Related Experiment Video
Updated: Jun 22, 2026
Establishing a Competing Risk Regression Nomogram Model for Survival Data
Published on: October 23, 2020
Multicentric Validation of a Complete Blood Count-Based Model for Breast Cancer Risk Stratification in Women Age
Daniella Castro Araújo1, Bruno Aragão Rocha2, Suzylaine da Silva Lima1
1Huna Ltd, São Paulo, Brazil.
Purpose:
Recent changes in breast cancer (BC) screening guidelines have extended eligibility to women age 40-49 years, yet risk stratification in this group remains challenging. This study evaluated the performance of a complete blood count (CBC)-based BC risk model, originally developed for women age 40-70 years, in women age 40-49 years across diverse clinical settings in Brazil.
Methods:
In this retrospective multicenter study, CBC data from 239,089 women age 40-49 years were analyzed. A ridge regression model using age, neutrophil-to-lymphocyte ratio (NLR), and RBC was evaluated. CBCs were stratified by timing relative to breast diagnostic procedures: same day, 1-90 days prior, and 91-180 days prior. Model performance was assessed using area under the receiver operating characteristic curve (AUC), sensitivity, and specificity, with leave-one-group-out cross-validation across five Brazilian clinical settings.
Results:
The CBC model showed stable performance when CBCs were obtained within 3 months before diagnosis (AUC = 0.63 ± 0.03 for same day; 0.61 ± 0.05 for 1-90 days prior). Performance declined in the 91- to 180-day interval (AUC = 0.58 ± 0.03). NLR increased significantly closer to diagnosis, driven mainly by neutrophilia. At 70% specificity, screening the top 30% identified 47% of cancers (relative risk = 2.07).
Conclusion:
The CBC-based model provides a low-cost approach to BC risk stratification in women age 40-49 years, particularly when CBCs are collected within 3 months of evaluation. This opportunistic strategy may be especially valuable in resource-constrained settings where routine mammographic screening is limited.
