Contrast-associated acute kidney injury in the critically ill: systematic review and Bayesian meta-analysis

Purpose: Critically ill patients, among whom acute kidney injury is common, are often considered particularly vulnerable to iodinated contrast medium nephrotoxicity. However, the attributable incidence remains uncertain given the paucity of observational studies including a control group. This study assessed acute kidney injury incidence attributable to iodinated contrast media in critically ill patients based on new data accounting for sample and effect size and including a control group. Methods: Systematic review of studies measuring incidence of acute kidney injury in critically ill patients following contrast medium exposure compared to matched unexposed patients. Patient-level meta-analysis implementing a Bayesian nested mixed effects multiple logistic regression model. Results: Ten studies were identified; only four took into account the baseline acute kidney injury risk, three by patient matching (560 patients). Objective meta-analysis of these three studies (vague and impartial a priori hypothesis concerning attributable acute kidney injury risk) did not find that iodinated contrast media increased the incidence of acute kidney injury (odds ratio 0.95, 95% highest posterior density interval 0.45–1.62). Bayesian analysis demonstrated that, to conclude in favor of a statistically significant incidence of acute kidney injury attributable to contrast media despite this observed lack of association, one’s a priori belief would have to be very strongly biased, assigning to previous uncontrolled reports 3–12 times the weight of evidence strength provided by the matched studies including a control group. Conclusions: Meta-analysis of matched cohort studies of iodinated contrast medium exposure does not support a significant incidence of acute kidney injury attributable to iodinated contrast media in critically ill patients.