Development and Validation of Interpretable Machine Learning Models Incorporating Paraspinal Muscle Quality for to Predict Cage Subsidence Risk Followingposterior Lumbar Interbody Fusion

The study focuses on identifying risk factors for cage subsidence after posterior lumbar interbody fusion (PLIF) and developing an interpretable machine learning model to predict these risks. It analyzes patients from two large teaching hospitals, using clinical, radiographic, and surgical parameters, including paraspinal muscle indices and bone density markers. A web-based application was developed to facilitate real-time clinical risk assessments using the machine learning model, enhancing surgical planning and reducing subsidence risks.

Study Overview

• Status: Completed
• Conditions: Machine Learning; Degenerative Lumbar Diseases; Cage
• Intervention / Treatment: Procedure: MR4

• Study Type: Observational
• Enrollment (Actual): 720

Contacts and Locations

Study Locations

• China
  • SuZhou
    • Jiangsu, SuZhou, China, 215006
      • The First Affiliated Hospital of Soochow University Medical Record and Imaging System

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Child; Adult; Older Adult
• Accepts Healthy Volunteers: Yes

• Sampling Method: Non-Probability Sample

Study Population

This retrospective study included patients with degenerative lumbar diseases who underwent single-level PLIF at the First and Second Affiliated Hospitals of Soochow University (January 2018-October 2023).A total of 620 patients from the First Affiliated Hospital comprised the training set, while 100 patients from the Second Affiliated Hospital formed the external validation set. The study adhered to the Declaration of Helsinki and was approved by the institutional ethics committee.

Description

Inclusion Criteria:

1. confirmed lumbar disc herniation, spinal stenosis, or spondylolisthesis based on clinical and imaging findings;
2. patients who failed conservative treatment for ≥3 months or experienced recurrence and underwent surgery for the first time;
3. minimum 12-month follow-up.

Exclusion Criteria:

1. prior spinal surgery;
2. spinal deformity or severe instability;
3. lumbar tuberculosis, infection, tumor, or severe bone destruction;
4. incomplete or lost follow-up.

Study Plan

How is the study designed?

Number of groups / cohorts

2

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
Non-Subsidence Group; Group A (n = 390): Lost IH value <2 mm.Cage subsidence was defined as a lost IH value ≥2 mm during the follow-up period	Procedure: MR4; The study is a clinical retrospective study and does not involve any interventional measures.
Subsidence Group; Lost IH value ≥2 mm.Cage subsidence was defined as a lost IH value ≥2 mm during the follow-up period	Procedure: MR4; The study is a clinical retrospective study and does not involve any interventional measures.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Bone density imaging indicator: Vertebral Bone Quality (VBQ)	VBQ is calculated by dividing the mean T1 signal intensity (L1-L4) by the cerebrospinal fluid (CSF) signal at L3	Preoperative measurement for PLIF (Posterior Lumbar Interbody Fusion)

Collaborators and Investigators

• Sponsor: Hao Liu

Study record dates

Study Major Dates

• Study Start (Actual): March 1, 2025
• Primary Completion (Actual): March 10, 2025
• Study Completion (Actual): March 15, 2025

Study Registration Dates

• First Submitted: March 17, 2025
• First Submitted That Met QC Criteria: March 17, 2025
• First Posted (Actual): March 25, 2025

Study Record Updates

• Last Update Posted (Actual): March 25, 2025
• Last Update Submitted That Met QC Criteria: March 17, 2025
• Last Verified: March 1, 2025

More Information

Terms related to this study

• Keywords: Machine learning; Cage subsidence; Degenerative lumbar disease
• Other Study ID Numbers: （2025）Lun Yan Grant No. 220 (Other Identifier: Ethics Committee of the First Affiliated Hospital of Soochow University)

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: YES
• IPD Plan Description: Before January 1,2026;in the form of paper. The original data of the trial will be uploaded to Genome Sequence Archive (GSA) (ngdc.cncb.ac.cn)

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No
• product manufactured in and exported from the U.S.: No