Endoscopic ultrasound-guided fine needle aspiration cytology (EUS-FNA) is considered to be a minimally invasive and safe technique, with low complication rates, for obtaining tissue samples from pancreatic lesions, and mediastinal and intra-abdominal nodes. We retrospectively reviewed the diagnostic accuracy of this method at a tertiary care hospital in Pakistan. The medical records of 393 consecutive patients referred to the gastroenterology service from August 2008 to September 2013 were reviewed in this retrospective study. The mean age of the patients at presentation was 51.2 years (range 6-89; 56.7% males). Major indications for referral were to establish a diagnosis in patients with a pancreatic mass or lymphadenopathy of unknown origin detected on CT scan or PET-CT (n = 235), to exclude lymph node metastasis in those with a known primary tumor (n = 114), and to rule out relapse of lymphoma after treatment or during follow up (n = 44). Of 393 patients, 181 had mediastinal lymphadenopathy, 115 had pancreatic lesions; 79 had intra-abdominal lymphadenopathy and 18 patients had miscellaneous lesions (gastric mass, splenic, retroperitoneal, esophageal, or adrenal lesions). A mean of 2.2 passes per patient were made; 85% of patients had ≤ 3 passes. The initial rapid onsite evaluation (ROSE) was positive in 363 of 393 (92.3%) patients, but on final cytopathological interpretation, 369 of 393 (93.9%) patients had an adequate specimen for diagnosis. Sufficient material for cytopathology was obtained in 369 patients (93.9%), whereas 24 patients (6.1%) had inadequate aspirate on EUS-FNA as assessed by final cytopathological interpretation. The concordance between final cytopathological diagnoses and ROSE was 98.2%. Follow-up data were available for 296 (75.3%) patients, whereas 97 patients were lost to follow up. Of these 296 patients, 283 of 296 patients were true positive, 9 of 296 were true negative, and 4 of 296 patients were false negative. The overall calculated sensitivity, specificity, positive predictive value, and negative predictive value were 98.6, 100, 100, and 69.2%, respectively. The major diagnoses that we obtained were granulomata (n = 114), adenocarcinoma (n = 104), reactive nodes (n = 35), metastatic carcinoma (n = 35), Hodgkin's lymphoma (HL; n = 19) and non-Hodgkin's lymphoma (NHL; n = 11), neuroendocrine tumors (n = 11), spindle cell lesions (n = 7), benign lesions (n = 9), cystic neoplasms (n = 6), solid pseudopapillary neoplasm (n = 3), and miscellaneous (n = 15). Only 3 of 393 patients (0.7%) suffered a complication as a result of EUS-FNA, and all three were successfully managed conservatively. EUS-FNA, combined with ROSE, is a safe and highly effective technique in experienced hands with excellent diagnostic accuracy and low complication rates.