Comparative effectiveness of maintenance doses of opioid agonist treatment among individuals with opioid use disorder: a target trial emulation protocol using a population-based observational study

Abstract

Introduction

Opioid agonist treatment (OAT) prescribing patterns have shifted in recent years in British Columbia (BC), Canada due to the increasingly toxic unregulated drug supply. Experimental evidence to support guidelines on the effectiveness of maintaining clients at different maintenance dosage levels is incomplete and outdated for the fentanyl era. Our objective is to assess the risk of treatment discontinuation and mortality among individuals receiving different maintenance dosage strategies for OAT with methadone, buprenorphine/naloxone or slow-release oral morphine (SROM) at the population level in BC, Canada.

Methods and analysis

We propose a retrospective population-level study of BC residents initiating OAT on methadone, buprenorphine/naloxone or SROM between 1 January 2010 and 31 December 2022 who were ≥18 years of age with no known pregnancy, no history of cancer diagnosis or receiving palliative care and not currently incarcerated. Our study will employ health administrative databases linked at the individual level to emulate a target trial per OAT type where individuals will be assigned to discrete maintenance dosing strategies, according to the full range observed in BC during the study period. Primary outcomes include treatment discontinuation and all-cause mortality. To determine the effectiveness of alternative maintenance doses, we will emulate a ‘per-protocol’ trial using a clone-censor-weight approach to adjust for measured time-dependent confounding by indication.

Ethics and dissemination

The protocol, cohort creation and analysis plan have been classified and approved as a quality improvement initiative by Providence Health Care Research Ethics Board and the Simon Fraser University Office of Research Ethics. All data are deidentified, securely stored and accessed in accordance with provincial privacy regulations. Results will be disseminated and shared with local advocacy groups and decision-makers, developers of national and international clinical guidelines, presented at national and international conferences and published in peer-reviewed journals electronically and in print.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• Comprehensive database with individual-level linkage under universal healthcare settings represents an ideal setting for direct comparison of opioid agonist treatment maintenance dose strategies at the population level and within key subgroups.

• A per-protocol analysis, using a clone-censor-weight approach, will provide a direct comparison of maintenance dose strategies administered in clinical practice in the past 12 years in British Columbia, Canada for the treatment of opioid use disorder.

• A range of sensitivity analyses will be executed to evaluate the robustness of the results.

Introduction

The goal of opioid agonist treatment (OAT) in the treatment of opioid use disorder (OUD) is to reduce opioid use and related harms.¹ An appropriate maintenance dose aims to control withdrawal symptoms and reduce cravings for opioids while avoiding intolerable and potentially deadly side effects. Ultimately, an individual’s daily OAT dosage should be personalised based on their tolerance and treatment goals.² Methadone and buprenorphine/naloxone are recommended OAT medications internationally,^1–14 while slow-release oral morphine (SROM) is restricted to select European jurisdictions and has only been recommended for the treatment of OUD in Canada since 2017.^1–4 In Canada, OAT is offered through specialised addiction treatment centres, community health centres, office-based settings and pharmacies, similar to the UK, Australia and New Zealand. However, in the USA, access to methadone is primarily restricted to federally regulated opioid treatment programmes.^{6 7}

Methadone’s efficacy as a therapeutic treatment for OUD is the most established among the various forms of OAT.¹⁵ Methadone is a full μ-opioid receptor agonist, which increases the risk for overdose during titration when used at doses above an individual’s tolerance or when combined with other central nervous depressants.¹⁶ Methadone has no maximum dose and can be titrated upwards as long as the patient’s corrected QT interval remains within normal limits. In contrast, buprenorphine, a partial μ-opioid receptor agonist with a ceiling effect with respect to respiratory depression (ie, increases past 24 mg do not increase the effects of respiratory depression), has a milder side effect profile, milder withdrawal symptoms and fewer drug interactions but can precipitate acute withdrawal in individuals using other opioids.¹⁶ Buprenorphine can be combined with naloxone, an opioid antagonist that blocks the effect of other opioids if injected but has poor oral absorption. This combination can be used to reduce the risk of misuse of the medication by injection and to prevent diversion. SROM is a long-acting, 24-hour formulation of oral morphine that is administered once daily and, like methadone, does not have a ceiling effect.² All three medications have demonstrated effectiveness in reducing the risk of overdose, morbidity and all-cause mortality among people with OUD.^17–23 However, the unregulated drug supply has changed dramatically since much of the evidence used to support current OAT dosing guidelines was produced.

Since the first overdose death involving fentanyl in BC was detected in 2012, its use has grown rapidly; fentanyl or its analogues have been detected in over 80% of all unregulated drug deaths each year since 2017.²⁴ Fentanyl and its analogues are estimated to be 40 times more potent than heroin²⁵ and 100–10 000 times more potent than morphine.^{25 26} The management of OUD with OAT is becoming increasingly difficult due to the widespread use of these high potency opioids.^{26 27} While research on effectiveness of OAT for people who use fentanyl is limited, case reports suggest higher maintenance dosing may be required to adequately manage cravings and withdrawal.^{28 29} Population-level studies have demonstrated that being on OAT provides substantial protection from fentanyl-related overdose,¹⁷ though retention in treatment has declined in the fentanyl era.³⁰ Since the introduction of fentanyl, daily dosing guidelines for OAT have not changed as a result of the dearth of evidence to support such adjustments. In BC, from 2014 to 2021, prescribers of OAT initiated both new and experienced OAT clients at higher doses than guideline recommendations, used combination therapies, titrated them more rapidly and maintained clients at higher doses,³¹ suggesting prescribers have begun to adapt to the needs of their clients based on changes in the unregulated drug supply regardless of clinical guidelines. While fentanyl and its analogues are detected in the majority of overdose deaths in Canada³² and roughly half of overdose deaths in the USA,³³ to date, the substance has not infiltrated unregulated markets in Europe, Australia and New Zealand to a comparable extent.^{34 35}

Canadian and international OAT guidelines

We summarise the recommendations for OAT maintenance dosing from a convenience sample of four Canadian and nine international guidelines for the clinical management of OUD^1–14 in table 1. All guidelines advise caution in prescribing towards and past the higher end of the recommended dosing range and to reserve these doses for individuals with very high known tolerance and very low risk of toxicity, often characterised by previous OAT experience and/or concurrent fentanyl use.^{2 6}

Table 1.

Comparison of maintenance dose guidelines for opioid agonist treatment*

	Recommended maintenance dose range for OAT (mg/day)
Guideline	Methadone	Buprenorphine/naloxone
BC, CAN (2017)	60–120 mg45 48 though higher doses may be required.51 67	12 mg/3 mg–16 mg/4mg,20 up to maximum of 24 mg/6 mg.96
BC, CAN (2023)	60–120 mg45 48 or higher51 67 though >150 mg may be required based on clinical experience in the fentanyl era.	24 mg/6 mg,96 up to maximum of 32 mg/8 mg to address high tolerance due to fentanyl and other highly potent synthetic opioids.97
CRISM, CAN (2018)	≥80 mg though >120 mg may be required for full opioid blockade and suppression of withdrawal.45 48 51 67 98	8–12 mg up to maximum of 24 mg of buprenorphine
CAMH, CAN (2021)	Clinical judgement	Clinical judgement to maintain optimal individualised daily dose, up to maximum of 24 mg buprenorphine. If >24 mg, inform patient of departure from approved doses and of the limited evidence of added benefit and possible increased risk for adverse events.
META-PHI, CAN (2021)*	≥100 mg for individuals using fentanyl daily, noting limited research on the effectiveness of doses >100 mg.49 50 99–101 Guideline focus group stated 120 mg minimum to reduce fentanyl use.	N/A
SAMHSA, USA (2021)	>60 mg, 80–120 mg102 is the typical range though wide variation.103	4 mg/1 mg–24 mg/6 mg* with target of 16 mg/4 mg
ASAM, USA (2020)	60–120 mg though some require higher or lower doses. Doses >120 mg not well studied but may be required due to fentanyl use.65	16 mg–24 mg of buprenorphine There is limited evidence regarding the relative efficacy of doses >24 mg, and the use of higher doses may increase the risk of diversion.
CSAM, California, USA (2019)	80–120 mg (or higher) Doses above 180 mg are not the norm, but sometimes appropriate.	4 mg/1 mg–24 mg/6 mg* with target of 16 mg/4 mg Doses >24 mg buprenorphine or 24 mg/6 mg buprenorphine/naloxone show no clinical advantage.
UK (2017)	60–120 mg45 though some require higher or lower doses. Lower doses may extinguish withdrawal symptoms but they may need higher dose to minimise craving.	12–16 mg, up to 32 mg of buprenorphine though some require higher or lower doses. Lower doses may extinguish withdrawal symptoms but they may need higher dose to minimise craving.
NICE CKS, UK (revised 2022)	60–120 mg	12–16 mg, up to 32 mg of buprenorphine
Australia (2014)	60–120 mg45 Based on clinical experience, doses >150 mg are generally associated with little additional benefit and increase the risk of adverse events.	12–24 mg, up to maximum of 32 mg104–107 or lower (4–8 mg) of buprenorphine Doses >16 mg are associated with increased duration of action, with little or no increase in opioid effect.107
New South Wales, AUS (2018)	60–100 mg Some require higher but doses >150 mg are generally associated with little additional benefit and the increase risk of adverse events.	12–24 mg, up to 32 mg or lower (4–8 mg) of buprenorphine 32 mg is the maximum recommended daily dose of buprenorphine; >32 mg is beyond the licensed limit.
Western Australia, AUS (2014)	60–100 mg > 100 mg may be required, but there is no evidence from outcome studies that >120 mg results in added benefit for most individuals.	12–20 mg of buprenorphine
New Zealand (2014)	60–120 mg	12–24 mg of buprenorphine

BC, CAN (2017); CRSIM, CAN (2018); CAMH, CAN (2021) provide no explicit guidance for maintenance dosing but state the average (mean) SROM dose prescribed ranges from 235 to 791 mg and the full range reported in the literature is 60–1200 mg. BC, CAN (2023) provides no explicit guidance but states the highest dose described in the literature is 1200 mg, and clinical experience indicates individuals often require doses >1200 mg due to high tolerance developed among those who use fentanyl.

*SROM is only licensed for the treatment of opioid use disorder in Canada.

ASAM, American Society of Addiction Medicine; BC, British Columbia; CAMH, Centre for Addiction and Mental Health; CKS, Clinical Knowledge Summaries; CRISM, Canadian Research Initiative on Substance Misuse; CSAM, California Society of Addiction Medicine; META-PHI, Mentoring, Education, and Clinical Tools for Addiction: Partners in Health Integration; N/A, not available; NICE, National Institute for Health and Care Excellence; OAT, opioid agonist treatment; SAMHSA, Substance Abuse and Mental Health Services Administration; SROM, slow-release oral morphine.

The 2017 BC guidelines suggested a typical therapeutic window for methadone of 60–120 mg/day, noting that sometimes doses over 120 mg may be necessary and that treatment outcomes are often better at higher doses.¹ The 2023 BC guidelines suggest that doses of 150 mg or higher may be required in patients who are dependent on highly potent opioids such as fentanyl, though recognise that this recommendation is based on clinical experience and that data supporting best practices is lacking.² Other guidelines from Canada (2018, 2021), USA (2019–2021), UK (2017, 2022), Australia (2014, 2018) and New Zealand (2014) commonly endorse the maintenance dose range of 60 mg–120 mg/day,3 4 7 9–11 14 with variation in specific recommendations. Many guidelines cite a lack of evidence on the effectiveness of doses above 100–120 mg/day5 6 11 12 14 or note that doses above 150 mg/day are generally associated with little additional benefit and increase the risk of adverse events such as overdose, suspected drug interactions or allergic reactions.^{11 13} Only one clinical guideline, produced in 2021 from Ontario, Canada, specifically focuses on the prescription of methadone for people who use fentanyl and emphasises that maintenance doses should be above 120 mg if needed.⁵

The 2017 BC guidelines advise a maintenance dose range between 12 mg/3 mg and 16 mg/4mg for buprenorphine/naloxone, up to a maximum of 24 mg/6 mg per day, with explicit justification required for higher doses.¹ The 2023 update increased this maximum to 32 mg/8 mg to address the high tolerance of individuals as a result of the high prevalence of fentanyl in the unregulated drug supply.² Guidelines from the UK (2017, 2022),^{9 10} Australia (2014, 2018)^11–13 and New Zealand (2014) generally align with BC’s range and maximum dosage. Other guidelines from Canada and the USA advise maintenance dose ranges of buprenorphine/naloxone between 4 mg/1 mg and 24 mg/6 mg, noting a target of 16 mg/4 mg due to the limited evidence of additional benefit with doses above 24 mg of buprenorphine,^{4 6 8} and possible increased risk for adverse events⁴ or diversion when not combined with naloxone.⁶

Finally, Canadian guidelines state the average dose ranges for SROM observed in the literature and do not provide explicit recommendations for maintenance dosages.^1–4 Based on clinical experience, the BC 2023 guidelines indicate that individuals may require doses above 1200 mg due to high tolerance developed as a result of fentanyl entering the unregulated drug supply and advise to use caution for side effects when prescribing higher doses (eg, above 1200 mg) and to clearly document the rationale.²

Impact of COVID-19 on policies for OAT provision

The COVID-19 pandemic prompted providers and policymakers to make changes to maintain accessibility to treatment while attempting to limit viral spread. Countries including Canada,^{36 37} the USA,^37–40 Australia,⁴¹ New Zealand⁴² and UK⁴³ relaxed guidelines to increase the accessibility of take-home doses for clients who did not previously meet criteria, increase the use of telehealth and medication delivery and decrease requirements for urine-drug testing. There were no reported changes to maintenance dosing guidelines due to the COVID-19 pandemic. While positive outcomes, such as increased patient satisfaction and treatment retention, have been reported by both clients and providers,^{36 39} the length of time COVID-era guidance was in effect varies based on the setting and service providers.^{40 44}

Available evidence on OAT maintenance dosing

Evidence cited in maintenance dosing guidelines for OAT consistently demonstrates a dose–response relationship with retention in treatment^{20 45–50} and suppression of unregulated drug use with higher daily doses of both methadone45 46 48–50 and buprenorphine.^{20 46 47} However, all evidence referenced in maintenance dosing guidelines predates the emergence of fentanyl in the unregulated drug supply.20 45–48 51

A widely cited 2003 systematic review of randomised trials and prospective studies found that higher daily doses of methadone were associated with longer retention in treatment, reduced unregulated opioid use and lower withdrawal symptoms. The results were valid up to 100 mg/day, after which the reliability of evidence was deemed lower.⁴⁵ Notably, the trials included in this review used fixed dosing, which does not reflect current clinical guidelines that call for patient-specific dose adjustments to eliminate withdrawal symptoms.² Though over 20 years old, Faggiano et al’s systematic review remains among the most highly cited articles to establish dosing guidance for methadone.1 2 9 11 Since this review, a 2009 population-level study from BC and a 2013 multisite randomised trial in the USA found that daily methadone doses above 120 mg were associated with longer retention in treatment.^{46 48}

A 2014 meta-analysis of randomised trials confirmed buprenorphine’s efficacy in retaining individuals at doses above 2 mg and suppressing unregulated opioid use at doses higher than 16 mg²⁰ but indicated an absence of evidence for the effectiveness of buprenorphine daily doses greater than 24 mg. Notably, generalisability is limited by the substantial heterogeneity across studies included in the meta-analysis with regard to participant inclusion and exclusion criteria, disease severity, study design, dosing protocols, observation times and how retention was measured.⁵² The claim of a lack of evidence to support buprenorphine daily doses above 24 mg has been challenged by an increasing body of peer-reviewed evidence supporting dose-dependent benefits up to at least 32 mg.⁵³ A randomised trial conducted with individuals enrolled in opioid treatment programmes in the USA has since demonstrated individuals receiving buprenorphine at 30–32 mg/day achieved longer retention and lower unregulated opiate use than those at lower daily doses.⁴⁶

Given the use of SROM for the treatment of OUD is limited to several European jurisdictions and Canada, the evidence base to support maintenance dosing for SROM is extremely limited in size and methodological quality relative to that of methadone and buprenorphine/naloxone.^{22 54–56} Though the relative effectiveness of different maintenance doses of SROM has not been established, the average daily SROM dose presented in the literature, recent to 2014, ranges from 235 mg/day to 791 mg/day and the full range of SROM daily doses described is 60–1200 mg.^{22 23 57} The present analysis presents a critical opportunity to generate evidence to support dosing decisions for SROM. More recently published evidence not cited by guidelines demonstrates a dose–response relationship of methadone^{58 59} and buprenorphine⁵⁸ with treatment retention, though are limited in sample size⁵⁸ or used data collected before fentanyl dominated the unregulated supply in BC in 2016.⁵⁹

Most of these studies did not investigate the effect of OAT maintenance dosages on all-cause mortality22 47 48 58 59 or could not sufficiently generate data to determine differential effects on mortality20 22 23 45 due to small sample sizes and truncated follow-up of randomised trials. This also limits their ability to capture longer periods of retention. Importantly, it may be deemed unethical to conduct a randomised control trial to investigate the effects of fixed dosing strategies as this would not be responsive to patient goals, needs, withdrawal symptoms or risks of harm. Using linked administrative health data, population-level studies can address these challenges to generate evidence-supporting maintenance dosing strategies according to real-world practice.^{60 61}

Maintenance dosing experience for people with lived experience

Qualitative studies assessing client and provider perspectives on OAT dosage suggest that clients may not be receiving adequate doses,^{5 62–64} with stigma being a commonly cited factor influencing treatment decisions among both providers and clients.^62–64 A qualitative study with methadone clients in Ontario, Canada found that stigma from friends, family and providers and a lack of awareness of methadone treatment among providers contributed to clients lowering their dose before they are ready or providers failing to prescribe high enough doses.⁶² Another study from Ukraine found that concerns among clients and prescribers associated with being on OAT for an extended period or prescribing higher OAT doses—especially above 120 mg of methadone—often led to fears of being accused of drug-seeking behaviour or of enabling drug use. These concerns often resulted in reduced retention and undermedication.⁶³ Participants of a focus group in Ontario conducted specifically to formulate methadone guidance for people who use fentanyl expressed that 120 mg of methadone is the minimum dose required to reduce fentanyl use.⁵ These first-hand accounts emphasise the need for up-to-date evidence to support treatment decisions for providers and clients and reduce stigmatisation.

Overall, the level of evidence to support OAT maintenance dosing guidelines is highly variable and most studies cited by clinical guidelines were conducted prior to the introduction of fentanyl and its analogues into the unregulated drug supply.^{25 26} Additionally, randomised trials and prospective studies are limited in their capacity to evaluate the effectiveness of OAT maintenance dosages on treatment retention and mortality. Considering the paucity of evidence available to guide dosing decisions in the fentanyl era, clinical experience and opinion^65–67 is heavily relied on in guideline development, highlighting the urgent need to establish recent, real-world evidence to inform clinical guidelines and support clinical decision-making. Therefore, the objective of this study is to determine the comparative effectiveness of different maintenance dosage strategies of OAT on treatment discontinuation and all-cause mortality among OAT clients in BC, Canada, 2010–2022.

Methods

Study population and data sources

We propose a protocol for a population-based retrospective observational study of all individuals receiving OAT (methadone, buprenorphine/naloxone or SROM) in BC from 1 January 2010 to 31 December 2022 (or latest available data at the time of analysis), aimed at generating real-world evidence. The population will be defined using Drug Identification Numbers specific to OAT medications in PharmaNet,⁶⁸ linked at the individual level with population-level administrative databases, including the Medical Services Plan (capturing physician billing records),⁶⁸ client roster (demographic and geographic information for the Ministry’s clients),⁶⁹ the Discharge Abstract Database (hospitalisations),⁷⁰ the National Ambulatory Care Reporting System database (emergency department visits),⁷¹ the Perinatal database (maternal and neonatal health for provincial births),⁷² Vital Statistics (death and their underlying causes),⁷³ BC Corrections (incarceration in provincial prisons)⁷⁴ and Social Development and Poverty Reduction (receipt of housing and income assistance).⁷⁵ The databases are linked using unique deidentified personal health numbers by the BC Ministry of Health.

Target trial emulation

We will emulate three target trials to compare the maximum maintenance doses of (1) methadone (2) buprenorphine/naloxone and (3) SROM as summarised in table 2. Emulating a target trial by imposing a maximum daily dosage may not pass through an institutional review board given recommendations of most guidelines to titrate dosage as needed to eliminate withdrawal symptoms. This represents a critical advantage of using observational data to emulate a target trial, which may not be feasible or ethical. Time zero will be set as the date of first OAT dispensation from a community-based pharmacy. BC residents who are 18 years of age or older with no history of cancer or palliative care, no current incarceration and with no known pregnancy (up to 3 months of postpartum) at OAT initiation will be included in the study. Pregnant individuals and those in palliative care will be excluded due to unique biological and clinical considerations that require tailored dosing strategies, which differ significantly from the broader OAT population.^76–79 We will consider both incident users (first OAT receipt since 1 January 1996) and prevalent new users (incident users and those with prior attempts at OAT). Treatment episodes initiated with multiple forms of the previously mentioned OAT prescribed concurrently will be excluded.

Table 2.

Key design components of the proposed emulated target trials on maintenance dose of opioid agonist treatment

Component	Target trial	Emulation using observational data
Eligibility criteria	Three target trials for individuals initiated on (1) methadone; (2) buprenorphine/naloxone; (3) slow-release morphine between 1 January 2010 and 31 December 2022 (or latest available data at the time of analysis) Incident user design: BC residents ≥18 years old with a diagnosis of opioid dependence and no prior attempt at OAT who were not incarcerated, pregnant (up to 3 months of postpartum) or had cancer or palliative care Prevalent new-user design: Selection criteria above, with no history of OAT within the past month	Same as target trial
Treatment strategies	Initiate any dose of OAT and titrate up and down to (1) methadone: (i) ≤60 mg; (ii) ≤80 mg; (iii) ≤120 mg; (iv) ≤150 mg; (v) ≤260 mg; (2) buprenorphine/naloxone: (i) ≤8 mg/2 mg; (ii) ≤12 mg/3 mg; (iii) ≤16 mg/4 mg; (iv) ≤24 mg/6 mg; (v) ≤32 mg/8 mg; (3) slow-release oral morphine: (i) ≤300 mg; (ii) ≤700 mg; (iii) ≤1200 mg; (iv) ≤2300 mg	Same as target trial, measured by average daily dose per week from PharmaNet
Assignment procedure	Individuals are randomly assigned to a treatment strategy at time zero and are aware of the assigned strategy	Each individual has a clone created for each strategy of the target trial.
Outcomes	Primary outcomes: OAT discontinuation defined as interruptions in prescribed doses lasting ≥5 days for methadone or slow-release oral morphine and ≥6 days for buprenorphine/naloxone, all-cause mortality identified from the provincial Vital Statistics database	Same as target trial
Follow-up	From OAT initiation to outcome, pregnancy, incarceration, cancer diagnoses, palliative care, medication switch, December 31 2022 or 24 months, whichever occurs first.	Same as target trial
Causal contrasts	Per-protocol effect	Per-protocol effect
Analysis plan	Per-protocol analysis: inverse probability weighted pooled logistic regression model with censoring at deviation from protocol. Weights estimated as a function of baseline and post-baseline covariates	Same as target trial, except the per-protocol analysis will be conducted in an expanded data set with replicates using a clone-censor-weight approach

The labels (1) (2) and (3) are used to distinguish the three trials, and the absence of numbers means the details apply to all three trials.

BC, British Columbia; OAT, opioid agonist treatment.

Outcomes and follow-up

Our primary outcomes are treatment discontinuation and all-cause mortality. OAT discontinuation is defined as breaks in days dispensed lasting 5 days or more for methadone or SROM and 6 days or more for buprenorphine/naloxone. These thresholds are defined based on BC guidelines on dose reversion following interruption.² We will assume that OAT receipt prior to hospitalisation is continued daily throughout the duration of hospitalisation.² We have detailed our methodology for OAT episode construction elsewhere.^{52 80} Additionally, we will consider overdose-related acute care visits (hospitalisation or emergency department visits due to poisoning by opioid, stimulant, sedatives or hallucinogens) as a secondary outcome. We will follow each individual from OAT initiation to outcome or censorship due to death (for the outcomes of treatment discontinuation or overdose-related acute care visit), pregnancy, incarceration, cancer diagnoses, palliative care, medication switch, end of study follow-up (31 December 2022) or 24 months, whichever occurs earlier. We will stratify all analyses according to the form of medication received at OAT initiation. If a client transitions from buprenorphine/naloxone to buprenorphine monotherapy (either monthly subcutaneous injection or 6-month subdermal implant), we will consider the transition as a continuation of treatment.²

Primary exposure

The primary exposure will be a categorical variable that indicates the maximum dose allowed during a follow-up period. This will allow an investigation of the consequences of receiving a lower dose than key dosing benchmarks that have been described in the literature and clinical guidelines. We will initially consider five maintenance-dosing strategies for methadone and buprenorphine/naloxone and four strategies for SROM, though we will consider intermediate thresholds if sufficient data are available:

1. Methadone: (1)≤60 mg; (2)≤80 mg; (3)≤120 mg; (4)≤150 mg; (5)≤260 mg.

2. Buprenorphine/naloxone: (1)≤8 mg/2 mg; (2)≤12 mg/3 mg; (3)≤16 mg/4 mg; (4)≤24 mg/6 mg; (5)≤32 mg/8 mg.

3. SROM: (1)≤300 mg; (2)≤700 mg; (3)≤1200 mg; (4)≤2300 mg.

The exposure categories are defined to accommodate the BC guideline recommendations for maintenance dosing on methadone and buprenorphine/naloxone and the empirical distribution of daily dosing levels for SROM. The threshold for the highest dose stratum is based on the empirical data (ie, 99thpercentile) and will be adjusted to have sufficient sample size. The dosage will be measured by the average daily dose per week from the PharmaNet database.

Since maintenance dosing is a time-dependent exposure and the observational data can be consistent with multiple strategies at time zero, we propose to employ the ‘clone-censor-weight’ approach.^{81 82} To emulate a ‘per-protocol’ target trial where each eligible individual is randomly assigned to one of the maintenance dose strategies, we will create five replicates of methadone and SROM episodes and four replicates for buprenorphine/naloxone episodes (one per strategy). When an individual’s data are no longer consistent with an assigned strategy, we will artificially censor the corresponding replicate at the point of deviation from protocol; that is, when they are given a daily dosage higher than indicated by an assigned strategy (figure 1).

Figure 1.

Example of censoring process for a hypothetical individual receiving methadone. When an individual’s data are no longer consistent with an assigned strategy after allowing time to titrate dose up and down to an assigned strategy from an initial dose, we artificially censor the corresponding replicate at the point of deviation from protocol. The horizontal line under each strategy indicates a follow-up time from time zero. (A): as a primary analysis, we compare methadone dose strategies based on a maximum as (1) ≤60 mg; (2) 60–≤80 mg; (3) 80–≤120 mg; (4) 120–≤150 mg; (5) 150–≤260 mg. We will censor when a client is given a dosage higher than indicated by an assigned strategy. The replicates with the strategies ≤60 mg and ≤80 mg are censored at week 6 and week 9, respectively, when the individual is given a higher dose than the strategy assigned. The replicates ≤120 mg, ≤150 mg and ≤260 mg are followed until the outcome occurs because the dose has not exceeded the maximum dose assigned; (B): as a sensitivity analysis, we compare methadone dose strategies based on dosing interval as (1) ≤60 mg; (2) 60–≤80 mg; (3) 80–≤120 mg; (4) 120–≤150 mg; (5) 150–≤260 mg. We will censor when (1) a client is given a dosage higher than indicated by an assigned strategy at least 2 weeks or (2) when a client is given the same dosage lower than the minimum threshold of the strategy assigned for at least 3 weeks The replicates with the strategies ≤60 mg and 60–≤80 mg are censored at week 9 and week 10, respectively, when the individual is given a higher dose than the strategy assigned for 2 weeks. The last two replicates with the strategies 120–≤150 mg and 150–≤260 mg are censored at week 11, at which 100 mg (ie, a dose lower than the strategy) is maintained for 3 weeks.

Statistical analysis

Inverse probability of censoring weights

Inverse probability of censorship weights will be constructed to account for the time-dependent confounding by indication (selection bias) introduced by artificial censoring. Assumptions needed to identify causal effects include: (1) exchangeability (ie, no unmeasured confounders); (2) correct model specification; (3) positivity (ie, non-zero probability of deviating from protocol); (4) consistency (ie, the observed outcome is identical to the counterfactual outcome under his/her observed exposure history). We will use a pooled logistic regression to predict the time-dependent probability of remaining uncensored for each treatment strategy conditional on their covariate history. The stabilised weights in each replicate will be defined as:

$${\displaystyle S{W}_t=\prod _{k=0}^t\frac{Pr[C_k=0|C_{k-1}=0,D_k=0,G_k=0,V]}{Pr[C_k=0|C_{k-1}=0,D_k=0,G_k=0,{\stackrel{-}{L}}_k]}}$$

where $C_t$ is the indicator of artificial censoring at time $t$; $D_t$ is the binary indicator of outcome at time $t$; $G_t$ is the indicator of other censoring at time $t$ (ie, pregnancy, incarceration, cancer diagnoses, palliative care, medication switch including multiple forms of OAT or end of study follow-up); ${\stackrel{-}{L}}_t$ is a vector of covariate history by time $t$ and $V$ is a vector of covariates measured at time zero (a subset of $L_0$).

Similarly, we will use a pooled logistic regression to predict the time-dependent probabilities of remaining uncensored due to pregnancy, incarceration, cancer diagnoses, palliative care or medication switch. The final weights will be the product of those two inverse probability of censoring weights, truncated at the 99th percentile.

In the outcome model, we will present HRs using a weighted pooled logistic regression model with robust variance to estimate the per-protocol effect of the maintenance dosing strategy. This weighted model will include as covariate weeks since time zero (week and week squared), treatment strategy and variables measured at time zero.

We estimate marginal cumulative incidence curves for each strategy. This will be done by computing the individual-specific predicted survival probabilities from the pooled logistic regression model, averaging these probabilities over all individuals and then computing the absolute risk of outcome (cumulative incidence) as one minus the averaged probability of survival, and the corresponding risk differences. We will allow for a time-varying HR by including interaction terms (linear and quadratic) between the strategy and the number of weeks since OAT initiation in the pooled logistic regression model. The 95% CI⁸³ for the risk ratio and differences will be estimated using 500 bootstrap samples.⁸⁴

Confounder selection

We previously identified potential confounders associated with OAT retention by conducting a systematic literature review.⁵² We propose to augment this listing with additional variables available in our linked administrative data, which we hypothesise to influence both maintenance dosing and the outcomes (table 3).^{52 85 86} These variables include sociodemographics, measures of disease severity and comorbidity, OAT-related factors and COVID-19-related measures. In order to account for potential changes in prescribing practices and explicit recommendations in BC in after March 2020, we will include virtual prescribing,^{87 88} indication of delivery of OAT⁸⁹ and concurrent dispensations of short-acting opioids through the provincial Prescribed Alternatives programme.^90–92 A recent population-level study has found that dispensations under the programme are associated with reduced overdose-related and all-cause mortality among people with OUD.⁹³ All time-varying covariates will be updated weekly.

Table 3.

Potential confounding variables affecting OAT maintenance dosing, opioid treatment retention and mortality

Covariate	Description	Data source
Demographic and social characteristics
Age	Age at OAT initiation	Client Roster
Sex	Female vs male	Client Roster
Living in rural region	Categorised by local health authority (LHA) at OAT initiation based on population (≤40 000) and geographical consideration (proximity to a larger population centres and health services)	Client Roster
Receipt of income assistance in the past year *	Plan C coverage (low-income Pharmacare coverage programme) or receipt of income assistance from SDPR	PNet, SDPR
Unstable housing in the past 5 years*	ICD-10: Z59.0, Z59.1 from physician billing or hospital discharge records108 or three consecutive records of no fixed address from SDPR	MSP, DAD, SDPR
Incarceration in the past year	Admission to BC correctional facilities	BC corrections
Time since OUD diagnosis	OUD diagnosis identified by ICD-9/10 codes and the first OAT receipt	MSP, DAD, NACRS PNet
Calendar year	Calendar time at OAT initiation	PNet
Concurrent condition†*
Serious mental disorder (ever)	Includes major depressive disorder,109 110 bipolar disorder110 and schizophrenia109 111	MSP, DAD, NACRS
Hepatitis C (ever)	Indication of care for hepatitis C112–114	MSP, DAD, NACRS
Other substance use disorders (ever)	Excludes opioid use disorder and alcohol use disorder115 116	MSP, DAD, NACRS
Chronic pain in the past year	Based on ICD-10 codes pertaining to pain117 118	MSP, DAD, NACRS
Respiratory disease (ever)	Asthma or chronic obstructive pulmonary disease109	MSP, DAD, NACRS, PNet
Alcohol use disorder (ever)	116 119	MSP, DAD, NACRS
Treated tobacco use disorder (past year)	Based on ICD-10 codes120 and indication of nicotine replacement therapy in PharmaNet	MSP, DAD, NACRS, PNet
Charlson Comorbidity Index (CCI)	Based on 17 CCI groups using ICD-10 codes in the past year from hospital discharge records115 121	DAD
Medical history*
Dispensations of psychiatric or sedative medication in the past month	Includes central nervous system (CNS) depressants	PNet
Total dose of opioid dispensations other than OAT in the past week	Total morphine equivalents of any short-acting opioid dispensation	PNet
Drug-related acute care visits past 12 months	Drug-related hospitalisation or ED visits	DAD, NACRS
OAT related factors
Funding of prescribers who initiated OAT	Alternative payment plan (APP)122 indicated by the absence of a Medical Services Plan billing record (capturing fee-for-service billing) at initial OAT dispensation.	MSP, PNet
Attachment to primary care providers*	Defined as ≥50% of physician billing attributed to a single identifiable general practitioner in the past 6 months123	MSP, PNet
Practice size of prescriber who initiated OAT	Number of unique OAT clients in the past 12 months determined by network analysis124	PNet
OAT take-home doses in the past week*	Receipt of take-home doses for 2 or more days	PNet
Urine drug test in the past week*	Receipt of urine drug test from physician billing records	MSP
Calendar year of OAT episode initiation	Linear and quadratic terms	PNet
Cumulative years of prior OAT exposure	Measured at OAT initiation (linear and quadratic terms)	PNet
OAT initiation dose	OAT dose at initiation	PNet
OAT interruption for 2–4 consecutive daily doses in the past week	Indicator of interruption for methadone or SROM as clients are required to receive a reduced dose after missed doses1	PNet
COVID-19 related factors (past month)*
OAT virtual prescribing	Virtual care indicated in physician billing records87 88	MSP
OAT pharmacy delivery	Indication of delivery of OAT in PharmaNet89	PNet
Receipt of prescriber safer supply (PSS)	Concurrent PSS dispensations in PharmaNet90	PNet

*Time-dependent variable.

†Concurrent condition identified via ICD-9/10 diagnostic codes.

BC, British Columbia; DAD, Discharge Abstract Database; ED, emergency department; ICD, International Classificaion of Diseases; MSP, Medical Services Plan; NACRS, National Ambulatory Care Reporting System database; OAT, opioid agonist treatment; OUD, opioid use disorder; PNet, PharmaNet; SDPR, Ministry of Social Development and Poverty Reduction.

Subgroup and sensitivity analysis

We will execute sensitivity analysis by restricting the cohort and study timeline and altering the definition of the exposure and outcomes (table 4). The criteria for the artificial censoring will be altered to allow for dose adjustments during titration. Specifically, we will censor a replicate when a client is given a dosage higher than indicated by an assigned strategy for at least 2 weeks, instead of during the first week. In addition, in order to compare alternative treatment strategies based on dosing interval, instead of a maximum (eg, methadone strategies (1) ≤60 mg; (2) 60–≤80 mg; (3) 80–≤120 mg; (4) 120–≤150 mg; (5) 150–≤260 mg), we will impose an additional condition for censoring: we will censor when a client is given the same dose lower than the minimum threshold of the strategy assigned for at least 3 weeks. Furthermore, as a sensitivity analysis of the mortality outcome, we will consider mortality while on treatment with a capture window up to 4 and 5 days (ie, allowed gap time) after methadone or SROM and buprenorphine/naloxone discontinuation, respectively.

Table 4.

Proposed subgroup and sensitivity analyses

Proposed sensitivity analysis	Rationale
1. Sample restriction
Inclusion criteria requiring past chronic pain	Individuals with prior indications of chronic pain may require different dosing strategies based on care requirements.
Inclusion criteria requiring past severe mental health condition	Individuals with mental health conditions may have different care requirements and results may be impacted based on differing care.
Individuals with history of prescription opioid use other than OAT prior to OUD diagnosis	To access the treatment effect among those who are at potentially greater risk of prescription opioids misuse.
Excluding OAT episode with concurrent dispensation of short-acting opioid at time zero and censor observations if initiated	To control for presence of short-acting opioid and its dose.
Individuals who initiate with Methadose for the trial of methadone	To account for the methadone reformulation.
Excluding individuals receiving OAT by primarily detox prescribers	To remove individuals receiving OAT for rapid detoxification.
Excluding OAT episode initiating immediately following hospital discharge	To access the treatment effect among OAT episodes with its first dose dispensed at community pharmacy.
Inclusion criteria requiring >1 year of prior cumulative OAT experience	To access the treatment effect among people with OAT experience.
Including all OAT episodes with 5/6 days of washout period	To account for episodes initiated within 30 days of previous OAT dispensation.
Extending a washout period to 90 days	To assess the sensitivity of gap times between episodes and account for any changes that may be impacted by more recent OAT experience.
2. Timeline restriction
After the date of the first death for which fentanyl was detected in the province (1 April 2012)	To account for period after fentanyl was detected
After the date of the declaration of the provincial public health emergency (14 April 2016)	To account for period after fentanyl contamination increased
Prior to the COVID-19 pandemic (up to 17 March 2020)	To explore the robustness of our finding prior to the COVID-19 era
During the COVID-19 pandemic (from 18 March 2020)	To explore the robustness of our finding during the COVID-19 era
3. Variable classification
Weekly dose measured by median daily dose per week	To assess the robustness of the exposure measurement
Episode discontinuation: >14 days	Alternative discontinuation thresholds have been defined at 15 days and 31 days in other studies and guidelines125 126 as opposed to discontinuation thresholds of 5 days (methadone or slow-release oral morphine) and 6 days (buprenorphine/naloxone)2
Episode discontinuation: >30 days
Alternative censoring rule: censor when a client is given a dosage higher than indicated for ≥2 weeks	To assess the treatment effect with a less stringent censoring rule allowing for dose adjustments during titration
Alternative treatment strategies based on dosing interval. We will censor when a client is given the same dose lower than the strategy assigned for ≥3 weeks	To assess the treatment effect with a more stringent censoring rule to compare alternative treatment strategies
No censoring at pregnancy, incarceration and receiving cancer or palliative care	To allow individuals to have such events during follow-up
Censoring at the point of switching to the extended-release formulation of buprenorphine	To assess the treatment effect among OAT episodes that are initiated with buprenorphine/naloxone without transitioning to the extended-release formulation
Alternative primary outcome: death with censoring at treatment discontinuation	To assess mortality while on treatment
Alternative secondary outcome: all-cause acute care visits	To assess the robustness of alternate definition of the secondary outcome related to drug overdose
Alternative secondary outcome: overdose-related acute care visits or overdose death (or overdose death only if sample size is sufficient)

OAT, opioid agonist treatment; OUD, opioid use disorder.

Patient and public involvement

While no patients were explicitly involved in the design of this study, its conception was influenced by prior engagement with local advocacy organisations of people who use drugs and people who have accessed OAT.⁹⁴ Qualitative feedback on this and other related objectives outlined in the parent grant R01DA050629 were used to prioritise this analysis, given the potential impact on client engagement. Findings will be shared in consultation with local advocacy groups following completion of the analysis.

Ethics and dissemination

Databases are available to the research team by BC Ministries of Health and Mental Health and Addiction as part of the response to the provincial opioid overdose public health emergency. This study is classified as a quality improvement initiative. Providence Health Care Research Institute and the Simon Fraser University Office of Research Ethics determined the analysis met criteria for exemption per Article 2.5 of the 2018 Tri-Council Policy Statement: Ethical Conduct for Research Involving Humans (Canadian Institutes of Health Research 2018). This study will follow international guidelines for study conduct and reporting, including Strengthening the Reporting of Observational Studies in Epidemiology guidelines.⁹⁵ All study data are stored and managed in a secure, access-controlled environment maintained by the Centre for Advancing Health Outcomes and accessed through approved data access protocols. Only authorised research personnel can access deidentified data. Data handling follows institutional and provincial privacy regulations. We will otherwise administer the ‘Risk of Bias in Non-Randomised Studies-of Interventions’ tool to a multidisciplinary scientific advisory committee for ex post evaluation. Results will be disseminated to local advocacy groups and decision-makers, national and international clinical guideline developers, presented at international conferences and published in peer-reviewed journals electronically and in print. This study will generate robust evidence on the efficacy of alternative daily maintenance dosage levels of OAT on treatment discontinuation or all-cause mortality, in the interest of improving the health outcomes of this population.

Ethics statements

Patient consent for publication

Not applicable.