205 Distinct Pain-Function Phenotypes in Elective Lumbar Fusion Patients Predict Long-Term Functional Outcomes: An Unsupervised Machine Learning Approach

View at publisher Summarise with VIRAT
Authors: Sameer C. Sundrani, Harsh Jain, Ranbir Ahluwalia, Tyler C. Zeoli, Iyan Younus, Soren Jonzzon, Hani Chanbour, Richard A. Berkman, Amir Mohamad Abtahi, Byron F. Stephens, Scott L. Zuckerman
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

INTRODUCTION: While lumbar fusion surgery is a common intervention for degenerative spine disease, predicting postoperative outcomes can be challenging. METHODS: A retrospective cohort study was performed for patients undergoing thoracic or lumbar fusion for degenerative disease from 2011-22. Inclusion criteria were: adults >18-years-old undergoing thoracic or lumbar fusion surgery for degenerative pathology with ≥ 2-year follow-up. Preoperative Visual Acuity Scale (VAS) leg/back pain scores, Oswestry Disability index (ODI), and motor deficits from patients were analyzed using Latent Profile Analysis (LPA). Long-term outcomes were compared across phenotypes using ANOVA with post-hoc Tukey tests. Multinomial regression identified significant phenotype predictors. RESULTS: A total of 194 patients undergoing lumbar fusion (age 61.6±10.7) years, 82 (41.8%) males, with mean follow-up of 6.7±2.9 years) were included. There were 127 (65.5%) single-level, 33 (17.0%) two-level, and 34 (17.5%) ≥ three-level fusions. LPA identified three symptomatologic phenotypes: 1) pain-dominant: predominantly high VAS pain scores (n=93), 2) baseline-functional: overall low ODI at baseline (n=77), and 3) motor dominant: the presence of a motor deficit at baseline (n=24). These profiles exhibited statistically significant differences in final postoperative ODI (p < 0.001), with the baseline-functional group demonstrating the least disability and lowest long-term ODI (19.6 ±17.0) compared to the pain-dominant (30.3±19.5) and motor-dominant (24.5±16.4) groups. CONCLUSIONS: Patients undergoing elective lumbar fusion for degenerative pathology can be classified into three clinically relevant preoperative phenotypes using unsupervised learning, which include: (1) predominantly high preoperative pain scores, (2) overall high function with low ODI, and (3) the presence of any motor deficit. the high function group had the best long-term outcomes, whereas the pain- and motor- dominant had the worst outcomes. This machine learning based classification system may serve as a valuable starting-point to prognosticate outcomes after elective lumbar fusion.

Fields of Study & Subjects

Spine and Intervertebral Disc Pathology Musculoskeletal pain and rehabilitation Cervical and Thoracic Myelopathy

Publication details

Journal / source
Neurosurgery
Volume
72
Issue
Supplement_1
Pages
44–44
Publication date
26 March 2026
Language
EN