Meta-analysis of the impact of human papillomavirus (HPV) on cancer risk and overall survival in head and neck squamous cell carcinomas (HNSCC)

Squamous cell carcinoma of the head and neck (HNSCC) has an estimated incidence of 35,310 new cases in 2008 in the United States, with an expected 7590 deaths due to these cancers. The male to female ratio is approximately 2:1[1]. A recent analysis of the Surveillance, Epidemiology and End Results (SEER) database showed in younger U.S. populations (ages 20-44 years) that while there has been an increase in incidence of tonsillar squamous cell carcinomas from 1973 to 2001, the incidence of squamous cell carcinomas in all other oral and pharyngeal sites remained constant or decreased[2]. A similar rise in the incidence of tonsillar squamous cell carcinomas from 1970-2002 has been shown in Sweden and has been associated with presence of human papillomavirus (HPV)[3]. For more than 30 years, certain genotypes of "high-risk" HPVs have been known to be involved in the pathogenesis of cervical cancers[4]. Recent studies have implicated high-risk HPV as a risk factor for HNSCC, independent of the traditional risk factors, which include tobacco abuse and ethanol consumption[5–7]. While high-risk HPV related HNSCC appear to be associated with certain sexual behaviors, such as oral sex and increasing numbers of sexual partners, there is a lack of association with smoking and drinking[5, 8]. Clinically, high-risk-HPV related HNSCC tend to present with lymph node positive disease and originate from the oropharynx, while histologically these tumors are usually high-grade and can be described as exhibiting a basaloid morphology [6, 9, 10]. On a molecular level, the HPV oncoproteins E6 and E7 are implied in tumorigenesis and are known to induce degradation of the tumor suppressors p53 and pRB, respectively[11, 12]. Another hallmark of HPV related HNSCC is lack of p53 mutations[13] and p16 protein overexpression[14–16], a result of loss of transcriptional repression which occurs as a response to upstream signals during early tumorigenesis[17]. Studies suggest that p16 immunohistochemical expression can function as a surrogate marker to predict treatment outcome in HNSCC[14, 18].

Recent emerging data indicates that HPV related HNSCC define a unique population of patients with distinct biology that likely should be treated separately from non-HPV related HNSCC[19]. The aims of this meta-analysis are to consolidate the available data, improve understanding of the biology of HPV related HNSCC, and thereby identify goals for future research and trial design in HNSCC.

We have conducted a literature review and large meta-analysis in HPV related HNSCC to determine the prevalence of the disease and assess the ramifications on survival and associated molecular biomarkers. Given prior published reports of a higher prevalence of HPV in carcinomas of the oropharynx, we concentrated this study on oropharyngeal carcinomas to enrich and focus the meta-analysis. Therefore, the number of studies included in our review is smaller than other previously reported meta-analyses with different inclusion criteria[30, 31]. Several other meta-analyses on HPV-related HNSCC have been published before[30–32], but to our knowledge, none of the prior reports have described within the same report, the impact of HPV on survival, response to therapy, as well as the association with molecular markers such as p16. In addition, our report includes recently published prospective studies that were not available at the time of publication of previous meta-analyses. Our data reports the prevalence of HPV-related HNSCC as 21.95% and is in accordance with other recently published reports which estimate the prevalence of HPV at 20-26% [25, 30–32]. As expected, we also confirmed a higher prevalence of HPV-related tumors in the oropharynx (40.97%) in agreement with prior studies[31, 33]. Virtually all of the studies included used PCR to detect HPV DNA, a method with slightly higher sensitivity than in situ hybridization[31]. The higher sensitivity of the detection method used suggests that the herein reported prevalence of HPV associated HNSCC is more likely to reflect the actual prevalence of HPV as compared to studies which use less sensitive methods, although given the paucity of data regarding the exact specificity of PCR in this setting does not enable a precise account of the number of false-positive patients included in these studies. The relatively wide range of the prevalence of HPV in HNSCC in the published literature confirms that in addition to geographical differences, the method of HPV detection is critical for accurate assessment of the HPV status.

In the current study we confirmed the role of p16 immunohistochemical expression as a surrogate marker for HPV infection in HNSCC, which is not only positively associated with HPV infection but also predicts for an improved survival outcome[14, 28, 29]. The HPV oncoprotein E7 is known to induce degradation of pRb, which in turn leads to upregulation of p16[34]. Therefore, overexpression of p16 is considered to be one of the molecular hallmarks of HPV positive HNSCC[4, 23, 35, 36] and has recently been shown to identify HPV related oropharyngeal cancers with favorable outcomes[16]. In a recent prospective trial, an analysis of tumor tissue (n = 323) collected from RTOG 0129 using HPV in situ hybridization (ISH) and p16 immunohistochemistry (IHC) demonstrated that p16 correlated well with HPV ISH with 96% of HPV positive tumors also expressing p16[19]. An additional 19% (n = 22) were p16 positive but HPV negative. Both HPV and p16 positive status predicted for an improved survival. This finding of p16 IHC positive tumors identifying more patients than HPV ISH was also seen in an additional study called HeadStart[37]. This trial compared HPV ISH and p16 IHC on tumor specimens from two arms of therapy, cisplatin-radiation versus cisplatin-tirapazamine-radiation, and suggested that p16 IHC may identify a higher percentage of HPV positive tumors than the HPV ISH assay. Given the debate regarding the best detection method for HPV[38], our data from the meta-analysis confirms that p16 IHC could be used in future in prospective trials as a reliable surrogate marker for HPV related HNSCC[19, 37, 39]. As the above two recent studies (RTOG and HeadStart) have only been reported in abstract form, we did not include them in our meta-analysis.

As HPV-related oropharyngeal cancers are reported to be rising in incidence[2, 3], it is important to consider treating this disease as a unique entity. Our data in congruence with other studies indicate that HPV positive HNSCC have a distinct biology, and are more responsive to treatment with radiotherapy and chemotherapy. In addition to treatment response, we and others have shown an improved survival in patients with HPV positive HNSCC, with a 60% lower risk of death. Our survival results are slightly different from other reports; a recent meta-analysis reported a death hazard ratio of 0.85. This discrepancy can be explained by the addition of the recent prospective trials[10, 14, 40, 41] in our analysis which were not included in prior reports. Our meta-analysis also confirms a better response to treatment in HPV positive HNSCC in accordance with recently published trials[10, 14, 39, 41, 42]. Recent reports have highlighted the importance of the negative influence of smoking on prognosis within HPV positive HNSCC[19, 43], and future studies are required to examine this relationship in a prospective manner.

There are several limitations inherent to our methodology, which was based on literature review, not individual patient data. While the total number of patients included in our meta-analysis is large, for many specific associations the number of the studies that actually could be combined to provide information is relatively small. This is a limitation which unfortunately cannot be overcome without access to primary data from all included publications. Also, there is overlap of the same patient series included in multiple reports. We tried to account for this fact as much as possible by reviewing the author list and institution of each paper and including only the largest study if several were apparently from the same group. Finally, given the heterogeneity of the kinds of studies our meta-analysis was based on, it is difficult to combine conclusions drawn, e.g. assessing HPV as risk factor for HNSCC from case series, retrospective and prospective trials. There is no valid formulation to give adequate weight to the results of each kind of the studies mentioned. In spite of the limitations inherent to this type of a meta-analysis, our results are in keeping with published literature and our current understanding of the natural biology of HPV-related HNSCC. We also performed as a validation control a meta-analysis for the established HNSCC risk factors smoking and alcohol consumption. Our results confirmed prior published results [20, 21].

For the future, there is a need for randomized prospective data to identify the optimal treatment for patients based on the tumor's HPV status and to also develop potential prevention programs. First, given the apparent improved response to current treatment modalities, patients with HPV-related HNSCC may be able to receive adequate curative intent therapy from a reduction in radiation or chemotherapy dose. These issues will require future clinical trial designs to either prospectively stratify patients by HPV status or to conduct trials which are dedicated to HPV positive HNSCC. In addition, it is necessary to determine the best method of detecting the HPV status and validate whether p16 overexpression should be utilized as a surrogate or in conjunction with direct HPV testing. Given the etiologic involvement of the HPV oncoproteins E6 and E7 during carcinogenesis of these tumors[44], it is also conceivable that intervention with anti-viral drugs may enhance sensitivity of these tumors to cytotoxic drugs, with the goal of reducing the required dose of chemotherapeutic agents and the associated side effects. Exploratory biomarker studies will be needed to identify prognostic and predictive markers in order to tailor specific therapies (such as biologic agents) to the biology of the patients' malignancy, resulting in better tolerability of treatment with improved outcomes.

The issue of preventing HPV associated HNSCC from occurring requires investigation as well. Two recently approved vaccines to prevent HPV-related cervical cancers[45, 46] cover the two serotypes that are most prevalent in HPV-related HNSCC (HPV-16 and 18). In the context of the findings of our study, it is conceivable that extending vaccination to both sexes would prevent a significant number of future oropharyngeal cancers in both men and women. Also, the hypothesized benefits of HPV-vaccination would potentially include a reduced incidence of other HPV-related cancers (i.e. vulvar, vaginal, anal, and penile cancers)[47]. On the other hand, a recent study showed that vaccination of males on a population basis might not be cost-effective[48]. Even in women, controversy exists regarding the maximum age at which the HPV-vaccine should be given, and a recent simulation model suggested that vaccination beyond 30 years of age might also not be cost-effective[49]. Given all these uncertainties, it will therefore be of great importance to assess in prospective trials that include both sexes the efficacy of these vaccines to prevent HPV-related HNSCC.

HPV-related HNSCC comprise about 25% of all HNSCC. They are predominantly tumors of the oropharynx, and exhibit a separate biologic behavior including improved response to (chemo)-radiation and survival compared to HPV-negative HNSCC. The optimal therapy for this subset of patients remains unclear, but the growing data supports the theory that this subset of patients should be treated differently than HPV negative patients. In addition to this avenue of clinical research, identifying peripheral surrogate biomarkers and possible screening and prevention methods is an important field of future investigation.