Background Gut microbiota composition can influence cancer immunotherapy response. Recent evidence suggests Helicobacter pylori infection may reduce immune checkpoint inhibitor (ICI) efficacy in lung cancer and melanoma, but thorough characterization of this association in patients with gastric cancer is lacking. We aimed to determine the impact of H. pylori on survival in this population.
Methods This single-center, retrospective study included all ICI-treated individuals with metastatic gastric cancer and documented H. pylori status at Memorial Sloan Kettering between July 2013 and October 2021. H. pylori-positive status was defined as history of infection obtained via breath test, stool antigen test, histopathology, and/or chart documentation. Negative status was defined as explicitly negative testing, histopathology, and/or chart documentation. Primary outcomes were progression-free survival (PFS) and overall survival (OS).
Results Of 215 included patients, 49 had documented history of H. pylori infection. Compared with H. pylori-negative patients, positive individuals tended to be younger, non-white, and Hispanic with non-cardia and intestinal-type gastric cancer. H. pylori-positive patients had significantly shorter median PFS (3.2 vs 6.8 months, HR 1.96, p<0.01) and OS (9.8 vs 17.9 months, HR 1.54, p=0.02). Multivariable analysis confirmed H. pylori infection as an independent predictor of PFS (HR 3.04, p<0.01) and OS (HR 2.24, p=0.01).
Conclusions In this largest study of its kind, H. pylori infection was associated with inferior survival in ICI-treated patients with gastric cancer. This suggests H. pylori status may be a prognostic marker of immune responsiveness. Future studies are needed to elucidate immunoregulatory mechanisms and whether treatment of active infections would improve immunotherapy outcomes.
- Immune Checkpoint Inhibitors
- Gastrointestinal Neoplasms
Data availability statement
Data are available upon reasonable request. De-identified data will be made available by the corresponding author upon reasonable request.
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
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
Helicobacter pylori has been shown to demonstrate significant immunomodulatory capacity in downregulating immune activity and upregulating programmed cell death-ligand 1 expression while also being linked to inferior survival in lung cancer and melanoma cohorts. Thorough characterization of H. pylori’s impact on immune checkpoint inhibitor (ICI) efficacy in large, diverse gastric cancer patient cohorts is lacking.
WHAT THIS STUDY ADDS
In patients with metastatic gastric cancer undergoing ICI therapy, H. pylori infection was found to be associated with significantly shorter progression-free survival and overall survival, which remained consistent when patients receiving concurrent chemotherapy were excluded, and both prior and active infection had a similar detrimental response to immunotherapy.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
In the largest study assessing the impact of H. pylori on ICI response in patients of any advanced cancer to date, H. pylori status may serve as an informative prognostic marker of immune responsiveness in patients with gastric cancer undergoing ICI therapy, especially in later treatment lines. The detection of H. pylori status may enhance the development of personalized cancer therapy approaches.
Immune checkpoint inhibitors (ICIs) via blockade of programmed cell death-1 (PD-1), programmed cell death-ligand 1 (PD-L1), and cytotoxic T-lymphocyte antigen 4 (CTLA-4) have transformed gastric cancer therapy—alone or in combination with chemotherapy.1–3 Despite these encouraging clinical prospects, only a fraction of patients derive significant benefit, immune-related adverse events are common, and predictive biomarkers capable of differentiating between potential responders and non-responders remain limited.4 Therefore, harnessing the maximal therapeutic potential from ICIs for the treatment of gastric cancer requires the evaluation of factors that may influence antitumor efficacy to determine optimal patient selection and treatment regimens. In response to this need, numerous studies have elucidated the influential role of gut microbiota composition on the effectiveness of anticancer immunotherapies.5–7 Low microbiome diversity and broad-spectrum antibiotic-induced dysbiosis have been previously found to weaken antitumoral immune activity and reduce the effect of ICIs.5 6 8
In addition to disruptions in the general microbiome population, the presence of specific bacterial strains has been previously shown to significantly influence ICI responses.5 6 9 A notable example is Helicobacter pylori (H. pylori), a bacterial pathogen present in approximately 50% of the human population and a major risk factor for gastric carcinogenesis.10 The immunomodulatory capacity of H. pylori’s virulence factors and their roles in downregulating T-cell proliferation, interfering with T-cell activation, avoiding recognition by pattern recognition receptors, and reducing pro-inflammatory and enhancing anti-inflammatory cytokine secretion have been well-established. These properties highlight H. pylori’s potential to alter systemic antitumor immune responses beyond its localized effects in the stomach.11–13 Furthermore, H. pylori’s virulence factors have not only been shown to mediate inhibition of immune activity, but also induce PD-L1 upregulation on infected gastric epithelial cells.14 Thus, H. pylori’s potential effects on immunotherapy response and its possible utility as a prognostic biomarker for predicting efficacy are of significant interest.
While recent evidence has demonstrated H. pylori-mediated decreased immune responses to ICIs in lung cancer, melanoma, and a small cohort of patients with gastric cancer,15–17 thorough characterization of H. pylori’s impact on ICI efficacy in large, diverse gastric cancer patient cohorts is lacking. Prophylactic eradication of H. pylori has been recommended to protect against gastric cancer recurrence after surgical resection,18 but the impact of prior versus active H. pylori infection on ICI treatment outcomes has yet to be assessed. In this study, we aimed to determine rates of prior and active H. pylori infection and their impact on progression-free survival (PFS) and overall survival (OS) among individuals with metastatic gastric cancer undergoing treatment with ICI.
This single-center, retrospective study included all individuals with stage IV gastric cancer undergoing treatment with ICIs and with documented H. pylori status at Memorial Sloan Kettering Cancer Center between July 2013 and October 2021. All patients had a stage IV diagnosis before ICI initiation of any treatment line. Patients with non-gastric primary tumors, including gastroesophageal junction tumors, those without documented H. pylori status in clinical history notes or pathology reports, those receiving only one dose of immunotherapy prior to death, and those lacking clinical follow-up data were excluded. Patient consent was not required due to the retrospective nature of the study.
H. pylori status
Positive H. pylori status was defined as a history of prior or active infection obtained via 13C-urea breath tests, H. pylori stool antigen tests and/or histopathology results extracted from medical charts, endoscopic procedures, and pathology reports. Prior infection was defined as positive H. pylori testing via at least one of the above modalities before gastric cancer diagnosis but explicitly negative testing at the time of cancer diagnosis. Negative H. pylori status was defined as explicitly negative infection testing via at least one of the above testing modalities at all time points. Patients for which H. pylori test results were unknown or not explicitly specified in the electronic medical record were excluded.
A combination of natural language processing and subsequent manual validation was used to comprehensively review medical charts, endoscopic procedures, and pathology reports and extract data on demographic characteristics, tumor locations, histopathological findings, germline analyses, H. pylori diagnosis and clinical outcomes. Tumor anatomic locations were classified according to corresponding International Classification of Diseases for Oncology, third edition (ICD-O-3) site codes (cardia 16.0; non-cardia 16.1–16.6; overlapping 16.8; and unspecified 16.9). Histology was primarily determined using ICD-O-3 histology codes for intestinal (8144/3, 8211/3, 8260/3) and diffuse (8142/3, 8145/3, 8490/3) subtypes. If no code was available, Lauren classification obtained from a total gastrectomy specimen or histological features biopsies from esophagogastroduodenoscopy (EGD) biopsies within the study period were used for histology classification. In these cases, intestinal histology was defined as a diagnosis of intestinal-type gastric cancer on the resection specimen or evidence of tubular, mucinous, or papillary histology on biopsy. Diffuse histology was defined as diffuse gastric cancer diagnosed from the resection specimen or signet ring cell or poorly-cohesive histology on biopsy. Positive PD-L1 expression was defined as a combined positive score ≥1. Proton pump inhibitor (PPI) usage was defined as use within≤30 days of ICI initiation.
The primary outcomes were PFS and OS for H. pylori-positive and negative groups. Secondary outcomes included PFS and OS stratified by the clinicopathologic and treatment differences between groups. Both OS and PFS were calculated from the date of ICI initiation until date of death or last follow-up for OS, or first progression, death, primary resection, or definitive radiation therapy, whichever occurred first, for PFS. Disease progression was determined via CT scans using the Response Evaluation Criteria in Solid Tumors V.1.119 or clinical progression by the treating oncologist. The end date for follow-up was July 31, 2022. All ICI regimens were initiated after stage IV gastric cancer diagnosis. An exploratory aim included evaluating differences in survival outcomes based on H. pylori infection timing of either prior to cancer diagnosis compared with on or after diagnosis.
Continuous variables were summarized as mean and SD if normally distributed and as median and IQR if not normally distributed. Categorical variables were summarized as counts and percentages. Student’s t-tests and Wilcoxon’s rank-sum tests were used to compare continuous variables while χ2 tests and Fisher’s exact tests were used for categorical variables. OS and PFS were estimated using Kaplan-Meier curves with distributions compared between groups using the log-rank test. Cox proportional risk regression models were used for multivariable analysis of predictors of OS and PFS. Results were presented as HRs with 95% CIs. For all analyses, an alpha of 0.05 was considered significant. Statistical calculations were performed using Stata Statistical Software: Release 17 (StataCorp, College Station, Texas, USA).
Patient and public involvement
No patients participated in the design of the study; however, the public is involved in dissemination of our results.
Of 215 patients with ICI-treated gastric cancer meeting the inclusion criteria, 49 (23%) had a documented history of H. pylori infection prior to or on cancer diagnosis. Demographic and cancer characteristics are summarized in table 1. Compared with H. pylori negative patients, positive patients tended to be younger (p<0.01), more likely to be of non-white (p=0.03) and Hispanic (p=0.04) backgrounds, and present with earlier stage malignancy at initial diagnosis (16% with stage I/II cancer vs 5% in H. pylori negative patients, p=0.03). The H. pylori-positive group had higher rates of non-cardia (73% vs 42%, p<0.01) and intestinal-type cancers (82% vs 65%, p=0.08) than the negative group. Eastern Cooperative Oncology Group Performance Status (ECOG PS), family history of gastric cancer, and molecular markers including PD-L1 expression, human epidermal growth factor receptor 2 (HER2) positivity, and microsatellite instability were comparable between patients with and without history of H. pylori infection.
Association of H. pylori status and survival outcomes
There were no significant differences in cancer treatment regimens, including ICI type, line or duration, concurrent chemotherapy, or primary surgery, between groups (table 2). Rates of PPI usage were not significantly different between groups. Twenty percent of patients had undergone surgical resection of their primary tumor. About one-third of patients in both groups received ICIs in the first-line setting (33% in negative patients and 31% in positive patients, p=0.61) and the median duration of therapy was 2.4 months. Fifty-four individuals (25%) received concurrent chemotherapy. All patients who only received one dose of immunotherapy prior to death were excluded from the subsequent survival analysis—in total 13 patients were excluded, all of whom were in the H. pylori-negative group.
The H. pylori-positive group had significantly shorter PFS than the negative group with an estimated median PFS of 3.2 months versus 6.8 months (HR 1.96; 95% CI 1.38 to 2.79) (figure 1A). OS was also significantly shorter in the H. pylori-positive group than negative group with an estimated median OS of 9.8 months versus 17.9 months (HR 1.54; 95% CI 1.09 to 2.19) (figure 1B).
When performing a subanalysis of only the 161 patients (37 H. pylori-positive and 124 negative) receiving ICIs alone without concurrent chemotherapy, baseline treatment characteristics were also comparable between groups (online supplemental table 1). Compared with the H. pylori negative patients, PFS continued to be shorter in H. pylori positive than negative patients (2.0 months vs 4.0 months; HR 1.93; 95% CI 1.28 to 2.89) (figure 2A). OS was also persistently shorter in the positive group when excluding those receiving concurrent chemotherapy (6.2 months vs 16.7 months; HR 1.63; 95% CI 1.09 to 2.45) (figure 2B).
Association of H. pylori status and survival outcomes by ICI line
Next, we assessed potential associations between H. pylori infection status and survival outcomes when stratified by ICI treatment line as summarized in online supplemental table 3. In those receiving first-line ICI therapy, H. pylori-positive patients had a median PFS of 6.9 months compared with 13.3 months in the H. pylori-negative group (HR 1.23; 95% CI 0.60 to 2.50) (figure 3A). Most of the individuals treated with ICIs in the first-line setting received concurrent chemotherapy (n=47). Notably, when conducting a subgroup analysis of only patients receiving first-line ICI therapy and concurrent chemotherapy, the H. pylori-positive had a significantly shorter median PFS of 4.9 months versus 11.7 months in the H. pylori-negative group (HR 2.43; 95% CI 1.07 to 5.51; online supplemental figure 1). Due to limited sample size, there was insufficient data to assess PFS in those treated in the first-line setting without concurrent chemotherapy.
PFS was shorter for both H. pylori-positive and negative groups in the setting of increasing ICI treatment line. Compared with H. pylori-negative patients, H. pylori-positive patients also had significantly shorter PFS on second-line (2.0 months vs 4.3 months; HR 2.93; 95% CI 1.21 to 7.10), third-line (2.0 months vs 4.3 months; HR 2.80; 95% CI 1.42 to 5.53), and ≥fourth-line (1.0 months vs 2.6 months; HR 2.00; 95% CI 1.11 to 4.00) ICIs (figure 3B–D). Similar trends of decreased OS in patients receiving later-line ICI therapy compared with first-line therapy were noted in both groups, with shorter OS also seen for H. pylori-positive patients in each ICI line (figure 4B–D), however these were only significant for the third-line subgroup. These findings stratified by treatment line persisted when patients receiving concurrent chemotherapy were excluded.
Predictors of survival
Multivariable Cox regression analysis was used to evaluate prognostic features for PFS and OS in the overall cohort. The multivariable model confirmed H. pylori status as being independently associated with shorter PFS (HR 3.04; 95% CI 1.67 to 5.54) and OS (HR 2.24; 95% CI 1.26 to 3.98) (table 3). Compared with first-line usage, later ICI treatment line was found to have progressively stronger associations with decreased PFS and OS. Another independent predictor of shorter PFS and OS included ECOG PS ≥2. In contrast, factors associated with significantly longer survival included positive PD-L1 status (HR 0.58; 95% CI 0.27 to 0.91 for PFS and HR 0.59; 95% CI 0.21 to 0.95 for OS) and positive HER2 status (PFS; HR 0.32, 95% CI 0.15 to 0.71 for PFS and HR 0.45; 95% CI 0.21 to 0.95 for OS). Microsatellite instability-high tumors were also linked to prolonged PFS (HR 0.21; 95% CI 0.15 to 0.71) and OS (HR 0.37; 95% CI 0.13 to 0.70). Neither PPI usage nor concurrent chemotherapy were found to be associated with PFS or OS.
Association of H. pylori timing and survival outcomes
In an exploratory analysis, we sought to evaluate the association between timing of H. pylori infection and survival outcomes. There were no differences in baseline demographic or cancer characteristics between the 27 patients with documented H. pylori infection before cancer diagnosis (prior infection group) and the 22 individuals with infections on or after diagnosis (active infection group) (online supplemental table 2). Data on H. pylori eradication and treatment was incomplete for several patients in each group, and so this characteristic was not included in the analysis. Gastric cancer treatment characteristics were also balanced between groups. Compared with the H. pylori-negative group, both the prior infection group and active infection group had significantly shorter PFS (2.0 months vs 6.8 months; HR 2.63; 95% CI 1.69 to 4.09; and 3.3 months vs 6.8 months; HR 1.48; 95% CI 1.01 to 2.42, respectively) (online supplemental figure 2A). Similarly, both the prior and active infection groups also had shorter OS compared with the H. pylori-negative patients (7.1 months vs 17.9 months; HR 1.71; 95% CI 1.11 to 2.64; and 9.8 months vs 17.9 months; HR 1.36; 95% CI 0.84 to 2.22, respectively) though the difference was only significant in the prior infection group (online supplemental figure 2B).
In the largest study assessing the impact of H. pylori on ICI response in patients of any advanced cancer to date, H. pylori infection was found to be associated with significantly shorter PFS and OS in patients with metastatic gastric cancer undergoing ICI therapy. The inferior survival noted in the overall and treatment line-specific cohorts persisted when patients receiving concurrent chemotherapy were excluded. Both prior and active infection with H. pylori had a similar detrimental response to ICI therapy. H. pylori infection may therefore serve as an important prognostic marker of immune responsiveness in patients, and future studies are needed to elucidate the specific mechanisms by which this immune regulation occurs and whether treatment of active infections would benefit immunotherapy outcomes.
While increased PD-L1 expression secondary to H. pylori infection was initially theorized to be a potentially favorable factor in individuals treated with ICIs,20 recent evidence suggests that H. pylori infection may adversely affect outcomes in this patient population.15–17 Our observation of decreased PFS and OS in H. pylori-positive patients compared with negative patients is consistent with recent evidence in populations of various cancer types. A study by Oster et al showed H. pylori-infected mice displayed reduced immune responses to ICIs compared with non-infected mice, and that H. pylori seropositivity was associated with shorter PFS and OS rates in humans with non-small-cell lung cancer.15 Another study in a cohort of 97 patients with advanced melanoma (21H. pylori-positive and 76 negative individuals) reported significantly shorter OS in H. pylori-positive patients treated with ICIs, further supporting this H. pylori-mediated immune blunting. They further observed that this negative impact seemed to occur independently of fecal microbiome composition.16 Finally, a smaller cohort of 77 patients with advanced gastric cancer (34H. pylori-positive and 43 negative individuals) also demonstrated inferior PFS and OS outcomes for H. pylori-positive individuals.17
In addition to validating these prior observations in the largest cohort to date, our study offers additional novel insights into the association between survival outcomes and H. pylori status. First, we demonstrate a persistent difference in outcomes between the H. pylori-positive and negative cohorts after excluding patients receiving concurrent chemotherapy. This finding suggests that improved survival in the negative cohort was mediated by a more efficacious response to ICIs, rather than response to concurrent chemotherapy. Moreover, establishing the independence of this association from the influence of chemotherapy is especially notable as previous studies have shown that chemotherapy-induced microbiome perturbations can decrease bacterial diversity thus permitting pathobionts to thrive, with such changes linked to increased immunotherapy toxicity and also modulation of drug efficacy which could further confound survival outcomes.5 6 21 22 Second, our findings of significantly worse PFS and OS in second or later lines of ICI therapy reciprocate existing studies conducted in a wide range of solid tumors.23 24 We also observed that H. pylori-positive patients treated in the first-line setting with concurrent chemotherapy had significantly worse survival. Moreover, after adjusting for line of therapy and use of concurrent chemotherapy the inferior outcomes of H. pylori-positive patients remain consistent. Taken together, our data suggests that screening for H. pylori infection may serve as an independent prognostic marker in this population.
Moreover, our study highlights a critical question that remains to be answered: if H. pylori infection is associated with worse survival outcomes, is there an advantage to treating active infection in patients with advanced cancer at the time of ICI initiation? In our study, both prior and active H. pylori infections were associated with inferior survival outcomes compared with the H. pylori-negative group. While we saw no meaningful difference between the active versus prior infection groups, our sample size was limited and the outcomes we observed may be influenced by factors such as the timing of the prior infection, timing of treatment, and the type of regimens used for eradication, which we were not able to assess due to limited data. Our observations are congruent with Oster et al’s mouse model findings, which demonstrated that eradication of H. pylori infection by antibiotic therapy did not revert the H. pylori-induced hyporesponsiveness to vaccine-based cancer immunotherapy.15 Nonetheless, prospective studies are needed to validate whether the timing of H. pylori infection plays a more influential role than the absolute presence of infection, particularly since multiple factors can influence immune responses in the clinical setting. For example, previous studies demonstrate that broad spectrum antibiotic-induced dysbiosis and low microbiome diversity can weaken antitumoral immune activity of ICIs and significantly shorten PFS and OS.5 6 8 25
The role of adjunctive therapies such as PPIs in suppressing H. pylori and mitigating its potentially detrimental effects is also unknown. Existing data regarding the impact of PPIs on immunotherapy outcomes are mixed with several studies suggesting a negative impact of PPIs on survival,26 27 while others have observed no impact.28 29 Our current study adds to the existing literature in observing no significant differences in baseline PPI usage between H. pylori groups nor associations of PPIs with survival outcomes. Moreover, the value of treating active H. pylori infection in patients with advanced cancer will need to be weighed against patient quality of life and medication adherence, given the large pill burden and potential side effects of treatment. Such tradeoffs may be worthwhile if treatment leads to meaningful survival benefits. Therefore, additional prospective studies are urgently needed to clarify the differential impact of prior and active H. pylori infection on immunotherapy responsiveness and to investigate the therapeutic impact of treating active infection. Finally, the development of hyperprogressive disease, defined as ≥twofold increase of the tumor growth rate before and after ICI therapy, has been identified in 20.5% of patients with ICI-treated gastric cancer in a recent large retrospective cohort30 and has been linked to with inferior survival. Dedicated future investigations are needed to thoroughly evaluate a potential association between H. pylori infection and an increased risk for developing hyperprogressive disease.
There are several limitations to this study. Due to its retrospective design, our results are unable to demonstrate causal relationships between H. pylori infection and decreased ICI efficacy. Further prospective studies are needed to validate our findings that H. pylori is associated with reduced benefit from ICIs and elucidate potential underlying mechanisms. This may reveal whether this association is due to ICI hyporesponsiveness induced by the bacterium itself or alternative explanations such as an adaptive immune deficit that tolerates H. pylori and leads to inferior ICI responses, host deficiencies in nutritional status or native microbiome diversity, or antibiotic-induced dysbiosis from prior eradication. While this is the largest study to date, we also recognize that sample size limited our ability to conduct more extensive subgroup analyses to better understand the impact of infection on ICI efficacy.
In conclusion, H. pylori infection was found to be associated with shorter PFS and OS in patients with advanced gastric cancer undergoing ICI therapy. This suggests that H. pylori status may serve as an informative predictive marker of immune responsiveness in this patient population and enhance the development of personalized cancer therapy approaches. Future studies are needed to elucidate the specific mechanisms by which this immune regulation occurs and whether treatment of active infections, even in the metastatic setting, would benefit immunotherapy outcomes.
Data availability statement
Data are available upon reasonable request. De-identified data will be made available by the corresponding author upon reasonable request.
Patient consent for publication
This study was approved by the Memorial Sloan Kettering Cancer Center Institutional Review Board - 19-205A(6).
Twitter @cowzerdarren, @DavidFaleckMD
Contributors PTM, SBM, DF, ML contributed to study concept and design. PTM, SBM, DC, SK, DF, ML acquired the data. PTM, SBM, DF, ML analyzed and interpreted the data. PTM, SK, ML performed statistical analyses. PTM, DF, ML drafted the manuscript. PTM, SBM, DC, SK, MS, YJ, DF, ML revised the manuscript. DF and ML supervised the study and are the guarantors of the manuscript. All authors approved of the manuscript for submission.
Funding National Cancer Institute of the National Institutes of Health - U01 CA265729; NIDDK of the National Institutes of Health - K08 DK125876; NIH/NCI Cancer Center - P30 CA008748.
Competing interests SBM has received research funding from Guardant Health, and consulted for Basilea, Bicara, Daiichi-Sankyo, Elevation Oncology, Purple Oncology, Amgen, Novartis, and Natera. MS serves on the advisory board for Boston Scientific, Mirai Medical, and Novo Nordisk. YJ has stock or ownership interest in RGENIX; has a consulting or advisory role with AstraZeneca, Basilea Pharmaceutica, Bayer, Bristol Myers Squibb, Daiichi Sankyo, Eli Lilly, Imugene, Merck Serono, Merck, Michael J Hennessy Associates, Paradigm Medical Communications, Pfizer, RGENIX, Seagen and Zymeworks; and has received research funding from Bayer, Bristol Myers Squibb, Cycle for Survival, Department of Defense, Eli Lilly, Fred’s Team, Genentech/Roche, Merck, NCI and RGENIX. DF has received consulting fees from AzurRx, Equillium, Kaleido Biosciences, Mallinckrodt Pharmaceuticals, and OnQuality Pharmaceuticals. All other authors have no conflicts of interest to disclose.
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.