Serologic Evidence of Human Papillomavirus 16 and 18 Infections and Risk of Prostate Cancer¹

Previous studies have found positive associations between risk of prostate cancer and indices of sexual activity (age at first intercourse and number of sexual partners) and sexually transmitted diseases such as gonorrhea and syphilis (1). In a previous report (2) from our population-based case-control study of prostate cancer, we also observed a positive association between history of gonorrhea (OR,³ 1.50; 95% CI, 1.02–2.18) and number of lifetime female sexual partners (≥30 versus 1; OR, 2.27; 95% CI, 1.42–3.62) and the disease.

One potential etiological factor of interest for prostate cancer in light of the associations with sexual practices is exposure to HPV. HPV has been positively associated with the development of several epithelial anogenital cancers (cervical, anal, penile, and vaginal cancers) and is sexually transmitted (3). Both HPV-16 and -18 have been associated with the development of these cancers, with HPV-16 accounting for approximately 50% of such cancers in a number of studies, whereas the association of HPV-18 varied geographically and was strongest for adenocarcinoma of the cervix (4). Because HPV-16 and -18 are transmitted sexually, and sexual factors have been linked to an elevated relative risk of prostate cancer, it has been hypothesized that HPV may also be a risk factor for prostate cancer.

Initial investigations of the possible relationship between HPV and prostate cancer involved attempts to detect HPV DNA in tumor tissue from prostate cancer patients compared with prostate tissue obtained from men with benign conditions (mostly benign prostatic hyperplasia). Strickler and Goedert (1) have reviewed these studies and noted that the results are inconsistent. Only 2 of 20 investigations found an association between tissue evidence of prior HPV infection and prostate cancer. Many of the studies were limited by small sample size and used different methods for detection of HPV.

A limited number of serological studies of antibodies to HPV-16 and -18 have also yielded inconsistent results. Two nested case-control studies (5, 6) observed increased relative risks for prostate cancer associated with HPV-16 or -18 antibodies, whereas no associations were observed in three other case-control studies (7, 8, 9). To further address the question of whether HPV plays a role in prostate cancer and to follow-up on our positive findings related to sexual history, we measured serological evidence of HPV-16 and -18 exposures in the context of a population-based study of middle-aged men.

The methods of this study have been described previously (2, 10). Briefly, cases and controls were Caucasian and African-American men who resided in King County in northwestern Washington state, and they were between 40 and 64 years of age at reference date (date of diagnosis for cases and a similar assigned date for controls that approximated the distribution of the cases’ diagnosis dates). Incident prostate cancer cases were diagnosed with histologically confirmed cancer between January 1, 1993 and December 31, 1996 and were ascertained through the Seattle-Puget Sound Surveillance, Epidemiology, and End Results cancer registry. All cases <60 years of age, 100% of African-American cases ages 60–64 years, and a 75% random sample of white cases ages 60–64 years were invited to participate. Of the 917 selected eligible cases, 753 (82.1%) were interviewed, and 642 (85.3%) provided a blood sample with sufficient sera to be tested for human papillomavirus. Controls of similar age (frequency-matched on 5-year age group) were identified in King County through random digit dialing (11). During the first step of random digit dialing, complete household census information was obtained for 94%, and of the 941 men deemed to be eligible, 703 (74.7%) were interviewed. Serological samples were obtained for 570 (81.1%) of interviewed control subjects.

Standardized questionnaires, administered in person by trained male interviewers, were used to ascertain exposure information. Subjects were asked about the following: demographic and medical factors; family cancer history; tobacco and alcohol consumption; prostate cancer screening history and lifetime sexual history, including sexual orientation, lifetime number of male and female sexual partners, frequency of sexual intercourse, and condom use during specified age intervals; and episodes of sexually transmitted diseases. Dietary information was obtained from a self-administered food frequency questionnaire, which has been validated against 24-h food recalls and 4-day food diaries (12).

Clinical information was obtained from the Seattle-Puget Sound Surveillance, Epidemiology, and End Results cancer registry. The Gleason system (13) was used to classify histological grade: scores of 2–4 are low grade (well differentiated); scores of 5–7 are moderate grade; and scores of 8–10 are high grade (poorly differentiated). Summary stage of disease used the best available clinical and pathological information obtained within 4 months of diagnosis date and was defined as localized disease confined to the prostate (stage A or B), regional disease spread beyond the prostatic capsule (stage C), and metastatic disease (stage D). Clinical data were used to assess disease severity, with more aggressive prostate cancer classified as tumors of high histological grade (Gleason scores 8–10) or regional or distant stage.

The laboratory was provided with batches of serum samples containing similar numbers of cases and controls. The person performing the assays was blinded with respect to the case/control status of the specimens.

To test for serum IgG antibodies to HPV-16 and -18, a capture antibody test similar to the HPV-16 assay described previously (14) was used. These antibodies have been shown to neutralize HPV-16 in an in vitro assay (15), and the antibody has been shown to inhibit the reactivity of most human sera (16). In brief, capture monoclonal antibodies were diluted in carbonate buffer and used to coat microtiter plates. The monoclonal antibodies used were H16.V5 (1:2000) and H18.J4 (1:2000), generously provided by Dr. Neil Christensen (Hershey, PA). Nonspecific binding was blocked by the addition of PBS with 5% goat serum (Sigma, St. Louis, MO) and 0.05% Tween 20 (Fisher Scientific, Fairlawn, NJ). Capsids were produced using recombinant baculovirus and partially purified on cesium chloride gradients as described previously (17). Capsids were added at an optimum dilution, which was determined using positive control serum. Human sera that had been frozen at −70°C was thawed, diluted 1:100 in blocking buffer, and tested in triplicate in wells containing HPV-16 or quadruplicate (HPV-18 capsids) and an equal number of wells containing capture antibody but no capsids. For the HPV-18 test, each serum sample was tested in two wells with antigen and two wells without antigen, in different positions on two plates, for a total of 8 wells/serum. Bound human IgG was detected using goat antihuman alkaline phosphatase (Roche, Indianapolis, IN), and the plates were read at 405 nm after a 1-h incubation at room temperature. The ELISA values were calculated as follows: ELISA value = natural logarithm of the median absorbance of wells with capsids minus the natural logarithm of the median absorbance of the wells without capsids. Sera that had a coefficient of variation greater than 20% for wells with or without antigen were retested. If the test and the retest gave inconsistent results, the sample was tested a third time. Control sera were from a natural history study of HPV among university women (14). The cutpoint was computed as 2 SDs above the mean of a set of 29 sera from women in whom no HPV DNA was detected and who also reported no previous male sexual partners.

For additional quality control, 40 blind duplicate serum samples were sent to the laboratory and distributed across all serum batches. Based on these samples, there was 97.5% agreement with respect to HPV-16 status and 95.0% agreement with respect to HPV-18 status.

Unconditional logistic regression was used to compute ORs and 95% CIs for the risk of prostate cancer associated with HPV-16 and -18 infection (18). Polytomous logistic regression (19) was used to evaluate the association between infection with HPV-16 or -18 and clinical characteristics of the tumor (grade, stage, and aggressiveness). The statistical significance of potential interactions was evaluated on a multiplicative scale by adding an interaction term to a model containing the main effect variables.

Established and suspected prostate cancer risk factors were examined for potential confounding effects on the HPV-prostate cancer association, including age at reference date, race, family history of prostate cancer in first-degree relative(s), height, weight, body mass index, exercise, cigarette smoking, alcohol consumption, dietary fat and vegetable consumption, total calorie intake, prostate cancer screening history, and sexual factors such as number of lifetime female partners and age at first intercourse. These variables were added one at a time to the model containing HPV-16 and/or -18 and age and were considered important if the factor changed the OR associated with HPV status by more than 10%. Only age (categorical) was retained in the final models.

Analyses were performed to determine the association between serological evidence of HPV-16 or -18 infection and number of lifetime female sexual partners and age at first intercourse among controls using the Cochran-Armitage test for trend in proportions (20). In addition, the relationship between HPV infection and self-reported history of gonorrhea, syphilis, and genital warts was assessed in controls using Fisher’s exact test (21).

Table 1 gives the distribution of potential confounders in subjects from whom serum samples were obtained. Prostate cancer cases were more frequently African American, had a family history of prostate cancer in a first-degree relative, were screened for prostate cancer using a PSA test, had more female sexual partners, had an earlier age at first intercourse, and had higher caloric intakes and lower total vegetable and cruciferous vegetable intakes.

There was no association between serological evidence of HPV-16 or -18 infection and the development of prostate cancer (Table 2), although subjects who showed antibody evidence of prior infection with both subtypes were at somewhat higher risk (OR, 1.94; 95% CI, 0.59–6.36) relative to men who tested negative for both subtypes.

Further analyses focused on clinical features of the prostate cancer cases. There did not appear to be an overall difference in the risk estimates for prostate cancer associated with HPV-16 and/or HPV-18 infection according to Gleason score, extent of disease at diagnosis, or a combined measure of tumor aggressiveness (Table 3). There was a suggestion that men who tested positive for either HPV subtype had a slightly higher relative risk (OR, 1.50) of having distant stage disease, although this difference was not significant and could represent a chance finding.

Lastly, we examined sexual behavior factors according to HPV status among controls. There was no trend in HPV-16 or -18 prevalence with number of lifetime female sexual partners (P = 0.33) or age at first intercourse (P = 0.43). There was also no association between self-reported history of gonorrhea (P = 0.14) or syphilis (P = 0.36) and HPV-16 or -18 infection status. There was, however, an association with self-reported history of genital warts (P = 0.03); only 6.7% of the controls with serological evidence of infection with HPV-16 or -18 reported a history of genital warts.

We did not observe an association between serological evidence of infection with HPV-16 (OR, 1.05; 95% CI, 0.71–1.57) or HPV-18 (OR, 1.36; 95% CI, 0.69–2.69) and the development of prostate cancer. This finding differs from those of previous nested case-control studies but is consistent with other case-control studies (Table 4). Although relative risks above 2.0 were observed for HPV-16 in two nested case-control studies (5, 6), these elevations were not statistically significant. Of the studies published to date, the only statistically significant result was for HPV-18 in the study by Dillner et al. (5).

Our findings are similar to those from the largest previously reported case-control study of HPV-16 and prostate cancer by Hayes et al. (9), in which no association was observed (OR, 1.4; 95% CI, 0.7–2.8). Two other small studies also found no relationship (7, 8).

One issue to consider in assessing the differences in study results is that of persistence of HPV seropositivity over time. The Dillner et al. (5) study ascertained prostate cancer cases with serum samples collected an average of 24 years before diagnosis, and the Hisada et al. (6) study used serum that was collected an average of 26 years before diagnosis. Previous studies have shown the persistence of antibodies to HPV for at least 3 (14, 22) to 9 (23) years. By examining the prevalence of HPV-1 antibodies in adults, and assuming that most plantar wart infections occur in childhood, it appears that HPV antibodies can persist for a decade (14). Similarly, the high prevalence of HPV antibodies in a sexually transmitted disease clinic population, many of whom had initiated sexual activity decades earlier, also led us to suggest that antibodies persist for at least a few decades.

Strickler and Goedert (1) suggest that lack of persistence of HPV antibodies is responsible for the difference in results between nested case-control studies and regular case-control studies because of a declining prevalence of HPV antibodies in older populations (24), but they do not mention the possibility of cohort effects. Differences in sexual practices in different birth cohorts could also contribute to these observations.

In case-control studies of anogenital cancers, there is an excess of antibodies in cases, with stronger associations observed for in situ disease than for invasive disease. One interpretation of these results is that antibodies may wane in individuals who do not have recurrent disease. The individuals with in situ disease would likely have had productive HPV infections, i.e., expression of the L1 protein, more recently than those with invasive disease; and similarly, individuals with invasive disease may have had HPV infection more recently than the controls. We would postulate that in the nested case-control studies, the association seen between HPV antibodies and prostate cancer may result from HPV antibodies being a surrogate for sexual activity.

In our study, which is the largest to investigate the relationship between HPV-16 and -18 and prostate cancer reported to date, there was 80% power (two-tailed α level of 0.05) to detect an OR of 1.67 or higher for HPV-16 and an OR of 2.34 or higher for HPV-18. Thus, we had sufficient power to detect ORs of the magnitude seen in the earlier studies (5, 6).

Another factor to consider when interpreting results is study design. Our study and the study of Hayes et al. (9) were population-based. Other case-control and cohort studies were either hospital-based or insurance plan-based. In contrast to the study of Hayes et al. (9), we determined HPV antibody status on the majority of interviewed study participants (85.3% of cases and 81.1% of controls). Hayes et al. (9) obtained blood samples from 57% of cases and 63% of controls. In the Hayes et al. study (9), men were not eligible for blood collection if they had been treated with chemotherapy, radiation therapy, hormone-ablation drugs, or orchiectomy. Our study may have a more representative group of patients because there were no exclusions based on treatment. Most cases in our study were treated with radical prostatectomy or radiation therapy.

We did not observe any association between lifetime number of sexual partners or age at first intercourse and HPV-16 or -18 infection status among controls. However, we did find a relationship between HPV status and self-reported history of genital warts (P = 0.03), which was only reported by 2.1% of the controls.

Our results show a lower prevalence of HPV infection than what is commonly observed in studies of women. We believe that we are able to accurately detect serological evidence of HPV in men because previous studies using our methods have found elevated relative risks for HPV infection and anal cancer (25) and oral cancer (26) in men. Other investigators have observed a lower prevalence of HPV infection in men than in women (27).

There are several potential limitations to our study. Interviews were completed for 82.1% of cases and 70.2% of control subjects, and blood specimens were not obtained on all interviewed subjects. Potential selection bias could have occurred if nonrespondents were different from respondents in terms of HPV infection status. We compared characteristics of subjects who provided blood samples with those who did not, and there were no differences in the distribution of risk factors for prostate cancer between the two groups.

In conclusion, we found no evidence to support the hypothesis that HPV-16 or -18 infection is related to the development of prostate cancer among middle-aged men. Numerous previous case-control studies have suggested that a sexually transmitted agent may be related to the development of prostate cancer due to the positive associations with age at first intercourse, number of sexual partners, and other sexual risk factors (1). The extent to which these sexual behaviors reflect potential exposure to a sexually transmitted agent(s) or differences in the hormonal milieu among men with differing sexual practices remains unclear. However, our results indicate that infection with HPV-16 and/or -18 does not explain the observed associations between sexual history and risk of prostate cancer.

We thank the study subjects who gave so generously of their time to participate in this study.