Pilot Biomarker Analysis and Decision Tree Algorithm Modeling of Patients with Chronic Subdural Hematomas

Abstract

The elderly population are at high risk for developing chronic subdural hematoma (cSDH). Surgical evacuation of cSDH is one of the most common procedures performed in neurosurgery. The present study aims to identify potential inflammatory biomarkers associated with its development and recurrence. Patients (>65 years of age) who presented with symptomatic cSDH (≥1 cm thickness or ≥5 mm midline shift [MLS]), requiring surgical intervention, were prospectively enrolled. The collected cSDH fluid was analyzed for inflammatory markers. Computed tomography (CT) scan data included pre-operative cSDH thickness and MLS. Outcome data included Glasgow Outcome Scale-Extended (GOS-E) score at 3, 6, and 12 months post-surgery, as well as cSDH recurrence. A decision tree model was used to determine the predictive power of extracted analytes for MLS, cSDH thickness, and recurrence. This pilot study includes 20 enrolled patients (mean age 77.9 ± 7.4 years and 85% falls). Rate of cSDH recurrence was 42%, with 21% requiring reoperation. Chemokine (C-X-C motif) ligand 9 (CXCL9) concentrations correlated with cSDH thickness (r = 0.975, p = 0.040). Interleukin (IL)-6 and vascular endothelial growth factor (VEGF)-A concentrations correlated with MLS (r = 0.974, p = 0.005; r = 0.472, p = 0.036, respectively). IL-5 concentrations correlated with more favorable GOS-E scores at 3, 6, and 12 months (r = 0.639, p = 0.006; r = 0.727, p = 0.003; r = 0.693, p = 0.026, respectively). Regulated on activation, normal T-cell expressed and secreted (RANTES) concentrations correlated with complete cSDH resolution (r = 0.514, p = 0.021). The decision tree model identified that higher concentrations of CXCL9 were predictive of MLS (risk ratio [RR] = 12.0), higher concentrations of IL-5 were predictive of cSDH thickness (RR = 4.5), and lower concentrations of RANTES were predictive of cSDH recurrence (RR = 2.2). CXCL9, IL-6, VEGF, IL-5, and RANTES are associated with recurrence after surgery and may be potential biomarkers for predicting cSDH recurrence and neurological outcomes.

Keywords: CT imaging; Glasgow Outcome Scale-Extended; biomarker; chronic subdural hematoma; inflammation; traumatic brain injury.

FIG. 1.

CXCL9 versus cSDH thickness. CXCL9, chemokine (C-X-C motif) ligand 9; cSDH, chronic subdural hematoma; pg, picogram; mL, milliliter; mm, millimeter .

FIG. 2.

VEGF-A versus MLS. VEGF-A, vascular endothelial growth factor; MLS, midline shift; pg, picogram; mL, milliliter; mm, millimeter.

FIG. 3.

Receiver operating characteristic (ROC) curve of CXCL9 concentrations to discriminate between cSDH patients with midline shift >10 mm. CXCL9, chemokine (C-X-C motif) ligand 9; cSDH, chronic subdural hematoma; mm, millimeter.

FIG. 4.

C5.1 decision tree demonstrating that CXCL9 concentrations >16,390.2 pg/mL were significantly associated with midline shift >10 mm in cSDH patients. CXCL9, chemokine (C-X-C motif) ligand 9; mm, millimeter; cSDH, chronic subdural hematoma; pg, picogram; mL, milliliter.

FIG. 5.

Receiver operating characteristic (ROC) curve of IL-5 concentrations to discriminate between cSDH patients with hematoma thickness >20 mm. IL-5, interleukin 5; cSDH, chronic subdural hematoma; mm, millimeter.

FIG. 6.

C5.1 decision tree demonstrating that IL-5 concentrations >22.215 pg/mL were significantly associated with hematoma thickness >20 mm in cSDH patients. IL-5, interleukin 5; mm, millimeter; cSDH, chronic subdural hematoma; pg, picogram; mL, milliliter.

FIG. 7.

C5.1 decision tree demonstrating that inflammatory biomarker concentrations (pg/mL) were significantly associated with cSDH recurrence. RANTES, regulated on activation, normal T-cell expressed and secreted; G-CSF, granulocyte colony-stimulating factor; PDGF-AA, platelet-derived growth factor-AA; pg, picogram; mL, milliliter.