Introduction

Opioid use disorder (OUD) is a chronic recurrent illness1 affecting an estimated 26.8 million people, with an increase in prevalence of over 47.3% over the past 20 years worldwide.2 Although opioid agonist treatment (OAT) is intended as a long-term, potentially indefinite treatment for persons with OUD, many persons may have a strong desire to eventually taper off treatment.3 OAT tapering involves a gradual reduction in daily OAT dosing to ultimately reach a state of opioid abstinence.4 Although neither British Columbia (BC) nor Canadian OUD treatment guidelines explicitly endorse tapering, they recommend additional psychosocial treatment, residential treatment and extended-release naltrexone among clients requesting a taper or not benefiting from OAT.1 5

Common reasons reported by people with lived experience to taper off OAT involve the increased risk of side effects from treatments (including drowsiness, cardiac arrhythmias, myocardial infarction6–8), expectations for recovery, pressure from family or peers (including concerns in mutual self-settings), inconvenience of going to pharmacy every day, financial or travel restrictions and the stigma of remaining in drug treatment.1 4 9–11 Qualitative studies assessing clients’ perspectives on OAT tapers have indicated that many people with OUD have a desire to eventually discontinue their OAT and achieve a state of abstinence from opioids. A qualitative study of 15 participants in Sweden indicated that clients’ perspectives on tapers are dependent on their duration and response to treatment.3 People in the first 2 years of treatment are less likely to consider tapering off OAT as cravings can be reduced or avoided by receiving an appropriate dose of OAT.3 During OAT, over 90% of participants in the study desired independence from their reliance on medications and the regimented lifestyle required in their OAT programmes, noting inconveniences caused by the environment at the treatment centre, limited programme opening hours, side effects of medications and a strong desire of being drug-free.3 12 13 According to a cross-sectional study that surveyed 145 people receiving OAT at public clinics in Sydney, Australia, 62% of participants expressed a strong desire to gradually taper off OAT over the coming 6 months.14 Additionally, 15% expressed moderate interest, whereas only 12% were not inclined to come off the treatment.14

Evidence regarding the optimal time to initiate a taper for clients who desire to achieve drug abstinence is limited and likely outdated for many settings, as it was exclusively generated prior to the introduction of fentanyl into the illicit drug supply. First, sustained abstinence following OAT taper completion is uncommon, with the risk of relapse being particularly high for clients who discontinue treatment within the first year.15 To this end, three clinical trials reported higher relapse rates after completing methadone taper, ranging from 53.1% to 66.7% of individuals relapsed by the first month and 61.1% to 89.2% by the sixth month after taper completion.1 16–18 Furthermore, the ideal duration of stability before a taper is initiated may be related to clients’ clinical course, severity, adherence to OAT treatment and other social factors such as family support and financial status, housing status and neighbourhood environment.1 4 19–23 A retrospective cohort study with 8996 clients who discontinued buprenorphine treatment found that individuals who were retained in treatment for 15–18 months had a lower probability of emergency department visits, hospitalisations and filling opioid prescriptions in the 6 months following discontinuation than individuals retained for 6–9 months.24 Similarly, a 12-year retrospective cohort study of 1308 individuals with buprenorphine found that 3.67% were observed to taper off buprenorphine, and of those who tapered, only 14% of individuals were able to successfully complete tapering within a follow-up period of 316 days.25

Several studies have demonstrated that initiating tapers after longer treatment durations is associated with a greater likelihood of taper completion. A retrospective study involving 853 individuals desiring to taper off methadone in Guangdong, China, showed that initiating tapers after 52 weeks increased the odds of taper completion.26 However, another retrospective study involving 14 602 individuals initiating methadone tapering in BC reported that the duration of methadone treatment had no independent effect on taper completion after adjusting for other covariates whereas taper duration was predictive.4

Specifically, besides identifying the optimal time to initiate tapering, tapering strategies related to the rate of OAT dosage reduction and concurrent treatments are known to influence the likelihood of successfully completing an OAT taper. A randomised controlled trial (RCT) of 63 participants with methadone maintenance treatment observed that over 30 weeks, tapering off methadone maintenance treatment with a weekly 3% reduction rate of the initial dose is more likely to achieve taper completion than a rapid strategy of 10% dose reduction per week (53% vs 24%).27 Studies from BC, Canada, and Guangdong, China, both demonstrated that more gradual and staggered dose decreases increased the likelihood of taper completion.4 26 A retrospective cohort study of 5774 individuals undergoing a buprenorphine taper reported that regardless of the tapering strategy, however, the time to taper initiation demonstrated an independent effect on the likelihood of successful taper completion,20 in contrast to the prior study from BC involving methadone.4 The independent effect of when the OAT taper is initiated or the possible effects of OAT medication type in the treatment course thus remains unclear.

One of the most common concerns among clients about tapering off OAT is a fear of withdrawal symptoms, whereas care providers generally fear a risk of relapse and opioid overdose. Concerns associated with a decreasing medication dose include adverse physical and psychological events such as persistent sleep disturbances, drug cravings, anxiety, myalgia, irritability, tremors, pupillary dilatation and fatigue, thus increasing the risk of relapse28–31 Tapering off OAT before achieving stabilisation, sometimes referred to as short-term detoxification treatment, increases the risk of overdose and can lead to a negative effect on the client’s health.31–34

While both rapid detoxification and tapering after a stabilisation period are techniques for opioid withdrawal management, they differ significantly in their timing and underlying philosophies. BC clinical guidelines for OUD define rapid detoxification as inpatient withdrawal management to detoxify individuals with opioids with a higher rate of completion of withdrawal.1 Compared with rapid detoxification, gradual tapering is associated with a higher rate of successful completion and fewer withdrawal symptoms.1 35 An RCT of 53 participants, aged 16–24 years, found that individuals on a 56-day buprenorphine taper (from medication induction to taper completion) had a substantially greater percentage of opioid-negative planned urine tests (35% vs 17%, p=0.039; Cohen’s d=0.57, 95% CI=0.02 to 1.13) and better treatment retention (37.5 vs 26.4 days, p=0.027; Cohen’s d=0.63, 95% CI=0.06 to 1.19) than those on a 28-day buprenorphine taper, however, outcomes in both arms were poor.36

Relatively little guidance, or supporting evidence, is provided on the appropriate time to initiate a taper in clinical best practices guidelines internationally (table 1). Of 11 national clinical guidelines on the clinical management of OUD in North America, Australia and UK, all recommend long-term OAT with no explicit endpoint in the treatment.1 5 15 21 37–46 Five of them, including BC, CAMH, ASAM, Queensland and New South Wales’ guidelines, recommend clients stay in OAT for at least 1 year prior to attempting a taper,1 37 39 40 45 46 however, no evidence was cited in support of this recommendation. Only the Australian national guideline cited four references in support of their recommendation; however, these references provided only indirect evidence to this end, demonstrating the benefits of longer treatment durations rather than explicitly comparing times to taper initiation among those wanting to taper.15 47–50

Table 1

Taper guidelines comparison for OAT

Taken together, the evidence used to guide strategies for transitioning off OAT among patients who desire to taper their medication is thus extremely limited. There is insufficient evidence assessing the feasibility of discontinuing OAT medications for stable patients without causing relapse.29 Thus far, the published literature has not described characteristics associated with an elective taper from any form of OAT among clients stabilised in treatment over an extended timeframe. As such, we will conduct a population-level retrospective observational study with the eight-linked provincial administrative database and aim to determine the safety and comparative effectiveness of alternative times from completed OAT induction at which a taper could be contemplated to maximise the likelihood of taper completion and minimise the risk of overdose mortality, as observed in clinical practice in BC, Canada.

MethodsStudy design

We aim to execute a population-level retrospective study using a linkage of eight provincial health administrative databases in BC, Canada, from 01 January 2010 to 17 March 2020. We propose a per-protocol study design among incident (first-time) users and prevalent new users (including repeated attempts) of OAT. Table 2 summarises the key elements of the proposed study design.

Table 2

Key components of the emulated target trial on OAT taper using the observational data

We will use the BC PharmaNet database51 (prescription dispensation records) to construct OAT episodes and identify tapering throughout the study period. Only dispensations provided to individuals in federal correctional facilities and during hospitalisation are not included in this database. This database is linked with data from seven population-level administrative databases, including the Discharge Abstract Database52 (hospitalisations records), Medical Services Plan53 (physician billing records), NACRS Database54 (emergency department visits), BC Provincial Corrections55 (imprisonment and release), Vital Statistics56 (death records and their underlying causes), Perinatal Care Database57 (maternal and infant healthcare) and BC Social Development and Poverty Reduction database58 (records of social supports). A de-identified personal health number is used to link these databases and is unique to each individual residing in BC.59 A summary of the data extraction timeline from the various linked databases can be found in online supplemental figure A1.

Study population

The study will include individuals aged 18 years or older without any diagnosis of cancer or history of receiving palliative care, who initiated OAT with either methadone or buprenorphine/naloxone from 01 January 2010 to 17 March 2020 and then completed induction before 17 March 2020. People who are incarcerated and women who are pregnant will be excluded from the analysis.

Study follow-up

Time zero is defined as the date of completed induction, which we operationalised as the date of reaching a minimum of 2 weeks of continuous OAT with no dose changes.60 The incident user cohort will include individuals engaged in OAT for the first time during the study period, without any prior OAT dispensations dating back to 01 January 1996. All analyses will be stratified according to the medication received at completed induction. Due to limited observational data, episodes initiated with slow-release oral morphine or injectable OAT will be excluded. Study follow-up will conclude at death, OAT dropout, pregnancy, receipt of cancer or palliative care, incarceration, out of province or the end of the study follow-up (17 March 2020), whichever occurs first.

Key measures

The primary exposure and outcome are not indicated directly within the linked databases and are therefore empirically derived. The primary exposure will be the time from completed induction (reaching a minimum of 2 weeks of continuous OAT with no dose changes) to taper initiation within a continuous treatment episode (no breaks in dispensed doses >5/6 days for methadone or buprenorphine–naloxone, respectively), defined as the difference between the date of completed induction and the date of taper initiation. Taper initiation is defined as the date of the second decrease in a client’s daily OAT dose after completed induction, without increases in the interim. Two consecutive daily dose decreases are used to distinguish taper starts from dose adjustments. The 5/6 day interval for treatment discontinuation was chosen to align with BC guidelines recommending the adjustment of starting doses after non-compliance durations to ensure safety.61 We will aim to compare different times from completed induction to taper initiation, including less than 3 months, 3–6 months, 6–12 months, 12–24 months and 24–48 months, pending sufficient statistical power. All analyses will be stratified by medication—including methadone and buprenorphine/naloxone. The primary outcomes will include completed taper and all-cause mortality during treatment. A completed taper is defined as reaching a final daily dose of ≤5 mg/day for methadone or ≤2 mg/day for buprenorphine–naloxone, with no subsequent OAT dispensation for at least 5 days for methadone or 6 days for buprenorphine/naloxone. Daily doses higher than these thresholds on the last dispensation are thus considered discontinuation prior to completing the taper. All-cause mortality is observed within 5/6 days of treatment discontinuation via linkage to Vital Statistics records.56 As our data do not identify OAT receipt in inpatient settings, we assumed individuals who began OAT prior to admission continued treatment during their hospitalisation. As both exposure and primary outcome were empirically derived, we propose sensitivity analysis on each to ensure results are robust to misclassification.

Analysis plan

We propose to conduct a per-protocol analysis to evaluate the comparative effectiveness of the time of OAT tapering initiation. We will present the effect of the taper initiation time on our outcomes using adjusted ORs from logistic regression models for completed taper and adjusted HRs from pooled logistic regression models for all-cause mortality.

Per-protocol approach

The per-protocol approach examines the treatment effect for individuals who follow the treatment they are assigned. To avoid possible selection bias, we require a time zero that is equal for all individuals, which will be the date of completed induction. However, as the time of taper initiation is not known at time zero, using future information to define exposure groups would introduce immortal time bias. Thus, we propose to use a clone-censoring-weight approach where replicates are made of individuals for each tapering strategy. Figure 1 illustrates time zero and comparison groups of the different OAT stabilised periods in the analysis. Strategies ‘a’ and ‘b’ denote two exposure strategies with the same time zero and different time points to initiate taper. ‘Taper initiation a’ and ‘taper initiation b’ indicate that tapering can be initiated at different time points following completed induction to exclude rapid detoxification episodes. We propose three censoring models, including one for protocol deviation (different time to taper initiation), one for OAT dropout and another for eligibility criteria violation (medication switch prior to taper initiation, death, pregnancy, out of province, receipt of palliative care or cancer and incarceration).

Figure 1

The time zero and different OAT stabilised comparison groups in the per-protocol analysis. IPCW estimations for deviation of the protocol will be calculated during the OAT-stabilised period up to the point of deviation. IPCW, inverse probability of censoring weights; OAT, opioids agonist treatment; t0, time zero; ta,b, taper completion; a, b: two cases.

Clone-censoring-weighting approach

The clone-censoring-weighting approach is an effective statistical method that avoids the immortal time and selection bias. The clone-censor-weighting approach replicates individuals to each different possible exposure group at time zero, which removes the possibility of immortal time bias over treatment assignment,62 63 and can also control for time-varying confounding effects that may influence the outcome. Individual replicates are censored when they no longer follow their assigned strategy (the time at which the taper initiated) or they drop out before initiating a taper. Informative bias will be introduced by artificial censoring and will be remediated by weighting the remaining clients by the inverse probability of being censored.64 Figure 2 uses two examples to illustrate the clone-censor approach in per-protocol analysis. In panel A, if an individual initiated tapering in the 30th month after time zero (completed induction), the individual will be censored for other tapering strategies. Panel B shows an individual initiating tapering in the 10th month. The individual will be censored for other strategies at the end of each time interval and censored at the same time with actual taper initiation for the strategy with a longer time interval. Per-protocol analysis requires censoring weights to control for the informative censoring imposed by the design; and inverse probability-of-censoring weights will be used to deal with the informative censoring.62 65

Figure 2

Clone-censor approach for per-protocol analysis. In panel A, if an individual initiates a taper in the 30th month after time zero (completed induction), he/she will be censored for other taper strategies. Panel B shows an individual initiating tapering in the 10th month. The individual will be censored for other strategies at the end of each time interval and censored at the same time with actual taper initiation for the strategy with a longer time interval.

Pooled logistic regression model

A pooled logistic regression model with censorship weights and robust variances will be implemented to adjust for time-dependent confounders at baseline (time zero) which affect both the time of taper initiation and the primary outcome.66 67 Robust variance estimates will be used to control for the cloned replicates and, in the prevalent new user design, repeated treatment attempts. A pooled logistic model will be applied for the time-to-event outcome of all-cause mortality. As mortality rates will be less than 10% of individuals at each time point, the risk ratio will approximate the HR.

Covariate selection

In observational contexts, the assumption of no uncontrolled confounding cannot be validated; however, we will account for all possible confounding variables accessible in our linked database.68 We executed a systematic literature review for articles published up to 3 March 2023 to identify potential confounding variables related to OAT completed tapers. The following search query string was used in PubMed: (‘opiate substitution treatment’ (Mesh) OR ‘opioid agonist treatment’ (Title/Abstract) OR ‘buprenorphine’ (Mesh) OR ‘methadone’ (Mesh)) AND (‘Inactivation, Metabolic’ (Mesh) OR completed taper (Title/Abstract) OR succeed taper (Title/Abstract) OR detox (Title/Abstract) OR detoxification (Title/Abstract) OR withdrawal management (Title/Abstract)) AND (factors (Title/Abstract) OR predictor (Title/Abstract) OR determinants (Title/Abstract)). A total of 110 articles were found by this query, and 9 were included in online supplemental table A1. We have reported factors related to completed OAT tapers by performing a systematic literature review for articles published until 22 April 2023. We propose to employ covariates of the systematic review for statistical analysis and follow these factors from our linked administrative data that were expected to impact both OAT taper initiation and taper completion. We propose to enhance this inventory by incorporating additional variables in our administrative database with the list of variables from the systematic review, which we hypothesise to affect both the time to initiate tapering and completed tapers. Therefore, only time-varying covariates measured prior to taper initiation will be updated on a weekly basis (online supplemental figure A2).

Subgroup and sensitivity analysis

We propose to perform subgroup and sensitivity analyses to evaluate the robustness of results and heterogeneity of treatment effects among key subgroups. We establish predetermined objectives that concentrate on limiting the study population, restricting the time frame of the study, categorising variables and specifying the model used in the analysis (table 3). Relevant findings from each analytic approach will be displayed in a forest plot. Any deviations from this protocol will be noted as such in final reports.

Table 3

Proposed subgroup and sensitivity analyses

Ethics and dissemination

BC Ministries of Health and Mental Health and Addiction provided the research team with access to this connected database as a response to the opioid overdose public health crisis in the province. The study was categorised as a quality improvement initiative. The analysis conducted by the Providence Health Care Research Institute and the Simon Fraser University Office of Research Ethics met the criteria for exemption as stated in Article 2.5 of the Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans.69

This study will adhere to international guidelines for conducting and reporting research, which includes Strengthening the Reporting of Observational Studies in Epidemiology.70 Additionally, a multidisciplinary scientific advisory committee will use the ‘Risk of Bias in Non-Randomised Studies-of Interventions’ tool for ex-post evaluation. The findings of the study will be electronically and physically published in peer-reviewed academic journals. The study will produce robust evidence regarding the effectiveness of different time intervals between OAT initiation and OAT tapers in real-world settings over the long term in the interest of enhancing OAT retention71 for these essential and life-saving72 medications.

Patient and public involvement

While there was no explicit involvement of OAT clients in the design of this study, the prioritisation of this analysis relative to other clinical questions was based on qualitative feedback from people with opioid use disorders received on this and other related objectives specified in the parent grant (R01DA050629). The results will be shared with local advocacy organisations representing people who use drugs and people who have received OAT, with their consultation. The potential impact on client engagement was a significant factor in this decision.