Machine Learning-Guided Survival Prediction and Treatment Sequencing in Advanced Soft Tissue Sarcoma Beyond...

Area of Science:

• Medical Oncology
• Clinical Research
• Cancer Therapeutics

Background:

• Optimal sequencing of later-line systemic therapy in advanced soft tissue sarcoma (STS) lacks robust evidence.
• Identifying prognostic factors for overall survival (OS) beyond second-line treatment is critical for improving patient outcomes.

Purpose of the Study:

• To identify prognostic factors for OS in advanced STS patients receiving third-line or later therapy.
• To explore the impact of systemic therapy sequencing on OS in routine clinical practice.

Main Methods:

• Retrospective analysis of 90 advanced STS patients undergoing third-line or later systemic therapy.
• Extreme Gradient Boosting (XGBoost) used to identify predictors of 1-year OS.
• Multivariable Cox models and sequencing analyses evaluated prognostic variables and treatment order.

Main Results:

• Progression on second-line therapy was associated with inferior OS (HR 2.31).
• Lipo-/leiomyosarcoma histology and time to progression ≥ 12 months on first-line therapy were linked to improved OS (HR 0.37 and HR 0.43, respectively).
• Sequencing analyses indicated sustained trabectedin activity, greater benefit from earlier pazopanib, and no line-dependent impact for gemcitabine + docetaxel.

Conclusions:

• Machine learning identified key predictors of survival in later-line STS.
• Prior treatment response and specific histology are strong determinants of outcomes.
• Exploratory analyses suggest that the timing of systemic therapy administration influences efficacy.