INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Bronchogenic carcinoma (CA) is the most common cause of cancer mortality in Western society and its incidence continues to rise (1). Because pulmonary resection is the only curative therapy available to patients with localized disease, it is important that these individuals be selected accurately and in the least invasive manner possible. The resectability of CA can be determined by mediastinoscopy, mediastinotomy, and/or thoracotomy. To avoid the morbidity, mortality, and cost of these procedures, transbronchial needle aspiration (TBNA) was adapted for the flexible bronchoscope (FB) in 1983 (2). The contribution of computed tomography (CT) to CA staging by TBNA was further emphasized by Wang who proposed bronchoscopic lymph node (LN) mapping to facilitate the process (3). We organized a multi-institutional study to more precisely determine the role of CT scanning with TBNA, using specific LN locations (both mediastinal and hilar) according to the recently proposed classification, and to identify the clinical and radiographic predictors of a positive aspirate.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Patients presenting for evaluation of suspected CA at the study centers (Easter Maine Medical Center and St. Joseph Hospital, Bangor, Maine; Cleveland Clinic, Cleveland, Ohio; Bellevue Hospital, New York University Medical Center, New York, New York; and The Regional Hospital, Ancona, Italy) between January 1, 1995 and December 31, 1996 were studied prospectively. Individuals ultimately found not to have CA, those with thoracic metastatic disease from a nonpulmonary primary site, and patients with a positive TBNA from an enlarged LN confluent with the tumor mass on CT were excluded from the study.

All patients had a standard chest radiograph to determine location of the tumor mass. CT scanning was conducted with contrast enhancement either with 10-mm sections from the thoracic inlet to the carina and inferior pulmonary veins to the diaphragm, with 5-mm sections from the carina to the level of the pulmonary veins, or with 7-mm sections on a spiral scanner. LN size was defined by measuring the short-axis diameter and was determined by a radiologist unfamiliar with the patient's clinical status. Size gradation was measured in 5-mm increments: Grade (Gr) 0: < 5 mm; Gr 1+: 5 to 9 mm; Gr 2+: 10 to 14 mm; Gr 3+: 15 to 19 mm; Gr 4+: 20 to 24 mm; Gr 5+: 25 to 29 mm; and Gr 6+: > 30 mm. If the tumor mass was confluent with the adjacent nodal group that was sampled, the results were excluded from statistical analysis. Location of the LN groups was determined after careful CT review as proposed by Wang: Ianterior carina; IIposterior carina; IIIright paratracheal; IVleft paratracheal; V right main bronchus; VIleft main bronchus; VIIright upper hilar; VIIIsubcarina; IXright lower hilar; Xsub subcarina; XIleft hilar (Figure 1) (3). Stations I-IV and VIII were considered mediastinal locations; VII, IX, and XI were hilar positions.

View larger version (119K): [in this window] [in a new window]   Figure 1.   Nomenclature for mediastinal and hilar lymph node locations used for transbronchial needle aspiration: Ao = aorta; PA = pulmonary artery. 1: Anterior carina. 2: Posterior carina. 3: Right paratracheal. 4: Left paratracheal (anteroposterior window). 5: Right main bronchus. 6: Left main bronchus. 7: Right upper hilar. 8: Subcarina. 9: Right lower hilar. 10: Sub subcarina. 11: Left hilar.

Flexible bronchoscopy with TBNA was conducted by standard techniques under conscious sedation by the investigators or their trained colleagues. Care was taken to perform TBNA before any distal airway examination or specimen collection. Suctioning was avoided to prevent inadvertent aspiration of respiratory secretions contaminated with exfoliated malignant cells. TBNA was performed with either no. 21 or no. 22 cytology needles or no. 19 gauge histology needles (Mill-Rose, Mentor, OH) at LN sites identified on chest CT at the discretion of the examining bronchoscopist as previously described (2). The Bangor investigator used a no. 21 (SW-121) in 167 patients, a no. 22 (SW-122) in 34 patients, a no. 19 (MW-319) in seven cases, and a no. 22 (SW-122) plus a no. 19 (MW-319) in three individuals. At the Cleveland Clinic, a no. 22 (SW-122) alone was employed in 20, a no. 19 (MW-319) alone in 38 patients, 35 had both no. 22 (MW-222) and no. 19 (MW-319). At New York University a no. 21 (SW-121) was used in two and no. 19 (MW-319) was used in 36. In Ancona, a no. 21 (SW-121) was exclusively employed in all of their 18 cases. Two to three aspirates were obtained from each LN site sampled. On-site cytologic analysis of the aspirates was not available at the time of the bronchoscopic examination. The bronchoscopic location of visible endobronchial lesions was identified. If tumor was present in more than one area, it was recorded as being situated in the most proximal location of the tracheobronchial tree.

Patients with a nondiagnostic bronchoscopic examination (including TBNA) underwent additional appropriate studies (e.g., thoracentesis, percutaneous needle aspiration, mediastinoscopy, mediastinotomy, video-assisted or standard thoracotomy) to establish a definitive diagnosis. Patients with limited stage non-small cell carcinoma (NSCC) had curative lung resection performed at the time of thoracotomy. All patients undergoing surgical mediastinal staging or thoracotomy had LN mapping performed according to the American Joint Cancer Committee (AJCC) classification. For comparative purposes the Wang bronchoscopic LN system was equated to the AJCC standard as follows: Ino equivalent, IIno equivalent, III4R, IV4L, V10R, VI10L, VII11R, VIII-7, IX11R, Xno equivalent, XI11L.

A cytologic specimen was considered a "true" positive only when it had large numbers of malignant cells mixed with lymphocytes based on the rare incidence of false-positives (4). When the histology needle was used (MW-319), aspirations were considered positive if either a histologic core of tissue or a cytologic sample obtained with the histology needle was reported to show malignant features. All specimens reported to have "suspicious," "few," or "rare" malignant cells were considered negative.

A negative TBNA with a negative surgical exploration (mediastinoscopy, mediastinotomy, or video-assisted or standard thoracotomy) was considered a true-negative. Because the majority of patients (287 of 360 [80%]) with a negative TBNA were not considered surgical candidates owing to locally extensive or distant metastatic disease, marginal ventilatory reserve, or severe coexistent medical illness, sensitivity determination was calculated on the basis of both best case (all negative TBNA without mediastinal surgical exploration excluded) and worst case (all negative TBNA without mediastinal surgical exploration considered false-negative) scenarios. Sensitivity was defined as true-positive/true-positive + false-negative, specificity as true-negative/true-negative + false-positive, positive predictive value as true-positive/true-positive + false-positive, and negative predictive value as true-negative/true-negative + false-negative.

Statistical analysis was performed with the use of Statistical Analysis System Software (SAS Institute, Cary, NC). Simple proportions were used to describe data on the type of lung cancer and how the diagnosis was established. Data were analyzed using the Mantel-Haenszel chi-square test except for cell sizes < 5 where the Fisher exact test was employed. A p value at < 0.05 indicated statistical significance (8). These tests were used to assess positivity of small cell carcinoma (SCC) and NSCC, the correlation of bronchoscopic tumor locations (Table 1) and radiographic tumor locations (Table 2). Correlation of LN size with TBNA positivity was determined for both SCC and NSCC using a linear regression line with a corresponding 95% confidence interval.

An attempt was made to use multivariate stepwise logistic regression analysis to relate LN size, study center, tumor laterality, TBNA needle size, tumor cell type, and LN location to the presence of a positive aspirate. Because this study was a clinical experiential trial rather than one with a rigid LN sampling protocol, there were an insufficient number of patients with identical data points to make this approach feasible. To circumvent this difficulty and examine these variables, logistic models were created from the four most important clinical sites (Stations I, III, IV, and VIII). In this way, we were able to confirm the most significant factors predictive of TBNA positivity.

The protocol for this study was approved by the institutional review board for human experimentation at Eastern Maine Medical Center, and written consent was obtained from all study participants.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

A total of 781 patients were initially bronchoscoped of whom 365 had CA (Table 3). Five had tumor confluent with the LN sampled by TBNA and were excluded from the analysis because of concern that the malignant tissue rather than the LN could have been exclusively or additionally sampled. The remaining 360, of whom 81 had SCC and 279 had NSCC, constituted the study population. CA was established by standard bronchoscopic techniques (brush, biopsy, and/or lavage) in 227 (63%), and was exclusively diagnostic of CA by TBNA (all other bronchoscopic samples were negative for malignancy) in 65 (18%). In the remainder of patients percutaneous needle aspiration in 33 (9%), thoracotomy in 19 (5%), mediastinoscopy in 9 (2.5%), video-assisted thoracoscopy in 2 (0.6%), sputum cytology in 2 (0.6%), supraclavicular lymph node biopsy in 1 (0.3%), median sternotomy in 1 (0.3%), and mediastinotomy in 1 (0.3%) established proof of malignancy.

TBNA was positive in 50 of 81 (62%) patients with SCC versus 135 of 279 (48%) individuals with NSCC (p = 0.034). TBNA was exclusively diagnostic in 19 of 81 (23%) patients with SCC and 46 of 279 (16%) individuals with NSCC (p = 0.15). In addition, TBNA proved mediastinal tumor extension in 39 other NSCC patients who were otherwise operable. It therefore precluded additional diagnostic surgery in a total of 104 of 360 (29%) patients.

TBNA was performed at 607 individual LN locations and 263 (43%) were positive. There was no difference in the diagnostic yield from mediastinal (225 of 529 [43%]) or hilar (27 of 59 [46%]) locations. Patients with SCC had positive aspirates from individual LN locations in 71 of 131 (54%) attempts compared with individuals with NSCC who had a positive aspiration in 192 of 476 (40%) attempts. This difference was significant (p = 0.005) and was related to the significance of positive mediastinal (63 of 112 [56%] SCC versus 162 of 407 [40%] NSCC) versus hilar (5 of 10 [50%] SCC versus 22 of 49 [45%] NSCC) aspirates.

Correlation of positive aspirates showed a linear relationship with LN size ranging from 5 to 9 mm (Gr 1+) up to 20 to 24 mm (Gr 4+) in size for both SCC and NSCC (Figure 2). Although none of the 54 cytologic aspirates from LN < 5 mm was positive, 15 of 103 (15%) samples from LN 5 to 9 mm (Gr 1+) were diagnostic of malignancy.

View larger version (37K): [in this window] [in a new window]   Figure 2.   TBNA of lymph nodes in small cell carcinoma (SCC) and non-small cell carcinoma (NSCC). Malignant aspirates were more frequent in SCC than NSCC and positivity increased in each among LN < 25 mm. All LN aspirates < 5 mm were negative.

Additional analysis of each of the LN groups indicated that among LN of equivalent size, right paratracheal and subcarinal LN aspirates were more likely to provide a positive cytology (Table 4). Needle size also appeared to have a significant influence on the frequency of positive TBNA recovery (p < 0.001). Aspirates were more likely to be positive with a histology needle (61 of 118 [57%]) than with a cytology needle (228 of 552 [41%]); however, none of the seven histologic aspirates from LN < 10 mm was positive (Figure 3).

View larger version (40K): [in this window] [in a new window]   Figure 3.   Percentage of positive TBNA among LN of increasing size. Histology aspirates in LN < 10 mm and cytology aspirates in LN < 5 mm were negative.

Correlation of TBNA results with both bronchoscopic and radiographic locations indicated that right-sided tumors were more often associated with a positive aspirate than those originating from the left lung (p = 0.002 to 0.01). These observations were statistically significant at upper and middle but not lower lobe locations (Tables 1 and 2). Because of low sample size, significant differences could not be confirmed from hilar aspirates. Multivariate analysis of aspirates from paratracheal and subcarinal LN sites indicated that LN size (p < 0.001) and right-sided tumor location (p = 0.02 to 0.05) were the two independent predictors of significance.

Because 273 of 360 (76%) patients were inoperable at the time of initial diagnosis, 259 of 344 (75%) negative aspirates could not be confirmed by direct surgical biopsy techniques. The overall sensitivity of TBNA was 70% in the best case scenario and 53% in the worst case scenario among all 607 aspirates performed. By limiting the analysis to LN  10 mm in size, sensitivity improved to 91% in the best case and 57% in the worst case and for LN  15 mm sensitivity was 95% and 62%, respectively.

Eighty-seven of 360 (24%) patients had mediastinal exploration. Among these, 141 TBNA aspirates at selected LN stations were obtained and 85 of them could be directly compared with the same lymph node locations examined surgically. There was one false-positive (specificity 99%) and 67 of 85 samples were true-negatives (negative predictive value 80%). By including the results of all positive mediastinal aspirates (225 positive TBNA + 17 false-negative TBNA), the sensitivity was 93% (positive predictive value 99%).

Utilization of TBNA at the four study centers is outlined in Table 5. Among the four institutions, there was considerable variability in the number of cases submitted, the size of LN sampled, the TBNA needle size employed, and the number and experience of the bronchoscopists performing the technique (Table 5). Even when comparing LN of equivalent size using a cytology needle, Cleveland examiners had a higher positivity rate than their Bangor counterpart (LN 2+, 4+, 6+: Bangor 17%, 64%, 71% versus Cleveland 46%, 84%, 84%). When comparing histologic needle use, New York and Cleveland practitioners had similar results (LN 2+, 4+, 6+: New York 80%, 75%, 88% versus Cleveland 64%, 82%, 94%).

Complications occurred in six of 360 (1.7%) patients. Bleeding occurred in four, in two of whom the efficacy of the bronchoscopic examination was felt to be compromised. No special therapy was necessary. One patient experienced pneumomediastinum and another had hypoxemia, both of which resolved with observation alone.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Bronchogenic carcinoma has now become a major public health problem in developed countries and is the leading cause of cancer mortality (1). While surgical excision remains the best hope for cure, appropriate pathological staging is necessary because most patients have mediastinal extension or distant metastatic disease at the time of initial diagnosis (9).

CT not only evaluates the extent of parenchymal disease, but is particularly advantageous in identifying hilar and mediastinal lymphadenopathy. Although the sensitivity of this radiographic technique is high, the specificity is low (10). TBNA, however, is very specific and reports of false-positives are rare (4, 5). Appropriate attention to proper technique can largely obviate this difficulty (2, 13). The complementary nature of these two diagnostic modalities, therefore, can greatly facilitate the staging of CA (3). Nonetheless, TBNA, adapted for FB in 1983, remains relatively underutilized. In fact, a survey by Prakash and coworkers indicated that only 12% of North American bronchoscopists routinely use TBNA and 29% use it only occasionally when evaluating thoracic malignancies (14). Reasons for this include operator inexperience, low yields, additional procedure time, FB damage, and concern regarding great vessel puncture.

This study was a prospective multi-institutional investigation involving both private and teaching hospitals and included bronchoscopists with a wide spectrum of experience, some of whom were still in training. The results, therefore, should be applicable to the majority of practitioners evaluating patients with CA. The findings in this study have not only confirmed, but also extended a number of observations noted by previous investigators. Whereas TBNA was positive in 62% of patients with SCC and 48% of patients with NSCC, it was able to exclusively establish the diagnosis 18% of cases. This latter value is higher than reported in previous series and may represent not only greater familiarity with the technique, but more importantly the value of preoperative CT to enhance the accuracy of the procedure (15). The finding that LN of equivalent size are more likely to be positive in patients with SCC than NSCC is noteworthy and may reflect either the biologic aggressiveness of SCC and/or conceivably decreased cellular adherence of these tumor cells.

Correlation with LN size showed progressively higher yields with increasing size up to 24 mm. Thereafter, larger sized LN did not yield a higher rate of aspirate positivity. In a series of CA patients reported by McCloud and coworkers, seven of 19 LN (37%) measuring 2 to 4 cm in patients with lung cancer were found to be free of malignant disease; the cause of lymphadenopathy being postobstructive pneumonia or previous granulomatous infection (19). We believe that very large LN may contain necrotic tumor which could also explain a negative TBNA in these cases. Interestingly, our study demonstrated that 15 of 103 (15%) small LN (5 to 9 mm) contained metastases. In this situation, however, TBNA cannot distinguish intra- versus extranodal disease so that a positive aspirate cannot categorically preclude surgical resection. In fact, upfront chemotherapy followed by surgery has achieved enhanced survival in a number of patients with Stage IIIa disease (20). Nonetheless, a study examining survival among patients evaluated by TBNA and controlled for an equivalent level of surgical staging suggested a poorer prognosis in those with positive mediastinal aspirates (23).

In a large retrospective study, TBNA of mediastinal aspirates were more likely to be positive in patients with tumors bronchoscopically localized to the right upper lobe (RUL) than those in the left upper lobe (LUL) (24). This study extends those observations, indicating that right-sided lesions identified by either FB or chest roentgenography are more likely to have a positive mediastinal aspirate than those originating from the left lung. The most likely explanation is the more extensive lymphatic network present in the right hemithorax (25). Whether laterality of CA has prognostic significance is uncertain; however, one study suggested that elderly patients with left-sided lesions may have enhanced survival (29).

Mediastinoscopy and mediastinotomy have a very high degree of sensitivity (87%) and specificity (100%) (30). These methods require general anesthesia, are expensive, have a morbidity of 1% and mortality of 0.2%. Schenk has reported his experience with a histology needle in which TBNA sensitivity is comparable to direct mediastinal exploration (6). Whereas the sensitivity in our series with LN  10 mm was 57% in the worst case scenario with a negative predictive value of 80%, it was 91% in the best case (positive predictive value 99%).

Comparing individual LN groups identified by the proposed bronchoscopic staging system at both hilar and mediastinal locations revealed that among LN  10 mm, Stations III and VIII (right paratrachea and subcarina) were most likely to yield a positive aspirate. Anatomic studies have previously demonstrated that subcarinal LN serve as a conduit of lingular and left lower lobe (LLL) lymphatics as well as the lymphatic drainage from the right lung to the right paratracheal chain (25). The prognostic significance of these infracarinal LN has been emphasized by Naruke (31). Because the aortic knob protrudes over the access site to the left paratracheal LN (Station IV), difficulty in accessing this location may explain the lower yields from this area.

In most cases, a negative aspirate could not be confirmed by direct biopsy techniques. This was because surgical exploration was precluded by: (1) the diagnosis of SCC (81 patients [23%]); (2) TBNA positivity of any mediastinal (N2) location (169 patients [47%]); (3) locally unresectable or distant metastatic disease (42 patients [12%]); or (4) severe coexistent pulmonary or cardiovascular disease (18 patients [5%]). Nonetheless in 85 aspirates from 83 patients, the corresponding LN group was directly examined at surgery. The specificity was 99% and the negative predictive value was 80%. Thus, TBNA proved to be highly accurate in this large series. It is difficult to definitively assess TBNA sensitivity in the particular subset of 83 patients because all but one had mediastinal exploration only after a negative TBNA at these locations, and, therefore the nominator of the fraction expressing sensitivity was zero. If one assumes, however, that all TBNAs were true-positives, the sensitivity would then be 93%.

One of the primary advantages of the bronchoscopic LN staging system is its complementarity with CT scanning which can be used as a roadmap to access well-defined LN locations (3). On-site cytologic analysis may shorten bronchoscopic examination time and reduce morbidity by precluding the need for endoscopic or transbronchial biopsy; however, this has not been proven. Further recent advances with the use of real-time CT imaging offer the potential for achieving additional accuracy with this technique; however, additional technical and personnel expense may preclude its widespread implementation (32, 33).

Stations V and VI (right and left main bronchial LN) are sites of potentially pivotal significance because involvement at these locations while juxtacarinal is not clearly mediastinal (N2) in location. While the number of cases was insufficient to permit definitive assessment at these two sites, in only one of 21 patients (5%) did TBNA sampling of these locations yield clinically significant information that was not otherwise obtained from TBNA of standard N2 sites.

As a clinical study this one has a number of limitations. Although patients were studied prospectively, 76% did not have surgical confirmation of their TBNA results. To do so would have sacrificed patient safety for study design. In addition, aspirate locations were determined by the examining bronchoscopist for similar reasons. It is, however, the only study that has compared identically localized LN locations (rather than pooled specimens from several mediastinal sites) albeit in 85 of a total of 519 mediastinal aspirates. Interinstitutional variability was notable in terms of LN sampled, operator experience, and frequency of use. Although some of the endoscopists were in training, the procedure was always conducted under the supervision of a bronchoscopist fully conversant with the technique. Utilization of the histology needle could only partially explain the differences in aspirate positivity, even when controlling for LN size. The explanation for some of these other inconsistent observations is not readily apparent.

While it is now clear that TBNA is the most operator- dependent of all bronchoscopic modalities, increasing experience and practice have been shown to improve yields (34). This study has demonstrated that the most effective use of TBNA mapping system is to clarify the importance of sampling those thoracic LN which are accessible with a TBNA needle, rather than to replace the standard AJCC classification of thoracic LN staging. TBNA used with CT in routine clinical practice can achieve a diagnostic yield exceeding 57% in the staging of CA, exclusively provide a diagnosis in 18%, and preclude additional thoracic surgery in 29% of patients. This represents a substantial advance in the staging of CA because it precludes operative study in a population known to have significant cardiopulmonary and vascular comorbidity.

The predictors of a positive TBNA are LN of increasing size, right-sided tumor locations, right paratracheal and subcarinal locations, the presence of SCC, and the use of a histology needle. Because complication rates are low (< 2%, none of which were serious) and specificity is high (99%), TBNA based on CT should be considered an integral part of the bronchoscopic examination when evaluating patients for malignant thoracic disease.