Abstract

Background
Endoleaks are a common complication following endovascular aneurysm repair (EVAR), occurring in up to 30% of patients and often necessitating reintervention. Accurate early prediction of endoleak could improve surveillance strategies, reduce unnecessary imaging, and enable timely intervention. Machine learning (ML) offers the potential to integrate complex clinical and imaging features to forecast endoleak risk. However, the performance and methodological quality of existing ML models remain unclear.

Method
We conducted a systematic review and meta-analysis. Databases were searched from inception to April 2025 for studies developing or validating ML models predicting endoleaks after EVAR. Eligible studies included adult patients with abdominal aortic aneurysm treated by EVAR. R(5.4.1) was used.

Results
Thirteen studies, encompassing 2,344 patients, which included 664 endoleak patients were included. A range of ML algorithms were reported, including logistic regression, random forests, support vector machines, and deep learning approaches. The pooled area under the curve (AUC) was 0.89 (95% CI 0.83–0.94), with sensitivity 0.86 (95% CI 0.75–0.92), specificity 0.87 (95% CI 0.83–0.91), and accuracy 0.85 (95% CI 0.80–0.90). Considerable heterogeneity was observed across studies (I² > 85%). All studies were rated as high risk of bias overall, mainly due to small single-centre designs, lack of external validation, and inconsistent handling of missing data.

Conclusion
ML-based prediction models demonstrate promising diagnostic performance in forecasting endoleak after EVAR. However, high risk of bias and substantial heterogeneity limit clinical applicability. Future research should prioritise multicentre prospective validation, methodological standardisation, and integration of prediction tools into routine post-EVAR surveillance.