Abstract P2-09-16: Comparison of prediction models of breast cancer in high risk populations?

Abstract

Abstract Breast cancer is the most frequent female cancer with 48 800 new cases each year in France. A French national screening program in women without genetic risk factors has been initiated in 2004. Women with BRCA 1 or 2 mutations are included in surveillance programs. Women without such mutations can be assessed for breast cancer risk using risk prediction models. The aim of this study is to compare the performance of three breast cancer risk prediction models: GAIL2, BODICEA and IBIS in a population of patients (pts) who developed breast cancer. Patients and Methods: The GAIL1 model was performed to 188 pts at high risk of breast cancer according to their family history and who developed breast cancer. Personal and familial data was regenerated at the day before the diagnosis of breast cancer and the three models were applied. From this population, 60 pts were selected: for whom all information necessary to use the 3 models was available and for whom the GAIL1 model provided sufficient variability in the relative risk of breast cancer. The GAIL2 model takes into account individual risk factors: age, menarches, parity, age at first birth, history of breast biopsy, atypical hyperplasia or in situ lobular carcinoma. The BODICEA model considers age and familial risk factors: number of related affected by breast ovarian, prostate and pancreas cancer and age at diagnosis. Both personal and familial risk factors are used for the IBIS model. We estimated lifetime breast cancer risk for each patient with the three models from the available software packages. We assessed Pearson's correlation coefficient between the three models. Results Median (range) age was 45 years (25-74). Risk prediction could not be evalauted by the GAIL2 model for 14 pts since they were less than 35 years of age. Lifetime breast cancer risk was 16.1% (4.4-38.7) for GAIL2, 11.6% (2.2-39.5) for BODICEA and 16.5% (3.5-36.3) for IBIS. Pearson's correlation coefficient between BODICEA and GAIL2, IBIS and GAIL2 and between BODICEA and IBIS were 0.36, 0.38 and 0.69 respectively. In most cases, IBIS risk predictions were higher than GAIL2's which were higher than BODICEA's as soon as lifetime risk was at least 20%. When the three models were applicable (46pts), IBIS estimated a higher risk in 31 cases (67%) versus 10 for GAIL2 (21.74%) and 5 for BODICEA (10.86%). The median (range) time for use of the tools per patient was 30 seconds (16-80) for GAIL2, 588 (198-1804) for BODICEA and 86 (46-135) for IBIS. Discussion Results in this selected population of pts who developed breast cancer show that IBIS seems to be the better performer to predict breast cancer risk:. Results are obtained faster and the risk predictions provide higher estimates. The GAIL2 model is quick and easy to use but with a limited number of items. Conversely, BODICEA requires a very large number of items, not always available, and does not consider incomplete data. In conclusion, IBIS model seems to be the most suitable for practical use in the evaluation of breast cancer risk. Citation Format: Mailliez A, Kramar A, Peyrat J-P, Revillion F, Bonneterre J. Comparison of prediction models of breast cancer in high risk populations?. [abstract]. In: Proceedings of the Thirty-Eighth Annual CTRC-AACR San Antonio Breast Cancer Symposium: 2015 Dec 8-12; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2016;76(4 Suppl):Abstract nr P2-09-16.