Who really benefits from chemotherapy in high-risk localized soft tissue sarcoma? A virtual twin analysis of ISG-STS 1001.

Abstract

11573 Background: Randomized trials of perioperative chemotherapy in high-risk localized extremity/truncal soft tissue sarcoma (etSTS) show modest average survival benefits, leaving uncertainty about which patients derive clinically meaningful benefit. Conventional hazard ratio-based reporting reflects population-level effects but offers limited guidance for individual decisions. Counterfactual approaches, such as the virtual twin framework, enable estimation of patient-specific outcomes under alternative treatment strategies, translating randomized evidence into clinically interpretable absolute benefit measures. Methods: Individual patient data from the ISG-STS 1001 randomized trial in high-risk localized etSTS with updated follow-up were analyzed using a virtual-twin framework with counterfactual predictions to estimate each patient’s outcomes under alternative treatment scenarios: standard epirubicin-ifosfamide (EI) vs histotype-tailored (HT) chemotherapy (proxy for no effective chemotherapy). Patients with myxoid liposarcoma were excluded. Cox proportional hazards models for overall survival (OS) and disease-free survival (DFS) were adjusted for pretreatment covariates. For each patient, counterfactual 5-year OS and DFS probabilities under EI and HT were estimated and individualized absolute treatment effects were defined as the difference between counterfactual predictions. Bootstrap resampling was used to quantify uncertainty and treatment benefit explored across levels of baseline risk and tumor characteristics. Results: The cohort included 218 pts (median follow-up 116 months). In covariate-adjusted models, mean absolute improvement in 5-year OS with EI was 7.5% (95%CI: 4.5%-10.2%). Absolute OS benefit was inversely correlated with Sarculator-predicted 10-year survival, indicating greater benefit among patients with poorer baseline prognoses. Pts in the highest Sarculator-predicted survival quartile (pOS >69%) had a mean 5-year OS benefit of 5.9% (95%CI: 3.5%-9.0%), with 63% achieving ≥5% benefit. In contrast, pts in the lowest predicted survival quartile (pOS <46%) had a mean benefit of 8.7% (95%CI: 5.3%-10.4%), with 98% achieving ≥5% benefit. Benefits varied by histology (undifferentiated pleomorphic sarcoma = 6.6%, leiomyosarcoma = 7.2%, synovial sarcoma = 8.3%, malignant peripheral nerve sheath tumor = 8.9%) and, within each subtype, was inversely correlated with Sarculator-predicted survival. Conclusions: Individualized counterfactual prediction reveals substantial heterogeneity in the absolute benefit of chemotherapy in high-risk localized etSTS, driven by baseline prognostic risk and histology. This framework provides clinically relevant, patient-centered estimates that support risk-adaptive decision-making and clarify chemotherapy benefits across risk groups beyond average trial effects.