Search

Background Endoscopic submucosal dissection (ESD) is widely used to treat gastrointestinal lesions. A traction-assisted (TA) strategy has been recently developed for ESD. In this study, we evaluated the safety and efficacy of TA-ESD compared with conventional ESD (C-ESD). Methods We searched PubMed, Embase, Web of Science, and the China National Knowledge Infrastructure to identify randomized clinical trials that compared TA-ESD and C-ESD. No filters for language or date of publication were used. Outcomes included complete resection rate, resected specimen size, procedure time, hemostasis usage (frequency), overall complication rate, perforation rate, and delayed bleeding rate. We used the mean difference (MD) for continuous outcomes in a random-effects model and Peto odds ratio (POR) for binary outcomes where any zero cell existed. Effect sizes and their 95 % confidence intervals (CIs) were determined. Results 12 out of 929 identified articles, including 1499 patients, were analyzed. According to pooled results, TA-ESD produced similar R0 resections to C-ESD, but its procedure time (minutes) was shorter than that of C-ESD (MD – 16.02, 95 %CI – 22.71 to – 9.33). Moreover, TA-ESD had a lower complication rate (POR 0.47, 95 %CI 0.29 to 0.76) and perforation rate (POR 0.24, 95 %CI 0.10 to 0.56) than C-ESD. A nonsignificant difference in delayed bleeding rate was observed, although there was a trend toward this being lower in TA-ESD than C-ESD (POR 0.90, 95 %CI 0.46 to 1.75, I 2 = 12 %). Conclusions The traction-assisted strategy improves safety and efficacy in the treatment of patients with ESD. However, we observed different effect sizes in the esophagus, stomach, and colorectum.

The purpose of this letter was to explore how lesion size affects the pooled effect of traction-assisted endoscopic submucosal dissection on procedure time. Our meta-regression showed that lesion size negatively associated with the effect of traction-assisted endoscopic submucosal dissection on procedure time (estimate point = − 1.02; 95% confidence interval, from − 1.58 to − 0.46). We also confirmed this result in different statistical models including fixed effect regression and two mixed effects regression models. Electronic supplementary material The online version of this article (10.1186/s12957-019-1699-0) contains supplementary material, which is available to authorized users.