Although screening for breast cancer is broadly utilized in the United States, the optimal approach in terms of examination modality, frequency, and starting age is unclear. Furthermore, there is no consensus on the benefit from earlier, more frequent, or additional modes of screening for women with high-risk for breast cancer. This thesis explores methods for finding optimal screening strategies by modeling the costs and benefits of breast cancer screening programs through a simulation-based and a theoretical approach, which incorporate some ignored or unexplored components. For the general population, cost-effectiveness studies evaluating breast cancer screening have focused on mammography alone, and have not assessed the impact of its combination with clinical breast exam, a cheaper and more practical exam. Costs incurred beyond screening including costs for work-up, biopsy, and treatment have often been ignored. We account for these factors in a comprehensive microsimulation analysis by modeling the natural history of the disease including age-specific incidence, sojourn time, and exam sensitivity and specificity, and evaluate the cost-effectiveness of a set of screening programs on a simulated cohort of women. Similarly, we focus on a special case of high-risk women where we include screening programs that begin at an earlier age, have more frequent examinations, and include combinations of magnetic-resonance imaging, which has recently been recommended for women at high-risk for breast cancer. The most cost-effective approach depends on how much society is willing to pay to save a year of life. Finally, we take a theoretical approach which is motivated by the idea that other optimal screening programs may exist outside a set of predetermined strategies used in a simulation-based approach. While empirical studies give a better idea of which screening schedules are useful, a theoretical approach may uncover a true optimal schedule. We assume a nonstable disease model and account for costs of screening and dollar value of benefit in a utility function under two different frameworks to estimate the optimal number or ages of examinations given certain general assumptions. We show that a solution exists for these models and demonstrate in numerical studies that they will give reasonable results.
