Different concentrations of albumin versus crystalloid in patients with severe sepsis and septic shock: a meta-analysis of randomized clinical trials

Abstract Background The best type of resuscitation fluids for severe sepsis and septic shock patients remains unclear. The aim of this study was to evalute the effacy of different concentrations of albumin on reducing the mortality rate of theses patients by meta-analysis. Methods PubMed, EMBASE and Web of Science databases were used for screening the relevant studies. Ramdonized controlled trials (RCTs) were eligible if they compared the effects of albumin with crystalloid on mortality in severe sepsis and septic shock patients. Data were examined and extracted by two reviewers independently. Any disagreements were resolved by consensus with or without the help from a third reviewer. Data including mortality, sample size of the patients and resuscitation endpoints were extracted. Meta-analysis was carried out using STATA version 16.0, based on the corresponding odds ratios with 95% confidence intervals. Results Eight studies with a total of 5124 severe sepsis patients and 3482 septic shock patients were included in this study. Compared with crystalloid, the use of albumin slightly reduced the 90-day mortality of severe sepsis patients [OR 0.91 (0.80, 1.02); P = 0.11] and significantly improved the outcome of septic shock patients [OR 0.85 (0.74, 0.99); P = 0.04]. Further analysis showed a potentially benenifical role of both 4–5% and 20% albumin on reducing the mortality of severe sepsis patients. The use of 20% albumin significantly decreased the 90-day mortality of septic shock patients [OR 0.81 (0.67, 0.98); P = 0.03], which was better than 4–5% albumin and crystalloid. Conclusions Albumin treatment, particularly 20% albumin, significantly reduced the 90-day mortality in septic shock patients. Either 4–5% or 20% of albumin may work better than crystalloid in improving the survival rate of patients with severe sepsis, but more relative RCTs are required for validation.