EMR is not inferior to ESD for early Barrett's and EGJ neoplasia: An extensive review on outcome, recurrence and complication rates.

Background and study aims In recent years, it has been reported that early Barrett's and esophagogastric junction (EGJ) neoplasia can be effectively and safely treated using endoscopic mucosal resection (EMR) and endoscopic submucosal dissection (ESD). Multiband mucosectomy (MBM) appears to be the safest EMR method. The aim of this systematic review is to assess the safety and efficacy of MBM compared with ESD for the treatment of early neoplasia in Barrett's or at the EGJ. Methods A literature review of studies published up to May 2013 on EMR and ESD for early Barrett's esophagus (BE) neoplasia and adenocarcinoma at the EGJ was performed through MEDLINE, EMBASE and the Cochrane Library. Results on outcome parameters such as number of curative resections, complications and procedure times are compared and reported. Results A total of 16 studies met the inclusion criteria for analysis in this study. There were no significant differences in recurrence rates when comparing EMR (10/380, 2.6 %) to ESD (1/333, 0.7 %) (OR 8.55; 95 %CI, 0.91 - 80.0, P = 0.06). All recurrences after EMR were treated with additional endoscopic resection. The risks of delayed bleeding, perforation and stricture rates in both groups were similar. The procedure was considerably less time-consuming in the EMR group (mean time 36.7 min, 95 %CI, 34.5 - 38.9) than in the ESD group (mean time 83.3 min, 95 %CI, 57.4 - 109.2). Conclusions The MBM technique for EMR is as effective as ESD when comparing outcomes related to recurrence and complication rates for the treatment of early Barrett's or EGJ neoplasia. The MBM technique is considerably less time-consuming.

Introduction

The incidence of adenocarcinoma around the esophagogastric junction (EGJ) as a consequence of gastroesophageal reflux disease (GERD) and Barrett’s esophagus (BE) has increased in Western countries over the past decades 1. Esophagectomy has long been regarded as the standard treatment following the detection of high-grade dysplasia (HGD) and early cancer (EC). However, esophagectomy is a complex surgical procedure with a reported mortality rate ranging between 3.0 % and 12.2 % 2. In recent years, endoscopic mucosal resection (EMR) was introduced for the treatment of HGD and EC (T1(m) adenocarcinoma) in patients with early Barrett’s neoplasia with reported 5-year survival rates exceeding 95 %. EMR is far less invasive than surgical resection3 4 5 6 7 and appears to be safe. The reported perforation rates using a capped-EMR technique range from 5 % to 7 % 8 9. The more recently introduced technique of multiband mucosectomy (MBM) appears to be even safer, with perforation rates reported in the range of 0 % to 1.2 %.8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 Importantly, endoscopic resection of early neoplasia in BE is associated with recurrence of metachronous neoplasia in remaining Barrett’s mucosa in up to 30 % of cases 24 25. Stepwise radical endoscopic resection (SRER) is a promising technique, particularly when using the MBM technique, and allows larger areas to be resected based on side-by-side piecemeal resections to eradicate not only the neoplastic lesion but all of the Barrett’s mucosa. The major drawback of circumferential resection is the high stricture rate. An alternative treatment is endoscopic submucosal dissection (ESD) which was originally introduced for the endoscopic treatment of early gastric cancer in Japan 26 27. ESD was developed for the en-bloc resection of large lesions 28 and enables precise histopathological assessment of specimens 29. ESD has emerged as the superior technique compared with piecemeal EMR when comparing recurrence rates in the endoscopic treatment of early gastric cancer, achieving overall 5-year recurrence-free rates of 100 % versus 82.5 %, respectively.30 It has recently been reported that ESD is used to treat early Barrett’s neoplasia and T1(m) adenocarcinoma at the EGJ 31 32 33 34 35 36.

To our knowledge, no literature is available which compares the efficacy of both techniques in the setting of distal esophageal and EGJ neoplasia. The aim of this literature review is to assess the safety and efficacy of MBM compared with ESD for the treatment of early neoplasia in Barrett’s and at the EGJ. The MBM technique was chosen because we consider this EMR technique to be superior to the capped-EMR technique when comparing perforation rates 8 9.

Methods

This literature review entails recently developed EMR techniques performed mainly by a multiband ligator device, 12 but also the cap technique, 37 and free-hand technique 12. For this purpose, we searched MEDLINE, EMBASE and the Cochrane Library to obtain all studies on EMR and/or ESD for Barrett’s esophagus neoplasia and adenocarcinoma at the EGJ that had been published up to May 2013. The following key words were used; “endoscopic mucosal resection (EMR),” “Barrett’s esophagus (BE),” “esophageal cancer,” “stepwise radical endoscopic resection (SRER),” “multiband mucosectomy (MBM),” “endoscopic submucosal dissection (ESD),” “esophagogastric junction (EGJ)” and “endoscopic mucosal resection (EMR).” All studies were screened according to the following inclusion and exclusion criteria.

In this study, we analyzed the MBM technique in particular because recent literature supports the view that MBM is safe and efficient for Barrett’s neoplasia at the EGJ in Western countries. There are no reports including only the MBM technique. MBM was not analyzed in direct comparison to other EMR techniques such as the capped-EMR owing to a lack of literature on other EMR techniques for Barrett’s neoplasia.

Studies on Barrett’s neoplasia and adenocarcinoma at the EGJ and lower esophagus.

Studies at least reporting on the multiband mucosectomy technique in EMR cohorts (multiband mucosectomy, the cap technique and free-hand technique).

Studies reporting clinical outcomes on the recurrence rate (local recurrence and distant metastasis), complete eradication rates, curative resection rates, complications (delayed bleeding, perforation and stricture), and procedure times.

Animal experiments

Case reports (less than five cases)

Review articles

Editorials

Abstract-only publications

Publications in a language other than English

Training program

Combination therapy with radiofrequency ablation (RFA)

Studies with less than 6 months of follow-up

Exclusion criteria

From the studies, we extracted the following information: first author, year of publication, country, research design, number of individuals in the EMR and ESD procedures, intervention types, follow-up period and the clinical outcomes. The reported clinical outcomes were the rates of recurrence (local recurrence and distant metastasis), complete eradication, curative resection, complications (bleeding, perforation and stricture), and procedure time. The term “complete eradication rate” is used to confirm the absence of neoplasia in any of the follow-up biopsy samples after several EMRs had been performed to eradicate Barrett’s neoplasia 38. Curative resection rates are histologically defined by a resection in which the lateral and vertical margins of the specimens are free of cancer and without submucosal invasion beyond the muscularis mucosae, lymphatic invasion, or vascular involvement 39 40.

EMR techniques

MBM was performed using the Duette Multiband Mucosectomy kit (Cook Endoscopy, Limerick, Ireland). This consists of a transparent cap with six rubber bands and an attachment for releasing wires, and a 5 – or 7-Fr hexagonally braided polypectomy snare. After applying markings surrounding the lesion, the tissue is sucked into the cap, and a rubber band is released creating a pseudo-polyp. The snare is placed under the rubber band and the pseudo-polyp is resected using pure coagulation current. The resected specimen is passed into the stomach and the adjacent mucosa subsequently resected in the same fashion until all markings have been included. Finally, the specimens are collected using a retrieval net 12 22. In the cap technique, a flexible oblique cap (diameter 18 mm) for en-bloc resection, piecemeal procedures or a standard hard cap (diameter 12.8 /14.8 /18 mm, MAJ-296 /297 or D206 – 5, Olympus Europe) are used. After mucosal marking, lesions are lifted by submucosal injection before being sucked into the cap. A preloaded snare in the rim of the cap is then pulled firmly and the lesion resected using EndoCut electrocoagulation 37. The free-hand technique is a standard lift and snare mucosectomy 10.

ESD technique

ESD procedures are performed using endo-knives, such as an insulation-tipped knife (IT knife) (KD-610L; Olympus Optical, Tokyo, Japan), IT knife2 (KD-611L; Olympus) or Flex knife (KD-630L; Olympus). A transparent hood (D-201 – 11804; Olympus) is attached to the tip of the endoscope. An electrosurgical generator (ICC200 or VIO300D; [ERBE Tubingen, Germany] or ESG100; [Olympus]) is connected to the endo-knife. Markings along the presumed cutting line are applied around the lesion. A saline solution with epinephrine solution (0.025 mg/ml) or a mixture of a glycerine solution with normal saline plus 5 % fructose (Glyceol; Chugai Pharmaceutical, Tokyo, Japan) or hyaluronic acid (MucoUp; Johnson & Johnson Japan, Tokyo, Japan) is injected into the submucosa for lifting. Circumferential cutting is performed using the endo-knife. Subsequent submucosal dissection is performed by using the endo-knife until achieving complete resection of the lesion 26 27 28 41 42.

Statistical analysis

The aim of the analysis was to compare the outcomes of two distinct techniques to treat Barrett’s esophagus neoplasia and adenocarcinoma at the EGJ, specifically MBM and ESD.

One approach to the analysis would be to use meta-analysis methods to combine the results from the different studies. However, several of the outcomes were binary in nature, and some of these outcomes did not occur in any of the patients in most studies. This prohibits the calculation of standard errors for the probability of the outcome occurring, which are required for a meta-analysis.

Instead, the original patient level data were recreated from the summaries reported in each paper. The occurrence of each outcome was compared between techniques using multilevel logistic regression. Two level models were used with individual patients nested within the study.

Results (Tables 1 – 4)

A total of 16 studies met the inclusion criteria for this study. Ten EMR studies originated as follows: five from the Netherlands 8 17 19 21 22, two from Germany 4 12, and one each from the USA 20, UK 23, and Australia 18. All six ESD studies that met the inclusion were from Japan 31 32 33 34 35 36. These studies entailed a total of 761 lesions in the EMR group and 335 lesions in the ESD group. All studies were published between January 2006 and May 2013.

Recurrence rates (local recurrence and distant metastasis) (Table 5)

Eleven studies in both groups 4 17 18 21 23 31 32 33 34 35 36 reported on recurrence rates. For ESD studies, the mean follow-up time was 28.7 months, while the equivalent number for the EMR studies was 25.6 months. Analysis showed that the recurrence rate was slightly higher in the EMR group (10/380, 2.8 %) compared with the ESD group (1/333, 0.3 %), but the difference did not reach statistical significance (odds ratio 8.55; 95 %CI, 0.91 – 80.0, P = 0.06). Two cases in the ESD group were excluded on account of having less than 6 months’ follow-up.

With regard to EMR procedures, complete eradication rates are described for the SRER procedures 35, while curative resection rates are reported for ESD procedures 36 37. Complete eradication rate in the EMR group was 363/380 (95.5 %). Curative resection rate in the ESD group was 253/335 (75.5 %). Non-curative resections were mainly because of submucosal invasions of more than 500 μm and/or lymphatic and venous invasion. Recurrence of metachronous neoplasia in the EMR group was managed by additional endoscopic resection 4 17 22. One case of distant metastasis was reported in the ESD group where the tumor depth was pT1sm and the patient declined additional surgical treatment 32.

Because of a lack of long-term data on prognosis and disease-free survival in the studies included, the results in this literature review only apply to a relatively short-term prognosis with a mean of 30 months ranging from 8 to 47 months.

Complication rates (Table  6)

Delayed bleeding rates

It proved impossible to compare rates of bleeding during the procedures as the individual definitions of acute bleeding are very different. In many instances, (small) bleedings are considered to be an integral part of the procedure and not a complication. Therefore, we focused on delayed bleeding. Fifteen studies 4 8 12 17 19 20 22 23 31 32 33 34 35 36 reported the occurrence of delayed bleedings with regard to 686 lesions in the EMR group and 335 lesions in the ESD group. The delayed bleeding rate in the EMR group (8/686, 1.2 %) was similar to that in the ESD group (7/335, 2.1 %), and the difference was not statistically significant (odds ratio 0.46; 95 %CI, 0.12 – 1.75, P = 0.26). All cases of delayed bleeding were effectively managed endoscopically. Blood transfusion was required in three cases 19 in the EMR group and in none in the ESD group.31 32 33 34

Perforation rates

Fifteen studies 4 8 12 17 19 20 22 23 31 32 33 34 35 36 reported on perforation rates. The perforation rate in the EMR group (8 /686, 1.2 %) was similar to that in the ESD group (5/335, 1.5 %), and the difference was not statistically significant (odds ratio 1.07; 95 %CI, 0.20 – 5.62, P = 0.94). In the EMR group, six patients were managed conservatively with clips (three cases), covered stents (two cases), and observation (one case) 17 22. Two patients were treated surgically (no detailed description). 8 17 In the ESD group, one patient was managed conservatively with clips 31 and the other four patients were not described 36.

Stricture rates

Fifteen studies 4 8 12 17 18 19 20 21 22 23 31 32 33 34 35 reported on stricture rates with regard to the treatment of 761 lesions in the EMR group and 157 lesions in the ESD group. There were three strictures reported in 456 EMRs for neoplastic lesions alone (no attempt was made to eradicate the whole Barrett’s segment). This accounts for a stricture rate of 0.7 % for lesional EMR of only neoplastic areas (odds ratio 0.21; 95 %CI, 0.03 – 1.41, P = 0.11). The overall stricture rate was higher in the EMR group when all SRER cases were included (170 /761, 22.3 %). In the SRER group, the stricture rate was very high (167 /305, 54.7 %). In the ESD group, seven strictures were reported in 207 cases (7 /207, 3.4 %). These results were similar to those in the EMR group. Symptomatic strictures required intervention with bougienage or balloon dilatation. Two cases in the SRER group were treated surgically on account of perforation after dilatation 12 17.

Procedure times ( Table 7)

Seven studies 8 22 31 33 34 35 36 reported on the procedure time including treatment of 82 lesions in the EMR group and 310 lesions in the ESD group. The analysis showed that the procedure time was less time-consuming in the EMR group (mean time 36.7 min, 95 %CI, 34.5 – 38.9) compared with the ESD group (mean time 83.3 min, 95 %CI, 57.4 – 109.2). The procedure time was analyzed as a continuous variable. To pool the results between different studies, information on the mean procedure time is required for each study as well as either the standard deviation or standard error. Since these data were not available, it is impossible to calculate the statistical significance.

Discussion

This review demonstrates that, when comparing immediate and short-term outcomes, EMR is not inferior to ESD for the treatment of early Barrett’s or EGJ neoplasia. The recurrence rate was slightly higher in the EMR group compared with the ESD group, but the difference was not statistically significant. More importantly, all recurrences in the EMR group were managed by additional endoscopic resections. SRER comprises complete resection to eradicate all intestinal metaplasia at risk of malignant degeneration. In SRER, the complete eradication rate was extremely high (95.5 %). The recurrence rate of intestinal dysplasia after SRER (2.8 %) was superior compared with conventional lesional EMR, which ranged from 14 % to 23 % of cases 11 40. However, these rates included metachronous neoplasia, and as a consequence, it is difficult to compare recurrence rates. All recurrences after EMR and SRER were treated with additional endoscopic resection. This approach to dealing with residual neoplasia is supported by a large German study that states that endoscopic resection should be accepted as the treatment of choice in most patients with high‐grade intraepithelial neoplasia (HGIN) and mucosal carcinoma in the esophagus. The rate of complete response was 96.6 %, and long-term complete response after re-treatment of metachronous neoplasia (21.5 %) was 94.5 % 43. Also, endoscopic therapy is highly effective and safe for patients with mucosal adenocarcinoma, with excellent long-term results. In an almost 5-year follow-up of 1000 patients treated with endoscopic resection, there was no mortality and less than 2 % had major complications. This study suggests that endoscopic therapy should become the standard of care for patients with mucosal adenocarcinoma 44. The results from these studies combined with our current study demonstrate no additional benefits from an oncological point of view of ESD over EMR in the treatment of early Barrett’s or EGJ neoplasia.

The risk of delayed bleeding and perforation rates in both groups was similar. Stricture formation is a common complication of endoscopic resection resulting in increased stricture rates with increasing proportions of the diameter resected. The analysis showed that the stricture rate was similar in both groups when comparing resection of the neoplastic lesion alone. Stricture rates increased rapidly in the SRER group when the complete Barrett’s mucosa was resected. Likewise, in the ESD group, the post-esophageal stricture rate may increase with larger proportions of the diameter being resected 33. Symptomatic strictures require intervention via bougienage or balloon dilatation, and are usually easily managed. Unfortunately, the number of dilatation sessions needed to manage these strictures was not reported in any of these studies.

The major drawback of ESD is the long procedure time, particularly in difficult positions, such as the EGJ. This can be a disadvantage in elderly patients or patients who are unable to undergo lengthy procedures or would require propofol sedation. We were unable to determine a statistically significant difference in procedure times between these two procedures because of the lack of data available on procedure times in the EMR studies. However, there was a huge difference in the small number of studies that did report on procedure times. We therefore believe that it is justified to say that, in general, the EMR procedure takes considerably less time to complete when compared with ESD.

It is difficult to achieve the same level of expertise in ESD techniques in Western countries as in Asian countries, mainly because the incidence of early gastric cancer is very low 1 45 46 47. Our review shows that the MBM technique, which is far easier to learn, is safe and as effective in treating early Barrett’s or EGJ neoplasia 8 9 12 13 14 15 16 17 18 19 20 21 22 23.

Some limitations of this literature review should be taken into consideration. First, all studies included were limited by the constraints of a non-randomized design. Second, all studies involved a non-concurrent comparison group. Third, the EMR studies were performed in Western countries, while the ESD studies were performed in Japan. Fourth, there is a lack of long-term data for both sets of studies. Fifth, the definition of therapeutic evaluation after each endoscopic treatment differs between EMR and ESD. Sixth, the results from the EMR studies are fairly heterogeneous and do not always include all parameters that were compared in this review, such as recurrence rates, complication rates, and procedure times. It is because of this that all six ESD papers are compared with different numbers of EMR reports throughout this study. Finally, several studies on EMR originate from a multicenter group, where it is possible that the data might be overlapping in these separate studies.

In conclusion, the MBM technique for EMR appears as effective as ESD when comparing important outcome parameters on the eradication of early Barrett’s or EGJ neoplasia. Our review supports the non-inferiority in oncological treatment in the short term where others have confirmed excellent results in the long term. There are no differences in outcome when comparing strictures, bleedings and perforation rates for both EMR and ESD in experienced hands. The MBM technique has considerable advantages in that it is easier to master, and is less time-consuming. Further studies involving randomized, controlled trials with the MBM technique versus ESD in early Barrett’s or EGJ neoplasia need to be performed to corroborate these results.

Recurrence rates, complete eradication rates/curative resection rates.

Author	Number	Recurrence rate	Follow-up	Range	Complete eradication
EMR group
Ell et al. 2007 4	100 (EMR) M/C	6 % (6 /100)	33 months (median)	range 2 – 83	99 % (99 /100)
Moss et al. 2010 18	75 (EMR) M/C	0 % (0 /35) (CBE)	31 months (mean)	range 3 – 68	94 % (33/35 CBE)
	5 not available	0 % (0 /35) (non-CBE)	31 months (mean)	range 3 – 89	89 % (31/35 non-CBE)
Thomas et al. 2009 23	16 (EMR) M	0 % (0 /16)	8 months (mean)	IQR 6 – 12	87.5 % (14/16)
Pouw et al. 2010 17	169 (EMR) M/C/F	1.8 % (3 /169)	32 months (median)	IQR 19 – 49	95.3 % (161/169)
van Vilsteren et al. 2011 21	25 (EMR) M/C/F	4 % (1 /25)	25 months (median)	IQR 19 – 29	100 % (25/25)
ESD group
Kakushima et al. 2006 31	30 (ESD)	0 % (0 /28) discarding 2 cases (follow-up less than 6 months)	14.6 months (mean)	range 6 – 31	70 % (21/30)
Yoshinaga et al. 2008 32	25 (ESD)	4 % (1 /25) including 1 recurrent case (declined surgery)	36.6 months (median)	range 4 – 94	72 % (18/25)
Hirasawa et al. 2010 33	58 (ESD)	0 % (0 /58)	30.6 months (median)	range 1.2 – 54.9	79 % (46/58)
Omae et al. 2013 34	44 (ESD)	0 % (0 /44)	33 months (mean)	range 6 – 64	84.1 % (37/44)
Imai et al. 2013 35	50 (ESD)	0 % (0 /50)	47 months (median)	range 22 – 97	72 % (36/50)
Hoteya et al. 2013 36	128 (ESD)	0 % (0 /128)	34 months (median)	range 2 – 96	74 % (95/128)

Abbreviations: C, cap; CBE, complete Barrett’s excision; F, free hand; IQR, interquartile range; M, multiband mucosectomy.

Complication rates.

Author	Number	Delayed bleeding	Perforation	Stricture	Stricture (exclusion of SRER)
EMR group
Soehendra et al. 2006 12	10 (EMR) M	0 % (0 /10)	0 % (0 /10)	70 % (7 /10)	0 % (0 /0)
Ell et al. 2007 4	100 (EMR) M/C	0 % (0 /100)	0 % (0 /100)	0 % (0 /100)	0 % (0 /100)
Peters et al. 2007 22	40 (EMR) M	0 % (0 /40)	0 % (0 /40)	0 % (0 /40)	0 % (0 /40)
Thomas et al. 2009 23	16 (EMR) M	0 % (0 /16)	0 % (0 /16)	0 % (0 /16)	0 % (0 /16)
Pouw et al. 2010 17	169 (EMR) M/C/F	1.8 % (3 /169)	2.4 % (4 /169)	50 % (84 /169)	0 % (0 /0)
Moss et al. 2010 18	75 (EMR) M/C	Not available	Not available	8 % (6 /75)	1 % (1 /70)
Pouw et al. 2011 8	42 (EMR) M	0 % (0 /42)	2 % (1 /42)	0 % (0 /42)	0 % (0 /42)
Alvarez Herrero et al. 201119	243 (EMR) M	2 % (5 /243)	0 % (0 /243)	13 % (33 /243)	0 % (0 /174)
Gerke et al. 2011 20	41 (EMR) M/C	0 % (0 /41)	4.9 % (2 /41)	44 % (18 /41)	14 % (2 /14)
van Vilsteren et al. 201121	25 (EMR) M/C/F	0 % (0 /25)	4 % (1 /25)	88 % (22 /25)	0 % (0 /0)
ESD group
Kakushima et al. 2006 31	30 (ESD) EGJ	0 % (0 /30)	3 % (1 /30)	3 % (1 /30)	3 % (1 /30)
Yoshinaga et al. 2008 32	25 (ESD) EGJ	0 % (0 /25)	0 % (0 /25)	8 % (2 /25)	8 % (2 /25)
Hirasawa et al. 2010 33	58 (ESD) EGJ	5 % (3 /58)	0 % (0 /58)	2 % (1 /58)	2 % (1 /58)
Omae et al. 2013 34	44 (ESD) EGJ	0 % (0 /44)	0 % (0 /44)	0 % (0 /44)	0 % (0 /44)
Imai et al. 2013 35	50 (ESD) EGJ	6 % (3 /50)	0 % (0 /50)	6 % (3 /50)	6 % (3 /50)
Hoteya et al. 2013 36	126 (ESD) EGJ	0.7 % (1 /128)	3 % (4 /128)	Not available	Not available

Procedure times.

Author	Methods	Procedure time	Median or mean	Standard deviation
EMR group
Peters et al. 2007 22	40 (EMR) M	37 min (range 28 – 58)	Median	No
Pouw et al. 20118	42 (EMR) M	34 min (IQR 20 – 52)	Median	No
ESD group
Kakushima et al. 2006 31	30 (ESD) EGJ	70 min (range 20 – 120)	Mean	No
Hirasawa et al. 2010 33	58 (ESD) EGJ	82 min (range 22 – 275)	Mean	No
Omae et al. 2013 34	44 (ESD) EGJ	121 min (range 49 – 272)	Median	No
Imai et al. 2013 35	50 (ESD) EGJ	42.5 min (range 10 – 157)	Median	No
Hoteya et al. 2013 36	128 (ESD) EGJ	102.6 min (range 32.6 – 171.4)	Mean	No

Average sizes of resected specimens.

Author	Number	Average size (resected specimen)	Median or mean
EMR group
Soehendra et al. 2006 12	10 (M)	14.3 ± 4.1 mm (range 7 – 22) (per specimen) mean × 2 piece (range1 – 5) median	Mean/median
Peters et al. 2007 22	40 (M)	17 mm (SD 6.3) (per specimen) × 6 piece (SD 3.5)	Mean
Moss et al. 2010 18	75 (EMR) M/C	14 mm (range 9 – 29) (per specimen) × 3 pieces (range 1 – 10)	Mean
Thomas et al. 2009 23	16 (EMR) M	3 cm (IQR 2 – 5)	Median
Pouw et al. 2010 17	169 (EMR) M/C/F	3 cm (range 2 – 5)	Median
Pouw et al. 2011 8	42 (M)	18 mm (range 15 – 20) (per specimen) × 5 piece (range 3 – 7)	Median
Alvarez Herrero et al. 2011 19	243 (M)	C4M6 cm (IQR C1 – 7 cm, M 3 – 8)	Median
Gerke et al. 2011 20	41 (M/C)	3 cm (range 1 – 8)	Mean
van Vilsteren et al. 2011 21	25 (EMR) M/C/F	C2M4 cm (range C1 – 3, M 2 – 5)	Median
ESD group
Kakushima et al. 2006 31	30 (ESD)	40.6 mm (range 20 – 80)	Mean
Yoshinaga et al. 2008 32	25 (ESD)	40 mm (range 25 – 70)	Mean
Hirasawa et al. 2010 33	58 (ESD)	37.7 mm (range 14 – 67)	Mean
Omae et al. 2013 34	44 (ESD)	35 mm (range 15 – 58)	Mean
Imai et al. 2013 35	50 (ESD)	40.5 mm (range 24 – 85)	Median
Hoteya et al. 2013 36	128 (ESD)	21.4 mm (range 2.6 – 37.8)	Mean

Recurrence rates.

Outcome	ESD		EMR		Odds ratio	P-value
	No. of studies	N (%)	No. of studies	N (%)	(95 % CI)
Recurrence rate	6	1 /333 (0.3 %)	5	10 /380 (2.6 %)	8.55 (0.91, 80.0)	0.06

Complication rates (delayed bleeding, perforation, and stricture).

Outcome	ESD		EMR		Odds Ratio	P-value
	No. of studies	N (%)	No. of studies	N (%)	(95 % CI)
Delayed bleeding	6	7 /335 (2.1 %)	9	8 /686 (1.2 %)	0.46 (0.12, 1.75)	0.26
Perforation	6	5 /335 (1.5 %)	9	8 /686 (1.2 %)	1.07 (0.20, 5.62)	0.94
Stricture (including SRER for EMR group)	5	7 /207 (3.4 %)	10	170 /761 (22.3 %)	5.38 (0.28, 105)	0.27
Stricture (EMR alone)	5	7 /207 (3.4 %)	7	3 /456 (0.7 %)	0.21 (0.03, 1.41)	0.11

Procedure times.

Method	No. of studies	Pooled procedure time (95 % CI)
EMR	2	36.7 (34.5, 38.9)
ESD	5	83.3 (57.4, 109.2)