Using ‘Big Data’ to explore unanswered questions in ovarian cancer epidemiology

Abstract

Background Each year nearly 300,000 women are diagnosed with ovarian cancer worldwide, and more than half of these women die within five years of their diagnosis. Clear evidence from epidemiological studies about factors affecting risk or survival is critical for efforts to prevent this cancer and improve prognosis. However, interpretation of findings has often been hampered by low sample sizes, biases in observational studies, and the historic practice of considering all ovarian cancers as one disease, although aetiological distinctions between the major histotypes are now recognised. Due partly to these limitations, it was unclear whether adiposity, greater height, or hysterectomy without oophorectomy are associated with ovarian cancer risk, and whether the use of common analgesic medications (aspirin and nonaspirin nonsteroidal anti-inflammatory drugs [NSAIDs], and acetaminophen) is associated with ovarian cancer survival. Each of these exposures affects a large, and growing, number of women globally. Aim The overall aim of this thesis was to address knowledge gaps in ovarian cancer epidemiology relating to several potential risk factors (adiposity, height, and hysterectomy) and exposures that may influence survival after diagnosis (common analgesic medications), using large complex datasets and non-traditional approaches. Methods The four studies included in this thesis each examined one exposure or several closely-related exposures in relation to ovarian cancer risk or survival, using two datasets and a variety of analytical methods. Two studies (examining adiposity and height in relation to ovarian cancer risk) applied Mendelian randomization (MR), a method which uses genetic variants as proxies for epidemiological exposures to overcome some of the biases that can affect observational research. Data were pooled from approximately 37,000 women (for adiposity analyses) or approximately 39,400 women (for height analyses) participating in 39 studies in the international Ovarian Cancer Association Consortium (OCAC). The third study used linked administrative data to follow approximately 840,000 Western Australian women over a 27-year period, and applied Cox proportional hazards regression with time-varying exposures and covariates to examine the association between hysterectomy and ovarian cancer risk. The final study used Cox proportional hazards regression to investigate the relationships between pre-diagnosis use of common analgesic medications and survival among approximately 7,700 women with ovarian cancer from 12 OCAC studies. ii Results The first study used MR to show that greater adiposity increases risk (about a 30% increase in risk per five units of body mass index) of ovarian cancers other than high-grade serous, but not the common and aggressive high-grade serous cancers. Abdominal adiposity (waist-hip ratio) was not associated with risk of either high-grade serous or non-high grade serous ovarian cancers. The second study indicated that greater height is associated with increased risk of invasive (a...