Article Text
PDF
Abstract
Introduction Sleep-time blood pressure correlates more strongly with adverse cardiovascular events than does daytime blood pressure. The BedMed trial evaluates whether bedtime antihypertensive administration, as compared with conventional morning use, reduces major adverse cardiovascular events.
Methods and analysis Design
Prospective randomised, open-label, blinded end-point trial.
Participants
Hypertensive primary care patients using blood pressure lowering medication and free from glaucoma.
Setting
Community primary care providers in 5 Canadian provinces (British Columbia, Alberta, Saskatchewan, Manitoba and Ontario) are mailing invitations to their eligible patients. Social media campaigns (Google, Facebook) are additionally running in the same provinces.
Intervention
Consenting participants are allocated via central randomisation to bedtime vs morning use of all antihypertensives.
Follow-up
(1) Telephone or email questionnaire at 1 week, 6 weeks, 6 months and every 6 months thereafter, and (2) accessing linked governmental healthcare databases tracking hospital and community medical services.
Primary outcome
Composite of all-cause death, or hospitalisation for myocardial infarction/acute-coronary syndrome, stroke or congestive heart failure.
Secondary outcomes
Each primary outcome element on its own, all-cause hospitalisation or emergency department visit, long-term care admission, non-vertebral fracture, new glaucoma diagnosis, 18-month cognitive decline from baseline (via Short Blessed Test).
Select other outcomes
Self-reported nocturia burden at 6 weeks and 6 months (no, minor or major burden), 1-year self-reported overall health score (EQ-5D-5L), self-reported falls, total cost of care (acute and community over study duration) and mean sleep-time systolic blood pressure after 6 months (via 24-hour monitor in a subset of 302 sequential participants).
Primary outcome analysis
Cox proportional hazards survival analysis.
Sample size
The trial will continue until a projected 254 primary outcome events have occurred.
Current status
Enrolment ongoing (3227 randomised to date).
Ethics and dissemination BedMed has ethics approval from six research ethics review boards and will publish results in a peer-reviewed journal.
Trial registration number NCT02990663.
- hypertension
- clinical pharmacology
- cardiology
- primary care
- clinical trials
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
Recruiting through primary care providers, having minimal exclusion criteria and reducing barriers to participation by communicating directly with participants, helps to ensure accurate data collection and good generalisability to primary care populations.
Beyond an assessment of efficacy, multiple potential harms are being evaluated.
Members of the public with hypertension are making substantial contributions to study design and conduct through our 10-member patient working group.
If we observe relative risk reductions for the primary outcome that are smaller than 17%, it is unlikely we will be able to declare those differences statistically significant with the planned sample size.
Introduction
Blood pressure (BP) normally exhibits a circadian rhythm with relatively lower pressures during sleep.1 Lack of this sleep time ‘dip’ correlates strongly with adverse cardiovascular events such as myocardial infarction (MI), stroke and congestive heart failure (CHF), and BP correlates most strongly with such events when measured at night (ie, during sleep).2–5 Given some antihypertensive medications might lower sleep time BP more effectively when administered at bedtime,6 administration time could conceivably alter the degree of cardiovascular risk reduction these medications provide.
In 2010, Spanish researchers published the first hypertension trial to compare bedtime with morning antihypertensive administration and examine mortality and morbidity outcomes.7 The results of this randomised controlled trial (RCT), the MAPEC trial (Monitorizacion Ambulatoria para Prediccion de Eventos Cardiovasculares, i.e. Ambulatory Blood Pressure Monitoring for Prediction of Cardiovascular Events), were striking, reporting a 61% relative reduction in a composite of major adverse cardiovascular events (MACE). Despite the obvious clinical importance of this finding, however, hypertension guidelines have yet to endorse bedtime prescribing.8–11 This presumably relates to concern over irregularities in the reporting of MAPEC’s results and methods.12 13 The MAPEC trial registry, for instance, was attributed to at least eight other RCTs,14–21 making it appear to describe a general programme of research, and not the methods of a single RCT. Following this, in 2019, the same principal investigator published another RCT favouring bedtime over morning antihypertensives, the Hygia trial, which reported a 45% relative reduction in MACE.22 Again, however, irregularities in the reporting of Hygia’s results and methods, including a lack of clarity over how randomisation and allocation were carried out, has led to calls for independent confirmation of these findings before bedtime prescribing of antihypertensives is embraced.13 23–25
BedMed is a large community-based RCT intended to replicate an MAPEC-like timing intervention in a hypertensive Canadian primary care population. BedMed randomises participants to take all existing BP medication (as tolerated) at bedtime, compared with conventional morning use, and tracks mortality and morbidity using regularly collected administrative health claims and participant self-report. This protocol is prepared in accordance with SPIRIT (Standard Protocol Items: Recommendations for Interventional Trials) guidelines.26
Objectives
Main
To determine whether a bedtime versus morning antihypertensive administration time influences mortality or cardiovascular morbidity.
Secondary
To determine whether a bedtime versus morning antihypertensive administration time adversely influences cognitive ability, visual acuity, risk of falls and fractures, or nocturia.
Methods
Trial design
BedMed is a phase 4 pragmatic clinical trial with an adaptive, event-driven, parallel enrolment, prospective randomised open blinded-endpoint design.27 Here, ‘adaptive’ refers to the potential future exclusion of new participants whose only antihypertensive is a diuretic, if adherence to bedtime allocation in such individuals is poor (see Adherence to bedtime diuretics substudy).
Recruitment began in March 2017, and the trial will continue until 254 primary outcome events have been observed (the number of events in MAPEC). Based on current ongoing enrollment (3227), and an observed 2.0% annual event rate, final analysis is anticipated in fall 2023.
Setting and recruitment
Pragmatic trials collaborative
Most recruitment (~78%) is through community family physicians (>400) who own and operate independent clinics. These providers are spread widely across five participating provinces (Alberta, British Columbia, Manitoba, Saskatchewan and Ontario), but affiliated with the Pragmatic Trials Collaborative (www.PragmaticTrials.ca), a practice-based research network which is coordinating the trial. Nurse practitioners with their own practice panel (seven at present) are also participating.
Each clinic uses their own electronic medical record to create a list of hypertensive patients and the primary care provider (PCP) removes those they consider palliative or incapable of informed consent. The study team then provides the clinic with recruitment envelopes, which the clinic addresses and mails to these potentially eligible patients. The envelopes contain (1) a letter of introduction from the patient’s PCP and (2) a pamphlet describing the trial and providing contact information (online supplemental files 1 and 2). Interested patients call the study team where research assistants answer questions, determine eligibility and obtain consent either in real-time via email (>80% of participants opt for this) or by letter-mail for handwritten consents.
Supplemental material
Supplemental material
Social Media
All hypertensive residents of our five participating provinces are eligible for BedMed, whether or not their PCP is involved. While this can happen through word-of-mouth, a social media campaign (Google and Facebook Ads) is being employed to inform the public about the trial. These Ads (online supplemental video 1) direct individuals to a landing page (https://bedmedstudy.ca/) providing trial information, a check of eligibility and telephone/email contact information for the study team.
Supplementary video
Trial population
Inclusion criteria
Clinician diagnosis of hypertension (by any physician or nurse practitioner).
Taking ≥1 BP-lowering medication once daily, or PCP willing to convert ≥1 BP-lowering medication to once daily.
≥18 years of age.
Community-dwelling (ie, not residing in a nursing home).
Exclusion criteria
Considered palliative or unable to consent by PCP.
Sleep disrupting shift work (more than three shifts/month during participant’s regular sleeping hours).
Glaucoma diagnosis, or using glaucoma medication (safety exclusion: nocturnal hypotension, which bedtime BP meds could worsen, has been associated with optic neuropathy in glaucoma patients).28–30
Randomisation and allocation
Consenting participants receive their random allocation to bedtime vs morning BP medications while dialoguing directly with a research assistant who has no preceding clinical interactions with that participant and who obtains their allocation (stratifying by province with random blocks of 10 or 12) from the central REDCap31 server’s randomisation module, ensuring irreversible and concealed allocation.
Intervention
Treatment
Use of all once-daily BP-lowering medication(s) at bedtime.
Control
Use of all once-daily BP-lowering medication(s) in the morning.
Implementation
Participants choose between having their PCP assist their timing change (using the PCP’s judgement on how and what to change), or being assisted by the research assistant with whom they are dialoging. Only PCPs assist with timing changes if participants describe heart disease, or if their BP medications include Tiazac XC or Diltiazem XC (which have delayed-release kinetics), furosemide, isosorbide mononitrate/dinitrate, or alpha blockers (medications whose timing decision may be more complicated). PCPs can convert twice daily medications to once daily alternatives, but this is not actively promoted.
Research assistants only change the timing of once daily medications, with a limit of one medication change per week (using the order ACE inhibitors angiotensin receptor blockers, calcium channel blockers, beta-blockers, diuretic-containing medications, other). They advise participants to make the switch by delaying the next dose until the allocated time, and continuing that schedule. If bedtime use is problematic, they ask participants to try taking their BP meds with dinner. If morning use is problematic, they ask participants to try taking it with lunch. Participants with regularly reversed sleep schedules (ie, sleeping during the day) take their BP medications when they get up, or when they go to bed, not according to the time of day.
At each follow-up, participants are asked about medication timing, and encouraged to adhere to allocation. No devices to separately monitor adherence are in use. As a memory aid, all participants are advised to place pill bottles near objects they use when transitioning to or from bed (eg, toothbrush, denture case, alarm clock), or to use an AM/PM dosette. If participants report a new diagnosis of glaucoma, they are advised to take their BP medications in the morning, regardless of allocation, to minimise the risk of optic neuropathy.
Follow-up and data management
Research assistant interactions
All participant interactions with research assistants are unblinded and recorded directly into the University of Alberta’s implementation of the REDCap data management platform.31 The following interactions are scheduled relative to the date of randomisation.
Baseline: Telephone interaction to (1) obtain baseline characteristics, (2) conduct the Short-Blessed Test to assess cognitive function, and (3) randomise the participant. May be split over multiple interactions (participant’s choice).
One week: Telephone interaction to troubleshoot timing change problems and encourage adherence.
Six weeks: Telephone interaction to gather information on adverse effects and outcomes.
Six months: Telephone interaction, or REDCap email survey (participant’s choice), to gather adverse effects and outcomes.
Twelve months: Same as 6 months+EQ-5D-5L quality-of-life survey (EuroQol Group’s health-related quality-of-life instrument).
Eighteen months: Same as 6 months+follow up Short Blessed Test (but available by telephone only).
Every 6 months thereafter: Same as 6 months.
Administrative claims data
All Canadian provinces have publicly funded healthcare systems and maintain linkable healthcare databases tracking medical services rendered during healthcare interactions for all their residents. This includes community physician services and diagnoses (whether by specialists or generalists), prescriptions dispensed, reasons for hospitalisation and vital statistics (ie, mortality). BedMed participants consent to these datasets being accessed and analysed to support the trial, providing both outcomes and baseline characteristics.
Twenty-four-hour ambulatory BP monitoring
To assess between-group differences in achieved BP, we intended to carry out 24-hour BP monitoring on a consecutive sample of 151 intervention and 151 control subjects residing in 6 Alberta communities at 6 months (providing 90% power to detect the difference in overnight systolic BP observed in MAPEC). Although we will be able to reach our intended sample size, the timing of these measurements has been substantially delayed for many participants due to both logistic hurdles, and the COVID-19 pandemic. Participants are provided a copy of their test results, which are also faxed (if they consent) to their PCP.
Outcomes
Unless otherwise stated, all outcomes are recorded over the duration of the study.
Primary
Major adverse cardiovascular events
Defined as first occurrence of either all-cause death or hospital admission/emergency department (ED) visit for acute coronary syndrome/MI, stroke CHF.
Secondary
Each component of the primary outcome individually.
All-cause hospitalisation/ED visit.
Long-term care (LTC) admission (ie, to nursing home or assisted living facility).
Non-vertebral fracture.
New glaucoma diagnosis
Cognitive decline at 18 months
Defined as ≥ 2-point worsening in cognitive performance compared to baseline, as measured by the Short Blessed Test.
Supplementary safety outcomes
Vision
Vision self-reported as ‘much worse’ compared with the last follow-up at any point, or ‘slightly worse’ than the last follow-up, on two or more occasions (Note: vision is reported, every 6 months, as either ‘unchanged’, ‘slightly worse’ or ‘much worse’ than the last follow-up).
Cognition
New ‘impairment consistent with dementia’ at 18 months (Short Blessed Test newly ≥10) or new diagnosis of dementia at any point during follow-up.
Symptomatic Hypotension
Self-reported light-headedness, or feeling faint without loss of consciousness, in the prior month.
Self-reported fainting (loss of consciousness) in the prior month.
Self-reported falling in the prior month.
Hip fracture.
(Note: at 6 weeks, 6 months and every 6 months thereafter, participants are separately asked whether they have felt lightheaded, fainted, or fallen in the last month).
Nocturia
Self-reported change from baseline in the number of overnight urinations per week (at 6 weeks and 6 months).
Self-reported nocturia burden in the prior month, recorded as no nocturia, or nocturia that is ‘no problem’, ‘minor problem’ or ‘major problem’ (at 6 weeks and 6 months).
Cost
Acute care costs (estimated from each hospital admission’s resource intensity weight and length of stay)*.
Total cost of care (acute care costs+medication costs+physician billings)*.
*All cost measures are derived entirely from administrative claims data, and not from self-report. If claims data is not available for some participants, they will be excluded from this analysis.
Exploratory
1. Self-reported overall health score (via EQ-5D-5L) at 12 months.
Process
Proportion of BP medication doses taken at the allocated time at 6 months (two times per day medications being considered as half dose in the AM and half dose in the PM for this calculation)†.
Sleep-time systolic BP after 6 months (consecutive sample of 302 Alberta residents)†.
†Although blinded to individual participant process outcomes, investigators are unblinded to the aggregated results for adherence to allocation time, and to the isolated results from the 24-hour BP assessments. This allows for consideration of protocol alterations should the intervention appear poorly applied. Investigators are otherwise fully blinded to all trial outcomes.
End-point adjudication
Administrative data
Administrative data derived outcomes will be identified using established and validated coding algorithms.32–34 Physicians providing these diagnoses are generally acute care providers (emergency physicians, hospitalists, specialists) who are unaffiliated with the BedMed trial.
Adjudication panel
Most primary and secondary outcomes are being collected in duplicate (ie, by administrative claims and participant self-reporting of the same events). This information will be reviewed by a panel of three physicians blinded to allocation. If the panel deems both data sources to be concordant, those events will be considered valid, and the event date in administrative claims will be used. When events are discordant (eg, only present in one of the two data sources or differing in diagnoses) the participant’s PCP will be contacted to provide the adjudication panel with more information, including their opinion on whether the event occurred. The exception is all-cause hospitalisation/ED visits, where we will preferentially use only administrative claims data, believing it to be highly accurate, and being more challenging to confirm with PCPs given the high number of such occurrences.
Sample size determination
BedMed is event driven, and originally sought to observe 406 primary outcome events before stopping. We chose this event target believing this was the largest number of events a network our size could detect with 3 years of observation. However, because patients receiving recruitment packages are less likely to enrol than expected (projected enrolment 12%, actual enrolment 6%), and because the overall annual event rate is at the low end of expectations (2.0%), we have reduced our event target for stopping to 254, which matches the number of events observed in MAPEC. Assuming meaningful covariates, 254 events should allow observed risk ratio differences of ~17% or larger to be declared statistically significant. To estimate when this number of events has likely been reached, Alberta Health Services is tracking the primary outcome event rate in Alberta BedMed participants on a quarterly basis. We then extrapolate this to the trial as a whole using the number of participant years of observation. At the current rate of events and enrolment, BedMed should conduct its final analysis in fall 2023.
Statistical analysis
Intention-to-treat assumptions
Lost to follow-up
If participants are lost to follow-up, but medical services continue to be recorded within administrative claims data, we will treat them as though they were still active in the study and censor survival data on the last date of medical services, or indication of death, whichever occurs later. If no such medical claims exist, data will be censored on the last day of successful telephone or email follow-up.
Withdrawal
Participants withdrawing from the study are asked to allow us to continue to follow their administrative claims data. If they agree (as the majority do), we will continue to use administrative claims outcomes for those individuals as per the loss to follow-up description. If they do not agree, survival data will be censored on the date of withdrawal.
Missing data
For each analysis, we will either impute a value from subsequent or preceding follow-up visits, or exclude a participant from analysis. How we deal with missing data will be specific to each analysis and prioritise either minimising bias, or being conservative when bias is unavoidable (ie, biasing against benefit and towards harm, for the intervention).
Non-adherence
Non-adherence to allocation will not exclude participants from analysis unless the outcome of interest is a harm that only makes sense to assess while on-treatment (eg, assessing how nocturia differs in diuretic users switched to bedtime, compared with non-diuretic users making the same switch).
Selecting regression covariates
Analyses of dichotomous outcomes will use a maximum of 1 covariate per 10 outcomes, and analyses of continuous outcomes will use a maximum of 1 covariate per 20 randomised subjects. The covariate list for each analysis is predefined in table 1, and all are measured at baseline. We will always use the maximum number of covariates possible, selected in the order given (ie, we will not undertake stepwise addition or subtraction).
Analysis plan
Subgroup analyses
We will repeat the primary outcome analysis for those with and without the following baseline characteristics: age ≥75, sex, physically frail (score ≥3 on physical frailty subscale of the Tilburg questionnaire), polypharmacy (≥5 medications), Overall Health Score ≤75, resistant hypertension (≥3 BP-lowering medications), CHF, diabetes, CAD (coronary artery disease), stroke or TIA, sleep apnoea, chronic kidney disease (with or without dialysis), sedentary (exercise 0 days per week).
Sensitivity analyses
We will present, according to treatment group, the baseline characteristics of those whose data was censored due to withdrawal or lost to follow-up, and compare these characteristics to those who were not censored in this way using Fisher’s exact test.
Patient and public involvement
Patient working group
BedMed has a 10-member patient working group helping to guide the trial. The group began meeting in 2016 prior to any recruitment to review and revise (1) recruitment materials, (2) phrasing of questions and (3) outcomes to be collected through self-report. Working group members have also assisted in hiring research staff, in further revising recruitment materials mid-study to increase enrolment, and in constructing a social media campaign. We anticipate working with our patient partners to make decisions, if needed, following our interim analysis in spring 2022, to interpret final results in 2023, and to help disseminate findings.
Patient-driven substudy
The draft BedMed protocol was presented in 2015 to a group of ~25 seniors prior to study registration and grant application. Feedback from this presentation resulted in the substudy to determine whether diuretics can be taken at bedtime without troublesome nocturia threatening adherence.
Substudies
Adherence to bedtime diuretics
Diuretics are widely believed to promote nocturia if taken later in the day, and are typically recommended for morning use only as a result.35 36 However, this recommendation is largely opinion based. Whether or not participants will adhere to bedtime diuretic dosing is unclear. To determine this, we will examine, at 6 weeks and 6 months, self-reported nocturia burden (no, minor, major), number of overnight urinations per week, and adherence to bedtime allocation, in the first 203 AM diuretic-only users randomised to bedtime and being followed for 6 months, and compare this to all those switching a single AM non-diuretic to bedtime during the same period. Assuming equal numbers in both groups, and 75% adherence to allocation time in non-diuretic users, this should provide 90% power to detect a 20% relative reduction in adherence in diuretic users.
Volunteer bias
Concern has been raised that randomised trial participants are poorly representative of real world populations.37–41 We will examine, using Alberta administrative claims, how baseline characteristics and preventive health behaviours differ in four distinct Alberta populations: (1) All BedMed-eligible patients attached to participating PCPs, (2) BedMed participants who enrolled after a PCP-letter, (3) BedMed participants responding to social media advertisement and (4) All BedMed-eligible Albertans. We will compare (1) Demographics (age, sex, postal code derived deprivation index, rural residence), (2) Comorbidities (diabetes, CAD, stroke, osteoarthritis, CHF, chronic obstructive pulmonary disease, dementia, hip fracture, CKD, dialysis, hospital admission in prior 6 months plus accompanying length of stay and resource intensity weighting), (3) Preventive therapies (prior 3 years shingles vaccine, statin use, osteoporosis medication), (4) Screening tests (prior 3 years PAP smear, colonoscopy, mammogram, FIT testing, PSA testing) and (5) clinical outcomes postrandomisation (death, BedMed primary outcome, all-cause hospitalisation or ED visit along with length of stay and resource intensity weighting, nursing home admission, new glaucoma diagnosis/treatment / surgery, hip fracture, and new dementia diagnosis). To substitute for the date of randomisation, we will use the date of PCP mailout for BedMed-eligible PCP-attached patients, and the date providing the same mean number of years of observation for all BedMed-eligible Albertans.
‘Nudge sentence’ recruitment strategy
Two years into recruiting, we hypothesised that altering the physician letter of introduction to state that a large number of people were already participating might improve the response rate. Online supplemental file 3 shows the new physician letter. The added wording states: ‘This study already has over 1700 Canadians with high blood pressure taking part. If you too choose to participate…’. As of March 2019, providers are given an equal number of both recruitment envelopes, sealed and shuffled together, for them to address and mail. Both letters are otherwise identical save for the date on the letter (odd numbered for the new version, even numbered for the original). Participants calling to enrol are asked the date on the letter to determine which version they are responding to, allowing a pseudorandom assessment of the ability of such a ‘nudge’ sentence to improve enrolment. This substudy will continue until recruitment ends, with sample size determined by the number of letters mailed during that interval.
Supplemental material
Early stopping
Independent data safety monitoring board
Outcomes from all provinces will be collected at the end of 2021. Each analysis described in this protocol will then be carried out, and presented to the Cochrane Hypertension Working Group (our independent data safety monitoring board, IDSMB).
Stopping Rules: If p is ≤0.001 for primary outcome benefit (the Haybittle-Peto boundary),42 or if p is ≤0.05 for harm, the IDSMB will apply clinical judgement and decide whether to recommend to the principal investigator that the trial be stopped early.
Competing studies
A trial similar to ours, the UK’s TIME trial,43 will likely release results ahead of BedMed. If convincing benefit is demonstrated, we will ask our IDSMB to weigh this new information and consider again whether early stopping is recommended.
Our group is also conducting a separate RCT of the same antihypertensive timing intervention in hypertensive LTC residents (BedMed-Frail).44 As both trials share the same IDSMB, interim data from both trials could be weighed in early stopping discussions for either trial.
Dissemination
Results will be published in a peer-reviewed journal, and summarised in knowledge translation vehicles targeted at PCPs, and the general public. We will also invite trial participants to a results webinar where they can directly pose questions to the principal investigator.
Ethics statements
Patient consent for publication
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Twitter @kimchspr, @DeeMangin, @DrRitaMc
Contributors BedMed was conceived and designed by SRG with input from MRK, GMA, JB, LG, AS, FAM, RSP, MDH, BM, KM, DPM, STW, RM, JPM, CN and TK. Each of these authors also participated as coapplicants on the grants that fund the trial, with SRG as the nominated principal applicant. Family physicians were recruited by SRG with help from MRK, GMA, TK, AS, BO'N, MG, DPM, DM, CM, STW and JEMK. DRT participates on the BedMed Patient Working Group and helped to revise recruitment materials, and to design the social media campaign. SRG and JB created the analysis plan. JB, SRG, AS, FAM and KM are coordinating access to administrative claims data. SRG and LSF wrote the draft manuscript, with all authors providing critical feedback.
Funding BedMed is funded by a Support for Patient Oriented Research (SPOR) Innovative Clinical Trial Multi-Year Grant (REF#: 151212) from the Canadian Institutes of Health Research (CIHR), and from a Partnership for Research and Innovation in the Health System (PRIHS) Grant from Alberta Innovates (REF#: 201 500 912 G2019000450). It also receives in-kind research assistance from Enhancing Alberta Primary Care Research Networks (EnACt), which itself is funded entirely by Alberta Innovates, and received pilot funding from the Northern Alberta Family Medicine Fund (University of Alberta Dept of Family Medicine funding).
Disclaimer Funders have no role in, or authority over, trial design, data collection, management, analysis, interpretation, writing, or decisions to submit for publication.
Competing interests None declared.
Provenance and peer review Not commissioned; peer reviewed for ethical and funding approval prior to submission.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.