Statistical analysis

J.B. and D.H. independently extracted data from included studies, cross-referenced for accuracy, and imported them into Review Manager version 5.3 (RevMan) for analysis.¹² Dichotomous outcomes were analyzed by the inverse variance random-effects method, and were expressed as risk ratios with 95% confidence intervals (CIs). Continuous outcomes were also analyzed by an inverse variance random-effects method, and were expressed as mean difference with 95% CIs. A random-effects model was used because of the suspected heterogeneity in types of continuous ultrafiltration methods used, underlying cardiac pathology, and patient risk profile. A meta-analysis was performed only if at least 2 included studies reported the same outcome. As stated in the prespecified protocol, any statistically significant difference in the primary and key secondary outcomes was deemed clinically relevant.

Statistical heterogeneity was measured by the χ² test (with P < 0.1 indicating significant heterogeneity) and was described by the I² statistic. I² > 75% suggests substantial heterogeneity, and outcomes that exhibit this pattern underwent investigation to better understand the root causes of the heterogeneity between studies. Reporting bias examination by a funnel plot analysis was completed if 10 or more studies reported on an outcome. One prespecified subgroup analysis was completed that differentiated patients by operative risk profile, as follows: low risk (STS or EuroSCORE II mortality risk score < 4) vs moderate or high risk (STS or EuroSCORE II mortality risk score > 4 or the presence of severe medical comorbidity or organ dysfunction). Examples of preoperative organ dysfunction include renal, cardiac, pulmonary, and hepatic failure. Test for subgroup interactions was completed using RevMan.¹²

A sensitivity analysis evaluated the meta-analysis results. Studies that were judged to have a high risk of bias, via the Cochrane RoB2 tool, were excluded from the pooled analysis for comparison with the primary results.