Metastasis to the mediastinal lymph nodes occurs in nearly half of all patients with non-small-cell lung cancer. The development of mediastinal metastasis is probably the most frequent deterrent to cure, even when the presentation is localized. If metastasis involves contralateral or large, bulky (>1.5 to 2.0 cm) subcarinal lymph nodes, surgery alone may not be curative [3-9]. As a result of recent revisions to the staging systems for lung cancer, ipsilateral mediastinal and subcarinal lymph node involvement is now classified as potentially resectable, N2 disease; contralateral mediastinal lymph node involvement (N3 disease) precludes resection [10-12].

Computed tomography of the chest is the current method by which mediastinal lymphadenopathy is detected in non-small-cell lung cancer. However, its sensitivity for detection of metastasis to the lymph nodes varies; accuracy in previous studies has ranged from 52% to 88% [13-23]. This inconsistency has been attributed to the variable correlation of lymph node size with the presence of malignancy. When enlarged contralateral or ipsilateral mediastinal lymph nodes are seen on computed tomography of the chest, standard practice is to determine more accurate staging by performing aspiration biopsy of the lymph node with computed tomographic guidance; bronchoscopy; or, less commonly, a transthoracic approach. If these procedures are unsuccessful, open biopsy is performed by using mediastinoscopy or limited thoracotomy [24-26]. If contralateral lymph nodes are positive for malignancy, surgical resection of the primary tumor is contraindicated.

The development of endoscopic ultrasonography has now made it possible to visualize, with high resolution, not only the gastrointestinal tract but also surrounding structures. Endoscopic ultrasonography has been shown to be superior to computed tomography in evaluating lymph nodes for metastases in esophageal, gastric, and pancreatic cancer [27-29]. Promising results for detecting posterior mediastinal lymph nodes in patients with lung cancer suggest a possible role for endoscopic ultrasonography in staging lymph nodes in patients with non-small-cell lung cancer [30-36]. Fine-needle aspiration biopsy guided by endoscopic ultrasonography was recently reported to further improve the accuracy of endoscopic ultrasonography in predicting malignancy of gastrointestinal masses, with rates as high as 87% to 91% [37-42].

We previously reported the results of endoscopic ultrasonography in 17 patients with lung cancer. This method was very accurate for detecting mediastinal lymphadenopathy; the overall accuracy was 71% compared with 41% for computed tomography (P = 0.032) [43]. During the initial study, however, fine-needle aspiration biopsy guided by endoscopic ultrasonography was not available. The goal of the present study was to prospectively evaluate the accuracy of endoscopic ultrasonography alone, endoscopic ultrasonography-guided fine-needle aspiration biopsy, and computed tomography of the chest in detecting mediastinal lymph node metastasis in patients with non-small-cell lung cancer.

Methods

Top Methods Results Discussion Author & Article Info References

Patient Selection

The study sample consisted of all patients presenting to the Indiana University Thoracic Oncology Program between July 1993 and June 1995 with a diagnosis of non-small-cell lung cancer. The study was approved by the institutional review board, and all enrolled patients gave informed consent. Patients were excluded if they had documented unresectable disease (that is, distant metastasis or locally advanced staged disease [stage III b]) as shown on computed tomography of the chest or if they had a serious medical illness and a life expectancy of less than 1 year. All patients underwent initial preoperative intravenous contrast-enhanced computed tomography of the chest; the axial technique was used in 60% of patients, and the spiral technique was used in 40%. Computed tomography was done at the referring hospital or Indiana University Medical Center; the scans were read at Indiana University Medical Center by a senior attending radiologist who has recognized expertise in this area and used currently accepted radiographic techniques to stage the tumor. The radiologist's determination of the benign or malignant nature of each lymph node was recorded on a preoperative computed tomography lymph node map; the American Thoracic Society mediastinal staging map (Figure 1) was used to describe the location of each node [12]. Any patient who had a questionably enlarged mediastinal lymph node (>1 cm in diameter) and was considered a surgical candidate was then scheduled for endoscopic ultrasonography.

View larger version (59K): [in this window] [in a new window]   Figure 1. American Thoracic Society scheme for mapping mediastinal lymphadenopathy by anatomic location, as seen from behind with endoscopic ultrasonography. Ao = aorta; D = diaphragm; E = esophagus; inf VC = inferior vena cava; L = left; I.PA = left pulmonary artery; PV = pulmonary vein; R = right; r.PA = right pulmonary artery;VC = superior vena cava. Numbers refer to specific mediastinal zones.

Endoscopic Ultrasonography

Endoscopic ultrasonography was performed in an outpatient setting on all patients by one of three experienced endosonographers; the radial scanning echoendoscope (GFUM-20, Olympus America, Melville, New York) or the linear-array scanning echoendoscope (FG32UA, Pentax, Orangeburg, New York) was used for all procedures. When done by an experienced operator, endoscopic ultrasonography is similar to standard upper endoscopy both in technique and duration of the procedure. When fine-needle aspiration biopsy is performed, the procedure is slightly prolonged. Patients were sedated with meperidine and midazolam, the doses of which were titrated to achieve adequate conscious sedation. The instrument was advanced into the stomach, and the celiac axis was imaged. The probe was then withdrawn to the gastroesophageal junction and slowly withdrawn at 1-cm intervals. Images were obtained with 7.5- and 12-MHz frequencies at each interval. All imaged mediastinal lymph nodes were mapped by location according to the American Thoracic Society classification scheme [12]. From these data, an objective assessment was made as to whether the mediastinal lymphadenopathy detected by endoscopic ultrasonography was benign or possibly malignant according to the following previously reported criteria for malignancy: round shape; sharp, distinct borders; hypoechoic texture; and a short-axis diameter greater than 5 mm [36-39]. Malignancy was suspected if all of these criteria were present. All patients who were studied before the availability of endoscopic ultrasonography-guided fine-needle aspiration biopsy underwent surgical resection, and endoscopic ultrasonographic findings were correlated to surgical pathologic findings. Patients who were studied after the advent of fine-needle aspiration biopsy and were found to have no suspicious lymph nodes by endoscopic ultrasonography were directly referred for surgery because there was no indication for aspiration biopsy.

Endoscopic Ultrasonography-Guided Fine-Needle Aspiration Biopsy

Endoscopic ultrasonography-guided aspiration biopsy became available after the first 17 patients were enrolled in our pilot study. All posterior mediastinal lymph nodes that were suspicious for malignant involvement according to the endoscopic ultrasonographic criteria were noted; selected nodes underwent biopsy during the same procedure. Many of the patients had more than one suspicious lymph node. We performed biopsy only on the most suspicious lymph node, which would have the greatest effect on clinical staging (that is, determination of whether the metastasis was contralateral or subcarinal). This technique for ultrasonography-guided aspiration biopsy was initially developed for use with the linear-array instrument (Figure 2) and is described elsewhere [40-42]. We recently reported a similar technique that uses a radial scanning echoendoscope (Figure 3) [37]. Ultrasonography-guided aspiration biopsy involves the insertion of an aspiration catheter needle device through the accessory channel port of the echoendoscope; the needle is then deployed into the lymph node to be sampled under endoscopic ultrasonographic guidance. Aspiration biopsy is done by introducing a specially designed fine-needle aspiration catheter system that consists of a 4-cm long, 23-gauge needle attached to a 180-cm long, 5-French aspiration catheter (Wilson-Cook, Winston-Salem, North Carolina); "in and out" movements of the catheter are used while the operator firmly grasps the catheter at the point at which it enters the accessory port.

View larger version (155K): [in this window] [in a new window]   Figure 2. A mediastinal lymph node as imaged with the linear-array endoscopic ultrasonography system. The needle is exiting the scope; the tip of the needle is in the center of the lymph node (arrow). Just inferior to the lymph node is the pulmonary artery. One advantage of the linear-array instrument is its Doppler capability, which allows precise imaging of regional blood vessels and safe positioning of the needle relative to the target lymph node.

View larger version (193K): [in this window] [in a new window]   Figure 3. Endoscopic ultrasonographic image obtained from the radial scanning instrument showing a large hypoechoic, oval subcarinal lymph node (LN) suspicious for metastatic involvement. The lymph node is adjacent to the aorta (Ao).

Preliminary cytologic findings were obtained during fine-needle aspiration biopsy by a cytopathologist who was present during the procedure. Before the sample was reviewed, Diff-Quik stain (Harleco, Gibbstown, New Jersey) was applied to the slide that contained the deposited specimen. Additional passes were made until a positive cytologic result or a negative result on an adequate tissue sample was obtained [37].

Patients who were considered eligible for surgical resection after staging by computed tomography and endoscopic ultrasonography (that is, patients with a negative result on aspiration biopsy of contralateral or bulky subcarinal lymph nodes or those with nodes that seemed to be benign according to endoscopic ultrasonographic criteria) underwent thoracotomy for pulmonary resection with ipsilateral mediastinal and subcarinal lymph node dissection. During mediastinal dissection, each lymph node was placed in the compartment of a box that was labeled according to the American Thoracic Society lymph node map; the compartment corresponded to the node's location of origin. Histopathologic findings were determined according to the location of the lymph nodes; this allowed exact correlation between results of endoscopic ultrasonography, computed tomography, and surgical pathology. The accuracy of computed tomography, endoscopic ultrasonography alone, and ultrasonography-guided aspiration biopsy of suspicious mediastinal lymph nodes in predicting malignant involvement was then compared with surgical pathologic findings.

Statistical Analysis

The operating characteristics (sensitivity, specificity, and positive and negative predictive values) and false-positive and false-negative rates of endoscopic ultrasonography and computed tomography in predicting nodal metastasis were first determined by comparing the results of these procedures with the results of surgical pathology and cytology of fine-needle aspirates. Confidence intervals were computed for this part of the study by using the "confin" software program (Stat-Exact, Inc., Cambridge, Massachusetts), which calculates exact CIs for proportions. This study group included patients who were evaluated by endoscopic ultrasonography alone and then had surgery (most, including the initial 17 patients from our pilot study, were seen before the development of ultrasonography-guided aspiration biopsy) and patients who were studied with imaging techniques followed by ultrasonography-guided aspiration biopsy (Figure 4).

View larger version (25K): [in this window] [in a new window]   Figure 4. Flow diagram showing how patients were studied on the basis of the study protocol. * = 8 subcarinal and 2 contralateral mediastinal lymph nodes; = 2 subcarinal and 12 contralateral mediastinal lymph nodes.

The operating characteristics and CIs for endoscopic ultrasonography with and without aspiration biopsy were then determined by using a multistage screening process previously described by Chinchilli [44]. In the comparison of sensitivities, patients with positive results (malignant cells) on pathology or cytology were included in the definition of the gold standard. In contrast, only patients with negative results on surgical pathology were included in the definition of the gold standard for the comparison of specificities.

We also retrospectively compared the cost of endoscopic ultrasonography (alone or with aspiration biopsy) with the cost of the more invasive staging procedures of mediastinoscopy and thoracotomy that were aborted (because of the presence of metastasis to lymph nodes) on the basis of the management options actually chosen for each patient.

Results

Top Methods Results Discussion Author & Article Info References

One hundred thirty patients with non-small-cell lung cancer were evaluated during the study period. In 52 of these patients, computed tomography showed possible mediastinal lymphadenopathy, and no evidence indicated metastatic disease elsewhere (this analysis includes the 17 patients in our pilot study [43]) (Figure 4). These 52 patients then had endoscopic ultrasonography. Twenty-four patients had endoscopic ultrasonography-guided fine-needle aspiration biopsy of suspicious mediastinal and subcarinal lymph nodes. A total of 24 lymph nodes (1 per patient)-14 contralateral nodes and 10 subcarinal nodes-were aspirated. Fourteen patients had positive cytologic findings on aspiration biopsy; this was accepted as representing definite nodal metastasis.

On the basis of preoperative evaluations, 21 patients were considered to have unresectable tumors. This group consisted of the 14 patients with malignant cytologic findings on ultrasonography-guided aspiration biopsy of contralateral or subcarinal lymph nodes and 7 patients in whom distant metastasis or advanced disease (T4) was found by endoscopic ultrasonography alone, computed tomography, or other diagnostic methods. Thirty-one patients ultimately had surgery (14 patients from our initial pilot group, 10 patients with benign cytologic findings on ultrasonography-guided aspiration biopsy, and 7 patients in whom aspiration biopsy was not performed because no suspicious lymph nodes were detected according to the ultrasonographic criteria) and therefore had surgical pathologic results available.

The predictive value of endoscopic ultrasonography for malignant involvement of lymph nodes in all patients who had surgical pathologic results is shown in Table 1. These data included the ultrasonographic results for the 14 patients who were ultimately found to have positive cytologic results on ultrasonography-guided aspiration biopsy and compare the nodal prediction provided by endoscopic ultrasonography to the gold standard of malignancy seen on cytology or pathology. Of the 45 patients evaluated, endoscopic ultrasonography alone accurately predicted mediastinal lymph node metastasis in 38 patients (overall accuracy, 84%). Three results were false negative, and four were false positive. The three false-negative results involved lymph nodes that were 8 mm in diameter or less, each of which had only a 1- to 2-mm focus of cancer on pathology. The four false-positive results involved lymph nodes that were more than 1.5 cm in diameter and appeared suspicious according to endoscopic ultrasonographic criteria but were shown by histopathology to have only reactive hyperplasia. Computed tomography also incorrectly determined that these same four nodes were positive. Endoscopic ultrasonography was best at accurately detecting mediastinal lymph node metastasis in the subcarinal (station 7), aortopulmonary window (station 5), paratracheal (station 4), and paraesophageal (station 8) regions (Figure 1).

Computed tomographic scans of the mediastinal lymph nodes in the same 45 patients were also compared with pathologic results (Table 1). Computed tomography correctly detected lymph node metastasis in 22 patients (overall accuracy, 49%; sensitivity, 64% [95% CI, 40.7% to 82.8%]). Fifteen false-positive results and 8 false-negative results occurred; all of the false-positive results were ascribed to the use of computed tomographic criteria that rely heavily on node size [13-23]. The combination of computed tomography and endoscopic ultrasonography did not improve overall accuracy beyond that seen with endoscopic ultrasonography alone because the three false-negative nodes on ultrasonography were also false negative on computed tomography; each of these nodes contained only a small focus of micrometastasis.

Table 2 and Table 3 show the diagnostic findings for the 24 patients who underwent ultrasonography-guided aspiration biopsy. Cytologic findings were positive in 14 patients; these results confirmed the presence of N3 disease in 12 patients and N2 disease (subcarinal involvement only) in 2 patients. Thus, these patients were ineligible for surgery. In 9 of the 10 patients who had negative cytologic findings and ultimately underwent surgery, pathologic results confirmed the ultrasonographic prediction of negative lymph nodes. In one patient with a negative cytologic result, a small focus of tumor invasion was seen in one 8-mm lymph node (that is, the cytologic result was false negative).

The sensitivities and specificities of endoscopic ultrasonography with and without aspiration biopsy and computed tomography were compared. Both the sensitivity and the specificity of endoscopic ultrasonography alone and with aspiration biopsy were significantly higher than those of computed tomography. On the basis of cytologic findings, ultrasonography-guided aspiration biopsy influenced the clinical management of 23 of 24 patients (95%) because it avoided any further staging attempts with mediastinoscopy in all 24 patients and avoided unnecessary surgery in the 14 patients who had positive cytologic results. The one patient with a false-negative result on cytology did not benefit from the ultrasonography-guided procedure and underwent unnecessary surgery (Table 3).

By using the Chinchilli approach, we estimated the sensitivities and specificities of the sequence of endoscopic ultrasonography with and without aspiration biopsy. The results of these calculations, shown in Table 4, revealed that the combined approach had a sensitivity of 95% (CI, 85% to 100%) and a specificity of 81% (CI, 64% to 98%).

We retrospectively compared the direct costs of ultrasonography-guided aspiration biopsy with the costs of mediastinoscopy and thoracotomy as they are currently used for detecting mediastinal lymph node involvement. The results of this comparison (Table 5) suggest that ultrasonography-guided aspiration biopsy may be a cost-saving method for obtaining accurate final staging of non-small-cell lung cancer.

Discussion

Top Methods Results Discussion Author & Article Info References

Our findings provide important information on the accuracy of endoscopic ultrasonography-guided fine-needle aspiration biopsy for diagnosing metastasis to mediastinal lymph nodes. They also support our earlier pilot data, which suggest that endoscopic ultrasonography is more accurate than computed tomography in detecting metastasis to posterior mediastinal lymph nodes in patients with non-small-cell lung cancer [43]. Current endoscopic ultrasonographic criteria used to identify malignant lymph nodes were correct in 84% of patients compared with 49% for computed tomographic criteria. The combination of computed tomography and endoscopic ultrasonography did not improve overall accuracy above that seen with endoscopic ultrasonography alone for detecting lymph node involvement, but it did help evaluate the extent of the lung cancer; detect distant metastasis not seen on endoscopic ultrasonography; and evaluate anterior and pretracheal nodes, which are not imaged by endoscopic ultrasonography.

Computed tomography of the chest is considered the gold standard in the preoperative evaluation of the mediastinum. However, this method relies primarily on lymph node size for determining metastatic involvement. Our data suggest that size alone is not an accurate criterion for assessing lymph node status. Size does not accurately differentiate inflammatory or reactive lymph nodes from malignant involvement and cannot detect normal-appearing lymph nodes that contain small foci of micrometastasis [13-24]. Furthermore, geographic factors influence the size of mediastinal lymph nodes. For example, in Indiana (an area in which histoplasmosis is endemic), the majority of the population has large mediastinal lymph nodes [36-40]. We postulate that this regional phenomenon may be partly responsible for the relatively low overall accuracy of computed tomography (49%) in evaluating such nodes in our patients.

Patients in whom non-small-cell lung cancer is diagnosed undergo routine staging with computed tomography of the chest and bronchoscopy. Bronchoscopic fine-needle aspiration biopsy is commonly used to evaluate suspicious paratracheal, hilar, and subcarinal lymph nodes seen on computed tomography for malignant involvement. The role of this technique in the diagnosis and staging of non-small-cell lung cancer is well established [45-50]. Malignant involvement of mediastinal lymph nodes, particularly bulky subcarinal or contralateral nodes, on bronchoscopic fine-needle aspiration biopsy of nodes found to be enlarged on computed tomography usually precludes surgery. When lymph node status according to both computed tomography and bronchoscopy is unknown, mediastinoscopy and, in some cases, thoracotomy are usually performed to clarify disease stage. These procedures require general anesthesia and inpatient recovery, thereby increasing the time, cost, and risk of the staging process. Another limitation is that mediastinoscopy cannot reliably evaluate subcarinal nodes.

With the advent of endoscopic ultrasonography, guided fine-needle aspiration sampling of suspicious lymph nodes became possible. We postulated that the use of this technology in conjunction with endoscopic ultrasonography staging could enhance the overall accuracy of detecting metastasis of non-small-cell lung cancer to mediastinal lymph nodes. Thus, once guided aspiration biopsy became available, all eligible patients presenting to our service underwent this procedure to confirm the existence of metastasis in any suspicious lymph node seen on endoscopic ultrasonography. These patients were then followed to assess the effect of the resulting cytologic findings on subsequent management. Our data indicate that this new technique represents an improvement over existing staging methods because it is less invasive and has a sensitivity of 95% for determining metastasis to posterior mediastinal lymph nodes. In light of these results, it seems reasonable to consider using endoscopic ultrasonography and ultrasonography-guided fine-needle aspiration biopsy in patients with suspicious posterior mediastinal lymph nodes; at least for now, it might be used in patients in whom previous lymph node sampling done by use of computed tomography or bronchoscopic techniques was unsuccessful. Mediastinoscopy or partial thoracotomy could be reserved for patients with enlarged anterior lymph nodes or suspicious nodes that were not successfully sampled by computed tomography, bronchoscopy, or endoscopic ultrasonography.

The overall diagnostic accuracy of endoscopic ultrasonography alone or with fine-needle aspiration biopsy was 96%; this method was significantly better than computed tomography for predicting mediastinal lymph node metastasis, particularly when metastases involved the subcarina and posterior mediastinum. Furthermore, the results of fine-needle aspiration biopsy contributed significantly to patient management because cancer in patients who had positive cytologic results was staged as pathologic III A disease (bulky subcarinal nodes). These patients were considered to be better served by protocols that included chemotherapy and radiation rather than surgical resection. No complications of guided fine-needle aspiration biopsy occurred in our study.

Despite its many potential advantages, endoscopic ultrasonography is not yet widely available, largely because of a lack of skilled endosonographers; the procedure can therefore be done only in academic or tertiary referral medical centers. It will be some time before many institutions have an opportunity to use this unique approach to staging non-small-cell lung cancer. Until then, they must rely on bronchoscopic or transthoracic fine-needle aspiration biopsy and mediastinoscopy, which, in experienced hands, contribute substantially to the detection of mediastinal lymph node metastasis and to the overall staging of non-small-cell lung cancer [51, 52]. The use of endobronchial sonography to evaluate bronchial carcinomas was recently reported [53]. Even more interesting is the development of an ultrasonic bronchoscope that can be used to diagnose lung masses [54]. Future use of this new generation of endobronchial instruments will probably enhance our ability to diagnose and locally stage lung tumors and to determine lymph node status.

Our study has some limitations. First, as noted, our patients had an increased regional prevalence of enlarged mediastinal and subcarinal lymph nodes. Because histoplasmosis is endemic in Indiana, much of the population is at risk for developing enlarged benign mediastinal nodes; this compounds the problem of staging in non-small-cell lung cancer. Most of these patients subsequently have more aggressive and invasive staging procedures. The high prevalence of benign enlarged nodes may have contributed to the low accuracy of computed tomography in our study sample; the diminished value of computed tomography in Indiana may therefore not apply to other parts of the United States. Conversely, if the diagnostic accuracy of computed tomography is improved in other regions, the advantages of endoscopic ultrasonography may be less obvious elsewhere. No data support this possibility, however, and endoscopic ultrasonography has not been systematically compared with computed tomography for overall accuracy in detecting mediastinal nodes. Studies must be performed to answer these questions and to verify whether endoscopic ultrasonography can detect lymph nodes not seen on computed tomography.

Second, we used what some might consider an unusual gold standard-the combination of fine-needle aspiration cytology and surgical pathology. We considered positive (malignant) results on cytology to be equal to positive results on pathology because false-positive cytologic results are rare and usually occur with inexperienced operators. Endoscopic ultrasonography-guided fine-needle aspiration biopsy done by using a 22-gauge needle provides ample tissue, in contrast to the limited amount of tissue usually obtained with the 25-gauge ("skinny") needles used for most cytologic sampling.

A third limitation is the inability of endoscopic ultrasonography to image the anterior mediastinum; thus, the use of this procedure is restricted primarily to the evaluation of the posterior mediastinum. Fourth, although we correlated our findings on computed tomography, endoscopic ultrasonography alone, and endoscopic ultrasonography-guided fine-needle aspiration biopsy with the nodes found on careful surgical dissection, certain nodes from difficult surgical resection may have been mismatched. Finally, our sample size is relatively small. We hope that future studies with larger samples, preferably drawn from multiple centers, will more clearly define the role of ultrasonography and endoscopic ultrasonography-guided fine-needle aspiration biopsy in the preoperative staging of non-small-cell lung cancer.

In summary, endoscopic ultrasonography by itself provided useful information about the status of posterior mediastinal lymph nodes in patients with non-small-cell lung cancer. The addition of transesophageal endoscopic ultrasonography-guided fine-needle aspiration biopsy further improved the overall accuracy of lymph node staging to 96%. Thus, endoscopic ultrasonography-guided fine-needle aspiration biopsy seems to be an important addition to the overall preoperative staging of patients with potentially resectable non-small-cell lung cancer through its ability to establish the pathologic diagnosis of N2 and N3 disease. Further investigations are now needed, including prospective comparisons of cost-effectiveness and outcome, examination of the role of spiral computed tomography, and the evaluation of patient populations in which enlarged mediastinal lymphadenopathy are uncommon (for example, in areas where histoplasmosis is not endemic) before the role of endoscopic ultrasonography alone or with fine-needle aspiration biopsy in the preoperative staging of non-small-cell lung cancer can be considered firmly established.

Dr. Savides: University of California at San Diego, 200 West Arbor Drive, San Diego, CA 92103.

Drs. Sandler, Kesler, Conces, Cummings, and Mathur and Ms. Bilderback: 550 North University Boulevard, Indiana University Medical Center, Indianapolis, IN 46202.

Dr. Ikenberry: 200 East Pennsylvania, Peoria, IL 45050.

Dr. Hawes: Medical University of South Carolina, 171 Ashley Avenue, Charleston, SC 29425.