External validation and recalibration of the Oswestry spinal risk index for survival prediction in patients with advanced cancers and spinal metastases

Objectives: Prognostic models for spinal metastases are widely used to guide management but were largely developed in high-resource settings and may not generalize to other populations. Prior validations show inconsistent performance, supporting the need for local validation and recalibration. This study aimed to externally validate six prognostic models for spinal metastases in an Indonesian cohort and to recalibrate the best-performing model. Methods: We externally validated six prognostic models (Bollen, Modified Bauer, New England Spinal Metastasis Score, Oswestry Spinal Risk Index [OSRI], Revised Tokuhashi, and Van der Linden) for predicting 3-, 6-, and 12-month survival. Performance was assessed using discrimination area under the receiver operating characteristic curve (AUC), calibration, and the Brier score, with recalibration applied to models with acceptable discrimination but poor calibration. Results: Two hundred patients were included, with a median overall survival of 5 months. The OSRI demonstrated acceptable discrimination across all time points (AUC 0.714–0.830), while the Bollen model showed acceptable discrimination only at 6 and 12 months (AUC 0.735 and 0.871). The remaining models demonstrated poor discrimination (AUC < 0.70). All models showed inadequate calibration. Recalibration of the OSRI improved calibration and Brier scores while preserving its discriminatory performance. Conclusions: The OSRI demonstrated the most consistent prognostic performance among the evaluated models, and recalibration improved its predictive performance in this cohort. However, further validation in larger, multicenter cohorts is required before the recalibrated OSRI can be considered for broader clinical implementation.