Shay SG, Chang E, Lewis MS, Wang MB. Characteristics of Human Papillomavirus–Associated Head and Neck Cancers in a Veteran Population. JAMA Otolaryngol Head Neck Surg. 2015;141(9):790-796. doi:10.1001/jamaoto.2015.1447
The US veteran population represents a unique cohort of patients in whom human papillomavirus (HPV)–related head and neck squamous cell carcinoma (HNSCC) has yet to be investigated.
To investigate the incidence and characteristics of HPV-positive HNSCC within the veteran population.
Design, Setting, and Participants
Retrospective medical record review including patients with HNSCC diagnosed between January 1, 2010, and November 15, 2013, from the VA Greater Los Angeles Healthcare System. Data were collected between November 16, 2013, and June 19, 2014, and analyzed between June 20, 2014, and March 26, 2015.
Chemoradiation therapy, radiation therapy, surgery, or no treatment.
Main Outcomes and Measures
We determined HPV positivity by p16 testing. Demographic and clinicopathologic information and overall survival were extracted from medical records.
We identified 150 patients with the diagnosis of HNSCC. Sixty-nine patients had HPV-positive tumors (46%), and 65 (43%) had HPV-negative tumors (16 did not have HPV testing). Age at diagnosis ranged from 44 to 94 years (mean, 64.6 [SD, 8.0] years), and median (interquartile range) follow-up was 16.7 (8.7-27.3) years. Tumor location differed significantly between the 2 groups, with an HPV-positive predominance in the oropharynx (43 of 57 [75%]; P < .001). The HPV-positive patients were more likely to be treated primarily with combined chemoradiation therapy than radiation therapy or surgery (P < .001). T4 tumors had a nearly 9 times greater rate of mortality compared with T1 tumors (HR, 8.52 [95% CI, 2.60-18.40; P < .001); N3 disease was associated with 7.18 times greater mortality (HR, 7.18 [95% CI, 1.99-12.26]; P < .001) compared with N1 disease; and M1 disease was associated with 6.0 times greater mortality (HR, 5.99 [95% CI, 2.59-13.81]; P < .001). There were 42 total deaths during follow-up, 25 in the HPV-negative group and 17 in the HPV-positive group, with a nonsignificantly higher overall survival among HPV-positive patients independent of alcohol or tobacco use history (P = .09).
Conclusions and Relevance
Previous studies have found that the proportion of HPV-positive HNSCC in the general population ranges between 20% and 75%. Although the incidence of HPV-positive HNSCC in the Veterans Affairs population is comparable, these patients have unique risk factors and demographic characteristics that may suggest different prognostic factors for HPV-positive HNSCC in this population. Nonetheless, HPV-positive tumors still seem to portend a better overall prognosis regardless of alcohol or tobacco history among the Veterans Affairs population.
Human papillomavirus (HPV) infection has been associated with a subgroup of head and neck squamous cell carcinoma (HNSCC), particularly with high-risk subtypes including HPV type 16.1- 5 There has been a steady increase in the incidence of HPV-positive HNSCC, which has been largely attributed to changing sexual practices, despite an overall declining incidence of HNSCC in the United States.6- 10 This increasing incidence of HPV-positive HNSCC has been linked to a concomitant improvement in overall population survival because HPV positivity has been found to be a robust predictor of prognosis independent of treatment regimen.11- 17
Whereas alcohol and tobacco consumption are well-established risk factors for HNSCC,18- 20 HPV-positive HNSCC has been predominantly found in patients who are nonsmokers and nondrinkers.21,22 Patients with HNSCC within the US veteran population possess unique risk factors. Most have a history of tobacco and alcohol use, poorer health status, and tend to present later in their disease course.19,23- 25 As such, HPV-positive HNSCC may behave differently as a result of these risk factors. The objective of this study was to investigate the incidence and characteristics of HPV-positive HNSCC within a veteran population. To our knowledge, our study is the first to investigate the incidence and risk factor profile of HPV-positive HNSCC within a veteran population and to attempt to assess the prognosis of HPV-positive HNSCC within this population.
Patients with HNSCC diagnosed between January 1, 2010, and November 15, 2013, at the VA Greater Los Angeles Healthcare System with available tissue for HPV testing were included. This retrospective cohort study was approved by the institutional review board at the VA Greater Los Angeles Healthcare System. Tissue samples without HPV and/or p16 test results at the time of initial diagnosis were retested. Immunohistochemical stain detection was performed using the DAKO FLEX System. Specifically, HPV detection was performed using DAKO panHPV immunohistochemical staining, which detects HPV serotypes 6, 11, 16, 18, 31, 33, 42, 51, 52, 56, and 58 using an antibody that reacts with a nonconformational, internal, linear epitope of a major capsid protein of HPV type 1 (Clone K1H8, DAKO). Immunohistochemical staining of p16INK4A was also performed using Ventana Medical Systems probes. For the purposes of this study, HPV positivity was determined by means of positive p16 test results in tumor specimens because overexpression of p16 is highly correlated with oncogenic transformation caused by persistent high-risk HPV infections.
Using the VA Greater Los Angeles Healthcare System computerized patient record system, relevant history, demographic, and clinicopathologic information was collected between November 16, 2013, and June 19, 2014, and analyzed between June 20, 2014, and March 26, 2015. If date of death was unavailable and the patient was either lost to follow-up after diagnosis and/or had no encounters in the computerized patient record system in the previous 12 months, the patient was categorized as death status unknown, with the last known encounter noted. Overall survival was used for all mortality comparisons because the deceased patients had a paucity of data regarding cause of death.
Nine covariates (potential confounders) were considered in the analysis: race, tobacco use, alcohol use, primary treatment, primary site (topography codes), T stage, N stage, M stage, and age at diagnosis.
Categorical variables were summarized as frequency (percentages) by HPV status and overall. Categorical variables were compared between the HPV-positive group and the HPV-negative group using the χ2 test. Age, a continuous variable, was summarized as mean (SD) by HPV status and overall and was compared between the groups using the 2-sample t test.
Survival curves were computed by level of each potential risk factor using the Kaplan-Meier (K-M) method. The K-M survival curves were compared between the HPV-positive and HPV-negative groups using the log rank test. The unadjusted assessments of each potential risk factor vs the rate of mortality were performed using the Cox regression model.
The relationship between HPV status and the rate of mortality was evaluated while controlling for multiple factors simultaneously using the multivariable Cox regression model. Because sample size was small, candidate predictors were selected on the basis of P < .10 in the bivariate analysis. Age was analyzed by quartile and as a continuous variable after confirming linearity. The final model was selected using the backward procedure for variable selection with liberal P < .25 as the retention criterion.
The interactions between HPV status and selected covariates including smoking, alcohol consumption, N stage, T stage, and age were tested by including the appropriate interaction terms into the multivariable model, but none of the interactions were found to be significant at P < .05. Hence, all the effects in the final model were assumed to be additive.
A total of 150 patients with the diagnosis of HNSCC during this period were identified (Table 1). All patients were male. Sixty-nine patients had HPV-positive tumors (46%), whereas 65 (43%) had HPV-negative tumors. Sixteen patients did not have HPV test results available and were subsequently excluded from the final analyses, for a total of 134 patients in our retrospective sample. The age range was 44 to 94 years (mean, 64.6 [SD, 8.0] years) at the time of diagnosis, and median (interquartile range) follow-up was 16.7 (8.7-27.3) months. Primary sites were larynx (6 of 34 [18%] HPV-positive), oral cavity (8 of 23 [35%] HPV-positive), oropharynx (43 of 57 [75%] HPV-positive), hypopharynx (9 of 16 [56%] HPV-positive), and nasopharynx (3 of 4 [75%] HPV-positive), with a statistically significant larger proportion of HPV-positive tumors found in the oropharynx (P < .001). There were 40 T1 tumors, 29 T2 tumors, 26 T3 tumors, 36 T4 tumors, and 3 patients without staging data (Table 2). Fifty-seven patients had N0 or N1 nodal disease, 73 had N2 or N3 disease, and 4 patients had no data on neck disease. One hundred three patients had no distant metastases, 7 had M1 disease, and 24 patients did not have M staging. Eighty-two patients were treated with primary combined chemoradiation therapy, 28 with primary radiation therapy, and 21 with primary surgical management (Table 3). Two of the 134 patients included in this study did not have records of treatment within the Veterans Affairs (VA) health care system, and 1 additional patient died prior to initiating therapy.
The HPV-positive group had a significantly higher proportion of white patients compared with the HPV-negative group (P = .03). The HPV-positive patients also were nonsignificantly younger than HPV-negative patients (P = .06). The HPV-positive patients were more likely to be treated with combined chemoradiation therapy and less likely to be treated primarily with radiation therapy or surgery (P < .001). Tumor location also significantly differed between the 2 groups (P < .001), with an HPV-positive predominance in the oropharynx (n = 43 [62%]). There were a total of 112 patients with reported smoking history (84%), 21 with no history of tobacco use (16%), and 1 patient with no reported data on tobacco use. There were a total of 92 patients with reported history of alcohol use (69%), and 41 patients without alcohol use history (31%). There was no statistically significant difference in alcohol and tobacco use between HPV-positive and HPV-negative patients.
The Figure displays the K-M survival curves by HPV status. Overall, there were a total of 42 deaths during follow-up, 25 in the HPV-negative group and 17 in the HPV-positive group. The 6-month, 1-year, and 2-year K-M survival estimates were 94%, 85%, and 74% in the HPV-positive group vs 88%, 77%, and 54% in the HPV-negative group. The HPV-positive group died at a nonsignificantly slower rate compared with the HPV-negative group (log rank test P = .09). The 6-month, 1-year, and 2-year K-M survival estimates in the overall sample of 150 patients were 89%, 77%, and 63%, respectively. Separate K-M survival estimates were performed comparing HPV status specifically within patients with oropharyngeal primary tumors; however, given the small sample size, no significant conclusions could be drawn from this analysis (P = .54).
Bivariate assessments of each potential risk factor vs the rate of mortality were performed. Human papillomavirus status, T stage, and N stage were found to be statistically significant risk factors in the bivariate analysis and were carried onto the final multivariate analyses. Race was also found to be a significant risk factor: African American patients had a 1.92 times greater rate of mortality compared with white patients, regardless of HPV status (hazard ratio [HR], 1.92 [95% CI, 1.10-3.42]; P = .02). Age was analyzed by quartiles based on the patient distribution: younger than 61 years at the time of diagnosis, 61 to 64 years, 65 to 68 years, and older than 68 years. Compared with age younger than 61 years at the time of diagnosis, patients aged 61 to 64 years had 1.17 times greater overall mortality (HR, 1.17 [95% CI, 0.49-2.75]; P = .73), 65 to 68 years had 1.40 times greater overall mortality (HR, 1.40 [95% CI, 0.55-3.55]; P = .48), and age older than 68 years was significantly associated with 2.82 times greater overall mortality (HR, 2.82 [95% CI, 1.23-6.45]; P = .01).
Bivariate analysis of current and former smoking and alcohol use at the time of diagnosis compared with reports of no prior smoking or alcohol history did not demonstrate a significant impact of smoking or alcohol use on mortality, regardless of HPV status. Both former smokers and current smokers had similar risk of mortality compared with nonsmokers (former vs never: unadjusted HR, 0.85 [95% CI, 0.37-1.95], P = .70; current vs never: unadjusted HR, 1.12 [95% CI, 0.53-2.37], P = .77). Both former alcohol drinkers and current alcohol drinkers had similar risk of mortality compared with nondrinkers (former vs never: unadjusted HR, 1.39 [95% CI, 0.68-2.82], P = .36; current vs never: unadjusted HR, 1.19 [95% CI, 0.61-2.32], P = .61). The relationship between HPV status and mortality according to smoking and drinking status (past, current, or never) was investigated before and after adjustment for age, N stage, and T stage under the Cox model. The cohort was classified into 5 mutually exclusive categories according to smoking and drinking status for the purpose of this analysis: (1) never smoker and never drinker, (2) past or current smoker and never drinker, (3) past or current drinker and never smoker, (4) past smoker and past drinker, and (5) current drinker and current smoker. Adjusted HRs ranged from 0.15 to 0.50 across the smoking and drinking categories, consistent with a protective effect of HPV positivity on mortality regardless of smoking and drinking. There was no statistical evidence that the relation between HPV status and mortality differed according to smoking or drinking (P = .68). Overall, there was no evidence that either smoking or drinking was associated with greater risk of mortality after adjustment for HPV status, age, T stage, and N stage.
In the multivariate analyses, the following candidate predictors were included in the multivariable analysis on the basis of the aforementioned bivariate screen: race, treatment modality, T stage, N stage, M stage, and age. Higher T stage, higher N stage, older age, and HPV-negative status were significantly associated with higher rates of mortality in the final multivariable model. Race, M stage, and treatment modality were not significant at P < .10 once the aforementioned factors were known and were therefore excluded from the final model. The final outcomes of the multivariate Cox model for HPV status, T stage, N stage, and age are presented in Table 4. Patients with positive HPV status had a lower rate of death compared with those with negative HPV status by approximately 56% (HR, 0.44 [95% CI, 0.23-0.86]; P = .02) after controlling for T stage, N stage, and age. Overall, patients with a higher T stage (T2, T3, or T4) had a significantly higher rate of mortality compared with patients with T1 tumors independent of N stage, age, and HPV status. In particular, patients with advanced primary tumor (T4) had a significantly greater rate of mortality compared with patients with T1 tumor disease (HR, 8.52 [95% CI, 2.60-18.40]; P = .001). Similarly, patients with advanced nodal metastases (N3) had a significantly greater rate of mortality compared with patients with no nodal disease (N0) while controlling for T stage, age, and HPV status (HR, 7.18 [95% CI, 2.54-20.28]; P < .001). On bivariate analysis, patients with distant metastases (M1) had a significantly greater rate of mortality compared with patients without distant metastases (M0) (HR, 5.99 [95% CI, 2.59-13.81]). For each 1-year increase in age, the rate of mortality increased by approximately 6% (HR, 1.06 [95% CI, 1.01-1.10] per year; P = .008) while controlling for N stage, T stage, and HPV status.
Whereas prior studies have demonstrated the predominance of HPV-positive HNSCC among a “new” population of younger men without a history of tobacco or alcohol use,6,16,22,26 the present study establishes HPV positivity as an important prognostic indicator even among high-risk groups with older patients and a high incidence of tobacco and/or alcohol use, such as a US veteran population. Findings in this study were consistent with prior studies regarding the behavior, location, and prognosis of HPV-positive HNSCC, suggesting that the current treatment and management modalities used to treat the “new” patients with HPV-positive HNSCC can also be applied to the “traditional” patient with HNSCC—the older man with a strong history of alcohol and tobacco exposure. Our study found that patients with HPV-positive HNSCC had better overall survival, as well as a lower rate of death, when compared with patients with HPV-negative HNSCC, and this was true across all T and N stages and age. Patients in the present study with HPV-positive HNSCC were predominantly treated with chemoradiation therapy. The improved survival of this cohort is consistent with previous studies that suggest that HPV-positive HNSCC tumors tend to be more sensitive to radiation therapy and chemotherapy compared with HPV-negative tumors.11,15
Similar to other studies,27 we also found that increased T and N stage,17,28,29 older age,30,31 and African American race32- 34 were risk factors for increased mortality, for both HPV-positive and HPV-negative cancers. As found in the present study, previous studies have suggested that African American patients fare more poorly compared with their white counterparts, presumably due to the larger proportion of white patients with HPV-positive tumors, thus lending a better prognosis to the latter group.32- 36 Whereas it is unclear what causes the racial disparity in HNSCC survival, it has been proposed that HPV status and differences in tumor stage at time of presentation may explain this disparity.
However, in contradiction to prior studies,14,19,27,37- 39 our data did not indicate that smoking or alcohol use had any significant effect on mortality for either HPV-positive or HPV-negative cancers. This is a striking and novel finding compared with multiple studies that have established tobacco and alcohol use as significant risks factors for HNSCC. Although our sample size was too small to elicit dose-response relationships between drinking or smoking and HNSCC, our findings suggest that the population of veterans is unique in the pathogenesis of HNSCC. Specifically, veterans who use the VA Health Administration services represent an even more selective patient population than that of veterans overall. Although all veterans who are honorably discharged are eligible for care through the VA system, priority is given to those with service-related disabilities, who are in certain subgroups (eg, prisoners of war), or with specific financial needs. Potentially as a result of these criteria, VA users nationally tend to represent a poorer, older population, often underemployed or unemployed, more likely African American, less educated with worse health status and more chronic medical conditions than the overall US population or veterans outside the VA system.23,25,40,41 As such, the VA population is unique, and the behavior of HPV-positive HNSCC in this veteran population may also be considered such.
Furthermore, specifically among HPV-positive cancers, our study also found no increased risk of mortality with increased smoking or alcohol use. This is consistent with prior findings by Applebaum et al,42 Hong et al,43 and D’Souza et al,2 who found no significant interaction between smoking or drinking and HPV status. This suggests that HPV-positive HNSCC represents a distinct form of HNSCC that is driven primarily by the carcinogenesis of HPV viral status, unrelated to smoking or drinking history. However, other studies have found synergistic effects of tobacco use and HPV positivity.44- 46 Until this relationship is further elucidated, smoking cessation should still be advocated, for both HNSCC risk and general health outcomes.
This study was limited by a small sample size. The findings of the present study need to be validated in a larger sample in future studies to further investigate the potential nonadditive effects among variables predicting mortality. Furthermore, due to limitations on records, not all patients had detailed follow-up to extract information related to recurrent disease. Further studies specifically including treatment failures in recurrent HNSCC would be valuable.
This study adds to a quickly growing body of literature characterizing the behavior of HPV-positive HNSCC. It is becoming increasingly clear that HPV-positive HNSCC is a distinct entity in pathogenesis, behavior, and prognosis from HPV-negative HNSCC, traditionally thought to be driven by a history of alcohol and tobacco use. Importantly, the strength of HPV status as a prognostic indicator for HNSCC regardless of patient age or social habits further emphasizes the need for standardized immunohistochemical staining for HPV and p16 for HNSCC tumors, particularly within the oropharynx, across all pathology laboratories. Detection of HPV and/or p16 status is invaluable to patient care and provides the patient with powerful information regarding his or her diagnosis and prognosis. Particularly in today’s setting in which the incidence of HPV-positive HNSCC is increasing and patients are becoming increasingly aware of HPV-positive HNSCC, HPV testing has proven to be a crucial diagnostic tool allowing for advancements in patient management.
The proportion of HPV-positive HNSCC in the general population ranges between 20% and 75%,6,7,9,12,14 and our study demonstrates a comparable incidence in a selected VA cohort. Although VA patients also have unique risk factors and demographic characteristics, the prognosis for HPV-positive HNSCC remains favorable in this population, even with the high incidence of smoking and alcohol use. Testing for HPV status of HNSCC, particularly oropharyngeal cancers, should be an important component in the diagnosis and management of these patients.
Submitted for Publication: January 22, 2015; final revision received June 15, 2015; accepted June 23, 2015.
Corresponding Author: Marilene B. Wang, MD, Department of Head and Neck Surgery, UCLA David Geffen School of Medicine, 200 UCLA Medical Plaza Ste 550, Los Angeles, CA 90095 (firstname.lastname@example.org).
Published Online: August 13, 2015. doi:10.1001/jamaoto.2015.1447.
Author Contributions: Drs Shay and Wang had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Shay, Lewis, Wang.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Shay, Lewis, Wang.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Shay.
Administrative, technical, or material support: Chang, Lewis, Wang.
Study supervision: Wang.
Conflict of Interest Disclosures: None reported.
Previous Presentation: This study was presented at the Fifth World Congress of the International Federation of Head and Neck Oncology Societies/American Head and Neck Society Annual Meeting; July 26-30, 2014; New York, New York.
Additional Contributions: Daniela Markovic, MS, University of California Department of Biomathematics, provided ancillary statistical analyses on the data set. She was compensated for her contributions.