Scholarlink

Background <p>Spinal malignant lesions are a common feature in multiple myeloma (MM) and often require surgical intervention. Accurate survival prediction is critical for guiding treatment decisions in these patients. While PATHFx is a widely used, machine-learning-based prognostic tool for skeletal metastases, it has not been validated for malignant bone lesions in MM so far.</p> Objective <p>To evaluate the predictive performance and clinical utility of PATHFx 3.0 in a well-characterized cohort of MM patients with spinal malignant lesions.</p> Methods <p>A retrospective cohort of 100 MM patients with radiologically confirmed spinal malignant lesions treated between 2009 and 2024 at a tertiary care center was analyzed. 51 patients underwent surgery for the local treatment of spinal lesions, while 49 were treated with non-operative treatment regimes. Clinical data were entered into PATHFx 3.0 to generate survival estimates at 1, 3, 6, 12, 18, and 24 months. Model performance was assessed using Receiver Operating Curves (ROC) with Area under the Curve (AUC), Brier scores, calibration plots, and decision curve analysis (DCA), and compared to the actual survival in this cohort.</p> Results <p>PATHFx achieved good discriminatory performance at all time points, with AUC values ranging from 0.72 (1 month) to 0.79 (18 months). Calibration improved with longer prediction intervals, and Brier scores ranged from 0.09 to 0.20, with best accuracy at 3 months. DCA showed net clinical benefit for all models except the 1-month estimate. The inclusion of both surgical and non-surgical patients enhanced the generalizability of results.</p> Conclusion <p>PATHFx 3.0 is a reliable and clinically useful tool for survival estimation in MM patients with spinal disease. Its flexible design and acceptable predictive performance support its use in multidisciplinary treatment planning, especially where traditional scoring systems fall short.</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

Predicting survival of patients with spinal involvement in multiple myeloma using PATHFx 3.0 – a validation study of 100 patients in Germany

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月27日

查看全文
  • Julian Kylies,
  • Elias Brauneck,
  • Tobias M. Ballhause,
  • Katja Weisel,
  • Markus Schomacher,
  • Malte Schroeder,
  • Peter Obid,
  • Leon-Gordian Leonhardt,
  • Lennart Viezens

摘要

Background

Spinal malignant lesions are a common feature in multiple myeloma (MM) and often require surgical intervention. Accurate survival prediction is critical for guiding treatment decisions in these patients. While PATHFx is a widely used, machine-learning-based prognostic tool for skeletal metastases, it has not been validated for malignant bone lesions in MM so far.

Objective

To evaluate the predictive performance and clinical utility of PATHFx 3.0 in a well-characterized cohort of MM patients with spinal malignant lesions.

Methods

A retrospective cohort of 100 MM patients with radiologically confirmed spinal malignant lesions treated between 2009 and 2024 at a tertiary care center was analyzed. 51 patients underwent surgery for the local treatment of spinal lesions, while 49 were treated with non-operative treatment regimes. Clinical data were entered into PATHFx 3.0 to generate survival estimates at 1, 3, 6, 12, 18, and 24 months. Model performance was assessed using Receiver Operating Curves (ROC) with Area under the Curve (AUC), Brier scores, calibration plots, and decision curve analysis (DCA), and compared to the actual survival in this cohort.

Results

PATHFx achieved good discriminatory performance at all time points, with AUC values ranging from 0.72 (1 month) to 0.79 (18 months). Calibration improved with longer prediction intervals, and Brier scores ranged from 0.09 to 0.20, with best accuracy at 3 months. DCA showed net clinical benefit for all models except the 1-month estimate. The inclusion of both surgical and non-surgical patients enhanced the generalizability of results.

Conclusion

PATHFx 3.0 is a reliable and clinically useful tool for survival estimation in MM patients with spinal disease. Its flexible design and acceptable predictive performance support its use in multidisciplinary treatment planning, especially where traditional scoring systems fall short.