Risk factors and survival by HPV-16 E6 and E7 antibody status in human papillomavirus positive head and neck cancer



High-risk human papillomavirus types (HPV-HR) are associated with head and neck cancer (HNC) risk and better survival. Most patients with HPV-HR DNA-positive tumors develop anti-HPV E6/E7 antibodies; however, it is unclear whether those who mount an immune response have similar risk factors or clinical outcomes as those who do not. HPV-16 DNA tumor-positive HNC cases were evaluated for HPV-16 E6 and E7 antibodies using a GST capture ELISA system. Among 57 HPV-16 DNA tumor-positive HNC cases, 67% were detected with HPV-16 E6 and/or E7 antibodies. Male gender (76% vs. 42%, p = 0.02), younger age (63% vs. 16%, p = 0.001) but not tobacco or alcohol were associated with E6 and/or E7 seropositivity. Seropositivity was associated more often with late stage (76%), poor grade (65%), positive nodes (82%). and in the oropharynx (82%), Median disease-specific and recurrence-free survival were longer in E6 and/or E7 seropositive compared to E6/E7-negative cases (2.2 years vs. 1.4 years, both outcomes), although results were not statistically significant. When examined jointly with p16 expression, E6 and/or E7-positive/p16-positive cases had better disease-specific (2.1 years vs. 1.1 years, p = 0.06) and recurrence-free (2.3 years vs. 1.1 years, p = 0.03) survival compared to E6-/E7-/p16- cases. These findings suggest there are 2 distinct HNC patient groups with HPV DNA-positive tumors, distinguishable by E6 and/or E7 antibody status. Differences in antibody status are associated with distinct risk factors and clinical outcomes. This information can be available as a simple blood test at initial presentation, before the removal of tissue through biopsy or surgery.

Large population studies that have examined the role of HPV in head and neck carcinoma (HNC) have found that whereas HPV DNA-negative tumor cases have a significantly greater risk of cancer associated with tobacco and alcohol, HPV DNA-positive cases do not.1–3 These investigations also have found that risk factors that are consistent with an HPV infection, number of sex partners, age at first intercourse and history of sexually transmitted infections are associated with an increased risk of an HPV-related head and neck tumor.3

Similar HPV-related risk factors have been shown for HPV-16 L1 antibody positive HNC patients based on antibodies to the L1 proteins and a few other high risk HPV types.3, 4 Because the HPV L1 antibodies are considered a lifetime measure of exposure to HPV infection from both genital and oral sites; however, the assay is not a suitable screening measure for identifying current HPV-related tumors. Antibody response has been shown to the oncogenic viral proteins E6 and/or E7 of HPV-16, the predominant HPV-HR type in HNC, and has been shown to provide a better, more sensitive marker for HPV-associated cervical and head and neck cancers.5–8 We have demonstrated that HPV-16 E6/E7 antibodies are rarely detected in those with an HPV infection alone but primarily in patients with HPV DNA-positive tumors of the head and neck. The E6/E7 antibody response had a much higher correlation with tumor HPV DNA presence and with survival than did the L1 antibodies. Anti-E6/E7 serology status thus might be useful in determining targeted treatment at the time of clinical diagnosis, before tissue removal.9 The HPV type-specific concordance between the E6 and/or E7 antibodies and HPV DNA tumor type has been high,6, 9 ranging between 65% and 100%5, 8, 9 either before or soon after treatment.

It is unknown why some patients who have an HPV DNA-positive tumor but are E6/E7 antibody negative do not elicit an immune response and whether this group of cases has risk factors or prognosis that are different from the HPV DNA tumor-positive/E6/E7-positive HNC cases. We hypothesize that these tumor positive/seronegative HPV cases have risk factors, pathology characteristics and clinical prognosis more like patients who are HPV DNA tumor-negative. In this investigation, we examined sera limited to HPV-16 DNA tumor-positive HNC patients to determine the E6 and E7 seropositive prevalence and whether risk factors or survival were different by serology status.

Material and Methods


All primary cases ages 18 and older with cancer of the oral cavity, oropharynx or larynx seen at the University of Iowa Hospitals, Department of Otolaryngology, between 2000 and 2004 were eligible for enrollment. Prior to enrollment, all patients were administered a university approved IRB form for informed consent. In this study we examined a subset of these HNC patients included in a larger study whose tumors were detected with HPV DNA and who had an assessment of antibodies to HPV-16 E6 and E7 oncoproteins and had follow-up survival available at the time of this analysis (N = 57). Four cases were excluded because they had no HPV DNA results although they did have E6 and E7 serology. Three of these had no surgery and thus no tissue collected for evaluation of tumor HPV DNA, and the fourth case had a tumor that was too small and no tissue was available to evaluate HPV DNA. Patients were administered an approved Human Subjects Review form prior to interview. Included were all oral cavity, oropharygeal and laryngeal tumors sites as defined by the American Joint Commission on Cancer.10

Data collection

Patients completed a risk factor questionnaire regarding HNC risk factors (demographics, tobacco and alcohol use, sexual behaviors and history of HPV-related lesions). A blood sample was taken at the initial visit prior to treatment. Tumor tissue was available from tumor blocks. Research staff reviewed the medical and pathology records and the Iowa NCI SEER Cancer Registry for prior cancer history, treatments, follow-up survival and recurrence status.8 A pathology form reported cancer site, histology, stage, grade and lymph node status.

Laboratory methods

Procedures performed on clinical samples have been described previously.5, 6, 8, 9 DNA was extracted from deparaffinized tumor tissue sections with a QIAGEN DNA Tissue Kit according to the manufacturer's instructions (QIAGEN, Valencia, CA). Laser-capture microdissection procedures11 were performed when tissue specimens had less than 10% of tumor tissue to verify that HPV-16 DNA was in the tumor cells and not in adjacent nontumorous epithelium. To reduce the risk of tumor contamination, control tissue from benign oral lesions inserted between each tumor block and assessed for HPV DNA. All controls were HPV DNA-negative. Sample preparation, PCR analyses, DNA hybridization, and HPV typing procedures for assessment of tumor tissue were based on a standard protocol.8, 12 All tumor tissue tested positive for the β-globin gene which is included to verify adequate DNA and adequacy of the PCR amplifications of the cellular DNA.13 DNA was PCR-amplified with MY09 and MY11 primers to detect HPV.14 An aliquot of the PCR product of each HPV-negative specimen was hybridized by the dot blot method with 32P-labeled probes for a more sensitive detection of HPV DNA. Positive samples underwent hemi-nested PCR-amplification with MY09 and GP5+ primers.15 DNA sequencing (Applied Biosystems-PE, Foster City, CA) was used to identify HPV types in comparison with GenBank sequences using the BLAST program.16

Presence of antibodies to HPV-16 oncoproteins E6 and E7 was assessed using the GST capture ELISA system.17 GST-E6 and GST-E7 fusion proteins were used as antigens and the background reactivity of serum samples was determined using GST only as antigen. Absorbances of background wells were subtracted from corresponding values obtained in the presence of viral antigen. The cut-off level above which the adjusted values were considered positive was every day calculated for each antigen separately as a mean absorbance plus 3 standard deviations after eliminating outliers. In all patients, optical densities were recalculated into ratios by division of the particular cutoff value. An OD ratio above 1 was defined as positive. All serum samples with an OD ratio between 1.0 and 1.1 were retested; those positive in all tests were counted positive. About 20% (n = 11) of randomly selected specimens were retested as well to confirm the results. The mean coefficient of interassay variability was 12.4%/11.9% for GST-E6 (retested samples/retested samples with OD index 1.0-1.1) and 12.8/16.1 for GST-E7 (same distribution).


Risk factors examined were age, gender, alcohol, tobacco, tumor site, grade, stage and treatment. Cochran-Mantel-Haenszel (CMH) methods were used to examine the association between the dichotomous dependent variable and risk factors for HNC (age, gender, tobacco and alcohol exposure). Odds ratios and 95% confidence intervals were calculated using Mantel-Haenszel or logit estimation procedures to measure the magnitude of the associations. Survival was measured in years from the date of diagnosis until death or until the patient was last known to be alive. In disease-specific analyses, patients who died from a cause other than HNC were censored and patients with an indeterminate cause of death were excluded. Recurrence was measured in years from the date of diagnosis until the date of recurrence or until the patient was last known to be disease free. Patients who were never free of disease were excluded and patients dying before disease recurrence were treated as censored observations. Date of death or date last known to be alive was obtained from the NCI Iowa SEER Cancer Registry, University Hospital Tumor Registry, and National Death Index which together identify >98% of follow-up. Kaplan-Meier (KM) plots were generated using product-limit estimation methods. Because of the small number of deaths [disease-specific mortality included only 8 patients, 5 positive (17%) and 3 negative (21%) patients), the software procedure did not calculate medians and confidence intervals (CI)]. Medians were calculated as the 50th percentile for years of follow-up. CIs for medians were generated using the methods of Hahn and Meeker.18 Median follow-up comparisons were analyzed using the median Wilcoxon test. Also, because there were so few deaths, Cox proportional hazard models assessing survival and recurrence with covariates/confounders failed to generate results. It was not possible to control for stage because all early stage cases survived the disease and too few had a recurrence in early stages. For tumor grade results, only 2 cases with poorly differentiated tumors died and only 2 had a recurrence. Thus 0 cell sizes in analyses including these variables prevented calculation of p-values and hazard ratios. Statistical significance was based on 2-tailed tests and p-values ≤0.05, but comparisons with p-values ≤0.10 were considered of interest for future hypothesis testing. Statistical analyses were performed using SAS 9.1.


Among HNC cases detected with HPV-16 DNA in their tumor tissue, 57 had data on HPV-16 E6 and E7 antibody status and follow-up survival available at the time of analysis. Because the E6 and the E7 antibodies had similar frequencies for each risk factor and pathological characteristic, and had a high concordance rate (79% had both antibodies), they were combined and defined as anti-HPV-16 antibody positive if the individual was E6 and/or E7 positive (E6+ and/or E7+). These cases were compared to those who were negative to both E6 and E7 (E6-/E7-). HPV-16 DNA tumor-positive cases were E6 and/or E7 seropositive in 67%. After adjustment for age, alcohol, and tobacco, E6 and/or E7 seropositive cases were more likely to be male (76% vs. 42%) and younger age (63% vs. 16%) but there were no differences in serology status by alcohol or tobacco use (Table 1). There was a significantly higher prevalence of E6 and/or E7 seropositivity among cases with greater number of lifetime sexual partners but not with practicing oral–genital sex. Seropositivity was more frequent in the orophayrnx than in the oral cavity and rare in laryngeal tumors. E6 and/or E7 seropositive cases also were more likely to be diagnosed with late stage disease, poor grade and nodal involvement. P16 overexpression (p16+), thought to be a marker of HPV-positive tumor status, was significantly more likely to be detected in E6 and/or E7 seropositive compared to seronegative cases (92% vs. 32%). Only 5% of all cases were p16- and HPV DNA+/E6/E7+.

Table 1. Characteristics of HPV-16 DNA + HNC cases by HPV-16 E6 and E7 status
inline image

Because the pathology profiles of the HPV DNA tumor-positive patients have been associated with better prognosis, we examined survival and recurrence associated with HNC risk factors, and p16 status. Disease-specific survival was available for 44 cases and 38 for recurrence-free survival. Six patients were never free of disease. Median disease-specific and recurrence-free survival were shorter in the E6/E7 seronegative (1.4 years, both outcomes, 95% CI: 0.9/0.7–2.8/2.8) compared with the E6 and/or E7 seropositive group (2.2 years, both outcomes, 95% CI: 1.4/1.2–2.9/3.0) although these results did not reach statistical significance (p > 0.1). Survival analysis was examined adjusting for age, tobacco and alcohol. Number of sex partners and nodal involvement were not adjusted because some cells were too small and prohibited generation of results. Total years of survival also showed no significant differences in disease specific or recurrence-free survival by serology status although the percentage was higher in the E6 and/or E7 seropositive group (88% vs. 75%, Table 2, Figs. 1 and 2). Better disease-specific (p = 0.04) and recurrence-free (p = 0.02) survival were found among those who had p16+ tumors.

Figure 1.

Kaplan-Meir disease-specific survival by E6/E7 status.

Figure 2.

Kaplan-Meir recurrence-free survival by E6/E7 status.

Table 2. Disease-specific and recurrence-free survival by HPV serology and p16 status
inline image

We next evaluated whether prediction of survival was different when HPV seropositivity status was evaluated with p16 expression, adjusting for age, alcohol and tobacco. Research in HNC has suggested that p16 overexpression may be a marker for HPV DNA positive tumors, which our current data support and a better predictor of prognosis. Compared with the group with the expected best survival, E6+ and/or E7+/p16+, the E6-/E7-/p16- group had less than half the disease-specific or recurrence-free median survival (Table 3). Consistent with these findings, total years of disease-specific and recurrence-free survival were near or statistically significantly lower in the E6-/E7-/p16- group compared to the E6+ and/or E7+/p16+ group (Table 4). There were too few numbers in the discordant groups to assess these outcomes; however, survival in these 2 groups, which included either E6+ and/or E7+ or p16+ status, also was better than in the E6-/E7-/p16- group.

Table 3. Median1 disease-specific and recurrence-free follow-up by HPV serology and p16 status
inline image
Table 4. Disease-specific and recurrence-free survival by joint HPV serology and p16 status
inline image

Treatment differences were examined in association with serologic and p16 status. Measures of HPV and p16 were not performed until some time after all treatment was completed; thus clinicians were not providing treatment based on known biomarker status. Patients who were E6+ and/or E7+ were more likely to have both surgery and radiation than surgery only (OR=4.6, p = 0.04). Those who were E6+ and/or E7+/p16+ had an even greater probability of having these 2 treatments (OR = 26.0, p = 0.002) compared with the E6-/E7-/p16- group. The E6+/E7+ and E6+/E7+/p16+ groups were significantly more likely to be detected with Stage IV disease than were the E6-/E7-(p16-) groups (p < 0.01). Because of significant treatment differences by site and because of the high association between HPV status, stage and tumor site, with 100% of oropharyngeal but only 52% of oral cavity cases detected in stages III/IV, treatment findings regarding E6/E7 are most likely confounded by site. Because of the relatively small sample, adjustment for stage or other potential confounders could not be performed.


This investigation demonstrates for the first time that there are 2 HPV DNA-positive HNC tumor groups distinguishable by E6 and/or E7 antibody status. Furthermore, differences in antibody status were associated with distinct risk factors and clinical outcomes even though all tumors were diagnosed with an HPV high-risk type. The distinctions in risk factors between the E6 and/or E7 antibody seropositive and seronegative HNC patients are similar to those that have been shown to distinguish HPV DNA tumor-positive from tumor-negative HNC patients: male gender, younger age, more sex partners and oral–genital sex, and oropharyngeal cancers.2, 3, 8, 19 Interestingly, the HPV DNA positive/seropositive cases were not different in tobacco or alcohol use from the DNA positive/seronegative group. In this study, E6 and/or E7 seropositive patients had pathobiological features similar to HPV DNA tumor-positive patients (worse stage, higher grade, nodal involvement and p16 overexpression) whereas the HPV DNA-positive/seronegative patients did not.11, 19, 20 Since HPV DNA-positive tumors are recognized as having significantly worse disease pathologic characteristics, our finding that the HPV E6/E7 seropositive group received more aggressive treatment than did the seronegative group was not unexpected.18, 20 Our results should lead to refinement of the definition of an HPV DNA-positive tumor and to its specific treatment since there are known differences in disease-specific and recurrence-free outcomes between the HPV DNA-positive and DNA-negative tumor cases.

Several previous studies suggest that HPV-16 DNA-positive HNC may fall into 2 biologically distinct groups. This investigation, and those of Smeets et al.21 in a small Dutch study and of Herrero in a multicenter IARC HNC study,1 suggest that only a subset of the HPV DNA tumor-positive cases are also E6 and/or E7 seropositive. HPV-16 DNA-positive/E6 and/or E7 RNA positive cases were found in 50% (12/24) of DNA-positive tumors in another Dutch series,21 and 43% (12/28) in a German series.22 HPV-associated molecular characteristics such as p16 up regulation also have been identified in HPV-16 DNA/E6 and/or E7 RNA-positive cases in these studies,22, 23 as have antibodies to HPV-16 E6 or E721 and in our current findings an association with oropharyngeal tumors. Braakhuis et al.23 found that among HPV-16 DNA-positive tumor cases which expressed HPV-16 E6/E7 RNA, all (100%) had p53 wild type whereas 3 of 8 (38%) HPV DNA-positive/RNA-negative cases had p53 mutations.23 Wiest et al.22 also found p53 wt at a high rate (91%, 11/12) in HPV-16 DNA-positive/RNA-positive cases but did not report findings for DNA-positive/RNA-negative cases. We found 68% of HPV-16 DNA-positive/E6 and/or E7 seropositive cases with wild type p53, like Braakhuis et al.23 based on tumor RNA, a much lower rate of HPV-16 DNA-positive//E6/E7-negative cases with p53 overexpression (32%). Our investigation and those of others21, 22 found that HPV DNA-positive/E6 and/or E7 RNA- or seronegative HNC cases were more often identified with cancer in the oral cavity or larynx and often were p16-negative (68%–83%). Thus, this group of patients has molecular, serologic and tumor characteristics similar to patients with HPV DNA-negative/E6 and/or E7 mRNA negative cases: the tumors are frequently found in these 2 HNC subsites (oral cavity and larynx), are p16-negative (81%-87%), and have a high rate of p53 mutations or p53 overexpression (75% in both Wiest et al.22 and Braakhuis,23 and 68% in our series). In the HPV DNA-positive/E6/E7 mRNA-negative cases, the viral genes are not transcriptionally active despite the presence of HPV genomes, and thus may not reflect biologically meaningful viral involvement.21 Therefore, patients with HPV-16 DNA-positive tumors that fail to elicit E6 and E7 antibodies may comprise those where viral E6 and E7 expression is no longer needed for tumor growth as well as those who harbor cancers that express the viral E6 and/or E7 oncoproteins yet do not mount antibody response against them.

In addition to the differences in molecular characteristics, several case-control investigations suggest that the risk of HNC in HPV DNA-positive/L1-seropositive individuals is not increased further by use or level of tobacco or alcohol whereas their use does increase the risk of HNC among HPV L1 seronegative cases.2, 24 Although the Dutch and the IARC multicenter study did not describe risk factors among HPV DNA-positive tumors by E6/E7 serology status, our data indicate that tobacco and alcohol do not play a distinct role between the DNA-positive/seropositive versus DNA-positive/seronegative HNC patients.

One of the potentially useful measures of an HPV DNA-positive tumor is use of a surrogate marker such as detection of antibodies to E6 and/or E7. E6/E7 serologic status could be assessed at the time of clinical diagnosis before tissue diagnosis by tests performed routinely in clinical pathology labs and lead to targeted treatment planning. Furthermore, E6/E7 expression has been associated with a high rate of p16 overexpression in HNC.21, 22 Of greater significance, prognostic findings based on E6 and E7 antibodies and p16 in this investigation were consistent with the results of previous studies of ours and Klussmann et al.20, 25 that examined the joint effect between HPV DNA tumor status and p16 overexpression. Both median and total disease-specific and recurrence-free survival were better in HPV seropositive/p16+ patients compared with HPV seronegative/p16-negative patients. Because of the relatively small number of cases and limited duration in survival, these findings need to be confirmed. Smeets et al.21 have proposed an algorithm for high throughput identification of HNC among those with clinically relevant HPV infection. First paraffin-embedded tissue is examined for p16 immunohistochemistry, followed by GP5+/6+-PCR on p16-positive HNC cases., producing 100% sensitivity and specificity in their study. This method would reduce limitations of single assay results. Unfortunately, it cannot be performed prior to treatment.

Evidence for continued HPV involvement in a tumor requires measuring levels of E6/E7 RNA to determine if the viral DNA is transcriptionally active. This is a critical determinant but necessitates adequate fresh frozen tumor tissue which is only available after treatment has been initiated. Furthermore, pretreatment and adequate tissue may not be available for patients first treated by radiation or chemotherapy. Use of an algorithm of detection methods like that proposed by Smeets et al.21 but which can be performed prior to treatment or immediately after surgery may provide a sensitive diagnostic procedure that does not lead to overtreating the patient. Distinctions in HPV E6/E7 antibody status may be one such method; our results suggests that E6/E7 seropositive- and seronegative-related HPV DNA tumors represent different tumor entities driven by different risk factors. Sensitivity and specificity of E6/E7 antibodies to detect HPV-driven HNC and their potential as a screening marker and prediction of clinical outcomes for this tumor subgroup remain to be determined in larger series of HNC cases. The assessment of HPV E6 and E7 antibody response has the potential to be developed as a clinical test for HPV-driven HNC tumors that is rapid and has high predictive value.


This study was supported by a grant from NIH NIDCR R01 DE11979 (EMS, LMR, THH, LPT), NIDCR R01 DE13110 (EMS, LMR, THH, LPT), and Veterans Affairs Merit Review Funds (LPT, THH).