Background: Rapid and accurate tissue diagnosis for a deep-seated malignancy would allow treating physicians to provide disease-specific interventions and help patients make early informed management decisions. Providing on-site tissue diagnosis for fine-needle aspirate samples obtained with endosonography would help develop such efficient patient management issues. Here we report our experience of prospectively providing on-site diagnosis on 485 endoscopic ultrasound fine-needle aspirate samples. Methods: Four hundred eighty-five endoscopic ultrasound fine-needle aspirates from the pancreas (n= 305), lymph nodes (n = 91), biliary tree (n = 47), liver (n = 15), gastrointestinal tract (n = 19), and adrenal gland (n = 8) were reviewed. For all aspirates, the cytologic diagnoses, both preliminary and final, were categorized into the following: positive for malignancy, positive for neoplastic process, suspicious for malignancy, atypical cells, reactive process, and nondiagnostic. Results: Of the 485 cases, 163 (33.6%) were diagnosed as benign, 43 (8.8%) as atypical, 21 (4.3%) as suspicious, 18 (3.7%) as positive for neoplasm, and 230 (47.4%) as malignant after final cytologic interpretation. A significantly (P < .001) higher degree of concordance was noted for unequivocal diagnosis of malignancy (196/198, 98.9%) vs nonmalignancy (200/250, 67.2%) between on-site and final cytologic diagnosis. Of the 52 discordant cases, 12 (2.6%) diagnoses were downgraded and 40 (8.9%) were upgraded from preliminary on-site diagnosis. Our overall sensitivity (87 vs 92), specificity (95% vs 100%), and accuracy (90% vs 94%) improved for final cytologic diagnosis. Conclusion: On-site diagnosis of malignancy could be used to initiate informed patient management decisions. Cases where a diagnosis of malignancy is not rendered at on-site interpretation need further cytologic evaluation. © 2007 Elsevier Inc. All rights reserved.