C.02 Predictors of survival in a surgical series of Metastatic Spinal Cord Compression and a complete external validation of 8 models in a prospective multi-centre study

Abstract

Background: We aimed to identify preoperative predictors of survival in Metastatic Epidural Spinal Cord Compression (MESCC) patients surgically treated, examine how these predictors relate to eight prognostic models, and to perform the first full external validation of these models in accordance with the TRIPOD statement. Methods: 142 surgically treated MESCC patients were enrolled in a prospective, multicenter cohort study and followed for 12 months or until death. Cox proportional hazards (PH) regressions were used. Non-collinear predictors with <10% missing data, ≥10 events per stratum and p<0.05 in univariable analysis were tested through a backward stepwise selection. For the original and revised Tokuhashi, Tomita, modified Bauer, van der Linden, Bartels, OSRI, Bartels and Bollen, we examined calibration and discrimination; survival stratified by risk groups with the Kaplan-Meier method and log-rank test. Results: Primary tumor, organ metastasis and SF-36v2 PC were associated with survival in multivariable analysis; corrected discrimination was 0.68. These three predictors were common to most current prognostic models. However, calibration was poor overall while discrimation was possibly helpful. Conclusions: Primary tumor type (breast, prostate and thyroid), absence of organ metastasis, and a lower degree of physical disability are preoperative predictors of longer survival in surgical MESCC patients. Clinicians should use these 8 prognostic models with caution.