Are Current Survival Prediction Tools Useful When Treating Subsequent Skeletal-related Events From Bone Metastases?

Area of Science:

• Oncology
• Orthopedics
• Machine Learning in Medicine

Background:

• Bone metastases present significant challenges in advanced cancer, impacting patient quality of life and survival.
• Accurate prognostic models are crucial for guiding treatment decisions, especially for patients experiencing subsequent skeletal-related events (SREs).
• The reliability of existing prognostic models, like Skeletal Oncology Research Group machine-learning algorithms (SORG-MLAs), for subsequent SREs is uncertain.

Purpose of the Study:

• To evaluate the accuracy and reliability of SORG-MLAs for predicting survival in patients with subsequent SREs.
• To assess the performance of SORG-MLAs specifically for patients treated with surgery or radiotherapy for subsequent spinal or extremity SREs.

Main Methods:

• Retrospective analysis of 584 patients with initial and subsequent SREs treated between 2010 and 2019.
• Patients were categorized into spine and extremity SRE subgroups.
• SORG-MLAs were used to predict survival at the time of the subsequent SRE, with performance assessed using AUC, Brier score, and decision curve analysis.

Main Results:

• SORG-MLAs demonstrated acceptable discrimination (AUCs 0.65–0.73) and good overall performance (Brier scores lower than null model) for both spine and extremity groups.
• The algorithms provided a net benefit in decision curve analysis for both cohorts.
• A significant overestimation of 1-year survival probabilities was observed in the spine group (median log(O:E) of -0.60).

Conclusions:

• SORG-MLAs are viable prediction tools for patients with subsequent SREs, offering satisfactory discrimination and net benefits.
• Clinicians should exercise caution when interpreting 1-year survival predictions for spinal SREs due to overestimation.
• There is a need for improved prognostic algorithms and innovative tools to better manage patients with subsequent SREs as lifespans increase.