Twenty-five years of breast cancer risk models and their applications

Abstract

I would like to thank the editors for the invitation to comment on “Projecting Individualized Probabilities of Developing Breast Cancer for White Females Who Are Being Examined Annually” ( 1 ) and to recognize coauthors of that article, Louise A. Brinton, David P. Byar, Donald K. Corle, Sylvan B. Green, Catherine Schairer, and John J. Mulvihill, who contributed key insights and analyses. Several features may explain the frequent citations to this paper. 1) It focused on a clinically crucial quantity, absolute risk, namely the probability that a woman with specific risk factors will develop breast cancer over a defined age interval, with allowance for competing risks. 2) Breast cancer is a common cancer for which preventive interventions have been developed. Risk models are most useful in connection with interventions because the absolute risk of breast cancer can be compared with that of other health outcomes in the presence and absence of intervention. 3) The model is simple: only age and answers to five questions about reproductive, family, and medical history are needed. 4) The model (sometimes called the “Gail model”) has been taken up by practitioners and is available at http://www.cancer.gov/bcrisktool/ as the National Cancer Institute’s (NCI’s) Breast Cancer Risk Assessment Tool (BCRAT). Currently this site is visited over three million times a year. I shall elaborate on the first two points, mention alternative models, and discuss prospects for improving risk models and other applications of absolute risk. Absolute risk is the probability that a woman with specific risk factors but without breast cancer at age a will be diagnosed with breast cancer in the age interval a<T≤a+τ , where τ is the duration of the risk projection interval. Consider a 40-year-old white woman who began menstruating at age 12 years, whose first live birth was at age 25 years, who has had no biopsies, and whose mother developed breast cancer. From BCRAT, her absolute risk of breast cancer in five years is 1.1% and her risk to age 90 years (“lifetime risk”) is 18.8%. In contrast, the average absolute risks for a 40-year-old white woman in the NCI’s Surveillance, Epidemiology and End Results (SEER) Program are, respectively, 0.6% and 12.4%. These absolute risk estimates are reduced by the chance that the woman might die from a competing cause of mortality before breast cancer develops. In the competing risks literature, the terms “crude risk” ( 2 ) and “cumulative incidence” ( 3 ) are sometimes used instead of absolute risk. Absolute risk is not the same as the “pure” cumulative risk that is often used for risk models, such as genetic models of breast cancer risk ( 4 , 5 ). Pure risk is the hypothetical risk of developing breast cancer over a defined time interval if there were no competing causes of mortality. Pure risk is therefore higher than absolute risk. For example, based on SEER rates in white women from 2007 to 2011, the absolute risk to age 90 years is 12.28%, but the pure risk is 15.74%, which is 28% higher. For short projection intervals, such as five years, competing mortality typically has little effect and pure risks are only slightly higher than absolute risks. But for longer intervals, absolute risk is the relevant quantity, because in fact women are subject to competing mortality risk. One minus the Kaplan-Meier estimate of the probability of remaining breast cancer–free estimates pure risk. Special calculations ( 1 , 6 ) are needed to estimate absolute risk. Risk models are used for counseling individual women and for developing and implementing public health prevention strategies. One of the most valuable uses of risk models is to give a woman a realistic risk estimate and perspective on how large this risk is in comparison with other risks. Indeed, BCRAT was initially developed for this purpose. Dr. Mulvihill, one of the authors of the 1989 article, thought that many of the women he was advising in a high-risk breast cancer clinic had unrealistically high perceptions of risk that could lead to ill-advised procedures, such as prophylactic mastectomy. Such perspective can also affect less drastic but nonetheless important management decisions. There has been disagreement over whether women in their forties should have routine mammographic screening. The American Cancer Society recommended it ( 7 ), whereas the US Preventive Services Task Force (USPSTF) wrote ( 8 ), “USPSTF recommends against routine screening mammography in women aged 40 to 49 years. The decision to start regular, biennial screening mammography before the age of 50 years should be an individual one and take into account patient context, including the patient’s values regarding specific benefits and harms.” This report disregarded risk assessment based on “demographic, physical, or historical risk factors,” stating that, “none conveys clinically important absolute increased risk for cancer” ( 8 ). Risk assessment can help women who are confuse