Article Text
PDF
Abstract
Objectives We used an international pharmacosurveillance network to estimate the rate and characteristics of antidepressant use in older adults in countries with more conservative (UK) and liberal depression guidelines (Canada, USA).
Setting Electronic health records and population-based administrative data from six jurisdictions in four countries (UK, Taiwan, USA and Canada).
Participants A historical cohort of older adults (≥65 years) who had a new episode of antidepressant use between 2009 and 2014.
Outcome measures The age and sex-standardised cumulative incidence of new episodes of antidepressant use in older adults was measured. Descriptive statistics were used to compare the proportion of new users by the antidepressant prescribed, therapeutic class, potential treatment indication and country, as well as the characteristics of the first treatment episode (standardised daily doses, duration and changes).
Results The incidence of antidepressant use between 2009 and 2014 varied from 4.7% (Montreal and Quebec City) to 18.6% (Taiwan). Tricyclic antidepressants (TCAs) were the most commonly used class in the UK (48.8%) and Taiwan (52.4%) compared with selective serotonin reuptake inhibitors (SSRIs) in North American jurisdictions (42.3%–53.3%). Chronic pain was the most common potential treatment indication (41.2%–68.2%). Among users with chronic pain, TCAs were used most frequently in the UK and Taiwan (55.2%–60.4%), whereas SSRIs were used most frequently in North America (33.5%–46.4%). Treatment was longer (252–525 vs 169–437 days), standardised doses were higher (0.7–1.3 vs 0.5–1.0) and treatment was more likely to be changed (31%–46% vs 21%–34%) among patients with depression (9.1%–43%) than those with chronic pain.
Conclusion Antidepressant use in older adults varied 24-fold by country, with the UK, which has the most conservative treatment guidelines, being among the lowest. Chronic pain was the most common potential treatment indication. Evaluation of real-world risks of TCAs is a priority for future research, given high rates of use and the potential for increased toxicity in older adults because of potent anticholinergic effects.
- antidepressants
- multinational comparison
- older adults
- epidemiology
- pharmacosurveillance
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
Strengths and limitations of this study
Uses an international pharmacosurveillance network to characterise antidepressant use across multiple jurisdictions.
Despite data harmonisation efforts, measurement and reporting issues may lead to arbitrary differences between regions and countries.
Treatment indication is unknown for all but one study cohort. Documentation of health problems in medical services billing data and electronic medical records was used as a proxy for treatment indication for remaining cohorts, which could have resulted in misclassification bias.
While some cohorts represented total populations or representative samples, others were assembled from practices supported by specific information technology systems.
Introduction
Depression affects an estimated 300 million people worldwide.1 In the past decade, many countries have reported a twofold to threefold increase in the use of antidepressant medications.2 The primary increase has been in newer antidepressants such as selective serotonin reuptake inhibitors (SSRIs), which are considered to have better efficacy than older drugs, with fewer side effects.3 Increasing prevalence of use of antidepressants has also been associated with an increase in the number of people started on therapy as well as its duration, with 40% of antidepressants being prescribed for more than 180 days.4 Antidepressants now represent one of the most commonly prescribed medications,5 6 especially among older adults, where the increase in use has been the greatest in the US population.5 7
The increasing use of antidepressants in older adults may be due to a number of factors. There has been long-standing advocacy for better recognition and treatment of depression in older adults.8 9 In addition, individuals with multiple chronic conditions are more likely to experience depression. In Canada, 12% of persons ≥65 years have two or more major chronic conditions compared with only 3.6% in younger adults; and 4% of adults with at least one major chronic disease have mood or anxiety disorders, which are other approved indications for antidepressant use.10 While increasing prevalence of depression, anxiety and other mood disorders in older adults may account for increasing use of antidepressants, there is also evidence that antidepressants are prescribed for unapproved indications that may not be supported by scientific evidence of efficacy.11–14 One recent study found that nearly 50% of antidepressants were prescribed for unapproved indications including chronic pain, tiredness and sleep disturbance,11 which are also more common in older adults.15
Taken together, these factors have led to increasing concern about overuse of antidepressants, particularly in older adults.16 This concern is heightened because of the reported association between antidepressant use and fall-related injuries, motor vehicle accidents, functional decline and mortality in older adults.17–19 Clinical practice guidelines for the management of depression vary considerably in their recommendations for first-line treatment,20–24 although most acknowledge the need to consider patient preferences (table 1). The UK NICE guidelines are the most conservative, recommending first-line treatment with psychotherapy such as cognitive behavioural therapy with pharmacotherapy added only if non-pharmacological therapy is unsuccessful.24 At the other end of the spectrum, the Canadian 2009 guidelines list pharmacotherapy as a first-line treatment for major depressive disorder, including SSRIs, serotonin-norepinephrine reuptake inhibitors (SNRIs), mirtazapine or bupropion.25–27 Of note, Canada ranks the third highest user of antidepressants among 23 Organisation for Economic Co-operation and Development countries, surpassed only by Australia and Iceland.2
Summary of depression guidelines
However, few data are available regarding how national guidelines may contribute to differences in how antidepressants are used for older adults. We took advantage of an international pharmacosurveillance network28 that uses electronic medical record (EMR) and population-based health administrative data to estimate the rate of antidepressant use in older adults in countries that had more conservative (UK) and lenient guidelines (Canada, USA, Taiwan). We evaluated differences in the choice of antidepressant, dose and duration of treatment, the prevalence of dose changes and switches in treatment of older adults with new episodes of antidepressant use, as well as differences in treatment approach for patients with a recorded diagnosis of depression, other mental health comorbidities and chronic pain.
Methods
Study design and population
To characterise antidepressant use in older adults, a cohort of individuals who were ≥65 years of age and had a new episode of antidepressant use between 1 January 2009 and 31 December 2014 following 2 years during which they had no prescription or dispensation of an antidepressant was assembled. Eligible patients were identified from EMR or population-based registries in a total of six jurisdictions in four countries (Canada, USA, UK and Taiwan).
Data sources
Canada
Quebec EMR cohort
Data were extracted from the Medical Office of the 21st Century (MOXXI) EMR, which includes a real-time linkage to the Quebec insurance agency, the Régie de l’assurance maladie du Québec (RAMQ) databases. This linkage provides historical and daily updates of patients’ received medical services (diagnosis, procedure, date, location, provider), prescriptions dispensed from community pharmacies for persons who are publicly insured (prescriber, pharmacy, drug, dose, dispensing date, duration, refills) and mortality. Approximately 110 primary care physicians in Quebec use MOXXI for approximately 90 000 of their patients, representing approximately 25% of the patients in the practice, and approximately 5% of the Quebec population in primary care practices.
Quebec administrative data cohort
Data were retrieved from the Montreal Population Health Record, a population-based 25% random sample of the 4.1 million residents of Montreal, Canada’s second largest city, which is dynamically updated each year to account for in and out-migration. Beneficiary characteristics were measured by linking the RAMQ beneficiary database (age, sex, postal code) with Statistics Canada’s census file (postal code level measures of socioeconomic status) and the Institute of Statistics birth and death registry (date of birth and death). From this population-based dataset, we identified all non-institutionalised persons ≥65 years of age between 2009 and 2014. All medical services (date, diagnosis, procedure), hospitalisations (admission and discharge date, primary and secondary diagnoses, in-hospital procedures) and medications (prescriber, pharmacy, drug, dose, dispensing date, duration, refills) received by members of the cohort were retrieved from the RAMQ database and those of the Quebec Ministry of Health.
Ontario administrative data cohort
Data were extracted from the Institute for Clinical Evaluative Sciences population-based repository of health records for the 13.6 million Ontario residents. The data are provided by the Ontario Ministry of Health and Long-Term Care, which pays for all essential health services. For the 2 053 588 Ontario non-institutionalised residents who were ≥65 years between 2007 and 2014, we randomly sampled 20% and extracted age, sex and date of death from the beneficiary registrants database, all billing claims for medical services provided by Ontario physicians (date of service, diagnosis, procedure code, physician provider), all claims for medications dispensed in the community by Ontario pharmacies (drug, date dispensed, prescriber, quantity, duration, number of refills) and data from discharge summaries of emergency department visits and hospital stays (date of admission and discharge, primary and secondary diagnoses and procedures).
USA
Boston, Massachusetts EMR cohort
Data were extracted from the Partners HealthCare Research Patient Data Registry (RPDR), which provides care for approximately 3 million of 5 million residents in Boston and surrounding areas. RPDR includes data from the longitudinal medical record (LMR), an internally developed, web-based, fully functional EMR that was in use during this period for the participating primary care clinics from Brigham and Women’s Hospital (BWH) and Massachusetts General Hospital (MGH), the two founding members of Partners Healthcare in Boston, Massachusetts. Data were retrieved from the structured clinical encounter information from RPDR for all outpatient primary care visits. To ensure complete follow-up, patients were eligible if they were seen in one of 37 BWH-affiliated or MGH-affiliated primary care or diabetes clinics.
UK
EMR cohort
Data were extracted from the Clinical Practice Research Datalink (CPRD), an anonymised LMRs database for primary care. The primary care EHR includes information on all primary care interactions, including documented health problems, visit notes, prescriptions and records of specialty referrals, as well as information on lab results, hospitalisations and death. The CPRD includes more than 4 million active patients, 76% of whom are in England, representing 6.9% of the 64 million primary care population in the UK, and around 600 primary care practices.29 Patients were included if their records met minimum quality standards and there were no gaps in registration. A 10% random sample of the 2 436 180 elderly patients in the database were extracted.
Taiwan
Administrative data cohort
Data were extracted from the Taiwanese National Health Insurance (NHI) system databases. The NHI provides health coverage to around 99% of Taiwan’s population of 23 million, providing comprehensive coverage for all health services, including dental, preventive, Western medicine and traditional Chinese medicine services. A random sample of 103 400 of the 3.6 million residents who were ≥65 years in Taiwan between 2007 and 2014 was extracted for this study. The NHI’s research databases include a registry for beneficiaries (eg, registrant’s age, sex and residence), an outpatient visit database (date and time of visit, ICD-9-CM codes of existing health problems, service provided), an inpatient visit database (date of hospitalisation and discharge, ICD-9-CM codes of existing health problems, procedure codes and dates) and a pharmacy database (drug prescribed, date, duration, dosage, prescribing physician, dispensing pharmacy).
Measurement
Patient characteristics
Age and sex
Date of birth and sex were retrieved from the administrative data in Canada and Taiwan as these data are verified at the time of enrollment in the health plan. For the USA and UK, these data were retrieved from the EMR.
Depression, other mental health comorbidities and chronic pain
We measured the existence of potential indications for antidepressant use11 12 including depression, other mental health conditions and chronic pain. Depression included mild, moderate, major single or recurrent depressive disorder with or without psychotic symptoms, adjustment reaction and mixed anxiety and depression. Other mental health conditions included anxiety, alcohol abuse, illicit drug use, attempted suicide, psychosis, schizophrenia and bipolar disorder. Anxiety included dissociative and somatoform disorders. We used standard diagnostic codes (ICD 9, ICD 10 and Read codes) retrieved from the EMR, and/or medical services claims, and hospitalisations in the 2 years prior to the first antidepressant prescription to measure these conditions (see online supplementary appendix).
Supplemental material
To measure chronic pain, we used a previously validated ICD9 and ICD10 code set for non-cancer pain.30 The concepts and definitions used in the development and validation of this code set were used to map to Read codes (appendix available on request). Chronic non-cancer pain conditions included lumbar pain, back disorders that are often associated with pain (eg, degenerative disc disease), neck and back problems (eg, spondylosis), fibromyalgia, complex regional pain syndromes (eg, mononeuritis), painful neuropathic disorders (eg, postherpetic neuralgia), pain disorders with psychosocial dysfunction and unclassified chronic pain problems.
Characterising antidepressant use
The characteristics of antidepressant use were compared in the population of seniors with a new episode of antidepressant use in each country, overall and then within subgroups that had a diagnosis of (a) depression, (b) chronic pain or (c) other mental health problems.
Classification of starting therapy
The Anatomic Therapeutic Classification system31 was used to map national drug names and identification numbers to a common nomenclature based on the antidepressants that were prescribed during the study time period. Therapeutic classes included tricyclic antidepressants (TCAs), SSRIs, SNRIs, serotonin antagonist and reuptake inhibitors (SARIs), norepinephrine–dopamine reuptake inhibitors (NDRIs), noradrenergic and specific serotonergic antidepressants, tetracyclic antidepressants, melatonergic agonists and NRIs. Monoamine oxidase inhibitors were excluded as they are rarely used and only for bipolar disorder. Starting therapy was classified by therapeutic class and drug, and by the number of antidepressants prescribed at treatment initiation.
Starting antidepressant dose
To enable comparisons in the dose prescribed among antidepressants, we created a standardised dose for each drug by dividing the prescribed dose by the WHO recommended daily dose for adults. The resulting indicator represents the proportion of the recommended daily adult dose that was prescribed. When more than one antidepressant was prescribed concurrently, we summed the standardised dose for each drug.
Analysis
The cumulative incidence of older adults with new episodes of antidepressant use in the period 2009 to 2014 was measured by dividing the number of new users by the total population that was ≥65 years during this time period. Rates were standardised for age and sex using the direct method and the UK as the reference population. New episodes of use were defined as those who had no prior prescription of an antidepressant in the 2 years preceding their first prescription. Descriptive statistics were used to compare the proportion of new antidepressant users by the antidepressant prescribed, therapeutic class, country and potential treatment indication, as well as the characteristics of the first treatment episode. As the coexistence of other mental health conditions may modify treatment choices, we used the prevalence ratio to estimate the concurrence of other mental health problems, chronic pain and depression. The prevalence ratio allowed us to compare between countries even if the prevalence of these conditions is different.32 It was calculated as the prevalence of the health problem of interest among individuals with depression, divided by the prevalence of the health problem among individuals without depression.
Patient and public involvement
Patients and members of the public were not involved in the development of research questions or outcome measures, design or implementation of this study. There are no plans to involve patients in dissemination of research results.
Results
The age-standardised cumulative incidence of antidepressant use in persons aged ≥65 in the period 2009–2014 was highest in Taiwan (18.6%) and Ontario, Canada (15.3%), and lowest in Quebec-Montreal (4.7%) and England (6.6%) (table 2). Simlarly, the overall prevalence of antidepressant use in the same time period was lowest in England (10.3%), and highest in Ontario, Canada (26.8%) and Taiwan (23.4%). The distribution of age was similar in all jurisdications.
Comparisons in the cumulative incidence and prevalence of episodes of antidepressant use in adults ≥65 years between regions and countries
Frequency distribution of the first antidepressant prescription by therapeutic class
Overall, there were substantial differences between jurisdictions in the choice of starting therapy for new users (table 3). Tricyclic antidepressants were the most commonly used class of antidepressants in England (48.8%) and Taiwan (52.4%) compared with SSRIs in North American jurisdictions (42.3%–53.3%). Notably, in the UK, approximately one-quarter of antidepressant prescriptions were for newer classes of antidepressants: serotonin antagonist and reuptake inhibitors (trazodone), norepinephrine–dopamine reuptake inhibitors (bupropion) and noradrenergic serotonin-specific antidepressants (mirtazapine).
Frequency distribution of the first antidepressant prescribed for new antidepressant users by therapeutic class by country and jurisdiction
Coexistence of depression, other mental health problems and chronic pain
The prevalence of a depression diagnosis among new antidepressant users was relatively low, varying from 9.1% (Taiwan) to 21.7% (USA). The exception was the EMR cohort in Quebec where the prevalence was 43.0% (table 4). Among new antidepressant users with a diagnosis of depression, there was a 30% (UK prevalence ratio: 1.3) to sevenfold (USA prevalence ratio: 7.3) increase in the likelihood of also having an alcohol or other substance abuse problem. Anxiety, which varied from being present in 12.1% of new antidepressant users in the UK to 47.8% of new users in Taiwan, was the most common mental health problem, and was also more likely to be present among persons with depression (Prevalence ratio: 1.3 [Ontario, Quebec] to 3.6 [USA]). Although less common in all jurisdictions, both major mental illness and suicide attempt were more likely to be present among antidepressant users with depression compared with those without depression. In contrast, chronic pain was the most prevalent problem among new antidepressant users (41.2% [UK] to 68.2% [Taiwan]), and was as likely to be present among those with or without depression (Prevalence ratio: 0.9–1.3) (table 4). For almost one-third of new antidepressant users, there was no documented diagnosis of depression, other mental health problems or chronic pain.
Prevalence of potential therapeutic indications: depression, other mental health problems and chronic pain among new antidepressant users, and the proportion of persons with depression who had other concurrent mental health problems or pain by jurisdiction
Choice of antidepresssant by potential treatment indication
Among new antidepressant users with a diagnosis of depression, SSRIs were selected as the starting therapy for 55.2% (Ontario) to 71.4% (UK), followed by SNRIs in Canadian jurisdictions and newer antidepressants in Boston, the UK and Taiwan (table 5). For antidepressant users with a diagnosis of chronic pain, starting therapy was predominantly with SSRIs in North American jurisdictions (33.5%–46.1%), and with TCAs in the UK and Taiwan (55.2%–60.4%). Selection of therapy for other mental health problems was a mixture of choices made for persons with a diagnosis of depression and chronic pain.
Starting therapy for new users of antidepressants by potential therapeutic indication by therapeutic class and country
Characteristics of the first treatment episode by potential treatment indication
Among patients with a diagnosis of depression who were started on antidepressants, 54% (Ontario) to 68% (Taiwan) remained on their initial course of antidepressant treatment (table 6A). For this group, the mean duration of treatment varied from 104.7 (Taiwan) to 340.9 (Quebec) days, and the mean dose varied from 0.6 of the recommended adult dose (Taiwan), to the recommended adult dose in the USA site (mean dose: 1.0). Among the 32% (Taiwan) to 46% (Ontario) with a change in treatment, the most common change in all jurisdictions was an increase in dose occurring in 20% (Taiwan) to 33% (Ontario) of patients, followed by a decrease in dose (13% [Taiwan] to 21% [Ontario]) and the addition of a drug from a different therapeutic class (6% [UK] to 18% [Ontario]). Among patients with a change in therapy in their initial treatment episode, the mean duration of treatment was almost double (mean duration 312.6 to 616.9 days) than for patients with no change, and the mean dose during treatment was substantially higher (mean standardised proportion of the adult dose: 0.8 to 1.4).
Characteristics of the first antidepressant treatment episode among persons with a diagnosis of depression
The first treatment episode for new antidepressant users with a diagnosis of chronic pain was substantially different from patients with a diagnosis of depression (table 6B). Overall, 66% (Ontario) to 79% (Taiwan) were kept on the inititial treatment regimen, for a much shorter duration than for new users with a diagnosis of depression (mean duration: 83.6 days [Taiwan] to 300.9 days [Quebec]), and at lower mean doses (mean standardised proportion of the adult dose: 0.4 [Taiwan, UK, Montreal] to 0.8 [USA]). For the 21%–34% of patients who had a change in treatment, the most common changes in all jurisdictions were an increase in dose (13% to 25%), a decrease in dose (9% to 14%) and the addition of another drug from a different therapeutic class (3% to 11%). For those with a change in treatment, the duration of the treatment episode was considerably longer in all jurisdictions (264.4–590.0 days), and the mean dose during the treatment episode was higher (0.5–1.1).
Characteristics of the first antidepressant treatment episode among persons with a diagnosis of chronic pain
Interpretation
This multinational study is the first to compare the use of antidepressants among comparable cohorts of new users in different countries. We found wide variation in the incidence and prevalence of antidepressant use; the lowest rates included the UK, where guidelines for pharmacotherapy use for depression were the most conservative. Choice of starting therapy also varied widely by country and condition, with the UK and Taiwan more likely to use TCAs for persons with a diagnosis of chronic pain, whereas SSRIs were the most common choice for both depression and pain in North American settings. Overall treatment duration was longer for patients with depression than chronic pain, mean treatment doses were higher and there were more likely to be changes in therapy.
Differences in the incidence of antidepressant use in different countries as well as the choice of antidepressant therapy may also be influenced by a variety of system-related factors. For example, direct to consumer advertising of prescription drugs is permitted in the USA but banned in Canada, Europe and the UK because these ads drive demand for specific treatments.33 Also, drugs available and covered by public insurance plans in the UK, Taiwan and Canada are influenced by rigorous reviews and guidelines for the use of new drugs and technologies by national agencies such as the National Institute for Health and Care Excellence in the UK.34–36
A surprisingly large percentage of older adults were prescribed antidepressants for conditions other than depression. We found that chronic pain was the most common problem documented, and it appeared to be the most likely reason for antidepressant use. Chronic pain has an estimated prevalence of 11.8%37–43.5%38 in adults, and increases with age, with an estimated 62% of those ≥75 years of age experiencing this problem.38 While use of antidepressants for chronic pain has been shown to be effective, particularly TCAs (amitriptyline) and SNRIs (duloxetine) for neurogenic pain,39 40 TCAs have been on the lists of potentially contraindicated medications for the elderly for over a decade because of their potent anticholinergic effects that can lead to cognitive impairment, and other avoidable morbidity.41 42 Moreover, only one antidepressant, duloxetine, has been approved for treatment of chronic pain in North America and Europe.11 Although the anticholinergic effects of TCAs have been well documented,43–45 the actual ‘real-world’ evidence of harm is limited,46 and should be addressed in future studies given the high rates of use.
Notably, in this study of older adults, depression was documented in only 2 in 10 users; the exception being in Quebec City/Montreal where 4 out of 10 patients were prescribed antidepressants for depression. There are known problems in underreporting, diagnosis and documentation of depression,47–49 which may account for some or all of these differences between sites. The MOXXI system used in the Quebec cohort likely provides an approximate estimate of the extent to which depression is undocumented. In this cohort, physicians are required to document the treatment indication for each drug prescribed,50 data that have been validated in prior studies.51 Results from this setting suggest that twice as many patients receive antidepressants for depression, whereas a similar proportion of new antidepressant users had a diagnosis of chronic pain as in other sites.
Also, of interest, among patients with documented depression, antidepressant use in all countries followed current guidelines for depression management.21 24 52 53 Choice of first-line therapy was predominantly SSRIs and SNRIs. Starting doses were lower than the recommended adult doses, mean duration of treatment was 251–525 days and the main changes in treatment were increases in doses and the addition of an antidepressant from a different therapeutic class. Not surprisingly, among patients where there was a change in treatment, both the average antidepressant dose was higher and treatment duration was longer, suggesting a lack of treatment effectiveness. The cohort in Taiwan was different in a number of respects including lower antidepressant doses, shorter treatment duration and fewer changes in therapy once started. The reason for these differences in approach is not clear. They may represent an innate conservatism in the use of antidepressants in the Taiwanese population where there is an acknowledged absence of evidence to support clinical guidelines,53 or known differences in genetic determinants of drug metabolism that have been established in population-based genomic studies of drug metabolism.54
This study has several limitations to consider in the interpretation of results. Efforts were made to harmonise data from different sources and countries; however, both measurement and reporting issues may lead to arbitrary differences between regions and countries. With the exception of the Quebec cohort, treatment indication is unknown. Documentation of health problems in medical services billing data and EMRs was used as a proxy for treatment indication, which based on past studies has a predictive value of 7.8%–80.3%.55 The resulting misclassification will attenuate differences observed in treatment approach between conditions. Moreover, diagnoses of health problems tend to be underreported in adminsitrative data. Prescription data were used in the three jurisdictions and may overestimate the incidence and prevalence of antidepressant use as approximately 37% of antidepressant prescriptions are not filled.56 The Ontario, Montreal and Taiwan cohorts represented total populations or representative samples, whereas the UK, Quebec and USA cohorts were assembled on the basis of practices supported by specific information technology systems. The trends observed in these more selected populations, however, are similar to those reported for antidepressant use in older adults in these countries.6 57–62
In conclusion, antidepressant use in older adults varies up to 24-fold by country (cumulative incidence from 0.7% to 17.0%), and chronic pain appears to be the most common treatment indication in all jurisdictions, even more so than depression. Evaluation of real-world risks and benefits of antidepressants by treatment indication should be a priority for future research. The relative contraindication of TCAs in older adults because of their potent anticholinergic effects needs to be assessed given their high rate of use and proven efficacy in pain management.39–42
References
Footnotes
Contributors RT, DWB, DLB, WD, AJF, JH and JL contributed to the study concept and design. NG, TS and BH contributed to the statistical analysis. RT drafted the manuscript. DWB, DLB, WD, AJF, NG, JH, BH, SK, JL and TS critically reviewed the manuscript for intellectual content. All authors read and approved the final manuscript.
Funding This project was funded by the Canadian Institutes of Health Research (CIHR) operating grant # MOP-111166.
Competing interests None declared.
Ethics approval This study received ethics approval from the McGill University Institutional Review Board (IRB # A01-B02-02A), the Institute for Clinical Evaluative Sciences, the Partners Human Research Committee IRB (Protocol # 2011P000773/PH) and the Independent Scientific Advisory Committee (ISAC) of CPRD (Protocol # 14_081R).
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement No additional are data available.
Patient consent for publication Not required.