Comparative study of pathological response evaluation systems after neoadjuvant chemotherapy for breast cancer: developing predictive models of multimodal ultrasound features including shear wave elastography combined with puncture pathology

Ying Duan,Xuemei Song,Ling Guan,Weili Wang,Bo Song,Yaqiong Kang,Yingying Jia,Yangyang Zhu,Fang Nie

DOI: https://doi.org/10.21037/qims-22-910

2023-04-26

Quantitative Imaging in Medicine and Surgery

Abstract:Background: This study created a predictive preoperative nomogram dependent on multimodal ultrasound characteristics and primary lesion biopsy results for various pathologic response assessment systems following neoadjuvant chemotherapy (NAC).Methods: This retrospective study included 145 breast cancer patients treated at Gansu Cancer Hospital between January 2021 and June 2022 who underwent shear wave elastography (SWE) prior to completing NAC. Intra- and peritumoral SWE features, including maximum (E_max), mean (E_mean), minimum (E_min), and standard deviation (E_sd) elasticity, were measured individually and linked with the Miller-Payne grading system and residual cancer burden (RCB) class. Univariate analysis was used for conventional ultrasound and puncture pathology. Binary logistic regression analysis was used to screen for independent risk factors and to develop a prediction model.Results: Intratumor E_mean and peritumoral E_sd differed significantly from the Miller-Payne grade [intratumor E_mean: r=0.129, 95% confidence interval (CI): -0.002 to 0.260; P=0.042; peritumoral E_sd: r=0.126, 95% CI: -0.010 to 0.254; P=0.047], RCB class (intratumor E_mean: r=-0.184, 95% CI: -0.318 to -0.047; P=0.004; peritumoral E_sd: r=-0.139, 95% CI: -0.265 to 0.000; P=0.029) and RCB score components (r=-0.277 to -0.139; P=0.001-0.041). Two prediction model nomograms-pathologic complete response (pCR)/non-pCR and good responder/nonresponder-for the RCB class were developed using binary logistic regression analysis for all significant variables in SWE, conventional ultrasound, and puncture results. The area under the receiver operating characteristic curve for the pCR/non-pCR and good responder/nonresponder models was 0.855 (95% CI: 0.787-0.922) and 0.845 (95% CI: 0.780-0.910), respectively. According to the calibration curve, the nomogram had excellent internal consistency between estimated and actual values.Conclusions: The preoperative nomogram can effectively guide clinicians to predict pathological response of breast cancer after NAC and has the potential to guide individualized treatment.

radiology, nuclear medicine & medical imaging