BACKGROUND. Endoscopic ultrasound (EUS) provides detailed imaging of both intramural and extramural structures within the abdomen and mediastinum. However, EUS is limited in its ability to differentiate an inflammatory/reactive process from a malignancy. Fine-needle aspiration biopsy (FNAB), coupled with EUS, allows for the sampling of the target lesion under ultrasound guidance in real time. To better evaluate the clinical utility and efficiency of EUS-FNAB, a retrospective analysis of the first 103 EUS-FNABs performed at our institute was undertaken. METHODS. EUS-FNABs was performed in 80 patients with 103 lesions. Both airdried and alcohol-fixed smears were prepared and stained with Diff-Quik (American Scientific Products, McGraw Park, IL) and Papanicolaou stains, respectively. In addition, ThinPrep slides (Cytyc, Boxborough, MA) and cell blocks, when additional material was available, were also prepared. Immunohistochemical stains were performed on cell blocks wherever required. Cytologic diagnoses were then correlated with the final diagnoses. The latter was based on histologic examination of biopsies/resected pathology materials (n = 54) and clinical follow up (n = 48). Follow-up information was not available for one lesion. RESULTS. Of 103 EUS-FNABs, 42 FNABs were from the pancreas, 38 from the lymph nodes (10 mediastinal and 28 intraabdominal), 10 from the gastrointestinal tract, 7 from the liver, 4 from the adrenal gland, 1 from the biliary tract, and 1 from a retroperitoneal mass. The mean number of passes to obtain diagnostic materials was 3.3. Of 103 EUS-FNABs, 45, 9, 6, and 37 were reported as malignant, suspicious, atypical, and benign, respectively. Six FNABs were nondiagnostic. The authors did not encounter any false-positive cases. There were three false-negative cases (two pancreatic carcinomas and one gastrointestinal stromal tumor of the stomach). No complications were encountered. The sensitivity, specificity, and accuracy were 71%, 100%, and 81%, respectively. If the FNABs that were classified as suspicious were considered as malignant, the sensitivity, specificity, and accuracy were 86%, 100%, and 91%, respectively. CONCLUSIONS. EUS-FNAB is a safe and accurate diagnostic procedure for the evaluation of intramural and extramural lesions of the gastrointestinal tract. In the majority of cases, it obviates the need for more invasive diagnostic procedures to obtain a tissue diagnosis. © 2002 American Cancer Society.