Accuracy of the Breast Cancer Surveillance Consortium Model Among Women with LCIS

Abstract

Purpose: The Breast Cancer Surveillance Consortium (BCSC) model predicts risk of invasive breast cancer risk based on age, race, family history, breast density, and history of benign breast disease, including lobular carcinoma in situ (LCIS). However, validation studies for this model included few women with LCIS. We sought to evaluate the accuracy of the BCSC model among this cohort.

Methods: Women with LCIS diagnosed between 1983 and 2017 were identified from a prospectively maintained database. The BCSC score was calculated; those with prior breast cancer, unknown breast density, age < 35 years or > 74 years, or with history of chemoprevention use were excluded. The Kaplan-Meier method was used to estimate incidence rates. Time-dependent receiver operating characteristic (ROC) analysis was used to analyze the discriminative capacity of the model.

Results: 1302 women with LCIS were included. At a median follow-up of 7 years, 152 women (12%) developed invasive cancer (6 with bilateral disease). Cumulative incidences of invasive breast cancer were 7.1% (95% CI 5.6-8.7) and 13.3% (95% CI 10.9-15.6), respectively, and the median BCSC risk scores were 4.9 and 10.4, respectively, at 5 and 10 years. The median 10-year BCSC score was significantly lower than the 10-year Tyrer-Cuzick score (10.4 vs 20.8, p < 0.001). The ROC curve scores (AUC) for BCSC at 5 and 10 years were 0.59 (95% CI 0.52-0.66) and 0.58 (95% CI 0.52-0.64), respectively.

Conclusion: The BCSC model has moderate accuracy in predicting invasive breast cancer risk among women with LCIS with fair discrimination for risk prediction between individuals.

Keywords: Breast Cancer Surveillance Consortium (BCSC) model; Breast cancer; Lobular carcinoma in situ (LCIS); Risk prediction.

Fig. 1

Cumulative incidence of invasive breast cancer. The 5– and 10– year cumulative incidences of invasive cancer were 7.1% (95% CI 5.6–8.7) and 13.3% (95% CI 10.9–15.6), respectively

Fig. 2

Distribution of 5-year and 10-year BCSC scores. The median 5- and 10-year BCSC risk scores were 4.9 and 10.4, respectively

Fig. 3

Comparison of 10-year Tyrer-Cuzick and BCSC scores. The median 10-year risk score was significantly lower with the BCSC model versus the Tyrer-Cuzick model (10.4 vs 20.8, respectively)

Fig. 4

Calibration of the BCSC risk score for predicting 5- and 10-year invasive cancer risk. Most Kaplan–Meier estimates were not significantly different from the BCSC model estimates, as evidenced by the intersection of the Kaplan–Meier confidence intervals with the line of equivalence

Fig. 5

Receiver Operating Characteristic curve for 5-year and 10-year BCSC scores. Graphical representation of ROC model at 5 years with AUC of 0.52 (95% CI 0.52– 0.66) and at 10 years with AUC of 0.58 (95% CI 0.52–0.64), respectively