A Scoping Review of Preoperative Machine Learning Models for Outcome Prediction After Endovascular Repair of Infrarenal Abdominal Aortic Aneurysms

Objective In this scoping review, we provide an overview of machine learning–based prediction models using preoperative data to forecast postoperative outcomes in patients undergoing elective endovascular aortic repair (EVAR) for infrarenal abdominal aortic aneurysms. Methods A systematic search was conducted in EMBASE, PubMed, and Web of Science for relevant studies published between 2010 and 13th of January, 2026 according to the PRISMA-ScR guidelines. Studies were selected based on their application of endovascular repair and machine learning (ML) methodologies. After independent screening by two reviewers, data extraction was performed in accordance with the CHARMS (Critical Appraisal and Data Extraction for Systematic Reviews of Prediction Modeling Studies) checklist. The quality of the studies was assessed using the PROBAST+AI (Prediction model Risk Of Bias Assessment Tool for Artificial Intelligence) framework. A narrative synthesis was performed due to heterogeneity in outcomes, input features, and modeling approaches. Results Of the 3 291 screened articles, a final number of 17 studies were included. The most common outcomes assessed were mortality and complication/adverse events, followed by endoleak, reintervention and sac remodeling prediction. ML models showed good predictive performance with little differences between ML techniques and, in comparative studies, frequently reported higher performances than traditional statistical approaches. However, performance differences were inconsistent and depended strongly on the type and richness of input features. Most commonly used ML methods were tree based models, followed by kernel based and boosting methods. Models most frequently relied on age, comorbidities, and smoking status for mortality and complication prediction. For endoleak and reintervention prediction anatomical importance was shown. Most studies emphasized model development over external validation or clinical implementation. Conclusion Machine learning holds significant promise for outcome prediction after EVAR for infra-renal aneurysms, but current models are limited by small sample sizes, lack of standardization, and insufficient validation. The capabilities of machine learning techniques can be further explored by incorporating multi-modal features. Moreover, prioritize the creation of standardized, multi-center datasets to enhance generalizability and support the development of robust, clinically applicable ML models.