Scholarlink

<p>The binary pathological complete response (pCR) status of triple-negative breast cancer (TNBC) patients after neoadjuvant chemotherapy (NAC) inadequately captures risk heterogeneity for prognosis. Therefore, this study aimed to develop a machine learning-based survival model to predict individualized overall survival (OS) in TNBC patients with pCR after NAC. This retrospective study utilized a Dutch nationwide cohort of TNBC patients, diagnosed between 2013 and 2021. We trained a survival model on the data using nested 5-fold cross-validation. Patients were stratified into a low- and high-risk group, based on the predicted survival probabilities. A total of 2642 patients with a pCR post-NAC were included, of whom 2620 were used in the final analysis. The survival model achieved good discrimination (C-index 0.754). Risk stratification identified a high-risk subgroup (<i>n</i> = 495, 18.9%) with significantly worse survival compared to the low-risk group (5-year OS 85.2% vs. 96.7%, log-rank <i>P</i> &lt; 0.001). Pre-treatment nodal status (cN) was the strongest predictor, followed by age and clinical tumor stage (cT). This nationwide analysis demonstrates that TNBC patients achieving pCR remain a heterogeneous group with distinct survival outcomes. Machine learning-based survival modeling may enable individualized risk prediction using routinely collected clinical variables and may guide future post-surgical treatment escalation for high-risk TNBC patients despite pCR.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine learning-based identification of high-risk triple-negative breast cancer patients despite pathological complete response after neoadjuvant chemotherapy

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年6月2日

查看全文
  • Koen Kwakkenbos,
  • Nadine S. van den Ende,
  • Bernadette A. M. Heemskerk-Gerritsen,
  • Jifke F. Veenland,
  • Agnes Jager,
  • Carolien H. M. van Deurzen

摘要

The binary pathological complete response (pCR) status of triple-negative breast cancer (TNBC) patients after neoadjuvant chemotherapy (NAC) inadequately captures risk heterogeneity for prognosis. Therefore, this study aimed to develop a machine learning-based survival model to predict individualized overall survival (OS) in TNBC patients with pCR after NAC. This retrospective study utilized a Dutch nationwide cohort of TNBC patients, diagnosed between 2013 and 2021. We trained a survival model on the data using nested 5-fold cross-validation. Patients were stratified into a low- and high-risk group, based on the predicted survival probabilities. A total of 2642 patients with a pCR post-NAC were included, of whom 2620 were used in the final analysis. The survival model achieved good discrimination (C-index 0.754). Risk stratification identified a high-risk subgroup (n = 495, 18.9%) with significantly worse survival compared to the low-risk group (5-year OS 85.2% vs. 96.7%, log-rank P < 0.001). Pre-treatment nodal status (cN) was the strongest predictor, followed by age and clinical tumor stage (cT). This nationwide analysis demonstrates that TNBC patients achieving pCR remain a heterogeneous group with distinct survival outcomes. Machine learning-based survival modeling may enable individualized risk prediction using routinely collected clinical variables and may guide future post-surgical treatment escalation for high-risk TNBC patients despite pCR.