Thirteen RCTs (1,910 participants) were included. Eight trials reported adequate methods for randomisation. Six trials reported adequate methods for allocation concealment. Only one trial reported blinding of outcome assessors. All trials were at low risk of bias for incomplete outcome data but none were completely free from risk of other sources of bias.
Compared with traditional care, ERAS was associated with shorter primary length of stay (MD 2.44 days, 95% CI 1.83 to 3.06; Ι²=88%; 11 RCTs), total length of stay (MD 2.39 days, 95% CI 1.09 to 3.70; Ι²=85%; seven RCTs) and total complications (RR 0.71, 95% CI 0.58 to 0.86; Ι²=65%; 13 RCTs). Statistical heterogeneity was significant for all these outcomes. There was no significant difference for readmission rates (RR 0.93, 95% CI 0.56 to 1.54; Ι²=0%; 11 RCTs). Times to first flatus and stool were shorter in the ERAS group. Mortality did not differ between groups.
Quality of evidence was rated low for primary hospital stay, total complications and mortality and rated moderate for total hospital stay and readmission rate. Results for other outcomes were reported. Funnel plots were asymmetrical for some outcomes, which indicated possible publication bias.