Abstract LB-327: Validation of a breast cancer risk prediction model in African American women

Existing breast cancer risk prediction models, such as the Breast Cancer Risk Assessment Tool (BCRAT), are less reliable for predicting risk in African American women than they are in white women. Gail et al. (2007) modified the BCRAT model for risk prediction of breast cancer in African American women based on data from the Women's Contraceptive and Reproductive Experiences (CARE) study; the CARE model uses information on a woman's age, age at menarche, family history of breast cancer, and number of previous breast biopsies to estimate absolute breast cancer risk. They validated the model among postmenopausal African American women aged 50 or older in the Women's Health Initiative and found it to be well calibrated but with limited discriminatory accuracy. We assessed calibration and discrimination of the CARE model in both pre- and postmenopausal women using data from the Black Women's Health Study (BWHS), an ongoing follow-up study of African American women 21-69 years of age at enrollment in 1995. We based our validation on 36,713 women who at baseline were 35 years and older, had no history of cancer, and had complete data on the risk factors in the model. We compared the expected number of invasive breast cancer cases predicted by the model to the observed number of cases identified in the BWHS overall and within subgroups of risk factors. We measured the discriminatory accuracy using the concordance statistic, the area under the receiver-operating curve. During a mean follow-up of 9.4 years, we identified 773 invasive breast cancer cases. The model predicted 691.9 cases, yielding an expected-to-observed ratio (E/O) of 0.90 (95% CI: 0.83, 0.96). Of the risk factors included in the CARE model, the E/O ratio was lowest among women with a previous biopsy (E/O = 0.76; 95% CI: 0.66, 0.88); the E/O ratio was close to 0.90 across categories of the other factors in the model. The CARE model does not include age at first birth, which was included in the BCRAT; we found the CARE model to be well calibrated among women with an age at first birth <25 years (E/O = 0.99; 95% CI: 0.90, 1.09) but not among women with an age at first birth ≥25 years (E/O = 0.73; 95% CI: 0.64, 0.83). In addition, risk was underestimated among women with a body mass index (BMI) at age 18 <20 kg/m*2 (E/O = 0.79; 95% CI: 0.71, 0.87). Results did not differ between premenopausal and postmenopausal women. The age-specific concordance statistic was 0.59 (95% CI: 0.54, 0.64), 0.58 (95% CI: 0.54, 0.61), 0.56 (95% CI: 0.52, 0.60), and 0.55 (95% CI: 0.49, 0.61) for women aged 35-39, 40-49, 50-59, and 60-69 years, respectively, which is comparable to the discriminatory accuracy reported in previous validations of the CARE and BCRAT models. The risk underestimation in particular subgroups of women suggests that the CARE model may be improved by the inclusion of age at first birth (as in the BCRAT) and perhaps other risk factors such as BMI at age 18.Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr LB-327. doi:1538-7445.AM2012-LB-327

Duke Scholars

Author Michael J Pencina Biostatistics & Bioinformatics, Division of Biostatistics