Sexually Transmissible Infections and Prostate Cancer Risk

Prostate cancer is the most common cancer among American men, but its risk factors, aside from older age, African American race, and family history, are poorly understood (1). Causes for the racial disparity in prostate cancer incidence are unknown. Cultural variations in environmental and lifestyle factors, including male sexual behavior, have been suspected to play a role (2, 3).

In a meta-analysis of mostly case-control studies by Dennis et al. (4), self-reported history of sexually transmissible infections (STIs), particularly syphilis and gonorrhea, was associated with increased prostate cancer risk [odds ratio (OR), 1.4; 95% confidence interval (95% CI), 1.2-1.7], as were an increased number of sexual partners and higher frequency of sexual activity (4). Combining self-reported and laboratory data, a subsequent meta-analysis by Taylor et al. (5) found elevated prostate cancer risks associated with gonorrhea, human papillomavirus (HPV), and any STIs (OR, 1.4-1.5, all statistically significant). However, subsequent large prospective studies [with 691 (6, 7), 738 (8), 804 (9), and 2,263 (10) prostate cancer cases] found no association for self-reported history of syphilis and gonorrhea (10), no association for seropositivity of HPV-16, HPV-18, and HPV-33 (7, 9), no association (7) or a reduced risk for Chlamydia trachomatis seropositivity (OR, 0.7; ref. 8), and a reduced risk for human herpesvirus-8 (HHV-8) seropositivity (OR, 0.7; ref. 7), although one found an increased risk for Trichomonas vaginalis seropositivity (OR, 1.4; ref. 6). In tissue-based studies, the presence of HPV [summarized by Taylor et al. (5)], C. trachomatis (11, 12), HHV-8 (13), herpes simplex virus (HSV)-2 (14, 15), and cytomegalovirus (CMV; refs. 14, 16) has been reported in prostate cancer or other prostate samples. However, other studies have failed to detect these agents (5, 17-20), potentially reflecting tumor sampling artifacts, DNA degradation, PCR contamination, or even a “hit-and-run” effect (21).

These prior studies focused on one or a few STIs at a time with limited ability to address joint effects. Furthermore, more data from large prospective investigations could help confirm or refute the suggestive case-control associations. We therefore examined prostate cancer risk associated with serum antibodies against C. trachomatis, HPV-16 and -18, HSV-2, CMV, and HHV-8, and with self-reported history of syphilis and gonorrhea, using prospectively collected serum specimens and questionnaire data from the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial.

The PLCO Cancer Screening Trial is a randomized community-based study designed to evaluate the effectiveness of cancer screening tests on site-specific cancer mortality. Detailed descriptions of the Trial were provided elsewhere (22-24). Briefly, about 77,000 men and 77,000 women ages 55 to 74 years were recruited to the Trial through direct mailings, advertisements, and other means. Enrollment took place between September 1993 and July 2001 at 10 screening centers (Washington, DC; Detroit, MI; Salt Lake City, UT; Denver, CO; Honolulu, HI; Minneapolis, MN; Marshfield, WI; Pittsburgh, PA; St. Louis, MO; and Birmingham, AL). Non-Hispanic Whites and Blacks comprised 88% and 5% of the participants, respectively. Approximately 50% of the participants were randomized to the intervention arm, which for males included periodic screening tests for prostate cancer (prostate-specific antigen annually in years 0-5 and digital rectal examination annually in years 0-3). Data on general risk factors and dietary intake were collected through questionnaires, and serial blood samples were obtained for etiologic and early marker studies (25, 26). Institutional review boards at each of the participating institutions approved the PLCO protocol and participants provided written informed consent.

Prostate cancer cases in the PLCO Trial were identified through multiple mechanisms and confirmed by pathologic records. A positive screening test for prostate cancer (prostate-specific antigen >4 ng/mL or a digital rectal examination suspicious for cancer) led to referral for clinical workup, which may include a biopsy. In addition, participants reported annually by mailed questionnaires whether they had a diagnosis of cancer in the previous year and the type and date of the diagnosis. Reports from physicians and relatives were also collected when available. The National Death Index and cancer registries as available were periodically searched to enhance completeness of endpoint ascertainment. For subjects with suspected or reported prostate cancer, medical records, including clinical and pathologic data pertinent to the date of diagnosis, stage and grade (Gleason score), etc., were obtained and coded by trained abstractors.

Subjects were eligible if they had been randomized to the intervention arm, were non-Hispanic White or Black men, had a valid PLCO screen for prostate cancer (prostate-specific antigen or digital rectal examination) between October 1993 and September 2001, completed the baseline risk factor questionnaire, provided a blood sample, signed the informed consent, and reported no prior history of prostate cancer (n = 28,243; refs. 27-36). Case subjects were men with a pathologically confirmed diagnosis of prostate adenocarcinoma. Among Whites, we included only cases diagnosed ≥1 year after their first valid prostate cancer screen (n = 765). Among Blacks, we had 60 cases diagnosed at least 1 year after the screening; to increase power, we also included 43 cases diagnosed at enrollment or within 1 year of screening for a total of 103 Black cases. Cases diagnosed with stage III to IV or Gleason score of ≥7 were considered aggressive. For comparison, 1,283 men without a diagnosis of prostate cancer at the time of case diagnosis were selected as controls using incidence density sampling, frequency matched by age (55-60), ethnicity (1.2:1 for Whites and 4:1 for Blacks), time since initial screening (1-year interval), and year of blood draw. All subjects were followed from their first valid prostate cancer screen to first occurrence of prostate cancer, loss-to-follow-up, death, or September 30, 2001, whichever came first. Both cases and controls had an average follow-up of 1.6 years with mean of 1.4 post-enrollment screening visits before diagnosis/selection.

At study entry, participants completed a risk factor questionnaire reporting whether a doctor had ever told them that they had syphilis or gonorrhea. We compared these cases both with controls in the nested case-control study and to all at-risk PLCO participants in the intervention arm. Other information collected in the baseline questionnaire included marital status, education, weight, height, use of nonsteroid anti-inflammatory drugs (NSAIDs), smoking behavior, family history of prostate cancer, and history of enlarged prostate or benign prostatic hypertrophy, prostate biopsy, and vasectomy.

Serum samples were collected from intervention arm participants at study entry, processed within 2 h of the blood draw, and stored at −70°C or lower. Aliquots of 100 μL serum were sent to five separate laboratories for assays, standardized with predetermined cutoff points, categorizing the infection status as positive, negative, or equivocal. Laboratory personnel were blinded to case-control status. Thirty-three masked replicate samples from four individuals were tested to assure satisfactory performance for each serologic assay before proceeding with the study samples. All assays, except HHV-8, were done in duplicates, and a third test was done for discordant calls; for each batch, internal positive and negative quality-control samples were tested to ensure quality performance of the assays. For HHV-8, three positive and three negative controls were run on each plate composed of 96 samples; the control values were required to be within a narrow range or the plate was repeated. We also interspersed 96 pairs of blinded duplicate samples (from 48 Whites and 48 Blacks) among the study samples to evaluate assay reproducibility. For 50 of the controls, we added a second sample collected 1 year after the baseline serum, placed immediately adjacent to a subject's corresponding baseline sample to test intraindividual variability.

C. trachomatis. Using a commercial ELISA (Medac), we measured IgA and IgG antibodies to a synthetic peptide derived from the major outer membrane protein of C. trachomatis that is common among all serotypes; sensitivity and specificity of these assays have been reported to range from 79% to 88% and 82% to 89%, respectively (8, 55). Also, we tested subjects seropositive for IgA or IgG antibodies to major outer membrane protein for IgG to C. trachomatis heat shock protein 60 as a presumptive indicator of repeated or persistent infection (56). An absorbance cutoff value of >1.1 indicated IgG and IgA antibodies to C. trachomatis and IgG antibodies to heat shock protein 60. HPV-16 and -18: We used a virus-like particle–based enzyme immunoassay to measure serum IgG against HPV-16 and -18 capsids as described previously (57). The virus-like particles were produced in the Trichoplusia ni (High-Five) cells (Invitrogen) from a recombinant baculovirus that expresses the L1 gene of HPV-16 or HPV-18 and purified by density-gradient ultracentrifugation as described previously (58). The cutoff for seropositivity was defined as an absorbance value greater than the mean ± 3 SDs of serum samples from children, ages 1 and 5 years, after exclusion of outliers. HSV-2: HSV-2 IgG antibodies were measured using a solid-phase enzymatic immunodot assay with a purified glycoprotein specific for HSV-2 as the antigen (59, 60). The assay has been tested widely for a large number of samples collected in the National Health and Nutrition Examination Survey II and III with sensitivity >98% and specificity >99% (59). CMV: CMV IgG antibodies were measured using commercially available Microparticle Enzyme Immunoassay (AxSYM CMV IgG assay; Abbott Laboratories) and the manufacturer-recommended cutoff for seropositivity. The presence of at least 15 antibody units per mL of sample is indicative of past or current infection with CMV. Results of ≥10 but <15 antibody units per mL are considered equivocal. The relative sensitivity was 99.7% and specificity was 100% according to the manufacturer (61, 62). HHV-8: Antibodies to HHV-8 were measured by ELISA to detect IgG against the HHV-8 K8.1 structural glycoprotein expressed during lytic viral replication as described previously (63, 64). The cutoff for seropositivity was calculated as the mean of the three negative controls on the plate plus 0.75 to correct for plate-to-plate variability.

We calculated modified κ coefficients for agreement between the blinded duplicates as suggested (65). To evaluate the correlation between the STIs, we calculated Spearman correlation coefficients and P values from χ² tests. Conditional logistic regression, conditioning on the four matching factors of age, ethnicity, time since initial screening, and year of blood draw, was used to estimate OR and 95% CI for the association with each STIs, including analysis on any STI. Additional adjustment of potential confounding factors, including family history of prostate cancer, education, NSAID use, smoking, and marital status, yielded virtually no change in results; thus, results were not presented herein. We also did unconditional logistic regression models and the results were very similar (data not shown). The small numbers of serologic tests classified as equivocal were analyzed as positives; in sensitivity analyses, exclusion of these samples yielded very similar results (data not shown). We constructed a summary variable for total number of past infections (range, 0-7) from the five serologically identified STIs (C. trachomatis by IgA or IgG, HPV-16/18, HSV-2, CMV, and HHV-8) and two self-reported histories (syphilis and gonorrhea). We used this variable to perform trend tests for the associations with prostate cancer risk in logistic regression models. To explore potential risk modifications by suspected prostate cancer risk factors or inflammation-related factors, we conducted stratified analyses using questionnaire data (e.g., education, marital status, family history of prostate cancer, history of hepatitis infection, and use of NSAID such as aspirin- or ibuprofen-containing products), clinical information (e.g., stage, Gleason score, and age at diagnosis), and genotypes (e.g., single nucleotide polymorphisms in RNASEL, NOS2A, NOS3, SOD1, SOD2, SOD3, IL1B, IL6, IL8, IL10, TNF, PTGS2, PPARD, and PPARG genes; refs. 28-30, 32, 34). To analyze associations with self-reported history of syphilis and gonorrhea among all at-risk PLCO participants in the intervention arm, we used Cox proportional hazards regression to estimate relative risk and 95% CI, adjusting for age, center, education, smoking status, NSAID use, and family history of prostate cancer. Two-sided P-values < 0.05 were considered statistically significant.

Table 1 summarizes the quality-control data for each serologic test. For both reproducibility of blinded duplicate samples (n = 96 pairs) and intraindividual agreement of samples collected 1 year apart (n = 50 pairs), κ values ≥0.7 were obtained for all assays, except the reproducibility of HPV-18 (κ = 0.6).

Prevalences of the 7 STIs were weakly correlated among controls, similarly for Whites and Blacks (Table 2). C. trachomatis IgA and IgG antibody positivities were significantly correlated and both were positively correlated with HSV-2 seropositivity; correlation with self-reported history of syphilis and gonorrhea was slightly stronger for C. trachomatis IgA than IgG. Likewise, seropositivity for HPV-16 and HPV-18 were significantly correlated and both were positively correlated with HSV-2 seropositivity; correlation with self-reported history of gonorrhea was slightly stronger for HPV-16 than HPV-18. Seroprevalences of CMV and HHV-8 appeared to be less correlated with those of other infections.

Cases and controls of both racial groups were similar with respect to body mass index, smoking behavior, use of NSAIDs (aspirin or ibuprofen), history of prostatitis, and prior vasectomy (Table 3). For both racial groups, cases were significantly more likely than controls to have a history of benign prostatic hypertrophy or prostate biopsy and had higher prostate-specific antigen levels and prevalence of prostate enlargement detected at study entry. Among Whites, a higher percentage of cases than controls reported a family history of prostate cancer in fathers or brothers, and cases were slightly more frequent than controls to report a college degree or higher education; these characteristics did not differ significantly among Blacks.

Among controls, Blacks had higher seroprevalences of all studied STIs, except HHV-8 (Table 4). Among Whites, none of the infections individually was significantly associated with prostate cancer risk. Among Blacks, IgA, but not IgG, seropositivity to C. trachomatis was associated with increased risk of prostate cancer; this association was slightly stronger (OR, 2.3; 95% CI, 1.2-4.1) after exclusion of men with weak-positive results. Among subjects seropositive for IgA or IgG antibodies to C. trachomatis, ∼40% of Whites and 74% of Blacks also had antibodies to C. trachomatis heat shock protein 60, but these percentages did not differ between cases and controls.

Among Whites, slightly more cases than controls were positive (seroprevalence or questionnaire-based) for at least one STI; positivity for any STI was associated with a 1.3-fold increased risk of prostate cancer (95% CI, 1.0-1.6; P = 0.05; Table 4). The dose-response trend for number of STIs was suggestive albeit nonsignificant. Among Blacks, nearly all subjects were positive for at least one STI; compared with men with none or one STI as the reference group, those with ≥2 prior STIs had a nonsignificant elevation in prostate cancer risk. Although limited by small numbers, we observed no obvious differential in risk patterns between Black cases who were diagnosed at least 1 year versus within 1 year after the serum collection.

Further exploration in White men showed no differences in risk patterns for subgroups defined by age at diagnosis, education, marital status, history of hepatitis infection, use of NSAIDs such as aspirin- or ibuprofen-containing products, genotype of selected variants in inflammation genes (P_Interaction < 0.05 for only 2 of the 34 variants examined), or aggressive (stage III-IV or Gleason score ≥7) versus nonaggressive cancer. Prior STI was associated with prostate cancer among subjects without a family history of prostate cancer (OR, 1.4; 95% CI, 1.1-1.8) but not among those with a family history (OR, 0.8; 95% CI, 0.3-2.3; P_Interaction = 0.04). Sample size for Black men was too small for stratified analysis.

In addition, compared with all at-risk PLCO participants in the intervention arm, prostate cancer cases had no excess of either syphilis (Whites: relative risk, 0.4; 95% CI, 0.1-1.3; Blacks: relative risk, 0.7; 95% CI, 0.3-1.8) or gonorrhea (Whites: relative risk, 1.0; 95% CI, 0.8-1.3; Blacks: relative risk, 0.9; 95% CI, 0.6-1.4), similar to the comparison with the nested controls (Table 4).

Circulating antibodies against infectious agents and self-reported history of syphilis and gonorrhea represent an individual's cumulative lifetime exposure or past infections and thus are particularly suited for risk evaluation of cancer with a long latent disease process. Our analysis is among the first to examine the effects of many STIs simultaneously in a large prospective study of prostate cancer risk. We found no strong association between prior evidence of specific STIs and prostate cancer, but having had any STI was associated with a modestly increased risk among Whites. In the smaller number of Black men studied, we found an increased risk associated with IgA antibody to C. trachomatis and a nonsignificant elevation with two or more prior STIs.

Race-specific seroprevalence estimates for some STIs were available from nationally representative populations, although further specification by gender and age were often not provided. Seroprevalences for Whites in this study were similar to National Health and Nutrition Examination Survey for CMV (78.2%; ref. 66), HPV-16 (8.0%; ref. 67), and HSV-2 (13.7%; ref. 68), whereas the Black controls had higher seroprevalences for the latter two agents [90.2% (66), 9.6% (67), and 40.3% (68), respectively, in the National Health and Nutrition Examination Survey]. Seroprevalence reports for HHV-8 were highly variable (69); our measure of HHV-8 K8.1 was higher for both Blacks and Whites than our laboratory had reported previously from National Health and Nutrition Examination Survey (2.1% and 1.5%, respectively; ref. 70), although it was similar to their alternative National Health and Nutrition Examination Survey estimate (7.1%, both races) based on a less-specific cutoff (70) as used in this study. For C. trachomatis IgG and HPV-18, our results were similar to findings from large population-based case-control studies of Whites [11% (8) and 3.8% (9), respectively]; national survey data were unavailable for Black men.

Compared with other STIs, characterization of syphilis and gonorrhea is relatively achievable through self-reported survey because these conditions are typically symptomatic in men and routinely investigated in clinical settings. The frequencies of self-reported history of syphilis and gonorrhea among our control subjects were similar to self-reports in two large U.S. studies [0.1% (71) and 0.2% (10) in Whites and 2% in Blacks (71)] and the seroprevalence of syphilis (0.6%, mixed races) in the 2001 to 2002 National Health and Nutrition Examination Survey.⁸

Although both serologic measures and questionnaire-based self-reports may have error, such errors should be nondifferential between cases and controls, as materials and data were prospectively collected and laboratory personnel were blinded to case-control status. To the extent that nondifferential misclassification occurred, our results may be biased toward the null.

Meta-analyses (4, 5) showed overall modest associations of prostate cancer with self-reported syphilis (OR, 1.4-2.3), gonorrhea (OR, 1.3-1.4), and history of any STIs (OR, 1.4-1.5). These analyses were based predominantly on case-control data for Caucasians; studies of African Americans, serologic measures, and prospective data have been lacking. A large population-based case-control study of prostate cancer in U.S. Blacks (479 cases) and Whites (502 cases) found a 1.6-fold risk associated with self-reported history of gonorrhea or syphilis and, in a subgroup with available serum (125 Black cases, 146 White cases), a 1.8-fold risk association with antibodies to syphilis (71); patterns of risk were similar for Whites and Blacks (71). The prospective Health Professionals' Follow-up Study (10), not included in either of these meta-analyses, investigated White men with a low prevalence of self-reported history of syphilis and gonorrhea and found no association of either with prostate cancer risk.

C. trachomatis is the most common bacterial STI (72) and may cause chronic persistent infections. In sexually active men, it is the main cause of nongonococcal urethritis and has been suspected as a probable cause of prostatitis (12). C. trachomatis IgG antibodies were not associated with prostate cancer in our study or in another prospective study in U.S. Whites (7), and an inverse association was reported in a Scandinavian study (8). However, our finding of a significant association of C. trachomatis IgA seropositivity with prostate cancer risk among Blacks is suggestive. Further investigation is warranted to illuminate whether the observed racial disparity was due to differences in seroprevalence, susceptibility to infection or cancer, other differences between the racial groups, or occurred by chance.

HPV and the herpesviruses HSV-2, CMV, and HHV-8 are all sexually transmissible, although nonsexual routes have also been established for CMV and suggested for HHV-8 (70, 40-43). Seroepidemiologic associations with prostate cancer have been mixed. A meta-analysis of eight retrospective and two prospective studies found a significant 1.4-fold increased risk of prostate cancer with HPV seropositivity (5), but two additional reports not included in the meta-analysis found no association with antibodies to HPV-16, HPV-18, or HPV-33 (7, 9). HSV-2 seropositivity was not associated with prostate cancer in a subsample of the Nordic nested case-control study (165 cases; ref. 45). Using small convenience samples, findings have been mixed for associations of CMV infection and prostate cancer risk (16, 46, 47). Findings also were mixed for HHV-8 seropositivity (48, 64), with the largest reported study (691 cases) showing a modest inverse association with antibodies to HHV-8 lytic antigens (7).

Other infections also were linked with prostate cancer risk. Seropositivity for T. vaginalis, a relatively prevalent but underrecognized STI, increased prostate cancer risk in the Health Professionals' Follow-up Study (6). A novel retrovirus named XMRV was discovered in prostate cancer tissues from men carrying genetic variants of RNASEL (49). Four or more years use of tetracycline, an antibiotic used to treat severe acne, was found to be associated with increased prostate cancer risk (50).

A potential explanation for these disparate associations is that the host response of infection, specifically inflammation, represents a final common pathway for prostate cancer risk. Prostate cancer tissues frequently contain activated inflammatory cells, including the proliferative inflammatory atrophy lesions thought to be precursors of early cancer development, and may have somatic mutations in inflammation-related markers (51, 52). Genetic and circulating markers of inflammation and host response to infection have been variably shown to increase prostate cancer risk, whereas intake of NSAIDs and antioxidants has been protective (53, 54). These considerations implicate chronic inflammation as a potential mechanism of prostate carcinogenesis. Alternatively, the diffuse associations with the various infections may be due to correlation with a true causal factor not directly measured.

Our study has several notable features that overcome important limitations of previous investigations. Cases and controls were identified from the same source population, which was screened for prostate cancer by a standardized procedure. Data and specimens were collected prospectively before diagnosis and were used to assess a large number of STIs simultaneously. However, prevalences of the individual infections were relatively low in our White participants and we had few Black participants, which limited our statistical power. We attempted to address this deficiency by including Black cases diagnosed within 1 year of blood draw. Potentially, these cases may be more slowly progressing or less clinically aggressive than cases arising at least 1 year after the initial screen; nevertheless, risk patterns did not grossly differ between these two groups of Black cases. White participants with at least one past infection had modestly increased prostate cancer risk; this pattern was not replicated among our Black participants, but very few had no past infections.

In summary, this large prospective study found no consistent association between history of individual STIs and prostate cancer risk. Our novel finding of an increased risk with C. trachomatis IgA antibodies was restricted to Blacks and requires confirmation. By studying many STIs simultaneously, we found a modest risk associated with any prior STI among Whites. The modest associations and lack of specificity parallel the variable findings from previous reports and may suggest involvement of a correlated factor or shared underlying response (e.g., inflammation) not directly measured. Further studies in high-risk populations would help advance our understanding of the role of STIs in the etiology of prostate cancer.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Drs. Christine Berg and Philip Prorok (Division of Cancer Prevention, National Cancer Institute), the Screening Center investigators and staff of the PLCO Cancer Screening Trial, Tom Riley and staff (Information Management Services), Barbara O'Brien and staff (Westat), the staff of the serologic testing laboratories, and the study participants for contributions to making this study possible.