Reports on the effective use of CT-guided automated needle biopsy in diagnosing pulmonary nodules are starting to be seen [1,2,3,4,5]. Because automated needle biopsies provide core specimens from the pulmonary nodules, diagnostic accuracy for malignant lesions by automated needle biopsy in the absence of a pathologist at the time of biopsy, which ranges from 82% to 97%, is comparable to that of aspiration biopsy in the presence of a pathologist [1, 2, 5]. And in cases of benign lesions, diagnostic accuracy of automated needle biopsy is in fact higher than that of aspiration biopsy, ranging from 71% to 100%, and is particularly useful in making specific diagnoses [1, 2, 5]. Moreover, the frequency of pneumothorax encountered when using automated needle biopsy is comparable to that for aspiration biopsy (9-54%) [1,2,3,4,5]. Many reports have discussed factors associated with pneumothorax in lung biopsies [6,7,8,9,10], but few investigations have dealt with the factors determining diagnostic accuracy [3, 11]. Hence, this study was designed to evaluate the factors influencing the diagnostic accuracy of CT-guided automated needle biopsy of the lung.

Between July 1995 and June 1998, 138 consecutive CT-guided automated needle biopsies of the lung were performed at our institution. Two biopsies were performed for 12 lesions, and three biopsies for one lesion. The study population included 75 men and 48 women (one woman having two lesions) with a mean age of 66 years (range, 28-83 years). There were 124 lesions (6-10 mm: 13 lesions, 15 biopsies; 11-20 mm: 51 lesions, 57 biopsies; 21-30 mm: 39 lesions, 45 biopsies; 31-50 mm: 15 lesions, 15 biopsies; 51-70 mm: six lesions, six biopsies). The mean lesion diameter (± SD) was 23.0 mm (± 12.0 mm; range, 6-70 mm). All biopsies were performed on request from pulmonary physicians or surgeons, and all lesions were scanned using routine CT at a thickness of 1 cm (ProSeed Accel; General Electric Yokogawa Medical Systems, Tokyo, Japan) before the automated needle biopsy. Maximum diameters of the lesions were measured on lung window settings. Pulmonary nodules less than or equal to 3.0 cm in diameter on routine CT were measured by further imaging at slice thicknesses of 2 mm, and pulmonary nodules greater than 3.1 cm in diameter were measured by imaging at slice thicknesses of 5 mm. Lesion depth, from the pleural surface and from the skin surface to the edge of the lesion, was calculated from CT images obtained immediately before automated needle biopsy.

All biopsies were performed by one experienced chest radiologist or other radiologists under his supervision. Before each biopsy, the pulmonary physician or surgeon requesting the automated needle biopsy explained the necessity, the method, and the risk of this examination to patients. Immediately before biopsy, a chest radiologist explained to patients the importance of patient cooperation; all patients were trained several times in regular breathing and breath-holding before the biopsy. Patients were placed in either a prone, a supine, or a lateral decubitus position, allowing penetration of the lesion from a position closest to the body surface. At the time of biopsy, images were obtained at 5-mm section thickness throughout the lesion. Localization was determined by CT imaging with laser lights and markers on the skin. All biopsies were performed using a detachable 18-gauge coaxial automated cutting needle biopsy system (ASAP Detachable; Medi-Tech/Boston Scientific, Watertown, MA). The depths from the skin to the pleura and from the pleura to the edge of the lesion were calculated from CT images. First, the introducer stylet-outer cannula combination was advanced to the calculated depth of the pleura. At this point, the positions of the needle tip and the lesion were confirmed using CT images. Next, the introducer stylet-outer cannula combination was advanced to the depth of the lesion. The position of the needle tip and the lesion was again checked on CT. After CT confirmation of adequate needle-tip position, the introducer stylet was removed from the outer cannula, and the inner central notched stylet was inserted in the outer cannula. The central notched stylet is detachable; the length of the notch measures 17 mm. Subsequently, the automated biopsy system was fired. The inner stylet containing the core biopsy specimen in its notch was removed, leaving the outer cannula in position. The introducer stylet was then reinserted in the outer cannula. Decisions to perform additional needle passes were based on visual inspection of the adequacy of the specimen and on confirmation of the positions of the needle tip and the lesion from CT images obtained immediately after removal of the initial specimen. Core specimens were submitted in 10% formalin for pathologic examination and diagnosis by several pathologists. Frozen section analyses were not performed at the time of biopsy.

All 138 CT biopsies provided histologic material. A positive automated needle biopsy result (including suspected malignancy) was considered true-positive when there was surgical confirmation, when biopsy of another site revealed cancer with the same histologic characteristics, or when the lesion increased in size and other proven metastases were found. A negative automated needle biopsy result was considered true-negative when there was surgical confirmation, when the lesion subsequently disappeared or decreased in size with or without administration of antibiotics, or when the lesion remained stable on follow-up CT for at least 18 months. Follow-up CT for cases with negative findings was scheduled at 3 and 6 months after the biopsy, and every 6 months thereafter, during which changes in lesion size were assessed. A positive automated needle biopsy result (including suspected malignancy) was considered false-positive if there was surgical resection with a benign diagnosis, if the lesion subsequently disappeared or decreased in size, if the second biopsy had negative findings, or if the lesion remained stable on follow-up CT for at least 18 months in the absence of extrapulmonary primary neoplasm or metastasis. A negative automated needle biopsy result was considered false-negative if there was surgical resection with a malignancy diagnosis, if the lesion increased in size, or if other proven metastases were found.

The length of biopsy procedures was calculated by recording the difference between the clock reading on the annotated CT image obtained with the first scout view (considered the starting time) and the time on the CT image obtained through the chest after removal of the guide needle (considered the ending time).

Pulmonary function tests were performed in 76 patients who were subject to thoracic surgical resection.

Automated needle biopsies were usually performed as outpatient procedures. However, patients living in areas without access to facilities capable of responding immediately to pneumothorax after biopsy, and those requesting admission for the procedure, were admitted to the hospital overnight. After the biopsy procedure, postbiopsy pneumothorax was routinely searched for on chest radiographs obtained 1 hr after and the morning after the procedure. Additional chest radiographs were obtained in patients with respiratory symptoms after biopsy. Patients with progressive respiratory symptoms or enlarging pneumothorax on sequential chest radiographs were treated with placement of a chest tube by a skilled thoracic surgeon. Patients with slight respiratory symptoms or slight pneumothorax that did not enlarge on sequential chest radiographs were treated conservatively.

Initial false-positive and false-negative results from lesions for which two or three biopsies were required were used to calculate the overall accuracy. The relationships between accuracy in diagnosis, sex, and lesion size were analyzed using the chi-square test. The relationships between accuracy in diagnosis and other quantitative variables were analyzed using the Student's t test. Significant difference was considered to be a p value of less than 0.05.

The 81 lesions diagnosed as malignant (Table 1) included 68 lesions given final diagnoses of malignancy through surgical resection, three lesions as metastases from primary malignant tumors in other organs, and 10 lesions through confirmed progression or by metastasis to other organs. Among the 43 lesions diagnosed as benign, eight lesions were confirmed through surgical resection, 33 lesions by lesion diminution or disappearance, and two lesions by stabilization for at least 18 months.

The mean diameter of all 124 lesions (± SD) was 23.0 ± 12.0 mm (range, 6-70 mm). Lesion depth from pleura averaged 16.1 ± 18.1 mm (range, 0-113 mm). Lesion depth from skin averaged 51.2 ± 20.9 mm (range, 4-126 mm). The mean procedure time was 22.7 ± 9.1 min (range, 8-60 min). The mean number of needle passes was 1.4 ± 0.5 (range, 1-2 passes). The mean forced expiratory volume (1) in 1 sec (FEV₁) and FEV₁ (%) were 2.1 ± 0.5 1 (range, 1.0-3.6 1) and 84.0 ± 14.4% (range, 52-161%), respectively. The mean FVC (1) and FVC (%) were 2.8 ± 0.7 1 (range, 1.7-4.7 1) and 99.9 ± 14.7% (range, 65-141%), respectively.

In the total 138 CT biopsy procedures in 123 patients (124 lesions), 91 biopsies were on lesions confirmed malignant in the final diagnosis, and 47 biopsies were on benign lesions (70 true-positive, 44 true-negative, three false-positive and 21 false-negative). Overall accuracy was 82.6%, sensitivity was 76.9%, specificity was 93.6%, positive predictive value was 95.9%, and negative predictive value was 67.7% (Table 2).

One of the three false-positive biopsy cases was diagnosed as granuloma in the first biopsy but was subsequently diagnosed as a suspected malignancy in the second biopsy. Results from bacterial cultivation performed from specimens collected in the second biopsy were defined as atypical Mycobacterium tuberculosis, and the lesion disappeared after 18 months of therapy. The second case was diagnosed as suspected malignancy in the first biopsy but was diagnosed as sclerosing hemangioma in the second biopsy. The lesion size remained stable and other metastatic lesions were not found on follow-up CT for more than 27 months, from which the lesion was considered benign. In the third false-positive biopsy case, surgical resection was performed after a diagnosis of leiomyosarcoma from the first biopsy; the lesion was later confirmed in the final postoperative diagnosis to have been leiomyoma. Among the 21 false-negative biopsy cases, two were diagnoses of hyperplasia of alveolar cells; seven were fibrous tissue; and others were scar, necrosis, and inflammation without malignancy. Among the 78 cases of bronchogenic carcinoma, histologic subtype was diagnosed in 77 (98.7%). Among the 43 lesions diagnosed as benign histologically, specific diagnoses were obtained in 28 cases (65.1%).

Repeated biopsies were performed for 13 lesions (12 lesions underwent two biopsies and one lesion underwent three biopsies). In the first biopsy, 10 of these lesions were found to be false-negative, two to be true-negative, and one to be false-positive. A second biopsy was indicated for one lesion that was false-positive at first biopsy from CT images that appeared benign. The other 12 lesions were subject to a second biopsy (including a third biopsy for one lesion) for suspected malignancy, also captured on CT. Five of 10 lesions that were false-negative at first biopsy became true-positive on the second biopsy (or the third biopsy for one lesion). One lesion that was false-positive at first biopsy became true-negative at second biopsy. Therefore, six (54.5%) of 11 lesions in which the results from the first biopsy were inaccurate were successfully diagnosed by two or more repeated biopsies.

We classified the cases into a “success” group (true-positive plus true-negative) and a “failure” group (false-positive plus false-negative) to evaluate the factors affecting diagnostic accuracy in CT-guided automated needle biopsy of the lung. Sex, patient age, lesion size, lesion depth from pleura, lesion depth from skin, length of biopsy procedure, needle passes, and pulmonary function tests were analyzed (Table 3). A significant difference in diagnostic accuracy was found for lesion size (p = 0.014). The mean lesion size in the success group (24.1 mm) was greater than that of the failure group (17.6 mm). There were no statistically significant differences between the other factors and diagnostic accuracy.

Diagnostic accuracy was then further examined in terms of the significant factor, lesion size. The accuracy rate was 100% for lesions 51-70 mm in diameter, with accuracy decreasing for smaller lesions, to 66.7% for lesions 6-10 mm in diameter (Table 4).

Postbiopsy pneumothorax occurred in 31 (22.5%) of the 138 needle biopsy procedures. In four cases (2.9%), chest tube placement was necessary. Postbiopsy hemoptysis occurred in three cases (2.2%), and pulmonary bleeding occurred in seven (5.1%). Hemoptysis was estimated to be less than 15 ml of blood in all cases, and none required specific treatment.

CT is superior to conventional radiography for detecting small pulmonary nodules. Once a lesion is found, the most important point is determining whether the lesion is malignant or benign. Even when a lesion appears to be either malignant or benign on radiography, histologic specimens or bacteriologic examinations should confirm the final diagnosis. In this respect, CT-guided automated needle biopsy, which obtains core specimens from the lesions, is an excellent method for confirming the diagnosis of pulmonary nodules. However, studies to date have not evaluated in detail the factors determining diagnostic accuracy in CT-guided automated needle biopsy of the lung. Thus, this study was undertaken to evaluate the factors related to diagnostic accuracy of this procedure.

In our study, lesion diameter was a significant factor for diagnostic accuracy in CT-guided automated needle biopsy of the lung. Diagnostic accuracy decreased in proportion with a decrease in lesion size.

Although many reports describe CT-guided automated needle biopsy as an accurate means of diagnosing pulmonary nodules, accuracy declines with a decrease in lesion size, making definitive diagnosis difficult. In a series of CT-guided aspiration biopsies (including pulmonary, pleural, and mediastinal lesions), van-Sonnenberg et al. [11] reported diagnostic accuracies of 90.0% for lesions 3.1-4.0 cm in diameter, 89.3% for lesions 2.1-3.0 cm in diameter, 83.9% for lesions 1.1-2.0 cm in diameter, and 73.9% for lesions 0.3-1.0 cm in diameter, recognizing a decrease in diagnostic accuracy accompanying a decrease in lesion size. In a study of diagnostic accuracy of CT-guided percutaneous needle aspiration biopsy classifying lesions into two groups by size, Li et al. [12] reported that the accuracies were 74% for lesions less than or equal to 1.5 cm in diameter and 96% for lesions greater than 1.5 cm in diameter, a statistically significant difference. We also found a significant difference in accuracy between the 67.6% (true-positive plus true-negative / number of biopsies: 25/37) for lesions less than or equal to 1.5 cm in diameter and 88.1% (89/101) for lesions greater than 1.5 cm in diameter (chi-square test, p < 0.01). Lucidarme et al. [5] (in CT-guided automated needle biopsy) reported no significant difference in accuracies of 81% for lesions less than or equal to 2.0 cm in diameter and 91% for lesions greater than 2.0 cm in diameter. Contrary to their findings, applying their 2-cm cutoff point to our study yielded a significant difference between accuracies of 76.4% (55/72) for lesions less than or equal to 2.0 cm in diameter and 89.4% (59/66) for those greater than 2.0 cm in diameter (chi-square test, p < 0.05). In studies of CT-guided automated needle biopsy in the absence of a trained cytopathologist similar to our study, Klein et al. [1] reported mean lesion size and diagnostic accuracy to be 2.9 cm and 88%, respectively; Haramati [2] reported 4 cm and 81%; and Lucidarme et al. reported 33.6 mm and 88%. One reason that may be ascribed to lower accuracy in CT-guided automated needle biopsy as the lesions become smaller is, as pointed out by Li et al. [12], sampling error. In other words, the smaller the lesion, the more difficult it becomes for the biopsy needle to hit the lesion, making it increasingly likely for specimens to be obtained from the periphery of the lesion rather than from the lesion itself, or for specimens to be inadequate in volume.

Previous studies have reported that diagnostic accuracy increases with an increase in the number of needle passes through the lesion. In the presence of a cytopathologist during biopsy procedures, Westcott et al. [13] found that additional passes performed on the same day increased diagnostic accuracy in some cases. Lucidarme et al. [5] reported that overall accuracy was 88% when the mean number of needle passes was 2.5, but that if only a single core specimen had been obtained routinely, the overall accuracy would have been 83%. In our study, the mean number of needle passes was 1.4 (maximum = 2), and no significance was seen between diagnostic accuracy and the number of needle passes. However, we believe that increasing the number of needle passes may have increased accuracy in our study, particularly for lesions less than or equal to 20 mm in diameter, for which a significant decrease was noted in diagnostic accuracy. As reported by Khouri et al. [14] and Williams et al. [15], we also confirmed that higher accuracy could be obtained by second biopsies. In our study, repeated biopsies on another day were conclusive for six (54.5%) of 11 lesions in which accurate diagnoses were not obtained in the first biopsy.

Li et al. [12] pointed out that a pneumothorax occurring before the needle pass of the lesion decreases diagnostic accuracy. This is because pneumothorax with partial collapse in the lung occurring at the time the needle passes the pleura displaces the lesion from the point of initial localization, making it more difficult to obtain specimens from the lesion, particularly when the lesion is small. However, in our study, there were only two cases in which pneumothorax large enough to displace the lesion occurred before needle pass. One case was true-positive and the other was false-negative, and we considered pneumothorax before needle pass to be an insignificant factor, given its low incidence. Some reports describe correlation between the prevalence of pneumothorax and pulmonary function test results [7,8,9,10]. However, our study showed no correlation between diagnostic accuracy and the pulmonary functions. Regarding other possible factors determining diagnostic accuracy, we initially expected a decrease in accuracy in proportion to the depth of the lesion from the pleura or skin surface. However, no significant difference was noted between the success and failure groups. As for age, although the mean age of the failure group was older by 4 years than the age of the success group, no significant correlation was found between the two groups.

Pulmonary nodules move with respiration. Patient cooperation is thus indispensable in performing conventional CT needle biopsies. Faint movement or unstable breath-holding during the biopsy renders the initial localization of the lesion inaccurate, making the needle biopsy more difficult, particularly with small lesions. If the lesion is positioned under the rib, needle passes can become difficult without controlled respiration by the patient. In this regard, as described by Moore [16], we also recognize patient cooperation to be one of the most important factors necessary for a successful procedure. For this reason, we fully explain the biopsy procedure and possible complications to our patients and have them practice breath-holding and regular respiration before biopsy.

Conventional CT-guided needle biopsies of the lung do not allow real-time visualization of the needle tip or the site of the lesion. This lack of real-time imaging capability is one reason for a decrease in diagnostic accuracy. Recently, real-time CT (i.e., CT fluoroscopy) was developed to overcome the limitations of conventional CT. Katada et al. [17] reported diagnostic accuracy for aspiration biopsy with CT fluoroscopy to be 100% for lesions greater than 11 mm in diameter, but 67% for those less than or equal to 10 mm in diameter, which is comparable to our result of 66.7% accuracy for small lesions biopsied using conventional CT. These figures show the difficulty of obtaining an accurate diagnosis in lesions less than or equal to 10 mm, even in CT fluoroscopy-guided needle biopsies. Moreover, radiation exposure to the operator's hand is a problem with the present CT fluoroscopy system.

In conclusion, examination of diagnostic accuracy in CT-guided automated needle biopsy of the lung revealed lesion size to be a significant factor in determining diagnostic accuracy, which decreased in proportion to lesion size. Although CT fluoroscopy enabling real-time imaging has the potential of increasing diagnostic accuracy, conventional CT guidance can achieve comparable results for lesions less than or equal to 10 mm. Our results highlight the difficulty of diagnosing small lesions, even with sufficient patient cooperation.

Address correspondence to H. Tsukada.