Correspondence to: Manolis Kogevinas, Centre for Research in Environmental Epidemiology (CREAL), Doctor Aiguader 88, 08003 Barcelona, Spain, Tel.: +34 93 214 73 32, Fax: +34 93 214 73 02, E-mail: email@example.com
An infectious etiology for bladder cancer has long been suspected. Merkel cell virus (MCV), BKV and JCV polyomaviruses are possible causative agents but data remain scarce. Therefore, we evaluated the seroresponse to these three polyomaviruses in association with bladder cancer risk. 1,135 incident bladder cancer subjects from five Spanish regions and 982 hospital controls matched by sex, age and region were included. 99% of cases were urothelial-cell carcinomas. Antibody response against MCV, BKV and JCV was measured by enzyme immunoassay using Virus-Like-Particles. Our results show a similar seroprevalence in cases and controls: 64/60% for BKV, 83/82% for MCV and 87/83% for JCV. However, among seropositive subjects, higher median seroreactivities were observed in cases compared to controls for BKV (0.84 vs. 0.70, p-value = 0.009) and MCV (1.81 vs. 0.65, p-value < 0.001). Increased bladder cancer risk was observed for BKV (OR = 1.4, 95%CI 1.04–1.8) and for MCV (OR = 1.5, 95%CI 1.2–1.9), when comparing highest to lowest seroreactivity tertiles. The associations of BKV and MCV with bladder cancer were independent of each other and neither smoking status nor disease stage and grade modified them. Furthermore, no association was observed between seroresponse to JCV and bladder cancer. Therefore, we conclude that BKV and MCV polyomavirus infection could be related to an increased bladder cancer risk.
In industrialized countries, urothelial-cell carcinoma (UCC, previously described as transitional-cell carcinoma) is the dominant histological type of malignant tumors of the urinary bladder. Cancer of the bladder is the fifth most common cancer for both sexes in Spain, and accounts for an estimated age-standarized incidence rate of around 105.000 new cases per year in Spain. Tobacco consumption, male gender, occupational exposure to aromatic amines and other chemicals, certain analgesics, drinking water contaminants and genetic susceptibility factors are among the main identified causes associated with UCC.[2-4] An infectious etiology, Schistosoma haematobium, has been linked to squamous cell carcinoma of the urinary bladder and has long been suspected also for UCC although evidence is scarce. Overall evidence, based on a number of case–control studies, suggests an increased risk for cystitis and nonspecific urinary tract infections but evidence is not consistent and may be potentially attributable to reverse causality.[3, 4]
Infection with polyomaviruses has been recently categorized by the International Agency for Research in Cancer as “probable carcinogen” in humans for Merkel cell polyomavirus (MCV) and Merkel cell carcinoma. Infections with BKV and JCV polyomaviruses have been classified as “possible carcinogens” with studies in humans showing inconsistent evidence for an association with cancers at various sites. The genomes of these three viruses code for oncogenic proteins in experimental studies but only MCV has been linked to human cancer development. Furthermore, although polyomaviruses are commonly identified in urine samples,[6, 7] little is known on their potential association with bladder cancer. BKV is known to produce a specific type of nephropathy which has enhanced the research on its possible association with bladder cancer, mostly in case-series studies. Two negative epidemiological studies have been reported; a small BKV serological nested case–control study within a cohort and a case–control study based on polyomavirus DNA detection in urine. Additionally, one study reported a higher incidence of bladder cancer in subjects with prior detection of decoy cells in the urine while another study found an almost null prevalence of large T antigen nuclear immunostaining in urothelial carcinoma samples from renal transplant recipients.
Therefore, we examined in a large case–control study in Spain the association between BKV, MCV and JCV seroreactivity with the risk of bladder UCC.
Material and Methods
The Spanish Bladder Cancer Study is a hospital-based case–control study including 18 hospitals within five areas of Spain (Asturias, Barcelona metropolitan area, Vallès/Bages, Alicante and Tenerife). Eligible cases were aged between 21 and 80 years old and had newly diagnosed, histologically confirmed and previously untreated carcinoma of the urinary bladder in 1998–2001. Tumor sections from each patient were reviewed by a panel of expert pathologists to confirm the diagnosis and to ensure uniformity of classification criteria, based on the 1998 system of WHO and the International Society of Urological Pathology.
Controls were selected from patients admitted to participating hospitals with diagnoses unrelated to known or suspected causes of bladder cancer. The distribution of hospital admissions for controls were: 39% inguinal hernia, 11% other abdominal surgery, 23% fractures, 6% other orthopedic problems, 12% hydrocele, 3% circulatory disorders, 2% dermatological disorders and 4% other diseases. Controls were individually matched to cases for age at interview within 5-year categories, sex, ethnic origin and region. Among the eligible subjects, 84% of cases and 88% of controls agreed to take part in the study and were interviewed. Of the 1,219 cases and 1,271 controls interviewed, 1,137 (93%) cases and 985 (77%) controls provided a blood sample for serology testing, collected prior to treatment. Five individuals (two cases, three controls) were removed from analysis because of missing information on smoking status. Thus, the final study population available for analysis was 1,135 cases and 982 controls.
Information on known or potential risk factors for bladder cancer was collected by means of computer-assisted personal interviews during the hospital admission. Detailed information on smoking habits, dietary factors, fluid intake, medical conditions, occupational and residential histories, family history of cancer and history of medication use (i.e., analgesics and nonsteroidal anti-inflammatory drugs) was collected. Based on smoking habits, participants were classified as never smokers if they had smoked fewer than 100 cigarettes in their lifetime and as ever smokers otherwise. Information on stage and grade of the tumor was extracted from clinical records and reevaluated through a group of expert pathologists. Subjects were classified by disease stage and grade into three subphenotypes: low-grade nonmuscle invasive (LG-NMI; TaG1/TaG2 papillary tumors, n = 613), high-grade nonmuscle invasive (HG-NMI; TaG3/T1G2/T1G3 papillary tumors, n = 214), and muscle-invasive (MI; T ≥ 2, n = 261). Informed consent was obtained from all participants and the protocol was reviewed and accepted by local ethics committees and the institutional review board of the United States National Cancer Institute.
Viral exposure was measured serologically for each virus using VP1 virus like particles (VLP) enzyme immunoassay as previously described.[13-16] Minor modifications included use of serum at a 1:400 dilution for MCV and 1:200 for BKV and JCV. Serum samples yielding an optical density (OD) value greater than the cut-off value (COV) of OD ≥0.220 for MCV, OD ≥0.200 for BKV or OD ≥0.160 for JCV, were considered as positive for IgG antibodies against each virus. These COVs were established as an OD greater than the mean seroreactivity results plus four standard deviations of serum samples from 63 children aged 2-years-old after exclusion of outliers (n = 7). Among those considered to be seropositive, OD values were also categorized into tertiles based on the control population data distribution to allow a categorical analysis.
Serology tests were performed initially in a subsample of 933 subjects (466 controls and 467 cases) for the three viruses. Afterwards, since no significant association was found for JCV, remaining samples of the study were tested only for BKV and MCV and, therefore, additionally adjusted by batch number.
To estimate the association of categorized exposure for polyomaviruses and bladder cancer risk, odds ratios (OR) and 95% confidence intervals (95%CI) were calculated using unconditional logistic regression, adjusting for potential confounding variables (i.e., age at interview, geographic region, gender, smoking habit and batch sampling). To test for linear trend, categorized variables were treated as continuous.
Generalized additive models (GAM) were used to evaluate the exposure–response curve independently of the exposure categories defined. Best fit model was examined through visual inspection, the Akaike Information Criterion-AIC and total gain (nonlinearity chi-square) of each model after applying up to four degrees of freedom.
To further explore the seroreactivity results in relation to a putative disease-related immunosuppression, population co-infected by both viruses was evaluated. Correlation between individual and tertiles categorized serological data for both viruses were analyzed by Spearman and chi-square test respectively. Adjustment by introduction of both viruses in the same logistic model was also done.
Multinomial logistic regression was used to calculate the OR and 95%CI for association estimates to each specific disease subphenotype compared to controls. A possible association of each polyomavirus seroreactivity with subphenotype risk among cancer subjects was assessed by adjusted logistic regression.
The association between seroreactivity and bladder cancer risk was evaluated also separately for ever smokers and never smokers. In addition, cross-product terms for smoking habit and each viral exposure, combined and separately, were added to the logistic model and its effect measured by likelihood ratio test. Finally, an analysis was conducted excluding the 16 cases that were classified as non-UCCs.
Significance level was established at 0.05 and all tests were two-sided. Analyses were conducted with Stata software, version 10.1.
A description of the study population of cases and controls is shown in Table 1. Most subjects were men (>85%), aged between 60 and 70 years (median age: 67 years) and with a low educational level as 80% of them had not finalized secondary education. Compared to controls, cases were more likely to be older, ever smokers and to live in a big city or metropolis.
Table 1. Descriptive characteristics of study population
The associations of BKV, MCV and JCV seropositivity with demographic and potential confounding factors in controls are shown in Table 2. Seroprevalence in the control population was 60% for BKV, 82% for MCV and 83% for JCV. BKV seroprevalence decreased with age and was lower among those living in villages. Differences for BKV seropositivity were also observed for recruitment region while no effect was observed for gender, education and smoking habit. MCV seropositivity was not associated to any of the characteristics studied. JCV seroprevalence was only associated to size of residence and, contrary to BKV, was highest among those living in villages.
Table 2. Descriptive characteristics of BKV, MCV and JCV polyomavirus seroprevalence among controls
Further exploration of polyomaviruses seroprevalences and seroreactivities was done in the control population. Age has been previously linked to polyomavirus quantitative serology response.[15, 17] However, we did not find any significant correlation for continuous data (r = 0.016, p-value= 0.653 for MCV and r = 0.074, p-value = 0.070 for BKV) nor for median age comparison of seroreactivity among tertiles (p-value = 0.338 and p-value = 0.103, for MCV and BKV respectively). Regarding a possible relation between viruses, there was no association for BKV and MCV seroprevalence (p-value = 0.90). Among control subjects seropositive for both viruses (n = 488), no correlation was observed between continuous data (r = −0.004, p-value = 0.926) neither an association between MCV and BKV serology tertiles (p-value = 0.170). Additionally, further analyses were done to rule out a different serological behavior by any of the hospital admission subgroup among controls. No differences in seroprevalence neither in seroreactivities were observed between control subgroups.
Table 3 shows the associations observed between bladder cancer and each polyomavirus seroreactivity. Cases had a higher seroprevalence for BKV polyomavirus than controls (64% vs. 60%), although this difference was only marginally significant (OR = 1.19, 95%CI = 0.99–1.43). No significant differences for MCV seroprevalence (83% vs. 82%) were observed while JCV seroprevalence was slightly higher in cases (87% vs. 83%, p-value = 0.114).
Table 3. Odds ratio and 95% confidence intervals (OR, 95%CI) of incident bladder cancer by seroprevalence to polyomaviruses and by seroreactivity among seroprevalent subjects
Among seropositive subjects, higher median seroreactivities were observed in cases both for BKV (0.84 vs. 0.70, p-value = 0.009) and MCV (1.81 vs. 0.65, p-value < 0.001) when compared to control subjects, and an increasing risk was observed by levels of seroreactivity evaluated as a continuous exposure variable through a GAM model. No association was found with JCV. Categorization of seroreactivity of BKV and MCV into tertiles confirmed this increase in bladder cancer risk after adjustment by possible confounding factors (Table 3). Risk of bladder cancer was significantly increased in subjects highly reactive to BKV (OR = 1.37, 95%CI = 1.04–1.80) when compared to those of low seroreactivity (first tertile) to this virus. A similar increased risk was observed in those highly reactive to MCV (OR = 1.48, 95%CI = 1.16–1.88) when compared to those of low seroreactivity (first tertile) to MCV. These results were not modified when models were mutually adjusted for the seroresponse to the other virus (Supporting Information Fig. S1) among subjects positive for serum antibodies to both viruses (n = 1,094). Evaluation of dose–response incorporating seronegative subjects as reference category resulted in even stronger trends for BKV (p-value = 0.007) and similar for MCV (p-value = 0.015). Sensitivity analysis by exclusion of cases with nonurothelial-cell bladder cancer (n = 16) did not change results.
ORs by disease subphenotype for high versus low seroreactivity to BKV and MCV are shown in Figure 1. Overall, seroreactivity was associated with an increased risk for all subphenotypes and, when comparing risks between subphenotypes, there was no consistent pattern of increased risk (p-values higher than 0.25) for any specific subphenotype. The same lack of a pattern by disease subphenotypes was observed for MCV and BKV seropositivity compared to seronegativity (p-values ≥ 0.3).
Adjustment by tobacco (never/ever smokers) did not substantially change the association of each viral exposure. The ORs for seroprevalence was examined by smoking status separately for never and ever smokers but there were no significant differences observed with a p-value for interaction of 0.206 for BKV and 0.389 for MCV.
The etiologic role of infections in urinary bladder cancer has been long discussed and has only been demonstrated for Schistosoma haematobium and squamous cell carcinoma. We show in this serology-based case–control study that higher seroreactivities for BKV and MCV polyomaviruses were significantly associated to UCC, the most common histological type of urinary bladder cancer in industrialized countries.
Two previous epidemiological studies evaluated the association between polyomaviruses and bladder cancer[8, 9] and did not find a significant association. Newton et al. performed a BKV serology nested case–control study within a cohort but the follow-up time was of only 1.5 years and only nine cases were evaluated. Polesel et al., in a larger sample, tested for BKV and MCV polyomaviruses DNA in urine and obtained negative results.
Our results are based on serology and, therefore, a positive seroprevalence does not imply viral presence in the tumor. However, two previous case-series studies performed in Italy by Monini et al. and Fioriti et al. detected BKV DNA in around 50% of bladder cancer tissues. The most recent and largest case-series study performed in the United States detected a DNA prevalence of only 5%. Regarding MCV presence in bladder cancer tissue, little data is available. To our knowledge, there is only one previous study where MCV DNA was detected in six/eight bladder cancer samples at a low viral DNA copy number. This is in contrast to the pattern observed in Merkel cell carcinoma series where high viral DNA copy numbers are normally observed.
Viral seroprevalence was similar in cases and controls but cases had a higher seroreactivity for each polyomavirus independently. However, IgG seroreactivity is difficult to interpret since it does not measure the innate response against an initial infection Therefore, these higher seroreactivities among cases could reflect the viral load at the time of initial infection and its subsequent steady-state level. They could also reflect a persistent viral replication and antigen stimulation. Finally, the high seroreactivity could also be related to a viral reactivation in temporal proximity to cancer development; the latter happening before and maybe contributing to bladder cancer pathogenesis or happening later as a result of an immune system impairment after the disease onset. The higher seroreactivity observed for not only one but two viruses, as well as our previous observation of a higher seroprevalence in lymphomas, could indicate reverse causality (high seroreactivity as a consequence of immune impairment) rather than a causative role. However, other viruses such as EBV or HPV are able to cause more than one cancer depending on the body location and geographic area. There are little data available and it cannot be excluded that a similar pattern could occur for MCV infection. On the other side, the lack of correlation between BKV and MCV viruses suggests that the immune response to these highly prevalent infections is independent; subjects with high seroreactivity to one virus did not show high seroreactivity to the other. This finding together with the lack of correlation of seroreactivity with disease subphenotypes, which show different tumor aggressiveness and therefore a different immunological impairment, supports that our observations with MCV and BKV are unlikely to be explained as a reverse causality effect due to the disease and a consequent immunological impairment, although not discarded.
Tobacco smoking is the most important risk factor for bladder cancer. Besides its carcinogenic effect, it plays also an important role in immunity through both a pro-inflammatory and suppressive effect[23, 24] although much of this evidence is mainly based on experimental research. However, in our data there was no indication that seroreactivity to MCV and BKV was related to tobacco smoking status.
Data on the control population provides some information on the natural history of these viruses. The seroprevalences observed for the viruses are similar to those observed by Touze et al. and Viscidi et al., who also used VLP-ELISA methodology in a similar age-range population. Our data also confirms the decrease observed by these authors on BKV seroprevalence at increasing age. Regarding risk factors for viral seropositivity, a relation has been observed with size of city of longest residence. Conversely, BKV and JCV share a natural history characterized by persistent kidney infections and periodic reactivations but only BKV has been associated to urinary tract diseases. Additionally, MCV has been detected in several tissues and biological samples, although the cells harboring the virus are unknown. Our data suggests that MCV could also be found in the urogenital tract although further studies are needed to unravel its natural history and tropism in this tissue.
Our study has some limitations. Control subjects in this study were hospital-based largely identified in orthopedics and surgery departments with minor, acute problems. Although it seems unlikely that our results could be due to largely modified seroresponses among the controls we cannot rule out that they may be more similar to our cases and not fully representatives of the general population; this might have led to an underestimation of the risk. The study could have benefited of urine samples as it would have allowed us to measure viral DNA excreted from the bladder but, unfortunately, these were not available. Further studies on viral DNA in tumor tissue should be conducted. Additionally, data on previous history of organ transplant other than kidney as well as previous and/or current intake of immunosuppressive drugs was not collected.
The major advantages of our study are its epidemiological design, the large study size, the fact that cases were identified and recruited immediately after their diagnosis, the blood sample collection before treatment and the extensive experience in serological analysis. Serological data explore lifetime exposure in a similar manner for both cases and controls and thus allows to adequately estimate measures of association.
This is the first study to provide evidence for an association between bladder cancer and BKV and MCV polyomavirus, particularly among subjects with high seroreactivity against any of these viruses. Further molecular studies are warranted.