Endoscopic Mucosal Resection (EMR) and Endoscopic Submucosal Dissection (ESD)

Updated: Aug 31, 2022
  • Author: Raymond E Kim, MD; Chief Editor: Kurt E Roberts, MD  more...
  • Print


Endoscopic mucosal resection

Endoscopic mucosal resection (EMR) is a technique used for the staging and treatment of superficial neoplasms of the gastrointestinal (GI) tract. This technique was first developed in Japan for the treatment of early gastric cancer (EGC) and has since spread in use throughout the world for various indications, including dysplastic Barrett mucosa and sessile colonic neoplasms. The utility of EMR rests in its ability to do the following:

  • Provide accurate histologic staging of superficial GI neoplasms
  • Provide a minimally invasive technique for removal of superficial malignancies

Several variations of EMR are currently used, including injection-assisted, cap-assisted, and ligation-assisted techniques. All adhere to the basic principles of identification and demarcation of the lesion, submucosal injection to lift the lesion, and endoscopic snare resection. By virtue of its overall safety and efficacy in appropriately selected patient populations, EMR has become firmly integrated into the diagnostic and treatment algorithms of superficial GI malignancies.

Some authors have also studied the use of EMR in a procedure known as antireflux mucosectomy (ARMS) to treat refractory gastroesophageal reflux disease (GERD). [1, 2]  

Endoscopic submucosal dissection

Endoscopic submucosal dissection (ESD) was developed to resect larger tumors and aid in achieving higher rates of en-bloc resection than would be possible with EMR. The main goal of ESD is to achieve an R0 resection. In the United States, ESD is primarily performed in select centers by specialized endoscopists experienced in this technique. ESD is generally indicated for the following [3] :

  • Tumors that are diagnosed as carcinomas with intramucosal–to–superficial submucosal invasion
  • Lesions with submucosal fibrosis that cannot be removed by EMR even if smaller than 20 mm
  • Cases where a snare is unlikely to enable a successful en-bloc resection with EMR
  • Removal of large polyps, early colorectal cancer, and those lesions that cannot be accessed transanally in patients who wish to avoid major surgical resection

Endoscopic and ultrasonographic characterization of lesions

Several classification systems for the staging of early GI cancers that may aid in the prediction of lymph node metastases have been developed. Much of this work was pioneered by Japanese gastroenterologists for the staging and treatment of EGC. The Japanese Society of Gastroenterology (JSGE), working from large databases of EGC resections, classified lesions according to their endoscopic features and the implied risk of mural invasion. [4]

The subsequent Paris classification, developed in 2002 at an international consensus meeting, echoed the structure of the JSGE system. In this classification, superficial (type 0) lesions are divided into polypoid (0-I) and nonpolypoid categories (0-II), which are further subcategorized as pedunculated (0-Ip), sessile (0-Is), slightly elevated (0-IIa), flat (0-IIb), slightly depressed (0-IIc), or excavated (0-III). [5]  (See the image below.)

Paris classification of superficial gastrointestin Paris classification of superficial gastrointestinal neoplasms.

In the Vienna classification, lesions are divided into two broad categories, noninvasive (low-grade dysplasia, high-grade dysplasia [HGD]) and invasive (intramucosal cancer, cancer that infiltrates the submucosa). [6]

The mucosal layer is divided into upper, middle, and lower layers: m1 (epithelium), m2 (lamina propria), and m3 (muscularis mucosae). The submucosa is similarly divided into three layers: sm1, sm2, and sm3. Submucosal tumor involvement of 500 μm or less below the muscularis mucosae is characterized as sm1 (superficial) disease, and involvement past 500 μm is categorized as sm2-3 (deep) disease. [5, 7]  The sm1 layer is further divided into sublayers a, b, and c on the basis of the lateral spread within the layer. (See the image below.)

Schematic of lesion penetration as determined by h Schematic of lesion penetration as determined by high-frequency endoscopic ultrasonography (EUS).

High-frequency (≥20 MHz) endoscopic ultrasonography (EUS) produces an image of the mucosal wall comprising nine separate layers differentiated by their echogenicity. By carefully examining the depth of lesion penetration into the mucosal and submucosal layers, one may determine the risk of lymph node metastases with greater precision. [8]

EUS has excellent sensitivity and specificity in accurately diagnosing the tumor depth and lymph node status of esophageal cancer [9] and is considered the most accurate imaging modality currently available. Its accuracy ranges from 75% to 82% for T1 disease, from 88% to 100% for T4 disease, and from 72% to 80% for lymph node involvement. [10]

A study assessing staging with radiologic and pathologic correlation in patients with esophageal cancer with lymph node metastases who were radiologically staged as N0 reported that whereas EUS, contrast-enhanced (CE) computed tomography (CT), and positron emission tomography (PET)/CT were all more likely to understage nodal disease, PET/CT was more likely to do so than EUS was. [11] The accuracy, sensitivity, and specificity of EUS for N0 vs N+ disease were 55.4%, 42.6% and 75%. Most lymph nodes (82%) were smaller than 6 mm, making direct visualization challenging with current imaging techniques (probably the main reason for the discrepancy between radiologic and pathologic staging). 

False-positive EUS images are attributed to peritumoral inflammation, whereas false-negative staging is often due to the microscopic spread of tumor not detectable by EUS. Micrometastases have been found in lymph nodes of early esophageal tumors (as many as 44% in one study [11] ​). Because of the limitations of the technology, EUS is subject to significant rates of false positive and negative disease; however, EUS followed by EMR and histopathologic analysis remains the standard of care for early esophageal cancer staging.



Endoscopic resection may be considered for definitive treatment of superficial premalignant and well-to-moderately differentiated malignant lesions of the GI tract in the absence of lymph node or distant metastases (T1mN0M0). They also play an integral role in the staging algorithm of early GI cancers by providing a larger resection specimen than standard forceps biopsy, allowing accurate T staging and establishing the presence of lymphovascular involvement. [12]  Furthermore, pathologic examination of resected specimens frequently leads to significant alterations in patient management. [13]

Esophageal lesions

Squamous cell carcinoma of esophagus

EMR is reserved for lesions smaller than 2 cm that are limited to esophageal mucosa (corresponding to stage T1a) and involve less than one third of the esophageal circumference. Larger lesions may be treated with ESD. Disease-specific survival rates are as high as 95%, and complication rates are low. [14, 15, 16]  

In a retrospective cohort of 300 patients with esophageal squamous cell carcinoma (SCC) who underwent ESD or EMR, local recurrence was 1% with ESD and 10% with EMR, and pathologic free margin was 78% with EMR and 97.4 with ESD; however, at 4-year follow-up, despite the significantly better disease-free survival with ESD, there was no difference in overall survival. [17] EMR and ESD may be performed for staging of lesions lacking architectural or EUS features suggestive of deep submucosal invasion.

Adenocarcinoma of esophagus

ESD can be used for early adenocarcinoma of the esophagogastric junction (EGJ). Several case series have retrospectively evaluated the role of ESD for non-Barrett adenocarcinoma of the EGJ. [18, 19, 20] The rate of en-bloc resection was 100%, with curative resection rates of 68-79%. Patients with noncurative resections were typically managed surgically. In studies involving mean follow-up durations of 15-30 months, patients with curative ESD resections had no local recurrences or metastatic cancer; in studies involving longer follow-up periods, a 5-year disease-specific survival rate of 100% was reproted for patients with curative resections.

Barrett esophagus with dysplasia and early adenocarcinoma

EMR should be used in segments of Barrett esophagus (BE) with mucosal irregularities, including nodularity, ulceration, or flat but irregular mucosal contour. This allows more accurate staging of the lesion, as well as potential therapeutic benefit if a lesion is completely resected. Accurate pathologic staging, the potential for cure, and a favorable safety profile are all reported advantages of EMR in this setting. [21, 22, 23, 24]

As with squamous neoplasms, EMR is considered definitive therapy for cancer if the lesion is well-to-moderately differentiated, is limited to the mucosal layer, is 2 cm or smaller, and has no lymphovascular invasion. If an EMR specimen is a definitively resected Barrett cancer or shows only HGD, endoscopic ablative therapy of the remaining BE should be performed. If the EMR specimen demonstrates neoplasia at the deep margin, residual neoplasia should be assumed, and surgical or systemic therapy (chemotherapy, radiation, or a combination of the two) should be considered. [25]

Investigations have demonstrated the efficacy of EMR applied to the treatment of larger lesions, as well as the complete eradication of BE with favorable results (albeit with potential increases in complications). [26]

Gastric lesions

Early gastric cancer

As an established alternative to surgery, guidelines for the use of EMR in the treatment of EGC have been published by the JSGE. [27]

Well-to-moderately differentiated tubular and papillary adenocarcinomas are classified as differentiated cancers, whereas signet-ring cell carcinomas and poorly differentiated adenocarcinomas are classified as undifferentiated cancers. Indications for endoscopic resection are classified as absolute, expanded, or out of indication and are based on the following:

  • Depth of invasion
  • Finding of ulceration (UL)
  • Lesion size
  • Classification as differentiated vs undifferentiated 

Absolute indications for endoscopic resection include the following:

  • Macroscopically intramucosal (cT1a) differentiated carcinomas measuring less than 2 cm in diameter
  • Macroscopic type does not matter but no ulceration scar (UL[–])

Expanded indications include the following:

  • UL(–) cT1a differentiated carcinomas greater than 2 cm in diameter
  • UL(+) cT1a differentiated carcinomas less than 3 cm in diameter
  • UL(–) cT1a undifferentiated carcinomas less than 2 cm in diameter

Factors determining whether EMR or ESD is indicated include the following:

  • Histopathologic type
  • Lesion size
  • Depth of invasion
  • Presence or absence of ulceration

Given the risk of incomplete resection with EMR for lesions with expanded indications, ESD is recommended by the JSGE as the preferred modality. There have been no randomized controlled trials comparing clinical outcomes between gastric EMR and ESD.

In a Japanese retrospective gastric cancer registry of 12,647 patients comparing short- and long-term outcomes of EMR and ESD, 81% of patients underwent ESD and 19% underwent EMR. [28] En-bloc and R0 resections were achieved in 79% and 89% of patients undergoing EMR and ESD, respectively. The total proportion of patients who underwent curative resection was 69%; 44% underwent curative resection for absolute-indication lesions, and 25% underwent curative resection for expanded-indication lesions.

In this study, the 5-year overall survival rate was 92% in patients with absolute indications and 90% in those with expanded indications after curative resection and 87% in those undergoing noncurative resection. [28] The 5-year disease-specific survival rates were 99.9% for absolute indications and curative resection, 99.7% for expanded indications with curative resection, and 98.7% for noncurative resections

Criteria for ESD for EGC have expanded, as reflected in Japanese guidelines. Data comparing outcomes after ESD versus surgery in these patients are limited. (See Outcomes.)

Duodenal lesions

EMR and ESD have been used in the treatment and staging of ampullary and periampullary adenomas, early carcinomas, and ampullary submucosal lesions. Additional applications include nonampullary adenomas, neuroendocrine tumors, and submucosal lesions. Available evidence suggests that duodenal EMR and ESD may be associated with a substantial rate of complications, particularly in the treatment of submucosal lesions.

Colorectal lesions

EMR and ESD are commonly used for the resection of laterally spreading benign or malignant lesions of the colon, including early-stage colon and rectal cancers, flat adenomas, large superficial colorectal tumors, and rectal carcinoids. Additional indications also include those patients who refuse surgical intervention or in whom significant comorbidities are prohibitive. Appropriate indications for EMR include the following [29] :

  • Well-differentiated or moderately differentiated tumors confined to the mucosa
  • Type 0-IIa lesions smaller than 2 cm
  • Type 0-IIb lesions smaller than 1 cm
  • Type 0-IIc lesions smaller than 1 cm

For large lesions whose size exceeds half of the circumference of the colorectal lumen, piecemeal EMR should be avoided, and ESD or a surgical resection should be performed. [30] ESD provides a higher curative resection rate with a lower recurrence rate for colorectal tumors larger than 2 cm.

In 2015, the Japan Gastroenterological Endoscopy Society (JGES) published guidelines for colorectal EMR/ESD, including the following recommendations [30] :

  • In nonmalignant lesions, endoscopic resection is recommended for adenomas 6 mm or larger
  • Resection is recommended for superficial depressed-type lesions (type 0-IIc) even when the lesion is 5 mm or smaller
  • Among early colorectal carcinomas (Tis/T1), lesions with little possibility of lymph node metastasis and a higher expectancy of curability with en-bloc resection on the basis of size and location are usually treated endoscopically
  • Obvious clinical T1b carcinomas (submucosal invasion depth ≥1000 μm) are usually treated surgically

When endoscopic treatment is performed for colorectal carcinomas, en-bloc resection is the principal approach; however, piecemeal EMR is permissible for some adenomatous and “carcinoma in adenoma” lesions when appropriately carried out and when the possibility of submucosal invasion can be definitively excluded.



Contraindications include the presence of or a high index of suspicion for lymph node or distant metastases.

Biopsy of the lesion before endoscopic resection (including ESD and EMR) may result in a false-positive result as a consequence of fibrosis at the biopsy site. Because the result of mucosal biopsies is unlikely to alter treatment algorithms, biopsy prior to referral for ESD/EMR should be avoided. If biopsy is undertaken, minimization of the time interval between biopsy and endoscpic resction may help reduce false-positive results. [31]

The nonlifting sign, defined as the failure of a lesion to elevate above the surrounding mucosa after submucosal injection underneath the lesion, has been established as a highly accurate and specific predictor of submucosal invasion by colonic lesions. [32] Regarding mucosal lesions with a nonlifting sign and residual or recurrent lesions, ESD often can resect those lesions for which EMR is generally difficult and for which en-bloc resection is desirable. [30]

The presence of coagulopathy is a relative contraindication, in view of the risk of bleeding. Any contraindication that applies to standard endoscopy (eg, severe cardiopulmonary comorbidities) is also applicable to ESD and EMR.

The key curative difference between surgical and endoscopic resection of cancer is the lack of lymph node dissection with endoscopic resection. Therefore, endoscopic resection is not considered in lesions with more than a negligible risk of lymph node metastasis. Tumor depth of invasion correlates with risk of lymph node metastasis (see Table 1 below). [33, 34]

Table 1. Lymph Node Metastasis Risk for T1 Tumor at Different Sites in Gastrointestinal Tract [33] (Open Table in a new window)

Depth Esophageal Squamous Cell Carcinoma Esophageal Adenocarcinoma Gastric Adenocarcinoma Colonic Adenocarcinoma
M1 0% 0% 0-0.4% 0%
M2 0.03%
M3 9% 0.02%
Sm1 15-20% 200 μm 1.9% (if lesion is < 2 cm) up to 500 μm < 3% 1000 μm < 1%

10% without risky signs

33% with risky signs

~20% 6%
Sm3 14%

Technical Considerations

Procedural planning

Given the increasing risk of lymph node metastases associated with deeper tumor invasion into the wall, determination of the depth of lesion penetration is an integral step before endoscopic resection is considered.

The following two methods can be used to predict the depth of penetration:

  • Analysis of the endoscopic mucosal features of the target lesion (see the first image below)
  • Pre–endoscopic resection routine EUS and high-frequency EUS using a miniprobe (see the second image below)
Endoscopic view of superficial gastric neoplasm. Endoscopic view of superficial gastric neoplasm.
Probe-based high-frequency ultrasound image of sup Probe-based high-frequency ultrasound image of superficial gastric neoplasm.

Certain lesion characteristics detectable on endoscopy (see Background) have been associated with risk of submucosal invasion, including the following [35] :

  • Depression of the lesion [5, 36] (type IIc in the Paris classification)
  • Size greater than 15 mm
  • Failure to lift with submucosal injection of saline (see Contraindications)
  • Smooth nonindented circumference for elevated (0-IIa) or depressed (0-IIc) lesions

For those lesions with submucosal invasion, the risk of lymph node metastases is further modified by the presence or absence of lymphovascular invasion on histologic examination. [37]  In one study of HGD and early cancer in BE, lymph node metastases were detected in 0%, 23%, and 69% of lesions within the sm1, sm2, and sm3 layers, respectively. [38]

Several retrospective studies of patients with T1b esophageal adenocarcinoma (EAC) undergoing esophagectomy have found that 16-18% of patients have lymph node metastasis and 27-34% have lymphovascular invasion. [39] ​ Studies comparing lymph node metastases between sm1 and sm2-3 disease, all retrospective and including small numbers of patients, have shown conflicting results. [40, 41, 42]

Similarly, in a study of more than 3000 patients who had undergone gastrectomy with lymph node dissection for EGC, no lymph node metastases were seen with well-differentiated to moderately differentiated lesions smaller than 3 cm that were confined to the mucosal or sm1 layers without lymphovascular invasion. [43]

Standard EUS, with a probe frequency of 7.5-12 MHz, is the most commonly used method of determining the presence of lymph node metastases immediately before EMR. Compared with high-frequency EUS, standard EUS provides greater tissue penetration, displaying a five-layer mucosal wall while allowing visualization of locoregional lymph nodes.

In evaluating for lymph node metastases in BE-HGD and early-stage Barrett-associated EAC, EUS demonstrated an accuracy of 93% in an early series. [38]  In a second study, EUS identified suspicious lymph node metastases in 28% of patients referred for EMR of BE-HGD and BE-EAC, and fine-needle aspiration (FNA) confirmed metastases in 71% of these cases. [44]

A subsequent study examined the accuracy of staging with radiologic and pathologic correlation in patients who had esophageal cancer with lymph node metastases and who were radiologically staged as N0. [11]  EUS correctly identified N0 vs N+ disease in 55.4% of cases. Its sensitivity for identifying N0 vs N+ disease was 42.6%, and its specificity was 75%.

The impact of the information provided by EUS beyond that of a careful endoscopic examination has been questioned. [45]  As a result, the use of EUS before endoscopic resection has been somewhat controversial. [46, 47]

Endoscopic submucosal dissection

The adoption of ESD in western practice has been limited by its technical difficulty, steep learning curve, and limited learning opportunities. ESD, however, provides several advantages over EMR for appropriately selected patients and is also a valuable technique in removing lesions with submucosal fibrosis from a prior resection. Because it is difficult to perform en-bloc resection with EMR for lesions larger than 20 mm, ESD is also more suitable for these lesions. The ability to predict depth of invasion in an attempt to decide whether to pursue EMR, ESD, or surgical resection remains difficult to achieve.

ESD is a technically demanding modality for endoscopic resection, carrying an increased risk of adverse events. In the case of colonic ESD, challenges include the following [48] :

  • Difficulties achieving scope stability as a consequence of looping
  • Peristalsis, flexures, and folds making manipulation of the knife challenging
  • Thin colonic wall allowing little margin of error during dissection
  • Risk of peritonitis from a perforation

Hybrid techniques using ESD to perform a circumferential mucosal incision with partial submucosal dissection, followed by EMR, can also be used where either complete ESD is too technically challenging or sufficient expertise to achieve en-bloc resection is lacking.

Resection of a lesion with ESD can result in the creation of a large mucosal defect that can cause delayed adverse events (eg, bleeding or perforation) after the procedure. There appears to be a learning curve with ESD, and complications decrease as the endoscopist's experience increases. When small perforations occur, endoscopic clipping or suturing can be used to close the site. In more severe cases or situations where endoscopic closure cannot be accomplished, surgical intervention is often needed.

A key technical consideration for ESD is procedure time. Because ESD is a technically challenging procedure, it can have a significantly longer procedure time than EMR does. In addition, patients may require admission for observation after ESD. Finally, the availability of expert endoscopists with experience in performing ESD is limited in Western practice, as is access to appropriate training opportunities.

Complication prevention

EMR and ESD should be performed by experienced operators in a center of expertise. Appropriate adjuncts to treatment should be readily available in the procedure room (see Periprocedural Care). Additionally, interventional radiology and surgical backup should be available in the event of uncontrolled hemorrhage or perforation.





In large retrospective series that evaluated ESD for non-Barrett EGJ adenocarcinoma, the rate of en-bloc resection was 100%, with curative resection rates of 68-79%. [18, 19, 20]  Patients with curative ESD resections had no local recurrences or metastatic cancer detected in mean follow-up durations of 15 months to 5 years.

A retrospective cohort study of 300 patients with early SCC who underwent ESD or EMR excluded all patients found to have histologic evidence of submucosal invasion. [17] This study reported en-bloc resection rates of 100% for ESD (mean lesion size, 30 mm) and 53% for EMR (mean lesion size, 20 mm). Local recurrence was lower in the ESD group (1%) than in the EMR group (10%). There was no difference in survival at a follow-up of more than 4 years.

Another retrospective series of 70 patients with SCCs 2 cm or larger showed that there was no recurrence of lesions treated by en-bloc resection and that lesions treated with EMR had a significantly higher recurrence rate than lesions treated with ESD. [49]



Large retrospective series evaluating ESD for early gastric adenocarcinoma reported en-bloc resection rates of 86-97%, R0 (negative lateral and vertical margins) rates of 88-93%, and local recurrence rates of approximately 1%. The studies showed a high 5-year overall survival (96-100%) with a remarkable 5-year disease-specific survival (99-100%). [50, 51, 52, 53, 54]


Two meta-analyses of ESD versus EMR for treatment of EGC reported en-bloc resection rates of 92% for ESD and 52% for EMR; R0 resection rates of 82-92% for ESD and 42-43% for EMR; and local recurrence rates of 0.8% for ESD and 5.0-6.4% for EMR. [55, 56]

ESD vs surgery

A single-center retrospective study (N = 308) from Japan compared the outcomes of ESD (n = 181) and surgery (n = 127) in patients with differentiated-type EGC meeting expanded-indication criteria. [57] Outcomes with a propensity-matched subset of these groups comprising 74 matched pairs were analyzed as well. Before matching, 5-year overall survival (OS) was not significantly different between groups (ESD, 99%; surgery, 92%; median follow-up, 43 mo). Adverse events were significantly more common with surgery (24% vs. 9%). In the propensity-matched cohort, 5-year OS was significantly higher after ESD than after surgery (97% vs. 86%; median follow-up, 43.5 mo).




Large retrospective series comparing EMR with ESD for colorectal tumors larger than 2 cm showed that in the ESD group, the lesions were larger (29-37 mm vs 22-28 mm), en-bloc resection rates were higher (84-95% vs 33-57%), and local recurrence rates were lower (0-2% vs 12-26%). [58, 59, 60, 61]

ESD vs transanal endoscopic microsurgery

One retrospective study compared ESD and transanal endoscopic microsurgery (TEM) for Tis or T1 rectal cancers. [62] Patients treated with ESD (n = 30) and those treated with TEM (n = 33) had similar rates of en-bloc resections (96.7% vs 100%) and R0 resections (96.7% vs 97.0%). Neither group had local recurrences or distant metastases. ESD was associated with a shorter total procedure time and a shorter hospital stay than TEM was.

A meta-analysis that included data from from 11 ESD and 10 TEM series (N = 2077) evaluated outcomes in the treatment of rectal neoplasms larger than 2 cm. [63] TEM was associated with higher rates of en-bloc resection (99% vs 88%) and R0 resection (89% vs 75%) than ESD was, but a higher percentage of lesions in the ESD group were cancers as compared with those in the TEM group, which were mostly adenomas. Also, there was a trend toward fewer local recurrences in the ESD group than in the TEM group (2.6% vs 5.2%). [63, 33]

ESD vs laparoscopy-assisted colorectal surgery including lymphadenectomy

A large retrospective series from the National Cancer Center in Tokyo evaluated 589 patients with colorectal cancer (297 patients with colorectal intramucosal or slightly submucosal invasive cancers undergoing ESD; 292 patients with T1 colorectal cancers undergoing laparoscopy-assisted colorectal surgery [LAC]). [64] In this study, ESD was associated with a lower complication rate than LAC, with favorable en-bloc and curative resection rates.