BACKGROUND. Although atypical or suspicious cytology may support a clinical diagnosis of a malignancy, it is often not sufficient for the implementation of therapy in patients with pancreatic carcinoma. Endoscopic ultrasound-guided fine-needle aspiration biopsy (EUS-FNAB) is a relatively new method for obtaining cytology samples, and one that may decrease the number of atypical/suspicious diagnoses. The goals of the current study were to prospectively evaluate the yield of EUS-FNAB in the diagnosis of patients presenting with solid pancreatic lesions and to evaluate the significance of atypical, suspicious, and false-negative aspirates. METHODS. All patients who presented with a solid pancreatic lesion and underwent EUS-FNAB over a 13-month period were included in the current study. One endoscopist performed all EUS-FNABs. On-site evaluation of specimen adequacy by a cytopathologist was available for each case. Follow-up included histologic correlation (n = 21) and clinical and/or imaging follow-up (n = 80), including 38 patients who died of the disease. RESULTS. EUS-FNABs were obtained from 101 patients (mean age, 62 ± 11.8 years; age range, 34-89 years). The male-to-female ratio was 2:1. Sixty-five percent of the lesions were located in the head of the pancreas, 12% were located in the uncinate, 17% were located in the body, and 6% were located in the tail. The mean size of the tumors was 3.3 cm (range, 1.3-7 cm). A median of 4 needle passes were performed (range, 1-11 needle passes). Sixty-two biopsies (61.4%) were interpreted as malignant on cytologic evaluation, 5 (5%) as suspicious for a malignancy, 6 (5.9%) as atypical/indeterminate, and 26 (25.7%) as benign processes. Of the 76 malignant lesions, 71 were adenocarcinomas, 3 were neuroendocrine tumors, 1 was a lymphoma, and 1 was a metastatic renal cell carcinoma. All except one of the suspicious/atypical aspirates were subsequently confirmed to be malignant. Agreement was complete for the atypical cases. Among the suspicious cases, 2 of the 5 were identified as carcinoma by one cytopathologist and as suspicious lesions by the other, yielding a 40% disagreement rate between the 2 cytopathologists. Therefore, for the 10 atypical or suspicious cases that later were confirmed to be malignant, the final diagnosis of malignant disease was not made due to scant cellularity that could be attributed to sampling error in 8 cases and to interpretative disagreement in 2 cases (20%). All four false-negative diagnoses were attributed to sampling error. Two percent of all biopsies were inadequate for interpretation. Of the 99 adequate specimens, 72 yielded true-positive results, 23 yielded true-negative results, and 4 yielded false-negative results. No false-positives were encountered. Therefore, the sensitivity, specificity, positive predictive value, and negative predictive value of EUS-FNAB for solid pancreatic masses were 94.7% (95% confidence interval [CI], 89.7-99.8%), 100%, 100%, and 85.2% (95% CI, 71.8-98.6%), respectively. CONCLUSIONS. EUS-FNAB is a safe and highly accurate method for tissue diagnosis of patients with solid pancreatic lesions. Patients with suspicious and atypical EUS-FNAB aspirates deserve further clinical evaluation. © 2003 American Cancer Society.