Current endoscopic ultrasound-guided approach to incidental subepithelial lesions: optimal or optional?

Subepithelial lesions (SEL) are identified during endoscopic procedures on a regular basis. They can occur anywhere in the gastrointestinal (GI) tract and are located beneath the normal epithelial layer, which explains why a tissue diagnosis is difficult to obtain with routine biopsies. Endoscopic ultrasound (EUS) is used to further characterize these lesions. EUS can distinguish intramural lesion from extramural compression. Furthermore, it allows allocation of intramural lesions to a specific layer of the GI wall and offers additional information as to whether a lesion could be benign or malignant. EUS also assists in choosing the optimal means of tissue acquisition. The choice of tissue acquisition is based on a number of factors, such as tumor size, EUS features, and location within the GI tract or within a specific layer of the GI wall. Furthermore, local expertise and patient factors should be considered when deciding whether tissue acquisition, surgical intervention or follow up is recommended. In this review we offer an EUS-guided approach to the evaluation of incidental SEL based on current evidence and point out areas of uncertainty, which explain why the proposed algorithmic approach may be optional rather than optimal.

Introduction

Subepithelial lesions (SEL) are regularly encountered as incidental findings during endoscopic procedures with an estimated incidence of approximately 1 in 300 patients [1]. Only a minority of SEL presents with clinical symptoms and complications such as vomiting, anemia, dysphagia, and gastrointestinal (GI) bleeding [2]. As the name suggests, these lesions are located beneath the normal epithelial layer and can originate from anywhere within the GI wall or tract, with the stomach being the organ most frequently involved [2]. Many SEL are benign lesions, such as lipomas, pancreatic rests, leiomyomas, schwannomas or duplication cysts. However, up to 13% of upper GI tract lesions are malignant (e.g. metastases or lymphomas) and an additional 8% at least have malignant potential, such as GI stromal tumors (GIST) [3]. It is therefore important to further characterize and manage these lesions accordingly. A stepwise approach, using endoscopy, endoscopic ultrasound (EUS) and various forms of tissue acquisition has been recommended by most experts [2,4-8]. In this review, we describe an EUS-guided approach of incidental SEL based on the current evidence. As new techniques for endoscopic tissue acquisition evolve and information on varying success rates with endoscopic tissue acquisition is gained, such an approach must be viewed against the background of local expertise and patient factors. A 3-step algorithmic approach to incidental SEL (Fig. 1) is offered to help guide the reader through this review and areas of uncertainty are discussed.

Figure 1

Stepwise algorithm for the work-up of an incidental upper gastrointestinal (GI) subepithelial lesion (SEL)

The diagnostic algorithm is explained in the manuscript. It applies mainly for upper GI-SEL. Exceptions, such as rectal lesions, small bowel lesions, suspected leiomyoma in the esophagus and suspected neuroendocrine tumors must be considered. Broken lines depict alternative options

Patient history, localization, size, endoscopic appearance/specific morphology, color, forceps palpation: consistency, gliding sign and pillow sign (suspected lipoma).

Step 1: Endoscopic assessment

Endoscopy is the first diagnostic modality used to assess incidental SEL. Although endoscopy usually cannot distinguish between different types of SEL, it can offer important clues to the correct diagnosis and further management. As the first step, endoscopy is used to describe the appearance of a lesion and its location. The location of a lesion within the GI tract already increases the pretest probability of certain lesions. For example, leiomyomas are the most common hypoechoic SEL in the 4th EUS layer in the esophagus, whereas in the stomach such lesions are likely to represent GIST. Brunner’s gland hyperplasia is commonly found in the duodenal bulb and neuroendocrine tumors represent up to 40% of SEL found in the rectum [9]. With the few exceptions mentioned below, all solid appearing SEL should be biopsied to rule out an epithelial origin. Mucosal biopsies are usually insufficient to obtain an adequate tissue specimen, unless an ulcer of a lesion is targeted. Ulcerations occur in inflammatory or rapidly growing lesions and are a typical feature of large gastric GIST (Fig. 2A). However, ulcerations can also occur in other SEL, such as benign inflammatory fibroid tumors (Fig. 2B) [10]. Caution is necessary when a biopsy is taken from an ulceration. Especially large ulcerated GIST can bleed severely with or without biopsy of the ulcer, because these lesions are often highly vascularized [11]. The closed biopsy forceps can be used to probe the lesion to assess its consistency (firm or soft). If the lesion has a yellow hue and can be easily indented with a closed biopsy forceps (positive pillow sign) the diagnosis of a lipoma can be made with an accuracy of >95% and no further evaluation is usually necessary (Fig. 1, 2C) [5]. In uncertain cases deep “bite on bite” biopsies or jumbo forceps biopsies are generally sufficient to unroof the lipoma and reveal the underlying tissue [4]. Biopsies should not be obtained if cystic or vascular lesions are suspected. Cystic lesions or lymphangiomas often appear slightly translucent and are easily compressible (Fig. 2D). Varices usually have a bluish tinge, are often serpiginous and also easily compressible. However, in rare cases isolated varices can be mistaken for SEL [12]. If there is doubt, EUS should be performed prior to taking a biopsy to rule out a vascular lesion. Biopsies should also be avoided in lesions which are likely to undergo thoracoscopic surgical enucleation, such as probable esophageal leiomyomas or GIST. This recommendation is made by authors who have reported that scarring after biopsy might hinder surgical enucleation [13,14]. Pancreatic rests often have a classic umbilicated endoscopic appearance and are commonly located within the gastric antrum (Fig. 2E). A typical endoscopic feature of granular cell tumors is their yellowish appearance, but in comparison to lipomas they are rather firm and most commonly located in the distal esophagus (Fig. 2F). In addition to inspection and biopsy, the size of the SEL should be estimated. The open biopsy forceps (usually 7-8 mm) can be used as a reference standard. Lesions ≤1 cm in size are rarely symptomatic or malignant and therefore endoscopic follow up in 1 year is recommended. If the lesion grows or reaches a size >1cm, further characterization is required. Unless a definite diagnosis can be made by endoscopic criteria (e.g. lipomas or definite varices), all lesions >1 cm should be evaluated by EUS as the next step. An exception to this rule is potential neuroendocrine tumors in the duodenum and rectum, which have a high prevalence in these locations and show potentially malignant behavior. Therefore, further work-up is recommended, even for subcentimeter SEL in these locations.

Figure 2

Endoscopic features of selected subepithelial lesions. (A) Gastric gastrointestinal stromal tumor with a central ulceration. (B) Inflammatory fibroid polyp with ulceration. (C) Lipoma in the transverse colon with a positive pillow sign after indentation with the closed biopsy forceps (arrow). (D) Lymphangioma with translucent appearance, the septated lesion appears lobulated on macroscopic inspection. (E) Pancreatic rest in the antrum with umbilication. (F) Granular cell tumor. A yellowish hue becomes particularly visible after superficial biopsy (small picture)

Mural versus extramural location, layer, size, contour, echogenicity (as compared with muscularis propria), echo-pattern (homogeneous, heterogeneous, specific features like halo or anechoic areas), vascularity (color Doppler, contrast enhancement).

Step 2: EUS criteria and classification

EUS is the second step in the evaluation of SEL and adds valuable information to guide further management. EUS can distinguish intramural lesions from extramural compression with nearly 90% accuracy [15]. Extramural compressions are most commonly caused by normal organs, such as the spleen or splenic artery, pancreas, gallbladder and the left lobe of the liver, but can also represent pathologic structures, such as (pseudo-) cysts, enlarged gallbladder, splenic artery aneurysms or tumors [15,16]. Further work-up and treatment of pathologic extramural lesions should be performed as clinically indicated, but is beyond the scope of this review.

Intramural SEL should be further characterized by EUS. The EUS layer, location within the GI tract and echo features can provide valuable information to establish the possible diagnosis of these lesions (Table 1, Fig. 3). In an algorithmic approach, hyperechoic and anechoic lesions are initially differentiated from hypoechoic, isoechoic or mixed echogenic lesions. Hyperechoic lesions are generally benign and most often represent lipomas. If there are no unusual (mixed) features within a hyperchoic lesion, no further work-up is needed. Anechoic lesions are fluid-filled structures which can represent vascular lesions (e.g. varices or hemangioma) or “cystic” lesions (e.g. duplication cysts, lymphangioma). Vascular lesions and “cystic” lesions are easily distinguished by a positive or negative Doppler signal, respectively. Mixed lesions with partially solid appearing components require further work-up, as these lesions can represent solid (mesenchymal) lesions with cystic degeneration, complicated cystic lesions or intramural abscesses. In the case of attempted fine needle aspiration (FNA) of such lesions, prophylactic antibiotics should be given and the cyst or cystic part should be aspirated completely to avoid septic complications [17-19].

Table 1

Typical endosonographic features of benign and malignant subepithelial lesions

Figure 3

Endosonographic features of selected subepithelial lesions. (A) Hyperechoic gastric lipoma, arising from the 3rd endoscopic ultrasound layer (by courtesy of Prof. Dr. Uwe Will, Gera). (B) Small gastric leiomyoma with calcifications (atypical), arising from 4th hypoechoic layer. (C) Small gastric gastrointestinal stromal tumor (GIST), arising from 4th hypoechoic layer (arrowhead). (D) Large gastric GIST with distinct heterogeneity and small anechoic cavities (arrow heads), highly vascularized (contrast-enhanced power Doppler), arising from 4th hypoechoic layer. (E) Rare case of esophageal leiomyomatosis. The hypoechoic tumor nodules arise from the 4th hypoechoic layer. (F) Large esophageal leiomyoma arising from 4th hypoechoic layer

For hypoechoic, isoechoic or mixed SEL, a specific diagnosis is required, because of their possible malignant potential [3,20-23]. The accuracy of EUS to distinguish benign from (pre-)malignant lesions is suboptimal [24,25].GIST present a particular challenge in this regard. GIST are mesenchymal tumors that originate from the interstitial cells of Cajal and are most commonly found in between the proper muscle layers (4th EUS layer) of the GI wall (Table 1). Although most small GIST (≤2 cm) behave in a benign fashion and are frequently found as incidental findings on autopsy series and surgical specimen, all GIST are thought to have at least malignant potential [26,27].Efforts have been made to identify EUS features that distinguish benign mesenchymal SEL (e.g. leiomyoma and Schwannoma) from GIST and benign from malignant GIST [28-31].

One retrospective study demonstrated that 4 EUS features (echogenicity compared with the muscularis propria, homogenicity, hyperechoic reflexes or a hypoechoic halo) may differentiate gastric GIST from gastric leiomyomas with a sensitivity and specificity of 89% and 86%, respectively [20]. Moreover, digital image analysis was able to differentiate GIST from leiomyoma with 91% accuracy [21].

Chak et al identified 4 criteria, namely size >4 cm, irregular extraluminal border, echogenic foci (>3 mm) and cystic spaces as independent predictors for malignancy of gastric SEL [28]. In another study, Palazzo et al confirmed a high positive predictive value for malignancy when extraluminal irregular margins, cystic spaces and also lymph nodes were present [29]. The latter is surprising because lymphatic spread is rare in malignant GIST. Despite the poor sensitivity of this finding, advanced disease can be expected, when it occurs [32,33]. Sepe and Brugge pointed out that large size of a SEL and irregular margins are the most consistent findings to suggest malignancy [34]. The sensitivity for the diagnosis of a malignant GIST increases if 2 or more of the above features occur simultaneously [28,29]. Although EUS criteria can give important clues to the etiology and risk of malignancy of a lesion, their accuracy is insufficient and therefore a tissue diagnosis is usually required. Although new EUS methods for tissue characterization, such as digital image analysis, real-time elastography and contrast-enhanced EUS, may be helpful in the differential diagnosis of SEL in the future, these are not validated yet [21,35,36]. Preliminary data suggest that contrast- enhanced EUS may visualize the strong neovascularization of intermediate and high-risk GIST and, therefore, may be used to predict the malignant potential of GI SEL [37,38].

The strength of EUS prior to tissue acquisition is its capability to guide the physician to the choice of an optimal tool of tissue acquisition (Fig. 1, 4). For example, small lesion within the 2nd or 3rd EUS layer can be completely resected by endoscopic (sub-) mucosal resection, whereas large lesions with EUS criteria for malignancy are best referred to surgery. Which tool for tissue acquisition is used depends on EUS criteria, location of the lesion within the GI tract, local expertise and patient preference.

Means of tissue acquisition depends on: Preliminary diagnosis, layer, specific localization, size, ulceration, vascularity, local expertise and patient factors.

Tools of tissue acquisition: Forceps biopsy (bite-on-bite), unroofing techniques, endoscopic resection techniques, EUS-FNA / fine needle biopsy (FNB), surgery.

Step 3: Tissue acquisition

Tissue acquisition is required to establish the correct diagnosis of hypo-, iso- or mixed-echoic SEL, to predict their malignant potential and to guide further management. Ideally, a histological specimen of sufficient size to perform immunohistochemistry (IHC) is obtained. IHC is an essential tool for further characterization of a lesion. This is especially important to distinguish between completely benign leiomyomas (positive stain for desmin and actin), Schwannoma (positive stain for S100) and potentially malignant GIST (positive stain for cluster of differentiation (CD) 117, CD 34, Platelet derived growth factor receptor/PDGFR-a or Diagnosed on GIST/DOG-1), when a hypoechoic lesion is identified in the 4th EUS layer [39]. Because SEL are located deep under the epithelial layer, tissue acquisition remains challenging. The options for sampling SEL include FNA, FNB techniques, such as 19G Trucut biopsy (TCB), endoscopic submucosal resection (ESMR) or dissection (ESD), jumbo forceps or bite-on-bite biopsies, unroofing techniques, endoscopic ligation, resection and tunneling techniques, and surgery. The diagnostic yield, possible challenges and risk of complications of each method are discussed below.

FNA and FNB techniques

A number of needles are available for EUS-guided tissue acquisition, such as FNA (e.g. 25G, 22G or 19G caliber needles), FNB with new reverse bevel 25G, 22G, and 19G needle systems and TCB (19G) [40]. Tissue acquisition with such methods, primarily FNA, has become the method of choice in the work-up of SEL in the majority of cases, especially if a lesion arises from the 4th EUS layer (Fig. 1). FNA/FNB or TCB is not recommended in large lesions (>3-5 cm) with malignant features, because these should be referred to surgery, unless FNA alters management decisions (e.g. neoadjuvant imatinib therapy for advanced GIST) [41]. Enough tissue should be obtained to perform IHC, which can best distinguish between different types of SEL, especially those originating from the 4th EUS layer. Ideally, a histological core specimen is obtained for IHC (Fig. 4 C-F). All needle systems are generally able to provide cytology and/or histology, although large bore (19G) needles are specifically designed to deliver larger core specimen [42]. FNA is performed by advancing the needle into the SEL under EUS guidance. Most investigators use a stylet for the initial puncture, but whether this improves sampling is still a matter of debate [43,44]. A recent randomized trial has shown no difference in the diagnostic yield, whether a stylet was used or not [45]. After removing the stylet, the FNA needle is advanced back and forth, preferably in a fanning technique [44]. Given the strong cohesiveness of mesenchymal tissue, suction should be applied to the syringe in the case of suspected mesenchymal tumors. The number of needle passes depends on the adequacy of the obtained tissue. Rapid onsite evaluation of the specimen by a cytologist has been shown to improve the yield of FNA by up to 10-30% [43].

Figure 4

Endoscopic ultrasound (EUS)-guided fine-needle aspiration of a gastric subepithelial tumor. (A) Endoscopic image of an incidental medium-sized subepithelial lesion (SEL) arising from the gastric cardia. (B) EUS reveals a 2 cm, hypoechoic, homogeneous, lobulated SEL with distinct margins, arising from the 4th EUS layer (M. propria, arrow heads). Leiomyoma is suspected. (C) EUS fine needle aspiration using a 22G aspiration needle. (D) Smear cytology (May-Grünwald-Giemsa, x200) and (E) Histology of core particles (Hematoxylin-Eosin, x200) show fascicles of spindle-shaped tumor cells, suggestive of a mesenchymal tumor. (F) Immunohistochemistry (Desmin, x 200: strongly positive; CD117: negative, not shown here) confirms benign leiomyoma. Surgery is not necessary (Photomicrographs D-F by courtesy of Dr. Stephan Wagner, Königs Wusterhausen, Germany)

Retrospective studies report yields between 64-100% for small (22 or 25G) needles (Table 2) [46-56]. However, the two largest studies only report a yield of 43-62% when “suspected” diagnoses (based on morphologic criteria without IHC) were excluded [52,54]. Furthermore, the highest yields were observed in the smallest studies with no more than 17 patients [47,51]. One prospective study by Akahoshi et al showed a diagnostic yield of 82% in 53 patients with the use of 22G FNA (Table 3) [57]. Likewise, diagnostic yields of 79-100% were observed in retrospective studies, which used 19G FNA needles. Again the highest yield was observed in a case series of 6 patients [56,58-60]. In prospective investigations by our group and Philipper et al (Table 3), 19G FNA had a yield of a definite IHC-based diagnosis of 52% and 34%, respectively [18,61]. Only one study by Larghi et al, which used a new forward-viewing EUS technique in some cases, reported a high yield of 82% with 19G FNA in a small subgroup of SEL (n=17) [62]. The same group used the same forward-viewing echoendoscope and a 19G aspiration needle in a retrospective study of 121 consecutive patients with GI SEL. Histologic examination of specimens including immunostaining was possible in 93.4% of patients [63].

Table 2

Retrospective studies: Yield of EUS FNA with small and large-bore needles

Table 3

Prospective studies: Yield of FNA or TCB (22G-, 19G FNA and 19G TCB)

19G TCB is another large-bore technique, designed to deliver a histological core specimen. This type of needle has a spring-loaded mechanism in the handle that permits automated specimen procurement. The needle assembly is first advanced into the lesion under EUS guidance. Then the cutting needle is advanced to its limit (2 cm) and the spring is released. This causes the cutting sheath to deploy over the tissue tray, generating a core biopsy. 19G TCB has a yield of 47-67% in “accessible” gastric SEL in retrospective studies (Table 2) [52,60]. In prospective studies the diagnostic yield of 19G TCB was 55-63% for “accessible” lesions [17,64,65]. Technical failures of the 19G TCB occurred in up to 14% in the study by Lee et al and were mostly related to difficulty with advancing the relatively stiff needle because of acute angulation of endoscope [65]. One concern using large bore needles in SEL with necrotic or cystic components is septic complications [17,18]. It is therefore our routine to administer prophylactic antibiotics in such cases. In contrast, risk of relevant hemorrhage or tumor seeding appears almost negligible [18,66,67]. Another limitation of all EUS needles including the 19G needles is the fact that mitotic counts, a factor for predicting the malignant potential of GIST, cannot be reliably assessed [18,67]. In a recent study the authors were able to establish a mitotic count on 50 HPF in 40.6% of biopsies but these counts were lower than those of the subsequent surgical specimen [63]. These limitations and the results of two prospective, comparative studies by Philipper et al and Fernandez-Esparach et al, question the superiority of large-bore needles over 22G FNA for the diagnosis of SEL and the former even suggests that 22G FNA may have a higher diagnostic yield [61,64]. Whether new large bore biopsy needles, such as the ProCore reverse bevel technology (Cook Endoscopy Inc, Limerick, Ireland) will deliver better results remains to be seen. One study has shown high accuracy for diverse GI lesions with this needle, with a yield of 82% in a subgroup of 11 SEL. However, technical difficulties were observed in 18% of all cases [59]. Therefore, the choice of the needle is still largely determined by the operator’s reference and technical aspects.

Jumbo forceps (bite-on-bite) biopsies and unroofing techniques

Large capacity (“jumbo”) forceps biopsies or bite-on-bite biopsies can be used for unroofing of the epithelial layer, allowing to dig into the submucosa (or deeper) in order to obtain tissue biopsies from a SEL. This technique can be particularly useful in those cases, in which SEL are located in the 2nd or 3rd EUS layer. One study performed jumbo forceps biopsy (Radial jaw 4, Boston Scientific Corp., Natick, MA, USA) followed by ESMR in lesions limited to the submucosa (3rd EUS layer) and found diagnostic yields of 17% and 87%, respectively [68]. Interestingly, all 23 lesions were benign lesions without malignant potential. Another study used the same jumbo forceps and aimed at clear visualization of the gastric SEL after unroofing it with the forceps [69]. Twenty-nine percent of the lesions were located in the submucosa and 71% in the muscularis propria. The authors reported a much higher yield of 92% with jumbo forceps biopsy compared to 58% with EUS FNA. A benefit of forceps biopsy was the fact that mitotic indices were reliably assessed in 89% all GIST in the study (29/33 GIST). In a multicenter, retrospective study, Buscaglia et al found mixed results with jumbo forceps biopsies, depending on the layer in which the lesion was located [70]. Jumbo forceps had a yield of 65% for lesion in the submucosa (2nd and 3rd EUS layer) but only of 40% in the muscularis propria (4th EUS layer), whereas the reverse was true for EUS FNA with yields of 38% and 57%, respectively. The large range in the diagnostic yield between 17% and 92% might be related to the type, size and location of the lesions or to differences in technique. For best biopsy results, it seems of particular importance that the SEL is thoroughly unroofed to ensure an adequate tissue diagnosis, especially in those lesions that are located in the 4th EUS layer. In a small retrospective series, Serna-Higuera et al used a 6-12 mm needle-knife incision with subsequent deep introduction of a jumbo forceps in a “button hole” technique to obtain tissue. They reported a histological diagnosis in 93% of the 14 retrospectively analyzed patients. Mitotic counts were determined in 71% of GIST with sufficient material (5/7 GIST). Recently, Tae et al used a similar technique in a prospective study, in which biopsies were taken after unroofing the SEL with a needle-knife. They reported a yield of 90% with a change in treatment plan in 35% of their cases [71]. To gain optimal access to small upper GI SEL (<2 cm), Lee et al used a snare for an endoscopic partial resection unroofing technique in a case series of 16 patients [72]. The diagnostic yield of the endoscopic partial resection unroofing technique was 94%, although oozing of blood occurred in 56% of these patients. Fortunately, successful hemostasis was achieved by using argon plasma coagulation in all cases.

Endoscopic ligation, resection and tunneling techniques

Endoscopic techniques to excise SEL in the submucosa include traditional ESMR with a snare or newer approaches using a transparent cap (ESMR-C) or ligation device (ESMR-L), which are most commonly used for small SEL (<2 cm) [73]. Initially, ESMR with a regular snare or a cap-fitted endoscopic snare device was proposed as a technique to completely excise strictly submucosal or pedunculated lesions, originating from the EUS layers 2 or 3 [74]. As mentioned above, ESMR has a high diagnostic yield for lesions in the submucosa, as it delivers a tissue specimen rather than biopsy fragments and it has the potential to completely excise the lesion [69]. A grasp-and-snare technique has recently been reported to perform full thickness resection of deeper lesions and closure with an over the scope clip (OTSC). The reported rate of complete resection was 86% and the procedure was performed under laparoscopic control [75]. However, this study has been criticized, because grasping the tumor with a tissue anchor can cause rupture, resulting in the inability to ensure the integrity of the tumor. Furthermore, there is potential risk of contamination to the peritoneal cavity and peritoneal infection and finally the OTSC technique cannot treat leaks larger than 2.5 cm [76].

ESMR-C uses a crescent-shaped snare that is placed in the distal ridge of a transparent cap. The lesion is usually lifted with a combination of dilute epinephrine (1:100,000) in a saline solution, mixed with indigo-carmine dye. The lifted lesion is then sucked into the cap, creating a pseudopolyp, captured by closing the snare and removed by electrocautery. This technique can remove small lesions completely in 87% of cases and has a high diagnostic yield of 95%, but is associated with a bleeding risk of 5% [77].

ESMR-L uses a ligation device with a cap to deploy an elastic band or loop around the base of an SEL after it is lifted and aspirated into the cap of the device. The lesion is subsequently decapitated and biopsied or completely removed with a snare using electrocautery. This technique mainly applies to small lesions that can be aspirated into the cap (<15-20 mm) [78,79]. Sun et al reported spontaneous sloughing of the lesion after banding in up to 95% of small gastric and 100% of duodenal GIST within 3-4 weeks [80,81]. Recently, Binmoeller et al used a modification of this technique, in which small SEL (<20 mm) are sucked into a transparent cap and ligated with a detachable loop, the so-called suck-ligate-unroof-biopsy technique. After ligation the SEL is unroofed with a needle-knife and biopsies are taken. When GIST were biopsied, the authors were able obtain enough tissue to analyze the mitotic count/50 HPF. Follow-up endoscopy at 3 and 6 months revealed no residual tumor in the 9 lesions examined. However, there is some concern of delayed bleeding approximately 7 days after the procedure when the lesions slough off and also of possible incomplete microscopic margins or spillage, especially when GIST are removed [80-82].

Lately, a plethora of mostly Asian studies have reported the successful removal of esophagogastric or gastric SEL in the muscularis propria by ESD [82-92]. Successful resection varies from 64-100% in these studies. Most of them are small retrospective case series. The 2 largest studies by Li et al and He et al reported complete resection in 92-94% of their patients with a perforation rate of 4-14%. The most recent studies even report ESD in lesions with a size of up to 5 cm. However, this approach had a high risk of perforation in 19% of cases. The high perforation rate and the risk of incomplete resection or tumor spillage which will turn potentially curable lesions into metastatic disease are a major concern with this technique. A recent analysis by Chun et al suggested that smaller tumor size ≤ 2cm and a positive “rolling sign” (a mobile mass under the mucosa) correlate significantly with complete resection and perforations seem to occur mostly within fixed lesions [93].

Submucosal tunneling techniques are the latest endoscopic innovation, aiming at complete resection of the SEL. A number of small case series have been published using this method [94-96]. The method was adapted from the peroral endoscopic myotomy procedure, which was recently inaugurated for the treatment of achalasia [97]. Using this technique the mucosa is lifted and incised a few centimeters proximal to the lesion to create an entry point into the submucosal layer for the endoscope. Then a submucosal tunnel is created using the ESD technique and the tumor is enucleated using electrocautery knives (triangle-tip or insulated-tip knives). Finally, the enucleated SEL is suctioned into a transparent cap on the endoscope and removed via the tunnel. The proximal entry site is subsequently closed with the use of endoclips. The reported success rates in small case series are 70-100%. Pneumoperitoneum occurred in 13-16% of cases, but patients were treated conservatively. In another retrospective study, a tunneling technique was used for esophageal leiomyomas [98]. Here, hemorrhage was reported in 3 of 18 patients. Again, an additional major concern with this method is incomplete resection, especially in the case of larger GIST, removed in a narrow working-space, such as the submucosal tunnel. These studies do not explicitly mention how large tumors of up to 4 cm can be safely retrieved en bloc after enucleation [97]. Certainly, such treatments are only possible by trained experts in specialized centers and cannot be recommended for broad application. Larger studies and comparisons with surgical procedures are necessary to establish the role of these new endoscopic procedures.

Surgery

An in-depth discussion of the surgical management of SEL is beyond the scope of this review. However, some basic principles to guide the endoscopist in the decision which patients to refer for surgery, are herein mentioned. The main indication for surgery is the removal of symptomatic lesions or (pre-) malignant lesions, most of which are GIST in the upper GI tract or neuroendocrine (carcinoid) tumors in the rectum and duodenum. The difficulty in deciding which patients should be directly referred for surgery lies in the fact that EUS criteria and means of tissue acquisition have the above-mentioned limitations and a definite tissue diagnosis before surgery is not always possible or desirable. For example, surgical enucleation of esophageal leiomyomas could be difficult after previous biopsy or other tissue acquisition, such as unsuccessful tunneling or EMR procedures.

Malignant lesions in the GI tract should always be treated by oncologic standards. The most common pre-malignant SEL are GIST and comprehensive guidelines are available for their management [99-101]. A Korean guideline strictly discourages endoscopic enucleation if a GIST is suspected [39]. Surgical resection is considered the primary approach of GIST management and should be performed in all GIST >2 cm [41]. GIST ≤2 cm in the stomach have a very low risk of malignancy and follow up can be discussed with the patient [59,102]. Conversely, GIST in the rectum have a high risk of malignancy and should be excised, even if they are smaller than 2 cm [39]. The goals of surgery include complete resection, avoidance of tumor rupture, and intra-operative staging to exclude metastatic disease [103]. Because GIST rarely metastasize to lymph nodes, lymphadenectomy is only warranted if suspicious nodes are present. Laparoscopic (wedge-) resection of gastric GIST is feasible if intra-abdominal tumor rupture or seeding is unlikely, but discouraged for large tumors with high risk of rupture [39,101].

Algorithmic approach

Multiple algorithms have been proposed for the stepwise assessment of SEL or suspected GIST by experts in the field [4,5,7-9,34,41,104]. All algorithms follow a stepwise EUS-guided approach, similar to our suggested 3-Step algorithm (Fig. 1). Most experts agree that endoscopic assessment of the lesion (Step 1; Fig. 1) is the first step to evaluate the SEL and that most lesions ≤1 cm can be followed annually, unless they grow [4,5,7-9]. Because small GIST ≤2 cm (in the stomach) are considered to have very low risk of malignant behavior, few authors even consider a size cut-off ≤2 cm appropriate for follow up [105-107]. It must be stressed that this size cut-off does not apply for rectal SEL, because these often represent GIST with high malignant potential or neuroendocrine tumors, which should be resected. It is also broadly accepted that the next step in the evaluation of SEL is EUS (Step 2; Fig. 1). EUS can clearly identify benign lesions, such as most hyperechoic (e.g. lipoma) or completely anechoic lesions (e.g. vessel, hemangioma, cysts or lymphangiomas) and it is used to determine the layer of origin.

The main difficulty with any algorithm for SEL is that all methods of tissue acquisition (Step 3; Fig. 1) have limitations. It is not surprising that, although FNA is considered the method of choice for tissue acquisition by the vast majority of EUS practitioners in the United States and Germany, only a minority will perform EUS FNA regularly when GIST are suspected [108,109]. We suggest an approach to tissue acquisition guided by the size of the lesion, layer of origin and location in the GI tract. Because rectal SEL have a higher risk of representing a malignant lesion, complete endoscopic or surgical resection is usually performed in these lesions. SEL in the small bowel, located distal to the duodenum, are also surgically removed, as they cannot be easily accessed by endoscopic interventional techniques. Longitudinal endoscopes, used for FNA-guided biopsy, have an oblique-viewing optic, which limits the applicability within the large bowel. This might change in the future, as front-viewing, forward-array endoscopes give better access to proximal colonic lesions [110]. Therefore, the algorithm should be considered mainly for SEL of the upper GI tract. Here, duodenal lesions might pose another exception as FNA or TCB have limited applicability in some locations in the duodenum and EMR or surgery might be preferred for lesions that cannot be clearly identified as benign lesions [111]. As shown in the algorithm (Step 3; Fig. 1), we follow an approach that considers symptoms, EUS layer, EUS features of malignancy and size in choosing the tool for tissue acquisition. Tissue acquisition in the 2nd or 3rd EUS layer can be performed by endoscopic resection techniques, unroofing/jumbo biopsy technique or FNA depending on the size and suspected EUS diagnosis (Table 1). SEL in the 4th EUS layer of the esophagus more frequently present leiomyomas and if surgical enucleation is considered, they should not be biopsied or partially resected prior to surgery. In the stomach, GIST are the most common entity in the 4th EUS layer. FNA/TCB is still considered the primary means of tissue acquisition in suspected GIST, unless the tumor is very large, symptomatic, has malignant features or is resected, regardless of FNA results. FNA best applies for 4th layer-lesions between 20-30 mm in size and possibly for those up to 50 mm if the presence of a non-GIST histology (e.g. leiomyoma) is likely, because this would change management. Very large suspected GIST should undergo FNA/TCB if neoadjuvant treatment with tyrosine kinase inhibitors is considered. Any lesion that grows during the follow up must be re-evaluated according to the algorithm. Although all confirmed GIST have potential for malignant transformation, the risk in gastric GIST ≤2 cm is considered to be very low and follow up is an option, especially in patients of advanced age or those with high surgical risk. We discourage endoscopic resection with ESD or tunneling procedures in suspected GIST, as these procedures carry a risk of tumor rupture and might turn a curable lesion into metastatic disease. Special consideration must always be given to the local availability of technique, operator expertise, surgical risk and patient wishes.

Concluding remarks

SEL are commonly identified during endoscopic procedures. These lesions can occur anywhere in the GI tract and EUS-guided tissue acquisition is used for further characterization. EUS cannot only distinguish intramural from extramural lesions, but also helps distinguish benign from potentially malignant lesions, even before tissue acquisition. The choice of tissue acquisition is based on a number of factors, such as tumor size, EUS features, tumor location within the GI tract or within a specific EUS layer. Furthermore, local expertise and patient factors must be considered when deciding whether tissue acquisition, surgical intervention or follow up is recommended. An algorithmic approach of the management of SEL is offered. However, the outlined challenges of all diagnostic methods explain why any algorithmic approach to SEL should be considered optional rather than optimal.