A case-cohort study of human herpesvirus 8 seropositivity and incident prostate cancer in Tobago

Background We previously reported a cross-sectional association between the presence of human herpesvirus 8 (HHV-8) serum antibodies and screen-detected prostate cancer in men living in Tobago. In the same study population, we examined the association between HHV-8 seropositivity and incident prostate cancer discovered at later screenings. Methods In 40-81 year-old men without prostate cancer discovered at initial digital rectal examination (DRE) and prostate-specific antigen (PSA) screening, a case-cohort design measured the association between baseline HHV-8 seropositivity (modified immunofluorescence assay for antibodies against HHV-8 lytic antigens) and incident prostate cancer detected at DRE and PSA screenings three or five years later. Results Analyses included 486 unique individuals, 96 incident prostate cancer cases, and 415 randomly selected subjects representing an at-risk cohort. By design, the random sub-cohort contained 25 incident prostate cancer cases. In the sub-cohort, the frequency of HHV-8 seropositivity increased across age groupings (40-49 years: 3.5%, 50-59 years: 13.6%, and ≥ 60 years: 22.9%). HHV-8 seropositivity was higher in men with elevated (≥ 4.0 ng/mL) than men with non-elevated PSA at initial screening (30.4% vs. 9.9% seropositive; crude odds ratio (OR) 3.96, 95% confidence interval (CI) 1.53-10.2; age-adjusted OR 2.42, 95% CI 0.91-6.47). HHV-8 seropositivity did not increase incident prostate cancer risk (age-adjusted hazard ratio (HR) 0.88, 95% CI 0.46-1.69). Conclusions Case-cohort analysis did not identify association between HHV-8 seropositivity and incident prostate cancer. However, analyses uncovered possible association between HHV-8 and PSA (a marker of prostate inflammation). Co-occurrence of HHV-8 seropositivity and PSA elevation may explain cross-sectional association between HHV-8 and PSA screen-detected prostate cancer.

HHV-8, a DNA virus, causes Kaposi's sarcoma and primary effusion lymphoma. In a high prostate cancer risk cohort of African-Caribbean men living on Tobago [24,25], we found an association between HHV-8 seropositivity and prostate cancer discovered as a result of an initial prostate cancer screening (odds ratio [OR] 2.24, 95% confidence interval [CI] 1.29-3.90) [9]. Four studies completed later in other population settings could not confirm an association between HHV-8 and prostate cancer [6,11,26,27]. Therefore, our current study re-examines this association in our Tobago study population, through consideration of the association between HHV-8 seropositivity and prostate cancer discovered, not as a result of the initial screening, but later as a result of subsequent screenings.

Study Population
The Tobago Prostate Survey is an ongoing populationbased longitudinal study of prostate cancer screening, as well as risk, in ≥ 40 year-old men living in Tobago [24]. Tobago is a small Caribbean island, 7 by 26 miles in size, with 8078 40-79 year-old men, according to a 2000 census [28]. The population as a whole is 89% African or Black and 7% mixed heritage by nationality or ethnicity [28]. Identification of study participants occurred through the agency of posters, flyers, public service announcements, public presentations, healthcare workers, private physicians, and word of mouth [24] Although an open cohort, this report included only men screened at Wave 1 and subsequently rescreened at Waves 2 and/or 3. Study procedures included risk factor questionnaires, blood collections, and prostate cancer screening examinations, with prostate specific antigen (PSA) serum concentrations ≥ 4 ng/mL or abnormal digital rectal examinations (DRE) prompting referral for ultrasound-guided trans-rectal prostate biopsy [24].
Wave 1 enrolled 3264 40-81 year-old men (97% selfreporting African descent). The current study excluded men missing Wave 1 PSA (n = 283), men with Wave 1 PSA ≥ 4.0 ng/mL not followed by prostate biopsy (n = 104), and men with prostate cancer detected at Wave 1 (n = 330), thereby leaving 2547 men at risk for prostate cancer at Wave 2 or Wave 3 ( Figure 1). The study design excluded 756 at-risk men, including 633 at-risk men without subsequent PSA at either Wave 2 or Wave 3, 108 men with a Wave 2 or Wave 3 PSA ≥ 4.0 ng/mL not followed by biopsy, and 15 men with a prostate cancer negative Wave 2 biopsy, but no Wave 3 biopsy for Wave 3 PSA ≥ 4.0 ng/mL associated with ≥ 1.0 ng/mL PSA increase between Waves 2 and 3 ( Figure 1). In the remaining 1791 at-risk men, a Wave 2 or Wave 3 biopsy completed before study closure (August 15, 2007) detected prostate cancer in 109 ( Figure 1).
To measure the association between Wave 1 HHV-8 seropositivity and prostate cancer detection from a Wave 2 or Wave 3 screening, we used a case-cohort study design that compared prostate cancer cases at Wave 2 or Wave 3 (n = 109; Gleason 6 -49%, Gleason 7 -45%, Gleason 8 or 9 -6%; pre-diagnostic screening PSA, mean 9.5 ng/mL, median 4.4 ng/mL) against a control group constructed as a simple random sample (n = 442) of the 1791 at-risk men screened for prostate cancer at Wave 2 or Wave 3. We used a case-cohort design because research costs prohibited determination of the HHV-8 status of all 1791 men in the at-risk group. As a result of the random selection procedure, 27 case men with prostate cancer at Wave 2 or Wave 3 entered the sub-cohort and contributed data as controls ( Figure 1). The 442 and 1349 randomly selected and excluded men were statistically similar with respect to age, education, marital status, prostate cancer family history, history of smoking, personal history of cancer, history of benign prostatic hypertrophy, and Wave 1 PSA and DRE results. The 442 men selected for the subcohort survived a median 4.9 years (5 th -95 th percentile 1.7-6.8 years) between Wave 1 and the last complete post-Wave 1 visit. For the 109 cases, a median 7.2 months (25 th -75 th percentile 2.5-13.8 months) elapsed between pre-diagnostic screening and confirmatory biopsy.
Study participants signed an informed consent approved by the Institutional Review Boards of the Tobago Division of Health and Social Services and the University of Pittsburgh.

Laboratory Methods
Laboratory assays used frozen serum samples (thawed once and never re-frozen) stored temporarily in a -20°C freezer at the Tobago Health Studies office in Scarborough, Tobago, and stored later in a -80°C freezer at the University of Pittsburgh, Department of Epidemiology. PSA measurements used either Abbott Diagnostics AxSYM ® or Siemens Healthcare Diagnostics ADVIA Centaur ® immunoassays. To detect serum antibodies against HHV-8 lytic antigens, an indirect immunofluorescence assay, as described elsewhere [29], used BCBL-1 cells containing the HHV-8 genome with a modified Rta gene inducible by doxycyclin [30]. Targeting fixed and permeabilized B cells that have been induced to replicate HHV-8, this assay potentially identifies any of the lytic proteins involved in HHV-8 replication. A single reader (FJJ), blinded to samples' prostate cancer case status, examined microscopic slides for fluorescence. A positive assay result required specific fluorescence at a 1:100 dilution. Each assay run included prostate cancer case and non-case subject sera and known HHV-8 positive and negative control sera. We tested each sample in duplicate on separate days with disagreements resolved by retesting on a third day. Estimates of assay sensitivity and specificity range between 53.4-89.9% and 96.6-97.5%, respectively [31]. Kappa agreement between first and second assay results was 0.76 (95% CI 0.71-0.81).

Data Analysis
We used the chi-square test to evaluate 1) the statistical significance of differences between at-risk men with incomplete vs. complete follow-up with respect to baseline factors such as age, education, marital status, family history of prostate cancer, history of smoking, cancer, and benign prostatic hypertrophy, and results from PSA and DRE screening and 2) the statistical significance of differences in the sub-cohort rates of HHV-8 seropositivity with respect to the same baseline factors. In analyses restricted to sub-cohort members, logistic regression estimated odds ratios to express strengths of association between Wave 1 PSA elevation and Wave 1 HHV-8 seropositivity, two factors determined at the same point in time. We used Cox proportional hazards models (Breslow weighted denominator method) for case-cohort designs to estimate hazard ratio [HR] measures of unadjusted and age-adjusted association between HHV-8 seropositivity measured at baseline (Wave 1) and prostate cancer detected later in time, at Wave 2 or Wave 3 [32]. These models compare prostate cancer cases detected at one or the other points in time (Wave 2 or Wave 3) with the appropriate risk set, constructed from sub-cohort members disease-free and available at Wave 2 or Wave 3. For prostate cancer cases in the sub-cohort, Cox models started follow-up at the Wave 1 screening date and censored follow-up at the Wave 2 or Wave 3 screening date that led to prostate cancer diagnosis. For non-cases in the sub-cohort, Cox models started follow-up at the Wave 1 screening date and censored follow-up at the date of the last Screened for incident prostate cancer n=1791 Anti-HHV-8 assay completed, n=96 Anti-HHV-8 assay completed, n=415 (includes n=25 with incident prostate cancer)  Table 1 summarizes the characteristics of all 3264 40-81

Results
year-old men enrolled at Wave 1. The current study excluded 717 men missing a Wave 1 PSA, men with Wave 1 PSA ≥ 4.0 ng/mL not followed by prostate biopsy, and men with prostate cancer detected at Wave 1, leaving 2547 screen-negative men at risk for prostate cancer at Wave 2 or Wave 3 ( Figure 1). Table 1 compares these 2547 at-risk men with the 717 men who were either inadequately screened at Wave 1 or discovered to have prostate cancer at Wave 1. At-risk men were younger and better educated (Table 1). At-risk men less often reported a history of smoking, cancer, or benign prostatic hypertrophy (Table 1). PSA values were lower and DRE results positive less often in at-risk men, as expected, since the group not at risk included men with prostate cancer detected as direct result of PSA elevation or DRE abnormality ( Table 1). The study design excluded 756 men from the group of 2547 at-risk men eligible for follow-up ( Figure 1). As shown in Table 1  of 109) men in the case group (Gleason 6 -49%, Gleason 7 -45%, Gleason 8 or 9 -6%; pre-diagnostic PSA, mean 10.0 ng/mL and median 4.4 ng/mL). Table 2 summarizes the baseline characteristics for these men with non-missing HHV-8. In addition, Table 2 compares sub-cohort rates of HHV-8 seropositivity according to the same baseline characteristics. Referenced against the subcohort, characteristics of the case group included older age, less frequent smoking history, more frequent benign prostatic hypertrophy history, more frequently elevated Wave 1 PSA (27.1% vs. 5.5% PSA ≥ 4 ng/mL), and more frequently positive Wave 1 prostate cancer screening (43.8% vs. 16.9% DRE or PSA positive). Case and subcohort HHV-8 seropositivity rates were 17.7% and 11.1%, respectively. When compared with 40-49 year-old subcohort men (3.5% HHV-8 seropositive), HHV-8 seropositivity was higher in 50-59 year-old sub-cohort men (13.6% HHV-8 seropositive) and higher yet in ≥ 60 yearold sub-cohort men (22.9% HHV-8 seropositive). HHV-8 seropositivity rates were lower in sub-cohort men with a history of smoking than those without (7.1% vs. 13.9%) and higher in sub-cohort men with a history of benign prostatic hypertrophy than those without (24.0% vs. 10.5%). HHV-8 seropositivity increased with Wave 1 PSA (7.6%, 12.1%, 12.7%, and 30.4% for PSA 0.0-0.9, 1.0-1.9, 2.0-3.9, and ≥ 4 ng/mL, respectively). HHV-8 seropositivity was higher in men with elevated (≥ 4.0 ng/mL) than men with non-elevated PSA (30.4% vs. 9.9% seropositive; crude OR 3.96, 95% CI 1.53-10.2; age-adjusted OR 2.42, 95% CI 0.91-6.47; data not shown). HHV-8 seropositivity was higher in sub-cohort men with a positive than in men with a negative Wave 1 prostate cancer screen result (20.0% vs. 9.3% seropositive). Table 3 compares Wave 1 HHV-8 seropositivity between the case and sub-cohort groups, according to age and Wave 1 PSA. Age-specific HHV-8 seropositivity rates were lower in case than sub-cohort men (40-49 years: 0.0% vs. 3.5% and 50-59 years: 10.8% vs. 13.6%), except in the oldest age group (≥ 60 years: 27.7% vs. 22.9%). In men with a non-elevated (< 4 ng/mL) Wave 1 PSA, the HHV-8 seropositivity rate was higher in the case group (17.1% vs. 9.9%). In men with an elevated (≥ 4 ng/mL) Wave 1 PSA, however, the HHV-8 seropositivity rate was lower in the case group (19.2% vs. 30.4%). In the two age sub-groups with appreciable HHV-8 seropositivity, age-specific HHV-8 seropositivity rates were not consistently higher or lower in case than sub-cohort men with non-elevated Wave 1 PSA (50-59 years: 11.5% vs. 13.7% and ≥ 60 years: 28.1% vs. 20.4%) and consistently lower in case than sub-cohort men with elevated Wave 1 PSA (50-59 years: 9.1% vs. 12.5% and ≥ 60 years: 26.7% vs. 41.7%). Table 4 shows associations, unadjusted and ageadjusted, between Wave 1 HHV-8 seropositivity and prostate cancer at Wave 2 or Wave 3, overall and in sub-groups defined by Wave 1 screen results. Though not statistically significant, HR point estimates indicate lower prostate cancer risk in HHV-8 seropositive men, overall (age-adjusted HR 0.88, 95% CI 0.46-1.69) and in HHV-8 seropositive men with elevated Wave 1 PSA (age-adjusted HR 0.39, 95% CI 0.10-1.63), and equivalent risk in HHV-8 seropositive men with non-elevated Wave 1 PSA (age-adjusted HR 1.03, 95% CI 0.49-2.16). In men eligible for prostate biopsy at Wave 1 (DRE or PSA positive), HHV-8 seropositivity reduced risk (ageadjusted HR 0.59, 95% CI 0.18-1.91) to a statistically insignificant level. In perhaps the most meaningful subgroup, men not eligible for prostate biopsy at Wave 1 (DRE not positive and PSA < 4.0 ng/mL), analyses supplied no evidence of association between seropositivity and prostate cancer risk (age-adjusted HR: 1.02, 95% CI 0.44-2.39). In men with non-elevated Wave 1 PSA, single-year-of-age-adjusted (continuous) risk estimates were HR 0.80 (95% CI 0.19-3.34) and HR 1.27 (95% CI 0.50-3.25) for the 50-59 and ≥ 60 year-old men, respectively (data not shown). Using all (96 case and 415 sub-cohort) men or only ≥ 45 year-old (95 case and 312 sub-cohort) men made no meaningful difference in the age-adjusted risk estimates (data not shown).

Discussion
Our previous study used an immunofluorescence assay to measure HHV-8 antibodies in 138 prostate cancer cases and in 140 age-matched controls [9]. HHV-8 seropositivity was significantly more frequent in cases than controls (39.9% vs. 22.9%, OR 2.24, 95% CI 1.29-3.90) [9]. Our previous study compared Wave 1 screendetected (DRE positive and/or PSA elevated) prostate cancer cases with DRE negative and PSA non-elevated controls. In the same Tobago study population, using a  Table 4). A positive association between HHV-8 seropositivity and prevalent prostate cancer in a cross-sectional study [9] and an inverse (though not statistically significant) association in a prospective study, an inverse association most evident in men sent for biopsy (e.g., PSA ≥ 4 ng/ mL, age-adjusted HR 0.39, 95% CI 0.10-1.63; Table 4), lead to the following speculation. HHV-8 may associate with factors, such as elevated PSA, that prompt biopsy and subsequent recognition of prostate cancer. In effect, HHV-8 may segregate men with manifest and emergent prostate cancer into two groups, HHV-8 seropositive prostate cancer detected immediately and HHV-8 seronegative prostate cancer detected later. This selection bias may explain opposing positive and negative HHV-8 associations seen with prevalent and incident prostate cancer, respectively. A similar selection bias may explain inverse associations between HHV-8 and prostate cancer observed in other prospective studies, as described below.
Four comparative studies of HHV-8 and prostate cancer have appeared [6,11,26,27] since our 2004 publication [9]. In a prospective study from Finland, ELISA detected serum antibodies against the HHV-8 ORF65 protein in 3 (1.8%) of 163 men with incident prostate cancer and in 7 (2.4%) of 288 age-matched men without cancer (OR 0.74, 95% CI 0.19-2.88; [27]). In a U.S. population-based case-control study, the immunofluorescence assay detected serum antibodies against HHV-8 lytic antigens less often in cases than controls (95 African-American cases and 75 controls: OR 0.56, 95% CI 0.28-1.14; 104 white cases and 80 controls: OR 0.71, 95% CI 0.36-1.43; [26]). In a study of 691 individually matched case-control pairs nested within the U.S. Health Professional Follow-up Study, the immunofluorescence assay detected plasma antibodies against lytic antigens less often in men diagnosed with prostate cancer, on average, 3.1 years later (OR 0.70, 95% CI 0.52-0.95; [6]). Finally, in the Prostate, Lung, Colorectal, and Ovarian (PLCO) Cancer Screening Trial, ELISA detected IgG antibodies against the HHV-8 K8.1 structural protein in study entry serum samples from 103 (13.5%) of 765 and 103 (11.3%) of 915 white prostate cancer cases  Analyses restricted to the sub-cohort showed strong association 1) between HHV-8 seropositivity and increasing age, a result also seen in Tobago women [33] and many other populations [34,35], and 2) between HHV-8 seropositivity and PSA elevation ≥ 4.0 ng/mL. Though not statistically significant (p = 0.17), ageadjusted geometric mean Wave 1 PSA was 18% higher in HHV-8 seropositive than seronegative sub-cohort men (data not shown). The age-adjusted odds of HHV-8 seropositivity was more than two-fold higher in subcohort men with elevated PSA than men with non-elevated PSA. Personal or environmental factors related to HHV-8 exposure or immune function may explain the age association with HHV-8 [33]. Accepting PSA as a marker of prostate inflammation, we postulate that the association between HHV-8 seropositivity and elevated PSA signifies either the effects of HHV-8 infection on prostate inflammation [19] or the effects of prostate inflammation on HHV-8 reactivation. PSA elevation has been observed in relation to other infectious disease agents [36,37].
Study strengths include unique population and setting (predominantly African ancestry Tobago residents [38]) and a control group large enough to estimate age-specific HHV-8 seroprevalence rates with acceptable precision. Study limitations include unavoidable misclassification according to prostate cancer outcome. DRE and PSA invariably miss instances of biopsy detectable prostate cancer. The Prostate Cancer Prevention Trial, for example, observed a 15% prostate cancer biopsy prevalence in men with PSA ≤4 ng/mL [39]. In addition, our study can not define the prostate cancer risk experience of men who did not return for repeat screening. Follow-up intervals much longer than our five-year interval between initial and final screening may be needed to detect a prostate cancer effect from any chronic inflammation caused by HHV-8 infection. Also, HHV-8 may cause inflammation and prostate cancer only in a relatively small genetically susceptible subgroup. Finally, a small case count limits, especially in sub-groups, the precision of our risk estimates. For example, in men not eligible for prostate biopsy at Wave 1, the 95% confidence interval embraced both 50% lower and 200% higher prostate cancer risks in relation to HHV-8 seropositivity.