Distinct postoperative quality of life trajectories after surgery for degenerative cervical myelopathy: a multicenter...

Postoperative outcomes in degenerative cervical myelopathy (DCM) vary considerably, yet few studies have characterized the heterogeneous recovery trajectories using longitudinal data.

To identify distinct postoperative quality of life (QOL) trajectories in DCM patients and determine baseline predictors of recovery patterns.

Prospective multicenter observational study.

A total of 977 patients undergoing surgery for DCM across 10 high-volume spine centers in Japan.

The QOL outcome measure comprised the Short Form-36 physical component summary (PCS) score. Functional outcomes were specifically captured through Japanese Orthopaedic Association Cervical Myelopathy Evaluation Questionnaire. Outcomes were measured at baseline, 6, 12, and 24 months postoperatively.

Latent growth mixture modeling was employed to classify patients into distinct postoperative recovery trajectories based on PCS trends. To identify independent predictors of trajectory membership, multinomial logistic regression was performed, with variable selection refined through least absolute shrinkage and selection operator regression (LASSO) regression. Model performance was assessed using area under the receiver operating characteristic curve (AUC) for discrimination and decile-based calibration plots with bootstrap validation.

A total of 4 distinct PCS recovery trajectories were identified: Low-to-High (L-H, 7.3%), High-to-High (H-H, 44.9%), Low-to-Low (L-L, 37.7%), and Initial-Decline (I-D, 10.1%). Preoperative lower extremity function emerged as the strongest predictor of trajectory class, reflecting the baseline QOL. Additional significant predictors included age, smoking history, symptom duration, and cervical spine function. Particularly, reduced cervical function at baseline was found to be a significant predictor of unfavorable QOL at 24 months. The prediction model demonstrated good discriminatory performance following least absolute shrinkage and selection operator (LASSO) regression for common classes (AUCs: H-H=0.86, L-L=0.80) and moderate performance for L-H class (AUC 0.74). However, accuracy was limited for the I-D class (AUC = 0.63), and calibration was compressed in rarer classes due to class imbalance.

Distinct patterns of postoperative recovery exist among DCM patients, with baseline physical function and patient characteristics significantly influencing QOL trajectory. While predictive models reliably distinguished major recovery patterns, less frequent trajectories, particularly those involving deterioration, were difficult to forecast. These findings support the utility of trajectory modeling and patient-reported outcome measures to enhance individualized surgical prognostication in DCM.