In staging bronchogenic carcinoma by transbronchial needle aspiration (TBNA), rigid histology needles are generally preferred to flexible cytology needles owing to the widespread opinion that rigid needles have higher diagnostic yield and less false-positive results. The objective of this study was to compare the efficacy and safety of the rigid and flexible TBNAs in staging bronchogenic carcinoma to establish whether a flexible cytology needle method can replace the rigid needle. A prospective study was conducted in 138 consecutive patients with extra- or endobronchial masses suggestive of bronchogenic carcinoma and amenable to surgical procedures. All 8 mm and larger paratracheal, carinal, hilar and/or main bronchial lymph nodes determined before bronchoscopy by computed tomography (CT) were sampled by successive 18-gauge rigid and 21-gauge flexible TBNAs in the same session. The anatomic landmarks were followed precisely during TBNAs, and a proper technique applied in sampling and specimen processing. Malignant lymph node involvement was specified in 97 (72%) cases of bronchogenic carcinoma by rigid, and in 89 (66%) by flexible TBNA. There were 4 (100%) benign cases (3 with tuberculosis and 1 with sarcoidosis) of 101 (73%) with positive rigid TBNAs (82 with histological and 19 with cytological specimens). TBNAs determined malignant lymph node involvement in a total of 104 (78%) patients. Of 30 TBNA-negative patients, 14 were proven to have false-negative TBNAs by mediastinoscopy/mediastinotomy/minithoracotomy, and 16 to have true-negative TBNAs by thoracotomy. Thoracotomy confirmed true positivity in 52 rigid and 49 flexible TBNAs, and false negativity in 4 rigid and 7 flexible TBNAs. Further staging was confirmed in these 7 cases. Four had proven false-negative results by both methods. The presence of small cell carcinoma (21) or N3 disease (27) presented a contraindication to thoracotomy in 48 TBNA-positive patients. Adequate-quality and malignant lymph node specimens were more frequently obtained by both techniques at advanced tumor and node stages. However, malignant lymph node invasion was significantly more frequent in rigid and flexible TBNA specimens only in the presence of advanced tumor status and abnormal endoscopic appearance. The sensitivities of rigid and flexible TBNAs were 74 and 70%, respectively (p > 0.05), but both had a specificity of 100%. Neither false-positive results nor serious complications other than hemorrhage of 30-100 ml (rigid: 5%, flexible: 2%) were encountered with either technique. These results indicate that in bronchogenic carcinoma, hilar and mediastinal lymph nodes can be staged by 21-gauge flexible TBNA (76%) as accurately as by 18-gauge rigid TBNA (79%) if a proper technique is applied and anatomic landmarks are followed precisely (p > 0.05).