Scholarlink

Purpose <p>Accurate prognostic prediction remains a significant challenge in the management of low-grade serous ovarian cancer (LGSOC), a rare and molecularly distinct histologic subtype. This study aimed to develop a prediction model visualized by nomogram to predict recurrence and survival outcomes in LGSOC patients.</p> Methods <p>A retrospective analysis was conducted on patients with LGSOC, using Cox regression to identify factors associated with recurrence and survival for further development of prediction model. The model’s accuracy and discriminative ability were assessed with area under the receiver operating characteristic curve (AUC) and calibration curves. The predictive performance of the model and International Federation of Gynecology and Obstetrics (FIGO) staging was compared using the concordance index (C-index), net reclassification improvement (NRI), and integrated discrimination improvement (IDI). Additionally, a deep-learning-based prediction model was developed through regression analysis, with performance evaluated via Kaplan-Meier analysis and C-index.</p> Results <p>A cohort of 155 patients with LGSOC was analyzed and four independent prognostic factors were identified and incorporated into a Cox regression-based model. The model demonstrated good calibration, as shown by calibration curves. Through internal validation, the model showed superior discriminatory ability over the FIGO staging system, with higher C-indexes for both disease-free survival (0.781 vs. 0.689) and overall survival (0.802 vs. 0.679), which was further confirmed by significant improvements in IDI and NRI. Additionally, the deep learning-based model based on this model was developed to evaluate potential non-linear relationships. This model achieved even higher predictive performance, with C-indexes of 0.907 for disease-free survival and 0.922 for overall survival.</p> Conclusion <p>We developed a risk prediction model, visualized by a clinically practical nomogram to predict recurrence and survival outcomes in LGSOC patients. Additionally, the deep learning-based prediction model based on neural networks was developed, providing improved prognostic evaluation for these patients.</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

Risk prediction model of survival in patients with low-grade serous ovarian cancer: a multicenter Cohort study

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年4月18日

查看全文
  • Ying Xue,
  • Zhongshao Chen,
  • Xuecheng Fang,
  • Ran Chu,
  • Mingbao Li,
  • Yuanming Shen,
  • Qin Yao,
  • Baochen Fu,
  • Tianyu Qin,
  • Li Li,
  • Xu Qiao

摘要

Purpose

Accurate prognostic prediction remains a significant challenge in the management of low-grade serous ovarian cancer (LGSOC), a rare and molecularly distinct histologic subtype. This study aimed to develop a prediction model visualized by nomogram to predict recurrence and survival outcomes in LGSOC patients.

Methods

A retrospective analysis was conducted on patients with LGSOC, using Cox regression to identify factors associated with recurrence and survival for further development of prediction model. The model’s accuracy and discriminative ability were assessed with area under the receiver operating characteristic curve (AUC) and calibration curves. The predictive performance of the model and International Federation of Gynecology and Obstetrics (FIGO) staging was compared using the concordance index (C-index), net reclassification improvement (NRI), and integrated discrimination improvement (IDI). Additionally, a deep-learning-based prediction model was developed through regression analysis, with performance evaluated via Kaplan-Meier analysis and C-index.

Results

A cohort of 155 patients with LGSOC was analyzed and four independent prognostic factors were identified and incorporated into a Cox regression-based model. The model demonstrated good calibration, as shown by calibration curves. Through internal validation, the model showed superior discriminatory ability over the FIGO staging system, with higher C-indexes for both disease-free survival (0.781 vs. 0.689) and overall survival (0.802 vs. 0.679), which was further confirmed by significant improvements in IDI and NRI. Additionally, the deep learning-based model based on this model was developed to evaluate potential non-linear relationships. This model achieved even higher predictive performance, with C-indexes of 0.907 for disease-free survival and 0.922 for overall survival.

Conclusion

We developed a risk prediction model, visualized by a clinically practical nomogram to predict recurrence and survival outcomes in LGSOC patients. Additionally, the deep learning-based prediction model based on neural networks was developed, providing improved prognostic evaluation for these patients.