Scotland's 2009–2015 methadone-prescription cohort: Quintiles for daily dose of prescribed methadone and risk of methadone-specific death

Aims As methadone clients age, their drug-related death (DRD) risks increase, more than doubling at 45+ years for methadone-specific DRDs. Methods Using Community Health Index (CHI) numbers, mortality to 31 December 2015 was ascertained for 36 347 methadone-prescription clients in Scotland during 2009–2015. Cohort entry, quantity of prescribed methadone and daily dose (actual or recovered by effective, simple rules) were defined by clients' first CHI-identified methadone prescription after 30 June 2009 and used in proportional hazards analysis. As custodian of death records, National Records of Scotland identified non-DRDs from DRDs. Methadone-specific DRD means methadone was implicated but neither heroin nor buprenorphine. Results The cohort's 192 928 person-years included 1857 non-DRDs and 1323 DRDs (42%), 546 of which were methadone specific. Actual/recovered daily dose was available for 26 533 (73%) clients who experienced 420 methadone-specific DRDs. Top quintile for daily dose at first CHI-identified methadone prescription was >90 mg. Age 45+ years at cohort-entry (hazard ratio vs 25–34 years: 3.1, 95% confidence interval: 2.4–4.2), top quintile for baseline daily dose of prescribed methadone (vs 50–70 mg: 1.9, 1.1–3.1) and being female (1.3, 1.0–1.6) significantly increased clients' risk of methadone-specific DRD. Conclusion Extra care is needed when methadone daily dose exceeds 90 mg. Females' higher risk for methadone-specific DRD is new and needs validation. Further analyses of prescribed daily dose linked to mortality for large cohorts of methadone clients are needed internationally, together with greater pharmacodynamic and pharmacokinetic understanding of methadone by age and sex. Balancing age-related risks is challenging for prescribers who manage chronic opiate dependency against additional uncertainty about the nature, strength and pharmacological characteristics of drugs from illegal markets.


| INTRODUCTION
By the 1990s, UK's heroin injector epidemics of the early 1980s were being countered by opioid substitution therapy, primarily methadone, which not only reduced drug-related deaths (DRDs) 1,2 but also bloodborne virus transmissions and criminality. 3,4 Scotland's methadoneprescription cohort 5 includes many who are former injectors, of whom at least half will be hepatitis C virus carriers 6,7 ; most smoke, and misuse of alcohol, psychiatric and physical comorbidities are also not uncommon. 8,9 Despite the remarkable fall in UK's DRD-rate per 1 million defined daily doses of methadone in the early 21 st century, 1 the present decade has seen sharply increased numbers of opioid-related deaths. 10 The UK's increase was anticipated by evidence syntheses [11][12][13] and by national record-linkage studies of virtual cohorts of opioid-dependent clients. 3,14 Both forewarned about demographic influences (sex, age group) on DRD rates, including a strong sex by age-group interaction in DRD-risk for opioid users. 5,14 Discovering that females' advantage, in terms of lower DRD-risk, diminished with age 14 is important for risk prediction 15 but was unrecognized in early systematic reviews. 16 We became concerned about a possible role for prescribed methadone in Scotland's rise in methadone deaths in the second decade of the 21 st century. 5 Indeed, Scotland's 2009-2013 methadoneprescription cohort demonstrated that clients' risk of methadonespecific DRD increased more strikingly with age than for all DRDs. 5 Quickly, the cohort of opioid users in England's 2005-09 National Drug Treatment Management System was used to validate the Scottish results. 17 This English record-linkage cohort could adjust for a triad of major behavioural risk factors (injector status, misuse of alcohol, misuse of benzodiazepines) but lacked information on the prescribed quantity or daily dose of methadone. Synthesis of the 2 UK studies suggested that the risk of methadone-specific DRD tripled by age 45+ years (95% confidence interval [CI]: 3.0 to 4.7) compared with 25-34 years. 17 Meanwhile, based on 87 DRDs in a primary care cohort, Hickman et al. 18 considered confounding between the choice of opioid substitution therapy (methadone vs buprenorphine) and the client's age group or number of comorbidities, both of which are potentially implicated in methadone-specific DRDs. 5,17,[19][20][21][22] Internationally, there is a dearth of information on the daily dose of prescribed methadone for large cohorts of opioid-dependent clients. 15,16 Our aim is to provide alternative proportional hazards (PH) analysis for methadone-specific DRDs in Scotland's 2009-2015 methadone-prescription cohort based on: i. clients' sex, age group at accrual, prescription source and baseline quintile for quantity (qQs) of prescribed methadone; or ii. clients' sex, age group at accrual, prescription source and baseline quintile for daily dose (dQs) of prescribed methadone.

| Definitions: DRDs
We applied the UK harmonized definition of DRD. 23 National Records of Scotland 10 provided information on the opioid specificity of Scotland's DRDs: methadone-specific DRDs: methadone was implicated in DRD but neither heroin/morphine nor buprenorphine implicated; heroin-specific DRDs: heroin/morphine was implicated in DRD but neither methadone nor buprenorphine implicated; and heroinmethadone DRDs: methadone and heroin/morphine both implicated in DRD but buprenorphine was not implicated.
In appraising which drugs are implicated as causal factors in any DRD and which, although present, probably did not contribute, What is already known about this subject • Opioid substitution therapy halves clients' drug-related death rate.
• However, age-related increases drug-related and methadonespecific deaths persist.
• Opioid cohorts lack substantial data on prescribed daily dose of methadone.

What this study adds
• Validated simple rules for recovery of prescribed daily dose of methadone.
• Top quintile for prescribed daily dose (>90 mg per day) associated with increased hazard for methadone-specific death vs 50-70 mg; 3 times greater risk of methadonespecific death at 45+ years vs 25-34 years; females may be at greater risk of methadone-specific death.
Scotland's pathologists are supported by having a national protocol for toxicological testing at forensic autopsies.

| Scotland's Community Health Index
Scotland's Community Health Index (CHI) is a register of all patients in National Health Service (NHS) Scotland, Scotland's publicly funded healthcare system. From birth, patients are identified by a 10-digit CHI number, usually the patient's date of birth (DDMMYY) followed by 4 digits: 2 randomly generated, the third identifying sex (odd for males), and the fourth a check digit. The CHI numbers are key to Scotland's trusted record linkage, 24 not least because deaths and hospitalizations are CHI identified.

| Scotland's methadone-prescription client cohort for 2009-2015: data sources and linkage
Methadone prescriptions for opioid substitution therapy fall within a specific classification category in Scotland's National Prescribing Information System 25 : most are CHI identified, see Figure 1. All give quantity of methadone prescribed and number of instalments by which the prescribed quantity is issued, as both are required for the reimbursement of pharmacists. Daily dose of prescribed methadone is not routinely available in electronic format. 5,26 But daily dose of prescribed methadone is available electronically for a subset of general practitioner (GP) prescriptions; and was extracted by Scotland's Information Services Division using Natural Language Programs applied to GPs' electronic messaging. 26 Instalment dispensing is different in Scotland: on regular GP prescriptions, any drug is allowed and can be for any duration, although good practice suggests a limit of 28 days. In England, instalment pre-

| Ensuring positivity
As in an earlier analysis, 5 because reimbursement date was substituted for missing prescription date, 60 days were added to all survival times. Positivity was assured for all except for 5 CHIidentified clients, who were excluded.

| Plausible upper bound for age at cohort entry
Ninety-two CHI-identified clients were excluded because age at cohort entry was >69 years. The 2 checks above together excluded 97 CHI-identified clients: 52/97 had died; 2/52 were DRDs. There remained 36 347 CHIidentified methadone-prescription clients.

| Simple rules for establishing daily dose: Derivation and validation
For this paper, we devised and verified simple rules for recovery of daily dose from quantity of prescribed methadone and number of instalments, see Appendix 1.
Simple rules (hereafter, recovery rules) were devised for establishing daily dose from quantity of methadone prescribed and number of instalments at first CHI-identified prescription. Not every instalment number had an acceptable rule, see below. Recovery rules were derived for CHI-identified clients who had most recently joined the cohort so that our derivation subcohort comprised:  For 26 533 (73%) clients, recovery rules allowed quintile for daily dose of prescribed methadone at first CHI-identified prescription (dQ) to be analysed. Quintiles partition clients into fifths according to prescribed daily dose: from the 20% receiving the lowest fifth of baseline prescribed daily doses (dQ1) to the top 20% of prescribed daily doses of methadone (dQ5). Using quintile indicators allows the association between hazard ratio (HR) and increasing baseline prescribed daily dose to be made explicit in PH regression analysis.

| Statistical analysis
After documenting the impact of exclusion criteria on prescriptions, clients and deaths, we summarize the performance of recovery rules for daily dose for Scotland's 2009-2015 methadone-prescription cohort. Next, for each covariate level, we provide DRD rates and methadone-specific DRD rates for the cohort as a whole; and when restricted to clients with actual or recovered daily dose at first CHIidentified prescription.
Using PH regression analysis, we assess how steeply HRs increase by age group at accrual for methadone-specific DRDs; and how influential-based on regression χ 2 values on 4 degrees of freedomare dQs vs qQs. Adjusted HRs and 95% CIs are estimated simultaneously, relative to each covariate's baseline category as shown in tables. Appendix 2 includes corresponding PH analyses for all DRDs which, unlike methadone-specific DRDs, require sex by age group interaction to be taken into account. All analyses were performed using STATA v15.1; STATA's stcox was used for PH analysis.

| Sensitivity analysis
Appendix 3 includes 3 types of sensitivity analysis. First, rather than age group at cohort entry, we fitted time-updated age group since a single transition to an older age group could have occurred. The second analysis focuses on GP clients solely as GP prescriptions alone were the basis for our recovery rules. Thirdly, the key PH analysis for methadone-specific DRDs that incorporated dose quintiles was

| Recovery rules and descriptive statistics
Eight accepted recovery rules for daily dose at cohort entry were generally of the form quantity prescribed divided by D(i) where the value for divisor D(i) depended on (i), the number of instalments (4,7,14,21,28,35,42 or 56), with extra conditions needed only when the number of instalments was 4 or 7 in order for at least 75% of actual daily doses to be recovered correctly, see Appendix 1: actual agreement rate was 88% overall.
Recovered or actual daily dose at cohort entry was available for 26 533 (73%) of clients in Scotland's 2009-2015 cohort, including for 7349 (58%) of the cohort's 12 743 CHI-identified clients whose prescriber was other source, see Table 1. Recovered or actual daily dose was available for 19 184 (81%) of 23 604 CHI-identified clients whose prescriber was GP.  Table 2.
Clients' non-DRD rate was 9.6 per 1,000 py vs their DRD rate of 6.9 per 1,000 py, both precisely estimated. The 65% of clients with a GP prescriber had lower DRD rate (and lower methadone-specific DRD rate) than clients whose prescriber was other source. Two-thirds of clients were male, for whom methadone-specific DRD rate was lower at 2.6 per 1000 py (95% CI: 2.4-2.9) than for females (3.2, 95% CI: 2.8-3.7).
Both DRD rate and methadone-specific DRD rate increased with age group at cohort entry, the latter more steeply. The modal age group at accrual to Scotland's 2009-2015 methadone-prescription cohort was 25-34 years (44% of clients) with only 8% of clients aged 15-24 years. Methadone-specific DRD rate was significantly higher for clients in the top quintile for prescribed quantity at accrual to the cohort (qQ5: 4.3 per 1000 py, 95% CI: 3.7-5.0) than for clients in the middle quintile (qQ3: 2.1 per 1000 py, 95% CI: 1.7-2.6).

| Adjusted HRs for methadone-specific DRDs in Scotland's 2009-2015 methadone-prescription cohort
Baseline quintile for prescribed quantity (Table 4) For methadone-specific DRDs, interaction between sex and agegroup is unnecessary (χ 2 on 3 degrees of freedom of 4.00, P = .026, see Appendix 2). Females have higher HR (1.4) than males; HRs increase very steeply with age group at cohort entry, being 3-fold higher for clients 45+ years and 2-fold greater for clients aged 35-44 years than if 25-34 years old at cohort entry. Only the top quintile for quantity of prescribed methadone was associated with a significantly increased HR compared to qQ3. Clients whose prescription source was non-GP had significantly higher methadone-specific risk (HR, 1.36); likewise, DRD risk (see Appendix 2: HR, 1.31). Baseline quintile for recovered or actual daily dose (

| Sensitivity analyses
See Appendix 3 for 3 sensitivity analyses. The first relates to current age group rather than age group at cohort entry (Table A4). Since clients' age group changes at most once during follow up, age effects sharpened only slightly with current age group as alternative to age group at accrual. The second focuses on GP clients only, as GP prescriptions were the basis for our recovery rules (Table A5), but still endorses Table 5. Thirdly, Appendix 3 re-estimates Table 5 for the subcohort of mainly prevalent CHI-identified clients whose cohort entry was in July to December 2009 (Table A6) vs later-recruited CHIidentified clients ( physical and psychiatric comorbidities may be less well known by other prescribers than by GPs.
For context, non-DRDs outnumbered DRDs by 3:1 for methadone-prescription clients aged 45 years and over.

| Key considerations
Notwithstanding the substantial reduction in harms (overdose deaths, criminality and blood-borne virus risks) that opioid substitution therapy has delivered for younger heroin users, the risk of methadonespecific DRD increases both as clients age into their 40s and 50s; and steadily with baseline quintile for daily dose of prescribed methadone.
Guidelines for methadone clients recommend a daily dose of 60-120 mg. 27 Adherence to prior guidance was checked by prescribing surveys [28][29][30][31]  In some individuals, females especially, 32 methadone (unlike buprenorphine 18 ) is associated with prolongation of the QTc interval leading to the development of Torsades de Pointes and cardiac arrest. 33 Undiagnosed QTc prolongation may manifest as methadonespecific DRDs.
Periodic electrocardiograms are recommended for clients receiving >100 mg of methadone daily, 33 but not achieved in practice. Other risk factors for QTc prolongation include comorbidities such as circulatory or liver disease; coprescribing for mental or physical ill health 34 ; use of both methadone and cocaine; and being female. 32,35 Could the latter partly explain our novel finding that females are at higher risk of methadone-specific DRD? [35][36][37][38] Confounding between methadone-specific DRD risk and the client's daily dose of prescribed methadone (dQ) cannot be excluded.
Balancing of prescribing risks is challenging in clinical practice, never more so than when managing chronic opiate dependency. Risks The above considerations produce an environment of high risk for overdose toxicity and sudden death. Information, as in this paper, which might mitigate some of these risks is, therefore, important.
We have raised a concern about higher DRD risks for methadone clients of non-GP prescribers: in particular, for methadoneprescription clients whose cohort entry date was during 2010-2015, a period when nonmedical prescribing in the management of substance misuse had expanded in Scotland, as elsewhere. 40 Scotland's GPs have not been reluctant to manage methadone prescriptions for clients with major comorbidities as around 80% of CHIidentified clients whose cohort entry was July-December 2009 had GP prescribers. In terms of risk mitigation, it might be helpful for GP Summary Care Records to be routinely available to specialist service prescribers so that they are aware of ageing clients' comorbidities.
In a changing landscape of funding, commissioning of services In deriving simple rules for recovery of daily dose, a limitation was that actual daily doses were available electronically from GP prescriptions only but recovery rules were applied to other-source prescriptions on the reasonable assumption that the same relationships hold. As a check, PH analyses using quintiles for daily dose were repeated for GP-prescribed clients only, and inferences were essentially unaltered.
Record-linkage studies have limited scope for resolving data queries. We took a harder line than previously on exclusion criteria, respectively for prescriptions and clients.
The need to substitute the later reimbursement date for missing first prescription date was a minor issue. More importantly, we did not know, and so could not analyse, when clients exited from methadone therapy as the date of their last CHI-identified methadone prescription does not exclude later non-CHI-identified prescriptions.
Hence, once included in the cohort, clients have remained in follow up.
Confounding between DRD-risk and the client's daily dose quintile cannot be ruled out. 18 Age >45 years and prescribed daily dose >90 mg may be markers for harder-to-support clients whose opioid dependency is chronic, who have physical comorbidities, notably circulatory and digestive system diseases, or coprescriptions for mental or physical ill health. 8 Finally, we did not request that methadone clients' coprescriptions 34 for benzodiazepines, antiviral medications, mirtazapine, amitriptyline, sertraline or macrolide antibiotics (to name but a few) be linked in because the added time and complexity would not have been warranted given that illicit supplies would have remained unaccounted for: and constitute most of the benzodiazepines present at Scotland's DRDs. 8

| CONCLUSIONS
Scotland's 2009-2015 methadone-prescription cohort helps to explain why UK official statistics on DRDs and opioid-specific deaths in the second decade of 21 st show stark increases by age-group, and disproportionately so for females. 10  Urgently, interventions are needed to de-escalate ageing methadone clients' risk not only of methadone-specific DRD but also of their major causes of non-DRDs. 19 Practitioners must balance: moderation of clients' daily dose of methadone, ideally to below 90 mg if clients are willing; review of circulatory or digestive comorbidities that, respectively, risk sudden death masquerading as methadone-specific DRD or prolongation of methadone's half-life; support for smoking cessation to manage better the client's respiratory and circulatory diseases; hepatitis C virus clearance by directly acting antiviral therapy; and review of medications prescribed for psychiatric and physical comorbidities for possible interactions with methadone. 34 Methadone exhibits large interindividual variation in response, a nar-  Baseline quintile for prescribed quantity: For all DRDs, Table A2 shows that the interaction between sex and age group at accrual is highly statistically significant (P = .0026) and signals that female clients' reduced DRD hazard (overall HR, 0.82) is reversed for older clients.