Predictors of oropharyngeal cancer survival in Europe

FIPV16-positive oropharyngeal cancer (OPC) patients experience better outcomes compared to HPV16-negative patients. Currently, strategies for ffeatment de-escalation are based on HPV status, smoking history and disease stage. However, the appropriate cut-point for smoking and the role of other non-clinical factors in OPC survival remains uncertain. We examined factors associated with OPC outcome in 32 I patients recruited in a large European multi-center study. Seropositivity for HPVI6 E6 was used as a marker of FIPVl6 positive cancer. Hazard ratios (HR) and confidence intervals (CI) were estimated using Cox proportional models adjusted for potential confounders. Overall 5-year survival following OPC diagnosis was 50%o. HPV16-positive OPC cases were at significantly lower risk of death (HR:0.49, 95ohCl:0.34- 0.70). A significant effect on OPC survival was apparent for female sex (HR 0.52: 95o/o CI; 0.3- 0.91) and being underweight at diagnosis (HR: 2.43, 95oh CI: 1.4- 4.22). A l0 pack year smoking history was not associated with survival. This study confirms that HPVI6 status is an independent prognostic factor for OPC survival while female sex lowers risk of death and being underweight at diagnosis increases the risk of death. Smoking was not an independent predictor of OPC survival. The role of patient-related factors in oropharyngeal cancer survival is unclear. Here we show that in addition HPVI6 status and disease stage, being female was associated with nearly 50% reduced risk and low BMI diagnosis almost doubled the risk of death following oropharyngeal cancer. of where the maximum score of eight reflected Body mass index was calculated based on weight measured at the time of BMI ranging fiom 18.5 to 25.0 was considered normal, below 18.5 underweight while >25.0 was considered overweight. Informed consent was obtained from ali participants in the study and the study was approved by the ethical review boards at the participating centers. Preffeatment serum samples were tested for HPV antibodies using the bead-based multiplex serology method [ 17, l8l. We have previously shown that HPV16 E6 antibody is a highly specific marker of HPVI6+ OPC with false-positive rates less than lYo[L1, 19-211. For comparison, 198 paraffin-embedded OPC tumor blocks were qualitatively evaluated for pl6 expression using the CINtec Histology Pl6INKaa Kit (9511, mtmlabs) following manufacturer's instructions. Expression was scored based on the percentage and intensity of nuclear or cl.toplasmic staining, A combined score of 4 or greater was considered positive for p1(tNraa overexpression. HPV genotyping was performed using the Type-Specific E7 PCR bead-based multiplex assay (TS-E7-MPG, IARC, France)[17, 22f . Given that HPV serology was available on all cases; our analysis defined a HPV+ tumor as HPVI6 E6 antibody positive. We subsequently compared this to tumor results based on p16 expression and HPV DNA. Participants underwent a one-time follow-up conducted between 2012 and 2015. Mortality data including cause and date of death were obtained from at least two information sources for 7 5Yo of cases, and one source for the remaining 25Yo.ln Prague and Aviano follow-up was completed by review of medical charts alone. In all other centers, medical chart reviews together with information from population-based registries at the regional or national level were used. at censor, or date of death (if applicable). Over 96%o of OPC patients' recruited in the study have complete follow-up. Overall survival (all-cause moftality) was evaluated using Cox proportional hazard models, predictors were explored for OPC overall and stratified by stage, sex and HPVI6 status. Mortality was also explored using Kaplan Meier curves. The joint effects on survival were considered by combining cofactors in interaction models. We used chi-squared tests to examine heterogeneity in hazard estimates. All statistical analyses were performed using statistical Statistical significance was at P less than 0.05. being female and OPC survival among HPVI6+ (HR: 0.30, 95% CI:0.09- 0.95) orHPVl6- OPC (HR:0.67,95yo CI:0.36- 1.37,(P-heterogeneity:0.22).Conversely, the increased risk of death associated with being underweight at diagnosis was consistent for stage I and II (HR: 1.51, 95o/oCl: 0.17-13.30), stage III (HR: 4.94, 95ohCl:1.77-13.81) and stage IV disease (HP.:2.22,95%Cl:1.08-4.55), with no heterogeneity in the estimates (p for heterogeneity: 0.39). We further examined these findings in the context of the recently proposed staging for HPV+ OPC that combines disease stage at diagnosis, age and smoking pack-years into four groups with differing prognosis; group I comprising of stage I-I[ OPC patients' with < 20 pack years, group II of stage I-III patients' with >20 pack years, group III were stage IV and <70 years old and group IV were stage IV with >70 years of age [0]. Consistent with the initial report, overall survival decreased across these groups (data not shown). Notably, female sex (HR: 0.54,95o/o CI: 0.33- 0.88) and being underweight at diagnosis (HR for: 2.59,95yo CI: 1.50- 4.46)remained consistently associated with OPC survival, independent of the suggested prognostic risk classification. death among HPV16+ OPC in this study was similar in magnitude to previous studies 231. Our results indicate that additional factors may have an important effect in predicting survival following OPC. In particular that women may have approximately 50% reduced risk of death compared to men following an OPC diagnosis. Importantly, this benefit was extended to late stages of disease and was consistent for both HPVI6+ and HPVI6- OPC suggesting the protective effect may be driven by inherent factors. Although an individual's gender has been long recognized as a key factor affecting cancer incidence, prognosis, and treatment response, the underlying mechanisms remain poorly understood. We also report that being underweight at diagnosis increases the risk of death following it following disease prior towards improved association with BMI at 30 years age and OPC survival in this not to note that sex independently associated even when considering recently proposed progrrostic risk classification OPC. These findings warrant further validation in independent studies. of the varying definitions for pack years increments of I pack year, 10 pack years or tbresholds of 10 and 20 pack years). Sequential adjustment of covariates indicates that tIPVl6 status and disease stage could potentially account for the survival differences due to smoking. Further, we found no interaction between smoking and HPV16 status in the risk of death following OPC. These results call for additional evidence in larger cases series in order to accurately establish the role of smoking and survival among HPVI6+ OPC. This study demonstrates a strong association between HPV 16 status and OPC survival. Even though p I 6 and HPV I 6 DNA each showed utility in predicting mortality, they both appeared to be sub-optimal proxies for HPV16 status.


Introduction
Cancers arising in the oral cavity and pharynx have an estimated global burden of 442,760 incident cases and 241,458 deaths each year[]. Tobacco smoking and alcohol consumption explain nearly 70%o of these cancers [2,3].1 Infection by human papillomavirus (HPV), particularly type 16 GIPVI6) causes a subset of cancers, particularly those arising at the tonsil, oropharynx, soft palate and base ofthe tongue (collectively referred to as oropharyngeal cancers-OPC) [4]. Further, HPVl6-positive (HPVl6+) OPC is described to be epidemiologically, molecularly and clinically distinct from HPVl6-negative (HPVI6-) OPCt5l. The increasing incidence of OPC in several Westem countries is athibuted to the HPV16+ fraction[6-9].
Since HPVI6+ OPC patients experience better survival outcomes compared to HPVI6patients, alternative stagidg has been recommendedtl0-l2]. Recurrence remains a concem and it is unclear which patients may benefit from deintensified treatment. Clinically, HPV status is ascerlained based on HPV DNA and p16 expression or pl6 expression alone. HPV status in combination with disease stage and patient smoking history @ased on a l0 or 20 pack year cut off) is currently used to classi$ patients into prognostic groups and to identifu candidates for deescalation of treatment[0, l3]. This scheme has rarely been verified. Further, the appropriate cut-point for pack years of smoking remains uncefiain. tn addition, the role of other non-clinical risk factors in OPC survival is not fully understood.
To address these knowledge gaps, we tested 321 oropharyngeal tumors in a large series of well characterized European patients for HPVI6 DNA, pl6 expression and the corresponding sera for HPV16 serology. Applying rigorous protocols of sample processing; we aimed to evaluate the role of HPV16 and other risk factors in predicting OPC survival and recurrence.

Methods
This analysis was based on cases from the European Alcohol Related Cgncers and Genetic susceptibility il Europe (ARCAGE) study, conducted across l0 countries in Europe using a standardized protocol [14]. Briefly, over 2000 incident cases of the oral cavity, pharynx. larynx, esophagus and matched controls were recruited during 2002 to 2005. This analysis included squamous cell carcinoma of ICD-O diagnoses C01,C02.4, C05.1-C05.2, C09, Cl0. All participants underwent personal interviews to record lifestyle exposures. All cases were histologically or cytologically confirmed primary cancers, and cancer stage was ascertained based on the sixth edition of the staging atlas developed by the American Joint Committee on Cancer (AJCC). Tobacco use was broadly categorized as ever or never smokers, ever smokers were defined as individuals who smoked any tobacco product at least once a week for a year. Pack years were calculated for all types oftobacco smoking based on clgarette equivalents. Ever drinkers were those who reported ever consumption of any alcoholic beverage and the consumption of all types of alcoholic beverages was estimated and the total frequency was expressed in terms of drinks of alcohol per day [5]. A weighted composite score of oral hygiene and dental care was constructed as described previously [ 1 6] and included denture wear, age at start of denture-wearing and gingival bleeding. A weighted dental care score was also constructed by combining the frequency of tooth cleaning, use of toothpaste, toothbrush or dental floss and frequency of dentist visits, where the maximum score of eight reflected poor dental care. Body mass index was calculated based on weight measured at the time of recruitment. BMI ranging fiom 18.5 to 25.0 was considered normal, below 18.5 underweight while >25.0 was considered overweight. Informed consent was obtained from ali participants in the study and the study was approved by the ethical review boards at the participating centers.
Preffeatment serum samples were tested for HPV antibodies using the bead-based multiplex serology method [ 17, l8l. We have previously shown that HPV16 E6 antibody is a highly specific marker of HPVI6+ OPC with falsepositive rates less than lYo[L1, 19-211. For comparison, 198 paraffin-embedded OPC tumor blocks were qualitatively evaluated for pl6 expression using the CINtec Histology Pl6INKaa Kit (9511, mtmlabs) following manufacturer's instructions. Expression was scored based on the percentage and intensity of nuclear or cl.toplasmic staining, A combined score of 4 or greater was considered positive for p1(tNraa overexpression. HPV genotyping was performed using the Type-Specific E7 PCR bead-based multiplex assay (TS-E7-MPG, IARC, France) [17, 22f . Given that HPV serology was available on all cases; our analysis defined a HPV+ tumor as HPVI6 E6 antibody positive. We subsequently compared this to tumor results based on p16 expression and HPV DNA.
Participants underwent a one-time follow-up conducted between 2012 and 2015. Mortality data including cause and date of death were obtained from at least two information sources for 7 5Yo of cases, and one source for the remaining 25Yo.ln Prague and Aviano follow-up was completed by review of medical charts alone. In all other centers, medical chart reviews together with information from population-based registries at the regional or national level were used.
In Athens, Barcelona and Manchester physicians were contacted to obtain patient outcome information, while in Oslo, Zagreb and Glasgow cancer registries were consulted. In Bremen, Turin, Padova and Dublin mortality registries were examined. End of follow-up was defined as the date of last confirmed contact, vital status at censor, or date of death (if applicable). Over 96%o of OPC patients' recruited in the study have complete follow-up. Overall survival (all-cause moftality) was evaluated using Cox proportional hazard models, predictors were explored for OPC overall and stratified by stage, sex and HPVI6 status. Mortality was also explored using Kaplan Meier curves.
The joint effects on survival were considered by combining cofactors in interaction models. We used chi-squared tests to examine heterogeneity in hazard estimates. All statistical analyses were performed using STATA statistical software, version 11 (StataCorp, College Station, TX), and all reported P values are two sided. Statistical significance was set at P less than 0.05.

Results
Subjects were primarily male (7'7%), ever smokers (90o ), alcohol drinkers (97%) and had a median age at diagnosis of 58. The vast majority of OPC were diagnosed at late stages; 49o/oinstage IV, 25Yoinstage III and26%oin stages I and II. 31% of OPC cases were HPVI6 E6 positive. OPC subjects were followed for 1257 person years during which 175 deaths occurred (mean follow-up of 3.92 years), of which 98 (560/o) were due to head and neck cancer.
Overall S-year survival tbllowing OPC diagnosis was 50% (g5% CI: 43.9-54.9). As described in table l, in the univariate analyses there was significantly lower risk of death among 99 HPVI6+ compared to 222 HPV-OPC  Table   2). Three year recurrence-free survival was 77o/o in HPVI6+ and 660/o in IIPV16-OPC (P:0'03, Figure 2 In this study, 198 corresponding tumors of 321 cases were tested for HPV DNA and p16 expression. Of these, 16 were excluded due to limited tissue. Of the remaining 156 tumors, 52 were HPV16 E6 seropositive of which 43 were positive for both HPV16 DNA and p16 expression while 44 were positive for p16 expression alone (Table 3).

Discussion
This study confirms I-IPVI6 status as an independent prognostic factor for overall survival and recurrence-free survival among OPC patients'. We also report that women are at lower risk of death (both [IPV+/FPV-), while being underweight at diagnosis increases the risk of death following OPC diagnosis.
The reduction in risk of death among HPV16+ OPC in this study was similar in magnitude to previous studies [3, 231. Our results indicate that additional factors may have an important effect in predicting survival following OPC. In particular that women may have approximately 50% reduced risk of death compared to men following an OPC diagnosis. Importantly, this benefit was extended to late stages of disease and was consistent for both HPVI6+ and HPVI6-OPC suggesting the protective effect may be driven by inherent factors. Although an individual's gender has been long recognized as a key factor affecting cancer incidence, prognosis, and treatment response, the underlying mechanisms remain poorly understood. We also report that being underweight at diagnosis increases the risk of death following OPC by over two-fold. This association was robust across stage, sex and HPVI6 status. Although it remains probable that low BMI at diagnosis reflects the disease itself, since cancer patients tend to lose weight rapidly following disease onset, it could also point towards the need for nutritional interventions prior to clinical treatment as a strategy towards potentially improved outcomes. The lack of association with low BMI at 30 years of age and OPC survival in this study (data not shown) supports this notion. lt remains important to note that both low BMI at diagnosis as well as female sex remained independently associated even when considering the recently proposed progrrostic risk classification for OPC. These findings warrant further validation in independent studies. Previous studies have found tobacco smoking to be independently associated with overall survival as well as progression-free survival among OPC patients[s, 10,13,24]. Smoking, in particular, 10 or 20 pack year thresholds in conjunction with"tumor stage and FIPV status have therefore been proposed for risk stratification and [0, l3]to identifu patients who may benefit from keatment de-escalation. In this study, the lack of association between smoking and OPC survival was consistently observed irrespective of the smoking parameter examined (i.e., smoking status, or varying definitions for pack years including increments of I pack year, 10 pack years or tbresholds of 10 and 20 pack years). Sequential adjustment of covariates indicates that tIPVl6 status and disease stage could potentially account for the survival differences due to smoking. Further, we found no interaction between smoking and HPV16 status in the risk of death following OPC. These results call for additional evidence in larger cases series in order to accurately establish the role of smoking and survival among HPVI6+ OPC.
This study demonstrates a strong association between HPV 16 status and OPC survival. Even though p I 6 and HPV I 6 DNA each showed utility in predicting mortality, they both appeared to be sub-optimal proxies for HPV16 status.
While the utility of tumor HPV16 DNA and pl6 as diagnostic markers of OPC is now well established [25-27], this study confirms the utility of HPVI6 E6 serology as a robust prognostic marker of OPC.
Despite the advantage of the multi-centric design and the large number of participants enrolled in the parent study, the OPC subset was limited to 321 cases. Even so, to the best of our knowledge, this remains among the largest analysis of this rare can€er with detailed information on risk factors, patient outcome as well as HPV16 status. Important strengths of this study are the centralized testing for HPV16, uniform collection of survey and cancer outcome information in all centers. Some of the limitations include the absence of detailed ffeatment information that could potentially impact survival. However, the majority of the patients' were treated with surgery alone or in conjunction with chemotherapy or radiation therapy (68%). Further, disease stage was missing on 16%o of OPC since the study was initially designed with a focus on risk factors for cancer occurrence, therefore comprehensive collection of clinical data was not emphasized.
In conclusion, this study confirms the markedly improved prognosis associated with HPV+ OPC in Europe. In addition, we report that being female and low BMI at diagnosis are important factors affecting overall OPC survival.
Further large studies that also include comprehensive mutational profiling will be required to better understand the mechanisms underlying these prognostic subgroups.