A multi-state survival model to identify risk factors for lethal ovarian cancer

Abstract

Background: Given primary prevention strategies for ovarian cancer, such as surgery and medications, have inherent risks, identifying those at high risk of lethal ovarian cancer is critical. We examined pre-diagnosis factors and risk of developing and dying from ovarian cancer among cancer-free women. Methods: Analyses were conducted in three 12-year periods from 1980 to 2017 in the Nurses’ Health Study (NHS) and NHSII cohorts. Potential risk factors were reproductive and hormonal variables, endometriosis history, smoking, low-dose aspirin, self-identified race, family history, depression, and adiposity over the life course. We used a multi-state survival model to estimate relative risks and 95% lower and upper confidence limits (RR, LCL-UCL) for lethal ovarian cancer among 211,420 cancer-free women, among whom 1,730 developed ovarian cancer and 660 died due to ovarian cancer in the same risk period as diagnosis. Results: Of the 22 exposures evaluated, 10 were associated with lethal ovarian cancer. For example, nulliparity had an amplified association with lethal ovarian cancer (1.62, 1.23-2.13) due to associations with both incidence and mortality in the same direction. Oral contraceptive use ≥10 versus 0 years was associated with lethal ovarian cancer (0.65, 0.43-0.97) primarily due to association with incidence while ≥20 versus 0 pack-years of smoking was associated with lethal ovarian cancer (1.25, 1.02-1.53) primarily due to the mortality relationship. Conclusions: Several reproductive factors, depression, and self-identified race were associated with risk of lethal ovarian cancer. Impact: Evaluations of lethal ovarian cancer risk must consider differential associations of exposures with incidence and mortality.