45 Subepithelial Tumors of the Gastrointestinal Tract
Jennifer Maranki and Stavros N. Stavropoulos
Subepithelial tumors (SETs) of the gastrointestinal (GI) tract are frequently encountered lesions and a common cause of referral for endoscopic ultrasound. While the overall prevalence is unknown, these lesions are often discovered during routine endoscopy or cross-sectional imaging performed for other reasons. On endoscopy, these are most frequently encountered in the stomach. These are often incorrectly referred to as “submucosal” tumors. The term “subepithelial” is more accurate as these tumors can be located in and/or arise from any of the layers deep to the epithelial layer of the mucosa (including the muscularis mucosae, submucosa, and muscularis propria).
SETs typically appear as a bulge within the GI lumen with normal overlying mucosa. Most of the tumors cause no symptoms, but can cause bleeding, obstruction, dysphagia, or, if in close proximity to the ampulla, jaundice, and pancreatitis. These lesions can be benign or malignant, and endoscopic ultrasound (EUS) with fine-needle aspiration (FNA) is an important modality in the diagnostic evaluation of these lesions. EUS is also useful for differentiating between mural tumors and extramural (or extrinsic) lesions. Up to 30% of suspected subepithelial mural tumors identified on routine endoscopy are found to be either extramural lesions or extrinsic compression by adjacent organs of vascular structures when evaluated with EUS. 1
A variety of subepithelial lesions present in the GI tract, including gastrointestinal stromal tumors (GISTs), leiomyomas, carcinoids, lipomas, pancreatic rests (ectopic pancreatic tissue), duplication cysts, schwannomas, or metastatic disease. Abnormal vasculature, such as pseudoaneurysms or varices, may also present as subepithelial lesions.
In this chapter, the types of SETs as well as their endoscopic and endosonographic characteristics are discussed. Histologic and immunohistochemical features are noted, as well as high-risk features. Methods of tissue acquisition, including EUS with FNA and core biopsy are compared and contrasted. Endoscopic approaches to tumor resection are reviewed, with advantages and disadvantages of each technique.
45.2 Types of SETs
Understanding of the histologic layers of the GI wall is essential in characterizing SETs. The wall is comprised of five layers identified using EUS imaging with frequencies of 5 to 12 MHz. The innermost (first) layer, characterized by a very thin hyperechoic band, represents the superficial mucosa/interface with the lumen. Below this, a thin hypoechoic band represents deep mucosa including the muscularis mucosae (second layer). The submucosa (third layer) appears as a hyperechoic band, and deep to this, the muscularis propria (fourth layer) appears as a thick hypoechoic band. The serosa/adventitia (fifth layer) is thin and hyperechoic, and often unable to be differentiated from surrounding structures. ▶Table 45.1 summarizes the most common types of SETs and their corresponding wall layers.
45.2.1 Gastrointestinal Stromal Tumors
GISTs are the most frequently encountered SET of the upper GI tract. 2 These tumors are believed to arise from the interstitial cells of Cajal, and 10 to 30% are malignant at the time of diagnosis. 3 Annually, over 5,000 cases of GISTs are diagnosed. 4 Small GISTs (< 10 mm) are common in the adult population, with autopsy data from Germany reporting gross detection in 22.5% of adults over the age of 50. 5 These data suggest that most small GISTs do not progress into large macroscopic tumors despite the presence of the c-kit and PDGFRA mutations. 6
Endoscopically, the lesions are most commonly encountered in the stomach. These vary is shape and size, and may have an overlying ulcer. On EUS, GISTs appear as hypoechoic masses arising from the muscularis propria (fourth layer) (▶Fig. 45.1) or muscularis mucosa (between second and third layer). These are similar in appearance to leiomyomas, and differentiating between the two requires immunohistochemical evaluation of specimens. 7 GISTs are described as c-kit (CD117)-positive mesenchymal spindle cell or epithelioid lesions. Over 95% are positive for c-kit (CD117). 8 , 9 The majority of GISTs are positive for CD34, whereas a minority (20–30%) is positive for spinal muscular atrophy (SMA), and the expression of the two may be reciprocal. Even fewer express the S100 protein, and fewer than 5% are positive for desmin. 10 , 11 , 12 A newer marker for GISTs, DOG1, has shown high sensitivity and specificity for GISTs, and is particularly useful for diagnosing GISTs that are c-kit negative. 6 , 13 , 14 , 15 , 16 , 17 GISTs typically are found in older individuals, with the stomach being the most common site (60–70%), followed by the small intestine (20–25%), colon and rectum (5%), and esophagus (< 5%). 12
Leiomyomas are benign smooth muscle tumors and appear endoscopically and endosonographically similar to GISTs, but are often smaller in size (typically < 2 cm). Occasionally, the tumors may become large. These are frequently encountered in the esophagus but can occur throughout the GI tract (▶Fig. 45.2). These are characterized by normal overlying mucosa that can easily be tented with biopsy forceps. Similar to GISTs, these may also have a denuded or necrotic central area, particularly with large lesions. On EUS, these appear as a hypoechoic lesion arising from either the second layer (muscularis mucosa) or fourth layer (muscularis propria). Because these are unable to be differentiated from GISTs without immunochemical staining, biopsy is often required to exclude a lesion that may require endoscopic or surgical resection. Unlike GISTs, leiomyomas do not stain positive for c-kit or DOG1.
Carcinoid tumors (▶Fig. 45.3, ▶Fig. 45.4) originate from neuroendocrine cells and are most frequently encountered in the small bowel. The ileum is the most common site, followed by the jejunum and duodenum. 18 Carcinoid tumors also occur in the stomach, and account for nearly 10% of all carcinoid tumors. In a retrospective, 50-year analysis of over 500 carcinoids, the male–female ratio has steadily decreased to 0.54. 19 Gastric carcinoids are classified into three types, depending on malignant potential. Type I gastric carcinoids are associated with chronic atrophic gastritis, hypergastrinemia, and pernicious anemia (▶Fig. 45.3). These occur most commonly in elderly female patients and have very low malignant potential. 20 Type II gastric carcinoids are associated with hypergastrinemia due to Zollinger–Ellison syndrome and multiple endocrine neoplasia type 1 (MEN 1). These have intermediate malignant potential. Type III gastric carcinoids are sporadic, not associated with hypergastrinemia, and have the highest risk of malignant transformation. 20
Endoscopically, carcinoids appear as sessile or semipedunculated lesions with normal overlying mucosa. 21 Carcinoid tumors may originate in the mucosal layer and penetrate into the submucosal layer, so they can often be diagnosed by mucosal biopsies. 22 On EUS examination, carcinoids appear as oval-shaped or round, hypoechoic, homogeneous tumors, most often arising from the submucosal layer (▶Fig. 45.3, ▶Fig. 45.4). 23
45.2.4 Other Subepithelial Lesions of the Gastrointestinal Tract
Other subepithelial lesions include lipomas, pancreatic rests, duplication cysts, schwannomas, and aberrant vasculature. Metastatic lesions may also present as subepithelial lesions.
Lipomas are usually asymptomatic and can occur anywhere within the GI tract. Rarely, they can cause bowel obstruction, bleeding, or intussusception. 8 These have a smooth mucosal surface and often have a yellowish hue to the mucosa. These may show a “pillow sign” when probed with closed biopsy forceps. On EUS, these appear as well-demarcated, oval, hyperechoic lesions arising from the submucosal layer (▶Fig. 45.5).
Pancreatic rests, or ectopic pancreatic tissue, are most frequently encountered in the stomach, most notably the antrum, but can occur throughout the GI tract. These do not typically cause symptoms, but can lead to ulceration and hemorrhage, esophageal or intestinal obstruction and intussusception, and biliary obstruction. 24 On standard endoscopy, pancreatic rests may have an umbilicated surface, sometimes described as “volcano like.” 25 Endoscopic ultrasound images reveal well-circumscribed, mixed echogenicity lesions in the submucosa that can involve the muscularis as well that may contain anechoic, serpiginous ductal structures (▶Fig. 45.6).
Duplication cysts are uncommon congenital anomalies of the foregut that are often asymptomatic. 26 These are typically incidentally found on standard endoscopy or cross-sectional imaging. These appear endoscopically as round bulges that may be compressible with normal or somewhat translucent overlying mucosa. On EUS, these are smooth, spherical or tubular in shape, anechoic, with a well-defined wall, and located in the submucosa, muscularis propria, or serosal layers. 27 In the mediastinum, FNA is avoided due to the risk of infection. 28 , 29
Schwannomas are benign nerve sheath tumors that can present with a variety of symptoms, including GI bleeding, abdominal pain, dysphagia, obstruction, and weight loss. 8 Histologically, they consist of spindle cells and epithelioid cells with peripheral lymphoid cuffing, and stain positive for S100 and vimentin. 30 , 31 Endoscopically, schwannomas may have a slightly yellowish hue that can result in misidentification as lipomas. Unlike lipomas, however, endosonographically, these tumors are very low echogenicity lesions arising from the fourth layer (muscularis propria). These typically do not have internal echogenic foci but may have a marginal hypoechoic halo. 32
Granular cell tumors are rare lesions occurring throughout the entire GI tract, with approximately one-third occurring in the esophagus. 33 These generally occur in the fourth, fifth, and sixth decades of life. 34 , 35 Granular cell tumors are usually solitary, but may be multifocal. These are usually incidentally found, but may cause symptoms of dysphagia, nausea, and regurgitation with lesions greater than 1 cm. 33 , 34 On EUS, lesions are typically less than 2 cm in size, with a homogeneous, hypoechoic lesion arising from the muscularis mucosae or submucosa. 36 Most esophageal granular cell tumors have a benign course, but cases of malignancy have been reported. It is generally accepted that tumors causing symptoms and those greater than 1 cm in size should be removed (endoscopically if possible). 33 , 37
Vascular abnormalities, such as varices or pseudoaneurysm, may also appear as subepithelial lesions. 38 Gastric varies often appear as a submucosal bulge, often in the cardia and fundus, or an enlarged gastric fold. Unlike esophageal varices, gastric varices often do not have a bluish hue because these sit deeper in the gastric wall. Gastric varices should be suspected not only in patients with portal hypertension due to advanced liver disease, portal vein thrombosis, and Budd–Chiari syndrome, but also in patients with pancreatic disease causing splenic vein thrombosis. On endoscopy, other clues to the presence of portal hypertension may be seen, such as portal hypertensive gastropathy or esophageal varices. On EUS, varices appear as anechoic tubular structures in the submucosa that exhibit flow on Doppler examination.
While metastatic spread to the wall of the GI tract is rare, a variety of malignancies have been shown to spread to the gastric wall. These include melanoma and carcinomas of the lung, breast, ovaries, and kidney. 2 , 39 , 40 These appear as hypoechoic lesions in any of the layers, can have an umbilicated “volcano-like” appearance, and can be diagnosed by EUS-guided FNA (EUS-FNA) or deep biopsies depending on their depth within the wall of the GI tract.
45.3 Risk Stratification of Subepithelial Tumors
While most SETs are benign, several are malignant or have malignant potential, including GISTs, carcinoids, and lymphoma.
While all GISTs are thought to have some degree of malignant potential, GIST lesions within the small bowel are higher risk than those in the stomach or rectum. 41 Several factors, including those based on overall size and mitotic rate, have been used to stratify the risk of malignancy (▶Table 45.2). 22 Endosonographic features that have been reported to be associated with increased risk of malignancy include lesions with cystic spaces (▶Fig. 45.7), an inhomogeneous echotexture, irregular borders, and the presence of enlarged lymph nodes. 42 , 43 , 44 , 45 Risk stratification is important in formulating an appropriate management strategy for GISTs. 46
Risk of progressive disease (%)
≤ 2 cm
≤ 5 per 50 HPFs
> 2 cm, ≤ 5 cm
≤ 5 per 50 HPFs
Very low (1.9)
> 5 cm, ≤ 10 cm
≤ 5 per 50 HPFs
> 10 cm
≤ 5 per 50 HPFs
≤ 2 cm
≤ 5 per 50 HPFs
> 2 cm, ≤ 5 cm
≤ 5 per 50 HPFs
> 5 cm, ≤ 10 cm
≤ 5 per 50 HPFs
> 10 cm
≤ 5 per 50 HPFs
HPFs, microscopic high-power field in tissue sections.
a Defined as metastasis or tumor-related deaths.
b Denotes tumor categories with very few patients.
Source: Data from Miettinen M, Lasota J. Semin Diagn Pathol 2006;23:70–83.
For carcinoids, risk of metastasis appears to be related to the site of the primary lesion. The rate of distant metastasis is higher in patients with small bowel carcinoids than in patients with neuroendocrine tumors (NETs) in other locations. Jejunoileal lesions are associated with a high rate of transmural invasion and aggressive behavior. 47 Further, lung and rectal carcinoids tend to remain local. 48 , 49
45.4 Methods for Tissue Acquisition
Pathologic diagnosis of SETs may be helpful in clinical decision making. A variety of techniques have been applied to obtain adequate tissue for histologic analysis, and include bite-on-bite mucosal biopsies with forceps, biopsy after mucosal incision, endoscopic submucosal resection, EUS-FNA, and EUS-guided core biopsy (EUS-FNB). Several large-bore needles and core biopsy needles are currently in use for tissue acquisition (▶Fig. 45.8).
45.4.1 Endoscopic Ultrasound-Guided Fine-Needle Aspiration
EUS-FNA is the most widely used method for tissue sampling, and is safe and effective. 2 , 6 Several studies have demonstrated the utility of EUS-FNA for diagnosing subepithelial lesions, and in particular, differentiating GISTs from leiomyomas. 50 , 51 , 52 , 53 , 54 , 55 This is especially true when on-site cytopathologic analysis of sample adequacy is undertaken. 56 In a retrospective study of 112 patients undergoing EUS-FNA of lesions arising from the fourth endosonographic layer, Hoda and colleagues reported that EUS-FNA was diagnostic in 61.6%, suspicious (spindle cells identified) in 22.3%, and nondiagnostic in 16.1%. 57 Mekky et al retrospectively analyzed 141 consecutive patients with gastric SETs who underwent EUS-FNA. 58 Results of the FNA were diagnostic or suggestive in over 82%, with adequate specimens being obtained in 83% of cases. Overall 49% had a definite final diagnosis, of which 60% were GISTs. EUS-FNA results were concordant with the final diagnosis in 66 of 69 lesions (accuracy rate 95.6%). Further, EUS-FNA demonstrated a sensitivity of 92.4% and specificity of 100% in differentiating benign from malignant lesions. Other studies have shown diagnostic yields ranging from 75 to 100% and sensitivities of EUS-FNA of greater than 80%. 59 , 60 , 61 , 62 , 63 Ando and colleagues reported on 49 patients with SETs originating from the muscularis propria, and found that for the diagnosis of malignant GIST, the accuracy, sensitivity, and specificity of EUS-FNA with the addition of immunohistochemical staining was 100%. 51 These data provide support to the use of EUS-FNA with immunohistochemical staining for the diagnosis of GISTs. 6
Based on the data reviewed, the overall rate of definitive histologic diagnosis with EUS/FNA is relatively modest at 60 to 80%, and the amount of tissue obtained, even if adequate for definitive histologic diagnosis, may be too limited for histologic risk stratification. One approach to further optimizing tissue acquisition has been to use a larger-bore needle. Disadvantages of this approach are that the 19-gauge needle can be difficult to advance through the scope and into the lesion, resulting in a technical failure, particularly in the setting of an angulated scope. Additionally, more bleeding may occur with a larger needle. Conversely, use of a smaller-gauge (25-gauge) needle has also been explored, with the thought that it is easy to maneuver and may cause less trauma to the sampled area. While the 22-gauge needle is most commonly used, several studies have assessed the use of both 19-or 25-gauge needles. 64 , 65 , 66 , 67 , 68 , 69
Other sampling techniques have also been studied. EUS-FNA using a forward-viewing echoendoscope is a modification that has been applied in hopes of increasing the diagnostic yield. Larghi et al reported on the use of a forw ard-viewing EUS scope in 121 consecutive patients undergoing EUS-FNA of SETs. 70 Full histologic assessment, including immunohistochemistry, was successful in 93.4%. In terms of neoplastic versus nonneoplastic disease, the sensitivity and specificity were 92.8 and 100%, respectively. Matsuzaki et al conducted a randomized crossover study comparing forward-viewing versus oblique-viewing echoendoscopes in 41 patients with subepithelial GI lesions. 71 There was no difference in diagnostic yield between the two groups, but the forward-viewing group had shorter procedure times (21 minutes versus 27 minutes, p = 0.009), and superior tissue sample areas (2.46 vs. 1 mm2, p = 0.046). ▶Table 45.3 summarizes the yield of EUS-guided sampling.