Establishing a Classification System for Predicting Flow-Limiting Dissection After Balloon Angioplasty Using...

Area of Science:

• Vascular Surgery
• Interventional Cardiology
• Machine Learning in Medicine

Background:

• Percutaneous transluminal angioplasty (PTA) is a primary treatment for peripheral arterial disease (PAD).
• Flow-limiting dissection (FLD) is a significant complication following PTA, potentially impacting treatment outcomes.
• Predicting FLD is crucial for optimizing patient management and procedural success.

Purpose of the Study:

• To develop and validate a machine learning-based classification system to predict FLD after PTA.
• To identify key predictors of FLD in patients undergoing femoropopliteal (FP) artery treatment.
• To create a user-friendly tool for surgical decision-making in PAD treatment.

Main Methods:

• A multi-center, retrospective cohort study of 407 patients undergoing FP artery PTA.
• Evaluation of preoperative CT angiography for chronic total occlusion (CTO), FP artery grading, and peripheral artery calcium scoring (PACSS).
• Training and validation of machine learning models, with SHapley Additive exPlanations used to identify predictors and develop the CTO-FP-PACSS classification system.

Main Results:

• The random forest model achieved an AUC of 0.82 for FLD prediction.
• CTO, PACSS, and FP grading were identified as the most influential predictors of FLD.
• The CTO-FP-PACSS classification system independently predicted FLD (multivariate HR 4.13, p<0.001).

Conclusions:

• The CTO-FP-PACSS classification system accurately predicts FLD after PTA in FP artery disease.
• This system offers a practical approach to guide treatment decisions, potentially reducing FLD incidence.
• The classification system can aid surgeons in choosing between PTA and adjunctive devices.