A submucosal lesion is located from deep mucosa to deeper serosa, so more accurate term is a subepithelial lesion.1 Gastric submucosal lesion is found in about 0.36% of cases by upper gastrointestinal endoscopies.2

When a submucosal lesion is suspected at endoscopy, the first step is to differentiate it from extramural compression, and endoscopic ultrasonography (EUS) is the most accurate tool for this job. The second step is to determine the nature of the submucosal tumor (SMT), for which also EUS is the most accurate tool. EUS shows typical findings for lipoma, duplication cyst and pancreatic nest,1 but in hypoechoic lesions, such as leiomyomas, gastrointestinal stromal tumors (GISTs) and schwannomas, EUS findings are not enough for definite diagnosis, which is why we need tissue acquisition from the submucosal lesion. Because it is located deeper than epithelium, we cannot take specimens from the subepithelial lesion using conventional endoscopic biopsy method. Various methods, such as bite-on-bite technique, endoscopic ultrasound-guided fine needle aspiration (EUS-FNA), EUS-guided Trucut biopsy (TCB), and EUS-guided fine needle biopsy (FNB), have been introduced to obtain samples of sub-epithelial lesion.

For gastrointestinal subepithelial lesions, bite-on-bite technique of two to eight bites using conventional-sized forceps yielded 38% of diagnostic rate (54% in the esophagus, 28% in the stomach and duodenum; p<0.019).3

For SMTs, cytology, cell block processing, and immunohistochemical stains are available using 22-gauge EUS-FNA needle.4 Immunohistochemical stains are helpful for cytopathologic differentiations. Mitotic counts and immunohistochemical stains usually cannot be performed on smears, and so it requires cell blocks.2 Cytological subclassification of submucosal spindle cell lesions by immunohistochemical staining on cell blocks correlate very well with the final diagnosis.5

All GISTs have some degree of malignant potential.6 Even small GISTs may present malignant features on histologic examination or biologic behavior. A lot of efforts have been made to increase the diagnostic yields of submucosal lesions, but each method has their own limitations.

EUS-FNA, first reported by Vilmann et al.7 in 1992, can obtain cell block specimen for histologic examinations and immunohistochemical staining. When processing EUS-FNA specimen, one slide is fixed by air drying, stained with modified Giemsa (Diff-Quick; Baxter Diagnostics, Inc. McGaw Park, IL, USA) and examined on-site for specimen adequacy. The other slides are fixed in the alcohol and stained using the Papanicolaou technique. Cell block material is fixed in 10% formalin, embedded in paraffin, cut thinly and then stained with hematoxylin and eosin. Diagnosis is done with smear cytology, histology, and immunohistochemical staining of the cell blocks.4 Yoshida et al.8 proposed several preparation methods of EUS-FNA specimen for good results. First, maintaining negative pressure of needle many times is necessary to get sufficient tissue because GIST cells have tight cell to cell adhesion. Second, aspirated materials should be pushed out from the needle under the saline solution so that they are not dried. Third, one drop of 1% albumin is helpful for preserving the morphology when cells are collected with cytospin. Fourth, cell degradation materials are also aspirated, especially in the presence of necrosis.

EUS findings are not sufficient to distinguish malignant and benign stromal cell tumors. EUS findings can diagnose malignant subepithelial lesions with a sensitivity of only 64% and a specificity of 80%,3 which is why a further pathologic diagnosis is needed. Echogenicity enables some differentiations between GISTs, leiomyomas, and schwannomas. It was found that GISTs have generally low but slightly higher echogenicity than the proper muscle layer. That of leiomyomas was equal to that of proper muscle layer and schwannoma showed extremely low echogenicity.9 Some EUS findings indicate increased risk of malignancy; tumors with more than 3 or 4 cm of size and irregular borders are the most important features.1 Other findings such as echogenic foci, cystic spaces, ulcerated mucosa, and lymph nodes with malignant pattern are not completely decided yet. EUS findings of high grade malignancy have 1) greater size, 2) lobulated forms, 3) irregular borders, and 4) echogenic foci more often than the low grade malignancy, but EUS is not sufficient to predict malignant pathological findings accurately.10

Inter-observer variation in EUS interpretation can be a factor contributing to the discrepancy of results. Inter-observer agreement was poor in terms of echogenicity, irregular border, presence of cystic spaces, and echogenic foci.1 There are criteria developed by combining EUS features; the criteria for predicting malignancy are tumor size >3 cm in homogeneous pattern, irregular outer margins and lymph node size >10 mm. The reported sensitivity and specificity of predicting malignancy were 64% and 80%, respectively, when meeting at least 2 of these criteria.11

Gastric tumors larger than 30 mm, with irregular borders, mucosal ulceration, and non-oval shape on EUS suggest high risk GISTs. The malignant potential of gastric SMT increases when tumor size is >30 mm; approximately 50% of cases with SMTs larger than 3 cm are malignant.12 Gastric ulcers were concomitant with SMT larger than 5 cm in 39% of cases; 37% of these patients presented with malignant tumor. Therefore, patients having both SMTs and ulcers should be closely monitored if they are not planned for surgery. Tumor size correlates with malignant potential. Tumors <2 cm in diameter without clinical signs of malignancy or complications are to be followed up at 12 month intervals conservatively.12 The current role of imaging studies for predicting malignant potential is of less importance, so we should consider that all GISTs have certain malignant potential.

Mazur and Clark13 first described GISTs in 1983 as tumors negatively stained with muscle and neuron markers. All stromal tumors of the gastrointestinal tract were classified as 1) smooth muscle type, 2) neural type, 3) combined type, and 4) uncommitted type. This definition was widely accepted and the uncommitted type has been used to describe GIST in a narrow sense.10 Because most GISTs were positive for C-KIT and CD34 staining, they were assumed to have originated from interstitial cell of Cajal. In the pathogenesis of GISTs, activation of kit signaling occurred often by mutations in the c-kit gene is the most important factor. Activation of KIT, which is the product of the c-kit gene, seems important for the growth of GISTs.10

Average diagnostic accuracy rate of EUS-FNA is 60% to 80% in SMTs.4 Recently Mekky et al.4 reported diagnostic utility of EUS-FNA in gastric SMTs. The sampling adequacy was 83% with average 2.5 passes. EUS-FNA results were diagnostic in 43.3%, suggestive in 39%, and nondiagnostic in 17.7%. EUS-FNA results were 95.6% accurate in differentiating potential malignant lesions. Heterogeneous echo pattern predicted sampling adequacy. Better results than previous studies were attributed to on-site cytological analysis.

Sampling adequacy increased with tumor size. There was an increase in sampling adequacy in SMT with a 95% yield for lesion size of >50 mm.4 Another study reported 100% yield of EUS-FNA with lesion size of >40 mm.18 Suzuki et al.6 reported 74.5% diagnostic yield of EUS-FNA for the diagnosis of gastric SMT. Patient age of under 60 years old and location of SMT at lower third area of the stomach were the predictive factors of inadequate tissue yield in EUS-FNA. SMT at lower stomach area was difficult to obtain adequate sample.6

Turhan et al.19 performed a prospective study of EUS-FNA for diagnosis of upper gastrointestinal submucosal lesions and reported 82.9%, 73.3%, 87.9%, 64.7%, and 80% for the sensitivity, specificity, positive, and negative predictive values and accuracy of EUS-FNA, respectively, for upper gastrointestinal tract submucosal mesenchymal tumors.

There are some cases of GISTs that are negative for KIT on cell blocks but positive on the resected surgical pathology. DOG1 (discovered on GIST1, anoctamin1, and ANO1) appears to be expressed highly in GISTs based on gene expression profiling.27 This gene encode a calcium activated chloride channel in the interstitial cell of Cajal, which has a critical role in peristalsis. DOG1 was the more sensitive marker than KIT in all GIST cytologic cell blocks.27

Applying continuous suction with a syringe during EUS-FNA improves the sensitivity of diagnosis of malignancy in patients with solid masses.21

In SMTs, endoscopic forceps biopsy using bite-on-bite technique can provide specimens adequate for diagnosis, so bite-on-bite biopsy should be the first diagnostic step if available.21

Diagnostic yield of EUS-FNA is moderate and limited by unsatisfactory immunostaining in some patients. It can be improved by obtaining samples for cytopathological plus histopathological examinations. The diagnostic yield of EUS-TCB is similar to that of EUS-FNA. There are some cases that EUS-guided sampling is not likely to impact management, so it is not indicated in patients with the followings: 1) plan of surgery for SMT related symptoms, 2) typical echo features of a lipoma, and 3) small (<2 cm) SMTs of the esophagus and the stomach. Also the clinical benefit of EUS-guided sampling in patients with hypoechoic esophageal or gastric SMTs >2 cm is usually limited and should not be overstated.28

EUS-guided sampling is indicated in the following situations: 1) SMTs with a presumptive diagnosis of unresectable GIST for which treatment with tyrosine kinase inhibitors is contemplated, 2) patient with previous history of malignancy with a SMT that may be consistent with a metastasis, 3) suspected diagnosis of lymphoma, neuroendocrine tumor or extrinsic tumor based on EUS, biological or clinical criteria.28

For duodenal and colorectal SMTs, there are not sufficient data to suggest any recommendation.

For definite diagnosis of submucosal lesions, tissue acquisition from submucosal lesion is necessary. EUS-FNA is a good method for the tissue diagnosis of submucosal lesions. EUS-TCB providing core tissue specimen is useful when immunohistochemical staining and tissue architectural details are required. There are problems of procedural difficulties, however, and a lot of efforts have been made to increase the diagnostic yield of EUS-FNA in submucosal lesions.

For accurate diagnosis of submucosal lesions using EUS-FNA specimen, both cytologic and histologic examinations including immunohistochemical stains are necessary. Recently, EUS-FNB needle, a new histologic needle with core trap, has been developed enabling acquisition of core specimens with promising results. To increase diagnostic accuracy of EUS-FNA, we should pay attention to appropriate EUS-FNA procedure and specimen processing. EUS is more accurate in the diagnosis of submucosal lesions if combined with cytology, histology, and immunohistochemical staining using cell blocks.