Article Text
Abstract
Background Endometrial cancers with more than one molecular feature—POLE mutations (POLEmut), mismatch repair protein deficiency (MMRd), p53 abnormality (p53abn)—are called ‘multiple classifiers’.
Objective To describe our cohort of multiple classifiers and to report the results of a review on their incidence and the techniques used to identify them.
Methods Multiple classifiers identified at the European Institute of Oncology, Milan, between April 2019 and Decmber 2022, were included. Clinicopathological, molecular characteristics, and oncologic outcomes were summarized and compared between single and multiple classifiers sharing common features. Studies on molecular classification of endometrial cancer were searched in the PubMed Database to collect data on the incidence of multiple classifiers and the techniques used for classification.
Results Among 422 patients, 48 (11.4%) were multiple classifiers: 15 (3.6%) POLEmut-p53abn, 2 (0.5%) POLEmut-MMRd, 28 (6.6%) MMRd-p53abn, and 3 (0.7%) POLEmut-MMRd-p53abn. MMRd-p53abn and MMRd differed in histotype (non-endometrioid: 14.8% vs 2.0%, p=0.006), grade (high-grade: 55.6% vs 22.2%, p=0.001), and MMR proteins expression, whereas they differed from p53abn in histotype (non-endometrioid: 14.8% vs 50.0%, p=0.006). POLEmut-p53abn and POLEmut differed only in grade (high-grade: 66.7% vs 22.7%, p=0.008), while they differed from p53abn in age (56.1 vs 66.7 years, p=0.003), stage (advanced: 6.7% vs 53.4%, p=0.001), and histotype (non-endometrioid: 6.7% vs 50.0%, p=0.002). Two (7.1%) patients with MMRd-p53abn, 4 (4.0%) with MMRd, and 25 (34.3%) with p53abn had a recurrence. No recurrences were observed in POLEmut-p53abn and POLEmut. TP53 sequencing allowed the detection of additional 7 (18.9%) multiple classifiers with normal p53 immunostaining. The incidence of multiple classifiers ranged from 1.8% to 9.8% in 10 published studies including >100 patients. When only p53 immunohistochemistry was performed, the highest incidence was 3.9%.
Conclusions The characteristics of POLEmut-p53abn resembled those of POLEmut, whereas MMRd-p53abn appeared to be intermediate between MMRd and p53abn. The high proportion of multiple classifiers may be related to the methods used for molecular classification, which included both p53 immunohistochemistry and TP53 sequencing.
- Endometrial Neoplasms
- Pathology
- Uterine Cancer
- Genital Neoplasms, Female
- Neoplastic Processes
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Multiple classifiers are endometrial cancers with more than one molecular feature (POLE mutation (POLEmut), MMR deficiency (MMRd), or p53 abnormality (p53abn)). Previous studies reported an incidence of 3–6% of all endometrial cancers. Based on preliminary data from a few retrospective studies, patients with POLE mutation are classified as POLEmut even in the presence of MMRd and/or p53 abnormality, while MMRd-p53abn are classified as MMRd.
WHAT THIS STUDY ADDS
In our prospective cohort, 11% of endometrial cancers were multiple classifiers. This finding, probably due to the thorough molecular analysis that included an assessment of both p53 expression and TP53 mutations, underscores the critical need to clarify the role of multiple classifiers. MMRd-p53abn had characteristics that appeared to be intermediate between those of MMRd and p53abn, whereas POLEmut-p53abn was similar to POLEmut endometrial cancer.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
Collaborative studies worldwide should aim to understand the role of multiple classifiers in predicting outcomes in endometrial cancer. The hierarchical classification approach may hinder an accurate identification of multiple classifiers, which instead requires simultaneous molecular analysis.
INTRODUCTION
The Cancer Genome Atlas Research (TCGA) Network classified endometrial cancers into four categories: DNA polymerase epsilon (POLE)/ultramutated, microsatellite instable/hypermutated, copy-number low, and copy-number high.1 Despite its prognostic significance, the application of this classification was impractical due to the complex and expensive technologies required. To bring it into practice, the TransPORTEC and Vancouver groups proposed surrogate biomarkers that could mimic the TCGA.2–7 Accordingly, POLE ultramutated can be identified by POLE exonuclease domain sequencing, microsatellite instable/hypermutated by mismatch repair protein immunohistochemistry(MMRd), and copy number high by p53 immunohistochemistry(p53abn). Endometrial cancers lacking these features are called ‘no specific molecular profile’ and resemble the copy-number low group. POLE exonuclease domain sequencing identifies all POLE ultramutated and MMR immunohistochemistry has a high concordance with the microsatellite instability assay. Although p53 immunostaining/TP53 mutation status is not equivalent to the copy-number high subgroup, it can identify cases with significantly worse outcomes and its clinical usefulness has already been accepted.3 8
Nowadays, guidelines recommend the use of this classification along with traditional histopathologic risk factors to better characterize endometrial cancer,9–13 and as a way to risk-stratify patients and to inform decisions about post-operative treatment.9 11 13
Most tumors are easily classified into one of the four categories due to the presence of only one molecular feature (hereafter referred to as ‘single classifiers’). Conversely, some tumors have more than one feature and are called ‘multiple classifiers’
In this study, we primarily aimed to describe the clinicopathological and molecular characteristics and the oncologic outcomes of our cohort of multiple classifiers. A secondary objective was to compare the characteristics of multiple classifiers with those of single classifiers that shared at least one molecular feature. Furthermore, we conducted a review of the literature on the incidence of multiple classifiers and the techniques used for molecular classification.
METHODS
Case Selection
From the prospectively collected database of endometrial cancers treated at the European Institute of Oncology, Milan, Italy, we identified consecutive cases with molecular analysis (performed systematically on all endometrial cancers since April 2019). Patients were included if they had undergone surgical staging between April 2019 and December 2022, regardless of histologic and clinical characteristics, and if POLE, microsatellite instability/MMR, and p53/TP53 status were known. Patients not consenting to data use for clinical research or not undergoing surgical staging were excluded. A subset of this cohort has been described in other studies.14 15
Molecular Analyses and Classification
POLE and TP53 mutations were identified by next-generation sequencing using a panel of 26 cancer-related genes. POLE mutations (exons 9, 13, 14) were classified according to the literature,16 whereas TP53 mutations were evaluated using the COSMIC, ClinVar and cBioPortal Databases. MMR immunohistochemistry was classified as proficient if MSH6, PMS2, MSH2, and MLH1 were expressed, MMRd if at least one protein expression was lost, and equivocal in cases of equivocal staining. The Idylla microsatellite instability assay (Biocartis, Mechelen, Belgium) for seven microsatellite regions (ACVR2A, BTBD7, DIDO1, MRE11, RYR3, SEC31A, SULF2) was also performed. In cases of discrepancy between immunohistochemistry and the Idylla assay, the Promega microsatellite instability analysis system (version 1.2) was used. A tumor was classified as MMRd if either MMR proteins were not expressed or microsatellite instability assessment was positive. Immunohistochemistry for p53 was classified as normal/wild type, aberrant (overexpression, null, cytoplasmic), or subclonal according to the literature.4 17 18 A tumor was considered p53abn if either p53 staining was aberrant or TP53 gene harbored a pathogenic/probably pathogenic mutation. Cases without any molecular features were considered as no specific molecular profile. The ProMisE and TransPORTEC molecular algorithms omitted microsatellite instability assessment and TP53 sequencing, both of which we performed on all patients.2 3
Cases were then classified as single classifiers (POLEmut, MMRd, p53abn, no specific molecular profile) or multiple classifiers (POLEmut-MMRd, POLEmut-p53abn, MMRd-p53abn, POLEmut-MMRd-p53abn).
Clinicopathological Characteristics and Oncologic Outcomes
Age at surgery and body mass index, International Federation of Gynecology and Obstetrics (FIGO) 2009 stage (I–II vs III–IV) and grade (G1–2 vs G3), histotype (endometrioid vs non-endometrioid), myometrial invasion (none vs <50% vs ≥50%), lymphovascular space invasion absent/focal vs diffuse), lymph node metastases (negative vs isolated tumor cells vs micro/macrometastases), risk groups according to the European Societies of Gynecological Oncology, Radiotherapy and Oncology, and Pathology (ESGO/ESTRO/ESP) guidelines, dates of first recurrence, and last follow-up were collected.9 19 20
The study was deemed exempt from ethical approval by the European Institute of Oncology ethics committee (UID2418). In accordance with journal guidelines, our data are available for independent analysis in the online supplemental material. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.21
Supplemental material
Statistical Analysis
Differences between categorical variables were assessed using the Χ2 or Fisher’s exact tests, with Bonferroni correction for multiple comparisons.22 One-way analysis of variance with Bonferroni correction was used for differences between continuous variables. Multiple pairwise comparisons were performed when a significant difference (p<0.05) was found between three subgroups. Median time to recurrence and follow-up and disease-free survival with 95% CI were estimated through the Kaplan-Meier method. A log-rank test was used to compare survival among groups. A p value <0.05 was considered significant. Analyses were performed with Stata Statistical Software: Release 17 (StataCorp. 2021. College Station, Texas, USA: StataCorp LLC).
Literature Review
The PubMed Database was searched. Articles were included if they were published in English in the last 10 years before August 4, 2022, and included endometrial cancer molecular classification. The search strategy was designed by an experienced librarian with input from the study authors. The actual approach, listing all search terms used and how they were combined, is available in the online supplemental appendix. Study screening and data extraction were performed by two independent reviewers (LADV and IB).
RESULTS
Between April 2019 and December 2022, 498 patients were treated at our institution. Among them, 422 met all the eligibility criteria, including complete molecular analysis: 374 (88.6%) single classifiers and 48 (11.4%) multiple classifiers. Specifically, 22 (5.2%) POLEmut, 99 (23.5%) MMRd, 180 (42.7%) no specific molecular profile, 73 (17.3%) p53abn, 2 (0.5%) POLEmut-MMRd, 15 (3.6%) POLEmut-p53abn, 28 (6.6%) MMRd-p53abn, 3 (0.7%) POLEmut-MMRd-p53abn (online supplemental table 2).
MMRd-p53abn
MMRd-p53abn was the largest group with 28 cases. They were mainly early-stage (60.7%), endometrioid (85.2%), high-grade (55.6%), invading<50% of the myometrium (74.1%), without diffuse lymphovascular space invasion (81.5%), and without nodal involvement (74.1%) (table 1). Based on their histomorphologic features, 53.6% were classified as low or intermediate risk according to ESGO/ESTRO/ESP guidelines.
Comparison of MMRd-p53abn with MMRd and p53abn
Compared with MMRd, MMRd-p53abn were more likely to be non-endometrioid (14.8% vs 2.0%, p=0.006) and high-grade (55.6% vs 22.2%, p=0.001). Compared with p53abn, multiple classifiers were less likely to be non-endometrioid (14.8% vs 50.0%, p=0.001) (table 1).
In terms of molecular characteristics, expression of MMR proteins differed significantly between MMRd-p53abn and MMRd. Multiple classifiers showed a lower proportion of abnormal expression of MLH1 (44.4% vs 71.4%, p=0.010) and PMS2 (44.4% vs 78.4%, p<0.001) and a higher proportion of MSH6 loss (55.6% vs 19.4%, p=0.001). Compared with p53abn endometrial cancer, p53 subclonal pattern was more frequent in MMRd-p53abn (18.2% vs 2.9%, p=0.005). Five (17.9%) MMRd-p53abn had more than one TP53 mutation, in contrast to only 3 (4.1%) p53abn.
Two (7.1%) recurrences were observed in MMRd-p53abn at 5.0 and 6.9 months after surgery, while MMRd and p53abn recurred in 4 (4.0%) and 25 (34.3%) cases, respectively, with a median time to recurrence of 8.8 and 8.4 months (online supplemental figure 2). Three-year disease-free survival was 87.8% (95% CI 59.6 to 96.8), 91.9% (95% CI 78.7 to 97.0), and 38.0% (95% CI 20.2 to 55.8) for MMRd-p53abn, MMRd, and p53abn, respectively (p<0.001). The median follow-up of patients without recurrences was 4.2, 11.6, and 10.8 months in MMRd-p53abn, MMRd, and p53abn, respectively.
POLEmut-p53abn
POLEmut-p53abn accounted for 15 cases. Most were early-stage (93.3%), endometrioid (93.3%), high-grade (66.7%), invading <50% of the myometrium (66.6%), and without diffuse lymphovascular space invasion (93.3%) or nodal involvement (100.0%) (table 2). Based on histomorphologic features, 66.6% were low or intermediate risk according to ESGO/ESTRO/ESP guidelines.
Comparison of POLEmut-p53abn with POLEmut and p53abn
POLEmut-p53abn differed from p53abn in age at surgery (56.1 vs 66.7 years, p=0.003), advanced-stage (6.7% vs 53.4%, p=0.001), and non-endometrioid histotype (6.7% vs 50.0%, p=0.002), showing similar characteristics to POLEmut tumors. The only exception was grade, as POLEmut-p53abn differed significantly from POLEmut (high-grade: 66.7% vs 22.7%, p=0.008) but not from p53abn (71.2%, p=0.724). Although not significant, most POLEmut and POLEmut-p53abn had no nodal metastases, whereas p53abn had 22.6% positive nodes.
In POLEmut-p53abn, POLE variants were 7 (46.7%) P286R, 4 (26.7%) V411L, 2 (13.3%) S297F, 1 (6.7%) S459F, 1 (6.7%) A456P. Similarly, P286R and V411L were the most common in POLEmut, accounting for 14 (63.6%) and 5 (22.7%) cases, respectively. POLEmut-p53abn showed fewer aberrant and more subclonal p53 stainings compared with p53abn endometrial cancers (p<0.001).
No recurrences were observed in 15 POLEmut-p53abn and 22 POLEmut during a median follow-up of 12.8 and 12.6 months, respectively (online supplemental figure 3).
POLEmut-MMRd and POLEmut-MMRd-p53abn
Two POLEmut-MMRd were both endometrioid, stage IA, low-grade, without lymphovascular space invasion. Isolated tumor cells in sentinel lymph nodes were identified in one of them. POLE mutations were S297F and P286R.
We identified three early-stage, high-grade, endometrioid or mixed histotypes POLEmut-MMRd-p53abn with six different TP53 mutations and a unique POLE mutation (V411L).
After a median follow-up of 17.0 months, no recurrences were observed in either of these two groups.
TP53 Sequencing and p53 Immunohistochemistry Discordance
TP53 sequencing and p53 immunostaining were both available in 364 (86.3%) cases with conflicting results (ie, one wild-type and the other abnormal) in 44 (12.1%) cases: 32/327 (9.8%) single and 12/37 (32.4%) multiple classifiers (online supplemental table 3). Sequencing allowed the identification of 7 (18.9%) additional cases of multiple classifiers that had normal p53 immunohistochemistry: 3 POLEmut-p53abn and 4 MMRd-p53abn.
Microsatellite Instability Assay and MMR Immunohistochemistry Discordance
Microsatellite instability assay and MMR immunostaining were both available in 395 (93.6%) cases and gave conflicting results (ie, MMRp-microsatellite instable or MMRd-microsatellite stable) in 20 (5.1%) cases: 16/355 (4.5%) single and 4/40 (10.0%) multiple classifiers (online supplemental table 4). MMR immunostaining allowed identification of 4 (10.0%) additional cases of multiple classifiers that were microsatellite stable: 2 MMRd-p53abn, 1 POLEmut-MMRd, and 1 POLEmut-MMRd-p53abn. Microsatellite instability assay did not identify any additional patients with normal MMR protein expression.
Literature Review
The PubMed search returned 192 records. Screening of the reference lists of the included publications identified four additional studies. We excluded 146 records based on titles and abstracts and 34 after full text screening. Ultimately, 16 published records from 15 studies were available for the analysis.1–6 16 23–31 The study flow diagram is reported in online supplemental figure 1.
All the records were published between 2013 and 2021, after the publication of the TCGA report.1 Fourteen studies were retrospective, 10 single-center and four multicenter. One study was described as single-center prospective.29 The largest study by León-Castillo et al16 23 pooled data from the PORTEC, Vancouver, TCGA, and two independent cohorts.1 4–6 32 33 Eight groups included all comers’ endometrial cancers, whereas the remaining analyzed high-risk populations (table 3).
Diagnostic Tests
Included studies performed Sanger or next-generation sequencing (exons 9, 13, 14, or 9–14) for POLE. MMR status was assessed by immunohistochemistry only in nine studies, microsatellite instability only in three and a combination of the two techniques in three studies. p53 status was assessed by immunohistochemistry only in seven studies, sequencing only in one and a combination of the two techniques in seven studies.
Incidence of Multiple-Classifiers
The incidence of multiple classifiers ranged from 1.8% to 14.3% (figure 1). In 10 studies with more than one hundred patients, the incidence was between 1.8% and 9.8%.1 6 Among the four studies that assessed p53 status by immunohistochemistry alone in more than 100 patients, the highest incidence was 3.9%.24 Conversely, the incidence was between 2.6% and 9.8% when TP53 sequencing was performed. The most frequent subgroup was MMRd-p53abn, followed by POLEmut-MMRd, POLEmut-p53abn, and POLEmut-MMRd-p53abn (figure 1 and table 3).
DISCUSSION
Summary of Main Results
In our cohort, approximately 1/10 endometrial cancer was a multiple classifier. The clinicopathological characteristics of MMRd-p53abn were intermediate between those of p53abn and MMRd for histotype, grade, and MMR proteins expression. POLEmut-p53abn resembled POLEmut in clinicopathological, molecular characteristics, and oncologic outcomes. Subclonal p53 expression was more frequent in multiple classifiers than in p53abn single classifiers.
Results in the Context of Published Literature
This is the first attempt to compare clinicopathological and molecular characteristics between single and multiple classifiers in a cohort of consecutive patients. With the exception of León-Castillo et al, who analyzed 137 multiple classifiers and will be the focus of our discussion, other studies identified only 12 or fewer cases.16 23
León-Castillo et al described 64 MMRd-p53abn as predominantly endometrioid, high-grade, early-stage, without lymphovascular space invasion, and with a high proportion of MSH6±MSH2 or single PMS2 loss. They also revealed a higher frequency of multiple TP53 mutations and subclonal p53abn staining compared with single-classifier p53abn. For genomic features and survival outcomes, MMRd-p53abn was more similar to single-classifier MMRd than to p53abn. The 5-year recurrence-free survival of stage I MMRd-p53abn was significantly different from that of p53abn (92.2% vs 70.8%). However, they did not compare MMRd-p53abn survival with MMRd. These findings are consistent with our results, but we also found a difference in grade and histology between MMRd-p53abn and MMRd, which should be further analyzed in subsequent studies.
Our results on POLEmut-p53abn are in agreement with those of Leon-Castillo et al. They described 31 cases as mostly endometrioid, early-stage, high-grade, without lymphovascular space invasion, and characterized by p53 subclonality and multiple TP53 mutations. POLEmut-p53abn clustered mainly with POLEmut rather than p53abn. Accordingly, 5-year recurrence-free survival was significantly different from p53abn when the analysis was limited to stage I disease (94.1% vs 70.8%). Similarly, our subset of POLEmut-p53abn showed an excellent prognosis and a high similarity to POLEmut in terms of clinicopathological and molecular characteristics.
Leon-Castillo et al also described 14 POLEmut-MMRd with a probably pathogenic POLE mutation, which had a 5-year recurrence-free survival of 92.3%. In addition, 12 endometrial cancers belonged to POLEmut-MMRd-p53abn, which were predominantly early-stage endometrioid, but two were classified as mixed histology. All patients showed p53 subclonality, and most cases clustered with POLEmut. In our cohort only two POLEmut-MMRd and three POLEmut-MMRd-p53abn were found. Therefore, we cannot draw any conclusions about these rare groups.
The high frequency of multiple classifiers found in our cohort is probably the result of the methodology used for molecular analyses.34 The combination of TP53 sequencing and p53 staining has increased the detection of multiple classifiers with p53 abnormalities, as occurred in other studies included in the review (figure 1 and table 3). Indeed, in our study, sequencing allowed the detection of an additional 18.9% of multiple classifiers that had a normal immunohistochemistry for p53.
TP53 mutation is probably a non-driver mutation in multiple classifier, caused by the POLE mutation and its ultramutated phenotype.17 In fact, one in three POLEmut-p53abn showed subclonal p53 expression, potentially indicating a later occurrence of TP53 mutation in tumor development, unable to affect the immunohistochemical pattern ubiquitously and resulting in a subclonal pattern.17 18 Similarly, TP53 mutations found in MMRd-p53abn are probably non-driver. However, based on our results, TP53 mutations seemed to influence the tumor characteristics of MMRd-p53abn more than those of POLEmut-p53abn. Furthermore, discordance between p53 immunohistochemistry and TP53 mutations is higher in multiple (32.4%) than in single classifiers (9.8%). Our findings are consistent with the results of Vermij et al who reported that 22/32 cases with discordance between p53 immunohistochemistry and TP53 gene status were either POLEmut or MMRd.18
Although a subclonal mutation is one of the possible causes of discordance between TP53 sequencing and p53 immunostaining, we acknowledge additional different explanations, such as mutations in non-sequenced exons, post-translational mechanism independent of the TP53 gene sequence, misinterpretation by the operator.35
Strengths and Weaknesses
This is one of the largest datasets on multiple classifiers based on consecutive patients, making the estimation of incidence reliable. The use of next-generation sequencing and microsatellite instability testing, combined with staining for p53 and MMR protein, is also a strength of the study, as it allowed the detection of otherwise missed multiple classifiers. However, further studies are needed to define the usefulness of this approach, as there is no strong evidence of its prognostic impact. The limited number of cases and short follow-up are obvious limitations (the last patients were treated in December 2022). Hence, survival data should be interpreted with caution.
Implications for Practice and Future Research
Hierarchical molecular algorithms, meaning that testing is stopped as soon as an abnormality is detected, prevents the identification of multiple classifiers. In addition, the use of the original ProMisE algorithm, which starts with the assessment of the MMR protein, hinders the identification of POLEmut-MMRd. Alternatively, to identify all multiple classifiers, a simultaneous approach performing all tests should be used. The algorithm proposed by our group in a previous work can help to reduce the cost of the simultaneous approach without missing multiple classifiers, at least in early-stage endometrial cancer.15
Evidence on Lynch syndrome in multiple classifiers is currently limited. Hence, screening should be recommended in MMRd patients, regardless of other molecular features.
Collaborative studies engaging research teams from around the world still need to address the prognostic role of multiple classifiers, since current evidence is based on a small number of patients.
Although the combined approach of TP53 sequencing/p53 immunohistochemistry may be highly informative from a scientific point of view, further evidence is needed to support its worldwide applicability. Given the routine inclusion of TP53 gene in a sequencing panel for POLE analysis, it is crucial to understand how to manage TP53 mutant endometrial cancers expressing a normal p53 protein.
CONCLUSIONS
Multiple classifiers accounted for 11% of our cohort. Compared with previous studies, the higher proportion of multiple classifiers may be related to the extensive molecular analysis we performed, which included evaluation of both p53 expression and TP53 mutations. The characteristics of MMRd-p53abn appeared to be intermediate between those of MMRd and p53abn for histotype, grade, and MMR proteins expression, whereas the characteristics of POLEmut-p53abn were similar to POLEmut.
Data availability statement
All data relevant to the study are included in the article or uploaded as supplementary information.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants but the European Institute of Oncology Review Board - Reference ID: UID 2418 exempted this study. Participants gave informed consent to participate in the study before taking part.
Acknowledgments
The authors are grateful to Cynthia Avallone for her assistance with the literature search.
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 @LuigiDEvitis, @gcarusomd, @Giovanni Aletti, @BoganiGiorgio, @Fmultinu, @IBetella
Contributors LADV, GS, GC, IB, FM, and CF conceived, planned, and summarized the project. Cases (with clinical data and outcomes) were contributed by MTA, AA, AG, VZ, GA, AMg, and NC. Dataset assembly/organization by LADV and GS, with molecular methodologies performed by CF, EG-R, MB, DV, AMr, GB, and NC who provided intellectual contribution and edits. LADV and FM performed the statistical analyses. All authors contributed to manuscript writing and editing. IB and FM were responsible for the overall content as guarantors.
Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.
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.