Modification of Ovarian, Endometrial, Colorectal, and Breast Cancer risks associated with Oral Contraceptive use
This evidence comes from a 2013 Agency for Healthcare Research and Quality Evidence Report on “Oral Contraceptive Use for the Primary Prevention of Ovarian Cancer.” This report synthesized the findings from published literature on OC use and risks of ovarian, breast, cervical, endometrial, and colorectal cancer. Several data gaps were identified, including a need to understand cancer risks in subpopulations, such as those stratified by family history of cancer, smoking, or obesity.
To address this data gap, we will examine how OC use and duration of use is associated with risks of several cancers: breast, ovarian, endometrial, and colorectal and whether these associations are modified by factors which themselves are modifiable, such as baseline body mass index, smoking status, alcohol use, and physical activity. In estimating OC-related cancer risks by common lifestyle factors from the same base-population, we will enable clinicians, researchers, and individuals to understand and compare these risks within one paper. Ultimately, we hope to contribute toward a body of literature that will enable clinicians to better counsel their patients and make evidence-based practice decisions with regard to the risks and benefits of taking OCs.
Evaluating effect modification via stratification reduces sample sizes/increases variation within strata. As such, examining modification will require a large sample size to maintain power. Thus, we propose to pool the PLCO data with that from the NIH-AARP Diet and Health Study. Pooling these two cohorts to pursue these questions is ideal; aside from the large number of participants in these studies, both collected similar information on OC use and duration of use as well as on the effect modifiers of interest and important confounders (e.g., age, race); furthermore, baseline information was collected in the same calendar period and age range for both studies.
Previous researchers have commented on OC use and these cancer risks using PLCO and NIH-AARP data (references available). However, none of these analyses reported on modification of these relationships by any factors other than parity or use of menopausal hormone therapy. To our knowledge, researchers have not used this data to explore OC use duration and risks of colorectal or breast cancers. One study used pooled PLCO and AARP data to report on breast, endometrial, and ovarian cancer risks associated with OC use. However, this study was limited to white women over the age of 50, dichotomized OC duration, focused on predictive versus causal modeling, and did not report risks for breast cancer associated with this exposure (Pfeiffer et al., 2013).
Our primary aim is to evaluate whether OC use associations with breast, ovarian, endometrial, and colorectal cancers are modified by factors which themselves are modifiable, such as baseline body mass index, smoking status, alcohol use, and physical activity. We will also consider assessing effect modification by family history of a given cancer (if available) or parity, as these may be strong risk factors for the cancers of interest. Analyses using Cox proportional hazards regression will be conducted for each individual cancer outcome. We will briefly present the associations for each cancer risk associated with OC use and duration of use. Models will be appropriately adjusted for potential confounders for each cancer of interest. If an effect of OC use or duration of use is detected, we will explore modification by the modifiers listed above. If an effect is detected for both OC use and duration of use, we will focus on modification of the duration effect. We will test the statistical significance of the effect modification via a likelihood ratio test and present stratified results regardless of the significance of these tests.
Britton Trabert, Ph.D. (co-PI); MEB, DCEG, NCI
Louise Brinton, Ph.D.; DCEG, NCI
Ruth Pfeiffer, Ph.D.; DCEG, NCI