An MRI-based multi-parameter clinical decision support system (NeoMDSS) for early prediction of pathological complete response after the first cycle of neoadjuvant therapy in breast cancer: A multi-center prospective observational cohort study.

Abstract

592 Background: Currently, there is no reliable and convenient way to predict response rates in early stages of neoadjuvant therapy (NAT) for breast cancer, which can affect the prognosis of non-responsive patients. Methods: In this prospective multi-center observational study, eligible participants who underwent NAT followed by radical surgery were recruited. All patients underwent magnetic resonance imaging (MRI) examination before and after the first cycle of NAT (1st-NAT). We analyzed the effects of clinicopathological features, such as histological grade, hormone receptor (HR) and HER2 status, Ki-67 expression, blood cell analysis, and MRI parameters including the reduction ratio of tumor diameter (ΔD%), apparent diffusion coefficient (ADC), and early enhancement ratio (EER) on NAT. By incorporating independent influencing factors, we developed a multiparametric clinical decision support system (NeoMDSS) model based on a retrospective cohort and validated its accuracy in both a prospective single-center internal cohort and a prospective multi-center external cohort. Relevant clinical trial details can be found under the identifier NCT04909554. Results: A total of 301 breast cancer patients were enrolled between January 2019 and December 2023, including 140 patients in the training cohort, 120 patients in the internal validation cohort, and 41 patients in the external validation cohort. The NeoMDSS model demonstrated excellent performance in predicting pathological complete response (pCR) at 1st-NAT, with an AUC of 0.874 (95% CI 0.813-0.935) in the training cohort, 0.845 (95% CI 0.771-0.919) in the internal validation cohort, and 0.867 (95% CI 0.742-0.992) in the external validation cohort. The NeoMDSS model exhibited a sensitivity, specificity, and accuracy of 80.0%, 81.3%, and 80.5%, respectively, in the internal validation cohort, and 87.0%, 83.3%, and 85.4%, respectively, in the external validation cohort. Calibration curves and decision curve analysis (DCA) further supported the clinical value of the NeoMDSS model. To facilitate the clinical application of the NeoMDSS model, a nomogram and an online website ( www.gdphneomdss.com ) were developed to calculate the probability of pCR, which the sensitivity, specificity, and accuracy of the NeoMDSS model were all 80.0% in the internal validation cohort and 80.0%, 81.3%, and 80.5% in the external validation cohort. Conclusions: The NeoMDSS model serves as a convenient tool in clinical practice at 1st-NAT to calculate and accurately predict the probability of pCR after NAT. This assists clinicians in determining whether to modify the treatment plan and improving personalized treatment for patients with poor response to NAT. Clinical trial information: NCT04909554 .