Sexually Transmitted Infections and Risk of Epithelial Ovarian Cancer: Results From the Finnish Maternity Cohort.

A case-control study was nested within the FMC of the Northern Finland Biobank Borealis. The FMC includes 2 million biospecimens from >90% of pregnancies in Finland between 1983 and 2016; samples were collected during the late weeks of the first trimester or the early weeks of the second trimester of pregnancy [ 24 ]. The pregnancy and perinatal data were obtained from the Finnish Medical Birth Registry [ 25 ]. Study participants were selected among the members of the cohort with (1) no history of twin pregnancies; (2) a blood sample obtained during the first trimester of a pregnancy leading to childbirth; and (3) no history of invasive cancer (except for nonmelanoma skin cancer) and/or borderline ovarian cancer before blood donation. EOC cases were identified through linkages with the Finnish Cancer Registry (founded in 1952). This registry is >95% complete, as the reporting of newly diagnosed cancers is mandatory [ 26 ]. FMC members who received an invasive EOC diagnosis after blood donation and before 31 December 2015 were identified as cases. Where cases had multiple serum samples in the FMC, the serum sample from the singleton pregnancy most proximate to diagnosis was used for the study. A total of 484 eligible cases were included. One control was selected from the eligible control pool based on the criteria above, matched on age at sampling (±6 months), date of sampling (±3 months), and where the data were available, parity at sampling (1, 2, or ≥3 children) and parity at diagnosis (1, 2, or ≥3 children) (n = 484). Morphology codes, provided by the Finnish Cancer Registry, were recoded according to the World Health Organization guidelines [ 27 ] to histologic subgroups of EOC as serous (n = 249), mucinous (n = 144), endometrioid (n = 71), clear cell (n = 20). Tumors classified as malignant Brenner (n = 1), and epithelial histology not otherwise specified (n = 9) were excluded. Data on smoking during pregnancy (yes/no) and number of pregnancies were obtained through linkage with the birth registry. The linkage of Finnish health registers was approved by the data protection authorities at the National Institute for Health and Welfare. The study design and analyses were approved by the Scientific Committee of Biobank Borealis of Northern Finland, University of Oulu, Finland. Serum levels of antibodies against CT, MG, HSV-2, and HPV antigens were measured using a multiplex, fluorescent bead-based assay as described previously [ 18 ]. Primary antibodies of interest for CT were pGP3, noted as a “gold standard” marker of ever infection [ 28 ], and cHSP60, a potential marker of persistent infection [ 29 ]. HtrA and TroA were evaluated as hypothesized markers of CT-related tubal damage [ 30 ]. Antibodies to L1 of HPV-16 and HPV-18 were deemed indicative of HPV-16 and/or -18 exposures [ 31 ]. All antibody analyses were performed at German Cancer Research Center in Heidelberg, Germany. Sera of the individually matched case and control subjects were included in the same laboratory batch, and laboratory technicians were blinded to the case or control status of the samples ( Supplementary Methods ). Conditional logistic regression was used to estimate the relative risk (RR) and 95% confidence intervals (CIs) for EOC overall, comparing individuals with positive versus negative serology using laboratory defined cutpoints. In analyses by histotype, we used an unconditional multinomial logistic regression model, adjusted for matching factors, to evaluate associations for serous, mucinous, and endometrioid/clear cell EOC; endometrioid and clear cell EOC were evaluated together due to limited number of cases in these subgroups and data suggesting similar associations for other risk factors with these subtypes [ 5 ]. Matching factors plus smoking and parity were the core adjustment set; we adjusted for pGP3 in a secondary analysis. We evaluated antibody levels by time between blood collection and diagnosis ( Supplementary Figure 1 ). Potential heterogeneity in associations with exposure to pGP3 and HSP60 alone versus exposure to multiple pathogens (eg, CT vs CT and MG) and lag time between blood donation and diagnosis (<10 years, ≥10 years) were evaluated in stratified analyses, and statistical heterogeneity in associations was evaluated using interaction terms and assessing the Wald P value. We further evaluated associations between CT-associated major outer membrane protein (MOMP)–A/HSP60 antibody level ratio, given prior findings on these markers [ 32 , 33 ]. In a secondary analysis, we dichotomized individuals seropositive for pGP3 and HSP60 markers by the laboratory cutpoint into categories of low and high positive based on prior evidence that chlamydial HSP60 antibodies are associated with TFI [ 34 ] and with higher antibody titers associated with greater TFI severity. For TroA and HtrA, we evaluated seropositive samples in quartiles of antibody level as an exploratory analysis to evaluate associations across levels of antibody response. The FMC was launched in 1983 [ 35 ], while the birth registry was launched in 1987 [ 25 ]. Thus, data on parity and smoking status from the birth registry were not available for participants before 1987 and were missing by design. There were further sporadic missing data (eg, women not disclosing smoking status). Overall, 60.9% of the study sample had missing smoking status and 42.4% had missing parity data. We used multiple imputation to address the missing data and allow for statistical adjustment for these variables ( Supplementary Methods ). All statistical analyses were performed in SAS version 9.4 software (SAS Institute, Cary, North Carolina). All statistical tests were 2-sided, and P < .05 was considered statistically significant.

Median age at blood sample collection was 31 years (range, 15–45 years). Cases were diagnosed at a median age of 45 years (range, 25–63 years), and median time between blood collection and diagnosis was 13 years (range, 0.25-30 years). A total of 52% (n = 249) of cases were diagnosed with serous tumors, with 19% (n = 91) clear cell/endometrioid and 29% (n = 144) mucinous. The youngest median age at diagnosis was observed for mucinous EOC (41 years [range, 25–58 years]); serous (median, 46 years [range, 26–63 years]) and clear cell/endometrioid cases (median, 47 years [range, 29–60 years]) had older age at diagnosis. Smoking prevalence was similar in cases and controls (17%–18%) ( Table 1 ; similar distribution in the subset with complete data). The proportion of smokers was higher in mucinous cases (28%) as compared to the serous (15%) or clear cell/endometrioid groups (10%), and cases with mucinous cancer had the lowest prevalence of multiparity (36%; serous, 41%; clear cell/endometrioid, 43%). Baseline Characteristics of Study Participants by Epithelial Ovarian Cancer Subtype, Finnish Maternity Cohort, 1983–2015 Percentages displayed in the table are for data after multiple imputation. Abbreviation: EOC, epithelial ovarian cancer. a Smoking data after multiple imputation; smoking data were missing for 60.9% of study samples. b Parity data after multiple imputation; parity data were missing for 42.4% of study samples. Seropositivity to at least 1 STI was observed in 324 (67%) cases and 312 (65%) controls. Positive serology for the pGP3 antibody was observed in 38% of cases and 40% of controls, with a total of 30% of cases and 32% of controls positive for pGP3 and both tubal damage markers (146 cases, 153 controls) ( Table 2 ). Similar prevalence of seropositivity in case and controls was observed for MG (140 [29%] cases, 127 [26%] controls) and HSV-2 (137 [28%] cases, 120 [25%] controls). Positive serology for HPV-16 and/or -18 was observed for 7% of cases (n = 34) and 9% of controls (n = 45) ( Supplementary Table 1 ). All but 5 subjects positive for both tubal damage markers were also positive for pGP3, as expected given that the tubal damage markers are CT specific. Among pGP3-seropositive women, the most frequent additional infection detected was MG (positive serology for both, 7.4% of cases and 7.9% of controls) ( Supplementary Table 2 ). Overall, characteristics of participants with positive versus negative serology were similar with respect to age at blood collection and age at diagnosis (absolute difference in age between positive and negative serology, ≤3 years; Supplementary Table 3 ). For current smokers, there was no significant differences by infection status. For all infections, seronegative women were more likely to have ≥2 births compared to seropositive women ( Supplementary Table 3 ). We observed no associations between CT infection–related serology measures and EOC risk overall (eg, pGP3: RR, 0.92 [95% CI, .72–1.19]; HSP60: RR, 1.04 [95% CI, .81–1.34]; Table 2 ). No significant associations were observed in analyses by histologic subtype ( P value for heterogeneity [ P het ] ≥ .16), with the exception for the combination of positive serology for pGP3 and both tubal damage markers (TroA and HtrA; P het < .0001); however, none of the individual associations was statistically significant ( Table 2 ). We further evaluated associations considering low and high positive levels of pGP3 and HSP60 ( Table 2 ) and TroA and HtrA serology in quartiles ( Supplementary Table 4 ); no associations were observed in these analyses. Adjusted Relative Risk for Epithelial Ovarian Cancer Overall and by Subtype by pGP3, HSP60, TroA, and HtrA and Mycoplasma genitalium Serologic Status in the Finnish Maternity Cohort, 1983–2015 Abbreviations: CI, confidence interval; EOC, epithelial ovarian cancer; HSP60, heat shock protein 60 variant 1; P het , P value for heterogeneity; Ref, reference category; RR, relative risk. a Frequency (No.) and percentage (%) are reported based on seropositivity (+) for individual markers or the combined marker (“Seropositive”). The reference category is seronegative for an individual marker or negative for the combined marker. b RRs and 95% CIs are from logistic regression models conditioned on matching factors and adjusted for age at diagnosis, sampling age, smoking (nonsmoker/smoker), and parity (1 birth/≥2 births). c RRs and 95% CIs are from unconditional logistic regression models adjusted for age at diagnosis, sampling age, smoking (nonsmoker/smoker), and parity (1 birth/≥2 births). d Tubal damage markers TroA and HtrA both present as well as pGP3 or HSP60 as indiciated. Subsequently, we evaluated associations between the other STIs and the risk of EOC overall and by histotype. Positive serology for MG was associated with significantly higher mucinous EOC risk (RR, 1.66 [95% CI, 1.09–2.54]; P het by histologic subtype = < 0.001) ( Table 2 ); the association was somewhat stronger in models adjusted for pGP3 infection (RR, 1.77 [95% CI, 1.16−2.69]). We observed no significant associations between HSV-2 and HPV serology and EOC risk ( Table 3 ). No statistically significant associations were observed in our secondary analyses on seropositivity to multiple STIs and EOC risk ( Table 3 ), nor were associations observed between MOMP-A/HSP60 antibody level ratios. No clear patterns emerged in analyses by time between blood collection and diagnosis for antibody levels ( Supplementary Figure 1 ) or risk associations ( Supplementary Table 5 ), though a significant association between MG and mucinous EOC was only observed among women diagnosed within 10 years of blood collection. Adjusted Relative Risk for Epithelial Ovarian Cancer Overall and by Subtype by Herpes Simplex Virus Type 2, Human Papillomavirus, and Multiple Sexually Transmitted Infections Serologic Status in the Finnish Maternity Cohort, 1983–2015 Abbreviations: CI, confidence interval; CT, Chlamydia trachomatis ; EOC, epithelial ovarian cancer; HPV, human papillomavirus type 16 or 18 L1; HSV-2, human simplex virus type 2; P het , P value for heterogeneity; Ref, reference category; RR, relative risk; STI, sexually transmitted infection. a Frequency (No.) and percentage (%) are reported based on seropositive (+) for individual markers or the combined marker (“Seropositive”). The reference category is seronegative for an individual marker or negative for the combined marker. b RRs and 95% CIs are from logistic regression models conditioned on matching factors and adjusted for age at diagnosis, sampling age, smoking (nonsmoker/smoker), and parity (1 birth/≥2 births). c RRs and 95% CIs are from unconditional logistic regression models adjusted for age at diagnosis, sampling age, smoking (nonsmoker/smoker), and parity (1 birth/≥2 births).

We observed no association between the evaluated CT markers and EOC risk, regardless of disease subtype. These results are in contrast to recent studies showing positive associations between EOC overall [ 17–20 ] or for specific EOC subtypes [ 17 , 22 ]. We found a positive association between MG seropositivity and mucinous EOC risk; no associations were observed for other histotypes. Seropositivity to multiple STIs was neither associated with increased EOC risk overall, nor by histologic subtype in this study. CT infection has various clinical manifestations. CT can partially evade the host immune system, provoke local inflammation through cytokine release, and inhibit host-cell apoptosis [ 36 ]. CT-related sequelae can result in damage and scarring of the fallopian tubes and other genital tract structures [ 14 ]. Infections with CT and MG result in a detectable, durable immune response [ 21 ]. There is some evidence that antibodies to chlamydial HSP60-1 are present in more severe, persistent disease [ 21–24 ]. Additionally, higher antibody titers postexposure may be associated with a more severe, ascending infection [ 34 ], progressing to further sequela such as PID. More recently, antibodies against HtrA and TroA were observed in individuals with CT infection–associated tubal damage and persistent infection [ 22 , 23 ]. We observed no associations between positive CT serology and EOC risk in the current study. Contrary to our findings, recent prospective studies found an increased risk of EOC overall in individuals seropositive for CT marker pGP3 [ 19 ] within the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial (PLCO) cohort (RR, 1.43 [95% CI, 0.78–2.63]) and the Nurses' Health Study (NHS/NHSII) cohort (RR, 2.07 [95% CI, 1.15–3.4]), although in the European Prospective Investigation into Cancer and Nutrition (EPIC) cohort only the association with mucinous disease was significant (EOC overall, 1.36 [1.13–1.64]; mucinous, 2.30 [1.22–4.32]). HSP60 seropositivity was associated with higher overall and serous EOC risk (eg, serous, 1.44 [1.12–1.85]) within the EPIC cohort [ 4 ], with no statistically significant associations observed in prior retrospective studies [ 18 , 22 , 37 ]. There are several potential explanations for the discrepant results between the current study and prior serologic studies. First, blood samples in the current study were collected exclusively during pregnancy. The findings in the NHS/II [ 19 ] and EPIC [ 17 ] cohorts were not modified by parity; thus, this is unlikely to fully explain the discrepancy in findings. FMC participants were younger than those of prior studies (eg, FMC median age at blood collection was 31 years vs 55–63.5 years for prior studies) [ 17–19 ]. Assuming that the majority of infections occurred at an earlier age (eg, according to the latest European Centre for Disease Prevention and Control report, 60% of CT cases were reported among 15- to 24-year-olds [ 38 ]), this is relevant given the dynamics of chlamydial antibody persistence in blood. About one-third of CT-exposed women have measurable CT antibodies 3–10 years after the infection. Repeated exposure boosts the preexisting antibody levels [ 39 ]. It is plausible that the association seen in the other studies was due to persistent antibodies in subjects with more severe infection and/or that the infection caused discernible elevation in risk manifesting at a more advanced age, once a certain time threshold is reached. We observed higher seropositivity prevalence in the current study, relative to prior studies, with 38.4% of participants seropositive for pGP3 (seroprevalence of 25%, 19.6%, and 30.8% in PLCO, NHS/II, and EPIC, respectively). Interestingly, in the EPIC study, the association of pGP3 seropositivity and EOC risk increased when the sample was restricted to women who were >60 years of age at blood collection [ 17 ]. Further unmeasured factors may explain the inconsistent findings, such as whether an STI diagnosis has ever been given, symptom severity, and treatment, as they influence complete infection clearance and thus decrease the chances of chronic infection, persistence, and associated sequelae [ 40 ]. Last, the exposure to different bacterial strains may depend on the geographic region, as different strains may show varying degrees of pathogenicity [ 41 ]. Given the tubal origins of the majority of EOC [ 3 ], the suggested CT-related tubal damage markers TroA and HtrA may be of relevance for EOC risk. Our study observed no significant associations between these markers and EOC risk. However, it should be noted that the TroA and HtrA assays have not yet been widely validated and, given the relatively high proportion of seropositive women (31.4%), these markers may have limited specificity. MG is associated with urogenital inflammatory conditions [ 42 ]. The immunopathogenicity of MG is not fully understood, though some experimental and clinical studies have shown that MG utilizes human mucin structures as entry points [ 43 ] and can persist through chromosomal aberrations, inducing a strong local inflammatory response [ 44 , 45 ]. These processes can induce malignant transformation of the cells. Positive MG serology was associated with 70% higher mucinous EOC risk in the current study, but not any other EOC subtypes. To our knowledge, this has not been observed previously. This finding was robust to adjustment for smoking and parity. Mechanisms linking MG specifically to mucinous EOC are not clear. Histopathologically, mucinous EOC resembles gastrointestinal tract tumors [ 27 ] and is suggested to originate from transitional epithelium at or near the tuboperitoneal junction. These cells might be vulnerable to infections, such as MG. However, further mechanistic research in this area is required. Previously, a suggestive positive association between MG exposure and overall EOC risk was observed in the NHS/II cohort (RR, 1.92 [95% CI, 0.78–4.72]) [ 19 ] and in the Polish case-control study (RR, 1.63 [95% CI, 1.20–2.22]) [ 18 ], albeit the findings in the latter were suggestive and not statistically significant in NHS/II and no associations were observed in EPIC or PLCO. Interestingly, in the analysis restricted to parous women within the NHS I/II cohort, the association between MG seropositivity and overall EOC risk became stronger (RR, 3.40 [95% CI, 1.25–9.27]) [ 19 ]. Notably, the MG seropositivity in our cohort is higher than in the prior prospective study, which previously observed a positive association with EOC (28% in the FMC vs 11.5% in NHS/II [ 19 ]). Beyond the primary bacterial infections of interest, we further evaluated the HSV-2 and HPV high-risk types and EOC risk. HPV-16 and -18 are established inducers of cervical transformation zone dysplasia and malignancy [ 17 ]. Consistent with previous epidemiologic and clinical studies [ 17–19 ], we observed no association between HSV-2, HPV-16, and HPV-18 and EOC risk. One exception is the EPIC study [ 17 ], which found a significant association between HSV-2 infection and endometrioid EOC. We observed no association in the subgroup of endometrioid and clear cell disease. This study has notable strengths and limitations. Cases in this study had a relatively young age at diagnosis, resulting in a higher proportion of nonserous histotypes than in past studies. There are limited data on the association of past STI exposure and ovarian cancer risk by EOC histotypes, with the EPIC cohort providing an evaluation in these subgroups [ 17 ], and other recent studies [ 18 , 19 ] assessing associations in subgroups of serous and nonserous EOC. While a limitation is that the power to detect associations was still limited, we provided needed evidence by histologic subgroup. To control for confounding, the models were adjusted for smoking and parity. Smoking has been identified as a risk factor for mucinous EOC [ 10 ]. Due to limited data availability, we were unable to adjust our models for risk factors such as BMI or oral contraceptive and menopausal hormone therapy use. Furthermore, data on health conditions other than cancer were not available. Given the minimal confounding observed in the NHS/II study (eg, CT: unadjusted RR, 2.04 [95% CI, 1.26–3.29]; adjusted RR, 2.07 [95% CI, 1.25–3.43]), we do not expect substantial residual confounding. We evaluated TroA and HtrA in exploratory analyses, classifying antibody levels in quartiles and restricting the positive subjects to those who were also pGP3/HSP60 positive. Tubal damage is one of the leading causes of infertility, and a limitation of the current study is the assessment of tubal damage markers in a cohort of pregnant, and thus fertile, women. While previous studies did not report heterogeneity of risk by parity status, the fully parous study sample limits the generalization of these findings and may have obscured relevant associations with these markers. Gestational age at blood collection data was not available; however, gestational age at blood collection would likely be a source of random error and not be associated with EOC. Finally, while our study was hypothesis based, we made multiple comparisons across markers and by subtype and, after adjustment for multiple testing, none of our findings would remain statistically significant. We observed no increased risk of EOC overall and by specific histotype for seropositivity to previously studied CT antigens. Higher risk of mucinous EOC was observed in participants seropositive for MG. Further experimental, epidemiologic, and clinical research is needed to determine the role of multiple and/or persistent CT infection, seropositivity to specific chlamydial antigens, and MG infection in EOC development with a focus on disease histotype. Future studies should examine the role of the gynecologic tract microbiome in inflammatory and oncogenic processes toward identifying potential opportunities for the prevention of this highly lethal disease through STI screening, treatment, and vaccine development.

Click here for additional data file.