The purpose of this study was to determine the diagnostic accuracy and frequency of complications of CT-guided transthoracic needle biopsy done with an automated biopsy system. Thirty-three consecutive biopsies were performed on 32 patients between February 1992 and July 1994 (mean age, 55 +/- 15 years; 18 men and 14 women). An 18-gauge (n = 28) or 20-gauge (n = 5) needle was used. Core specimens were submitted for pathologic examination in 10% formalin. No cytopathologist or frozen section analysis was available at the time of biopsy. All biopsies but one were performed by one chest radiologist. Thirty-one lung biopsies and two mediastinal biopsies yielded a mean lesion size of 4.0 cm (range, 1.2-13.0 cm). Postbiopsy pneumothorax occurred in three (9%) of 33 biopsies; none of the pneumothoraces required placement of a chest tube. The mean number of needle passes was 1.3 (+/- 0.6). Thirty biopsies (91%) yielded sufficient tissue for pathologic evaluation. The diagnoses included carcinoma in 14 cases, acute or chronic pneumonia in 4 cases, non-Hodgkin's lymphoma in two cases, and Kaposi's sarcoma, plasma cell granuloma, hypersensitivity pneumonitis, Pneumocystis carinii pneumonia, and fibrosis in one case each. One patient with a 1.5-cm nodule stable for 1 year on CT had fibrosis and chronic inflammation found on needle biopsy, and the nodule was considered benign. Overall, biopsies in 12 (80%) of 15 patients without carcinoma were diagnostic. In three patients, the tissue obtained was not representative of the underlying abnormality. The biopsy specimen showed only inflammatory changes in two patients who ultimately had proved carcinoma. One patient with multiple pulmonary infarcts due to tumor emboli showed evidence of only pulmonary infarct on biopsy. Three patients had insufficient tissue for analysis; none of the three had malignant tumor on follow-up. The sensitivity of CT-guided automated needle biopsy of the chest was 84%. CT-guided transthoracic needle biopsy of the chest done with an automated biopsy system is safe, with a pneumothorax rate comparable to that of skinny needle aspiration. An overall accurate tissue diagnosis was made in 26 (81%) of 32 patients. Biopsies in 12 (80%) of 15 patients without carcinoma were diagnostic, which compares favorably with the reported accuracy of skinny needle aspiration.