Book Description

Breast cancer is the most common non-skin cancer and a leading cause of death for U.S. women population. Depending on the personal and hereditary factors, each woman has a different risk of developing a breast cancer and the focus of recent studies has been on identifying the high-risk patients and reducing the mortality and adverse effects due to a breast cancer. For high-risk women, we can prevent breast cancer mortality either by aggressive screening and early detection or by taking preventative treatments such as hormonal therapy (i.e. Tamoxifen, Raloxifene, Exemestane) or a risk reduction surgery (i.e. Bilateral Prophylactic Mastectomy). In Chapter 3, we focus on the problem faced by clinicians and patients who are at higher risk of breast cancer: should a high-risk patient choose to go for a preventative treatment given her current risk estimates, age and treatment history? We formulate this optimal decision-making problem for preventing breast cancer as a discrete time, finite-horizon Markov Decision Process (MDP) model. The objective of this model is to maximize the total expected quality-adjusted life years (QALYs) for a patient while assuming that the decision makers are risk neutral. We use data from literature and Surveillance Epidemiology and End Results (SEER) database for estimating the model parameters. The interventions we consider in the model are tamoxifen and mastectomy for pre-menopausal women and tamoxifen, raloxifene, exemestane and mastectomy for post-menopausal women. The optimal policy in the preliminary results shows that the patients who have 47% or more risk of breast cancer should have a mastectomy if they are between 35-40 years old and have no previous history of preventative treatment. The risk threshold goes up with age for the mastectomy decisions. Among the pharmacological interventions, tamoxifen is optimal for pre-menopausal women around the age of 40 with a lifetime risk of 39% or above which also corresponds to a 5-year risk of 1.6% and above around the age of 40. Exemestane is optimal for post-menopausal women with a lifetime risk of 13% (corresponding 5-year risk of 1%) or above after age of 51. Raloxifene is optimal only when we have elevated risks after age 65 or older. In Chapter 4, we consider optimal breast cancer risk reduction policies when direct and indirect costs associated with the treatments are considered. Although exemestane is the dominant recommendation in QALY maximizing model, the results of the cost model show that it is not cost-efficient up to willingness-to-pay (WTP) ratio of $150,000. We also demonstrated that the optimal policy is very sensitive to the discount factor from which most preventative interventions suffer. In Chapter 5, we investigate the effect of risk-averse behavior on the optimal policies for breast cancer risk reduction. We use two main approaches to model risk aversion. First, we use utility functions with different forms to see the effect of aversion from uncertainty. In the second approach, we utilize dynamic risk measures and explore cases in which aversion is from the cancer risk. While purely risk-averse utility functions lead to myopic decision making, increasing risk-aversion coefficient for dynamic risk measures result in more aggressive treatment policies.