Pharmacological interventions for protecting renal function after cardiac surgery: a Bayesian network meta-analysis of comparative effectiveness.

Many drugs have been investigated as potentially protective of renal function after cardiac surgery. However, their comparative effectiveness has not been established. We performed an arm-based hierarchical Bayesian network meta-analysis including 95 randomised controlled trials with 28,833 participants, which allowed us to compare some agents not previously compared directly. Renal outcomes, including: the incidence of postoperative renal dysfunction and haemodialysis; serum creatinine level at 24 hours postoperatively; all-cause mortality; and length of hospital and ICU stay, were compared. Exploratory meta-regression was conducted for potential effect modifiers. A random effects model was selected according to the evaluation of model fit by deviance information criteria. Atrial natriuretic peptide (odds ratio (95%CrI) 0.28 (0.17-0.48); moderate-quality evidence), B-type natriuretic peptide, dexmedetomidine, levosimendan and N-acetyl cysteine significantly decreased the rate of postoperative renal dysfunction compared with placebo. Atrial natriuretic peptide (OR (95%CrI) 0.24 (0.10-0.58); low-quality evidence), B-type natriuretic peptide, and dexamethasone significantly decreased the need for haemodialysis. Levosimendan significantly decreased mortality, OR (95%CrI) 0.49 (0.27-0.91); low-quality evidence). The benefit of atrial natriuretic peptide was still apparent when baseline renal function was normal. None of the potential effect modifiers were significantly correlated with our renal outcomes. Atrial natriuretic peptide was ranked best regarding renal dysfunction, haemodialysis and length of hospital stay. Levosimendan was ranked best regarding mortality and ICU stay. However, our results should be interpreted cautiously given the assumptions made about transitivity and consistency.