DESIGNING RISK-TAILORED BREAST CANCER SCREENING POLICIES CONSIDERING IMPERFECT ADHERENCE

Monday, October 25, 2010
Sheraton Hall E/F (Sheraton Centre Toronto Hotel)
Turgay Ayer, MS1, Oguzhan Alagoz, PhD1, Natasha K. Stout, Ph.D.2 and Elizabeth S. Burnside, MD, MPH, MS3, (1)University of Wisconsin, Madison, WI, (2)Department of Population Medicine, Boston, MA, (3)University of Wisconsin-Madison, Madison, WI

Purpose:   Mammography screening has resulted in improvements in breast cancer mortality; however, optimal screening policies to achieve maximal benefit for individual women are difficult to formulate due to the complexities of the problem. Further, differences in adherence behavior to screening strategies, which may alter the benefit of a strategy, are rarely considered. The purpose of this study is to analyze the effects of imperfect adherence on individual risk-tailored mammography screening policies using a decision-analytic modeling framework.

Method: We formulate a finite-horizon partially observable Markov decision process (POMDP) model to examine optimal risk-tailored screening policies that consider different adherence behaviors. Model inputs include state-transition probabilities and rewards, estimated using the University of Wisconsin Breast Cancer Simulation. Our POMDP model incorporates unobservable disease progression, two methods of detection (self or screen), mammography test characteristics and variable levels of adherence for women.

Result: We demonstrate that optimal screening decisions are determined by the trade-off between a woman's breast cancer risk and her likely adherence to screening recommendations. Compared with many existing guidelines, our analysis recommends significantly less frequent screening for low-risk women with a high likelihood of adherence. For example, consider two 40-year-old women who have identical current breast risks (0.4%) but different adherence rates (60% and 90%). We find that the optimal screening policy for the woman with low adherence rate is biennial screening until age 50 as long as her cancer risk is lower than 1.5%. On the other hand, the optimal screening policy for the woman with high likelihood of adherence is to screen every 3 years until age 45 as long as her cancer risk is lower than 0.5%, and biennial screening between ages 45-50 as long as her cancer risk is lower than 2.6%. Our results show that increasing adherence rate from 60% to 90% in a low-risk woman would result in approximately 11 days of life-savings. These savings increase with the increased risk.

Conclusion: In addition to tailoring screening by breast cancer risk, consideration of a woman’s likelihood of adherence to a screening recommendation might significantly increase the effectiveness of a breast cancer screening program. Designing interventions to improve adherence has the potential to increase women’s health and investigating their cost-effectiveness is a next step of our research.

Candidate for the Lee B. Lusted Student Prize Competition