The international surgical journal with global reach

This is the Scientific Surgery Archive, which contains all randomized clinical trials in surgery that have been identified by searching the top 50 English language medical journal issues since January 1998. Compiled by Jonothan J. Earnshaw, former Editor-in-Chief, BJS

Systematic review of mortality risk prediction models in the era of endovascular abdominal aortic aneurysm surgery. BJS 2017; 104: 964-976.

Published: 13th June 2017

Authors: N. Lijftogt, T. W. F. Luijnenburg, A. C. Vahl, E. D. Wilschut, V. J. Leijdekkers, M. F. Fiocco et al.

Background

The introduction of endovascular aneurysm repair (EVAR) has reduced perioperative mortality after abdominal aortic aneurysm (AAA) surgery. The objective of this systematic review was to assess existing mortality risk prediction models, and identify which are most useful for patients undergoing AAA repair by either EVAR or open surgical repair.

Method

A systematic search of the literature was conducted for perioperative mortality risk prediction models for patients with AAA published since 2006. PRISMA guidelines were used; quality was appraised, and data were extracted and interpreted following the CHARMS guidelines.

Results

Some 3903 studies were identified, of which 27 were selected. A total of 13 risk prediction models have been developed and directly validated. Most models were based on a UK or US population. The best performing models regarding both applicability and discrimination were the perioperative British Aneurysm Repair score (C‐statistic 0·83) and the preoperative Vascular Biochemistry and Haematology Outcome Model (C‐statistic 0·85), but both lacked substantial external validation.

Conclusion

Mortality risk prediction in AAA surgery has been modelled extensively, but many of these models are weak methodologically and have highly variable performance across different populations. New models are unlikely to be helpful; instead case‐mix correction should be modelled and adapted to the population of interest using the relevant mortality predictors.

Full text