# Comprehensive Analysis and Prediction of HER2-Targeted Therapy Insensitivity Among HER2-Positive Breast Cancer Patients Undergoing Neoadjuvant Treatment

## TL;DR

This study identifies factors predicting poor response to HER2-targeted breast cancer treatment and develops a model to help clinicians tailor treatment strategies.

## Contribution

A predictive model using baseline clinical data to estimate neoadjuvant therapy sensitivity in HER2-positive breast cancer patients.

## Key findings

- 3660 (28.1%) of patients achieved pathological complete response.
- NAT-insensitive patients had worse overall survival and distinct clinical characteristics.
- The predictive model demonstrated reasonable performance with AUCs of 0.762 and 0.776 in training and validation cohorts.

## Abstract

Neoadjuvant therapy combining chemotherapy with HER2-targeted drugs has become a standard treatment for many patients with HER2-positive early-stage breast cancer. However, a proportion of patients have shown limited response to this treatment and may subsequently experience less favorable long-term outcomes. Identifying patients who are less likely to benefit before treatment begins could help clinicians optimize treatment strategies. In this study, we analyzed data from 13,004 patients with HER2-positive breast cancer who received neoadjuvant therapy using the National Cancer Database, a large nationwide oncology registry in the United States. We examined clinical and pathological factors associated with treatment responses and exploratorily constructed a prediction model using routinely available baseline information. Our findings have highlighted substantial variability in treatment sensitivity and could provide a practical tool that may help estimate the likelihood of treatment responses and facilitate baseline risk stratification.

Purpose: HER2-targeted therapy has been incorporated into the standard neoadjuvant treatment (NAT) regimen for HER2-positive early-stage breast cancer, yet a subset of patients have shown a limited pathological response. This study aimed to evaluate clinicopathological factors associated with NAT sensitivity and to develop a predictive model. Methods: This retrospective study included 13,004 HER2-positive breast cancer patients from the National Cancer Database (2010–2022) who received neoadjuvant chemotherapy plus HER2-targeted therapy. Pathological complete response (pCR) was defined as no residual invasive carcinoma in the breast and axillary lymph nodes (ypT0/is, ypN0). NAT sensitivity was additionally defined using clinical-to-pathologic stage migration according to the AJCC 8th edition criteria. Baseline characteristics and overall survival (OS) were compared between NAT-sensitive and NAT-insensitive groups. A multivariable logistic regression model was developed based on age, clinical T stage, clinical N stage, histologic subtype, tumor grade, and hormone receptor (HR) status. Model performance was assessed using the area under the receiver operating characteristic curve and calibration curves. Results: Among the patients included, 3660 (28.1%) achieved pCR. Based on the predefined stage-based criteria, 10,451 (80.4%) were classified as NAT-sensitive and 2553 (19.6%) as NAT-insensitive. NAT-insensitive patients were older and more likely to present with clinical T1c and node-negative disease, whereas NAT-sensitive patients more frequently had higher clinical T and N stages. HR-positive and lower tumor grades were significantly associated with treatment insensitivity. NAT-insensitive patients demonstrated significantly worse OS compared with NAT-sensitive patients (p < 0.001). The predictive model showed acceptable discrimination with AUCs of 0.762 in the training cohort and 0.776 in the validation cohort, demonstrating good calibration. Conclusions: NAT sensitivity in HER2-positive early-stage breast cancer exhibited substantial biological and clinical heterogeneity in real-world practice. A younger age, higher clinical stage, invasive ductal histology, higher tumor grade, and HR-negative status were associated with improved responses. A predictive model based on routinely available baseline variables demonstrated reasonable performance for estimating treatment sensitivity, supporting its potential utility for baseline risk stratification pending external validation.