Article Text

Original research
Long-term prognosis of low high-sensitivity cardiac troponin T in the emergency department compared with the general population
  1. Love Cyon1,2,
  2. Erik Kadesjö1,2,
  3. Gustaf Edgren2,3,
  4. Andreas Roos1,2
  1. 1Department of Emergency and Reparative Medicine, Karolinska University Hospital, Stockholm, Sweden
  2. 2Department of Medicine, Solna, Karolinska Institutet, Stockholm, Sweden
  3. 3Department of Cardiology, Södersjukhuset, Stockholm, Sweden
  1. Correspondence to Dr Love Cyon, Department of Emergency and Reparative Medicine, Karolinska University Hospital, Stockholm, Sweden; love.cyon{at}ki.se

Abstract

Background Long-term prognosis associated with low–high-sensitivity cardiac troponin T (hs-cTnT) concentrations in patients with chest pain is unknown. We investigated these prognostic implications compared with the general population.

Methods All first visits to seven emergency departments (ED)s in Sweden were included from 9 December 2010 to 31 August, 2017 by patients presenting with chest pain and at least one hs-cTnT measured. Patients with myocardial injury (any hs-cTnT >14 ng/L), including patients with myocardial infarction (MI) were excluded. Standardised mortality ratios (SMRs) and standardised incidence ratios (SIRs) were calculated as the ratio of the number of observed to expected events. The expected number was computed by multiplying the 1-year calendar period-specific, age-specific and sex-specific follow-up time in the cohort with the corresponding incidence in the general population. HRs were calculated for all-cause mortality and major adverse cardiovascular events (MACE), defined as acute MI, heart failure hospitalisation, cerebrovascular stroke or cardiovascular death, between patients with undetectable (<5 ng/L) and low (5–14 ng/L) hs-cTnT.

Results A total of 1 11 916 patients were included, of whom 69 090 (62%) and 42 826 (38%) had peak hs-cTnT concentrations of <5 and 5–14 ng/L. Patients with undetectable peak hs-cTnT had a lower mortality risk compared with the general Swedish population (SMR 0.83, 95% CI 0.79 to 0.87), with lower risks observed in all patients ≥65 years of age, but a slightly higher risk of being diagnosed with a future MI (SIR 1.39, 95% CI 1.32 to 1.47). The adjusted risk of a first MACE associated with low versus undetectable peak hs-cTnT was 1.6-fold (HR 1.61, 95% CI 1.53 to 1.70).

Conclusion Patients with chest pain and undetectable hs-cTnT have an overall lower risk of death compared with the general population, with risks being highly age dependent. Detectable hs-cTnT concentrations are still associated with increased long-term cardiovascular risks.

  • Chest Pain
  • Myocardial Infarction
  • Acute Coronary Syndrome
  • Biomarkers

Data availability statement

No data are available.

http://creativecommons.org/licenses/by-nc/4.0/

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

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

WHAT IS ALREADY KNOWN ON THIS TOPIC

  • The short-term prognosis in patients with chest pain in the emergency department (ED) with low or undetectable high-sensitivity cardiac troponin T (hs-cTnT) concentrations is excellent and permits a safe-rule out of myocardial infarction (MI).

  • There is a lack of data regarding long-term prognostic implications in these patients compared with the general population.

WHAT THIS STUDY ADDS

  • In a large cohort of patients with chest pain as the chief complaint in the ED, we found an up to 20% lower risk of death in patients with undetectable hs-cTnT concentrations compared with the age-standardised general population.

  • The risks were highly age dependent, with older patients generally having a more favourable prognosis, and detectable hs-cTnT concentrations below the 99th percentile upper reference limit were still associated with long-term risks of death and major adverse cardiovascular events.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

  • Our study provides novel findings that could prove useful in communicating risks with patients in emergency care by specifically delineating and relating individual to age-standardised and sex-standardised long-term risks.

  • Although the absolute long-term cardiovascular risks were overall low, the observed relative risk increases associated with detectable hs-cTnT concentrations should merit further attention on risk profiles and characteristics in younger patient segments for targeting of preventative strategies.

Introduction

The introduction of high-sensitivity cardiac troponin (hs-cTn) in clinical practice has led to several improvements in the early diagnosis and care of patients with myocardial infarction (MI).1 2 The use of these assays has improved the early diagnosis of MI in patients presenting with chest pain, and in parallel allowed the development of several novel cTn-based strategies that permit a safe and rapid rule-out of MI already at presentation.3

Several studies conducted in the general population have consistently found a graded association between hs-cTn concentrations and long-term risk of cardiovascular events and death, even at concentrations well below the upper reference limit for different assays.4 5 The long-term prognosis in patients with chest pain in the emergency department (ED) with low or undetectable high-sensitivity cardiac troponin T (hs-cTnT) concentrations is less extensively covered, with a lack of data on prognostic implications compared with the general population.

In this study, we aimed to investigate the long-term mortality and cardiovascular risk in patients with chest pain in ED without myocardial injury in relation to the general population.

Methods

Data sources

The study was conducted at Karolinska Institutet, Sweden and was based on data from seven large EDs in Stockholm and Göteborg, Sweden, from 9 December 2010 to 31 August 2017. Data from each hospital’s electronic medical record system were used to identify all ED visits in patients ≥18 years of age, and laboratory data were obtained from the local clinical chemistry databases. The Elecsys 2010 system (Roche Diagnostics, Mannheim, Germany) was used for hs-cTnT analyses at all hospitals. Data were sent to the Swedish National Board of Health and Welfare for record linkage, providing information on comorbidities, medications and cause-specific mortality from the National Patient Register (NPR), the Prescribed Drug Register and the Cause of Death Register, respectively.6–8 Further details about MI diagnoses were obtained from the Swedish Web system for Enhancement and Development of Evidence-based care in Heart disease Evaluated According to Recommended Therapies (SWEDEHEART) register.9 Data on age-stratified and sex-stratified annual mortality rates in the general population were extracted from Statistics Sweden, which is a government agency that produces official Swedish population statistics, while information on MI rates was obtained according to official data published by the Swedish Board of Health and Welfare (for description of data sources, see online supplemental material).10

The study was approved by the Human Research Ethics Committee of Stockholm, approval number 2022-06523-02 and adhered to the principles of the Helsinki Declaration. Informed consent was not required for this study. The reporting of this observational cohort study followed the STROBE guidelines.11

Study population

All patients ≥18 years of age in the study base who had a first visit to the ED with chest pain as the primary complaint, as registered by the triage nurse in the ED according to the rapid emergency triage and treatment system (RETTS), and at least one hs-cTnT concentration measured concurrently were eligible for inclusion (figure 1). These visits were defined as the index visits, including subsequent in-hospital stay among admitted patients. RETTS is a triage tool used on a nationwide level by the attending nurses in the ED that assign each patient a primary complaint for ED sorting based on most alarming symptoms.12 Eligible patients were included according to peak hs-cTnT concentrations, defined as the highest hs-cTnT concentrations measured at any time point during the index visit, including measurements registered during hospital stay in admitted patients. The distribution of peak hs-cTnT concentrations among eligible patients is presented in online supplemental figure 1. Patients with a final diagnosis of an acute MI, coded according to the 10th revision of the International Classification of Disease (ICD-10) in the NPR, and/or a diagnosis registered in the SWEDEHEART-register, that is, both ST-elevation MI (STEMI) and non-STEMI-patients, were subsequently excluded as well as all patients with peak hs-cTnT concentrations >99th percentile cut-off value for the hs-cTnT assay used at all sites (14 ng/L).13

Figure 1

Selection of study population. hs-cTnT, high-sensitivity cardiac troponin T.

Definitions

Comorbidities were defined according to discharge diagnoses registered prior to the index visit, coded according to the ICD-10 in the NPR. Myocardial injury was defined according to the fourth universal definition of MI.14 Ongoing medication was defined as two or more dispensed medications during the year preceding the index date.

Exposure

The exposure was defined according to peak hs-cTnT concentrations measured during the index visit and was categorised as low (hs-cTnT 5–14 ng/L) or undetectable (hs-cTnT<5 ng/L). Patients with undetectable hs-cTnT concentrations were used as the reference group in internal comparisons.

Outcomes

The primary outcome was all-cause mortality, with follow-up starting at the index date. The secondary outcome was the first occurrence of major adverse cardiovascular event (MACE) in patients who survived the index visit and was defined according to ICD-codes registered in the NPR as an acute MI (ICD-codes I21 or I22), heart failure hospitalisation (ICD-code I50), cerebrovascular stroke (ICD-codes I60-I64) or cardiovascular death (any code in the I-chapter in the Cause of Death Register). End of follow-up for cause-specific mortality and MACE was 31 December, 2020.

Statistical methods

Continuous variables are presented as means and SD or medians with IQR, as appropriate, and categorical variables as absolute numbers and proportions.

Standardised mortality rates and mortality ratios (SMRs) were calculated as the ratio of the number of observed to expected deaths. The expected number was computed by multiplying the 1-year calendar period-specific, age-specific and sex-specific follow-up time in the cohort with the corresponding incidence in the general population. This method enabled the estimation of expected events by utilising the age-specific and sex-specific incidence rates prevalent in the Swedish population during the entire follow-up period. Similarly, standardised incidence rates and ratios (SIRs) for MI were estimated using information on annual MI rates in the general population, and including multiple events during the follow-up. For both SMR and SIRs, 95% confidence limits were constructed assuming a Poisson distribution. Interaction terms between hs-cTnT concentrations and age were computed to assess age differences in associations between hs-cTnT and the outcomes.

Cox proportional hazards models were used to estimate hazard ratios with 95% CIs for all-cause mortality and MACE. Models were conducted unadjusted and adjusted for age (expressed as a restricted cubic spline, with three evenly placed knots), sex (as a categorical variable), estimated glomerular filtration rate (categorised as <30, 30–59 and ≥60 mL/min/1.73 m2), and a number of comorbidities including; prior MI, heart failure, stroke, chronic obstructive pulmonary disease, atrial fibrillation, diabetes and hypertension, all coded as individual binary factors. In addition, we also included prior treatment with the following cardiovascular medications: aspirin, P2Y12-inhibitors (clopidogrel, prasugrel, dipyramidol and ticagrelor), oral anticoagulants (Warfarin and direct oral anticoagulants (DOAC)), beta-blockers, angiotensin-converting enzyme inhibitor/angiotensin receptor blockers and statins, also coded as binary factors. Adjusted all-cause cumulative mortality and MACE, with corresponding adjusted cumulative differences were calculated using all covariates in the regression models.

All statistical analyses were conducted using SAS Statistical Analysis Software, V.9.4 (Cary, North Carolina).

Results

Study population

The final study population consisted of 1 11 916 patients, of whom 69 090 (62%) and 42 826 (38%) had peak hs-cTnT concentrations of <5 and 5–14 ng/L, respectively (table 1). Patients with hs-cTnT concentrations <5 ng/L were younger (44 vs 62 years), more likely to be women (57% vs 41%), had a lower prevalence of cardiovascular comorbidities and were less frequently treated with platelet inhibitors (5.1% vs 20%) and other cardiovascular medications compared with patients with hs-cTnT concentrations of 5–14 ng/L. A larger proportion of patients with undetectable hs-cTnT were also discharged directly from the ED (88% vs 74%) (online supplemental table 1). The most common discharge diagnosis in both groups was a symptom diagnosis, including shortness of breath and non-specific chest pain.

Table 1

Baseline characteristics in patients with low concentrations of hs-cTnT in the emergency department

Risk of death and myocardial infarction compared with the Swedish population

A total of 1769 (2.6%) and 5534 (13%) patients with hs-cTnT <5 ng/L and 5–14 ng/L died during a median follow-up of 6.4 years (IQR 4.8–8.1), with 30-day mortality being 0.03% and 0.2% in patients with hs-cTnT <5 ng/L and 5–14 ng/L, respectively (table 2). The overall standardised risk of death was almost 20% lower compared with the general population in patients with undetectable hs-cTnT concentrations (SMR 0.83, 95% CI 0.79 to 0.87), while no such difference was observed in patients with hs-cTnT of 5–14 ng/L (SMR 1.02, 95% CI 0.99 to 1.05) (table 2).

Table 2

Risks with low hs-cTnT concentrations and no initial myocardial infarction compared with the general population

The association between peak hs-cTnT and mortality differed with age (pinteraction<0.001), with peak hs-cTnT concentrations <5 ng/L being associated with lower standardised risks of death in both men and women aged ≥65 of age, while risks were similar or slightly higher than in the general population in younger age groups (figure 2, online supplemental table 2, online supplemental figure 2). A similar pattern was seen also for patients with hs-cTnT concentrations of 5–14 ng/L, where patients ≥75 years of age had a similar or lower standardised mortality compared with the Swedish population (figure 2, online supplemental table 2, online supplemental figure 2).

Figure 2

Outcomes in patients with low or undetectable hs-cTnT concentrations compared with the Swedish population stratified by sex. Upper left: standardised mortality ratios in patients with hs-cTnT concentrations<5 ng/L. Upper right: standardised mortality ratios in patients with hs-cTnT concentrations of 5–14 ng/L. Lower left: standardised incidence ratios for myocardial infarction in patients with hs-cTnT concentrations<5 ng/L. Lower right: standardised incidence ratios for myocardial infarction in patients with hs-cTnT concentrations of 5–14 ng/L. hs-cTnT, high-sensitivity cardiac troponin T; SIR, standardised incidence ratio; SMR, standardised mortality ratio.

A total of 1281 (1.9%) and 2580 (6.0%) MIs occurred in patients with hs-cTnT concentrations of <5 ng/L and 5–14 ng/L, respectively. The corresponding 30-day incidence rates were 0.03% and 0.2%, respectively (table 2). SIRs were 1.39 (95% CI 1.32 to 1.47) for hs-cTnT <5 ng/L and 1.56 (95% CI 1.50 to 1.62) for hs-cTnT 5–14 ng/L, and the highest risks were observed in younger age groups (pinteraction<0.001) (figure 2, online supplemental table 2, online supplemental figure 2). The age-stratified and sex-stratified annual mortality and MI rates in the general population during the follow-up are presented in online supplemental table 3.

Outcomes in patients with a diagnosis of unstable angina pectoris

Altogether, 879 (0.8%) of all patients had a final diagnosis of unstable angina pectoris (UAP) according to ICD-codes (code I20.0) registered in the NPR at the index visit (online supplemental table 4). Patients with UAP had similar or slightly lower standardised mortality, and higher standardised MI risk, compared with the reference population. Corresponding risks for patients without UAP were similar as in the main analyses throughout.

Long-term all-cause and cause-specific mortality

The 1-year crude cumulative mortality was 1.99% (95% CI:1.86 to 2.12) and 0.33% (95% CI 0.29 to 0.37) in the hs-cTnT 5–14 ng/L and hs-cTnT <5 ng/L group, while corresponding numbers at 5 years were 9.07% (95% CI 8.81 to 9.35) and 1.55% (95% CI 1.46 to 1.65), respectively (table 3). The overall adjusted mortality was 1.7-fold increased (HR 1.70, 95% CI 1.60 to 1.80) among patients with hs-cTnT concentrations of 5–14 ng/L, corresponding to adjusted cumulative incidence differences of 0.4% (95% CI 0.4 to 0.5), 1.9% (95% CI 1.7 to 2.1) and 3.8% (95% CI 3.3 to 4.2) at 1, 5 and 9 years of follow-up, respectively (table 3, figure 3).

Figure 3

Adjusted cumulative survival and survival free of MACE with corresponding adjusted cumulative incidence differences. (A) Adjusted cumulative survival with corresponding adjusted cumulative incidence differences. (B) Adjusted cumulative survival free of MACE with corresponding adjusted cumulative incidence differences. hs-cTnT, high-sensitivity cardiac troponin T; MACE, major adverse cardiovascular event.

Table 3

Long-term all-cause mortality and risk of major adverse cardiovascular events in patients with low hs-cTnT concentrations

A larger proportion of deaths occurring in patients with hs-cTnT of 5–14 ng/L were cardiovascular (29% vs 19%) (online supplemental table 5, online supplemental figure 3). In both groups, the most common cardiovascular cause of death was ischaemic heart disease followed by heart failure, while the most common non-cardiovascular cause of death was cancer disease.

Risk of major adverse cardiovascular events

During a median follow-up of 6.2 years (IQR 4.6–7.9), 2326 (3.3%) and 6276 (15%) first MACEs occurred in patients with hs-cTnT <5 ng/L and 5–14 ng/L, respectively (table 3). The adjusted risk of MACE associated with hs-cTnT concentrations of 5–14 ng/L was 1.6-fold (HR 1.61 95% CI 1.53 to 1.70), with adjusted difference in MACE-free survival compared with patients with hs-cTnT <5 ng/L at 1, 5 and 9 years of 0.8% (95% CI 0.7 to 0.9), 2.4% (95% CI 2.1 to 2.7) and 4.0% (95% CI 3.5 to 4.4), respectively (figure 3). The observed associations were similar in sensitivity analyses when including all-cause mortality to the composite endpoint (online supplemental table 6). The highest risks associated with measurable hs-cTnT concentrations were observed for heart failure hospitalisation and cardiovascular death, for which risks were doubled compared with unmeasurable concentrations (online supplemental figure 4).

Discussion

In a large cohort of patients presenting in the ED with chest pain but no evidence of myocardial injury, we report three major findings.

First, we found that the overall standardised risk of death was almost 20% lower in patients with undetectable hs-cTnT concentrations in comparison to the general population. This association was highly age dependent, with hs-cTnT concentrations of <5 ng/L being associated with lower risks of death in ages above 65 years in both men and women, while risks were similar or slightly higher than in the general population in younger age groups. These observations were paralleled by a correspondingly increased all-cause mortality in younger age groups, likely reflecting a very low expected mortality in these ages and where even one visit to the ED may be indicative of a poorer general health compared with the age-stratified general population. Our results on both absolute and relative risks of long-term all-cause and cardiovascular mortality associated with detectable hs-cTnT concentrations are similar to those reported by others in different study settings.15–17 The present study extends the findings from prior investigations by comparing risks with those observed in the general population and subsequently provides novel insights that could prove very useful in communicating risks with patients by specifically delineating and relating individual to age-standardised and sex-standardised long-term risks. However, whether any approach incorporating such strategy could increase patient awareness and subsequently avoidance of unnecessary healthcare resource utilisation, including revisits to the ED, remains unknown.18

Second, the standardised incidence of MI was increased by 56% and 39% in patients with hs-cTnT 5–14 ng/L and hs-cTnT<5 ng/L, respectively, as compared with the general population. Findings should, however, be interpreted with caution and weighted against the overall low absolute number of events and the extremely low associated short-term risks, with only 22 (0.03%) and 105 (0.2%) MIs occurring within 30 days in patients with hs-cTnT <5 ng/L and 5–14 ng/L, respectively. These observations further corroborate prior studies on the excellent short-term prognosis in this patient segment, allowing for a safe and rapid rule-out of MI.17 19 The increased long-term risk of MI may instead be explainable by patient characteristics constituting mutual determinants associated with both emergency healthcare utilisation and long-term cardiovascular risk. Besides established cardiovascular comorbidities and awareness of individual cardiovascular risk profile, such factors may also involve the presence of psychiatric disorders, and structural and social factors such as socioeconomic status.20 21 Furthermore, experiencing cardiac symptoms may generally be associated with future cardiovascular disease, although the most common discharge diagnosis was non-specific chest pain. Also, we speculate that mediators for exposure to emergency healthcare utilisation might also influence the probability of being diagnosed with disease in general. Finally, we cannot exclude that competing risk may have contributed to the different directions of standardised risk of MI and mortality compared with the general population.22

Third, we found that the long-term risk of death and MACE was significantly higher in patients with hs-cTnT concentrations 5–14 ng/L compared with <5 ng/L. Of note, elevated relative risks remained evident even after 5 years of follow-up and persisted for all MACE as separate endpoints, with the strongest relative risk increase observed for heart failure hospitalisation. This finding is consistent with prior knowledge on persistently detectable hs-cTnT concentrations being more closely linked to manifestations of structural and functional heart disease and the risk of future heart failure hospitalisation than by indices of atherosclerosis and the risk of ischaemic events.16 23 These observations further support the potential utility of hs-cTn measurements, in combination with established predictive models, both for improving stratification of long-term cardiovascular risks and for targeting and monitoring prevention strategies.24

Hs-cTnT measurements may be responsive to preventative interventions, for example, lipid-lowering therapy, that change in parallel to modification of risk.25 However, the clinical utility and cost-effectiveness of incorporating hs-cTn measurements in such strategies to identify patients with or at risk for subclinical cardiac dysfunction is still unknown.24 Of note, careful considerations of selecting patients suitable for further investigations are needed, considering that non-cardiovascular death was still more common than cardiovascular death in both groups of hs-cTnT concentrations, and the numbers of cardiovascular events were overall low. Considering the age differences in risks found in this study, such strategies might primarily be preferable for younger patients with chronically detectable hs-cTnT concentrations without previously known cardiovascular disease.

Strengths

The data used in this study were retrieved from Swedish national sources with high accuracy and coverage, with virtually no loss to follow-up on outcome data.6 7 Inclusion was made from several hospitals in different parts of Sweden servicing patients with different socioeconomic distributions ensuring a high generalisability. Finally, by using RETTS to identify patients eligible for inclusion, the final cohort consisted of patients with chest pain of numerous aetiologies, which further increased the generalisability of the study findings.

Limitations

Eligible patients were identified by using information on chest pain according to RETTS, but we cannot exclude potential differences in registration processes between sites that may have affected the case mix. We used data from national administrative databases with high precision of registry codes, but these sources may still involve inaccuracies such as misclassification. In parallel, medical records were not available for review, which impeded us to ascertain the correctness of MI diagnoses based on a central adjudication process. Furthermore, analytical and biological variability at the low-end of measurable hs-cTnT concentrations may have biased results by misclassification of the exposure.26

We did not have information on background characteristics such as cardiovascular comorbidities and use of cardiovascular medication in the general population, which subsequently limited the ability to further explore the implications of the observed risk differences with the study population. In parallel, we did not have data on heart failure hospitalisations or cause-specific mortality in the general population, which impeded us to explore any standardised risk differences for these outcomes. Data on medication were only available based on prescriptions and the effects of in-hospital treatment have not been adjusted for in this study. Also, we did not have information on time from symptom onset until ED presentation. However, our studies relied on real-world data based on patient management according to clinical practice including considerations of symptom duration, and subsequently we believe that the findings would be generalisable to similar clinical contexts.

Conclusions

In a large cohort of patients with chest pain in the ED, we found an up to 20% lower risk of death in patients with undetectable hs-cTnT concentrations. These risks were highly age dependent, with older patients generally having a more favourable prognosis. Furthermore, detectable hs-cTnT concentrations in the normal range were still associated with long-term risks of death and MACE.

Data availability statement

No data are available.

Ethics statements

Patient consent for publication

Ethics approval

This study involves human participants and was approved by Human Research Ethics Committee of Stockholm, Sweden 2022-06523-02. Informed consent was not required for this study according to the Human Research Ethics Committee of Stockholm due to the retrospective nature of the study and large number of study participants.

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

  • Contributors All authors made substantial contribution to the design of the study protocol and in data analysis. The manuscript was critically revised and finally approved by each author. This paper is not under consideration elsewhere and none of the content has been previously published. All authors have read and approved the manuscript and the guarantor AR accepts full responsibility for the work and the conduct of the study, had access to the data, and controlled the decision to publish.

  • Funding No specific funding was obtained for this study. The work was supported by the Stockholm County Council (20200935 to AR, 20190354 to EK), the Swedish Research Council (2017-01954 to GE) and Region Stockholm (clinical research appointment for GE). The sponsors had no role in the design or conduct of this study.

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer-reviewed.

  • 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.