A Novel Method to Classify Cervical Incomplete Spinal Cord Injury Based on Potential for Recovery: A Group-Based Trajectory Analysis.

Abstract

The outcomes of cervical incomplete spinal cord injury (SCI) are heterogeneous. This study sought to dissociate subgroups of cervical incomplete SCI patients with distinct longitudinal temporal profiles of recovery in upper limb motor function. Patients with cervical incomplete SCI (American Spinal Injury Association Impairment Scale [AIS] B-D; C1-C8) were identified from four prospective, multi-center SCI datasets. A group-based trajectory model was fit to longitudinal upper extremity motor scores out to 1 year. Multi-variable multinomial logistic regression was performed to identify features that characterize each trajectory group. A classification system for predicting trajectory group at baseline was developed by recursive partitioning. In total, 801 patients were eligible. Four distinct trajectory groups were identified: 1) "Poor outcome": Severe injury, very minimal recovery; 2) "Moderate recovery": Moderate-to-severe injury, moderate recovery; most recovery occurs by 6 months, with mild, gradual recovery continuing thereafter; 3) "Good recovery": Moderate injury, good recovery; most recovery occurs by 3 months, with mild, gradual recovery continuing thereafter; and 4) "Excellent outcome": Mild injury, recovery to normal/near-normal by 3 months. On adjusted analyses, older age was associated with lower likelihood of "excellent outcome" (p = 0.020). AIS C and D injuries were associated with "moderate recovery," "good recovery," and "excellent outcome" (p < 0.001). Mid-cervical injuries occurred more frequently in "moderate recovery," "good recovery," and "excellent outcome" (p < 0.001) groups. Early surgical decompression (< 24 h) was associated with increased propensity for "good recovery" (p = 0.039) and "excellent outcome" (p = 0.048). A classification model based on recursive partitioning could predict trajectory group using age, AIS grade, and neurological level with an area under the curve of 0.81. Patients with cervical incomplete SCI demonstrate distinct temporal profiles of recovery in upper limb motor function. The trajectory a patient is likely to follow may be predicted at baseline with fair accuracy.