Personalized recurrence risk prediction in early-stage breast cancer through an integrative mathematical model based on MammaPrint®, radiotherapy, phenotype, and clinicopathological factors.

Purpose: To build a mathematical model with predictive capacity for the risk of recurrence in patients with early-stage breast cancer, based on four variables: the result of the MammaPrint®(MMP) test, postoperative radiotherapy (RT), tumor phenotype, and clinicopathological criteria. To estimate overall survival functions stratified by the dose of radiotherapy received.

Methods: A retrospective cohort of 156 patients with early-stage breast cancer was analyzed. Patients were classified according to their MammaPrint®genomic risk (ultralow, low, or high). Multivariate logistic regression using Firth's method was employed to evaluate the risk of recurrence, adjusting for biologically effective dose (BED), molecular subtype, their MammaPrint®classification and clinico-pathologic features. Receiver operating characteristic (ROC) analysis was used to assess model discrimination. The Kaplan-Meier method was used to estimate overall survival (OS) functions. To assess statistically significant differences in survival between patient groups, the log-rank test was applied.

Results: The predictive model, incorporating BED, genomic risk, molecular phenotype, and clinico-pathological classification, showed good calibration and discrimination (AUC: 0.755). The evaluation of OS according to the different BED levels provides clearer results regarding the clinical benefit of radiotherapy. This study reports statistically significant differences when comparing the group without radiotherapy (BED = 0 Gy) to the low-dose group (BED < 60 Gy), with a p-value of 0.0475.

Conclusion: The predictive model fitted using Firth's penalized logistic regression demonstrated an adequate discriminative ability (AUC = 0.755). MMP was the variable with the greatest weight, followed by RT. These variables allow for a more accurate prediction of recurrence risk than traditional clinicopathological factors, supporting their value in the personalization of treatment. This study reports statistically significant differences when comparing the group without radiotherapy (BED = 0 Gy) to the low-dose group (BED < 60 Gy), with a p-value of 0.0475.