We conducted a randomized, controlled trial to prospectively confirm that ultrasound-directed transbronchial needle aspiration (USTBNA) results in: (1) improved sensitivity for detecting lymph nodes involved with neoplasm, and (2) a decreased number of aspirates needed to achieve a diagnosis as compared with standard transbronchial needle aspiration (TBNA). The study was conducted in a tertiary medical center on patients undergoing fiberoptic bronchoscopy in the evaluation of enlarged mediastinal lymph nodes. USTBNA or TBNA were followed by rapid, on-site cytopathology examination of the collected specimens. Measurements included the (1) age and sex of the patient, prior diagnosis of cancer, nodal short-axis diameter and node location as determined by computerized tomography (CT), and endobronchial abnormalities at bronchoscopy; (2) number, order, and location of transbronchial aspirates and results of on-site evaluation; (3) results of surgical exploration in patients with negative transbronchial needle aspiration; (4) sensitivity, specificity, and diagnostic accuracy of USTBNA and TBNA; (5) number of aspirates required for successful lymph node aspiration as well as for a diagnosis of cancer for both USTBNA and TBNA; and (6) multiple logistic regression analysis to determine the significance of combinations of clinical predictors and needle aspirate results. Eighty-two bronchoscopic examinations were performed on 80 patients. We found no significant difference between USTBNA and TBNA in sensitivity (82.6% versus 90.5%, respectively), specificity (100% for both), or diagnostic accuracy (86.7% versus 91.7%, respectively). The sensitivity, specificity, and diagnostic accuracy of USTBNA and TBNA were similarly high, regardless of node location (paratracheal or subcarinal). A decrease in the number of aspirates required for lymph node sampling approached statistical significance for all USTBNAs as compared with TBNAs (2.03 +/- 0.19 versus 2.62 +/- 0.25, p = 0.06), but this was not demonstrated for the number required to confirm cancer (1.95 +/- 0.47 versus 2.68 +/- 0.21, p = 0.17). The number of aspirates to successful lymph node aspiration decreased with USTBNA versus TBNA in paratracheal lymph nodes (2.00 +/- 0.20 versus 2.91 +/- 0.34, p = 0.03), but not to a diagnosis of cancer (1.93 +/- 0.25 versus 3.00 +/- 0.58, p = 0.11). No difference was seen in the number of aspirates for subcarinal nodes. The number of TBNA attempts for paratracheal lymph node sampling was inversely correlated with node size (r = 0.48, p = 0.02). No such relation was seen with USTBNA of paratracheal nodes (r = 0.09, p = 0.66), TBNA of subcarinal nodes, or USTBNA of subcarinal nodes. A similar relation was seen between the number of aspirates to a diagnosis of cancer. On multiple logistic regression analysis, a positive transbronchial aspirate was associated only with a larger lymph node and history of prior cancer. We conclude that: (1) in the setting of on-site cytopathology, transbronchial needle aspiration has a high sensitivity, specificity, and diagnostic accuracy in the evaluation of enlarged mediastinal lymph nodes suspected of harboring malignancy; (2) mediastinal anatomy, including vascular structures and lymph nodes, is clearly imaged with endobronchial ultrasonography; (3) a greater short-axis diameter of the mediastinal lymph node and history of a prior malignancy increase the likelihood of a positive transbronchial aspiration; (4) USTBNA exhibits a similarly high diagnostic yield to TBNA in the setting of rapid on-site cytopathology evaluation; (5) USTBNA decreases the number of aspirates required for paratracheal lymph node sampling, which may be particularly useful in sampling smaller paratracheal nodes or at institutions that do not utilize rapid on-site cytopathology evaluation.