Validating the use of contrast-induced nephropathy prediction models in endovascular aneurysm repairs

Abstract

Background: Existing risk prediction models for contrast-induced nephropathy (CIN) are based on studies for percutaneous coronary interventions, with none validated for use in vascular procedures. We aim to validate existing CIN prediction models in patients who underwent aortic endovascular aneurysm repair (EVAR).

Methods: A retrospective review of 216 patients who underwent EVAR between January 2008 and December 2015 was undertaken. Incidence of acute kidney injuries at 24, 48, and 72 hours and at follow-up were evaluated. Of 12 CIN prediction models within the literature, 8 were suitable for validation in patients who underwent EVAR and validation was performed with C-statistics.

Results: There were 216 EVARs performed within the study period. The mean patients age was 73 years and 162 (75%) were performed in an elective setting. Percentage of preoperative chronic kidney disease stages 1 to 5 were 16%, 42%, 31%, 6%, and 5%, respectively. The mean intraprocedure contrast volume used was 280 mL. Incidence of acute kidney injuries at 24, 48, and 72 hours and at follow-up were 8%, 12%, 11%, and 6%, respectively. Three percent of patients became dialysis dependent. Validation of the eight existing CIN predication models reveal area under curve C-statistics between 0.61 and 0.75 (P = .026 to P < .001). Five of the 8 had good discriminative ability (C-statistics of >0.70) and the CIN prediction models by Mehran and Tziakas had the highest C-statistics at 0.75 (P < .001).

Conclusions: In our study population, 8 of 12 CIN prediction models within the literature were validated for use in patients undergoing EVAR and five are useful in identifying patients at risk for CIN.

Keywords: Acute kidney injury; Aortic aneurysm; Contrast-induced nephropathy; Endovascular procedures.