Construction of a prognostic prediction model for HER2-positive stage Ⅲ breast cancer with neoadjuvant therapy based on bioinformatics methods / 肿瘤研究与临床

ABSTRACT

Objective:To explore the factors affecting the prognosis of human epidermal growth factor receptor 2 (HER2)-positive stage Ⅲ breast cancer patients receiving neoadjuvant therapy (NAT) based on bioinformatics methods, and to construct a prognostic prediction model.Methods:Clinical data of HER2-positive stage Ⅲ breast cancer patients with NAT registered from January 2010 to December 2019 were collected using National Cancer Institute Surveillance, Epidemiology, and End Results (SEER) database*Stat 8.3.9 software. Data on general information and clinicopathological characteristics of patients were obtained. Patients were divided into modeling and validation groups using R 4.2.1 software createDataPartition tool in 7∶3. Cox proportional hazards model was used to analyze the influencing factors of breast cancer patient-specific survival (BCSS), and the BCSS prediction nomogram model was constructed using R 4.2.1 software survival package and rms package. The predictive efficacy of the nomogram model was assessed using consistency index (C-index), receiver operating characteristic (ROC) curve and area under the curve (AUC), and survival calibration curve; the optimal cut-off value of the prognostic-related scores was calculated using the X-tile software, and the survival curves were plotted using Kaplan-Meier method. Decision curve analysis (DCA) was used to observe the net benefit of the nomogram model.Results:A total of 1 790 HER2-positive stage Ⅲ breast cancer patients receiving NAT were enrolled, and they were randomly divided into a modeling group (1 250 patients) and a validation group (540 patients) according to the ratio of 7∶3, and the differences in the general information of patients between the two groups were not statistically significant (all P > 0.05). The result of multivariate Cox regression analysis showed that no radiotherapy, histologic grade Ⅲ-Ⅳ, estrogen receptor (ER) negative, progesterone receptor (PR) negative, stage T 4, lymph node metastasis rate of 1%-30%, lymph node metastasis rate of 31%-60%, and lymph node metastasis rate of > 61% were the independent risk factors for BCSS (all P < 0.05). The nomogram model for prognosis of HER2-positive stage Ⅲ breast cancer patients with NAT was constructed based on the above factors. After internal validation, the C-index of the modeling group was 0.732 (95% CI 0.724-0.739), and the C-index of the validation group was 0.728 (95% CI 0.721-0.737). The AUC of the modeling group for predicting the rates of BCSS at 3, 6 and 9 years was 0.724, 0.719 and 0.775, and the AUC of the validation group was 0.773, 0.699 and 0.758. The survival calibration curves showed that the 3-, 6- and 9-year BCSS rates in the modeling and validation groups were highly fitted to the actual observations. The total score of the nomogram model of patients included in the SEER database was calculated, and the optimal cut-off value for predicting prognosis were 80.0 points and 142.0 points. According to the optimal cut-off value, the patients were divided into low-risk group (total score ≤80.0 points), intermediate-risk group (total score > 80.0 points and ≤142.0 points) and high-risk group (total score >142.0 points), and they were also classified into stage Ⅲ A, stage Ⅲ B and stage Ⅲ C according to the American Joint Committee on Cancer (AJCC) staging, the differences in BSCC among these groups were statistically significant (all P < 0.001), and the nomogram model risk stratification was more effective in differentiating BCSS. The DCA curves showed that compared with AJCC staging and the two extremes of treatment modalities (complete treatment and no treatment at all), the nomogram model had a higher net gain in predicting BCSS of HER2-positive stage Ⅲ breast cancer patients with NAT. Conclusions:The prognostic factors affecting HER2-positive stage Ⅲ breast cancer patients with NAT are whether or not they are treated with radiotherapy, histologic grading, ER status, PR status, T stage, and lymph node metastasis rate. The constructed nomogram model has better efficacy and greater benefit than traditional AJCC staging in predicting BCSS.