Fluid type and the use of renal replacement therapy in sepsis: a systematic review and network meta-analysis
Additional Document Info
Fluid resuscitation, along with the early administration of antibiotics, is the cornerstone of treatment for patients with sepsis. However, whether differences in resuscitation fluids impact on the requirements for renal replacement therapy (RRT) remains unclear. To examine this issue, we performed a network meta-analysis (NMA), including direct and indirect comparisons, that addressed the effect of different resuscitation fluids on the use of RRT in patients with sepsis. The data sources MEDLINE, EMBASE, ACPJC, CINAHL and Cochrane Central Register were searched up to March 2014. Eligible studies included randomized trials reported in any language that enrolled adult patients with sepsis or septic shock and addressed the use of RRT associated with alternative resuscitation fluids. The risk of bias for individual studies and the overall certainty of the evidence were assessed. Ten studies (6664 patients) that included a total of nine direct comparisons were assessed. NMA at the four-node level showed that an increased risk of receiving RRT was associated with fluid resuscitation with starch versus crystalloid [odds ratio (OR) 1.39, 95% credibility interval (CrI) 1.17-1.66, high certainty]. The data suggested no difference between fluid resuscitation with albumin and crystalloid (OR 1.04, 95% CrI 0.78-1.38, moderate certainty) or starch (OR 0.74, 95% CrI 0.53-1.04, low certainty). NMA at the six-node level showed a decreased risk of receiving RRT with balanced crystalloid compared to heavy starch (OR 0.50, 95% CrI 0.34-0.74, moderate certainty) or light starch (OR 0.70, 95% CrI 0.49-0.99, high certainty). There was no significant difference between balanced crystalloid and saline (OR 0.85, 95% CrI 0.56-1.30, low certainty) or albumin (OR 0.82, 95% CrI 0.49-1.37, low certainty). Of note, these trials vary in terms of case mix, fluids evaluated, duration of fluid exposure and risk of bias. Imprecise estimates contributed to low confidence in most estimates of effect. Among the patients with sepsis, fluid resuscitation with crystalloids compared to starch resulted in reduced use of RRT; the same may be true for albumin versus starch.