Endoscopic ultrasound (EUS) plays an important role in the characterization of subepithelial lesions by providing information regarding their layer of origin, echogenicity, size, and local invasion. However, sonographic characteristics alone may be insufficient to distinguish subepithelial and hypoechoic third or fourth layer lesions. Several studies estimate the accuracy of EUS for the diagnosis of subepithelial lesions to be 43% to 48%.1,2 There is currently no consensus regarding the best management strategy for diagnosing and treating small subepithelial lesions.3 Operative planning, the type of surgery, and the follow-up period vary dramatically with respect to the diagnosis.

EUS-guided fine needle aspiration (EUS-FNA) is a means of sampling for cytologic confirmation. In addition, EUS-FNA in combination with immunohistochemistry (IHC) assists in differentiating various subepithelial tumors. Previous studies have demonstrated that EUS-FNA accurately diagnoses subepithelial lesions with an accuracy of 80% to 98% compared with the final diagnosis, whereas inadequate cytologic sampling occurs in 10% to 18% of procedures.4-7 An accurate EUS-FNA diagnosis requires an adequate cytologic specimen for pathologic evaluation; it is also important to decrease the frequency of nondiagnostic results and diagnostic errors.8

The purpose of this study was to assess the performance characteristics, clinical role, and accuracy of EUS-FNA for the diagnosis of gastrointestinal (GI) intramural masses, and to estimate its clinical impact on patient management.

Approval for this retrospective study was obtained from Institutional Review Boards of The Catholic University of Korea College of Medicine. Between June 2009 and November 2011, 46 consecutive patients who underwent EUS-FNA of upper GI intramural masses were identified from the medical records at Seoul St. Mary's Hospital. The following data were collected: patient demographics, location of the lesion, endoscopic and EUS characteristics of the lesion (size, echogenecity, and wall layer of origin), size of needle, number of needle passes, cytologic diagnosis from FNA, and final pathology based on a resected specimen, when available.

Following conventional esophagogastroduodenoscopy, EUS examinations were performed by using curvilinear array echoendoscopes (GFUCT-240; Olympus, Tokyo, Japan) connected to a US scanning system (SSD 5500; Aloka, Tokyo, Japan) by a single experienced endosonographer (Y.K. Cho). All samplings were achieved by using 22-gauge needles (Echotip; Wilson-Cook Inc., Winston-Salem, NC, USA; or Easyshot; Olymphus Medical System Corp., Ltd., Tokyo, Japan). Patients were followed up for 24 hours after the procedure to detect procedure-related complications. Because no on-site cytopathologist was available, at least three or four passes were conducted for each target lesion. Additional FNA passes were performed based on the endoscopist's gross assessment of sample adequacy. Ten to 20 repetitions of back and forth movements were performed per single needle pass. Cytologic samples were processed by the same experienced cytopathologist. Aspirates were placed on glass slides, and alcohol-fixed smears were prepared for cytologic examination. The remaining material was then flushed from the needle and collected in CytoLyt preservative (Cytyc Corp., Marlborough, MA, USA) for cell block preparation.

We assessed the performance characteristics, clinical role, and accuracy of EUS-FNA in evaluating intramural masses. We report a yield of 87% for EUS-FNA sampling of submucosal lesions in the upper GI tract. Although our overall yield was similar to that in other recent studies, the proportion of diagnoses that were based on cytology and IHC was lower. In our study, spindle cell tumors comprised 33% of all EUS-FNA diagnoses (15 of 46 lesions). Twenty percent of spindle cell lesions were adequately stained for these antibodies. There are many reports of mesenchymal cell tumors in the literature. For example, a study from Japan reported a sample adequacy rate of 83% for gastric submucosal tumors.10 However, this included 39% of samples that met suggestive cytologic criteria without IHC staining. We suspect that, in most cases, this is related to the absence of on-site cytopathology and small size of lesions. The actual number of passes required for diagnosis may be less relevant when an on-site cytopathologist confirms specimen adequacy.

The level of experience of the endoscopist, the needle gauge, the number of needle passes attempted, the location of the lesion, and the on-site presence of a cytotechnologist or pathologist may affect cytologic adequacy.11-13 In this study, neither the number of needle passes nor the size of the lesion was related to diagnostic yield. The median size of lesions that were insufficiently sampled or subject to diagnostic error was 2×1.5 cm. Most lesions in this study were approximately 2 cm (range, 1 to 20.6) and therefore insufficient for Trucut biopsy. Many lesions were located in the stomach, duodenum, and angulated portion, which also restricted the use of 19-gauge needles and Trucut biopsy because of potential handling difficulties. All procedures were performed with a 22-gauge needle, which facilitated adequate sampling irrespective of lesion size and location. However, this size may be insufficient to obtain a cell block, which is useful for diagnosing the subtypes of submucosal tumors. The reason for comparatively lower yields of the esophageal lesions is unknown; the esophagus is straighter and therefore allows easier needle handling compared with other organs such as the stomach. Because failed procedures were performed during the early stages of the learning curve, this may have affected cytologic outcomes.2

The limited value of EUS-FNA for the differential diagnosis of subepithelial masses was primarily the result of the rate of nondiagnosis and diagnostic error. The former is determined by technical failures, unsatisfactory specimens, and atypical and suspicious diagnoses; the latter is determined by false-positive and false-negative diagnoses, and errors in interpretation.8 Discordance between cytologic results and the final pathologic diagnosis may critically influence and misdirect patient management. In this study, the diagnostic accuracy of EUS-FNA for differentiating benign lesions from potentially malignant lesions and malignant lesions was 82%. There were three cases exhibiting discordance with the final pathology. Of these, two cases in which the cytologic diagnosis was benign were confirmed clinically to be a duodenal neuroendocrine tumor and ampulla myoblastic tumor, respectively, and one patient with a presumed GIST in the gastric antrum with atypical spindle cells was diagnosed with differentiated adenocarcinoma. Pathologic review suggested that atypical spindle cells were rarely obtained from the stromal portion of adenocarcinomas. This was possibly because of difficulties in communication with pathologists and insufficient specimens for IHC. In contrast, high-grade GIST should be distinguished from poorly differentiated carcinoma based on spindle cell features. The use of immunocytochemistry may be helpful in making this differential diagnosis. Tumor cells from GIST are positive for c-kit, whereas smooth muscle cells from the bowel wall and from spindle cell carcinoma are negative for this marker. Spindle cell carcinoma is positive for cytokeratin expression, whereas GIST is not. A duodenal submucosal tumor was initially misdiagnosed as ectopic pancreas tissue by using EUS. Subsequent surgical resection revealed a well-differentiated neuroendocrine tumor. Retrospective review revealed that the aspirate contained pancreatic ductal epithelium. To reduce the false-negative rate, it is critical to obtain a sufficient sample for cytology. In addition, the provision of sufficient information to the pathologist is also important, in particular when no on-site pathologist is available.

The decision to resect an intramural lesion also depends on the findings from EUS. Considering the morphologic EUS features of a tumor in combination with its cytohistologic FNA assessment and location enhances the power and accuracy of the preoperative diagnosis of subepithelial masses. However, EUS imaging alone is unable to differentiate hypoechoic third layer masses. If EUS-FNA cytology is able to accurately differentiate potentially malignant lesions, unnecessary procedures would be avoided, particularly for lesions of borderline size (i.e., 1 to 2 cm). National Comprehensive Cancer Network guidelines regarding GIST recommend resection of masses of >2 cm, and clinical follow-up for lesions of <1 cm. The lack of guidelines for lesions of medium size may lead to confusion regarding patient management.3,14 To date, there is no consensus regarding the indications for use of EUS-guided tissue sampling. In our study, 73% of sampling results influenced clinical judgment. In particular, unnecessary surgery was avoided in the majority of patients. The role of EUS-FNA in clinical management depends on the appropriateness of the indication. To avoid unnecessary procedures, EUS-FNA should be performed in accordance with the appropriate indication to positively influence the clinical outcome.

In conclusion, EUS-FNA exhibited a high diagnostic yield for most small intramural and extraintestinal lesions, and a diagnostic accuracy of 82% for differentiating malignant lesions. The results of most procedures influenced clinical judgment. However, EUS-FNA has some limitations for the differential diagnosis of intramural masses, primarily because of nondiagnostic sampling (9%), and more importantly, the likelihood of diagnostic error (6%); these factors may influence the clinical utility of EUS-FNA. Better tissue acquisition techniques and more accurate pathologic interpretation are necessary for improved differential diagnosis.