Human papillomavirus seropositivity and risks of head and neck cancer
We examined antibody response to VLP HPV-16, HPV-16 E6 and E7 antibodies as potential seromarkers of HPV-related head and neck cancer (HNC). The study included 204 HNC cases and 326 controls evaluated for HPV presence in sera using ELISAs for anti-HPV VLP antibodies and HPV-16 E6 and/or E7 antibodies, and in tumor tissue using PCR and DNA sequencing. Anti-HPV-16 VLP was detected in 33.8% of cases and 22.4% of controls, anti-E6 in 20.6% of cases and 0.9% of controls and anti-E7 in 18.6% of cases and 0.6% of controls. HPV-16 DNA was detected in 26.1% of tumors. The adjusted risk of HNC was elevated among those seropositive for HPV-16 VLP (odds ratio (OR) = 1.7, 1.1–2.5), E6 (OR = 32.8, 9.7–110.8) or E7 (OR = 37.5, 8.7–161.2). Compared to HPV DNA-negative/seronegative cases, tumor HPV-16 cases had increased risk of detection with anti-VLP antibodies (OR = 6.8, 3.1–14.9). The odds were more pronounced among cases seropositive for E6 (OR = 69.0, 19.3–247) or E7 (OR = 50.1, 14.7–171). Antibodies against E6 or E7 were associated with risk of cancer in the oral cavity (OR = 5.1, 1.2–22.4) and oropharynx (OR = 72.8, 16.0–330), and with disease characteristics: stage, grade and nodal status. Anti-E6 and/or E7 antibodies were found in 74% of tumor HPV-16 positive cases but in only 5% of tumor HPV-negative cases (K =0.7, 0.6–0.8) suggesting good correlation between the serologic marker and HPV tumor status. Antibodies to HPV-16 E6 and/or E7 represent a more specific biomarker than anti-HPV-16 VLP of an HPV-related HNC. Because of the survival advantage of HPV-related HNC, HPV-16 E6/E7 detection may be useful in therapy targeted for HPV-related tumors. © 2006 Wiley-Liss, Inc.
Virtually all cervical carcinomas and more than 25% of head and neck cancers (HNC) are associated with HPV high-risk (HR) types.1, 2, 3 Two serologic markers have been used as surrogates of HPV infection: antibodies to HPV-derived virus-like particles (VLP) and to HPV-16 E6 or E7 oncoproteins. Most investigations have used anti-VLP antibodies which are considered markers of changes that occur early in the development of tumorigenesis and also markers of lifetime exposure to an HPV infection.4, 5, 6 Although their presence may not represent a current HPV-related malignancy, the magnitude of risk between the presence of anti-VLPs for HPV-16,18 or 33 and cervical cancer is 2–3 fold.4, 5, 7 The anti-HPV-16 or 18 E6/E7 antibodies have been evaluated more recently to represent specific markers of a current HPV-associated cancer and they are overexpressed in the tumor.8, 9, 10, 11 The low prevalence of the anti-E6/E7 antibodies in healthy individuals without an HPV-related tumor has been verified in several large studies of adult populations.12, 13 In these studies only 2–3% of healthy individuals showed serological response to HPV-16 or HPV-18 E6 and/or E7 oncoproteins, whereas these rates increased with severity of cervical dysplasia and have been detected in 25–54% of cases with cervical cancer.13 However, it remains unclear whether serologic tests may be useful as biomarkers of early detection of HPV-related tumors.
Few studies have evaluated serologic markers of HPV infection in HNC tumors. In a case–control study, Schwartz et al.14 found that 70% of HPV-16 tumor positive HNC cases were antibody positive to HPV-16. In contrast Zumbach et al.15 found only 37% of HPV-16 DNA positive tumors with anti-HPV-16 E6 and/or E7 antibodies from sera taken during or after diagnosis. An international case–control study conducted by IARC12 is the only investigation to date that has examined both anti-HPV VLP antibodies and anti-HPV-16 E6 and E7 antibodies as risk factors in the same patients. They found an increased risk of cancer in the oral cavity and oropharynx based on both types of seropositive markers with a slightly higher risk in those with antibody response to HPV-16 E6 or E7 than to HPV-16 VLP, but seropositivity was low in both assays, <14%. The association between these HPV serologic measures and HPV DNA detected in the tumors was limited by the quality and quantity of specimens collected, differences in recruitment of patients from diverse geographic locations, wide variation in the prevalence of HNC risk factor exposures, HPV, tobacco and alcohol type, dose, and duration, and in differences in site-specific frequencies of HNC. These variations can diminish the power of a study to find statistically and clinically significant findings, particularly where the risk factor of interest, HPV, is strongly associated with sexual behaviors. Thus, studies of serologic markers of HPV raise several questions about their use as predictors of HPV-associated head and neck tumors. In this study, we tested the prevalence of antibodies to HPV-16 VLP and HPV-16 E6 and E7 in a group of HNC cases and controls from a single geographic location with a similar cultural background thus minimizing diversity in HNC etiologic exposures. Using laser capture microdissection to verify that the viral DNA was integrated into the tumor cells and not extrachromosomal, we evaluated the serologic markers as predictors of an HPV-HR tumor.
All primary cases of the oral cavity or oropharynx seen at the University of Iowa Hospitals, Department of Otolaryngology, between 2000 and 2004 were eligible for enrollment. Patients were administered a University approved Human Subjects Review form prior to interview. Included were all oral cavity and pharyngeal sites as defined by the American Joint Commission on Cancer,16 excluding nasopharynx. The data are reported for all histologies combined because the results for SCC only cases were similar to those reported among all histologies. Other histologies (n = 21) included 1 acinic cell carcinoma, 4 adenocarcinoma, 4 adenoid cystic carcinoma, 2 basal cell adenocarcinoma, 1 carcinoma NOS, 1 polymorphous adenocarcinoma, 7 mucoepidermoid carcinoma and 1 verrucous carcinoma. Exclusion criteria included patients who were mentally not capable of consent, less than 18 years old, or too ill (7.1%), refused (16.7%), or missed being interviewed (13.3% including unavailable blood samples due to emergency surgery), leaving 204 eligible cases included in this analysis. Controls were frequency matched for gender and 5-year age groups to cases and were recruited from clinics in Family Medicine and in Internal Medicine at the same hospitals as the cases where they were seeking routine screening or medical visits for prescriptions. Like cases, the majority of controls reside outside the county in which the hospital is located. Because of the rural nature of the study area, lack of health care services and specialties leads many individuals to seek routine medical care at this tertiary health care system. Exclusion criteria in the nondiseased control patients included under age 18, prior history of HNC, or treatment for acute or chronic serious diseases. This information was verified from hospital medical records and the Iowa NCI-SEER Cancer Registry. Of those excluded, 3.1% did not complete the specimen collection requirements, 1.1% failed to get blood drawn and 2.6% refused to participate. There were 326 controls enrolled.
Participants completed a risk factor questionnaire with information about sociodemographics, medical history, sexual practices, smoking habits, alcohol use and history of HPV-related diseases and oral lesions. A medical form was completed by the clinical research staff based on a face-to-face interview with the patient supplemented by information in the medical record. These data included prior cancer history, cancer site and treatments. Never tobacco or alcohol use was defined as not having used these substances on a regular basis during their lifetime for 1 year or more. Former use was defined as not having used tobacco or alcohol at least 1 year prior to diagnosis or interview. Current users included those who used tobacco or alcohol currently or less than 1 year prior to the time of diagnosis or interview. Ever users included those who had used tobacco products or alcohol for at least a year or longer in their life. Ever tobacco use required at least half a pack of cigarettes, one or more pipesful, one small can of snuff, or one pouch of chewing tobacco a week for a year or longer. Alcohol use required an average of at least 1 drink a week for a year or longer of any alcoholic beverage. Tobacco equivalent units were: 1 pack-year = 1 pack of cigarettes/day/year, with 20 cigarettes in a pack. Equivalence conversion of tobacco products were: 1 cigar = 1 pipeful = 1 ounce of snuff or chewing tobacco = 1 cigarette. Alcohol exposure was defined in a similar manner with exposure expressed as the average number of drinks/week. One drink was equivalent to a 12-ounce can/bottle of beer = 4-ounce glass of wine = 1[1/2]-ounce shot of liquor. Light users of tobacco were defined as those who had ≤30 packyears (or equivalent) and heavy users were >30 packyears. Light alcohol users drank ≤21 drinks/week on average and heavy users drank >21 drinks/week.
These procedures performed on clinical samples have been described previously.1, 17, 18, 19 Among the HNC cases, 176 had paraffin-embedded tumor tissue available at the time of analysis and prepared as previously described.1 Extraction of DNA from deparaffinized tissue sections was performed with a QIAGEN DNA Tissue Kit, according to the manufacturer's instructions (QIAGEN, Valencia, CA). The DNA was eluted from the column with 100 μl of QIAamp buffer AE and stored at −20°C in preparation for the PCR amplifications. Laser-capture microdissection (LCM) procedures19 were performed when tissue specimens had less than 10% of tumor tissue or when they were initially found to contain an HPV nononcogenic, low-risk type (HPV-LR; 1 case). The procedure verified that the HPV DNA was in the tumor cells and not in adjacent nontumorous epithelium except for the HPV-LR type detected which was not found by LCM to be in the tumor cells. LCM was performed in a separate laboratory on coded slides to assure independent, blinded results. As few as 20 microdissected cells routinely yielded sufficient DNA for analysis. HPV-16 tumor positive HNC case status is based on the paraffin-block tissue findings.
Sample preparation, PCR analyses, DNA hybridization and HPV typing procedures for assessment of exfoliated oral cells and tumor tissue were based on a standard protocol described previously.1, 18 Each PCR reaction included primers to amplify the β-globin gene20 to verify adequate DNA and adequacy of the PCR amplifications of the cellular DNA. All tumor tissue tested positive for the β-globin gene. Two percent of the DNA (typically 50–200 ng) extracted from tumor tissue specimens was PCR-amplified with MY09 and MY11 primers21 to detect HPV and a primer (HMB01) designed to amplify HPV-51 to improve detection of this type.22 The PCR product was first examined for the presence of a 400 nt HPV amplified band on ethidium bromide stained 1% agarose gels. An aliquot of the PCR product of all those specimens that were HPV-negative after staining was hybridized by the dot blot method with 32P-labeled probes for a more sensitive detection of HPV DNA. Those samples positive only after the membrane hybridization underwent heminested PCR-amplification with MY09 and GP5+primers.23 DNA sequencing was used to determine the HPV types in each specimen, performed by dye-termination on a DNA sequencer (Applied Biosystems-PE). The nucleotide sequences were compared to GenBank sequences using the BLAST program.24
Detection of HPV-specific antibodies
Presence of antibodies to antigens derived from HPV-specific proteins was evaluated using the enzyme-linked immunosorbent assay (ELISA). Pseudovirions mimicking HPV types 16, 18, 31 and 33, prepared in baculovirus expression system, were used as antigens to detect antibodies to capsid proteins.25 Presence of antibodies to oncoproteins E6 and E7 derived from type 16 was assessed using glutathione-S-transferase (GST) capture ELISA system.26
To detect the presence of anti-VLP antibodies, wells of microtiter plates (Polysorp; NUNC, Roskilde, Denmark) were coated with purified VLP in PBS at 37°C for 2 hr and at 4°C overnight. All subsequent incubations were performed at 37°C for 1 hr. To remove unbound reagents, wells were repeatedly washed using automatic microplate washer 1575 (Bio-Rad Laboratories, Hercules, CA) with buffer A (PBS, 0.21 mol/1 NaCl and 0.1% Triton X-100). Nonspecific binding sites were blocked by incubation with 1% BSA in PBS and the wells were subsequently incubated in duplicate with human sera diluted 1:25 in buffer A with 1% of BSA. Following incubation, antibodies bound were detected with swine antihuman IgG covalently linked to horseradish peroxidase (SEVAPHARMA, Prague, Czech Republic) and the reaction visualized by adding 100 μl of o-phenylenediamine containing substrate solution. The color reaction was stopped by 100 μl of 2 mol/l H2SO4 and optical densities (ODs) at 492 nm were read with an ELx808 Ultra Microplate Reader (Bio-Tek Instruments, Winooski, VT).
For detection of anti HPV-16 E6 or E7 antibodies, specific antigens were prepared as overexpressed bacterial fusion proteins containing GST at their N-terminus and C-terminal undecapeptide of the SV40 large T-antigen as tagging epitope at their C-terminus.26 The wells of microtiter plates (Polysorp; NUNC, Roskilde, Denmark) were coated with glutathion-casein conjugate in bicarbonate buffer, pH 9.6 at 4°C overnight. All subsequent incubations were performed at 37°C for 1 hr. To remove unbound reagents, wells were repeatedly washed using automatic microplate washer 1575 (Bio-Rad Laboratories, Hercules, CA) with washing buffer (PBS, 0.05% Tween-20). The nonspecific binding sites were blocked by incubation with buffer B (0.2% casein in PBS, 0.05% Tween-20), followed by incubation with cleared lysates from E. coli overexpressing GST-fusion proteins or GST alone in buffer B (0.25 mg total protein/ml). Sera were diluted in buffer B with 0.25 mg/ml GST lysate, incubated for 30 min at room temperature and added in duplicates to antigen-coated wells. Following incubation, antibodies bound were detected with donkey anti-human IgG (H and L chain), covalently linked to horseradish peroxidase (Jackson ImmunoResearch Laboratories, West Grove, PA) in buffer B and the reaction visualized using tetramethylbenzidine in substrate solution. The color reaction was stopped by 1 mol/l H2SO4 and ODs at 450/630 nm were read with an ELx808 Ultra Microplate Reader.
Background reactivity of samples was determined in case of anti-VLP antibodies in wells without antigen, in case of anti-E6/E7 antibodies in wells coated with GST antigen. Their absorbances were subtracted from corresponding values obtained in the presence of HPV-specific antigen. Control sera known to be positive and negative were tested on each plate. The cutoff level, above which OD values were considered positive, was calculated separately for each antigen on a collection of 10 standard sera from regular blood donors known to be HPV-antibody negative, included on each plate. The cutoff was represented as a mean absorbance plus 3 SD of these sera. All samples 10% above the cutoff value were tested twice more; those positive in all three runs were counted positive. About 20% randomly selected specimens were retested to confirm the results and the findings remained unchanged. In all cases, OD were recalculated into ratios by division of the particular cutoff value. An OD ratio above 1 was defined as positive.
The Wilcoxon rank sum test was used to compare quantitative variables between 2 groups of patients (e.g. cases vs. controls, seropositive cases vs. seronegative cases). The χ2 or Fisher's exact test was used to compare categorical variables among groups. Multiple logistic regression was used to estimate the odds ratios (ORs) of HNC after adjusting for age as a continuous variable, tobacco packyears (never, 0–30, >30) and average number of drinks per week (never, 1–21, >21). The cut points examined for tobacco and alcohol usage were based on the distributions among controls. Lack of linearity for continuous variables in the logistic regression models was assessed as discussed by Hosmer and Lemeshow.27 The referent group used in analyses to examine the association between HPV seropositivity status and risk of HNC included only individuals who were seronegative for the particular antigen being examined. Effect modification between HPV seropositivity status and alcohol or tobacco usage was examined on a multiplicative scale by including the appropriate interaction term into the logistic regression models. ORs of HPV seropositivity among cases also were adjusted for age, tobacco packyears and average drinks/week. Additional adjustment including number of lifetime sex partners and gender was examined but estimates were similar to those presented. To assess the potential prognostic value of the biomarkers, specificity of the tests was calculated. The specificity of the test is the probability that the test classifies people without the disease as negative and determines whether the frequency or proportion of false positives will be clinically acceptable. The confidence interval around the proportion of negative controls was calculated using interval estimation methods for binomial proportions.28 The Wald statistic was used to determine p-values based on logistic regression analyses to adjust for age, tobacco and alcohol use. The κ statistic was used to measure agreement for HPV positivity status between serology vs. tumor DNA samples. Statistical analyses were performed using SAS version 8.2.29
Case–control and site-specific HNC risks
There were 204 cases and 326 age–gender frequency matched controls in the study (Table I). Among the cases, 69.6% had cancers of the oral cavity and 30.4% had cancers of the oropharynx. Males were more likely than females to be diagnosed with a HNC and the average age of cases and controls was similar (59.5 years). There was no difference between cases and controls in frequencies of sexual behaviors that have been associated with an HPV-HR infection in HNC: number of lifetime sex partners or oral–genital sex. Compared to controls, HNC cases were more likely to be heavy users of both tobacco and alcohol. Male cases were significantly more likely than male controls to smoke heavily/drink lightly or smoke and drink heavily (p < 0.05 each; data not shown). Among females, heavy drinking was significantly associated with risk of cancer (p < 0.05) whereas tobacco use was not.
Table I. Characteristics of HNC Cases and Controls1
| Male||127 (62.2)||213 (65.3)|| |
| Female||77 (37.8)||113 (34.7)|| |
| ≤55 years||88 (43.1)||123 (37.7)|| |
| >55 years||116 (56.9)||203 (62.3)|| |
|No. of partners|
| 0–3||96 (50.8)||166 (53.2)||1.0|
| ≥4||93 (49.2)||146 (46.8)||0.8 (0.5–1.2)|
| Yes||48 (53.3)||142 (51.6)||0.9 (0.5–1.6)|
| No||42 (46.7)||133 (48.4)||1.0|
| Never/Never||40 (19.9)||76 (23.8)||1.0|
| ≤30/≤21||27 (13.4)||60 (18.8)||0.9 (0.5–1.6)|
| ≤30/>21||13 (6.5)||18 (5.6)||1.4 (0.6–3.1)|
| >30/≤21||32 (15.9)||42 (13.2)||1.5 (0.8–2.6)|
| >30/>21||53 (26.4)||19 (6.0)||5.3 (2.8–10.2)|
After adjusting for age, tobacco and alcohol, heavy drinking (OR = 2.8, 1.5–5.2) or heavy use of both tobacco and alcohol (OR = 4.0, 2.0–7.7) increased the risk of tumors of the oral cavity compared to never tobacco/alcohol users. Cases with cancer at this site also were more likely to report a prior diagnosis of oral leukoplakia (OR = 10.6, 2.9–38.5). An increased risk of cancer of the oropharynx was found in younger age (≤55 years vs. >55 years; OR = 2.6, 1.4–4.9) and site-specific risk was elevated in both light (OR = 2.7, 1.0–7.0) and heavy tobacco users (OR = 5.6, 2.2–14.3) and light (OR = 2.2, 0.99–5.1) and heavy (OR = 4.3, 1.8–10.6) alcohol users compared to never users.
Anti-HPV-HR VLP and anti-E6 and/or E7 antibodies in cases and controls
The prevalence and risk of detection with serologic markers against HPV-16, 18, 31 and 33 VLP, and HPV-16 E6 and E7 proteins are shown in Table II. The comparison group was defined as those who did not have the particular antibody present. However, we found few differences in results reported in Table II when the comparison group included only those with no other HPV-HR VLP or E6/E7 antibodies (data not shown). Cases had higher positivity rates than did controls for all seromarkers and were more likely to be detected with any of the anti-HPV-HR VLP antibodies (46.1% vs. 40.2%). Antibodies to HPV-16 VLP were the most prevalent (33.8% in cases vs. 22.4% in controls) and risk of cancer was significantly elevated among those who were seropositive for HPV-16 VLP. Cases had higher average OD ratios than controls (1.9 vs. 1.6, p = 0.01). Site-specific frequencies also were higher in cases than controls (Table II).
Table II. Risk of HNC Associated With HPV-HR VLPs and HPV-16 E6/E7 Antibodies
| Negative3||110 (53.9)4||195 (59.8)||1.0||83 (58.4)||1.0||27 (43.6)||1.0|
| Positive||94 (46.1)||131 (40.2)||1.3 [0.9–1.8]5||59 (41.6)||1.0 [0.7–1.6]||35 (56.4)||2.1 [1.1–3.8]|
| Negative3||135 (66.2)||253 (77.6)||1.0||104 (73.2)||1.0||31 (50.0)||1.0|
| Positive||69 (33.8)||73 (22.4)||1.7 [1.1–2.5]||38 (26.8)||1.2 [0.7–2.0]||31 (50.0)||3.5 [1.9–6.5]|
| Negative3||168 (82.4)||279 (85.6)||1.0||119 (83.8)||1.0||49 (79.0)||1.0|
| Positive||36 (17.7)||47 (14.4)||1.2 [0.7–2.0]||23 (16.2)||1.2 [0.7–2.1]||13 (21.0)||1.5 [0.7–3.3]|
| Negative3||146 (71.6)||253 (77.6)||1.0||103 (72.5)||1.0||43 (69.4)||1.0|
| Positive||58 (28.4)||73 (22.4)||1.2 [0.8–1.9]||39 (27.5)||1.2 [0.8–2.0]||19 (30.7)||1.4 [0.7–2.7]|
| Negative3||158 (77.5)||264 (81.0)||1.0||117 (82.4)||1.0||41 (66.1)||1.0|
| Positive||46 (22.5)||62 (19.0)||1.1 [0.7–1.8]||25 (17.6)||0.8 [0.5–1.4]||21 (33.9)||2.2 [1.1–4.2]|
|VLP HPV-HR & E6/E7 Negative|
| Negative6||93 (60.4)||193 (59.9)||1.0||80 (60.6)||1.0||13 (59.1)||1.0|
| Positive||61 (39.6)||129 (40.1)||0.9 [0.6–1.4]||52 (39.4)||0.97 [0.6–1.5]||9 (40.9)||0.9 [0.4–2.5]|
| Negative6||162 (79.4)||323 (99.1)||1.0||133 (93.7)||1.0||29 (46.8)||1.0|
| Positive||42 (20.6)||3 (0.9)||32.8 [9.7–110.8]||9 (6.3)||6.2 [1.5–25.4]||33 (53.2)||229.8 [54.1–976]|
| Negative6||166 (81.4)||324 (99.4)||1.0||137 (96.5)||1.0||29 (46.8)||1.0|
| Positive||38 (18.6)||2 (0.6)||37.5 [8.7–161.2]||5 (3.5)||3.0 [0.5–16.9]||33 (53.2)||342.2 [61.1–∞]|
|HPV-16 E6 or E7|
| Negative6||154 (75.5)||322 (98.8)||1.0||132 (93.0)||1.0||22 (35.5)||1.0|
| Either E6 or E7||20 (9.8)||3 (0.9)||15.0 [4.2–53.4]||6 (4.2)||5.1 [1.2–22.4]||14 (22.6)||72.8 [16.0–330]|
| Both||30 (14.7)||1 (0.3)||70.4 [9.3–533.0]||4 (2.8)||4.7 [0.5–45.3]||26 (41.9)||384 [49.3–∞]|
| Either or both positive||50 (24.5)||4 (1.2)||28.8 [9.9–84.0]||10 (11.1)||4.9 [1.4–17.2]||40 (64.5)||231 [62.0–859]|
Anti-HPV-16 E6 antibody was detected in 20.6% of cases and 0.9% of controls, anti-E7 was detected in 18.6% of cases and 0.6% of controls, and those with either antibody or both were at higher risk of HNC compared to those without the antibody (Table II). Antibodies against either or both E6 and E7 also were associated with a significant risk for site-specific HNC of the oral cavity and oropharynx. Risk of cancer was much greater when individuals were detected with anti-HPV-16 E6 and/or E7 than when individuals were detected with anti-HPV-16 VLP. Controls were almost never detected with anti-E6 or E7 antibodies in contrast to anti-HPV-16 VLP antibodies. All but one of the controls detected with the E6/E7 protein-derived markers had an OD ratio slightly exceeding the minimal positive cutoff. The single high OD positive control (>2.0 for E6 and E7) was male who might have undiagnosed HPV-HR disease elsewhere in the body. The fact that few healthy individuals were detected with these antibodies also explains the wide CIs. To assess whether the associations observed between antibody response to HPV-16 capsids or to E6/E7 proteins and case–control status were affected by sexual transmission of HPV, we included number of sexual partners into the models in Table II, but the ORs remained unchanged.
Cancer of the oral cavity had much lower seropositive rates than did those in the oropharynx (Table II) Risk of cancer detection in the oropharynx was much greater than the risk in the oral cavity for both HPV-16 serologic markers. Anti-HPV-16 VLP was not elevated in oral cavity cancers compared to controls but risk was increased among those detected with anti-E6 or anti-E7. Although risk of malignancy in the oropharynx was significantly elevated among those seropositive for anti- HPV-16 VLP, risk was much greater among those detected with anti-E6 or E7 oncoproteins.
Significant differences in HPV seropositivity were detected between cases and controls by gender. Compared to male controls, male cases were significantly more likely to be detected with anti-HPV-16 VLP (OR = 1.9, 1.1–3.3) and anti-HPV-33 VLP (OR = 1.8, 0.99–3.2), whereas in females there was no difference in type-specific anti-VLP antibodies between cases and controls (HPV-16: OR = 1.4, 0.7–2.7; HPV-33: 0.6, 0.3–1.2). Risks associated with anti-E6 and/or E7 antibodies could not be computed because few controls were detected with these antibodies. There was little difference in the percentage of male and female HNC cases detected with anti-HPV-16 VLP (35% vs. 33%) in contrast to anti-E6 and/or E7 (32% vs. 12%). The gender association with E6/E7 seropositivity may be due to men having a higher frequency of oropharyngeal cancers than women (38.6% vs. 16.9%; p = 0.001) and cancers of the oropharynx have a higher prevalence of anti-HPV E6/E7 antibodies than those of the oral cavity. In addition, men were significantly more likely than women to have ≥4 sex partners (p = 0.01), thus increasing their risk of HPV acquisition.
Joint associations were examined between seropositivity to HPV-16 VLP or E6 and/or E7 with tobacco or alcohol and risk of HNC (Table III). There was a significant interaction between HPV-16 VLP seropositivity and tobacco (ever versus never use: p = 0.02). Compared with never tobacco users/HPV-16 VLP seronegative, the ever tobacco users/HPV-16 VLP seropositive were at increased risk of HNC, whereas the never tobacco users/HPV-16 VLP seropositive or ever tobacco/HPV-16 VLP seronegative were not. There was no significant interaction between anti-HPV-16 E6/E7 positivity and tobacco (p = 0.6), but there were significant risks of HNC among the never tobacco/E6/E7 positive group and for the ever tobacco/E6/E7 seropositive group compared to never tobacco users/E6/E7 seronegative group. Compared to the never tobacco/E6/E7 seronegative group, the adjusted OR for the ever tobacco/E6/E7 seronegative group was 1.0. These findings suggest that the joint effect of tobacco and HPV-16 E6/E7 seropositivity may be more than an additive effect of the two exposures, but the low number of E6/E7 seropositive controls limits the conclusions. There were no significant interactions with alcohol for either HPV-16 VLP or E6/E7 seropositivity (data not shown).
Table III. Risk of Head and Neck Cancer Associated with HPV-16 VLP and E6 and/or E7 by Tobacco and Alcohol Use
| Negative||Never||43 (21.1)2||94 (28.8)||1.0|| ||p = 0.02|
|Ever||92 (45.1)||159 (48.8)||0.9||0.5–1.4|
| Positive||Never||12 (5.9)||31 (9.5)||0.8||0.4–1.7|
|Ever||57 (27.9)||42 (12.9)||2.0||1.1–3.7|
|E6 and/or E7|
| Negative||Never||48 (23.5)||124 (38.0)||1.0|| ||p = 0.65|
|Ever||106 (52.0)||198 (60.7)||1.0||0.6–1.5|
| Positive||Never||7 (3.4)||1 (0.3)||17.3||2.1–146|
|Ever||43 (21.1)||3 (0.9)||29.3||8.4–102|
Risk factor comparison between HPV-16 VLP and E6 and/or E7 seropositivity status
None of the risk factors shown in Table I was significantly different between either the HPV-16 VLP or E6/E7 seropositive and seronegative controls (data not shown).
Table IV shows differences in risk factors and tumor characteristics between cases positive for anti-HPV-16 VLP or anti-HPV-16 E6 and/or E7 compared to cases negative for the specific antibody after controlling for potential confounders. There were no significant differences in risk factors between anti-HPV-16 VLP negative or positive cases. In contrast, seropositivity to HPV-16 E6/E7 was significantly greater for all risk factors: male gender, younger age, greater number of lifetime sex partners, oral–genital sex, and categories of low, moderate and high tobacco/alcohol dose-durations compared to never users of both. Among those detected with anti-HPV-16 VLP, only tumors in the oropharynx showed a near significantly elevated risk for pathologic outcomes, whereas all tumor characteristics were significantly elevated in cases seropositive for HPV-16 E6/E7: oropharynx, stage, nodes and grade. In addition, those with anti-E7 antibodies had 2–3 fold higher risks than did those with anti-E6 antibodies associated with worse stage, nodal status and grade but not tumor site (data not shown).
Table IV. Demographic and Risk Factors Associated with HPV-16 VLP and HPV-16 E6 and/or E7 Antibodies in Cases1
| Male||83 (61.5)||44 (63.8)||0.7 [0.3–1.4]4||86 (55.8)||41 (82.0)||3.0 [1.2–7.1]|
| Female||52 (38.5)||25 (36.2)||1.0||68 (44.2)||9 (18.0)||1.0|
| ≤55||52 (38.5)||36 (52.2)||1.8 [0.97–3.3]||56 (36.4)||32 (64.0)||3.3 [1.7–6.6]|
| >55||83 (61.5)||33 (47.8)||1.0||98 (63.6)||18 (36.0)||1.0|
|No. of partners|
| 0–3||72 (58.1)||24 (36.9)||1.0||82 (58.2)||14 (29.2)||1.0|
| ≥4||52 (41.9)||41 (63.1)||1.7 [0.8–3.6]||59 (41.8)||34 (70.8)||2.3 [1.03–5.4]|
| Yes||32 (51.6)||16 (57.1)||0.6 [0.2–2.0]||39 (57.4)||19 (86.4)||7.8 [1.6–38.4]|
| No||30 (48.4)||12 (42.9)||1.0||29 (42.6)||3 (13.6)||1.0|
| Never, Never||30 (28.3)||10 (17.0)||1.0||37 (24.3)||3 (6.1)||1.0|
| ≤306, ≤217||16 (15.1)||11 (18.6)||1.9 [0.6–5.5]||17 (11.2)||10 (20.4)||6.7 [1.6–28.1]|
| ≤30, >21||9 (8.5)||4 (6.8)||1.3 [0.7–2.6]||9 (7.1)||4 (10.3)||8.9 [1.0–30.9]|
| >30, ≤21||18 (17.0)||14 (23.7)||2.5 [0.9–6.9]||23 (15.1)||9 (18.4)||5.4 [1.3–22.3]|
| >30, >21||33 (31.3)||20 (33.9)||1.7 [0.7–4.3]||40 (26.3)||13 (26.5)||3.9 [1.1–15.1]|
| Oral cavity||104 (77.0)||38 (55.1)||1.0||132 (85.7)||10 (20.0)||1.0|
| Oropharynx||31 (23.0)||31 (44.9)||2.0 [0.99–4.0]||22 (14.3)||40 (80.0)||21.7 [9.0–52.4]|
| Well/Mod||91 (76.5)||47 (77.0)||1.0||113 (82.5)||25 (58.1)||1.0|
| Poorly/Undiff||28 (23.5)||14 (23.0)||0.8 [0.4–1.7]||24 (17.5)||18 (41.9)||3.0 [1.3–6.5]|
| 0, I, II||49 (38.3)||16 (23.5)||1.0||59 (40.1)||6 (12.2)||1.0|
| III, IV||79 (61.7)||52 (76.5)||1.7 [0.9–3.5]||88 (59.9)||43 (87.8)||3.9 [1.5–10.2]|
| Yes||59 (45.7)||43 (63.2)||1.6 [0.9–3.1]||85 (57.4)||39 (79.6)||4.3 [1.9–9.6]|
| No||70 (54.3)||25 (36.8)||1.0||63 (42.6)||10 (20.4)||1.0|
Association between HPV status in serology compared to tumors
At the time of the study analysis, DNA from tumor tissue was available from 176 cases to test for HPV. HPV-16 DNA was the only type detected in tumors (26.1%). Table V shows the association between HPV status by serologic assay compared to tumor tissue. In the case–control comparisons, cases whose tumor tissue contained HPV-16 were significantly more likely than controls to be detected with anti-HPV-16 VLP antibodies, anti-E6 or anti-E7 proteins (p < 0.0001 each), whereas cases with HPV-negative tumors were not more likely than controls to be detected with HPV antibodies (p = 0.65/0.10/0.34).
Table V. Association Between HPV DNA in Tumors with Seropositivity Based on HPV-16 VLP and HPV-16 E6 and/or E7 Antibodies
| Negative1||253 (77.6)||16 (34.8)||102 (78.5)|
| Positive||73 (22.4)||30 (65.2)||28 (21.5)|
| p-value (case–control comparison)23|| ||<0.00014||0.655|
| Adj. OR267 (case–case comparison)||6.8 (3.1–14.9)|
| Negative8||323 (99.1)||16 (34.8)||126 (96.9)|
| Positive||3 (0.9)||30 (65.2)||4 (3.1)|
| p-value23 (case–control)|| ||<0.00014||0.105|
| Adj. OR267 (case–case comparison)||69.0 (19.3–246.5)|
| Negative9||324 (99.4)||18 (39.1)||126 (96.9)|
| Positive||2 (0.6)||28 (60.9)||4 (3.1)|
| p-value23 (case–control)|| ||<0.00014||0.345|
| Adj.OR267 (case–case comparison)||50.1 (14.7–170.8)|
In the case–case comparisons, HPV-16 seropositivity rates for each of the serology assays were similar for cases with HPV-16 positive tumor tissue (61–65%). Compared to tumor HPV-negative cases, tumor positive cases were at significantly increased risk of detection with anti-HPV-16 VLP (OR = 6.8) but the risk of seropositivity was greater for E6 (OR = 69.0) or E7 (OR = 50.1). Among E6 seropositive cases, there was a significant difference in median OD ratios between HPV DNA positive and negative tumors (OD = 5.3 vs. 1.4, p = 0.02; data not shown), whereas among E7 seropositive cases there was no difference in E7 OD values (OD ratio = 5.5 vs. 3.8, p = 0.7) between the HPV DNA positive and negative tumors.
The specificity of the HPV serology tests was higher in those not detected with anti-HPV-16 E6 or E7 than it was in those not detected with anti-HPV-16 VLP. Among tumor HPV-positive cases, 65% had antibodies to HPV-16 VLP compared to 22% of tumor HPV-negative cases (K = 0.4, 0.3–0.6). In contrast, antibodies to HPV-16 E6 and/or E7 were found in 74% of HPV-16 tumor positive cases but in only 5% of the HPV DNA negative cases (K =0.7, 0.6–0.8). Among cases seronegative for HPV-16 E6 and E7 antigens whose tumors were HPV-16 positive (n = 12), 8 were seropositive for HPV-16, 31 or 33 VLP and in 4 cases no HPV-related antibodies were found. Thus seropositivity for HPV-16 with E6 or E7 was a better predictor of HPV-16 DNA in tumor tissue than was anti-HPV-16 VLP.
We compared two serologic measures of HPV infection as potential biomarkers of HPV-related HNC risk in a large case–control study. The risk of an HPV-related tumor was much greater and the specificity of the test much higher among those not detected with anti-HPV-16 E6 and/or E7 antibodies compared to anti-HPV-16 VLP. Our findings support the hypothesis that E6/E7 proteins are markers of HPV-transformed tumor cells inducing an anti-E6/E7 antibody response,8 whereas among healthy patients without tumors, E6/E7 expression is an uncommon occurrence. Anti-E6/E7 antibodies also were significantly more likely than anti-HPV-16 VLP antibodies to predict HPV-related tumor characteristics: stage, tumor site, grade and nodal status (data not shown). Different findings between the serologic assays may be expected since anti-VLP presence reflects changes that occur earlier in the infection process and over a longer time period than does the presence of anti-E6/E7 oncoproteins which are markers of invasion and later events in the disease process.8, 11, 15, 30, 31
Few studies have evaluated both serologic assays in the same HNC cases and controls, yet both measures have been used to predict risk of HPV-related HNC. Differences in study findings regarding risk prediction may be due to the percentage of specific HNC sites since sites are known to have different HPV positivity rates, or due to differences in the characteristics of the assays. In a large study of anti-HPV VLP, Mork et al.32 reported that the adjusted odds among HNC cases (all sites combined) who were HPV-16 VLP seropositive was twice that of controls (12% vs. 7%; OR = 2.2). The IARC multicenter study12 also found a low frequency of anti-HPV-16 VLP but detected a significant increased risk of site-specific cancers (controls: 6%; oral cavity cases: 9%, OR = 1.5; oropharyngeal cases: 13%. OR = 3.5). The site-specific anti-HPV-16 VLP frequencies in that study contrast to the much higher prevalence rates in our data although interestingly the ORs are comparable: 27% in oral cavity, OR = 1.2; 50% in oropharynx, OR = 3.5. We detected a higher prevalence of anti-HPV-16 VLP in HNC cases and controls (34 and 22%) but similar overall risk of cancer (OR = 1.7) as did Schwartz et al. (OR = 2.3; Ref.14).
Using HPV-16 E6 and/or E7 proteins, we found that 11–65% of HNC site-specific cancer cases were seropositive compared to only 5–12% in the IARC multicenter analysis,12 and 12% in Zumbach et al.15 In contrast, Herrero et al. and Zumbach et al.12, 15 consistently found low positivity rates of anti-HPV-16 E6 and/or E7 (<1.0–3.5%) among healthy individuals without HNC cancer. Although the anti-E6/E7 positive rate in the IARC investigation was similar in controls to our population (anti-E6 and/or E7: 1.6% vs. 1.2%), the site-specific rates and associated cancer risks were much lower than ours for both cancers of the oral cavity (anti-E6 and/or E7: 4.6% vs. 11.1%; OR = 2.9 vs. OR = 4.9) and oropharynx (11.9% vs. 64.5%; OR = 9.2 versus OR=231). These differences remain to be clarified; however, it is evident that presence of anti-E6 or E7 antibodies signifies a significant risk of being detected with an HPV-related cancer.
It may be argued that seropositivity to HPV VLP primarily reflects associations with prior anogenital infections where higher rates of HPV-HR mucosotropic infections are detected in comparison to rates of viral infection in head and neck tumors; and that seropositivity does not reflect current disease. It also must be noted that some individuals do not seroconvert or maintain a detectable level of HPV antibodies over time. Nonetheless, the risk of an HPV-related HNC cancer in this study and elsewhere14 based on anti-VLP antibody presence was significantly elevated in head and neck tumors that were positive for HPV-16 DNA compared to HPV DNA negative tumors. The difference in the anti-VLP and anti-E6/E7 results is in the level of risk, with anti-E6/E7 antibody positivity conferring a 7–10 fold higher risk of an HPV-related cancer than anti-VLP antibody positivity. Although we found little difference in the HPV-16 biopsy positive/seropositive rates (61–65%) based on anti-VLP or anti-E6/E7 antibodies, it was among the tumor HPV negative cases that a difference in the frequency of seropositivity was noted for E6 or E7 (3%) in contrast to VLP positive/tumor negative cases (22%). It is possible that even though we used LCM, there may have been false negative HPV results in the tumor tissue. Nonetheless, we consistently found that those who were tumor HPV-negative were not at a significantly increased risk of detection with anti-HPV-16 antibodies compared to controls, regardless of serologic test. Furthermore, we established a significant association between HPV-16 positive oral cells and seropositivity and as we have shown previously, HPV-HR positive oral cells are significantly predictive of having an HPV-related HNC (data not shown).17 This is the first study to examine the association between these two biomarkers of HPV-associated HNC risk. Thus, a strength of our findings is in the consistency of the association among all the biomarkers.
Since it is known that HPV-related HNCs have significantly better survival rates despite the fact that they are detected primarily in late stage disease, the ability to predict an HPV-HR tumor at diagnosis may be a significant factor in determining treatment.33, 34, 35, 36, 37 Studies suggest that HPV-16 head and neck tumors are more susceptible to radiation treatment.34, 35, 36, 37 Schwartz et al.36 observed better survival among HNC cases who received radiation and who also were infected with HPV-16 (HR = 0.3, 0.1–0.8), whereas those who were HPV-negative had no survival advantage with the same treatment.
Knowledge of HPV serologic status also may be useful in influencing further diagnostic evaluation of an HPV-related cancer in asymptomatic individuals. In studies of cervical dysplasia, women with persistent HPV-16 VLP specific antibodies were at increased risk of developing cancer compared to those whose antibodies cleared.38 Anti-E6 or E7 antibody responses also have been shown to occur only shortly before diagnosis of cervical cancer30 or have a high prevalence in the cervix of women with HPV-16 infection by Hybrid Capture and diagnosed with an abnormality of the transformation zone.31 Prospective seroepidemiology studies are needed to determine whether these antibody tests are useful as an early diagnostic marker for preneoplastic or undetected early stage head and neck tumors. Currently no highly sensitive, predictive biomarkers exist for these cancers.