Article Text

Download PDFPDF

Original research
Comparison of active tuberculosis occurrence associated with Janus kinase inhibitors and biological DMARDs in rheumatoid arthritis
  1. Yun-Kyoung Song1,
  2. Jaehee Lee2,
  3. Junwoo Jo3 and
  4. Jin-Won Kwon4
  1. 1College of Pharmacy, The Catholic University of Korea-Songsin Campus, Bucheon, Gyeonggido, Korea (the Republic of)
  2. 2Kyungpook National University School of Medicine, Daegu, Korea (the Republic of)
  3. 3Department of Statistics, Kyungpook National University, Daegu, Korea (the Republic of)
  4. 4BK21 FOUR Community-Based Intelligent Novel Drug Discovery Education Unit, College of Pharmacy and Research Institute of Pharmaceutical Sciences, Kyungpook National University, Daegu, Korea (the Republic of)
  1. Correspondence to Professor Jin-Won Kwon; jwkwon{at}knu.ac.kr

Abstract

Objectives This study aimed to evaluate the risk of tuberculosis associated with the use of Janus kinase (JAK) inhibitors or biological disease-modifying antirheumatic drugs (bDMARDs) in patients diagnosed with rheumatoid arthritis (RA) in South Korea.

Methods In this nationwide matched-cohort study, we retrospectively identified adult patients with new-onset RA from the National Health Insurance Service database who were prescribed bDMARDs or JAK inhibitors and recently underwent latent tuberculosis infection (LTBI) screening during 2012‒2021, and followed them up until the end of 2022 for the development of active tuberculosis. HRs were estimated using Cox proportional hazards regression in a propensity score-matched cohort.

Results Among 16 760 matched patients with RA (3352 JAK inhibitor users and 13 408 bDMARD users), 18.8% received tuberculosis preventive therapy for LTBI. Overall, JAK inhibitor users had a significantly lower risk of tuberculosis than bDMARD users (HR (95% CI)=0.37 (0.22 to 0.62)). Among the patients treated for LTBI, patients with low treatment adherence had a significantly higher risk than those without LTBI (HR (95% CI)=2.78 (1.74 to 4.44)). Patients without LTBI and using JAK inhibitors had a significantly lower risk of tuberculosis across all ages and sexes compared with bDMARD users.

Conclusion Patients with RA using JAK inhibitors have a significantly lower risk of active tuberculosis than bDMARD users in South Korea; however, patients with RA having LTBI are equally at risk regardless of the treatment received (JAK inhibitor vs bDMARD). Therefore, vigilant tuberculosis monitoring, especially in patients with low adherence to tuberculosis preventive therapy, is essential.

  • Arthritis, Rheumatoid
  • Tuberculosis
  • Epidemiology

Data availability statement

Data may be obtained from a third party and are not publicly available. The data that support the findings of this study are available from the Health Insurance Review and Assessment (HIRA) but restrictions apply to the availability of these data, which were used under licence for the current study, and so are not publicly available (https://opendata.hira.or.kr/).

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

  • There are growing concerns about the risk of infection, particularly tuberculosis, in patients with rheumatoid arthritis (RA) taking high-efficacy disease-modifying antirheumatic drugs (DMARDs), such as biological DMARDs (bDMARDs) or Janus kinase (JAK) inhibitors, due to disease-associated immune alteration and treatment-related immune suppression.

  • Little evidence is available about the comparative risk of tuberculosis between JAK inhibitors and bDMARDs using patients with RA in this high-to-intermediate tuberculosis burden area of South Korea.

WHAT THIS STUDY ADDS

  • Among 16 760 matched patients, JAK inhibitor users had a significantly lower risk of tuberculosis than bDMARD users.

  • Among patients with latent tuberculosis infection (LTBI), those with low adherence to tuberculosis preventive therapy had a significantly increased risk of active tuberculosis than those without LTBI.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

  • Tuberculosis risk was lower with the use of JAK inhibitors than with bDMARDs in an intermediate tuberculosis burden area like South Korea, but patients with RA with LTBI must be closely monitored for medication adherence to ensure successful RA treatment.

Introduction

Rheumatoid arthritis (RA) is a chronic autoimmune inflammatory disease that primarily affects the synovial joints; however, chronic inflammation not only causes joint failure but also a wide spectrum of comorbidities including infection.1 Pharmacological management of RA is largely via the lifetime use of disease-modifying antirheumatic drugs (DMARDs), which can be either biological or synthetic. Janus kinase (JAK) inhibitors are an important alternative to biological DMARDs (bDMARDs) for patients with poor prognostic factors who have failed initial treatment with conventional synthetic DMARDs (csDMARDs).2 3

Lately, there have been growing concerns about high infection rates, particularly for tuberculosis (TB), in patients with RA taking high-efficacy DMARDs, such as bDMARDs or JAK inhibitors, which may be attributed to both disease-associated immune alteration and treatment-related immune suppression.4–8 The recent guidelines of the American College of Rheumatology and the Asia‐Pacific League of Associations for Rheumatology have recommended TB screening in patients initiating bDMARDs or JAK inhibitors.9 10 Despite implementing comprehensive TB screening strategies, a significantly higher risk of TB (ie, 30-fold) was observed in Koreans with RA being treated with tumour necrosis factor (TNF) inhibitors than the general population during the period of 2001–2005, whereas only a 4-fold increased risk was reported in the Spanish population during the years 1990 through 2000.11 12 Therefore, it is only reasonable to examine the risk of TB associated with these drugs in countries with intermediate-to-high TB burden including South Korea than in those with low TB incidence.13 14

The risk of TB associated with JAK inhibitor use is largely dependent on the background prevalence of TB in the region where the drug is being used.6 The increase in TB incidence with JAK inhibitor use was comparable with bDMARDs in patients with RA from Western countries.15 16 In contrast, South Korea has an intermediate-to-high burden of TB than other countries in the Organization for Economic Co-operation and Development, with the total TB incidence rate (IR) being 44 per 100 000 population.17 However, there is little evidence of the comparative risk of TB associated with the use of between JAK inhibitors versus bDMARDs in patients with RA in this region, with the existing results being largely conflicting. Effective disease control using JAK inhibitors while minimising their negative effects on TB is extremely important in patients with RA; therefore, this study aimed to evaluate the risk of TB associated with the use of JAK inhibitors and bDMARDs in Korean patients diagnosed with RA.

Methods

Study design and data source

This nationwide retrospective cohort study used the South Korean national insurance reimbursement claims data recorded by the Health Insurance Review and Assessment Service (HIRA) from 2010 to 2022. This dataset was initially created for health insurance reimbursement but was eventually provided for research purposes officially.18 As an institution, the HIRA reviews medical expenses for most Korean citizens according to the social health insurance benefit guidelines (all residents in South Korea are covered by social health insurance or Medicaid).19 20 The data included information on patients’ demographics, diagnosis as per the International Classification of Diseases and Related Health Problems-10th revision (ICD-10), diagnosis or treatment procedure, prescription and medical expenses.

To ensure patient anonymity and confidentiality, raw data were not directly released to individual researcher computers, and analysis was conducted via online virtual space.

Study population

Adult patients diagnosed with RA (ICD-10 code: M05 or M06) who were recently prescribed bDMARD (bDMARD group) or JAK inhibitor (JAK inhibitor group) with more than two continuous prescriptions between 1 January 2012 and 31 December 2021 were included. The index date was defined as the first date of prescription of bDMARDs or JAK inhibitors. The bDMARDs included TNF (infliximab, etanercept, adalimumab, golimumab and certolizumab pegol) and non-TNF (rituximab, abatacept, anakinra and tocilizumab) inhibitors, while JAK inhibitors included tofacitinib, baricitinib and upadacitinib. For identifying new users of bDMARDs or JAK inhibitors, as well as detecting the presence of latent TB infection (LTBI) screening, we used a 2-year history period prior to the index period. The exclusion criteria were (1) a history of any bDMARD or JAK prescription, (2) a history of active TB diagnosis (ICD-10 code: A15−A19 or U843), (3) no history of LTBI screening (X-ray and interferon-gamma release assays (IGRAs) or Mantoux test), or (4) a history of anti-TB drug prescription, except isoniazid or rifampicin, 2 years before the index date. Patients with inaccurate information in the claims data before the index (eg, death before the index date) or those below 20 years of age on the index date were also excluded.

Exposure data

The period of exposure to bDMARDs or JAK inhibitors was determined using the prescription date and days of drug supply. The administration interval was also considered for bDMARDs based on the drug label provided by the Korean Ministry of Food and Drug Safety. If the drug discontinuation period was less than 6 months (ie, 6-month window period), it was judged as continuous administration. Patients were followed up from the index date to the following censoring events, whichever occurred first: (1) switching from JAK inhibitor to bDMARD or vice versa, (2) active TB detection and (3) end of the study (31 December 2022). The maximum follow-up period was restricted to 5 years as JAK inhibitors were launched later than bDMARDs.

Study outcome and confounding variables

The primary study outcome was active TB occurrence, which was defined as the prescription of three or more anti-TB drugs within a year from TB diagnosis with an ICD-10 code of A15−A19 or U843. For the 2-year history period prior to the index date, we recorded the following baseline characteristics potentially associated with the study outcomes and RA severity: age on the index date, sex, insurance type, presence or absence of LTBI, comorbidities and RA medications (such as csDMARDs and corticosteroids). The presence of LTBI was defined as a prescription of TB preventive therapy (TPT) given after LTBI screening with X-ray, IGRA or Mantoux testing during the 2-year history period due to the LTBI diagnosis or spontaneously healed TB history. Patients with RA with abnormal chest radiographs resulting from spontaneously healed TB (findings included non-calcified nodules with distinct margins, discrete linear or reticular fibrotic scars, and fibrotic linear opacity) without adequate treatment history were included in the LTBI treatment group; these patients suggested to receive the medication for LTBI regardless of IGRA or Mantoux test results in accordance with the Korean guideline for TB.21 In South Korea, LTBI screening is mandatory before the prescription of bDMARDs or JAK inhibitors, and patients with positive results should be prescribed TPT before initiating bDMARDs or JAK inhibitors.9 20–22 The medication possession ratios (MPRs) of LTBI drugs in this study were calculated by dividing the total number of days of drug supply by 270 days for isoniazid, 120 days for rifampin and 90 days for isoniazid/rifampin; a value of ≥80% was considered to indicate adherence.21 23

Statistical analysis

After exact matching between the groups using the presence of LTBI, propensity score (PS) matching was performed to compare TB risk between the bDMARD and JAK inhibitor groups using the following variables: age groups, sex, insurance type, individual comorbidities and individual RA medications, including csDMARDs and corticosteroids. JAK inhibitor users were matched at a 1:4 ratio to bDMARD users using the greedy calliper matching algorithm.24

Descriptive statistics were used to summarise the data. Categorical data were expressed using frequency and percentage, and continuous data as mean and SD. Fisher’s exact test or Χ2 test was used to compare categorical data, whereas the t-test was used for continuous data. The IRs and 95% CIs for the occurrence of active TB were calculated. The Kaplan-Meier curve was plotted to analyse the cumulative incidence of TB. Cox proportional hazards regression was used to estimate HRs and 95% CI after adjusting for confounding variables in the entire cohort or PS-matched cohort. Further subgroup analysis was conducted according to the presence of LTBI, bDMARD type (receptor-Fc fusion protein etanercept; monoclonal antibodies such as infliximab, adalimumab and golimumab; and non-TNF inhibitors), age (<65 and ≥65 years), sex, TB-related comorbidities, including chronic lung disease, diabetes or cancer, and prednisolone equivalent cumulative dose (≥525.5 and <525.5 mg). Lastly, sensitivity analyses were conducted to evaluate the robustness of the study results when the window period and maximum follow-up period were modified to 12 months and until the end of the study, respectively. Statistical significance was set at a two-sided p value of <0.05. All statistical analyses were conducted using SAS V.9.4 (SAS Institute).

Results

Demographic characteristics

A total of 30 142 patients with RA were prescribed bDMARDs or JAK inhibitors between 2012 and 2021, of which 4636 were excluded according to the aforementioned exclusion criteria. The eligible study cohort consisted of 25 506 patients before PS matching (figure 1). The mean age of the JAK inhibitor group (55.4±12.2 years) was significantly higher than that of the bDMARD group (50.6±14.9 years) (see online supplemental table 1).

Figure 1

Study cohort selection process. bDMARD, biological disease-modifying antirheumatic drug; JAK, Janus kinase; RA, rheumatoid arthritis; TB, tuberculosis.

After the 1:4 PS matching, all JAK inhibitor users were matched with 13 408 bDMARD users; both groups were well balanced when considering the standardised mean difference values of all matched variables to be <0.1. In the PS-matched cohort, the mean ages of JAK inhibitor and bDMARD users were 55.4±12.2 years and 54.7±12.9 years, respectively; approximately 77% of patients were aged below 65 years in both groups. Of the PS-matched study cohort, 18.8% were on TPT for the diagnosis of LTBI. Either prednisolone or triamcinolone was mainly prescribed to the patients as corticosteroids; the average cumulative prednisolone equivalent doses in the JAK inhibitor and bDMARD groups during the past 2 years were 383.6±462.5 mg and 463.6±553.4 mg, respectively. On the index date, TNF inhibitor was prescribed to 72.5% of the bDMARD patients, whereas tofacitinib or baricitinib was prescribed to 86.7% of the JAK inhibitor patients (table 1).

Table 1

Demographic characteristics of the propensity score-matched cohort of patients with RA newly prescribed JAK inhibitor or bDMARD drugs

Factors associated with the development of active TB

The use of JAK inhibitor immediately after csDMARDs tended to significantly decrease the risk compared with the use of bDMARDs with adjusted HR (95% CI) of 0.37 (0.22 to 0.62) in the PS-matched cohort. Of the patients with LTBI, only 0.9% with high adherence to LTBI drugs (n=24 of 2550) developed active TB, whereas 3.4% with low adherence (n=20 of 595) developed active TB demonstrating a significantly higher risk (adjusted HR (95% CI)=2.78 (1.74 to 4.44)). As the age of the patients or the cumulative dose of corticosteroid increased, the risk of active TB also heightened (adjusted HR (95% CI)=4.74 (2.33 to 9.63) in patients aged ≥65 years and 1.58 (1.17 to 2.13), respectively; table 2). This tendency was also observed consistently across the entire cohort (online supplemental table 2). The following factors were also associated with an increased risk of active TB: Medicaid benefits and leflunomide use as csDMARD before the prescription of bDMARD or JAK inhibitor (adjusted HRs (95% CIs)=1.72 (1.10 to 2.69) and 1.40 (1.05 to 1.87), respectively; table 2).

Table 2

Factors associated with the occurrence of TB in the propensity score-matched RA cohort classified as per JAK inhibitor or bDMARD prescription

Comparative risk of active TB between JAK inhibitor and bDMARD users

The mean follow-up times from the start of JAK inhibitor and bDMARD use to the censoring date were 2.32±1.28 years and 2.99±1.72 years, respectively, in the PS-matched cohort. Regardless of the exposure variance studied, the incidence of active TB was significantly higher in the bDMARD group than in the JAK inhibitor group (p<0.001; table 3 and figure 2A). In the JAK inhibitor group, 0.4% of the patients (n=15 of 3352) had active TB with an overall IR per 100 person-years (95% CI) of 0.19 (0.09 to 0.29), whereas 1.4% of patients who were prescribed bDMARDs immediately after treatment with csDMARDs had active TB with an IR per 100 person-years (95% CI) of 0.46 (0.39 to 0.52). The median times to onset of the infection were 51.1 and 43.3 weeks in the JAK inhibitor and bDMARD groups, respectively. The risk of active TB was significantly lower in the JAK inhibitor group compared with the bDMARD group (adjusted HR (95% CI)=0.37 (0.22 to 0.62)). The results were robust in different window periods or follow-up times.

Table 3

Comparative risks of TB occurrence between JAK inhibitors and bDMARD using patients with RA in the propensity score-matched cohort

Figure 2

Cumulative incidence of tuberculosis (TB) in the propensity score-matched RA cohort using (A) JAK inhibitor or bDMARD (B) classified according to the presence of LTBI. bDMARD, biological disease-modifying antirheumatic drug; JAK, Janus kinase; LTBI, latent TB infection; RA, rheumatoid arthritis.

In patients without LTBI, the JAK inhibitor users had a significantly reduced risk compared with the bDMARD group (figure 2B); however, when comparing patients with LTBI, both groups had a comparable risk of developing TB. These findings were also reflected in the subgroup analysis according to the LTBI status (ie, the subgroups with and without LTBI had an HR (95% CI) of 0.83 (0.36 to 1.88) and 0.24 (0.12 to 0.48), respectively; figure 3). Furthermore, within the cohort of patients with LTBI, there was no significant variance in TB risk between the JAK inhibitor and bDMARD groups in both high and low adherence to TPT subgroups. Among patients with low adherence to TPT, active TB occurred mostly within 6 months of the index date (online supplemental figure 1). Figure 3 also illustrates the relatively low TB risk in the JAK inhibitor users compared with bDMARD users in most subgroups, except for the subgroup receiving high doses of steroids. Compared with patients prescribed monoclonal antibodies against TNF (such as infliximab, adalimumab and golimumab), patients prescribed JAK inhibitors had a significantly lower risk of developing TB (HR (95% CI)=0.25 (0.15 to 0.44)). However, no significant difference was observed in the risk between the users of JAK inhibitors and non-TNF inhibitors (HR (95% CI)=0.66 (0.36 to 1.10)).

Figure 3

Subgroup analysis for active tuberculosis (TB) occurrence in the propensity score-matched RA cohort prescribed JAK inhibitors and bDMARDs. bDMARDs, biological disease-modifying antirheumatic drugs; JAK, Janus kinase; RA, rheumatoid arthritis; TNF, tumour necrosis factor.

Discussion

To the best of our knowledge, this is the first large population-based cohort study to compare the impact of JAK inhibitors and bDMARDs on the risk of active TB in routine care patients with new-onset RA from an intermediate TB burden area like South Korea. This study found that JAK inhibitor users have a significantly decreased risk compared with bDMARD users. Despite the recommendation of TB screening for patients with RA who are new or continuing users of JAK inhibitors, so far, only very limited analyses have been done to evaluate the risk associated with the use of JAK inhibitors and bDMARDs.9 10 Using a TNF inhibitor might increase the risk of TB development than using a non-TNF inhibitor or tofacitinib.25 26 However, due to the small sample size, short follow-up period and very low TB incidence in these previous studies, no statistical analysis results were obtained. Taiwan has a relative lower annual incidence of TB as compared with South Korea, with the incidence being further lower in patients with RA; however, there was no significant difference in the TB risk between tofacitinib or TNF inhibitor users.17 27 28 TNF inhibitors are known to increase the risk of active TB by promoting granuloma-inducing pathogens, inducing a differential downregulation of interferon-γ production in response to TB antigens, and exerting a distinct effect on CD8+ effector T-cells responsible for antimicrobial activity.8 29 30 Theoretically, other biological agents used for RA, such as rituximab, abatacept or tocilizumab, may cause or worsen mycobacterial disease during their use. The mechanism of tofacitinib-related TB development may be associated with the diminished production of interferon-γ and interleukin-17 by CD4+ T-cell.8

Several risk factors for TB were identified among our study patients with RA, which included older age, Medicaid benefits and high prednisolone doses. These findings are consistent with the existing literature.7 31–34 Among the csDMARDs, leflunomide use was found to be associated with increased TB risk, as observed in Swedish patients with RA.35 Patients with LTBI are a potentially large source of active TB infection, 5–10% of whom develop active TB and most cases occur within the first 5 years after the initial infection.36 In the present study, 1.4% of patients with RA on LTBI treatment developed active TB within the 5 years, and the risk was significantly higher in these patients compared with those without LTBI, which concurs with previously reported results.37 However, previous studies have reported similar TB risks were also observed between patients with RA with and without LTBI.38 39 Several factors may contribute to this discrepancy, including variations in the definition of positive LTBI, differences in LTBI screening methods and variations in medication adherence among patients with LTBI. In this study, the LTBI treatment group comprised of patients with RA with abnormal chest radiographs with spontaneously healed TB lesions without adequate treatment, as per the Korean guidelines for TB management.21 Because a history of TB was associated with 5.4 times increased risk of TB development, the inclusion of patients with a TB history may have affected the significantly increased TB risk in the LTBI treatment group in this study.31 Among patients in the LTBI treatment group, a high TB risk was mainly observed in patients with an MPR <80%, in which TB is more likely to occur during TPT. Furthermore, no significant difference was observed in the risk of TB between patients with high medication adherence and those without LTBI. It has been reported that in the general South Korean population, 2.0‒2.2% of patients with LTBI who did not take or complete the preventive therapy developed TB compared with only 0.4% of those who completed treatment.32

The TB risk associated with the use of JAK inhibitors and bDMARDs was different in the patients with LTBI and those without LTBI. Since users of these drugs had similar TB risk in subjects with LTBI, TB screening in all patients planning to use JAK inhibitors might be important considering the very high IR of TB in South Korea.7 33 In particular, increased attention should be paid to the occurrence of TB during the early treatment phase, as the incidence of TB tends to be higher in the initial stages of bDMARD or JAK inhibitor treatment in patients with LTBI. The decreased risk of TB associated with JAK inhibitors compared with monoclonal anti-TNF-α antibodies is consistent with the existing literature, which demonstrated that these TNF inhibitors predispose the patient to a higher TB risk than the other TNF inhibitors.29 In contrast, tocilizumab, a non-TNF inhibitor, was not associated with a significantly increased risk of TB.40 While there is limited information on other non-TNF inhibitors, the relatively low TB risk might be attributed to the absence of differences in TB risk associated with the use of JAK inhibitor or non-TNF inhibitors.

This study has several limitations. First, diseases were defined using diagnosis codes from the ICD-10 coding system. Because these codes may potentially lead to misclassification, the use of TB medications was also considered. Second, despite employing PS matching, differences in demographic and clinical characteristics between bDMARDs and JAK inhibitors may have existed due to unmeasured confounding factors. Lastly, the claims data do not provide information on vaccination status for TB. In South Korea, TB vaccination is mandatory, with the vaccination rate estimated at 98% in 2021.41 42

In conclusion, this study demonstrated a statistically lower risk of active TB in patients using JAK inhibitors for the treatment of RA compared with those using bDMARDs in South Korea, which is an intermediate TB burden area. Patients with LTBI using JAK inhibitors or bDMARDs with low adherence to TPT should be closely monitored for the development of TB as they face a relatively higher risk of developing TB compared with those with high adherence to TPT or patients without LTBI.

Data availability statement

Data may be obtained from a third party and are not publicly available. The data that support the findings of this study are available from the Health Insurance Review and Assessment (HIRA) but restrictions apply to the availability of these data, which were used under licence for the current study, and so are not publicly available (https://opendata.hira.or.kr/).

Ethics statements

Patient consent for publication

Ethics approval

Since all patient data were anonymised and de-identified by a randomised identification number for research purposes, the current study was exempted from ethical review by the Institutional Review Board (IRB) of Kyungpook National University (IRB no. 2022-0182, 5 July 2022).

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 Conceptualisation—Y-KS and J-WK. Data curation—J-WK and Y-KS. Formal analysis—JC. Methodology—JL, JC, Y-KS and J-WK. Project administration—J-WK. Supervision—J-WK. Writing (original draft)—Y-KS and J-WK. Writing (review and editing)—JL, JC, Y-KS and J-WK.,

    Guarantor—J-WK.

  • Funding This research was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (2022R1A2C1004822). This research was also supported by BK21 FOUR Community-Based Intelligent Novel Drug Discovery Education Unit, Kyungpook National University.

  • Competing interests None declared.

  • 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.