Establishment of a Classification System and Postoperative Risk Warning Model for Patients Undergoing Bariatric Metabolic Surgery for Severe Obesity

Condition(s) studied

Study summary

This study aims to establish a classification system for patients undergoing metabolic surgery for severe obesity by constructing a prospective cohort of 2,000 patients and collecting clinical and biological data at multiple time points before and after surgery. By analyzing clinical, laboratory, and multi-omics characteristics, the study will identify indicators associated with postoperative adverse events and develop a risk warning model using machine learning algorithms. Ultimately, an intelligent digital system will be developed based on the classification criteria and risk model, integrating surgical classification and risk alert functions to provide real-time feedback, supporting clinicians and patients in optimizing postoperative treatment and risk management.

Eligibility

Primary outcome measure(s)

• Number of Participants Stratified into Distinct Clusters Using Unsupervised Clustering Algorithm Based on BMI, Comorbidity Count, Inflammatory Markers, and Proteomics Profiles — From enrollment to the end of follow-up at 2 years
  Participants will be stratified at baseline using an unsupervised clustering algorithm (e.g., k-means) based on the following input variables:（1）Body mass index (BMI, kg/m²)；（2）Comorbidity count (number of chronic diseases at enrollment)；（3）Inflammatory markers (e.g., CRP in mg/L, IL-6 in pg/mL)；（4）Proteomics features (relative expression intensity from LC-MS) The clustering procedure will produce a categorical variable assigning each participant to one of 3-5 data-driven subtypes. The total number of participants in each subtype group will be reported.
• Accuracy, Sensitivity, and Specificity of the Postoperative Risk Prediction Model for Adverse Outcomes — Up to 24 months after surgery
  Model performance will be assessed for predicting postoperative adverse outcomes within 12 months after bariatric surgery. Predictors include demographic data, intraoperative parameters, early postoperative recovery data, and stratification subtype. Model discrimination will be evaluated using area under the receiver operating characteristic curve (AUC-ROC), and calibration will be assessed with calibration plots and Hosmer-Lemeshow test. Sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) will be reported using confusion matrix analysis in internal validation (e.g., bootstrapping or 10-fold cross-validation).

Trial sites (1)

More The Third Xiangya Hospital of Central South University trials in China

Other trials for the same condition