Biologically relevant HPV infection, defined as presence of HPV DNA by ISH and/or PCR and p16 over-expression, was identified in 55% of patients diagnosed with OPC in South Wales (UK), 2001–2006. The survival advantage afforded by HPV in a ‘real world’ population of patients with OPC, including those managed palliatively, is clearly demonstrated as is the effect of HPV on the long-term clinical behaviour of the disease. The effect of poor quality DNA in fixed pathological specimens on the diagnostic and prognostic utility of DNA-based HPV testing methods, including ISH, is shown for the first time. P16 expression is not affected by DNA quality and may be utilized as a single marker of HPV infection in clinical practice, although a composite definition of HPV positivity is recommended for accurate HPV prevalence reporting.
HPV prevalence rates differ between different geographical regions and time periods. The rate in this study (55%) is consistent with international (51.2% (88/172)) and US series (59.4% (192/323)) collected between 2002–2005 [7, 17]. It also adds to a picture of regional and temporal variation in HPV prevalence across the UK where rates of 37.5% (33/88) and 42.7% (77/180) have been reported [12, 18]. The number of ‘equivocal’ cases with discrepant HPV DNA and p16 testing results was significantly lower than in some other studies [7, 18], suggesting that a testing algorithm combining PCR and ISH increases sensitivity for HPV DNA detection . Discordant HPV DNA and p16 testing results occurred in 6-7% of cases showing that p16 alone is not sufficient for studies that aim to accurately report HPV prevalence.
Poor quality DNA significantly reduced HPV prevalence estimates using PCR and ISH-based techniques, because of the occurrence of false negative results in samples containing degraded DNA. Although PCR-based testing protocols routinely incorporate assessment of DNA quality, DNA-based ISH techniques do not, and therefore risk under-estimating HPV prevalence; this may partly explain the lower sensitivity reported for ISH compared to other HPV detection methods in previous studies . However a recently developed RNA-based ISH test for HPV does include a control for sample quality and shows considerable promise as a diagnostic marker for OPC . The three HPV testing methods evaluated in this study were all good markers of survival, with no test performing significantly better than another. Hazard ratios (HR) for death were: 0.24 for p16, 0.27 for ISH, 0.29 for GP5+/6+ PCR and 0.22 for the composite definition of HPV positivity. Poor quality DNA reduced the prognostic value of DNA-based HPV testing methods and when poor quality samples were excluded, the prognostic value of GP5+/6+ PCR was similar to that of p16 and the composite marker (0.22). The effect of poor quality DNA on prevalence and prognostication is reduced by using the composite definition of HPV-positivity.
The clinical implications of HPV positivity in this unselected population of patients are clear. HPV positivity was associated with a 78% reduction in death rate (HR 0.22) and a 75% reduction in rate of progression, relapse or death (HR 0.25). This effect is greater than in many clinical trial cohorts, due in part (but not entirely) to the inclusion of palliative patients. Survival of radically treated HPV-positive patients was comparable to that reported in a large US study; 3y OS was 82.6% (95% CI: 73.7 to 91.5) compared to 82.4% (95% CI: 77.2-87.6) and 3y PFS was 72.5% (95% CI: 61.9 to 83.0) compared to 73.7% (95% CI: 67.7 to 79.8). Survival of radically treated HPV-negative patients was however significantly worse; 3y OS was 39.6% (95% CI: 32.5 to 46.7) compared to 57.1% (95% CI: 48.1-66.1) and 3y PFS was 31.2% (95% CI: 18.3 to 44.1) compared to 43.4% (95% CI: 34.4-52.4) . Similarly low survival figures for HPV-negative OPC have been reported previously  and poor performance status (~30% would have been excluded from the US study on this basis) and infrequent use of concurrent chemotherapy (<30% vs 100% in the US study) may have affected outcome in this study. Their prognosis was poor regardless of whether they were treated with primary surgery or RT/CRT.
The excellent outcomes of HPV-positive patients were independent of smoking status or treatment method. Retrospective analyses have suggested that smoking can negatively affect survival in some HPV-positive patients [5, 7, 21], and this data has influenced the design of several clinical trials. Although the relatively small cohort (n = 117) with known smoking history, crude definition of smoking and/or large effect of HPV status on outcome may have masked the effect of smoking in this study, it is possible that the effect of smoking, particularly past smoking, on outcome from HPV-positive OPC has previously been over-estimated, and this issue must be addressed prospectively in future studies. There was a trend for more HPV-positive patients to undergo primary surgery in this study (p = 0.07); although HPV status was unknown when treatment decisions were made, it is likely that selection of younger, fitter patients for surgery, resulted in preferential selection of HPV-positive patients. This highlights the dangers of comparing outcomes from non-randomized studies of surgery and RT/CRT, without knowledge of HPV status; randomized trials with mandatory HPV testing are required to assess treatments for OPC in future.
Improved outcomes in HPV-positive patients reflected better locoregional control rates. In contrast, rates of distant metastases occurring on follow-up were similar in both HPV-positive and negative patients. The occurrence of second HPV16-positive primaries, both in the tongue base and nasopharynx (EBV-negative) in this study is intriguing. Second HPV-associated cancers in the tonsils and nasopharynx and HPV-positive/EBV-negative nasopharyngeal carcinomas have previously been reported [22–24] and it is possible that the lymphoid tissue throughout Waldeyer’s ring is particularly susceptible to HPV-induced transformation. Second primaries occurring in patients with a history of HPV-positive OPC should be tested for HPV and further studies to investigate the frequency and timing of HPV-positive second H&N primaries are required to inform future follow-up protocols.
There are several potential limitations to the study. Histology blocks for 83% of OPC patients presenting across South Wales over the study period were included. There were several reasons why other cases were not included: blocks were not collected from a number of smaller centres, there was limited collection from one major centre due to logistical difficulty in identifying the relevant cases, mismatches were observed in coding between registry and pathology databases, and some blocks were missing from pathology archives. There is no reason to suspect systematic bias in the sample, especially given the multi-factorial reasons for samples not being included, but the potential for some bias cannot be completely excluded. The proportion of HPV-positive OPC is likely to have increased since 2006 and the sample has limited geographical representation, thus whilst it adds to the picture of HPV prevalence in the UK, caution should be exercised in generalising the findings.