yes/no) and baseline benzodiazepine use (yes/no).
Survival analysis with a time-dependent Cox proportional hazard model (35) was used to test the effects of treatment group assignment (CTN0027: buprenorphine; CTN0027: methadone; CTN0030: buprenorphine; CTN0051: extended-release naltrexone; CTN0051: buprenorphine) on drug overdose. The model controlled for sociodemographic variables (sex, age, race/ethnicity), any self-reported use, in the 28 days before baseline, of heroin, prescription opioids, and benzodiazepines (yes/no for each), and three time-varying covariates—dummy variables for not having been inducted on the assigned medication at the time of overdose, having stopped the assigned medication after initial induction and before end of treatment, and number of days stopping the assigned medication. Two cases without information on baseline substance use were excluded from these analyses.
A sensitivity analysis confirmed the conclusion from the Cox proportional hazard model with a competing risk model (the Fine-Gray model) (36), which aims to correctly estimate the probability of an event (here, overdose) irrespective of the outcome of another variable (competing event; here, study dropout). If the censoring is confounded with the outcome, the competing risk model estimates the marginal probability of overdose in the presence of a competing event (study dropout), and thus the overdose probability is unbiased. In addition, we repeated the proportional hazard model testing the effects of treatment group assignment and other covariates on the subsample of participants who received at least one dose of the assigned study medication. Statistical analyses were performed with proc phreg in SAS, version 9.4.
Results
Overdose events and the substance involved in the overdose are tabulated in Table 1. We identified 57 likely overdose events total, and opioids were either the only substance or one of multiple substances involved in 28 of these events. Among the events without opioid involvement, the majority were likely benzodiazepine overdoses (N=7), followed by cocaine/stimulant overdoses (N=4), and one polysubstance overdose involving benzodiazepines, alcohol, and a cold and flu medication that contains acetaminophen, dextromethorphan, and doxylamine. The 57 overdose events were experienced by 51 participants. Forty-six participants experienced one overdose, four participants experienced two overdoses, and one participant experienced three overdoses during their time in the study. Twelve “likely overdoses” occurred after week 24 (two in CTN0027 in weeks 27 and 29, and 10 in CTN0051 in weeks 26, 28, 29, 29, 31, 34, 36, 36, 36, and 47) and were not counted as overdose events in subsequent analyses.
| Study | Total (N) | |||
|---|---|---|---|---|
| Event and Substance | CTN0027 (N) | CTN0030 (N) | CTN0051 (N) | |
| Likely overdose event | 17 | 5 | 35 | 57 |
| Substance involved | ||||
| Opioid | 5 | 3 | 20 | 28 |
| Other | 8 | 2 | 2 | 12 |
| Unknown | 4 | 0 | 13 | 17 |
a
Events were categorized as a likely overdose if “overdose” or “opioid overdose” was the Medical Dictionary for Regulatory Activities (MedDRA) preferred term for the event. Alternatively, “overdose” or “opioid overdose” was not the MedDRA preferred term for the event, but the word “overdose” appeared in the listing on the adverse event form or in the serious adverse event narrative (34).
Overall, 1.78% of participants (39 of 2,199) experienced at least one overdose event by the time of their last follow-up or week 24, whichever occurred first. Table 2 summarizes the overdose events by sociodemographic characteristics, treatment group, baseline substance use, and lifetime anxiety, panic, or major depression diagnosis.
| Drug Overdoses | Log-Rank Test | |||||
|---|---|---|---|---|---|---|
| Characteristic | Overall N | N | % | χ2 | df | p |
| Total | 2,199 | 39 | 1.77 | |||
| Sex | 2.79 | 1 | 0.0950 | |||
| Male | 1,471 | 31 | 2.11 | |||
| Female | 728 | 8 | 1.10 | |||
| Ethnicity | 0.43 | 4 | 0.5106 | |||
| Hispanic | 323 | 7 | 2.17 | |||
| Non-Hispanic | 1,876 | 32 | 1.71 | |||
| Race | 2.71 | 2 | 0.2576 | |||
| White | 1,653 | 31 | 1.88 | |||
| Black | 175 | 5 | 2.86 | |||
| Other | 371 | 3 | 0.81 | |||
| Assigned treatment (study) | 21.62 | 4 | 0.0002 | |||
| Buprenorphine (CTN0027) | 740 | 7 | 0.95 | |||
| Methadone (CTN0027) | 529 | 8 | 1.51 | |||
| Buprenorphine (CTN0030) | 360 | 3 | 0.83 | |||
| Buprenorphine (CTN0051) | 287 | 6 | 2.09 | |||
| Extended-release naltrexone (CTN0051) | 283 | 15 | 5.30 | |||
| Baseline heroin useb | 1.30 | 1 | 0.2547 | |||
| Yes | 1,725 | 33 | 1.91 | |||
| Noc | 472 | 6 | 1.27 | |||
| Baseline prescription opioid useb | 1.61 | 1 | 0.2043 | |||
| Yes | 848 | 12 | 1.42 | |||
| No | 1,349 | 27 | 2.00 | |||
| Baseline benzodiazepine useb | 13.84 | 1 | 0.0002 | |||
| Yes | 472 | 18 | 3.81 | |||
| No | 1,725 | 21 | 1.22 | |||
| Baseline heroin and benzodiazepine useb | 17.28 | 1 | <0.0001 | |||
| Yes | 330 | 15 | 4.55 | |||
| No | 1,867 | 24 | 1.29 | |||
| Baseline cannabis useb | 2.33 | 1 | 0.1268 | |||
| Yes | 747 | 18 | 2.41 | |||
| No | 1,450 | 21 | 1.45 | |||
| Baseline stimulant useb,d | 0.25 | 1 | 0.6178 | |||
| Yes | 1,000 | 23 | 2.30 | |||
| No | 1,197 | 16 | 1.33 | |||
| Baseline alcohol intoxicatione | 1.10 | 1 | 0.2935 | |||
| Yes | 442 | 11 | 2.49 | |||
| No | 1,755 | 28 | 1.60 | |||
| Baseline nicotine dependencef | 0.05 | 1 | 0.8169 | |||
| Yes | 364 | 7 | 1.92 | |||
| No | 1,828 | 32 | 1.75 | |||
| Lifetime diagnosis of anxiety or panic disorder or major depression | 2.22 | 1 | 0.1363 | |||
| Yes | 999 | 23 | 2.30 | |||
| No | 1,200 | 16 | 1.33 | |||
a
The table includes only the first overdose for participants who had multiple events, and cases without an overdose event by study week 24 were censored.
b
Any use in the 28 days before baseline, assessed with the timeline followback.
c
Including 446 prescription opioid users without heroin use.
d
Including cocaine, amphetamine, and methamphetamine use.
e
Any alcohol intoxication, defined as ≥3 drinks per day, in the 28 days before baseline, assessed with Addiction Severity Index–Lite in CTN0030 and the timeline followback in CTN0027 and CTN0051.
f
Fagerström Test for Nicotine Dependence score ≥7.
The observed percentages of having an overdose event in the 24 weeks after randomization by medication group are presented in Table 3. Overall, the observed rate of having an overdose event was 5.30% (15/283) among patients assigned to naltrexone, 1.51% (8/529) among those assigned to methadone, and 1.15% (16/1,387) among those assigned to buprenorphine. Self-reported use of benzodiazepines at baseline was associated with an increased observed percentage of having an overdose event in all three groups (Table 3). Notably, more than one-fourth of patients (27.9%) assigned to extended-release naltrexone never initiated the medication, compared with only 2.2% of those assigned to buprenorphine and 1.7% of those assigned to methadone. Among those who initiated extended-release naltrexone, the overdose rate was 3.9% (8/204), compared with 8.9% (7/79) among patients who failed to initiate.
| Methadone | Buprenorphine | Extended-Release Naltrexone | |||||||
|---|---|---|---|---|---|---|---|---|---|
| Measure | Total (N) | Overdose (N) | % | Total (N) | Overdose (N) | % | Total (N) | Overdose (N) | % |
| Overall | 529 | 8 | 1.51 | 1,387 | 16 | 1.15 | 283 | 15 | 5.30 |
| Benzodiazepine usea | |||||||||
| Yes | 81 | 6 | 7.41 | 324 | 6 | 1.85 | 67 | 6 | 8.96 |
| No | 447 | 2 | 0.45 | 1,062 | 10 | 0.94 | 216 | 9 | 4.17 |
| Inducted on medication | |||||||||
| Yesb | 520 | 8 | 1.54 | 1,357 | 15 | 1.11 | 204 | 8 | 3.92 |
| No | 9 | 0 | 0.00 | 30 | 1 | 3.33 | 79 | 7 | 8.86 |
a
Any use in the 28 days before baseline, assessed with the timeline followback.
b
The participant received at least one dose of the assigned study medication.
Fifteen participants who experienced an overdose by week 24 stopped their medication prior to the event. Among participants assigned to methadone (eight overdose events total), one participant stopped the medication 1 day before the overdose event. In the three buprenorphine groups (16 overdose events total), eight participants stopped their medication 1 (N=2), 16, 21, 30, 91, 96, and 119 days before the overdose event. In the naltrexone group (15 overdose events total), six participants stopped their medication 5, 14, 27, 45, 59, and 121 days before the overdose event. An additional eight participants (one assigned to buprenorphine and seven to naltrexone) were not inducted on their assigned study medication before the overdose event.
A proportional hazard model (Table 4) testing the effects of study medication group assignment (five groups) on overdose, controlling for sociodemographic variables, baseline use of benzodiazepines, heroin, and prescription opioids, and three time-varying medication adherence covariates, revealed significant effects of baseline benzodiazepine use (χ2=13.46, p<0.001), noninduction to the assigned medication (χ2=10.81, p=0.001), and stopping the assigned medication after induction (χ2=8.00, p=0.005) on overdose. The overall effect of naltrexone assignment on overdose was not significant (χ2=8.63, df=4, p=0.071). Those who reported use of benzodiazepines in the 28 days before baseline (hazard ratio=3.36, 95% CI=1.76, 6.42, p<0.001) and those who either were never inducted on their assigned study medication (hazard ratio=6.64, 95% CI=2.12, 19.54, p=0.001) or stopped their medication after initial induction (hazard ratio=4.04, 95% CI=1.54, 10.65, p<0.001) had a significantly higher probability of experiencing an overdose event. However, the number of days stopping the assigned medication did not have a significant effect on overdose risk. In addition, there was no significant interaction between assigned medication and having stopped the medication (χ2=0.498, df=2, p=0.780) or between the assigned medication and baseline benzodiazepine use (χ2=5.55, df=2, p=0.062).
| Variable | Hazard Ratio | 95% CI | χ2 | Pr > χ2 |
|---|---|---|---|---|
| Assigned treatment (study) | ||||
| Extended-release naltrexone (CTN0051) (ref) | ||||
| Buprenorphine (CTN0027) | 0.68 | 0.23, 2.01 | 0.47 | 0.491 |
| Methadone (CTN0027) | 0.87 | 0.29, 2.57 | 0.07 | 0.796 |
| Buprenorphine (CTN0030) | 0.12 | 0.02, 0.64 | 6.15 | 0.013 |
| Buprenorphine (CTN0051) | 0.36 | 0.14, 0.97 | 4.08 | 0.043 |
| Female | 0.60 | 0.27, 1.31 | 1.66 | 0.197 |
| Age | 0.96 | 0.93, 1.00 | 3.82 | 0.051 |
| Hispanic | 1.86 | 0.75, 4.58 | 1.82 | 0.178 |
| Race | ||||
| White (ref) | ||||
| Black | 1.91 | 0.67, 5.44 | 1.45 | 0.228 |
| Other | 0.30 | 0.09, 1.06 | 3.48 | 0.062 |
| Baseline substance usea | ||||
| Heroin | 0.56 | 0.16, 1.95 | 0.83 | 0.362 |
| Benzodiazepines | 3.36 | 1.76, 6.42 | 13.46 | 0.000 |
| Prescription opioids | 0.55 | 0.23, 1.29 | 1.91 | 0.167 |
| Medication adherence | ||||
| Not inductedb | 6.44 | 2.12, 19.54 | 10.81 | 0.001 |
| Stoppedc | 4.04 | 1.54, 10.65 | 8.00 | 0.005 |
| Stop daysd | 1.00 | 0.98, 1.01 | 0.01 | 0.941 |
a
Any use in the 28 days before baseline, assessed with the timeline followback.
b
Time-varying dummy variable for not having been inducted on the assigned medication at the time of overdose.
c
Time-varying dummy variable for having stopped the assigned medication before the overdose event or end of treatment.
d
Time-varying continuous variable for number of days stopping the assigned medication before the overdose event or end of treatment.
A sensitivity analysis on a competing risk model to estimate the marginal probability of overdose in the presence of a competing event (study dropout) did not alter the main findings (see Table S1 in the online supplement ). Note that this model does not include the three time-varying medication adherence covariates. In a competing risk model on the subsample that initiated their assigned medication, the main effect of study medication group assignment (five groups) on overdose was no longer significant (χ2=5.99, df=4, p=0.200). Another sensitivity analysis on a proportional hazard model on the subsample that initiated their assigned medication also found that those who reported benzodiazepine use in the 28 days before baseline and/or stopped their assigned medication after initial induction had a significantly higher probability of experiencing an overdose event. However, overdose rates did not differ by treatment group (see Table S2 in the online supplement ).
Discussion
The aim of this study was to estimate the risk of overdose events once a patient is diagnosed with OUD and engaged in MOUD treatment and to test whether the assignment to a medication (methadone, buprenorphine, extended-release naltrexone) influences this risk. Our data, drawn from three previously completed clinical trials, indicate that patients undergoing MOUD treatment have a nonnegligible risk of overdose events in the first 24 weeks of treatment (39 of 2,199 patients; 1.78%). The overall prevalence of overdose was comparable with that in a nationally representative sample of commercial insurance and Medicare Advantage enrollees with OUD, of whom 1.7% (N=707) experienced an overdose within a 12-week follow-up period (25).
Most overdoses by week 24 were observed in CTN0051. A defining feature of the study was recruitment of patients from inpatient detoxification units who, compared with patients initiating outpatient medication treatment, may be at greater risk of early “treatment failure” (i.e., dropout or induction failure) and overdose after leaving the controlled setting (37, 38). Compared with CTN0051 and CTN0027 (27, 29), both of which included patients who primarily used heroin, notably fewer overdoses were observed among CTN0030 participants, who were primary prescription opioid users seeking outpatient treatment (28). Prescription opioid use has been associated with lower health risks and less severe patterns of psychopathology compared with heroin use (39, 40). In addition, both CTN0030 and CTN0027 were conducted earlier than CTN0051 and likely predate the occurrence of fentanyl in the heroin supply (41).
In addition to sample differences, differences in medication dispensing need to be considered. In CTN0027, study medication was dispensed daily (methadone) or three times weekly (buprenorphine; clinic schedule permitting). In CTN0030, buprenorphine was dispensed during weekly follow-up visits. And in CTN0051, buprenorphine was dispensed to participants weekly until week 4 and then biweekly, and extended-release naltrexone injections were given approximately every 4 weeks or sooner if clinically indicated, with a minimum of 21 days between injections. Different dispensing schedules and methods may have an impact on medication adherence and overdose risk, respectively. In our data, we observed more patients who stopped their medication prior to an overdose event in the buprenorphine group (eight of 16 participants who overdosed) and the naltrexone group (six of 15 participants who overdosed) compared with the methadone group (one of eight participants who overdosed). Treatment with naltrexone, along with the required detoxification period before initiation, increases sensitivity to the effects of opioids, and when naltrexone is stopped, the patient is in a nontolerant state, with the accompanying risk of overdose that is clearly articulated in the official prescribing information for extended-release naltrexone (42). In our model, discontinuation of medication after induction, but not length of discontinuation, predicted overdose. Therefore, any interruption in treatment may substantially increase the risk of overdose, regardless of MOUD option (16, 17). However, the confidence intervals of medication adherence effects on overdose were wide, and these findings should be viewed as hypotheses to be tested on new data in future research. To achieve longer MOUD retention and better adherence in clinical practice, it is also important to identify and address demand-side barriers, such as patient cost sharing (43), structural barriers such as accessibility (44), and socioeconomic barriers such as unemployment and housing instability (45).
Among the three medication groups, assignment to extended-release naltrexone was associated with the highest observed rate of having an overdose event within 24 weeks. Retrospective cohort studies have suggested that treatment with methadone and buprenorphine, but not naltrexone, are associated with a reduced risk of overdose (23, 25, 46). However, no previous large clinical trial data set has provided a clear signal that treatment with extended-release naltrexone increases the risk of overdose compared with other MOUD options. In the primary intent-to-treat analyses of CTN0051, there were more overdoses among patients on extended-release injection naltrexone than among those on buprenorphine, but the difference did not reach significance (29). A secondary analysis of CTN0051 (47) suggests a higher risk of overdose on extended-release naltrexone, although much of this risk was attributable to naltrexone induction failure (34). The present analysis used a larger sample, which included data from CTN0051 and two other studies and therefore had enhanced power to detect a risk difference. However, in our model controlling for sociodemographic variables, baseline use of opioids and benzodiazepines, and medication adherence (induction failure, medication discontinuation, and length of time off medication), the overall effect of naltrexone assignment on overdose was nonsignificant.
Naltrexone is more difficult to initiate than methadone or buprenorphine (29), and the present results suggest that the observed increased overdose risk associated with naltrexone assignment is attributable, at least in part, to failed induction attempts. This assumption is corroborated by our sensitivity analysis on a proportional hazard model after first induction on the assigned medication, which showed no significant difference in overdose rate between treatment groups. If naltrexone initiation is not feasible, clinicians should encourage patients to pivot to another MOUD option. For example, buprenorphine can be easily started when an attempt to initiate naltrexone appears to be failing (48), and it would protect against overdose.
Finally, benzodiazepine use in the 28 days before starting MOUD was a predictor of overdose risk during treatment. This finding extends previous results showing that benzodiazepines, either coprescribed (49) or used illicitly (50, 51), are associated with an increased risk of overdose mortality among people with OUD. A comparison of blood toxicology in drug-related fatalities among patients with OUD treated with methadone, buprenorphine, or naltrexone identified opioids and benzodiazepines in more than three-fourths of postmortem blood samples, with no difference between medication groups (52). Adding to these results, we did not find a significant interaction effect on overdose risk between assigned treatment and baseline benzodiazepine use. Benzodiazepines could be a marker for co-occurring mental health problems associated with overdose risk (18), and they increase the respiratory depressant effect of opioids (53).
Among this study’s strengths is that the data set represents the largest individual-level clinical trial data set in OUD treatment, harmonized through the multi-institutional CTN infrastructure, with commonly collected baseline predictors that are practical and interpretable in clinical practice. The randomized treatment assignment offered some protection against unmeasured confounding common in models derived from observational data sets. However, whereas participants were randomized to condition, they were not randomized to trial, and because not all trials included all MOUD options, there is the possibility of unmeasured confounding between those nonoverlapping study arms.
We took a systematic team approach to identifying overdose events in available safety monitoring data. However, a previous in-depth examination of adverse event and serious adverse event forms in CTN0051 suggests that there are some likely overdose events that are described with terminology other than “overdose” (34). In addition, standard adverse event and serious adverse event assessments do not specifically probe for overdose events, but rather for adverse events in general, and some participants may have neglected to report nonfatal overdoses. Moreover, for some fatalities in the trials, cause of death was unknown or not specified. Thus, the events examined here may be an underestimate of the true number of fatal and nonfatal overdoses in the samples. In addition, we were limited to the adverse event description provided by site clinicians based on all available clinical information, with varying detail and sources (i.e., some participants received care for their overdose and hospital records were available, while other events were self-reported or reported by relatives). Of note, some adverse event/serious adverse event records explicitly listed overdose events as suicide attempts; however, overall, there was not enough information to clearly differentiate between overdoses that were suicide attempts and those that were “accidental.” As previously recommended (34), employing a specific overdose survey instrument in future trials and systematically probing for and characterizing overdose events at each visit (including substances involved and suicidal intent) would enhance our ability to track and subsequently predict overdoses among patients with OUD during MOUD treatment (54).
The application of our model results also has some real-world challenges. Many individuals with OUD receive intermittent treatment and do not return to the clinic on a regular basis, and their risk of experiencing an overdose event may be poorly estimated by the current set of models. CTN trial participants received reimbursements for attending scheduled visits, and outreach efforts were made to keep them engaged with the study (e.g., calls and appointment reminders by study staff on a regular basis). Without these research-specific retention efforts, community patients may be at higher risk of treatment nonadherence or dropout, and consequently at higher risk of overdose. Additionally, participants who discontinued study medication may have started treatment with another MOUD option; however, use of nonstudy or nonassigned medication and adherence (i.e., periods on and off another medication) were not consistently tracked in all three trials, and we were unable to include this information in our models. Finally, other factors, such as prior hospitalizations with a diagnosis of opioid poisoning or cardiovascular problems (55), might be relevant in identifying patients at risk of experiencing an overdose event during MOUD treatment, but were not measured consistently in the clinical trials included in our analyses.
Conclusions
Patients undergoing MOUD treatment remain at risk of overdose events in the first 24 weeks after seeking treatment. The strongest message from these data is that patients who fail to initiate medication, or stop their medication, are at greater risk of experiencing an overdose event. The pharmacology of methadone, buprenorphine, and naltrexone is such that they all substantially lower overdose risk if taken as prescribed. Patients should be educated about overdose risk, the protective effect of MOUD, and the danger of discontinuing medication. Benzodiazepine use is also a signal of risk, and patients taking benzodiazepines should be evaluated and treated for mental health problems as part of an effort to wean them off benzodiazepines. The risk of overdose after discontinuing naltrexone may be greater than for other medications, although the present data are not definitive on this point, and the overall effect of naltrexone assignment on overdose was not statistically significant. Future large trials should implement more systematic assessments of overdose events based on a clear operationalization of overdose, querying actively rather than relying on spontaneous report, with detailed characterization of the event, including the substances involved and whether there was suicidal intent.
Supplementary Material
File (appi.ajp.20220312.ds001.pdf)
- View/Download
- 51.38 KB
REFERENCES
1.
Bohnert ASB, Valenstein M, Bair MJ, et al: Association between opioid prescribing patterns and opioid overdose-related deaths. JAMA 2011; 305:1315–1321
2.
Mattson CL, Tanz LJ, Quinn K, et al: Trends and geographic patterns in drug and synthetic opioid overdose deaths—United States, 2013–2019. MMWR Morb Mortal Wkly Rep 2021; 70:202–207
3.
Centers for Disease Control and Prevention: Vital signs: overdoses of prescription opioid pain relievers—United States, 1999–2008. MMWR Morb Mortal Wkly Rep 2011; 60:1487–1492
4.
Jones CM, Logan J, Gladden RM, et al: Vital signs: demographic and substance use trends among heroin users—United States, 2002–2013. MMWR Morb Mortal Wkly Rep 2015; 64:719–725
5.
O’Donnell JK, Gladden RM, Seth P: Trends in deaths involving heroin and synthetic opioids excluding methadone, and law enforcement drug product reports, by census region—United States, 2006–2015. MMWR Morb Mortal Wkly Rep 2017; 66:897–903
6.
Ma J, Bao Y-P, Wang R-J, et al: Effects of medication-assisted treatment on mortality among opioids users: a systematic review and meta-analysis. Mol Psychiatry 2019; 24:1868–1883
7.
Volkow ND, Frieden TR, Hyde PS, et al: Medication-assisted therapies: tackling the opioid-overdose epidemic. N Engl J Med 2014; 370:2063–2066
8.
Schwartz RP, Gryczynski J, O’Grady KE, et al: Opioid agonist treatments and heroin overdose deaths in Baltimore, Maryland, 1995–2009. Am J Public Health 2013; 103:917–922
9.
Bao Y, Pan Y, Taylor A, et al: Prescription drug monitoring programs are associated with sustained reductions in opioid prescribing by physicians. Health Aff 2016; 35:1045–1051
10.
Wheeler E, Jones TS, Gilbert MK, et al: Opioid overdose prevention programs providing naloxone to laypersons—United States, 2014. MMWR Morb Mortal Wkly Rep 2015; 64:631–635
11.
Chen Y, Wang Y, Nielsen S, et al: A systematic review of opioid overdose interventions delivered within emergency departments. Drug Alcohol Depend 2020; 213:108009
12.
Walley AY, Lodi S, Li Y, et al: Association between mortality rates and medication and residential treatment after in‐patient medically managed opioid withdrawal: a cohort analysis. Addiction 2020; 115:1496–1508
13.
Evans E, Li L, Min J, et al: Mortality among individuals accessing pharmacological treatment for opioid dependence in California, 2006–10. Addiction 2015; 110:996–1005
14.
Davoli M, Bargagli AM, Perucci CA, et al: Risk of fatal overdose during and after specialist drug treatment: the VEdeTTE study, a national multi-site prospective cohort study. Addiction 2007; 102:1954–1959
15.
Krawczyk N, Mojtabai R, Stuart EA, et al: Opioid agonist treatment and fatal overdose risk in a state‐wide US population receiving opioid use disorder services. Addiction 2020; 115:1683–1694
16.
Sordo L, Barrio G, Bravo MJ, et al: Mortality risk during and after opioid substitution treatment: systematic review and meta-analysis of cohort studies. BMJ 2017; 357:j1550
17.
Cornish R, Macleod J, Strang J, et al: Risk of death during and after opiate substitution treatment in primary care: prospective observational study in UK General Practice Research Database. BMJ 2010; 341:c5475
18.
Glanz JM, Narwaney KJ, Mueller SR, et al: Prediction model for two-year risk of opioid overdose among patients prescribed chronic opioid therapy. J Gen Intern Med 2018; 33:1646–1653
19.
Ferris LM, Saloner B, Krawczyk N, et al: Predicting opioid overdose deaths using Prescription Drug Monitoring Program data. Am J Prev Med 2019; 57:e211–e217
20.
Oliva EM, Bowe T, Tavakoli S, et al: Development and applications of the Veterans Health Administration’s Stratification Tool for Opioid Risk Mitigation (STORM) to improve opioid safety and prevent overdose and suicide. Psychol Serv 2017; 14:34–49
21.
Saloner B, Chang H-Y, Krawczyk N, et al: Predictive modeling of opioid overdose using linked statewide medical and criminal justice data. JAMA Psychiatry 2020; 77:1155–1162
22.
Kimber J, Larney S, Hickman M, et al: Mortality risk of opioid substitution therapy with methadone versus buprenorphine: a retrospective cohort study. Lancet Psychiatry 2015; 2:901–908
23.
Morgan JR, Schackman BR, Weinstein ZM, et al: Overdose following initiation of naltrexone and buprenorphine medication treatment for opioid use disorder in a United States commercially insured cohort. Drug Alcohol Depend 2019; 200:34–39
24.
Kelty E, Joyce D, Hulse G: A retrospective cohort study of mortality rates in patients with an opioid use disorder treated with implant naltrexone, oral methadone, or sublingual buprenorphine. Am J Drug Alcohol Abuse 2019; 45:285–291
25.
Wakeman SE, Larochelle MR, Ameli O, et al: Comparative effectiveness of different treatment pathways for opioid use disorder. JAMA Netw Open 2020; 3:e1920622
26.
Leece P, Cavacuiti C, Macdonald EM, et al: Predictors of opioid-related death during methadone therapy. J Subst Abuse Treat 2015; 57:30–35
27.
Saxon AJ, Ling W, Hillhouse M, et al: Buprenorphine/naloxone and methadone effects on laboratory indices of liver health: a randomized trial. Drug Alcohol Depend 2013; 128:71–76
28.
Weiss RD, Potter JS, Provost SE, et al: A multi-site, two-phase, Prescription Opioid Addiction Treatment Study (POATS): rationale, design, and methodology. Contemp Clin Trials 2010; 31:189–199
29.
Lee JD, Nunes EV, Novo P, et al: Comparative effectiveness of extended-release naltrexone versus buprenorphine-naloxone for opioid relapse prevention (X:BOT): a multicentre, open-label, randomised controlled trial. Lancet 2018; 391:309–318
30.
Sobell LC, Sobell MB: Timeline follow-back: a technique for assessing self-reported alcohol consumption, in Measuring Alcohol Consumption Psychosocial and Biochemical Methods. Edited by Litten RZ, Allen JP. Totowa, NJ, Humana Press, 1992, pp 41–72
31.
McLellan AT, Cacciola JS, Zanis D: The Addiction Severity Index–Lite. Philadelphia, Center for Studies of Addiction, University of Pennsylvania/Philadelphia VA Medical Center, 1997
32.
Moolchan ET, Radzius A, Epstein DH, et al: The Fagerstrom Test for Nicotine Dependence and the Diagnostic Interview Schedule: do they diagnose the same smokers? Addict Behav 2002; 27:101–113
33.
Landis JR, Koch GG: An application of hierarchical kappa-type statistics in the assessment of majority agreement among multiple observers. Biometrics 1977; 33:363–374
34.
Lee JD, Nunes EV, Novo P, et al: Commentary on Ajazi et al (2021) Re-analysis of the X:BOT Trial. J Addict Med 2022; 16:382–385
35.
Therneau T, Grambsch P: Modeling Survival Data: Extending the Cox Model. New York, Springer, 2000
36.
Fine JP, Gray RJ: A proportional hazards model for the subdistribution of a competing risk. J Am Stat Assoc 1999; 94:496–509
37.
Nunes EV, Gordon M, Friedmann PD, et al: Relapse to opioid use disorder after inpatient treatment: protective effect of injection naltrexone. J Subst Abuse Treat 2018; 85:49–55
38.
Sigmon SC, Bisaga A, Nunes EV, et al: Opioid detoxification and naltrexone induction strategies: recommendations for clinical practice. Am J Drug Alcohol Abuse 2012; 38:187–199
39.
Fischer B, Patra J, Cruz MF, et al: Comparing heroin users and prescription opioid users in a Canadian multi-site population of illicit opioid users. Drug Alcohol Rev 2008; 27:625–632
40.
Wu L-T, Woody GE, Yang C, et al: How do prescription opioid users differ from users of heroin or other drugs in psychopathology: results from the National Epidemiologic Survey on Alcohol and Related Conditions. J Addict Med 2011; 5:28–35
41.
Gladden RM, Martinez P, Seth P: Fentanyl law enforcement submissions and increases in synthetic opioid-involved overdose deaths—27 states, 2013–2014. MMWR Morb Mortal Wkly Rep 2016; 65:837–843
42.
Alkermes: Prescribing information: Vivitrol (naltrexone for extended-release injectable suspension). https://www.accessdata.fda.gov/drugsatfda_docs/label/2010/021897s015lbl.pdf
43.
Dunphy C, Peterson C, Zhang K, et al: Do out-of-pocket costs influence retention and adherence to medications for opioid use disorder? Drug Alcohol Depend 2021; 225:108784
44.
Hoffman KA, Baker R, Kunkel LE, et al: Barriers and facilitators to recruitment and enrollment of HIV-infected individuals with opioid use disorder in a clinical trial. BMC Health Serv Res 2019; 19:862
45.
Saloner B, Lê Cook B: Blacks and Hispanics are less likely than Whites to complete addiction treatment, largely due to socioeconomic factors. Health Aff 2013; 32:135–145
46.
Larochelle MR, Bernson D, Land T, et al: Medication for opioid use disorder after nonfatal opioid overdose and association with mortality: a cohort study. Ann Intern Med 2018; 169:137–145
47.
Ajazi EM, Dasgupta N, Marshall SW, et al: Revisiting the X:BOT naltrexone clinical trial using a comprehensive survival analysis. J Addict Med 2022; 16:440–445
48.
Shulman M, Hu MC, Nunes EV, et al: Initiating long-acting injectable naltrexone: procedure for a second attempt after a first attempt fails. J Addict Med 2022; 16:588–591
49.
Macleod J, Steer C, Tilling K, et al: Prescription of benzodiazepines, Z-drugs, and gabapentinoids and mortality risk in people receiving opioid agonist treatment: observational study based on the UK Clinical Practice Research Datalink and Office for National Statistics death records. PLoS Med 2019; 16:e1002965
50.
Jones CM, McAninch JK: Emergency department visits and overdose deaths from combined use of opioids and benzodiazepines. Am J Prev Med 2015; 49:493–501
51.
Liu S, O’Donnell J, Gladden RM, et al: Trends in nonfatal and fatal overdoses involving benzodiazepines—38 states and the District of Columbia, 2019–2020. MMWR Morb Mortal Wkly Rep 2021; 70:1136–1141
52.
Kelty E, Hulse G, Joyce D: A comparison of blood toxicology in fatalities involving alcohol and other drugs in patients with an opioid use disorder treated with methadone, buprenorphine, and implant naltrexone. Am J Drug Alcohol Abuse 2020; 46:241–250
53.
Gudin JA, Mogali S, Jones JD, et al: Risks, management, and monitoring of combination opioid, benzodiazepines, and/or alcohol use. Postgrad Med 2013; 125:115–130
54.
Connery HS, Taghian N, Kim J, et al: Suicidal motivations reported by opioid overdose survivors: a cross-sectional study of adults with opioid use disorder. Drug Alcohol Depend 2019; 205:107612
55.
Kelty E, Hulse G: Fatal and non-fatal opioid overdose in opioid dependent patients treated with methadone, buprenorphine or implant naltrexone. Int J Drug Policy 2017; 46:54–60
Information & Authors
Information
Published In
History
Received: 6 April 2022
Revision received: 24 May 2022
Accepted: 27 June 2022
Published online: 9 March 2023
Published in print: May 01, 2023
Keywords
Authors
Funding Information
Supported by NIDA grant 5UG1DA013035.Dr. Nunes has served as an investigator on studies funded by Braeburn-Camurus and BrainsWay; he has served as an unpaid consultant for Alkermes, Braeburn-Camurus, Invidior, and Pear Therapeutics; and studies on which he has been an investigator have received in-kind donations of medication or digital therapeutic applications from Alkermes, Braeburn, Chess Health, Indivior, and Pear Therapeutics. The other authors report no financial relationships with commercial interests.
Metrics & Citations
Metrics
Citations
Export Citations
If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download.
For more information or tips please see 'Downloading to a citation manager' in the Help menu.
View Options
View options
PDF/EPUB
View PDF/EPUBLogin options
Already a subscriber? Access your subscription through your login credentials or your institution for full access to this article.
Personal login Institutional Login Open Athens loginNot a subscriber?
PsychiatryOnline subscription options offer access to the DSM-5-TR® library, books, journals, CME, and patient resources. This all-in-one virtual library provides psychiatrists and mental health professionals with key resources for diagnosis, treatment, research, and professional development.
Need more help? PsychiatryOnline Customer Service may be reached by emailing [email protected] or by calling 800-368-5777 (in the U.S.) or 703-907-7322 (outside the U.S.).