HPV-related oropharynx cancer in the United Kingdom – an evolution in the understanding of disease etiology

A rising incidence of oropharyngeal squamous cell carcinoma (OPSCC) incidence has occurred throughout the developed world, where it has been attributed to an increasing impact of human papillomavirus (HPV) on disease etiology. This report presents the findings of a multicenter cross-sectional retrospective study aimed at determining the proportion of HPV-positive and HPV-negative OPSCC within the United Kingdom (UK). Archival tumor tissue blocks from 1602 patients previously diagnosed with OPSCC (2002-2011) were collated from 11 centers. HPV status was determined with 3 validated commercial tests to provide valid data for 1474 cases in total. Corresponding national incidence data from the same decade were obtained from UK Cancer registries. The overall proportion of HPV+ OPSCC between 2002-2011 was 51.8% (95% CI:49.3, 54.4) and this remained unchanged throughout the decade (unadjusted risk ratio:1.00 (95% CI:0.99, 1.02). However, over the same period, the incidence of OPSCC in the broader UK population underwent a 2-fold increase (age standardised rate (ASR) 2002:2.1 (95% CI:1.9, 2.2); 2011:4.1(95% CI:4.0, 4.3)). Although the number of OPSCC diagnosed within the UK from 2002-2011 nearly doubled, the proportion of HPV+ cases remained static at ~50%. Our results argue that the rapidly increasing incidence of OPSCC in the UK cannot be solely attributable to the influence of HPV. The parallel increase in HPV+ and HPV- cases we documented warrants further investigation, so that appropriate future prevention strategies for both types of disease can be implemented.


Introduction
The developed world has experienced a dramatic rise in oropharyngeal squamous cell carcinoma (OPSCC) incidence. [1,2] In England, the Age-Standardised incidence Rate (ASR) for OPSCC approximately tripled in men (from 2.0 to 5.8) and doubled in women (from 0.8 to 1.7) between 1995 and 2011. [3] Associations between tobacco and alcohol consumption and OPSCC are well established, [4] however sexual behaviour is also a risk factor, with lifetime number of oral sex partners recognised as the behavioural measure most strongly associated with OPSCC development. [5] Changes in sexual behaviour appear to underlie the increasing proportion of OPSCC attributable to oncogenic Human papillomavirus (HPV). [1,4,6] Several North American and European studies have confirmed sharp rises in HPV-induced OPSCC incidence, although the exact proportion of HPV-positive tumours within the total disease burden varies considerably by geographical region. [7][8][9][10][11] In the UK, the proportion of OPSCC attributable to HPV has been assessed in several single-centre studies, however each was small, applied diverse methodology and had restricted geographical coverage. [12][13][14] The current pan-UK study aimed to assess the proportion of OPSCC attributable to HPV infection in a large contemporary sample (2002-2011 inclusive) using robust, standardised methods. There is a pressing need for these data to facilitate health economic analyses and to inform evidence-based policy-making with regard to prophylactic male HPV vaccination, as has recently been implemented in Australia. [15,16]

Case selection
The study received Research Ethics Committee approval (REC 11/NQ/0452). Northern Ireland samples were accessed under approvals from the Northern Ireland Biobank (NIB 11/001). OPSCC was defined as cancers involving the base of tongue (C01), soft palate and uvula (C05.1 & C05.2), tonsil (C09) and oropharynx-not-otherwise-specified (C10.9). OPSCC cases diagnosed between 2002 and 2011 (inclusive) were collected from 11 recruiting centres distributed across the UK to ensure results were not distorted by effects in one area or centre (Belfast, Bristol, Cardiff, Coventry, Edinburgh, Liverpool, London, Manchester, Newcastle, Poole and Southampton). The overall target sample-size (1710) was sufficient to allow comparison of prevalence between years with 7.5% precision. The number of samples per centre was determined pragmatically, based on the number of cases seen annually at each centre. To avoid selection bias, the first 17 cases per year (11 cases for Coventry and Bristol) with available formalin-fixed paraffin embedded (FFPE) tumour blocks were included (irrespective of the definitive treatment modality employed). A representative FFPE block, either from diagnostic or resection specimen, was selected. Gender, age at diagnosis, year of diagnosis and histological diagnosis, including anatomical subsite classification, were recorded.

HPV testing
Sections of each FFPE block were taken for DNA analysis. To prevent DNA contamination, the microtome was thoroughly cleaned between specimens and a new blade used for each block. Tissue microarrays (TMA) were constructed for p16 immunohistochemistry (IHC) and high risk HPV DNA insitu hybridisation (ISH) testing as previously described. [17] Following construction, haematoxylin and eosin-stained sections of the TMAs were analysed to confirm accuracy of sampling. Samples were considered adequate only if all three TMA cores included tumour.
p16 IHC was undertaken as a surrogate marker of HPV oncogene expression [18]  Statistical analysis The characteristics of included cases (gender, age at diagnosis, oropharyngeal subsite, year of diagnosis and study centre) were described using frequencies/percentages or means/standard deviations as appropriate. Characteristics were compared with excluded cases, and with Cancer Registry data using t-tests or chi-squared tests as appropriate. The proportion of HPV-associated cases was calculated for the whole sample, then for each subset (with 95% confidence intervals).
Age was the only continuous variable, and was categorised into five groups. The proportion of cases positive by p16 IHC, high-risk HPV DNA ISH and HPV PCR was also calculated to allow comparison with studies reporting these endpoints, as was the prevalence of HPV types among cancers caused by a single HPV type. Trends in the proportion of HPV-associated cancers over time were assessed using Poisson regression with robust error variance. [22] This approach was used because odds ratios (obtained from logistic regression) are poor approximations of risk ratios if the outcome prevalence is high. Models were fitted before and after adjusting for sample characteristics. Finally, HPVpositive proportions determined in the current study were applied to the UK incidence data to estimate the burden of oropharyngeal cancers caused by HPV over the period.  Figure 1 illustrates the application of study inclusion criteria and the HPV diagnostic testing algorithm. Valid results were obtained for 1474 cases obtained from 11 centres. Sample characteristics (gender, age at diagnosis, oropharyngeal subsite, year of diagnosis and study centre) are shown in Table 1. The mean age of patients was 59.3 years, 75.0% of patients were male, and the majority of cases (57.9%) were tonsil cancers. Invalid results were obtained for 55 patients and these were excluded from the analysis. The age, gender and subsite distribution were similar for the included and excluded samples, and excluded samples were evenly distributed across the study period. The reasons for exclusion were either absence of tumour in the TMA cores or loss of TMA cores during processing for staining.

Case characteristics
The age and gender distributions were compared between OPSCC cases included in the current study, and those reported by UK Cancer Registries for the same period (Supplementary Table S1).

HPV Prevalence
The prevalence of HPV infection in OPSCC was 51.8% (95% CI:49.3, 54.4). The prevalence of HPV infection within specific subgroups is shown in Table 2 To estimate the burden of oropharyngeal cancers caused by HPV over this period, the proportions determined in the current study were applied to the incidence data ( Figure 4). Figure 4 highlights increasing incidence of both HPV-positive and negative OPSCC, especially in men. It was notable that incidence curves for HPV-negative OPSCC and mouth cancers show very similar trends. With regard to non-HPV associated head and neck cancers in males, substantial absolute increases in ASR were observed for HPV-negative OPSCC, and for mouth cancers, with a smaller increase in laryngeal cancers (1.72, 1.14, and 0.73 /100,000 respectively).

Discussion
This first nation-wide study investigating the prevalence of HPV in OPSCC within the UK showed that 51.8% (95% CI: disease incidence that has not previously been reported and show that, in the UK at least, the increasing incidence of OPSCC, cannot be explained solely by an increase in HPV-associated disease. The strengths of the study include: large sample size; broad geographical representation; rigorous and systematic case selection; and, use of well-validated commercial tests to identify HPV tumour status. The results are likely to reflect national trends, although there is potential for variation in OPSCC incidence, and in HPV prevalence, between different geographical areas in the UK. To assess potential bias in case selection, the records supplied by each centre were formally reviewed. This showed that FFPE blocks from only nine patients were unavailable due to use in other clinical studies. The study group included a higher proportion of younger patients (45-54.9 yrs) relative to the OPSCC population, but given the younger mean age for HPV-positive patients, this would be more likely to result in overestimation of the proportion of HPV-associated disease, rather than underestimation. The HPV testing regime included three independent, well-validated, commercial tests (IHC, ISH and PCR), performed in independent laboratories and the three tests showed highly similar trends in HPV prevalence (Figure 2). The analysis of data pertaining to behavioural factors, such as smoking and sexual history, could potentially have allowed further interpretation of our results, however due to the retrospective nature of sample and data collection these data could not be reliably obtained.
It is important to stress that conclusions based on the data presented should not be generalized beyond the UK. Substantial variation has been reported in the proportion of OPSCC attributable to HPV between countries and time periods. [10,11] This is likely to be a reflection of variations in multiple factors, including sexual behaviour and rates of genital HPV infection, as well as tobacco and alcohol consumption. This highlights that trends in the aetiology of OPSCC must be considered in a population-specific manner. Previous small, single centre studies from the UK reported HPV prevalence rates in OPSCC of 37.5% (95% CI:28, 48%), 42.7% (95% CI:36, 50%), and 55% (95% CI:45, 66%). [12][13][14] The current study is consistent with these, but is based on a much larger sample with broader geographical representation, including centres in all four countries of the UK. We observed a consistent proportion of HPV-positive OPSCC over time from 2002 to 2011, against a background of increasing incidence. This contrasts with previous data, detailing an increased proportion of HPVpositive OPSCC associated with increasing incidence of OPSCC overall. [8] Recent data from North America [10] and Stockholm, [23] however, suggest that plateaus in the proportion of HPV-positive OPSCC and HPV-positive tonsillar squamous cell carcinoma respectively, have been observed from the year 2000 onwards.
The absence of change in the proportion of HPV-associated disease, despite a continued rise in incidence of OPSCC, implies that HPV-negative OPSCC, traditionally associated primarily with smoking [24], is also increasing in incidence. However, this contrasts with the more modest increase in incidence of other smoking-related head and neck malignancies, such as laryngeal cancer, [25] ( Figure 3) and suggests that another risk factor, in addition to HPV and smoking, may contribute to the increased overall incidence of OPSCC. Prior tonsillectomy appears to reduce risk of tonsillar carcinoma, [26,27] but while the current UK tonsillectomy rate is approximately 75% lower than in the 1950s, [28] the absence of a disproportionate increase in OPSCCs specifically involving the tonsils, in our results and other published data, [25] suggests this is not a major contributory factor to the increasing incidence of OPSCC. The observed increases in ASR among the different subsites (non-HPV OPSCC, mouth, and laryngeal) may reflect the degree of exposure of specific anatomical sites to individual carcinogens, including alcohol and tobacco smoke. In an analysis of Dutch HNSCC incidence, van Monsjou et al. [29] suggested that behavioural changes in the post-World War II generation included reduced smoking rates coupled with significant rises in alcohol consumption and they suggested that excessive alcohol intake may be a more critical risk factor for OPSCC than smoking. In the UK, smoking rates have declined from 46% of adults in 1974 to 20% in 2010 [30,31].
However, since 1950, per capita alcohol consumption has increased from 3.9 L/year to a peak of 9.4 L/year in 2004 [32]. This appears consistent with the increasing incidence of cancers at sites with greater exposure to alcohol (e.g mouth) but smaller increases at sites more strongly associated with tobacco smoking (e.g. larynx).
It is probable that gender-neutral prophylactic HPV vaccination could prevent HPV-positive OPSCC. [33] Indeed national bodies, such as the Joint Committee on Vaccination and Immunisation in the UK, are currently considering extending prophylactic HPV vaccination to include boys, as well as girls. Our data substantially expand the evidence base available to inform decisions such as this, particularly when viewed in the context of a projected substantial continued rise in OPSCC incidence of up to 239% in the next 20 years. [34] The current bivalent and quadrivalent HPV vaccines protect against infection with oncogenic HPV types 16 and 18; these types were present in 714/764 cases were associated with HPV types included in current vaccines.
The parallel increase in both HPV-positive and HPV-negative tumours should be of concern to those involved in the clinical management of OPSCC, and to public health officials charged with developing strategies to reduce incidence. The data presented highlight that in the UK, increases in OPSCC incidence are not entirely due to HPV-associated disease. However, these findings should not be extrapolated to other developed world populations, rather, they emphasise the need to assess the aetiology of head and neck cancers, and oropharyngeal cancers in particular, on a population specific basis. Year of diagnosis        For each year, the gender-specific proportion of HPV positive samples was multiplied by the gender-specific incidence to estimate ASR for both HPV positive and negative OPSCC. Incidence of cancers of the mouth is also shown.