Article Text

PDF

Continuing reductions in HPV 16/18 in a population with high coverage of bivalent HPV vaccination in England: an ongoing cross-sectional study
  1. David Mesher1,
  2. Kavita Panwar2,
  3. Sara L Thomas3,
  4. Simon Beddows2,
  5. Kate Soldan1
  1. 1HIV & STI Department, Centre for Infectious Disease Surveillance and Control, Public Health England, London, UK
  2. 2Virus Reference Department, Public Health England, London, UK
  3. 3Department of Epidemiology & Population Health, London School of Hygiene and Tropical Medicine, London, UK
  1. Correspondence to David Mesher; david.mesher{at}phe.gov.uk

Abstract

Objectives The human papillomavirus (HPV) immunisation programme in England was introduced in 2008. Monitoring changes in type-specific HPV prevalence allows assessment of the population impact of this vaccination programme.

Methods Residual vulva-vaginal swab specimens were collected from young sexually active women (aged 16–24 years) attending for chlamydia screening across England. Specimens were collected between 2010 and 2013 for type-specific HPV-DNA testing. HPV prevalence was compared to a similar survey conducted in 2008 prior to the introduction of HPV vaccination.

Results A total of 7321 specimens collected in the postvaccination period, and 2354 specimens from the prevaccination period were included in this analysis. Among the individuals aged 16–18 years, with an estimated vaccination coverage of 67%, the prevalence of HPV16/18 infection decreased from 17.6% in 2008 to 6.1% in the postvaccination period. Within the postvaccination period, there was a trend towards lower HPV16/18 prevalence with higher vaccination coverage and increasing time since vaccine introduction from 8.5% in the period 2–3 years postvaccination to 4.0% in the period 4–5 years postvaccination. The prevalence of HPV31 reduced from 3.7% in the prevaccination period to 0.9% after vaccine introduction, although this no longer reached statistical significance after additional consideration of the uncertainty due to the assay change. Smaller reductions were seen in the individuals aged 19–21 years with lower estimated vaccination coverage, but there was no evidence of a reduction in the older unvaccinated women. Some overall increase in non-vaccine types was seen in the youngest age groups (ORs (95% CI); 1.3 (1.0 to 1.7) and 1.5 (1.1 to 2.0) for individuals aged 16–18 and 19–21 years, respectively, when adjusted for known population changes and the change in assay) although this should be interpreted with caution given the potential unmasking effect.

Conclusions These data demonstrate a reduction in the HPV vaccine types in the age group with the highest HPV vaccination coverage.

  • EPIDEMIOLOGY

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 and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/

Statistics from Altmetric.com

Strengths and limitations of this study

  • We conducted human papillomavirus (HPV) surveillance among a large number of young women attending for chlamydia screening, with HPV type-specific testing performed for almost 10,000 women.

  • The large sample size of this study has allowed us to consider the population impact of the bivalent HPV vaccine against the two vaccine types, and against cross-protective HPV types.

  • We demonstrate continued decreases in the prevalence of vaccine-targeted HPV types over time up to 4 years after the introduction of the bivalent vaccination programme.

  • Analyses compare data from repeat cross-sectional surveys. Therefore, unrecorded changes in the population characteristics may have resulted in a change in HPV prevalence which is unrelated to HPV vaccination.

Introduction

Persistent infection with a high-risk (HR) human papillomavirus (HPV) type is a necessary cause of cervical cancer, and has been shown to be associated with other cancers in men and women.1 ,2 Two of these HR-HPV types, HPV16 and HPV18, are present in around 70–80% of cervical cancers.3 ,4 Infection with low-risk (LR) HPV6 or HPV11 has been shown to be associated with the vast majority of genital warts.5

HPV vaccination of young females has been introduced widely in developed countries as well as in some developing countries6 since 2007, using the first two licensed vaccines (a bivalent HPV16/18 vaccine and quadrivalent HPV6/11/16/18 vaccine). In late 2008, the UK began providing HPV vaccination, free at the point of delivery, routinely to 12-year-old females, and catch-up vaccination to females up to and including 17-year-olds. The bivalent vaccine was offered until September 2012 when the programme changed to offer the quadrivalent vaccine. Throughout the UK, over 80% of females eligible for routine vaccination each year have completed the three-dose course.7–9 Three-dose coverage within the catch-up ages has been lower, with average coverage of 73% for individuals aged 14–15 years, and 45% for 16–17 years,7 although this is still higher than in most other countries.10–12

In 2013, we reported findings from our surveillance of type-specific HPV infections in sexually active young females in England, showing evidence of substantially lower HPV16/18 prevalence in the first 4000 postvaccination period specimens tested compared with prevaccination prevalence.13 Reductions in the prevalence of HPV16/18 following the introduction of HPV vaccination have also been shown in Australia,14 the USA,15–17 Scotland18 and Sweden.19

Some cross-protection against non-vaccine HR-HPV types closely related to HPV16/18 has been demonstrated in clinical trials of both vaccines (specifically, HPV31, HPV33 and HPV45 for the bivalent vaccine, and HPV31 for the quadrivalent vaccine),20–22 and has been observed for the bivalent vaccine by ongoing surveillance of young women undergoing cervical screening in Scotland.18 Ongoing surveillance for changes in the prevalence of other non-vaccine HPV types is also prudent. These changes could result from vaccination due to cross-protection against non-vaccine HR-HPV types (ie, causing decreases in prevalence) or due to type replacement (ie, causing increases in prevalence).

We report further findings from our ongoing HPV surveillance (now over 7000 postvaccination specimens) in our high-coverage population, including changes in vaccine and non-vaccine types. We aimed to determine to what extent any such observed changes were likely to have resulted from vaccination, rather than be due to methodological reasons (eg, assay performance, unmasking), or a result of other factors such as changes in sexual behaviour over time.

Methods

The methods of specimen selection, collection and testing, and the characteristics of the study population have been described previously.13 ,23 Briefly, residual vulva-vaginal swab specimens were collected via 10 laboratories from young women aged 16–24 years undergoing chlamydia screening at general practice, community and sexual health services (CaSH, otherwise known as family planning), and youth clinics. Residual specimens were all sent for HPV testing at the Virus Reference Department laboratory at Public Health England (PHE). In England, chlamydia screening is recommended for all sexually active men and women under 25 years old annually, and on partner change, irrespective of symptoms or perceived risk. Demographic data were reported separately and linked to the specimens received at the PHE Centre for Infectious Disease Surveillance and Control. Prior to testing for HPV DNA, specimens were unlinked from any patient-identifiable data and anonymised. This study was reviewed and approved by the South East Research Ethics Committee (REC reference: 10/H1102/7). Individual patient consent was not required as this study tested anonymised specimens (with no patient-identifiable data) as part of Public Health Surveillance conducted to monitor the HPV vaccination programme.

Prevaccination-period specimens were collected between January and September 2008, prior to the introduction of the national HPV vaccination programme in England. Postvaccination-period specimens were collected between October 2010 and April 2013, and divided into two periods, 2–3 (ie, 2010–2011) and 4–5 years (ie, 2012–2013) postvaccination.

Postvaccination specimens were tested for type-specific HPV DNA to detect 13 HR types (HPV16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59 and 68), five possible HR types (HPV26, 53, 70, 73 and 82) and two LR types (HPV6 and 11) using an in-house multiplex PCR and Luminex-based genotyping test with pyruvate dehydrogenase (PDH) detection for sample integrity.24 Prevaccination specimens were tested by Hybrid Capture 2 (HC2) HPV DNA test using the Combined Probe Cocktail Method to detect HR and possible HR types (as above) and five LR types (6, 11, 42, 43 and 44) and genotyped by the Linear Array HPV Genotyping (LA) test (Roche Molecular Systems) if HC2 positive.

HPV prevalence was calculated for each individual HPV type. We also calculated combined HPV prevalence, restricted to types included in the prevaccination and postvaccination assays, for (1) any HR-HPV type; (2) the HR-HPV types included in the current vaccines: HPV16 and/or 18; (3) the additional HR-HPV types included in the nonavalent HPV vaccine25: HPV31, 33, 45, 52 and/or 58; (4) the HPV types for which there is some evidence of cross-protection from clinical trials: HPV31, 33 and/or 45 and (5) the non-vaccine HR-HPV types (ie, HR types not including HPV16 or 18). Changes in prevalence between the prevaccination and postvaccination (combined) periods were compared using ORs calculated using a logistic regression model. Trends over time were assessed by including three time periods (prevaccination, 2–3 years postvaccination, 4–5 years postvaccination) as an ordered continuous variable. Adjusted ORs were calculated adjusting for age, testing venue type and chlamydia positivity (as a marker for sexual behaviour). To account for the change in assay between the prevaccination and postvaccination periods, we used type-specific sensitivity and specificity estimates from a validation study (428 prevaccination specimens, retested through the postvaccination testing system13). The logitem command in Stata was used to adjust for the different sensitivity and specificity of the assay used in the prevaccination period. This command performs logistic regression when the binary outcome is measured with uncertainty. This adjustment did not account for the uncertainty surrounding the sensitivity and specificity estimates, hence, bootstrapping techniques were used to incorporate this additional uncertainty to the SEs of the ORs. A similar statistical model to adjust for assay change was not available for use with prevalence ratios (PR). However, PRs adjusted for age, testing venue type, and chlamydia positivity were also calculated using a log binomial model and the results compared with the equivalent ORs.

In order to further address concerns about changes in prevalence of HPV in the postvaccination period unrelated to vaccine introduction (eg, changes in sexual behaviour not addressed by adjustment of chlamydia positivity or residual changes in assay sensitivity), we also compared prevaccination and postvaccination type-specific prevalence when restricted to specimens with at least one HR-HPV type detected. This enabled assessment of changes in the relative, rather than absolute, frequency of specific HR types. To give a simple example, if prevaccination specimens comprised 20% HR type positivity overall (10% prevalence of type 31 and 10% prevalence of type 33), and postvaccination specimens comprised 30% HR type positivity (15% prevalence of type 31 and 15% prevalence of type 33), the absolute frequency of each type would increase postvaccination. However, the relative prevalence of each type within the HR-HPVs positives would remain the same.

Vaccination coverage by age and time period was derived from published data.7–9 ,26

Results

Demographics and characteristics

Results were analysed from 2354 prevaccination specimens and 7321 postvaccination specimens: 3602 (49.2%) from 2–3 years postvaccination, and 3719 (50.8%) from 4–5 years postvaccination. The characteristics of study participants were similar in the prevaccination and postvaccination periods (table 1), except there were more specimens from women of non-white ethnic groups, and fewer specimens collected from youth clinics in the postvaccination collection (7.3% vs 17.6%, and 24.1% vs 3.1%, respectively). Furthermore, two laboratories, Leeds and Lewisham, included in only the postvaccination period, had notably higher chlamydia positivity rates than the other laboratories (22.4% and 8.4%, respectively). Data from these laboratories were excluded in sensitivity analyses. In the postvaccination period, the estimated vaccination coverage in the surveillance population, based on nationally reported data, was 67.2%, 30.7% and 0.6% for individuals aged 16–18, 19–21 and 22–24 years, respectively (table 2 for estimates by time period).

Table 1

Characteristics of women included in the prevaccination and postvaccination surveys

Table 2

Estimates of prevalence of HPV types by age in prevaccination and postvaccination periods

HR HPV16 and/or 18 infection

In the youngest age group (16–18 years), the prevalence of HPV16/18 was 17.6% in the prevaccination period compared to 8.5% in the period 2–3 years postvaccination, and 4.0% 4–5 years postvaccination (p value for trend <0.001; table 2 and figure 1). This corresponds to an overall reduction of 66% comparing the prevaccination prevalence to the combined postvaccination prevalence. A trend was also seen in individuals aged 19–21 years with a prevaccination prevalence of 16.9% compared to 14.2% in the period 2–3 years postvaccination, and 8.7% 4–5 years postvaccination (p value for trend <0.001; combined reduction between the prevaccination and postvaccination periods of 31%). However, there was no decrease in the prevalence of HPV16 and/or HPV18 in the oldest age group, who were largely unvaccinated. There was a slight decrease in HPV18 infection in the oldest age group, but this difference was no longer seen once adjustment was made for changes in population and HPV assay (data not shown).

Figure 1

Prevaccination and postvaccination prevalence of human papillomavirus (HPV) types by age. Percentages in square brackets represent estimated three-dose HPV vaccination coverage for individuals aged 16–18, 19–21 and 22–24 years, respectively. HR, high risk.

The adjusted ORs for the postvaccination periods (combined) compared with the prevaccination period were 0.3 (95% CI 0.2 to 0.4), 0.6 (95% CI 0.5 to 0.9) and 1.1 (95% CI 0.8 to 1.7) for individuals aged 16–18, 19–21 and 22–24 years, respectively (table 3).

Table 3

Prevalence and OR of HPV infection in the postvaccination period compared to prevaccination, by age group

HR HPV31, 33 and/or 45 infection (cross-protective HPV types)

The prevalence of HPV31, 33 and/or 45 among the individuals aged 16–18 years was 8.4% in the prevaccination period, 6.9% in the period 2–3 years postvaccination, and 5.8% 4–5 years postvaccination. After adjusting for demographics and the change in HPV assay, the adjusted OR postvaccination (combined) was 0.9 (95% CI 0.5 to 1.5), p value=0.58 (tables 2 and 3). The prevalence of HPV31 in this age group reduced from 3.7% in the prevaccination period to 0.9% in the combined postvaccination period. This reduction did not reach statistical significance after adjustment for the known population changes and the assay change (adjusted OR 0.4 (95% CI 0.2 to 2.9), p value=0.21): there was no evidence of a reduction in the overall prevalence of HPV33 or of HPV45. Among women aged 16–18 years with at least one non-vaccine HR-HPV type detected, the prevalence of HPV31/33/45 was 48% lower in the period 4–5 years postvaccination compared to the prevaccination period, with a reduction from 14.9% to 3.7% for HPV31, 9.6% to 7.9% for HPV33, and 11.5% to 6.5% for HPV45 (see online supplementary table S1). In the older age groups, with lower vaccination coverage, there was no evidence of a reduction in these three HPV types between the prevaccination and postvaccination periods.

Non-vaccine HR-HPV types

There was an increase in the prevalence of non-vaccine HR-HPV types between the prevaccination and postvaccination periods at all ages (24.9–33.7%, 26.9–39.6% and 26.4–32.9% for individuals aged 16–18, 19–21 and 22–24 years, respectively). After adjustment for age, venue type, chlamydia positivity and the change in assay, the adjusted ORs comparing the prevaccination and postvaccination prevalence were 1.3 (1.0 to 1.7), 1.5 (1.1 to 2.0) and 1.2 (0.9 to 1.6) for individuals aged 16–18, 19–21 and 22–24 years, respectively (table 3). There was also evidence for increases in the prevalence of the additional nonavalent HR-HPV types in the age group of 19–21 years. Adjusted ORs were 1.2 (0.9 to 1.7), 1.5 (1.1 to 2.2) and 1.3 (0.9 to 2.0), respectively. This increase in non-vaccine HR-HPV types was only seen between the prevaccination and postvaccination combined periods. Within the postvaccination period, there was no evidence of a change in the prevalence of these HR-HPV types over time (table 2).

The type-specific prevalence of HPV58 was similar in the prevaccination and postvaccination period for all age groups. However, there was an increase in the prevalence of HPV52 even after adjustment (adjusted OR 1.7 (1.0 to 3.2) and 2.4 (1.4 to 4.7), respectively) for individuals aged 16–18 and 19–21 years, and a borderline increase for individuals aged 22–24 years (1.6 (0.9 to 3.6)).

LR HPV6 and/or 11 infection

Similar to the non-vaccine HR types, there was a significant increase in the prevalence of HPV6/11 in the postvaccination period among women aged 16–18 years (5.8% prevaccination vs 8.3% postvaccination; adjusted OR 1.9 (1.1 to 3.4)). There was also a slight increase in the LR types for individuals aged 19–21 years (5.8% vs 7.6%, respectively; adjusted OR 1.4 (0.8 to 2.6)) although after adjustment for age, venue type, chlamydia positivity and the change in assay, this was not significant (p=0.15). There was no evidence of a change in the prevalence of HPV6/11 in the older age group (22–24-years, 4.4% prevaccination vs 4.3% postvaccination; adjusted OR 1.2 (0.6 to 4.1)).

Repeating analyses using PRs instead of ORs (adjusted for all factors except for assay change) gave very similar results for all HPV types (results not shown).

Discussion

This surveillance of young sexually active women undergoing chlamydia screening has demonstrated continuing reductions in the prevalence of the HPV vaccine types following the introduction of a high-coverage national HPV vaccination programme as well as some evidence of overall reductions in HPV31 (the closely related HPV type with strongest evidence of cross-protection from the bivalent vaccine clinical trials22). Encouragingly, these reductions are more marked in the later postvaccination period with higher estimated vaccination coverage. Use of bootstrapping techniques to account for the uncertainty of the specificity and sensitivity estimates from the validation study provided conservative estimates with wider CIs. Once we accounted for this additional uncertainty, the reduction in the prevalence of HPV31 postvaccination no longer reached statistical significance.

The percentage reductions between the postvaccination and prevaccination periods among the youngest two age groups were very similar to the estimated vaccine coverage. If all the reduction in prevalence was due to a direct effect of vaccination, this would be consistent with close to 100% vaccine effectiveness. Such high vaccine effectiveness is unlikely given that women included in this surveillance were largely vaccinated as part of the catch-up programme, and almost certainly some of those vaccinated would have had an existing HPV infection. These high reductions could be partly explained by the fact that nationally published data that was used to estimate vaccination coverage is based on reported data on vaccination administration. A recent study of serological markers has suggested that these administration data may be under-reporting HPV vaccinations among women eligible for vaccination as part of the catch-up programme (D Mesher, E Stanford, J White, et al. HPV serology testing confirms high HPV immunisation coverage in England. Submitted for publication 2015). This would mean that the vaccination coverage we had estimated for our surveillance population would have been a slight under-estimate. This would be more consistent with the relatively high overall reductions in HPV16/18 we observed although it is most likely that these are due to a combination of both higher vaccination coverage and some herd protection effect.

This surveillance makes use of a large sample of residual specimens taken for chlamydia screening and tested anonymously for HPV-DNA infection. Young women attending for chlamydia screening have higher risks of chlamydia, and therefore, probably for HPV infection, than the general population. The reductions in the HPV vaccine types (HPV16/18) observed here, therefore, reassures that benefits of HPV vaccination have not been inequitably biased to lower risk individuals.

The observation that the reductions in HPV16/18 were only seen in the age groups eligible for national HPV vaccination, and reduced further in the later postvaccination period (ie, were proportionate to estimated vaccination coverage), strongly suggests that the changes seen are attributable to vaccination.

If increases in the other HR-HPV types were restricted to the younger age groups, or were greater in the later postvaccination period with higher vaccination coverage, then this could raise suspicion of potential type replacement. However, the increases seen in the non-vaccine HPV types were seen in all age groups, including the older unvaccinated women, which suggest that these increases are unlikely to be due to type replacement, and are more likely a result of limitations in our study. First, comparison of HPV prevalence between the prevaccination and postvaccination periods were adjusted for age, venue type and chlamydia positivity (as a marker of sexual behaviour). However, other changes in the population characteristics (or sexual behaviour not captured by chlamydia positivity) may have resulted in a change in prevalence of the non-vaccine HR-HPV types. If women in the postvaccination period were at a higher risk of HPV infection then this could have underestimated the potential effect of HPV vaccination on the HPV vaccine types. Analyses restricted to women with at least one HR-HPV type show larger declines for HPV31 and evidence of a reduction in HPV45 which would support this hypothesis. Furthermore, these analyses restricted to HR-HPV-positive specimens show little difference in relative prevalence of HPV52 or 58, which strengthens our conclusion that these increases are probably not due to type replacement. Second, there was a change in the assay used between the postvaccination and prevaccination periods, but no change in the assay used throughout the postvaccination period; hence, continued reductions in the vaccine HPV types within the postvaccination period cannot be affected by this. However, it was necessary to adjust ORs comparing the prevaccination and postvaccination periods for the different assays used. Finally, broad-spectrum assays, such as those used in our study, can lack sensitivity to detect individual HPV types at low copy number in the presence of other HPV types. Therefore, the decrease in multiple HPV infections due to the reduction in HPV16 and 18 following vaccination could lead to an apparent, artificial increase in the prevalence of certain non-vaccine HPV types (ie, unmasking). Given that the increases in certain HPV types were apparent between the prevaccination and postvaccination periods, but remained relatively stable within the postvaccination period, this suggests that unmasking is not playing a huge role in these increases. However, while adjustment was made for the change in assay between the two periods, to what extent increases in non-vaccine types are due to temporal changes, changes in the population undergoing chlamydia screening, or changes in the detection accuracy of assays, is still somewhat unclear.

In England, the quadrivalent vaccine was introduced to the national HPV immunisation programme from 2012 as part of routine vaccination of 12-year-old girls. At the time this surveillance was conducted, the oldest women vaccinated with the quadrivalent vaccine as part of the national programme would have been 14 years old, hence, too young to be included in this surveillance (conducted among individual aged 16–24 years). Therefore, all women included in this surveillance who were vaccinated as part of the national immunisation programme would have received the bivalent HPV vaccine. We were unable to link these specimens to individual HPV vaccination status, and coverage estimates were derived from published data. This meant that we considered population-level impact of HPV vaccination rather than direct calculation of vaccine effectiveness. Our findings of reductions in the prevalence of the HPV vaccine types are consistent with surveillance conducted in other countries, although changes in the prevalence of non-vaccine HR-HPV types varied.27 Tabrizi et al14 showed a 77% reduction in the prevalence of HPV vaccine types among young women attending for a Pap test in Australia. In the USA, reductions in the prevalence of HPV vaccine types were 56% in individuals aged 14–19 years, despite a low self-reported vaccination coverage (34% with one or more doses). 17 In Sweden, surveillance also among women attending for chlamydia screening found a reduction of 42% in HPV16 and 46% in HPV18 among females aged 13–22 years, and also a slight increase in HPV52 and 56.19 In Scotland, where cervical screening is offered from age 20 years, a 54% reduction in the vaccine types has been shown in individuals aged 20 years, as well as a 48% reduction in the cross-protection types HPV31, 33 and 45.18

We have analysed HPV type-specific prevalence among almost 10,000 women over a period of 5 years. These data provide clear evidence of a reduction in the HPV vaccine types, and a suggestion of a reduction in HPV31, a closely related HPV type, since the introduction of the HPV immunisation programme in England. This will both inform future decisions regarding HPV vaccination in England and be of interest to other countries seeking to monitor the impact of HPV vaccination.

Acknowledgments

The authors are grateful to the staff at the participating laboratories who have provided residual specimens for testing: Bridget Reed, Deborah Blundell, Ian Robinson and Mike Rothburn at University Hospital Aintree; Heather Etherington, Amanda Ronson-Binns and Susan Smith at Leeds Teaching Hospital; Nick Doorbar and David Frodsham at University Hospital of North Staffordshire; Gail Carr and Laura Ryall at Public Health Laboratory, Cambridge, Addenbrooke's Hospital; Samir Dervisevic and Emma Meader at Norfolk and Norwich University Hospital; Roberta Bourlet and Marie Payne at East Kent Hospitals University; Kevin Chittock and Emma Hurley at Source Bioscience; Allyson Lloyd and Colin Walker at Queen Alexandra Hospital; Vic Ellis and Kathy Pollard at Royal Cornwall Hospital; Caroline Carder at University College London Hospital; Ruth Hardwick, Tacim Karadag and Paul Michalczyk at University Hospital Lewisham. The authors also thank the Chlamydia Testing Activity Dataset (CTAD) team, particularly Alizera Talebi, Emma Hollis, Ana Harb and Paul Vanta for assistance with conducting the data linking. They also thank Tracey Cairns and Krishna Gupta for helping in data processing.

References

View Abstract
  • 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 This surveillance was initiated and designed by KS. DM and KS were responsible for the sample collection and data management. SB and KP performed the HPV testing. DM conducted the statistical analysis. SLT contributed to the data analysis and interpretation. DM, KS and SLT wrote the first draft of the manuscript. All authors contributed to and approved the final draft.

  • Funding This work was supported by Public Health England.

  • Competing interests None declared.

  • Ethics approval South East Research Ethics Committee.

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

  • Data sharing statement No additional data are available.

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.