MATERIALS AND METHODS. We retrospectively reviewed a consecutive series

Huanqi Li1 Phillip M. Boiselle1 2 Jo-Anne 0. Shepard1 Beatrice Trotman-Dickenson1 Theresa C. McLoud1 Received January 2, 1996; accepted after revision Febru ary 19, 1996. tchest Division, Department of Radiology, Massachusefls Leneral Hospital, Boston, MA 02114. 2Presentaddress:Chest Division, Department of Diag. nostic Imaging. Temple University, 3401 N. Broad St., Philadelphia, PA 19140. Address correspondence to P.M. Boiselle. AJR1996;167:105-109 0361-803X/96/1671-105 American Roentgen Ray Society Diagnostic Accuracy and Safety of CT-Guided Percutaneous Needle Aspiration Biopsy of the Lung: Comparison of Small and Large Pulmonary OBJECTIVE. The purpose of this study was to compare the diagnostic accuracy and safety of CT-guided percutaneous needle aspiration biopsy of pulmonary nodules less than or equal to I.5 cm in diameter with those of nodules greater than I.5 cm in diameter. MATERIALS AND METHODS. We retrospectively reviewed a consecutive series of 97 patients who underwent CT-guided percutaneous needle aspiration biopsy of a lung nodule and then surgical resection (n = 95) or autopsy (n = 2). By examining CT images, we classified 27 nodules as small (l.5 cm) and 70 nodules as large (>1.5 cm). Diagnostic accuracy was calculated by comparing cytologic diagnoses based on biopsy with final diagnoses based on histologic findings from surgery or autopsy. Each case was reviewed for possible complications, including pneumothorax and chest tube placement. RESULTS. The diagnostic accuracy of CT-guided percutaneous needle aspiration biopsy of large nodules was 96%. The diagnostic accuracy for small nodules was 74%, a statistically significant difference (p <.05). The prevalences of pneumothorax in our population were nearly identical for small and large nodules (22 and 21 %, respectively). The prevalence of chest tube placement in our population was approximately 2%. The prevalences of chest tube placement were 0% for small nodules and 3% for large nodules. CONCLUSION. CT-guided percutaneous needle aspiration biopsy is significantly less accurate for small pulmonary nodules than for large pulmonary nodules, but the complication rates for both are low. P ercutaneous needle aspiration biopsy of the lung is a relalively safe and accurate means of diagnosing benign and malignant lesions of the chest [1-7]. Such biopsies of small pulmonary nodules may provide an early diagnosis of lung cancer, thus enabling effective intervention and increasing the potential for surgical cure. CT is better able than conventional radiography or fluoroscopy to detect small nodules, and CT-guided percutaneous necdie aspiration biopsy of pulmonary lesions as small as 3 mm has been described in the literature [5]. At our institution, CTguided percutaneous needle aspiration biopsy of small lung nodules, many of which would not be visible by conventional radiography or fluoroscopy, is routinely performed. We wondered whether differences exist in the diagnostic accuracy and complication rate of CT-guided percutaneous needle aspiration biopsy of small and large pulmonary nodules. The purpose of this study was to compare the diagnostic accuracy and safety of CT-guided percutaneous needle aspiration biopsy of pulmonary nodules less than or equal to 1.5 cm in diameter with those of pulmonary nodules greater than I.5 cm in diameter. Materials and Methods We retrospectively reviewed a consecutive series of 97 patients at our institution who underwent CTguided percutaneous needle aspiration biopsy of a pulmonary nodule and then surgical resection (n = 95) or autopsy (n = 2). All biopsies were performed during the 18-month period from January 1994 through June 1995. The study population included AJR:167, July 1996 105

Li et ai. 50 men and 47 women, who were 39-88 years old (mean, 66.8 years old) and who had been referred for biopsy of a pulmonary nodule. At our institution, all pulmonary nodules are biopsied with CT guidance rather than fluoroscopic guidance. All nodules were imaged on a CT scanner (Advantage: General Electric Medical Systems, Milwaukee, WI). We measured the long-axis diameter of the nodules on lung window settings. less than or equal to I.5 cm in diameter were measured on images obtained with a slice thickness of either 3 or 5 mm, and nodules greater than 1.5 cm in diameter were measured on images obtained with a slice thickness of either 5 or 10 mm. Lesions less than or equal to I.5 cm in longaxis dimension were categorized as small nodules (n = 27). They ranged in size from 4 mm to I.5 cm (mean, 1.15 cm; median, 1.20 cm). Lesions greater than I.5 cm in diameter were categorized as large nodules (ii = 70) and ranged in size from 1.6 to 8.2 cm (mean, 3. 17 cm; median, 3.09 cm). Radiologists performed biopsies in accordance with the protocol ofthe thoracic radiology division at our institution. Biopsies were performed by experienced chest radiologists, chest radiology fellows, or radiology residents under the supervision of a chest radiologist. Biopsies were assigned to staff radiologists and radiology fellows on a random basis, regardless of the size of the nodule or the anticipated difficulty of the biopsy. Before each procedure, informed consent was obtained. Patients were placed in a prone. supine, or lateral decubitus position, depending on the location of the lesion. Localization was performed by CT imaging with laser lights and a grid system. Local anesthesia with 1% lidocaine (Xylocaine; Astra USA, Westborough, MA) was administered subcutaneously. All biopsies were perlbrmed by a coaxial technique with a 19-gauge introducer needie (Greene needle: Cook, Bloomington, IN) and a 22-gauge aspirating needle (Greene needle: Cook). The number of pleural passes was recorded for each case. Before an aspirate was obtained, a CT image was obtained to document successful placement of the I 9-gauge introducer needle within the lesion. A cytotechnologist and a staff cytopathologist were present at all biopsies. The specimens obtained were placed on sterile glass slides and immediately smeared by the cytotechnologist. Rapid staining of the specimens was performed with a modified hematoxylin-eosin technique, and an opinion was rendered as to the adequacy of each cytologic specimen for diagnosis. When specimens were not sufficient for diagnosis, additional aspirates were obtained. After removal of the biopsy needle, patients were immediately placed in a puncture side-down position, as described by Moore et al. [8], and remained in the radiology observation area on a stretcher, under the supervision of a nurse, for 3 hr. Talking and coughing were discouraged. With patients erect, we obtained chest radiographs I and 3 hr after biopsy. We recorded the presence of pneumothorax at the time of biopsy, as detected on CT scans, and after biopsy, as detected on chest radiographs. Patients with pneumothoraces were administered oxygen by nasal cannula to speed resorption of the pneumothoraces 131. Patients with enlarging pneumothoraces on serial chest radiographs and those with symptomatic pneumothoraces were treated with placement of a chest tube. All final diagnoses were based on pathologic findings from surgery (ii = 95) or autopsy (,z = 2). The mean times from biopsy to surgery or autopsy were 33.2 days for patients with small nodules (range, 1-120 days) and 27.8 days for patients with large nodules (range, 1-308 days). During the time course of the study, a total of 202 patients underwent CT-guided biopsy of a lung nodule. Ninety-seven of these patients underwent surgical resection (ii = 95) or autopsy (ii = 2) and were included in the study population. One hundred five patients did not undergo surgery or autopsy and thus were not included in the study population. In six patients, a specific benign diagnosis was made by percutaneous needle aspiration biopsy. Because these patients did not undergo subsequent surgery or autopsy, they were not included in the study population. The benign diagnoses included hamartoma (0 = I ), desquamative interstitial pneumonitis (ii = 1), infection (n = I ), and fibrous scar (is = 3). Sixtysix patients underwent percutaneous needle aspiration biopsy of a nodule with a cytologic diagnosis positive for malignancy but did not undergo surgical resection of the nodule or autopsy. In most cases, the patients were not considered to be surgical candidates on the basis of either coexisting medical illnesses or the presence of metastatic disease. Thirty-three patients underwent percutaneous needie aspiration biopsy of a nodule with a cytologic diagnosis negative for malignancy but not diagnostic of a specific benign entity but did not undergo surgery or autopsy. Biopsy results were classified as ( I ) positive for malignancy (cytology showed diagnostically malignant cells); (2) suspicious for malignancy (cytology revealed atypical cells); and (3) negative for malignancy (cytology did not show atypical or malignant cells but was not diagnostic of a specific benign entity, such as granuloma or hamartoma). Cases with cytologic findings considered suspicious for malignancy were classified as positive for malignancy. Diagnostic accuracy was calculated by cornparing cytologic diagnoses based on biopsy with final diagnoses based on histologic findings from surgery or autopsy. Sensitivity and specificity were calculated for nodules proven malignant at surgery or autopsy. The chi-square test was used to assess the statistical significance of our results. Results Final diagnoses determined by surgery (ii = 95) or autopsy (n = 2) are shown in Table 1. Both small- and large-nodule groups predominantly comprised malignant lesions, which accounted for 23 (85%) of 27 small nodules and 62 (89%) of 70 large nodules. All 23 cases of malignant small nodules represented primary lung cancers; 56 (90%) of 62 cases of malignant large nodules represented primary lung cancers, and the remaining six (10%) were metastatic lesions. Overall, 12 ( I 2%) of 97 nodules represented benign lesions (Table 2). The categorization of biopsy results as positive for malignancy, suspicious for malignancy, and negative for malignancy is shown in Table 3 for both small and large pulmonary If.1U1Final Diagnosis nodules. Diagnoses Small No.of: Large Small-cell 3 9 12 Non-small-cell 20 47 67 Metastases 0 6 6 Benign lesions 4 8 12 27 70 97 IfFina1 Benign Diagnoses.. No.of Diagnosis Hamartoma 2 Granuloma 1 Infection (tuberculosis or 3 fungal infection) Bronchiolitis obliterans with 2 organizing pneumonia Bronchiectasis 1 Inflammation (nonspecific) 3 12 Cogi#{233} Results of v.t:ippercutaneous Needle Aspiration Biopsy. ;, :.,- No.of: ii*tp il :. : SmaII Larej. F : #{149}. i; 1-:i - ) Y ij:74.:rh; 97 106 AJR:167, July 1996

Percutaneous Needle Aspiration Biopsy of Lung The diagnostic accuracy, sensitivity, and specificity of percutaneous needle aspiration biopsy of small and large nodules are shown in Table 4. An accurate diagnosis was made for 67 (96%) of 70 large pulmonary nodules and for 20 (74%) of 27 small pulmonary nodules. a statistically significant difference (p <.05). We found a similar difference in sensitivity: 72% for small nodules and 94% for large nodules (/) <.05). Iccuracy, Specificity, and ISensitivityof Percutaneous eedle Aspiration Biopsy of Small and Large.. Value(%)for: p (Chi ar Snall r,lare I Square ;N#{243}dul#{233}s1 Nodul#{233}s4,TestLI --- :..I.,,..;, Accuracy. :t*w -. % ; <O5 Percutaneous Needle Aspiration Biopsy Results for False-Negative Cases Patient Sex Age (Yr) Nodule ame er,cm, D.. Lesion Biopsy Result. Final Result Female 53 1.0 Mixed inflammatory cells Adeno Male 61 1.5 Reactive pneumocytes Adeno Female 57 1.0 Fibrous and amorphous necrotic debris Adeno Male 64 1.1 Mildly atypical cells Adeno Male 70 1.3 Reactive bronchial cells Carcinoid tumor Male 67 2.6 Granular cells and benign mesothelial cells Large-cell Female 63 2.0 Reactive bronchial cells Adeno Female 75 3.2 Fragments of cells Leiomyosarcoma Female 63 2.9 Inflammatory cells and epithelial_histiocytes A false-negative diagnosis occurred in nine of 97 cases, for an overall false-negative rate of 9%. The false-negative rates were 19% (5127) for small nodules and 6% (4170) for large nodules. in addition to the nine false-negative cases of malignant neoplasms, there were two cases of benign neoplasms that were not diagnosed by percutaneous needle aspiration biopsy before surgery. Both patients had hamartomas, and both patients had small nodules that measured 1.0 and 1.2 cm in diameter. The percutaneous needle aspiration biopsy results and final diagnoses for the false-negative cases are shown in Table 5. None of the nine patients with false-negative diagnoses underwent repeat percutaneous needle aspiration biopsy. Three of the nine patients subsequently underwent video-assisted thoracoscopic surgery, and the remaining six underwent thomcotomy. There were no false-positive diagnoses. The number of true-positive and falsenegative diagnoses for small nodules according to specific size (in millimeters) is shown in Table 6. Carcinoid tumor t:1iprevaience of Pneumothorax Small Large 27 70 No. of Patients With Pneumothorax 6 15 Prevalence (%) 22 21 97 21 22 A single pleural pass was made in all 27 patients with small nodules and in 70 (93%) of 75 patients with large nodules. The prevalence of pneumothorax for our study group is shown in Table 7. The overall prevalence of pneumothorax was 22% (21/ 97); prevalences for patients with small and large nodules were similar (22 and 21%, respectively). Overall, only two of the 21 cases complicated by pneumothorax required chest tube placement, resulting in a chest tube placement rate of approximately 2%. Both patients requiring chest tube placement had pulmonary nodules classified as large. Therefore, the chest tube placement rate for patients with large nodules was 3%, and that for patients with small nodules was 0%. No fatalities occurred after biopsy in either group of patients. Discussion Percutaneous needle aspiration biopsy is a relatively safe and accurate means of diagnosing benign and malignant lesions of the chest [1-7]. Reported accuracies range from 64% to 97% [1-7], and major complications are rare [1-7]. Pneumothorax is the most common complication after percutaneous needle aspiration biopsy [3], with reported rates of 19-44% 11-7); the range of reported rates of chest tube insertion is 1.6-14.3% I I -7]. The results of this study indicate that the accuracies of CT-guided percutaneous necdle aspiration biopsy of small and large nodules differ significantly. However, we found no significant difference in the complication rates. The accuracies of percutaneous needle aspiration biopsy were 74% for small nodules and 96% for large nodules, a statistically significant difference. Despite this difference, the accuracy of percutaneous needle aspiration biopsy for small nodules falls within the range of accuracies reported in the literature [1-7]. Previous studies have reported accuracies for percutaneous needle aspiration biopsy of small and large nodules that conflict with one another [5, 9, 101. In a fluoroscopically guided biopsy series, Poe and Tobin [9] reported no significant difference in accuracy for small and large nodules. However, in that study, larger lesions predominated and the researchers defined small nodules as being less than 4 cm in diameter. Although the criteria for small nodules are arbitrary, AJR:167, July 1996

Li et al. we defined such nodules as being less than of small nodules is likely secondary to two series of fluoroscopically guided percutane- or equal to 1.5 cm in diameter. Our rationale main factors. The first factor is sampling ous needle aspiration biopsies of 64 small for this size criterion is that in our clinical error, because it may be more difficult to (2 cm in diameter) and 61 large nodules by experience, on the basis of size alone, nodules greater than 1.5 cm in diameter usually do not pose technical difficulty, whereas smaller nodules often are technically challenging. In fact, the average size of large nodules in our series was 3.17 cm, a value smaller than the size criterion for large nodules in the series of Poe and Tobin [9]. A series of fluoroscopically guided biopsies by Hayata et al. [ 10] suggested possible differences in accuracy for small and large nodules. After defining small nodules as being less than 2 cm in diameter, these researchers found accuracies of 75.0 and 89.6% for small and large nodules, respectively. The series of Hayata et al. [10] included only peripheral nodules, all of which were malignant; our series included central and peripheral nodules, the majority of which were malignant (86%). In the series of Hayata et al. [10], only 19% ofnodules were classified as small; our series included a higher percentage of small nodules, 28%, although this percentage still was considerably smaller than the percentage of large nodules. The present study and previous studies also differed in technique: biopsies were performed with fluoroscopic guidance in previous studies. CT is more effective than fluoroscopy at detecting small pulmonary lesions. In a series of CT-guided biopsies of small and difficult thoracic lesions, vansonnenberg et al. [5] also suggested that the accuracies of CT-guided percutaneous needle aspiration biopsy may differ for lesions of different sizes. In that series, vansonnenberg et al. [5] found accuracies of 73.9% for lesions less than or equal to 1.0 cm in diameter, 83.9% for lesions 1.1-2.0 cm in diameter, and approximately 90% for lesions greater than 2.0 cm in diameter. However, their series had a selecflon bias: all lesions were initially considered for fluoroscopically guided biopsy, but only lesions considered too difficult or too small for fluoroscopically guided biopsy were included in the series. Also, their series included pulmonary nodules, pleural masses, and mediastinal lesions. Thus, our series differed in two respects: first, all lesions were biopsied with CT guidance, and second, only pulmonary nodules were included. The reason for the lower accuracy of CTguided percutaneous needle aspiration biopsy obtain representative tissue from a small lesion than from a large one. Aspirates may more readily be obtained from different sites within larger lesions. A second factor is the increased technical difficulty of carrying out biopsies of small nodules, particularly in cases complicated by pneumothorax. When a pneumothorax occurs during CT-guided biopsy, the initial localization may become inaccurate, making relocalization more difficult [5], especially with small nodules. Of particular interest in our series was the high rate of false-negative diagnoses among patients with small nodules (19%). Among patients with large nodules, that rate was 6%. None of the nine patients with falsenegative diagnoses in our series underwent repeat percutaneous needle aspiration biopsy. All patients with false-negative biopsies underwent surgical resection, either videoassisted thoracoscopic surgery or thoracotomy. Another point of interest in our study was the finding that all 12 cases with cytologic findings suspicious of but not diagnostic for malignancy were proven malignant at surgery. Similarly, in the biopsy series of Poe and Tobin [9], Westcou [1], and Conces et al. [1 1], all cases categorized as suspicious for malignancy were proven malignant. The percentage of cases classified as suspicious for malignancy in our series (12%) resembles that in the study of Poe and Tobin (9.5%) but is considerably higher than those in the studies of Westcott (1.5%) and Conces et al. ( 1.8%). The higher percentage of cases classified as suspicious for malignancy in our series may reflect the relative conservatism of the cytopathologists at our institution. The lower percentage of falsepositive diagnoses in our series (0%) than in those of Westcott (4%) and Conces et al. (3.3%) supports this assumption. Our results and those of Poe and Tobin, Westcott, and Conces et al. suggest that a cytologic diagnosis of suspicious for malignancy is highly correlated with proven malignancy at surgery; however, a definitive diagnosis of malignancy, including cell type, remains important in planning appropriate therapy. The pneumothorax rates for patients with small and large nodules in our series were nearly equal (22 and 21%, respectively). These findings are similar to those from a Jereb [12] that showed no difference in the rate or degree of complication for patients with small or large nodules. However, a direct comparison of the two studies is difficult because of differences in biopsy technique and methods of pneumothorax detection. We found no significant difference in chest tube placement rates for patients with small and large nodules; in fact, only two of 97 patients (-2%) required chest tube placement, and both had large nodules. The low chest tube placement rate in our series is likely related to our stringent pneumothorax precautions, including puncture side-down positioning [8]. Thus, on the basis of the acceptable accuracy rate and low complication rate for biopsy of small nodules in our study, our results suggest that CT-guided percutaneous needle aspiration biopsy should be considered a primary diagnostic procedure in the evaluation of small nodules. We did not identify a critical size below which biopsy of a nodule under CT guidance should not be considered; in fact, a true-positive diagnosis was made in all five patients with nodules less than 10 mm in diameter. Our study has two important limitations. First, only patients who underwent percutaneous needle aspiration biopsy and then surgical resection or autopsy were included in our study. Although this study design provided a pathologic gold standard for final diagnoses, it excluded patients with lesions definitively diagnosed as benign by percutaneous needle aspiration biopsy, as well as patients who had biopsy results negative for malignancy but not diagnostic of a specific benign entity and who did not undergo surgical resection of the nodules. Thus, the diagnostic accuracy results are applicable only to malignant lesions. This study design also excluded patients who had percutaneous needle aspiration biopsy results diagnostic for malignancy and who were not surgery candidates, either because of coexisting medical illness or widespread metastatic disease. Second, none of the patients with mitial false-negative biopsies underwent repeat percutaneous needle aspiration biopsy. Therefore, we are unable to assess the value of repeat biopsies for initial false-negative diagnoses. 108 AJR:167, July 1996

Percutaneous Needle Aspiration Biopsy of Lung In conclusion, CT-guided percutaneous needle aspiration biopsy is significantly less accurate for small pulmonary nodules than for large pulmonary nodules, but the complication rates for both are similarly low. References 1. Westcott JL. Direct percutaneous needle aspiration of localized pulmonary lesions: results in 422 patients.radiology 1980;l37:31-35 2. Perlmutt LM. Percutaneous transthoracic needle aspiration: a review. AiR 1989;l52:45l-455 3. Shepard JO. Complications of percutaneous needle aspiration biopsy of the chest: prevention and management. Semin Inrervent Radiol 1994; 11:181-185 4. Westcott JL. Percutaneous transthoracic needle biopsy. Radiology 1988;l69:593-601 5. vansonnenberg E, Casola G. Ho M, et al. Difficult thoracic lesions: CT-guided biopsy experience in 150 cases. Radiology 1988;167:457-461 6. Khouri NE Stitik FP, Erozan YS, et al. Transthoracic needle aspiration biopsy of benign and malignant lung lesions. AiR 1985;144:28l-288 7. Stanley JH. Fish GD, Andriole JG. et al. Lung lesions: cytologic diagnosis by fine-needle biopsy. Radiology 1987;l62:389-391 8. Moore EH, Shepard JO, McLoud TC. Templeton PA, KosiukJP. Positional precautions in needle aspiration lung biopsy. Radiology 1990;175:733-735 9. Poe RH, Tobin RE. Sensitivity and specificity of needle biopsy in lung malignancy. Am Rev RespirDis 1980;l22:725-729 10. Hayata Y, Oho K, Ichiba M, Goya Y, Hayashi T. Percutaneous puncture for cytologic diagnosis: its diagnostic value for small peripheral pulmonary. Acta Cytol 1973;l7:469-475 11. Conces DJ, Schwenk R, Doering PR, GIant MD. Thoracic needle biopsy: improved results using a team approach. Chest 1987;91:813-816 12. Jereb M. The usefulness of needle biopsy in chest lesions of different sizes and locations. Radiology 1980;l34:13-15 AJR:167, July 1996 109