Routine Laboratory Parameters in the Differentiation of Spinal Cord Infarction and Seronegative Acute Myelitis: A Retrospective Study

Background/Objectives: Overlapping clinical and imaging features complicate the early differentiation between spontaneous spinal cord infarction (SCI) and seronegative acute myelitis (SAM). We aimed to compare the clinical and laboratory characteristics of SCI and SAM and, as an exploratory analysis, to develop and internally validate multivariable diagnostic models evaluating whether routinely available laboratory parameters provide additional discriminatory information beyond selected clinical features. Methods: We retrospectively analyzed 75 patients (34 SCI, 41 SAM) treated between January 2017 and June 2025. Between-group laboratory comparisons were adjusted for multiple testing using the Benjamini–Hochberg false-discovery-rate procedure. Penalized logistic regression analyses were performed using clinical variables, laboratory variables, and their combination, with candidate variables defined on clinical and data-quality grounds rather than by univariate statistical significance. Model discrimination was evaluated using repeated nested cross-validation. Bootstrap optimism correction was additionally performed for the full-data combined model, and decision curve analysis was conducted as an exploratory secondary analysis. Results: SCI patients were older, more often male, and had higher prevalence of hypertension and diabetes. Radicular pain was markedly more common in SCI. After Benjamini–Hochberg correction, five laboratory parameters remained significantly different between groups: CRP, triglycerides, monocyte percentage, and absolute monocyte count were higher in SCI, whereas HDL cholesterol was lower. In the combined penalized analysis, age, hypertension, radicular pain, CRP, triglycerides, absolute monocyte count, and absolute eosinophil count were retained. The clinical, laboratory, and combined models yielded AUCs of 0.821 (95% CI 0.718–0.910), 0.756 (95% CI 0.633–0.865), and 0.871 (95% CI 0.776–0.947), respectively. Compared with the clinical model, the combined model showed a modest increase in discrimination (ΔAUC 0.050, 95% CI 0.005–0.101, p = 0.030). Bootstrap internal validation of the full-data combined model yielded an optimism-corrected AUC of 0.902, compared with an apparent AUC of 0.937, with a mean optimism of 0.035. Exploratory decision curve analysis showed a potential net benefit of the combined model across a range of threshold probabilities. Conclusions: Spontaneous SCI and SAM showed distinct clinical and laboratory profiles in this retrospective cohort. Routine laboratory parameters provided modest additional discriminatory information beyond clinical features, and the combined model maintained discrimination during internal validation. These findings are exploratory and require validation in larger, independent cohorts before clinical application.