Article Text

Timing of HPV vaccination as adjuvant treatment of CIN2+ recurrence in women undergoing surgical excision: a meta-analysis and meta-regression
  1. Marek Petráš1,
  2. Vladimír Dvořák2,
  3. Danuše Lomozová1,
  4. Roman Máčalík1,
  5. Sylva Neradová1,
  6. Pavel Dlouhý3,
  7. Jana Malinová4,
  8. Jozef Rosina5,6,
  9. Ivana Králová Lesná7,8
  1. 1 Department of Epidemiology and Biostatistics, Charles University, Prague, Czech Republic
  2. 2 Center of Ambulatory Gynecology and Primary Care, Brno, Czech Republic
  3. 3 Department of Hygiene, Charles University, Prague, Czech Republic
  4. 4 Královské Vinohrady University Hospital, Prague, Czech Republic
  5. 5 Department of Medical Biophysics and Medical Informatics, Charles University, Prague, Czech Republic
  6. 6 Department of Health Care and Population Protection, Ceské vysoké ucení technické v Praze Fakulta biomedicinského inzenyrstvi, Kladno, Czech Republic
  7. 7 Institute for Clinical and Experimental Medicine, Prague, Czech Republic
  8. 8 Department of Anesthesia, Charles University First Faculty of Medicine, Prague, Czech Republic
  1. Correspondence to Dr Marek Petráš, Department of Epidemiology and Biostatistics, Charles University, Prague 100 00, Czech Republic; marek.petras{at}


Objective The main aim was to determine the overall vaccine effectiveness (VE) against recurrent cervical intraepithelial neoplasia grade 2 or worse (CIN2+) including specific VE associated with timing of human papillomavirus (HPV) vaccination using data from published studies.

Design Meta-analysis and meta-regression.

Data sources A computerised literature search was undertaken using the MEDLINE, EMBASE, International Pharmaceutical Abstracts, Derwent Drug File, ProQuest Science and Technology, Cochrane and MedRxiv databases. To be eligible, the studies, with no language restrictions, had to be published between 1 January 2001 and 25 May 2023.

Review methods Included were studies with an unvaccinated reference group that assessed CIN2+ recurrence irrespective of the HPV genotype in women undergoing conisation provided. The present study was carried out in compliance with the Preferred Reporting Items for Systematic Review and Meta-Analyses and Meta-analysis Of Observational Studies in Epidemiology guidelines. The risk of study bias was assessed using the Newcastle–Ottawa Quality Assessment Scale. The Grading of Recommendations Assessment, Development, and Evaluation guidelines were used to assess the strength of evidence for the primary outcome. Data synthesis was conducted using meta-analysis and meta-regression.

Results Out of a total of 14 322 publications, 20 studies with a total of 21 estimates were included. The overall VE against recurrent CIN2+ irrespective of the HPV genotype achieved 69.5% (95% CI: 54.7% to 79.5%). While the HPV vaccine valency, follow-up duration, type of study including its risk of bias had no effect on VE, the highest VE of 78.1% (95% CI: 68.7% to 84.7%) was reported for women receiving their first dose not earlier than the day of excision. This outcome was supported by additional analyses and a VE prediction interval ranging from 67.1% to 85.4%.

Conclusions The outcome of this meta-analysis and meta-regression convincingly showed the beneficial effect of post-excisional HPV vaccination against CIN2+ recurrence. Studies published to date have been unable to determine whether or not vaccination, completed or initiated before conisation, would be associated with more favourable results.

PROSPERO registration number CRD42022353530.

  • genital neoplasms, female
  • vaccination
  • human papillomavirus

Data availability statement

All data relevant to the study are included in the article or uploaded as supplemental information.

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:

Statistics from

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.


  • Women undergoing excisional treatment of high-grade cervical intraepithelial neoplasia (CIN2+) are at increased risk of recurrence that can be reduced by vaccination against human papillomavirus (HPV) infection. The outcomes of recent studies have suggested different rates of reduction of CIN2+ recurrence, a finding that may have been influenced by the timing of HPV vaccination.


  • In this meta-regression study analysing data of 20 eligible studies, a 78% vaccine-adjusted or pooled vaccine effectiveness against recurrent high-grade cervical dysplasia was found in women vaccinated with an HPV vaccine, irrespective of its type, after surgical excision. Additional analyses confirmed a consistent outcome, as also documented by the prediction interval ranging from 67% to 85%. Initiating HPV vaccination together with or early after conisation can obviously improve the outcome of adjuvant therapy and reduce the risk of recurrence.


  • The outcomes of our quantitative synthesis conclusively showed the benefit of post-surgical HPV immunisation against recurrent CIN2+. This conclusion can serve as useful information for clinical practice.


All currently available human papillomavirus (HPV) vaccines were originally evaluated in clinical trials, licensed and recommended for the primary prevention of the risk of cervical intraepithelial neoplasia (CIN) in women. Despite mounting evidence of the real-world impact or the effectiveness of vaccination on the incidence of HPV-associated diseases in women and men obtained from more than 200 nationwide studies, attention has recently focused on the benefit of vaccination in specific high-risk populations, particularly in women undergoing excisional CIN treatment. Although cervical conisation, as a crucial method for diagnosing and treating CIN, is highly effective, there are still 5–25% of patients suffering from recurrent high-grade lesions after treatment.1–3 Moreover, they can be at increased risk of the undesirable effect of offspring prematurity after repeat conisation.4–10 Therefore, vaccination as adjuvant treatment might be helpful in reducing the risk of recurrent CIN after surgical excision.

We decided to conduct a new specific meta-analysis (PAVIVE, papillomavirus vaccine effectiveness) of data from studies assessing only recurrent high-risk CIN (CIN2+) in women undergoing surgical excision with or without HPV vaccination. The primary objective of the present study was to determine both the overall vaccine effectiveness (VE) and specific VE based on vaccine valency or timing of vaccination with respect to surgical intervention. We were interested to know whether or not vaccination completed before conisation or initiated before or after conisation could have a different impact on CIN recurrence. Therefore, we followed up on the recently published meta-analyses and reviews in an attempt to establish, specifically, an association (if any) between timing of vaccination and surgical excision with regard to the outcome.11–14

Material and methods

Our quantitative analyses were performed in compliance with the flow charts and checklists of the Preferred Reporting Items for Systematic Review and Meta-Analyses guidelines (online supplemental material) and Meta-analysis Of Observational Studies in Epidemiology guidelines (online supplemental material).15–17 PAVIVE was registered in the International prospective register of systematic reviews (CRD42022353530) and the first study outcome was focused on the specific studies assessing HPV vaccination as adjuvant treatment of surgical excision in women with CIN2+.

Supplemental material

Literature search

A computerised search of the relevant literature was undertaken using the MEDLINE, Excerpta Medica dataBASE, International Pharmaceutical Abstracts, Derwent Drug File and ProQuest Science and Technology databases on the PubMed, STNext and Cochrane platforms and MedRxiv, a free online archive of complete preprints. Eligible studies, with no language restriction, had to be published between 1 January 2001 and 25 May 2023. Moreover, a recursive manual search of references in published review or meta-analysis articles was undertaken to identify additional studies.

Study selection

Relevant publications had to contain the keywords “immunization”, “HPV” and “effectiveness”” or their synonyms. The full search formulas are shown in online supplemental material. Eligible studies had to meet the following inclusion criteria: (1) observational studies or clinical trials with unvaccinated women as a reference group; (2) not immunocompromised women undergoing conisation for cervical dysplasia; (3) exposure defined by HPV vaccination; (4) outcome assessed by CIN2+ recurrence; (5) stated effect size (ES) or data for OR calculation including the CI. The excluded studies investigated dysplasia outside of the cervix (vulval, vaginal, anal, etc), HIV-positive women only or no recurrent diseases.

Data extraction

The VE estimates including the 95% CI were extracted directly from the selected studies. If missing, the VE estimates were obtained from the reported ESs irrespective of how they were expressed, such as OR, HR, risk ratio or incidence rate ratio. The total of vaccinated and unvaccinated women including the number of treatment failure cases was extracted from the studies to document the overall sample size of enrolled women in each particular study. Furthermore, these data were also used to estimate VE from the calculated OR where appropriate (ie, missing ES or VE).

Additional collected data included the timing of immunisation, HPV vaccine used, follow-up duration, study type and HPV genotypes of CIN2+. The timing of vaccination was stratified as previous vaccination (complete vaccination prior to conisation), before and/or after surgical excision (studies investigating the effect of mixed-start vaccination, that is, initiated both prior to and after conisation) and exclusively after conisation (vaccination not performed earlier than the day of conisation). Furthermore, the full vaccination was always defined by three received doses in all studies included in this meta-analysis. The VE estimate was assigned to a bivalent, quadrivalent or nonavalent vaccine if more than 68% of the women had been vaccinated with the respective HPV vaccine; otherwise, the VE estimate was allocated to the category of any HPV vaccine with unspecified valency. This approach is consistent with the conventional characterisation of a study population expressed by the range of a mean and ±SD applicable to 68% of study participants.18 The follow-up duration was categorised by the longest study period specified by the authors. If the period was not reported, the information was compensated for by the third quartile or by the sum of mean and SD of follow-up. The study types considered in our meta-analysis included a cohort study, case–control study and randomised clinical trial including its post-hoc analysis. The rates of CIN2+ recurrence were assigned to either vaccine-related or any HPV genotypes.

Quality assessment

To evaluate a study, data about its characteristics, participants, interventions, comparison and outcome were extracted. The quality of each study was evaluated using the Newcastle–Ottawa Quality Assessment Scale (NOS) in three domains: selection of participants, comparability of vaccinated versus unvaccinated women and outcome of interest expressed by the risk of bias (RoB).19 For the purpose of this meta-analysis, studies of ≥8 stars of score were regarded as ones at low RoB. Moreover, this scoring was employed as another variable of meta-regression outcome.

The Grading of Recommendations Assessment, Development, and Evaluation (GRADE) criteria were used to rate the quality of the outcome resulting from this quantitative synthesis.20 Serious limitations, indirectness, inconsistency and imprecision were assessed using the scores of the NOS domains of selection, outcome and comparability, the inconsistency index (I2) and the SE achieved in studies included in the overall or subgroup results of our meta-analysis. Moreover, the strength of evidence was supported by undetected publication bias, the prediction interval with a positive lower limit and at least 10 analysed estimates.21 The effect of small studies as well as an estimate of missing studies were only of informative value with no impact on the quality assessment of the outcome. More details are available in online supplemental material.

Data analysis

The meta-analysis was conducted to establish the pooled overall VE from precomputed ESs and their SEs obtained from both the upper and lower limits of the 95% CI in the metric closest to normality with an acceptable difference of lower limit <±0.05 against the value declared by study. The ES included any ratio assessing the relationship of HPV vaccination and CIN2+ recurrences, and mutual deviations among different ratios were considered neglected. VE was calculated as follows: (1–ES)×100%. Given the assumed heterogeneity of studies, quantitative synthesis was primarily conducted using the random-effects model (DerSimonian-Laird method) and heterogeneity was tested by I2.22

The primary analysis was complemented with meta-regression to evaluate the dependence of log-transformed ESs on the investigated categorical variables. The limited number of estimates extracted from the studies made it possible to conduct bivariate regression always including the timing of vaccination and another variable, that is, the HPV vaccine used, follow-up duration, type of study or RoB. The reference subgroup of each variable was the one that included the highest number of estimates and the results were expressed by regression coefficients and adjusted VE including the 95% CI.

Additional analyses were performed to support the strength of evidence: (1) meta-analyses using a fixed-effects model (inverse variance method, I–V) helped to identify a potential publication bias, (2) the effect of small studies determined by a meta-regression model with Egger’s test, (3) the summary effect of asymmetry with estimation of potentially missing studies using the trim-and-fill method, and (4) the prediction interval estimated based on the SEs of heterogeneity of studies.22 23 More details and outcomes of these analyses are reported in the online supplemental material.

Statistical analyses were performed using STATA V.17.0 (StataCorp 2021, Stata Statistical Software, College Station, Texas, USA) with a significance level of α=0.05 with a two-tailed 95% CI.


Our literature search identified a total of 14 322 studies, of which number 205 met the inclusion criteria for HPV VE. A total of 20 eligible studies were eventually included in the quantitative synthesis that met the inclusion criteria of the present study (figure 1).

While all eligible studies were designed to investigate CIN2+ recurrence caused by any HPV genotype (20 studies), only seven of them reported cases of recurrence associated with vaccine-related genotypes of 16 and 18. Therefore, the final quantitative synthesis involved studies investigating VE against recurrence irrespective of the HPV genotype (table 1 and online supplemental material). As only one study assessed adjuvant treatment with a nonavalent HPV vaccine, the result was added to that of a quadrivalent one of the same manufacturer. The study outcome was assigned to either a bivalent (HPV2) and a quadri/nonavalent (HPV4+) or unspecified vaccine. The follow-up duration was stratified to ≤2 years, 3–4 years and ≥5 years of follow-up. Overall, a total of 21 estimates were extracted from 20 studies because one study separately assessed preconisation CIN of grades 2 and 3. A total of 15 estimates were related to patients with only CIN2+ at conisation and three estimates were assigned to those mostly with CIN2+ (>90%).

Table 1

Characteristics of included studies

CIN recurrence in HPV-vaccinated women after surgical excision

The pooled VE against recurrent CIN2+ irrespective of the HPV genotype achieved 69.5% (95% CI: 54.7% to 79.5%) independently of the investigated variables (figure 2). The outcome was similar whether or not the women had been immunised with an HPV2 or HPV4+ vaccine, as documented by the regression coefficient of –0.10 (95% CI: –1.09 to 0.89) or by the timing-adjusted VE of 80.8% (95% CI: 43.6% to 93.4%) for HPV2 and 78.8% (95% CI: 67.7% to 86.1%) for HPV4+ (figure 3 and online supplemental material).

Figure 2

Forest plot of HPV vaccine effectiveness against CIN2+ recurrence (results from a random-effects model including I2, index of inconsistency and p value). Particulars of the forest plot: the diamonds indicate pooled effectiveness, with lateral points indicating the 95% CI; dashed lines indicate the point of pooled effectiveness. CIN2+, high-grade cervical intraepithelial neoplasia; F/S, failed/successful cases; HPV, human papillomavirus.

Figure 3

Adjusted VE dependence on predefined variables expressed by the meta-regression coefficients (effectiveness of timing immunisation was adjusted for HPV vaccine while that of any other variables was adjusted for timing of immunisation) HPV2/HPV4+, bivalent/quadrivalent or nonavalent human papillomavirus vaccine; VE, vaccine effectiveness.

Initiation of immunisation with regard to surgical excision was associated with different rates of reduction in CIN2+ recurrence. The highest effectiveness of 78.1% (95% CI: 68.7% to 84.7%) was achieved in women receiving their first vaccine dose not earlier than the day of conisation. The VE decreased to 75.9% (95% CI: 63.5% to 84.1%) if only women with CIN2+ before conisation were included as demonstrated by our sensitivity analysis (online supplemental material). The risk of recurrent CIN2+ was reduced by 49.8% in previously immunised women and by 47.8% in studies with mixed-start vaccination relative to conisation (table 2). Similar results were confirmed by the regression coefficients (figure 3) and VE adjusted by the vaccine used, follow-up duration, study type and RoB (online supplemental material).

Table 2

Overall and stratified pooled vaccine effectiveness (VE) by the investigated variables including the prediction interval, publication bias and quality assessment studies

No effect of follow-up duration was documented because the pooled VE ranged between 68.4% and 72.3% from ≤2 years up to ≥5 years with no statistical significance as reported by the 95% CI of regression coefficients (figure 3). Similarly, study type and RoB scoring did not show any impact on the VE depending on vaccination time. An additional analysis of sensitivity performed in the specific set of estimates extracted only from studies at low RoB showed 76.3% effectiveness (95% CI: 64.9% to 84.0%) in 99% of women with preconisation CIN2+ who had received mostly an HPV4+ vaccine and were followed up for at least 3 years. This result was consistent as documented by the prediction interval ranging from 62.0% to 85.2% (table 2).

Quality of evidence

The strength of evidence supported by a sufficient number of at least 10 VE estimates was determined for those immunised with an HPV4+ vaccine, vaccinated post-excision with follow-up duration of 3–4 years as well as for overall outcome. Only cohort studies or studies at low RoB were represented by more than 10 estimates (table 2).

The overall pooled outcome was burden by the very low quality of studies evaluated using the GRADE criteria and detected publication bias (a significant difference between the results of both models, that is, fixed-effects and random-effects models). Moreover, the uncertainty of the result was supported by the prediction interval (–22.8% to 92.4%).

Subgrouped outcomes of women vaccinated with an HPV4+ vaccine (75.9%) or those in cohort studies (65.4%) came from studies of low or very low quality. Therefore, these results were considered only supportive with no strength of evidence.

The pooled VE of post-excision vaccination or of the follow-up duration of 3–4 years was obtained from moderate-quality studies. The results did not exhibit publication bias and, moreover, no heterogeneity of studies was revealed. However, the effect of small studies assessing post-excision vaccination and missing studies was detected and the imputed studies to decrease pooled effectiveness to 75.0% (95% CI: 64.7% to 82.2%) for post-excision vaccination and 66.2% (50.5% to 76.9%) for the 3–4 years of duration (online supplemental material). Nevertheless, the prediction intervals demonstrated the stability of the pooled VE because the lower limits of this interval were >50% (table 2).

A total of 15 studies at low RoB reported pooled VE with no publication bias and with stability confirmed by the prediction interval of 59.4% to 78.3%. Even if the effect of small studies and potentially missing studies was found, the pooled VE of 70.3% decreased to 69.2% after accounting for two imputed hypothetical estimates. Both these effects could be deemed not serious. Similarly, the specific pooled VE related to post-excision of women with previous CIN2+ and immunised mostly with an HPV4+ vaccine followed up for at least 3 years conducted in estimates extracted from nine studies at low RoB was not influenced by publication bias or the effect of small studies. The imputed studies decreased the original VE of 76.3% to 74.4%, an outcome in line with the prediction interval of 62.0% to 85.2%.


The present meta-analysis and meta-regression showed that conisation supported by three-dose vaccination against HPV can reduce the recurrence of CIN2+ in women by more than 69% compared with excisional treatment only. This result is consistent with previous findings of similar meta-analyses conducted in 2020–2022 and involving 9–11 studies with pooled VE outcomes ranging between 46 and 65% regardless of timing of vaccination.11 12 24–26 It is likely that the nine additional studies in our PAVIVE Study did not affect the overall result showing robustness and benefit of HPV vaccination against CIN2+ recurrence irrespective of the HPV genotype. Moreover, another two meta-analyses evaluating recurrence rates for only vaccine genotypes of HPV16 and 18, confirmed a 63–73% reduction in CIN2+ recurrence rates.12 25 Given the small number of seven studies focused on HPV genotypes of 16 and 18, we decided not to investigate them in this meta-analysis because no new finding could be found versus current results of published meta-analyses.

We intentionally did not include a study conducted in HIV-positive women that showed no effect of vaccination, and this finding was consistent with results of recurrent anal intraepithelial neoplasia in vaccinated HIV-positive men.27 28 Overall, this finding does not support HPV vaccination as adjuvant treatment in HIV-positive persons.

The effect of HPV vaccination was not influenced by the commercial HPV vaccines used or their valency as well as by follow-up duration, as demonstrated by coefficients of meta-regression. Furthermore, no different impact of observational studies and clinical trials including their post-hoc analyses on VE was observed. Even if the studies at low RoB provided results of higher quality, the pooled VE estimated from studies at moderate-to-high RoB was not significantly different.

We believe to be the first to show that the timing of full vaccination could be an important factor having a major impact on VE. The benefit rose significantly to 78% in women vaccinated post-excision versus those vaccinated before and/or after surgical excision (48%).

Current data support three-dose vaccination initiated on the day of conisation or later since the effectiveness of such vaccination could be at least 51% (lower limit of prediction interval) independently of the HPV vaccine used and ≥3 years of follow-up, as suggested by the strength of evidence for results found in studies at low RoB. Moreover, the finding remained unchanged after adjustment for the HPV vaccine used, RoB of study, follow-up duration or type of study, as documented by regression coefficients.

It is not quite clear why the timing of HPV vaccine administration could have a variable impact on VE. HPV infection seems to be connected with the epithelial microenvironment and stimulate pro-inflammatory cytokines.29 However, as surgical excision of CIN2+ can reduce these cytokines to levels similar to those found in non-infected women, one could expect post-conisation vaccination will also be beneficial.30 Another explanation could be that it is conisation itself that can trigger some degree of inflammation as an injury. As suggested by a recent study, if vaccination is performed in the presence of a low level of local inflammation, the immune response could be stronger.31 The reason why pre-excision vaccination has a smaller effect in women with cervical dysplasia may be the precancerous environment that blunts the antiviral response and activates multiple immunosuppressive factors causing immune escape.32

These mentioned mechanisms could possibly explain the different effects depending on the time of vaccination. However, we can only speculate which and/or if any of those or other explanations fit reality. The results of our study including those furnished by supportive additional analyses confirmed a benefit of early post-conisation HPV immunisation in clinically immunocompetent women to reduce the risk of recurrent CIN2+. Despite this, given the small number of studies assessing preconisation immunisation and mixed start of vaccination times with respect to conisation in studies with one control group, we cannot conclusively claim that immunisation initiated or completed before conisation is associated with an inferior outcome.


Limitations of this study could include use of estimates based on ES involving different ratios of association, that is, OR, risk ratio, HR or incidence rate ratio. Nevertheless, this approach is generally acceptable, and it uses both adjusted and crude estimates related to each particular study. Another approach can apply extracted data of the number of exposed and unexposed participants with failed and successful intervention to estimate a particular ratio. Even if this method evaluates every estimate in the same way, the risk of distorting the original result is likely to be high. Therefore, we believe that the approach we adopted was the most suitable one.

Another limitation was the number of estimates stratified by categorical variables allowing to perform only bivariate regression where the timing of vaccination was always applied as an independent variable. The outcomes of these regression models can be considered satisfactory since the model appropriateness was confirmed by the coefficient of determination significantly higher than zero.

A potential shortcoming of this study could be the inclusion of studies assessing patients with CIN of any grade at conisation. However, we do not think this could have substantially impacted the outcomes because a specific analysis of sensitivity in 99% of women with CIN2+ at surgical excision and vaccinated after conisation showed an effect similar to that seen in women independently of the preconisation grade of CIN. Moreover, it was also supported by the sensitivity analysis conducted exclusively in women with high-grade cervical dysplasia.

Unfortunately, most of the selected studies failed to specify the immunocompetent status of female participants. As a result, the absence of immunosuppression was only speculative, without verification. Nevertheless, we believe that this assumption could be justified since one study in HIV-positive women showed no effect of HPV vaccination on CIN recurrence reduction.


This meta-analysis conclusively showed the robust and great benefit of post-surgical HPV immunisation against recurrent CIN2+ independently of HPV vaccine, follow-up duration, type of study and level of their RoB. The 78% reduction in recurrence rates was achieved in women with vaccination initiated close to the day of conisation.

Abstract translation

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.

Data availability statement

All data relevant to the study are included in the article or uploaded as supplemental information.

Ethics statements

Patient consent for publication

Ethics approval

Not applicable.


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.


  • Handling editor Nadja A Vielot

  • Contributors The study was conceived by MP, and designed by MP, IKL and VD. The data were acquired and collated by MP, RM, SN, DL and JM, and analysed by MP. Funding was arranged by PD. Administrative and material support was negotiated by JR. The manuscript was drafted and revised critically for important intellectual content by all authors (MP, IKL, VD, RM, SN, JM, PD, JR, DL). All authors gave final approval of the version to be published. MP accepts full responsibility for the work and/or the conduct of the study, had access to the data, and controlled the decision to publish.

  • Funding This work was supported by the Cooperatio 31 fund, Health Sciences, Charles University, Prague, Czech Republic.

  • 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.