Introduction: This study was designed to compare the Breast Cancer Risk Assessment Tool (BCRAT; Gail), International Breast Intervention Study (IBIS; Tyrer-Cuzick), and BRCAPRO breast cancer risk assessment models using data from the Marin Women's Study, a cohort of women within Marin County, California, with high rates of breast cancer, nulliparity, and delayed childbirth. Existing models have not been well-validated in these high-risk populations.
Methods: Discrimination was assessed using the area under the receiver operating characteristic curve (AUC) and calibration by estimating the ratio of expected-to-observed (E/O) cases. The models were assessed using data from 12,843 participants, of whom 203 had developed cancer during a 5-year period. All tests of statistical significance were 2-sided.
Results: The IBIS model achieved an AUC of 0.65 (95% confidence interval [CI], 0.61-0.68) compared with 0.62 (95% CI, 0.59-0.66) for BCRAT and 0.60 (95% CI, 0.56-0.63) for BRCAPRO. The corresponding estimated E/O ratios for the models were 1.08 (95% CI, 0.95-1.25), 0.81 (95% CI, 0.71-0.93), and 0.59 (95% CI, 0.52-0.68). In women with age at first birth > 30 years, the AUC for the IBIS, BCRAT, and BRCAPRO models was 0.69 (95% CI, 0.62-0.75), 0.63 (95% CI, 0.56-0.70), and 0.62 (95% CI, 0.56-0.68) and the E/O ratio was 1.15 (95% CI, 0.89-1.47), 0.81 (95% CI, 0.63-1.05), and 0.53 (95% CI, 0.41-0.68), respectively.
Conclusions: The IBIS model was well calibrated for the high-risk Marin mammography population and demonstrated the best calibration of the 3 models in nulliparous women. The IBIS model also achieved the greatest overall discrimination and displayed superior discrimination for women with age at first birth > 30 years.
Keywords: Breast cancer risk prediction; Calibration; Discrimination; Nulliparity; Risk factors.
Copyright © 2014 Elsevier Inc. All rights reserved.