190 Machine Learning-Based Cluster Analysis Identifies Four Unique Phenotypes of Degenerative Cervical Myelopathy Patients With Distinct Clinical Profiles and Long-Term Functional Outcomes

INTRODUCTION: Degenerative cervical myelopathy (DCM), the predominant cause of spinal cord dysfunction among adults, exhibits a diverse range of symptom. Traditional classification using mJOA have not been able to address its complex heterogeneity. This study employs machine learning-based clustering algorithms to identify distinct patient clinical profiles and functional trajectories following surgery METHODS: We implemented both latent profile analysis and k-means clustering on aggregated data from three large DCM trials. Key covariates, including the Nurick score, NDI (neck disability index), neck pain, and motor and sensory scores, were employed for clustering. Outcome differences among identified phenotypes were assessed using ANOVA, followed by posthoc Tukey test. RESULTS: A total of 1,047 DCM patients (mean [SD] age: 56.80 [11.39] years) had complete one year outcome assessment. Both LPA and k-means clustering identified four DCM patient phenotypes: “severe multimodal impairment” (n=286), “minimal impairment” (n=116), motor-dominant” (n=88) and “pain-dominant” (n=557) groups. The “severe multimodal impairment group”, comprising of frail elderly patients, demonstrated the worst overall 1-year outcomes (SF-36 PCS mean [SD]: 40.01 [9.75]; SF-36 MCS mean [SD]: 46.08 [11.50]), but exhibited the most substantial neurological recovery after surgery (DmJOA mean [SD]: 3.83 [2.98]). A higher frailty score and a positive smoking status predicted membership in phenotype 1 (“severe multimodal impairment” group). CONCLUSIONS: Unsupervised learning algorithms applied to baseline DCM symptoms enabled the prediction of distinct patient phenotypes. The concept of symptom clustering, provides a valuable framework for uncovering novel DCM subpopulations, enhancing patient identification beyond the use of a single patient reported outcome measure such as the mJOA.