Evaluation of five different aneurysm scoring systems to predict mortality in ruptured abdominal aortic aneurysm patients

Abstract

Background: Ruptured abdominal aortic aneurysms (RAAAs) are associated with a high overall mortality (up to 25% to 35%) ≤30 days when offered surgical treatment. Risk prediction models can provide valuable information on surgical risks, guide clinical decision making, and help identify patients who should not be operated on to prevent futile surgery. Finally, they can be used to evaluate clinical outcome. Different aneurysm scores are available. New ones (with only four parameters) are being developed, such as the Dutch Aneurysm Score (DAS). This study analyzed and compared these scoring models.

Methods: The study selected consecutive patients who presented with RAAA in two large vascular centers (Medisch Centrum Alkmaar and St. Antonius Nieuwegein) between 2005 and 2015. Variables necessary to retrospectively evaluate the scoring systems were registered in the patients' medical files. The discriminatory power and calibration were assessed using the receiver operating characteristic curve and the Hosmer-Lemeshow χ² test.

Results: The study included 347 consecutive patients with RAAA. There were 298 men (85.9%), and the mean ± standard deviation age was 72.6 ± 8.1 years. The receiver operating characteristic curves were developed for the DAS, Glasgow Aneurysm Score (GAS), Edinburgh Ruptured Aneurysm Score, Vancouver Scoring System (VSS), and Hardman Index. The area under the curve was better for the VSS (0.716; 95% confidence interval [CI], 0.647-0.786) than for the other scoring systems. Areas under the curve for the DAS (0.664; 95% CI, 0.592-0.736), Hardman Index (0.664; 95% CI, 0.592-0.736), Edinburgh Ruptured Aneurysm Score (0.621; 95% CI, 0.543-0.700), and GAS (0.591; 95% CI, 0.517-0.665) were slightly smaller, although only the difference between the VSS and GAS was statistically significant. Calibration showed a good fit for all models.

Conclusions: The performance of the tested models for the prediction of mortality in RAAA patients was comparable, with only a statistically significant difference between the VSS and the GAS in favor of the VSS. However, an almost perfect prediction is needed to withhold intervention, and no existing scoring system is capable of that.