METHODS: Systematic review and meta-analysis with random-effects inverse variance model. PubMed, Cochrane library, EMBASE, LILACS, and Web of Science were searched from inception to October 2016. Published and unpublished randomized controlled trials in any language that enrolled critically ill and/or perioperative adult patients and compared a low- to a highchloride solution for volume maintenance or resuscitation. The primary outcomes were mortality and RRT use. We conducted trial sequential analyses and assessed risk of bias of individual trials and the overall quality of evidence. Fifteen trials with 4067 patients, most at low risk of bias, were identified. Of those, only 11 and 10 trials had data on mortality and RRT use, respectively. A total of 3710 patients were included in the mortality analysis and 3724 in the RRT analysis.
RESULTS: No statistically significant impact on mortality (odds ratio, 0.90; 95% confidence interval, 0.69–1.17; P = .44; I 2 = 0%) or RRT use (odds ratio, 1.12; 95% confidence interval, 0.80–1.58; P = .52; I 2 = 0%) was found. Overall quality of evidence was low for both primary outcomes. Trial sequential analyses highlighted that the sample size needed was much larger than that available for properly powered outcome assessment.
CONCLUSIONS: The current evidence on low- versus high-chloride solutions for unselected critically ill or perioperative adult patients demonstrates no benefit, but suffers from considerable imprecision. We noted a limited exposure volume for study fluids and a relatively low risk of the populations in each study. Together with the relatively small pooled sample size, these data leave us underpowered to detect potentially important differences. Results from well-conducted, adequately powered randomized controlled trials examining sufficiently large fluid exposure are necessary.