External validation of a three-variable Ewing sarcoma prognostic score and incremental value of the Glasgow prognostic score in adults: a multicenter cohort study.

OBJECTIVE

To externally validate a three-variable Ewing sarcoma score and assess the incremental prognostic value of the Glasgow prognostic score (GPS) in adults.

METHODS

This retrospective cohort included 133 adults treated at 10 Turkish centers (2010-2026). The Ewing score used metastases, tumor diameter >5 cm, and leukocytes >11 000/mm3; GPS used C-reactive protein >10 mg/L and albumin <3.5 g/dl. Outcomes were overall (OS) and event-free survival (EFS). Discrimination was assessed using the concordance index and time-dependent area under the curve. Center-clustered Cox models included the Ewing score and ordinal GPS. Bootstrap uncertainty was estimated within centers.

RESULTS

During a median 81.1-month follow-up, 67 deaths and 82 EFS events occurred. Ewing score concordance indices were 0.668 (95% bootstrap confidence interval [CI]: 0.613-0.721) for OS and 0.688 (95% CI: 0.641-0.734) for EFS. High-risk patients had inferior outcomes, but low- and intermediate-risk groups were not separated. Each one-level increase in GPS was associated with shorter OS (hazard ratio [HR]: 2.67, 95% CI: 1.95-3.65) and EFS (HR: 2.31, 95% CI: 1.44-3.69). GPS increased the concordance index by 0.091 (95% bootstrap CI: 0.041-0.143) for OS and 0.070 (95% bootstrap CI: 0.027-0.115) for EFS. Its OS association was strongest within 24 months and attenuated thereafter.

CONCLUSIONS

The Ewing score identified a high-risk population but incompletely separated lower-risk patients. GPS added independent prognostic information, but the extension requires external validation before clinical use.