Twelve studies (3,806 gastric lesions; ESD = 1,734; EMR = 2,072), comprising three non-concurrent cohort studies (that compared a prospective treatment group with a comparison group in the past) and nine retrospective cohort studies, were included in the meta-analysis. All studies were considered to be well-conducted; no further quality details were reported.
En bloc resection rate:
ESD had a significantly higher rate than EMR (OR 8.43, 95% CI 5.20 to 13.67; 10 studies; Ι²=60%). Results were similar in subgroup analyses.
Complete resection:
ESD had a significantly higher rate or complete resection than EMR (OR 8.54, 95% CI 4.44 to 16.45; seven studies; Ι²=83%). Results were similar in subgroup analyses.
Curative resection:
ESD resulted in a statistically significant higher rate than EMR (OR 3.28, 95% CI 1.95 to 5.54; five studies; Ι²=68%). Sensitivity analysis did not alter the main finding.
Local recurrence:
A significantly lower recurrence rate was reported for ESD than EMR (RR 0.13, 95% CI 0.04 to 0.41; nine studies; Ι²=47%). subgroup analyses showed this was statistically significant only at the two to four year observation period (four studies; Ι²=23%). ESD was statistically superior only for tumours larger than 10 mm (two studies).
Mortality:
ESD resulted in a lower mortality rate than EMR, but this was not statistically significant (two studies: Ι²=0%).
Bleeding:
Intra-operative bleeding was higher in the ESD group (RR 2.16, 95% CI 1.14 to 4.09; three studies), but there was no statistically significant difference between groups for overall bleeding (12 studies; Ι² = 48%). Subgroup analyses showed a significantly higher bleeding risk for ESD with the mixed knife than for EMR, but ESD with other knife types showed no significant difference in bleeding rates compared with EMR.
Perforation:
Perforation rate was significantly higher in the ESD group (RR 3.58, 95%CI 1.95 to 6.55; 12 studies; Ι²=30%), subgroup analysis showed that this difference was only statistically significant for patients with primary disease (nine studies; Ι²=43%).
Resection time:
ESD was associated with a significantly longer operation time than EMR (SMD 1.55, 95% CI 0.74 to 2.37; seven studies; Ι²=97%)
There was no evidence of publication bias.