Modification of the Association Between Frequent Aspirin Use and Ovarian Cancer Risk: A Meta-Analysis Using Individual-Level Data From Two Ovarian Cancer Consortia

PURPOSE Frequent aspirin use has been associated with reduced ovarian cancer risk, but no study has comprehensively assessed for effect modification. We leveraged harmonized, individual-level data from 17 studies to examine the association between frequent aspirin use and ovarian cancer risk, overall and across subgroups of women with other ovarian cancer risk factors. METHODS Nine cohort studies from the Ovarian Cancer Cohort Consortium (n = 2,600 cases) and eight case-control studies from the Ovarian Cancer Association Consortium (n = 5,726 cases) were included. We used Cox regression and logistic regression to assess study-specific associations between frequent aspirin use (≥ 6 days/week) and ovarian cancer risk and combined study-specific estimates using random-effects meta-analysis. We conducted analyses within subgroups defined by individual ovarian cancer risk factors (endometriosis, obesity, family history of breast/ovarian cancer, nulliparity, oral contraceptive use, and tubal ligation) and by number of risk factors (0, 1, and ≥ 2). RESULTS Overall, frequent aspirin use was associated with a 13% reduction in ovarian cancer risk (95% CI, 6 to 20), with no significant heterogeneity by study design (P = .48) or histotype (P = .60). Although no association was observed among women with endometriosis, consistent risk reductions were observed among all other subgroups defined by ovarian cancer risk factors (relative risks ranging from 0.79 to 0.93, all P-heterogeneity > .05), including women with ≥ 2 risk factors (relative risk, 0.81; 95% CI, 0.73 to 0.90). CONCLUSION This study, the largest to-date on aspirin use and ovarian cancer, provides evidence that frequent aspirin use is associated with lower ovarian cancer risk regardless of the presence of most other ovarian cancer risk factors. Risk reductions were also observed among women with multiple risk factors, providing proof of principle that chemoprevention programs with frequent aspirin use could target higher-risk subgroups.


INTRODUCTION
Ovarian cancer is the most fatal gynecologic cancer, largely because of nonspecific symptom presentation and lack of early detection strategies. 1 Chemoprevention holds promise but remains an understudied paradigm to reduce ovarian cancer burden. 2 Chronic inflammation likely plays a key role in ovarian carcinogenesis, 3 as factors associated with epithelial disruption from ovulation, 4,5 inflammation-related exposures such as endometriosis and pelvic inflammatory disease, 6,7 and circulating biomarkers of inflammation 8,9 are associated with ovarian cancer risk. Anti-inflammatory medications such as aspirin may lower risk of ovarian cancer development via inhibition of the cyclooxygenase enzymes, leading to decreased synthesis of inflammatory mediators, or via cyclooxygenase-independent pathways including inhibition of Wnt/b-catenin and NF-kb. 10 A growing body of evidence supports a role of aspirin in reducing ovarian cancer risk. Pooled secondary analyses of randomized controlled trials of aspirin for cardiovascular disease prevention have noted a decreased risk of female reproductive cancers with $ 3 years of aspirin use, although too few ovarian cancer cases were diagnosed in these trial populations to draw inferences for ovarian cancer specifically. 11 In the observational setting, individual study results have been mixed, [12][13][14][15][16][17][18][19][20][21][22][23] but metaanalyses 24 and pooled analyses of cohort 25 and casecontrol 26 studies have found that aspirin may reduce ovarian cancer risk by 10%-20%, particularly when used frequently (ie, daily or almost daily).
However, although aspirin use appears to be one of the few modifiable protective factors for ovarian cancer, population-wide chemoprevention programs are generally considered infeasible because of the low incidence of ovarian cancer and the known risk of bleeding conferred by frequent aspirin use. 27 Instead, such programs will likely need to focus on subgroups of women at elevated risk of ovarian cancer. 28 Established factors that increase ovarian cancer risk include a family history of breast or ovarian cancer, endometriosis, and obesity, whereas factors that decrease risk include parity, oral contraceptive use, and tubal ligation. Whether frequent aspirin use reduces risk of ovarian cancer among subgroups of women defined by these risk factors is unknown, and extremely large, wellpowered studies are needed.
In this study, we leveraged harmonized, individual-level data from two ovarian cancer consortia that previously reported on frequent aspirin and ovarian cancer risk 25,26 to comprehensively assess this association across key subgroups of interest. By meta-analyzing results from these 17 studies, we aimed to test for the consistency of the association across study design and personal characteristics and provide the most precise estimates of the aspirinovarian cancer association to date.

Study Populations
We analyzed individual-level data from prospective cohort studies from the Ovarian Cancer Cohort Consortium (OC3) 29 and population-based case-control studies from the Ovarian Cancer Association Consortium (OCAC). Studies were included if they collected information on frequency of aspirin use; this resulted in the inclusion of nine cohort and eight case-control studies, a subset of the studies included in previous aspirin analyses from these consortia. 25,26 The cohort studies (NIH-AARP Diet and Health Study, 16,30 Cancer Prevention Study II Nutrition Cohort, 31,32 California Teachers Study, 33 Iowa Women's Health Study, 19 Nurses' Health Study, 18 Nurses' Health Study II, 18 Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial, 34 Sister Study, 35 and Vitamins and Lifestyle Cohort 36,37 ) were all US-based cohorts; our analysis included women from these cohorts with at least one intact ovary, no history of cancer at baseline, and nonmissing age and frequency of aspirin use. The case-control studies (Australian Ovarian Cancer Study, 22  All participating studies obtained institutional review board approval at their respective institutions. Participants provided written informed consent or implicit consent through return of study questionnaires. The coordinating centers for OC3 (Brigham and Women's Hospital, Harvard T. H. Chan School of Public Health) and OCAC (Duke University)

CONTEXT Key Objective
To determine whether the association between frequent aspirin use and ovarian cancer risk is modified by established ovarian cancer risk factors (endometriosis, obesity, family history of breast/ovarian cancer, parity, oral contraceptive use, and tubal ligation).

Knowledge Generated
In combined analyses of individual participant data from 17 study populations, frequent aspirin use was associated with a 13% reduction in ovarian cancer risk overall. Consistent risk reductions were observed across most subgroups of women with other ovarian cancer risk factors, with the exception of endometriosis. Among women with two or more risk factors, frequent aspirin use was associated with a 19% reduction in ovarian cancer risk. Relevance This study confirms the association of frequent aspirin use with decreased risk of ovarian cancer. The use of aspirin for ovarian cancer chemoprevention may best be targeted to higher-risk women with two or more ovarian cancer risk factors, to maximize the population-level benefit/risk ratio.
received institutional review board approval from their institutions and participating registries as required for data acquisition, pooling, and harmonization.

Study Variables
Given previous findings that frequent aspirin use was most strongly associated with ovarian cancer risk, 25,26 our primary exposure was frequent aspirin use, which was self-reported in all included studies (Appendix Tables A1  and A2, online only). Frequent aspirin use was harmonized across the study populations to indicate aspirin use for $ 6 days/week or $ 28 days/month and for a duration of $ 6 months. Frequent aspirin use was defined irrespective of dose, as few studies collected data on aspirin dose. Women who reported less frequent or no aspirin use were combined to form the reference group.
Other covariates were centrally harmonized at the coordinating centers of OC3 and OCAC. 6,29,[47][48][49] For the cohort studies, aspirin use and other covariates were assessed at enrollment or at a subsequent questionnaire cycle, which then became the baseline for this analysis. For the case-control studies, covariates were ascertained at enrollment.
Our primary outcome was invasive epithelial ovarian, fallopian tube, or peritoneal cancer. In the cohort studies, cases were identified through linkage to cancer registries or medical chart review. 29 Nonepithelial tumors and tumors of low malignant potential/borderline were excluded. Case ascertainment for the case-control studies included linkage to cancer registries or hospital registries, surgical treatment centers, gynecologic oncology centers, physician offices, and/or pathology databases. 47 We also examined associations for the most common ovarian cancer histotypes, including high-grade serous, mucinous, endometrioid, clear cell, and other epithelial tumors. Very few low-grade serous tumors were observed in these study populations; these tumors were consequently excluded from histotypespecific analyses.

Statistical Analysis
For each cohort study, hazard ratios (HRs) and 95% CIs comparing frequent aspirin use to nonfrequent use were calculated using Cox proportional hazards regression. Women entered the analysis at age at study entry and contributed person-time until first diagnosis of ovarian cancer, death, or end of follow-up. Models were adjusted for baseline age, number of full-term births (none, one, two, three, or $ four), duration of oral contraception use (never, # 1, . 1-5, . 5-10, or . 10 years), duration of menopausal hormone therapy use (premenopausal, never, # 5, . 5-10, or . 10 years), and body mass index (BMI, , 20, 20 to , 25, 25 to , 30, 30 to , 35, or $ 35 kg/m 2 ). For each case-control study, odds ratios (ORs) and 95% CIs were calculated using logistic regression, adjusting for the same covariates. Study-specific HRs and ORs were calculated overall as well as for subgroups defined by age at baseline (cohort studies) or diagnosis/index date (case-control studies; , 50, 50-59, 60-69, or $ 70 years), history of endometriosis (yes or no), obesity (BMI $ 30 or , 30 kg/m 2 ), parity (parous or nulliparous), family history of breast or ovarian cancer (yes or no), duration of oral contraceptive use (never, , 5, or $ 5 years), tubal ligation (yes or no), and nonaspirin nonsteroidal anti-inflammatory drug (NSAID) use (yes or no). Study-specific effect estimates, overall and for each subgroup, were combined using random effects metaanalysis to generate summary relative risks (RRs).
We also calculated RRs within subgroups defined by an ovarian cancer risk score (range, 0-6, categorized as 0, 1, and $ 2), with each ovarian cancer risk factor (endometriosis, obesity, nulliparity, family history of breast or ovarian cancer, no oral contraceptive use, and no tubal ligation) contributing one point to this score. Before using this score, we confirmed that the risk score was positively associated with ovarian cancer risk (RR for a score of 1 v 0: 1.20, 95% CI, 1.10 to 1.30; RR for a score of $ 2 v 0: 1.78, 95% CI, 1.64 to 1.94). Risk score analyses were adjusted for age and duration of menopausal hormone therapy use.
To examine associations by ovarian cancer histotype, we conducted competing risks Cox regression using an augmented data approach with the pooled cohort data, 50 and polytomous logistic regression with the pooled case-control data, 51,52 adjusting for study and the same covariates as above. We conducted pooled instead of study-specific analyses because of the smaller number of events by histotype. The results from the cohort and case-control analyses were combined using random effects metaanalysis.
We examined heterogeneity in effect estimates by study, study design, subgroup, and histotype using Cochran's Q tests. 53 The number needed to treat (NNT) to prevent one ovarian cancer was calculated using the observed 10-year absolute risk of ovarian cancer among nonaspirin users in the cohort studies and the combined cohort and casecontrol summary RRs. 54 All statistical tests were two-sided, and P values , .05 were considered statistically significant. Study-specific and pooled analyses were conducted in SAS 9.4, meta-analyses were conducted using the meta command in Stata 16, and figures were generated in R 4.0.2.

RESULTS
In the nine cohort studies, there were 491,651 women at risk. Using the combined cohort and case-control data, when we examined associations within subgroups defined by factors that increase ovarian cancer risk, we observed possible effect modification by history of endometriosis (Fig 1). Among women without endometriosis, frequent aspirin use was associated with reduced ovarian cancer risk (RR, 0.82; 95% CI, 0.73 to 0.92), whereas no risk reduction was observed among women with endometriosis (RR, 1.15; 95% CI, 0.80 to 1.65; P-heterogeneity 5 .08). However, the CI for the latter effect estimate was large because of the small number of women with endometriosis (prevalence range, 1%-9% in the cohort studies, 3%-11% among OCAC controls). Frequent aspirin use was associated with lower ovarian cancer risk regardless of obesity, although the association was slightly stronger among obese women (RR, 0.79; 95% CI, 0.67 to 0.93) than among nonobese women  , and risk score (n 5 17). Models were adjusted for age, parity, duration of oral contraceptive use, duration of menopausal hormone therapy use, and BMI. Models stratified by risk score were adjusted for age and duration of menopausal hormone therapy use. Participants with missing data on these covariates (, 10% for all covariates except duration of menopausal hormone therapy use) were retained in the models using missing indicators. We also conducted a complete case analysis and the results were unchanged. a P value for heterogeneity by study design. b P value for heterogeneity by subgroup. BMI, body mass index; OC, oral contraceptive; OC3, Ovarian Cancer Cohort Consortium; OCAC, Ovarian Cancer Association Consortium; RR, relative risk.
Associations were similar across strata of family history of breast or ovarian cancer (RR, 0.86; 95% CI, 0.79 to 0.94 for women without a family history, RR, 0.88; 95% CI, 0.72 to 1.06 for women with a family history, P-heterogeneity 5 .88).
We did not observe effect modification by age at enrollment in the cohort (P-heterogeneity 5 .78) or case-control (P-heterogeneity 5 .26) studies (Table 1). However, in the case-control studies, there was possible strengthening of the association with age, with the strongest inverse association observed among women age 70 years or older at diagnosis/enrollment (OR, 0.72; 95% CI, 0.57 to 0.91).
In general, associations with frequent aspirin use were similar for all ovarian cancer histotypes (Fig 2, Appendix  Table A3, online only). Risk reductions were particularly robust for high-grade serous ovarian cancer, both overall (RR, 0.86; 95% CI, 0.78 to 0.94) and across subgroups defined by ovarian cancer risk factors. For women with endometriosis, although there was no association between frequent aspirin use and ovarian cancer overall, there was suggestion of an inverse association with endometrioid ovarian cancer, the histotype most strongly associated with endometriosis.
In the cohort studies, 21% of women had none of the six ovarian cancer risk factors, 42% had one risk factor, and 37% had $ two. In the case-control studies, the corresponding percentages of women with zero, one, and $ two risk factors were 8%, 28%, and 64% for cases, and 12%, 37%, and 51% for controls. In analyses stratified by the number of risk factors (ie, the ovarian cancer risk score), frequent aspirin use was inversely associated with ovarian cancer risk among women at higher risk of ovarian cancer because of the presence of $ two risk factors (RR, 0.81; 95% CI, 0.73 to 0.90, Fig 1). The protective association for these higher-risk women was consistent across histotypes (Fig 2, Appendix Table A3, P-heterogeneity 5 .42). Among the higher-risk women, the NNT to prevent one ovarian cancer within 10 years was 970 (95% CI, 683 to 1,843; Appendix Table A4, online only). By contrast, the NNT for all women in the study population regardless of risk score was 1784 (95% CI, 1,160 to 3,866).

DISCUSSION
In this analysis of data from two ovarian cancer consortia, frequent aspirin use was associated with a 13% reduction in ovarian cancer risk overall. A similar risk reduction was observed for high-grade serous ovarian cancer, the most common and one of the most fatal histotypes, which is important because most established risk factors are more weakly associated with high-grade serous ovarian cancers. 6 The consistency of the frequent aspirin use-ovarian cancer association across the individual case-control and cohort study populations was notable and provides strong Importantly, our study also found that established ovarian cancer risk factors do not modify the association between frequent aspirin use and ovarian cancer risk. There was a suggestion of effect modification by endometriosis, with an inverse association observed for women without but not with self-reported endometriosis, but this was likely driven by the small number of women with endometriosis and the limited power to detect associations within this subgroup.
Additionally, there was no effect modification by endometriosis for endometrioid or clear cell tumors, the two specific histotypes for which endometriosis is a risk factor. 6 Risk reductions associated with frequent aspirin use were otherwise consistent across subgroups defined by factors that increase (obesity and family history of breast/ovarian cancer) and decrease (parity, oral contraceptive use, and tubal ligation) ovarian cancer risk. The lack of effect modification by adiposity is particularly notable, given that other studies have reported aspirin to be more weakly associated with reduced cardiovascular disease and colorectal cancer risk 55,56 and more strongly associated with reduced endometrial cancer risk 57 among obese individuals; this could suggest that aspirin's mechanism of action for preventing cardiovascular disease and these other cancers may differ from that preventing ovarian cancer.
There was possible effect modification by the ovarian cancer risk score, with a null association observed among women with zero ovarian cancer risk factors. However, the results for women with zero risk factors were inconclusive, given the small number of cases and heterogeneity in the studyspecific results for this subgroup. More critically, we observed a clear inverse association between frequent aspirin use and ovarian cancer among women with multiple ovarian cancer risk factors. These results are important, given that any implementation of aspirin use for ovarian cancer chemoprevention will likely need to focus on specific high-risk subgroups. 28 Our study suggests that frequent aspirin use is protective among women at increased risk of ovarian cancer because of the presence of established epidemiologic risk factors, with a lower NNT among women with $ 2 risk factors, and that targeting chemoprevention programs to women with epidemiologic risk factors may thus be a viable strategy.
To our knowledge, this study is the largest to date on aspirin and ovarian cancer risk and the first to examine effect modification by a comprehensive set of ovarian cancer risk factors. Previous studies of aspirin, examined alone or combined with other NSAIDs, have also reported no effect modification by BMI, 17,18 parity, 12,16,18,46 or oral contraceptive use, 16,18,46 but these individual studies were only powered to detect very strong differences. One study observed a possible stronger association between daily aspirin use and ovarian cancer risk with increasing BMI, 58 a trend that was mirrored in our study, although our study suggests that frequent aspirin may still be modestly protective among nonobese women. Our study confirms and expands upon these prior studies by combining the existing observational data, which facilitated a well-powered analysis, even among subgroups. Access to the individual-level data from each study allowed us to apply a standardized analytic approach, assess associations by histotype, and focus specifically on frequent aspirin use, the pattern of use that appears most protective against ovarian cancer. 25,26 Although we combined results across study design, such pooling was necessary to obtain sufficient power to test for effect modification. Moreover, formal comparison of the cohort and case-control results revealed no meaningful or statistically significant differences. There may have been some bias because of the use of observational data, but research has found that observational studies of aspirin and cancer can recapitulate randomized controlled trial findings when there is detailed recording of aspirin use and careful adjustment for confounders. 59 We were unable to examine associations specifically for low-dose aspirin, which has been more strongly associated with reduced ovarian cancer risk in prior studies, 13,26 but frequent aspirin use was highly correlated with low-dose use in the studies with dosage information available (r 5 0.97 in OC3, r 5 0.82 in OCAC controls). We did not have data on indication for aspirin use or age at initiation of use, both of which require further study. We did not look at associations among women at increased risk of ovarian cancer because of common or rare genetic variants as genetic data were not available for all included studies; whether aspirin reduces risk among women with highly penetrant mutations (ie, BRCA1/BRCA2 or Lynch syndrome) will require examination in specialized study populations. Finally, when calculating the NNT, we were unable to incorporate associations for precise durations of aspirin use because of the use of observational data. The NNT also does not account for the known risks associated with frequent aspirin use, and further research on the net benefits and harms is needed.
In conclusion, this study, the largest to date on frequent aspirin use and ovarian cancer, supports a 13% reduction in ovarian cancer risk with frequent aspirin use, with a 14% reduction for high-grade serous carcinoma, the most common and one of the more lethal histotypes. Similar risk reductions were observed across subgroups defined by established epidemiologic risk factors, and no subgroup experienced a significant increased risk with aspirin use. These results suggest that primary prevention of ovarian cancer is an added benefit of frequent aspirin use that could be incorporated into composite risk-benefit calculations. Given that women with elevated ovarian cancer risk because of epidemiologic risk factors also benefit and that the NNT to prevent one ovarian cancer is lower for higher-risk women, future work should explore how chemoprevention programs with aspirin could complement existing preventive strategies, which are currently limited to women with the highest risk (ie, prophylactic salpingo-oophorectomy for BRCA1/2 carriers) and target additional high-risk subgroups to maximize population-level impact and minimize risks.

AFFILIATIONS DISCLAIMER
The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This is a US Government work. There are no restrictions on its use.          NOTE. Models were adjusted for age, parity, duration of oral contraceptive use, duration of menopausal hormone therapy use, BMI, and study. Models stratified by risk score were adjusted for age, duration of menopausal hormone therapy use, and study. For mucinous ovarian cancers, the relative risk for women with ovarian cancer risk score 5 0 was unable to be estimated.

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST
Abbreviations: BMI, body mass index; OC, oral contraceptive; OC3, Ovarian Cancer Cohort Consortium; OCAC, Ovarian Cancer Association Consortium; RR, relative risk. Combined cohort and case-control summary RRs for the association between frequent aspirin use and ovarian cancer risk. c NNT 5 1/(S(t) RR -S(t)), where S(t) 5 1-to 10-year cumulative incidence of ovarian cancer in nonaspirin users. 54 d Given that the 95% CI for the RR overlaps 1, we cannot preclude the possibility that frequent aspirin use is associated with harm (ie, the full 95% CI extends to include the possibility of a positive NNH). 60