Publications

In November 2013, the U.S. Preventive Services Task Force issued a guideline on medications for risk reduction of primary breast cancer in women. Although mammography can detect early cases, it cannot prevent development of breast cancer. Tamoxifen and raloxifene are selective estrogen receptor modulators that have been shown to reduce the risk for estrogen receptor-positive breast cancer and are approved by the U.S. Food and Drug Administration (FDA) for this indication. However, neither medication reduces the risk for estrogen receptor-negative breast cancer or all-cause mortality. The Task Force concluded that postmenopausal women with an estimated 5-year risk for breast cancer of 3% or greater will probably have more net benefit than harm and recommends that clinicians engage in shared, informed decision making about these medications. The American Society of Clinical Oncology issued a practice guideline on use of pharmacologic interventions for breast cancer in 2013. It recommends that women aged 35 years or older at increased risk, defined as a 5-year absolute risk for breast cancer of 1.66% or greater, discuss breast cancer prevention medications with their primary care practitioner. The Society includes the aromatase inhibitor exemestane in addition to tamoxifen and raloxifene as a breast cancer prevention medication, although exemestane is not FDA approved for this indication. Here, an oncologist and an internist discuss how they would balance these recommendations and what they would suggest for an individual patient.

BACKGROUND: The Breast Cancer Risk Assessment Tool (BCRAT, "Gail model") is commonly used for breast cancer prediction; however, it has not been validated for women age 75 years and older.

METHODS: We used Nurses' Health Study (NHS) data beginning in 2004 and Women's Health Initiative (WHI) data beginning in 2005 to compare BCRAT's performance among women age 75 years and older with that in women age 55 to 74 years in predicting five-year breast cancer incidence. BCRAT risk factors include: age, race/ethnicity, age at menarche, age at first birth, family history, history of benign breast biopsy, and atypia. We examined BCRAT's calibration by age by comparing expected/observed (E/O) ratios of breast cancer incidence. We examined discrimination by computing c-statistics for the model by age. All statistical tests were two-sided.

RESULTS: Seventy-three thousand seventy-two NHS and 97 081 WHI women participated. NHS participants were more likely to be non-Hispanic white (96.2% vs 84.7% in WHI, P < .001) and were less likely to develop breast cancer (1.8% vs 2.0%, P = .02). E/O ratios by age in NHS were 1.16 (95% confidence interval [CI] = 1.09 to 1.23, age 57-74 years) and 1.31 (95% CI = 1.18 to 1.45, age ≥ 75 years, P = .02), and in WHI 1.03 (95% CI = 0.97 to 1.09, age 55-74 years) and 1.10 (95% CI = 1.00 to 1.21, age ≥ 75 years, P = .21). E/O ratio 95% confidence intervals crossed one among women age 75 years and older when samples were limited to women who underwent mammography and were without significant illness. C-statistics ranged between 0.56 and 0.58 in both cohorts regardless of age.

CONCLUSIONS: BCRAT accurately predicted breast cancer for women age 75 years and older who underwent mammography and were without significant illness but had modest discrimination. Models that consider individual competing risks of non-breast cancer death may improve breast cancer risk prediction for older women.

PURPOSE: Accurate risk assessment is necessary for decision-making around breast cancer prevention. We aimed to develop a breast cancer prediction model for postmenopausal women that would take into account their individualized competing risk of non-breast cancer death.

METHODS: We included 73,066 women who completed the 2004 Nurses' Health Study (NHS) questionnaire (all ≥57 years) and followed participants until May 2014. We considered 17 breast cancer risk factors (health behaviors, demographics, family history, reproductive factors) and 7 risk factors for non-breast cancer death (comorbidities, functional dependency) and mammography use. We used competing risk regression to identify factors independently associated with breast cancer. We validated the final model by examining calibration (expected-to-observed ratio of breast cancer incidence, E/O) and discrimination (c-statistic) using 74,887 subjects from the Women's Health Initiative Extension Study (WHI-ES; all were ≥55 years and followed for 5 years).

RESULTS: Within 5 years, 1.8 % of NHS participants were diagnosed with breast cancer (vs. 2.0 % in WHI-ES, p = 0.02), and 6.6 % experienced non-breast cancer death (vs. 5.2 % in WHI-ES, p < 0.001). Using a model selection procedure which incorporated the Akaike Information Criterion, c-statistic, statistical significance, and clinical judgement, our final model included 9 breast cancer risk factors, 5 comorbidities, functional dependency, and mammography use. The model's c-statistic was 0.61 (95 % CI [0.60-0.63]) in NHS and 0.57 (0.55-0.58) in WHI-ES. On average, our model under predicted breast cancer in WHI-ES (E/O 0.92 [0.88-0.97]).

CONCLUSIONS: We developed a novel prediction model that factors in postmenopausal women's individualized competing risks of non-breast cancer death when estimating breast cancer risk.

The population is aging, and breast cancer incidence increases with age, peaking between the ages of 75 and 79. However, it is not known whether mammography screening helps women aged 75 and older live longer because they have not been included in randomized controlled trials evaluating mammography screening. Guidelines recommend that older women with less than a 10-year life expectancy not be screened because it takes approximately 10 years before a screen-detected breast cancer may affect an older woman's survival. Guidelines recommend that clinicians discuss the benefits and risks of screening with women aged 75 and older with a life expectancy of 10 years or longer to help them elicit their values and preferences. It is estimated that two of 1,000 women who continue to be screened every other year from age 70 to 79 may avoid breast cancer death, but 12% to 27% of these women will experience a false-positive test, and 10% to 20% of women who experience a false-positive test will undergo a breast biopsy. In addition, approximately 30% of screen-detected cancers would not otherwise have shown up in an older woman's lifetime, yet nearly all older women undergo treatment for these breast cancers, and the risks of treatment increase with age. To inform decision-making, tools are available to estimate life expectancy and to educate older women about the benefits and harms of mammography screening. Guides are also available to help clinicians discuss stopping screening with older women with less than a 10-year life expectancy. Ideally, screening decisions would consider an older woman's life expectancy, breast cancer risk, and her values and preferences.

Guidelines for optimal cancer screening in older adults remain unclear, particularly for adults over the age of 75. While cancer screening in older adults may benefit some in good health, it may cause unnecessary burdens in others with limited life expectancy. Thus, a systematic approach to enable individualized cancer screening decisions in older adults is needed. We suggest a framework that guides such decisions through evidence-based approaches from multiple interactions, and that involves the patient, clinician, and healthcare system. An individualized approach considers differences in disease risk rather than the chronological age of the patient. This paper presents a comprehensive framework that depicts the independent and converging levels of influences on individualized cancer screening decisions in older adults. This Individualized Decisions for Screening (IDS) framework recognizes the reality of these interrelationships, including the tensions that arise when behaviors and outcomes are valued differently at the patient, clinician, and healthcare organization levels. Person-centered approaches are essential to advancing multilevel research of individualized cancer screening decisions among older adults.

OBJECTIVES: To examine receipt of colorectal cancer (CRC) screening according to age and life expectancy (LE) in adults aged 65 and older.

DESIGN: Population-based survey.

SETTING: United States.

PARTICIPANTS: Community dwelling adults aged 65 and older who participated in the 2008 or 2010 National Health Interview Survey (N = 7,747).

MEASUREMENTS: Receipt of CRC screening (e.g., colonoscopy within 10 years) was examined according to age and LE (≥10 and <10 years), adjusting for sociodemographic characteristics and survey year. Frequency of CRC screening was also examined according to age and LE at time of screening (e.g., age at colonoscopy rather than at interview). Participants screened when they were aged 75 and older or had less than a 10-year LE were considered to have received screening inconsistent with guidelines.

RESULTS: Overall, 38.5% of participants had less than a 10-year LE; 40.2% were aged 75 and older, and 56.3% had received recent CRC screening (90.1% by colonoscopy). CRC screening was higher in 2010 (58.9%) than 2008 (53.7%, P <.001) and was associated with longer LE and younger age, although 51.1% of adults aged 75 and older reported receiving CRC screening, as did 50.9% of adults with less than a 10-year LE. Based on age and LE at time of screening (rather than at interview), 28.4% of CRC screening of adults aged 65 and older was targeted to those aged 75 and older and those with less than a 10-year LE. Of adults aged 65 to 75 with a 10-year LE or more (adults recommended for screening by guidelines), 39.2% had not recently been screened.

CONCLUSION: Older adults with little chance of benefit because of limited LE commonly undergo CRC screening, whereas many adults aged 65 to 75 with a 10-year LE or greater are not screened.