Department of Otolaryngology-Head and Neck Surgery, Johns Hopkins Medical Institutions, Baltimore, Maryland
Department of Oncology, Johns Hopkins Medical Institutions, Baltimore, Maryland
Corresponding author: Sara I. Pai, MD, PhD, Departments of Otolaryngology-Head and Neck Surgery and Oncology, Johns Hopkins Medical Institutions, 601 N. Caroline Street, JHOC Sixth Floor, Baltimore, MD 21287; Fax: (410) 955-0035; email@example.com
A subset of patients with human papillomavirus (HPV)-associated oropharyngeal squamous cell carcinoma (HPV-OSCC) experience poor clinical outcomes. The authors of this report explored prognostic risk factors for overall survival (OS) and recurrence-free survival (RFS).
Patients with incident HPV-OSCC who received treatment at the Johns Hopkins Hospital between 1997 and 2008 and who had tissue available for HPV testing as well as demographic and clinicopathologic information (N = 176) were included. Tissue was tested for HPV by in situ hybridization (ISH) and/or p16 immunohistochemistry. Demographic and clinicopathologic information was extracted from medical records.
In total, 157 of 176 patients (90%) with OSCC had HPV-associated disease (HPV-OSCC). In the patients with HPV-OSCC, the 3-year and 5-year OS rates were 93% (95% confidence interval [CI], 88%-98%) and 89% (95% CI, 81%-97%), respectively. Shorter survival was observed among older patients (hazard ratio [HR], 2.33 per 10-year increase; 95% CI, 1.05-5.16 per 10-year increase; P = .038), patients with advanced clinical T classification (HR, 5.78; 95% CI, 1.60-20.8; P = .007), and patients who were currently using tobacco (HR, 4.38; 95% CI, 1.07-18.0; P = .04). Disease recurrence was associated with advanced clinical T-classification (HR, 8.32; 95% CI, 3.06-23; P < .0001), current/former alcohol use (HR, 13; 95% CI, 1.33-120; P = .03), and unmarried status (HR, 3.28; 95% CI, 1.20-9.00; P = .02). Patients who remained recurrence free for 5 years had an 8.6% chance of recurrence by 10 years (1-sided 95% CI upper bound, 19%; P = .088).
Despite decreases in per capita smoking prevalence and alcohol consumption in the United States over the past several decades, the incidence of oropharyngeal squamous cell carcinomas (OSCC) has steadily increased.[1-3] The increasing incidence in OSCC is attributed to a growing number of cancers caused by the human papillomavirus (HPV).[3, 4] Indeed, recent research has indicated that HPV-associated OSCC (HPV-OSCC) is a distinct etiologic entity, and this subset of head and neck cancers is associated with an improved prognosis compared with HPV-negative head and neck cancers.[5-13] Although the improved survival rates of patients with HPV-OSCC have been well established, clinically, there is a patient subset that still experiences poor outcomes.
A recent survival analysis demonstrated important prognostic differences between HPV-positive and HPV-negative OSCC. However, prognostic factors for patients with HPV-OSCC are less clear but of significant interest given the consideration of de-escalation treatment strategies for this patient population. Therefore, we analyzed predictors of survival in a series of patients with HPV-OSCC who were treated at a single academic institution.
MATERIALS AND METHODS
Patients who were treated for OSCC (International Classification of Diseases for Oncology codes C01.9, C02.4, and C09.0-C10.9) at the Johns Hopkins Hospital between 1997 and 2008 were screened for eligibility. Patients were included if they had a histologically confirmed diagnosis of incident OSCC and had tissue available for HPV testing. A modest number of individuals (n = 46) who did not have tumor tissue available for testing but had previously recorded HPV and/or p16 results also were included.
The medical records of all eligible patients were retrospectively reviewed to collect patient demographic information (age, sex, race [as specified by patient during hospital registration], marital status) and clinicopathologic information (site of tumor, date of cancer diagnosis, tumor stage, tobacco and/or alcohol use, primary treatment modality, treatment duration, recurrence, HPV status, and date of last follow-up or death). Follow-up was defined as the time elapsed between the date when treatment was completed and either the last recorded date of clinic follow-up or the date of death. If the demographic and/or clinicopathologic information could not be extracted from the medical records, then patients were excluded from the study. Survival data were confirmed with the National Death Index database. Institutional Review Board approval was obtained to perform this study, and an exemption was obtained to get informed consent from patients retrospectively, because patients were deidentified after demographic extraction from the medical records.
Determination of HPV status
Tumor HPV status was determined using either HPV-16 in situ hybridization (ISH), p16 immunohistochemistry (IHC), or both. All slides were reviewed by a head and neck pathologist (W.H.W.), and HPV-16 ISH and p16 IHC status was assessed. HPV tumor status was reported as positive if either ISH or IHC tests were positive. For cases in which there was a discrepancy between HPV-16 ISH and p16 IHC results, a positive result by either test was considered HPV-associated.
HPV-16 in situ hybridization
Briefly, 5-μm-thick, paraffin-embedded tumor sections were deparaffinized. Heat-induced target retrieval was performed in citrate buffer. The samples then were digested with Proteinase K (20 μg/mL; Roche Diagnostics, Indianapolis, Ind) and hybridized with a biotinylated HPV-16 DNA-specific probe (GenPoint; Dako, Carpinteria, Calif). Streptavidin-horseradish peroxidase complex, biotinyl tyramide, and streptavidin-horseradish peroxidase complex were consecutively applied for signal amplification. Samples were incubated with chromogenic substrate diaminobenzidine. Punctate hybridization signals that localized to the tumor cell nuclei defined an HPV-positive tumor.
Briefly, 5-μm-thick, paraffin-embedded tumor sections were deparaffinized. Target retrieval was performed using heat-induced epitope retrieval with citrate buffer. Tissue sections were subsequently incubated with p16 monoclonal mouse antibody (1:500 dilution; MTM Laboratories, Heidelberg, Germany). Visualization of p16 antibody binding was achieved using the avidin-biotin-peroxidase technique (Dako LSAB Kit; Dako Cytomation, Carpinteria, Calif). Expression of p16 was scored as positive if strong and diffuse nuclear and cytoplasmic staining was present in ≥70% of the tumor.
Descriptive statistics were used to characterize the study cohort. Differences in patient demographic and clinicopathologic characteristics between men and women who were diagnosed with HPV-OSCC were compared using the Wilcoxon rank-sum test for continuous variables and the Fisher exact test for categorical variables. Overall survival (OS) was computed from the date treatment ended to the date of death from any cause and was censored at the last follow-up at which the patient was known to be alive. Recurrence-free survival (RFS) was calculated from the date treatment ended to the date of disease recurrence and was censored at the earliest of the date of last follow-up or the date of death before recurrence.
The associations of demographic and clinical factors with the risk of death or disease recurrence in the HPV-positive cohort were evaluated first using univariate Cox proportional hazards models and then using multivariable models. Unadjusted (univariate analysis) and adjusted (multivariate analysis) hazard ratios (HRs) and corresponding 95% confidence intervals (CIs) were estimated. In the multivariable analysis, race (white, black, or other), sex, age (continuous), marital status (married or unmarried [single, divorced, or widowed]), alcohol use (never or former/current use), tobacco use (never, former, or current use), primary therapy (chemoradiation, radiation, and/or surgery), and clinical T classification (Tx/T1/T2 or T3/T4) and N classification (Nx/N0/N1 or N2/N3) were considered in the initial models. All variables with P values < .15 were retained in the final model using a backward elimination procedure. Because race and sex were of interest, these 2 variables were retained in the model regardless of statistical significance.
OS and RFS were summarized using nonparametric Kaplan-Meier survival curves as well as directly adjusted survival curves, which averaged the individual predicted survival curves based on unstratified Cox models. The method of direct adjustment was used to control for possible confounders caused by an imbalance of patient characteristics between groups. The analyses described above were performed separately for OS and RFS. The proportional hazards assumption was verified by creating a time-dependent covariate for each predictor and testing the time-dependent covariates all at once for significance in the Cox model, and no violations were noted. The 5-year conditional RFS of patients who were free of disease 5 years after treatment was estimated as the 10-year RFS divided by the 5-year RFS, and the corresponding standard error was estimated with a variation of the Greenwood formula. All statistical tests were 2-sided, unless otherwise indicated, and were considered statistically significant at P < .05. The analyses were carried out using SAS version 9.2 (SAS Institute Inc., Cary, NC) and R statistical software version 2.12.0 (R Foundation for Statistical Computing, Vienna, Austria).
Four hundred forty-five patients with incident OSCC were treated at the Johns Hopkins Hospital between 1997 and 2008. Of these patients, 176 (40%) had tissue available for HPV testing and complete demographic and clinicopathologic information and, thus, were included in this analysis.
We tested 176 OSCC samples for HPV status, including 130 tumors that were tested by both HPV-16 ISH and p16 IHC, 43 tumors that were tested by HPV-16 ISH only, and 3 tumors that were tested by p16 IHC only. Of the 130 samples that were tested by both HPV-16 ISH and p16 IHC, 11 had discordant results (10 were HPV-16 ISH-negative/p16-positive, and 1 was HPV-16 ISH-positive/p16 negative). A positive result by either ISH or IHC was considered HPV-associated (90%; 157 of 176 samples). HPV positivity was 90% (120 of 133 samples) by p16 IHC testing and lowered to 83% (144 of 173 samples) with HPV-16 ISH testing. HPV-16 ISH is considered a highly specific test but does not capture disease caused by other oncogenic HPV types, whereas p16 IHC is a highly sensitive method, and this difference accounts for the higher HPV-positive rate using the latter testing method.
Demographic and Clinicopathologic Characteristics
Patients with HPV-OSCC were primarily men (87%; 136 of 157 patients) and white (78%; 122 of 157 patients), and the median patient age was 54 years (range, 28.7-80.3 years). HPV-positive patients were more likely to be married (79.6% vs 20.4%; P = .042), never tobacco users (40.0% vs 15.2%; P = .017), and diagnosed with advanced N classification (77.1% vs 22.9%; P = .007) compared with HPV-negative patients. We had a mixed population of patients who received treatment with various modalities: 51% (80 of 157 patients) of HPV-OSCC patients underwent open surgery, and 49% (77 of 157 patients) received nonsurgical therapy. Of the 80 surgical patients, 79 patients (98.8%) also received adjuvant therapy, including 40 patients (50%) who received adjuvant concurrent chemoradiation therapy and 39 patients (48.8%) who received adjuvant radiation therapy alone. Demographic and clinicopathologic characteristics of the HPV-positive and HPV-negative patients are presented in Table 1.
Table 1. Patient Demographic and Clinicopathologic Characteristics by Human Papillomavirus Status
Abbreviations: HPV, human papilloma virus; SD, standard deviation.
A nonparametric Wilcoxon rank-sum test was used for continuous variables, and the Fisher exact test was used for categorical variables.
Age at end of treatment, y
Mean ± SD
54.5 ± 9.0
57.7 ± 10.7
54.1 ± 8.7
Clinical T stage
Clinical N stage
Clinical M stage
Although the patient population was primarily white and male, there were 11 black patients and 4 patients of another race. In addition, there were 21 women (13%) diagnosed with HPV-OSCC. Demographic and clinicopathologic characteristics of the HPV-positive men and women are presented in Table 2. Analysis of RFS and OS in these HPV-positive patients was performed to determine risk factors that contributed to a poor prognosis.
Table 2. Demographic and Clinicopathologic Characteristics Among Human Papillomavirus-Positive Patients
A nonparametric Wilcoxon rank-sum test was used for continuous age, and the Fisher exact test was used for other categorical variables.
Age at end of treatment, y
Mean ± SD
54.1 ± 8.7
54.4 ± 10.7
54.1 ± 8.4
Clinical T stage
Clinical N stage
Clinical M stage
Overall Survival Analysis
The 3-year and 5-year OS rates were 93% (95% CI, 88%-98%) and 89% (95% CI, 81%-97%), respectively (Fig. 1A). The median follow-up was 42 months (range, 1.8-167.0 months). It is noteworthy that the risk of disease-specific death stabilized between years 2 through 5 and then increased between years 5 through 10. In univariate analysis, OS for HPV-positive patients was reduced among older patients (HR, 2.56 per 10-year increase in age; 95% CI, 1.41-4.67 per 10-year increase in age; P = .002), patients with advanced clinical T-classification (HR, 4.77; 95% CI, 1.72-13.2; P = .003), and unmarried patients (HR, 3.19, CI, 1.28-7.95; P = .01) (Table 3).
Table 3. Association of Risk Factors With Overall Survival and Recurrence-Free Survival Among Human Papillomavirus-Positive Patients
In multivariate analysis of the HPV-positive patients, OS was worse among older patients (HR, 2.33 per 10-year increase; 95% CI, 1.05-5.16; P = .038), patients with advanced clinical T-classification (HR, 5.78; 95% CI, 1.60-20.8; P = .007), patients with less advanced clinical N stage (HR, 4.94; 95% CI, 1.53-16.0; P = .007), and current tobacco users compared with never tobacco users (HR, 4.38; 95% CI, 1.07-18.0; P = .04) (Table 3). Alcohol use and primary therapy modality were not associated with differences in OS. Sex and race did not have statistically significant differences in survival; however, for women and nonwhite patients, all HRs were >2.0 (Table 3).
The impact of sex on OS was further evaluated with Kaplan-Meier curves. There was a trend for women (N = 21) to have a greater risk of death compared with men (N = 136) in both adjusted analysis (HR, 2.11; 95% CI, 0.55-8.0; P = .274) (Fig. 1B) and unadjusted analysis (HR, 1.98; 95% CI, 0.71-5.5; P = .19); however, this difference did not reach statistical significance.
The effect of race on OS in HPV-positive patients also was explored. A slightly higher but not statistically different risk of death for patients with HPV-OSCC was observed in adjusted analysis (HR, 2.14; 95% CI, 0.23-20; P = .51) among blacks (N = 11) compared with whites (N = 142). The small number of deaths among blacks (N = 1) compared with whites (N = 12) during a similar follow-up period (median follow-up: black patients, 48 months; white patients, 42 months) limited the power of that analysis. A significant increased risk of death was observed among patients of other race (HR, 16.5; 95% CI, 2.15-126; P = .007), although this was based on only 4 patients (Table 3). A sensitivity analysis was performed to determine whether this small group affected our results, and the results remained unchanged with the exclusion of these 4 patients of other race.
Recurrence-Free Survival Analysis
The 3-year and 5-year RFS rates were 89% (95% CI, 82%-95%) and 86% (95% CI, 76%-96%), respectively (Fig. 2A). To examine whether patients who had been free of disease 5 years after completion of treatment were still at risk for recurrence, we estimated their 5-year conditional RFS. We observed that, having remained recurrence-free for 5 years (N = 40), patients had a cumulative risk of recurrence of 8.6% by 10 years (1-sided 95% CI upper bound is 19%; P = .088), suggesting an appreciable level of risk of recurrence for the subsequent 5 years. We confirmed that all recurrences were confirmed related to their primary head and neck cancer through histopathologic evaluation and HPV-16 ISH testing of the diagnostic biopsies at the recurrent sites, all of which had integrated HPV-16 DNA. Recurrences occurred in the following locoregional sites: oropharynx, parotid, neck, hard palate, tongue, tonsil, and retromolar trigone. Distant site involvement included the lung (n = 9), brain (n = 1), and temporal bone (n = 1).
In univariate analysis, variables associated with RFS included advanced clinical T-classification (HR, 7.30; 95% CI, 2.95-18.0; P < .0001) and unmarried status (HR, 2.63; 95% CI, 1.05-6.62; P = .04) (Table 3). In multivariate analysis, advanced clinical T-classification (HR, 8.32; 95% CI, 3.06-23; P < .0001), unmarried status (HR, 3.28; 95% CI, 1.20-9.00; P = .02), and current/former alcohol use (HR, 13; 95% CI, 1.33-120; P = .03) compared with never alcohol use retained a significant association with increased risk of disease recurrence (Table 3). In contrast to risk factors for OS, older age, lower N stage, and current tobacco use were not associated with RFS.
No difference in RFS was observed in men compared with women in adjusted analysis (HR, 0.87; 95% CI, 0.23-3.31; P = .84) (Fig. 2B) or unadjusted analysis (HR, 1.51; 95% CI, 0.50-4.54; P = .46). When evaluating race, the risk of recurrence was higher in blacks (HR, 4.77; 95% CI, 0.95-24; P = .06) compared with whites in an analysis adjusted for sex, marital status, alcohol use, and clinical T-classification, although the difference did not reach statistical significance. A significant increased risk of recurrence also was observed among patients of other race (HR, 47; 95% CI, 4.20-528; P = .002); however, once again, this was based on a limited number of patients (Table 3). Similar to the OS analysis, a sensitivity analysis excluding the 4 patients of other race was performed to determine whether this small group affected our results, and the results remained unchanged.
Although patients who have HPV-OSCC have an improved prognosis compared with patients who have HPV-negative OSCC, there is a subset of patients with HPV-OSCC who experience a poor clinical outcome. Identification of the risk factors for a poor outcome among patients with HPV-OSCC is of particular interest, because it would facilitate the stratification of those patients who may benefit from more aggressive intervention or, conversely, those who may benefit from de-escalation treatment approaches.
To identify risk factors for poor outcome in patients with HPV-OSCC, we retrospectively analyzed OS and RFS rates in 157 patients with HPV-OSCC who received treatment at the Johns Hopkins Hospital. To date, this is one of the largest studies evaluating predictors of survival among HPV-positive patients, some of whom have been followed for more than 10 years after definitive treatment. We demonstrated a 5-year OS rate of 89%, which was reduced in older patients, those with advanced clinical T-classification, and current tobacco users, and an RFS of rate 86%, which was reduced in patients with advanced clinical T-classification, current/former alcohol users, and unmarried patients.
Maxwell et al reported that tobacco use in patients with HPV-OSCC was associated with a higher risk of disease recurrence compared with never-users of tobacco. Our study provides further evidence that current tobacco use increases the risk of poor OS for patients with HPV-OSCC. It is unclear whether there is a true interaction between tobacco use and HPV. It has been hypothesized that the direct effect of tobacco on the pathogenesis of cancer is secondary to the carcinogenic properties of tobacco smoke, which can increase the risk of developing genetic alterations. Tobacco is also considered immunosuppressive; and, in the setting of a virus-related cancer, tobacco may inhibit the body's natural ability to mount effective immunologic responses to eradicate the virus and any virus-related cancer cells.
Advanced clinical T-classification has been reported as a significant risk factor for recurrence and death in patients with OSCC. Studies among patients with OSCC who received concurrent chemoradiation or underwent transoral laser microsurgery demonstrated poorer survival for those who had T3/T4 tumors versus those who had T1/T2 tumors, similar to what we observed in the current study. We also observed that lower clinical N stage was a poor prognostic factor, which may be explained in part by the clinical characteristics of the HPV-OSCC patient population. Patients with HPV-OSCC often have cystic cervical lymph nodes, which can result in a higher clinical N stage. In our study, the HPV-positive patients were diagnosed with a more advanced N stage compared with HPV-negative patients (77.1% vs 22.9%; P = .007); and, among the patients with a higher N stage, 89% had a lower T-classification (Tx, T1-T2). Given the association of higher N stage with lower T-classification in our study group, the prognostic effect of T-classification and N stage may have been confounded. Additional studies will be required to evaluate the prognostic effect of N stage in patients with HPV-OSCC.
We had a mixed population of patients treated with various modalities. Fifty-one percent of patients underwent open surgeries, and 49% received treatment with nonsurgical approaches. Postoperative adjuvant concurrent chemoradiation therapy is often administered based on aggressive pathologic features, including extracapsular extension (ECE). Although ECE was not evaluated specifically in our study, of the 80 surgical patients who were included, 50% received adjuvant concurrent chemoradiation therapy, and 48.8% received adjuvant radiation therapy alone. On the basis of previously published studies, ECE in head and neck squamous cell cancer usually predicts worse outcomes. However, in the era of HPV-OSCC, this is an area of controversy. A recent study evaluated 101 cases of OSCC in which 90% of the cancers were p16-positive. ECE, defined as extension into soft tissue, was not significantly associated with poorer OS (P = .14), disease-free survival (P = .2), or disease-specific survival (P = .09) in multivariate analysis. Those authors concluded that the impact of ECE in lymph node metastases was limited in OSCC. Thus, ECE may not be an applicable prognostic factor for HPV-OSCC. These results are intriguing, and ECE, as well as other traditionally aggressive pathologic features, in HPV-OSCC needs to be evaluated in future studies.
Head and neck squamous cell cancer is traditionally associated with recurrence rates of 50%, and approximately 80% to 90% of tumors recur within the first 2 to 4 years after definitive therapy.[24-31] These rates of recurrence have changed in the era of HPV-OSCC. In the HPV-OSCC patient population, it has been documented that recurrences occur at approximately 13.6% of local sites and 8.7% of distant sites within 3 years after the completion of definitive treatment. One study reported on the risk factors for distant metastasis in HPV-OSCC and observed that patients with N2c disease had reduced distant control if they received radiotherapy alone, suggesting that these patients may not be candidates for deintensification strategies that do not include chemotherapy. In our study, we reported a cumulative incidence of recurrence in HPV-OSCC of 14% after 5 years; and, in those patients who were recurrence-free for 5 years, the cumulative risk of recurrence at 10 years was 8.6%. This finding may provide us with further observational insight into the pathophysiology and natural course of this disease process. The indolent growth of these tumors may explain the increased number of recurrences 10 years post-treatment compared with HPV-negative head and neck squamous cell cancer. Nonetheless, these observations need to be further confirmed in additional studies to determine whether extended cancer surveillance is warranted in this cancer population.
Prior studies suggested that marital status had prognostic significance in patients with head and neck cancer, with partnered patients experiencing lower rates of recurrence and better OS compared with unpartnered patients.[33-35] We observed a similar trend toward poorer OS and significant recurrence rates in unmarried patients versus married patients diagnosed with HPV-OSCC. These findings highlight the finding that social support continues to be an important consideration in caring for head and neck cancers patients, independent of HPV status.
Differences in survival according to sex have been difficult to study, because men have a higher incidence of HPV-OSCC than women.[4, 13] Our study is one of the largest series to date comparing the survival of women and men with HPV-OSCC, and the survival of women did not differ statistically from that of men, suggesting that HPV-OSCC is a distinct clinical and pathologic entity driven predominantly by tumor biology rather than differences in sex.
Our institution is a large referral base that treats patients from diverse geographic and demographic backgrounds. Having all patients treated at a single institution has the inherent advantage that all patients are treated and clinically monitored for recurrence in a uniform manner. An additional strength of the study is that all of these tumors were tested for HPV status. Overall, our findings suggest that demographic and clinical features like advanced patient age, tobacco and alcohol use, advanced clinical T-classification, and unmarried status, which have been reported as poor prognostic factors for patients with head and neck cancer, continue to be applicable prognostic markers in patients with HPV-associated OSCC, independent of treatment modality. It is noteworthy that we observed a trend toward an increased risk of recurrence after the conventional 5-year cancer surveillance period subsequent to definitive therapy, and this trend needs to be investigated further to determine whether extended cancer surveillance is warranted in this cancer population.
This work was supported by the Head and Neck Specialized Center of Research (SPORE) Program, National Institute of Dental and Craniofacial Research, National Institutes of Health (P50 DE019032 [to W.H.W. and S.I.P.]).