Validation of the PERSARC scoring system for overall survival in Malaysian patients with soft tissue sarcoma: A retrospective cohort study.

Abstract

PurposeTo evaluate the reliability and calibration of the Personalised Sarcoma Care (PERSARC) scoring system in predicting overall survival (OS) for Malaysian patients with high-grade soft tissue sarcoma (STS) of the extremities, and to determine whether PERSARC can effectively stratify patients into risk groups to guide treatment decisions in a Malaysian clinical setting.MethodsWe conducted a retrospective observational study at a tertiary sarcoma treatment centre in Malaysia, including 115 patients with primary high-grade extremity STS treated between January 2002 and December 2019. PERSARC-predicted 5-year overall survival was compared with observed survival. In accordance with TRIPOD recommendations for external validation, model performance was assessed using discrimination (Harrell's C-index with bootstrapped confidence intervals and time-dependent area under the curve at 3, 5 and 10 years) and calibration (calibration plot of predicted versus Kaplan-Meier observed survival by quartile of predicted risk, calibration slope, observed-to-expected ratio, and Brier score). A secondary exploratory analysis examined prognostic factors for overall survival using Cox proportional hazards regression.ResultsOver a median follow-up of 7.0 years, 42 of 115 patients (36.5%) died. PERSARC showed good discrimination, with a Harrell's C-index of 0.80 (95% CI 0.73-0.86) and time-dependent AUCs of 0.80, 0.84 and 0.78 at 3, 5 and 10 years, respectively. Risk stratification was clear, with observed 5-year overall survival of 40.0% in the group predicted to have less than 50% 5-year survival, versus 88.8% in the group predicted to exceed it (log-rank p < 0.001). Calibration, however, was imperfect: PERSARC systematically underestimated observed survival across the entire range of predicted risk (observed-to-expected ratio 1.12; calibration slope 0.78, 95% CI 0.53-1.03), with the largest discrepancy in the poorest-prognosis quartile (predicted 22.4% versus observed 37.9%). In the secondary analysis, age, tumour size and distant metastasis were independently associated with mortality, with distant metastasis carrying the greatest hazard.ConclusionsIn this single-centre Malaysian cohort, PERSARC effectively discriminated between high- and low-risk patients with high-grade extremity STS, supporting its potential value as a structured aid to multidisciplinary discussion and shared decision-making. Calibration was less satisfactory, with consistent underestimation of survival. Therefore, absolute PERSARC-derived survival probabilities should be interpreted with caution in this population. Given the retrospective, single-centre design and modest sample size, these findings should be regarded as preliminary evidence supporting the applicability of PERSARC in Malaysian patients, pending larger multicentre validation and local recalibration.