Quantitative Spinal Cord Deformation Is Associated with Neurologic Status in High-Grade Thoracic Metastatic Epidural Spinal Cord Compression: A Retrospective Surgical Cohort

Abstract

Background/Objectives: Neurologic presentation varies among patients with high-grade metastatic epidural spinal cord compression (MESCC), even within Bilsky grades 2–3. We examined whether quantitative preservation of spinal cord caliber is associated with neurologic intactness at presentation. Methods: We retrospectively studied 73 adults who underwent surgery for thoracic Bilsky grade 2–3 MESCC between June 2020 and December 2025. Preoperative neurologic status was classified as Frankel E versus Frankel A-D. Blinded observers measured anteroposterior and transverse cord diameters at maximal compression and at the first tumor-free cranial and caudal pedicle levels. A normalized two-adjacent cord compression ratio (CCR) was available in 71 patients. Modified Poisson regression estimated prevalence ratios for Frankel-E status, adjusted for Bilsky grade; logistic, Firth, and cranial-reference analyses were used as sensitivity analyses. Results: Twenty-two patients (30.1%) were Frankel E. Two-adjacent CCR was higher in Frankel-E than Frankel A-D patients (median 0.521 vs. 0.439; p = 0.013). Each 0.10 increase in CCR was associated with greater prevalence of Frankel-E status after Bilsky adjustment (PR 1.23, 95% CI 1.04–1.47; p = 0.018). Logistic, Firth, and cranial-reference analyses were generally consistent in direction; the solid-tumor-only estimate also remained positive but was less precise and included the null. Interobserver agreement for CCR was acceptable but imprecise (ICC[A,1] = 0.758, 95% CI 0.433–0.911). Conclusions: In this surgically selected cohort, normalized cord-caliber preservation was associated with neurologic intactness after accounting for Bilsky grade. CCR should be regarded as a research phenotype rather than an individual-patient predictor or treatment threshold, pending prospective validation in broader surgical and nonsurgical populations.