Diagnostic Value of EBUS-TBNA for Lung Cancer with Non-Enlarged Lymph Nodes: A Study in a Tuberculosis-Endemic Country

Background In tuberculosis (TB)-endemic areas, contrast-enhanced computed tomography (CT) and positron emission tomography (PET) findings of lung cancer patients with non-enlarged lymph nodes are frequently discrepant. Endobronchial ultrasound-guided transbronchial aspiration (EBUS-TBNA) enables real-time nodal sampling, and thereby improves nodal diagnosis accuracy. This study aimed to compare the accuracy of nodal diagnosis by using EBUS-TBNA, and PET. Methods We studied 43 lung cancer patients with CT-defined non-enlarged mediastinal and hilar lymph nodes and examined 78 lymph nodes using EBUS-TBNA. Results The sensitivity, specificity, positive predictive value, and negative predictive value of EBUS-TBNA were 80.6%, 100%, 100%, and 85.7%, respectively. PET had low specificity (18.9%) and a low positive predictive value (44.4%). The diagnostic accuracy of EBUS-TBNA was higher than that of PET (91% vs. 47.4%; p<0.001). Compared to CT-based nodal assessment, PET yielded a positive diagnostic impact in 36.9% nodes, a negative diagnostic impact in 46.2% nodes, and no diagnostic impact in 16.9% nodes. Patients with lymph nodes showing negative PET diagnostic impact had a high incidence of previous pulmonary TB. Multivariate analysis indicated that detection of hilar nodes on PET was an independent predictor of negative diagnostic impact of PET. Conclusion In a TB-endemic area with a condition of CT-defined non-enlarged lymph node, the negative diagnostic impact of PET limits its clinical usefulness for nodal staging; therefore, EBUS-TBNA, which facilitates direct diagnosis, is preferred.


Introduction
Lung cancer remains a fatal disease worldwide, and surgical treatment may offer long-term survival. However, the indication and outcome of surgical resection depend on accurate preoperative staging of the cancer and the extent of intraoperative lymph node dissection [1,2,3]. Therefore, accurate lymph node staging in patients with non-small cell lung cancer (NSCLC) is crucial for planning the optimal treatment. Conventional contrast-enhanced computed tomography (CT) has been used to identify enlarged lymph nodes; lymph nodes .1 cm are defined as nodal metastatic lesions. However, because it is moderately sensitive and pecific, contrast-enhanced CT carries a substantial risk of understaging small nodal metastasis and overstaging inflammatory lymphadenitis [4].
Positron emission tomography (PET) using fluorine-18 fluorodeoxyglucose ( 18 F-FDG) provides functional information about tumor metabolism and has been used as a non-invasive alternative to contrast-enhanced CT for nodal staging in NSCLC [5,6,7]. In some countries, FDG-PET is increasingly being used for lymph node staging in NSCLC when lymph node enlargement is not detected by CT. Therefore, the accuracy afforded by FDG-PET may substantially alter the treatment strategy in institutions in which mediastinoscopy is unavailable for lymph node sampling. However, abnormal FDG uptake is believed to frequently occur in granulomatous and inflammatory diseases [8,9]. In tuberculosis (TB)-endemic areas such as Eastern Asia, FDG-PET has been reported to have low specificity and a positive predictive value in nodal staging in NSCLC [10,11,12]. Therefore, in such areas, the PET scan does not appear to replace mediastinoscopy for nodal staging of NSCLC [11], especially in the case of potentially operable patients without enlarged mediastinal lymph nodes.
The recently developed endobronchial ultrasound-guided transbronchial aspiration (EBUS-TBNA) procedure enables direct and real-time aspiration of mediastinal and hilar lymph nodes and is a less invasive alternative to mediastinoscopy for nodal staging [13,14]. Unlike CT and FDG-PET, in which lymph node stage is determined on the basis of the results of image analysis, EBUS-TBNA enables direct nodal sampling; therefore, nodal staging can be performed on the basis of the results of cytological or pathological examination. However, EBUS-TBNA has been reported to have variable sensitivity and negative predictive value [13,15], especially in cases wherein the size of the lymph nodes has been reduced by induction chemotherapy [16]. However, EBUS-TBNA has been reported to show high sensitivity and specificity for detecting small nodal metastasis in NSCLC patients without significant mediastinal lymph node enlargement on CT [17,18]. Therefore, whether EBUS-TBNA shows high diagnostic accuracy for nodal staging in lung cancer patients without enlarged mediastinal lymph nodes on CT in a TB-endemic country-a condition for which FDG-PET has shown much negative diagnostic impact, i.e. false-positive results-remains to be clarified.
In this study, we primarily aimed to compare the diagnostic accuracy of EBUS-TBNA and FDG-PET in lung cancer patients with CT-defined non-enlarged mediastinal and hilar lymph nodes. Further, we also aimed to look at the diagnostic impact of FDG-PET in this condition, and investigated the characteristics of lymph nodes which accounted for either negative or positive diagnostic impact on FDG-PET.

Study subjects
Between January 2007 and January 2009, nodal sampling by using EBUS-TBNA was performed in 43 lung cancer patients with non-enlarged mediastinal and hilar lymph nodes defined by contrast-enhanced CT examination. Written informed consent was obtained from all study subjects and the institutional review board of Chang Gung Memorial Hospital approved the study (No. 98-3639A3).

Imaging techniques
Contrast-enhanced CT was initially performed in the 45 patients included in the study to examine lung parenchymal and mediastinal lesions. CT images were obtained from the thoracic inlet to the adrenal gland, and the images were assessed by an experienced radiologist who was blinded to the FDG-PET and pathology reports. Non-enlarged mediastinal and hilar lymph nodes were defined as lymph nodes with short-axis diameters #10 mm on contrast-enhanced CT of the chest. The attending physician referred patients for FDG-PET by considering each patient's will and payment of personal medical insurance. Each patient was instructed to fast for 6 h before undergoing PET and was intravenously injected with 370 MBq (10 mCi) of 18 F-FDG 60 min before the examination. 18 F-FDG PET images were obtained using an ECAT EXACT HR+ camera (Siemens/CTI Inc., Knoxville, TN) by using full width at half-maximum of 4.5 mm and a transaxial field of view of 15 cm. Transaxial, sagittal, and coronal tomographs were assessed primarily to determine obvious abnormal foci of increased FDG uptake. All the images were double-read by two nuclear physicians with more than 6 years experience in PET interpretation, and the discrepancies were resolved by discussion. A visual analysis score (VAS) using a 4-point scale was primarily used to determine 18 F-FDG uptake in the lymph nodes, and the standardized uptake value was used as an accessory reference. The 4-point scale considered both the intensity of 18 F-FDG uptake in the lymph nodes and the contrast between the background mediastinum and lymph nodes showing 18 F-FDG accumulation. Accordingly, the 4point scale was developed as follows: 1, intensity of focal uptake of 18 F-FDG lower than that of mediastinal uptake (benign lesion); 2, intensity of focal uptake of 18 F-FDG equal to that of mediastinal uptake (equivocal for malignancy); 3, intensity of focal uptake of 18 F-FDG greater than that of mediastinal uptake (possibility of malignancy); and 4, intensity of focal uptake of 18 F-FDG significantly greater than that of mediastinal uptake (high probability of malignancy). A VAS of 3 or 4 was considered positive, whereas a score of 1 or 2 was considered negative. PET scan results of 32 patients were available at the time of data analysis. Two patients were excluded from EBUS-TBNA test due to metastatic disease noted by PET. Compared to CT-based nodal assessment, we interpreted a true-positive PET result as having positive diagnostic impact and a false-positive PET result as having negative diagnostic impact, categorizing either a true-negative or false-negative PET result as having no diagnostic impact (i.e. PET showed a same nodal assessment as CT).

EBUS-TBNA
All of the patients underwent TBNA via a flexible ultrasoundguided bronchoscope with a linear scanning probe on the tip (BF-UC206F-OL8, Olympus) EBUS-TBNA were performed by two well-trained operators both have had experience in more than 30 cases, and the bronchoscopic operators were blinded to the PET scan result. The curved probe scanned in a direction parallel to that of the bronchoscope insertion, and the obtained images were linked to an ultrasound scanner (EU-2000C, Olympus) with Doppler-flow imaging. Conscious sedation was induced in all patients by using midazolam before the bronchoscope was inserted. Ultrasound examination of the hilar lymph node (stations 10, 11) was followed by subcarinal (station 7) and mediastinal lymph nodes (stations 4, 2, 1). When a lymph node was visualized, the picture was recorded and selected cursors were used to measure its size 2-dimensionally. Lymph nodes with short-axis diameters .5 mm were selected for subsequent real-time EBUSguided TBNA with a 22-gauge needle (NA-201SX-4022, Olympus). The needle was passed at least 2 times for each targeted lymph node. An internal sheath was placed inside the needle to avoid contamination of the bronchial epithelium during each puncture, and the sheath was removed after passage of the needle into the targeted lymph node. Cytological examination of the specimens was performed by a pathologist blinded to the patients' clinical histories and imaging results. Pathological examination of the specimen was also performed when a tissue core was obtained by TBNA.

Statistical analysis
The result of each diagnostic modality was compared with that of the surgical specimen obtained by extensive lymph node dissection in 30 (69.8%) patients. Mediastinoscopy was performed to approach N2 level lymph nodes in 5 patients; in them, biopsy specimens were used for standard comparison. Eight patients refused any sort of approach for diagnosis of mediastinal lymph node involvement or treatment. Neither did they receive surgery nor chemotherapy or radiotherapy. A series of CT scan image follow-up was done in those patients; in that, 20% increase in short-axis of a lymph node within 3 months was used to defined pre-existing metastasis The sensitivity, specificity, positive predictive value, and negative predictive value of each diagnostic modality were calculated according to the standard definition. The chi-squared test was used to compare differences between groups.
All reported p values were two-sided, and a p value ,0.05 was considered statistically significant.

Diagnostic value of PET and EBUS-TBNA
The results of each diagnostic modality on a per lymph node basis are shown in Table 2. The sensitivity, specificity, positive predictive value, and negative predictive value were 80.6%, 100%, 100%, and 85.7%, for EBUS-TBNA and 85.7%, 18.9%, 44.4%, and 63.6% for PET, respectively. Thus, in our study, the diagnostic accuracy of EBUS-TBNA was higher than that of PET (91% vs. 47.4%; p,0.001). Further, when primary tumor site was considered during interpretation of the PET results, the contralateral lymph nodes with 18 F-FDG uptake scores of 3 or 4 were deemed PET negative. Thus, the accuracy of the modified method in which PET results were interpreted on the basis of the primary tumor site was still lower than that of the EBUS-TBNA results (52.3% vs. 91%, p,0.001). When lymph nodes were positive for malignancy by EBUS-TBNA irrespective of PET findings (PET-positive, n = 19; PET-negative, n = 3), the predictive value of EBUS-TBNA for lymph nodes diagnosis was 100% (Fig. 1). When lymph nodes were negative for malignancy by both EBUS-TBNA and PET (n = 8), the predictive value of both tests were 87.5%. In the case of EBUS-TBNA-negative and PETpositive lymph nodes (n = 35), predictive value was 85.7% in favor of the negative EBUS-TBNA result.

Diagnostic impact of PET
Compared to CT-based nodal assessment, 24 nodes detected by PET were true-positive, yielding a 36.9% positive impact incidence (Table 3). Thirty nodes detected by PET were falsepositive, which accounted for a 46.2% negative impact incidence. Seven and 4 lymph nodes were PET-true-negative and PET-falsenegative, respectively, and were together responsible for a 16.9% incidence of no impact. Patients with lymph nodes yielding negative PET scan diagnostic impact were thoroughly investigated to rule out the presence of pulmonary diseases other than lung cancer. Of the 17 patients examined, 7 (41.2%) had previous pulmonary TB, 3 (17.6%) had pneumoconiosis, 1 (5.9%) had pulmonary fibrosis, and 6 (35.3%) had no history of pulmonary disease. Rrepresentative cases of PET-negative diagnostic impact due to previous pulmonary TB with granulomatous nodal inflammation and anthracosilicosis diagnosed by EBUS-TBNAare shown in Fig. 2 and 3.

Factors predicting the diagnostic impact of PET
The characteristics of lymph nodes yielding PET-positive impact were compared to those yielding PET-negative impact (Table 4). Four variables, namely lymph node size, lymph node location, PET scale, and presence of calcification were reviewed as factors predicting the diagnostic impact of PET. Among the PETpositive impact lymph nodes, 21 (87.5%) were located in the mediastinum and 3 (12.5%) were located in the hilum. In contrast, in the PET-negative impact group, 14 (46.7%) lymph nodes were located in the mediastinum and 16 (53.3%) were located in the

Diagnostic value of PET in determining distant metastasis
In 32 patients receiving PET exam, 3 were suspicious of distant metastasis. Two of them were excluded from EBUS-TBNA study due to lumbar spine metastasis proved by pathology, and the other one was proved to have cholecystitis. Hence; in the present study, the sensitivity, specificity and accuracy for PET to identified distant metastasis were 100%, 96.7% and 96.9%, respectively.

Discussion
In our study, EBUS-TBNA had higher accuracy than PET for nodal diagnosis of lung cancer with CT-defined non-enlarged lymph nodes. In addition, presence of an 18 F-FDG-positive hilar lymph node was an independent predictor for PET-negative diagnostic impact. Moreover, patients with lymph nodes yielding a PET-negative diagnostic impact showed a high incidence of inflammatory lung disease, especially those with previous pulmonary TB.
The strength and limitation of EBUS-TBNA in diagnosing lymph nodes metastasis has been reported in either TB or non-TB endemic country. The specificity was 100% in general; with a wide range of sensitivity from 69% to 94% [14,15,17,19,20,21]. However, the diagnostic performance of EBUS-TBNA in lung cancer with non-enlarged lymph nodes has only been primarily studied in non-TB endemic country. Herth et al. reported a 89%  to 92% sensitivity of EBUS-TBNA for determining the nodal status of NSCLC patients in whom the mediastinum appears normal on radiography [17,18], a condition defined by lymph nodes with short-axis diameters ,1 cm on contrast-enhanced CT. The current study conducted in a TB-endemic country also primarily aimed at lung cancer patients with non-enlarged lymph nodes. The result has shown an 80.6% sensitivity of EBUS-TBNA, which was comparable to the historical results in non-TB endemic area.. We reported a 53.8% negative predictive value of CTdefined negative mediastinal and hilar lymph nodes, and this value is not satisfactory but is similar to the result reported in another study [22]. Furthermore, our results show that PET has low diagnostic accuracy (47.7%) for determining nodal status, which can be mainly attributed to a high rate of false-positive results (46.2%), a negative diagnostic impact of PET frequently happened in TB-endemic areas.
The development of FDG-PET is considered major progress in the field of oncology that has led to advancement of assessment of the functional and molecular bases of tumors. However, F 18uptake in activated white blood cells interferes with that in tumor cells; therefore, false-positive findings in inflammatory and granulomatous lung diseases can be attributed to interference in F 18 -uptake [23]. Several reports from Asian countries indicate that PET has a 15-25% chance of overstaging potentially operable NSCLC [11,12]. Management of these patients would be substantially altered accordingly if an invasive technique for nodal staging were unavailable [24,25]. A history of pulmonary TB, despite inactive infection at the time of assessment, was reported to be the leading cause for the false-positive PET results [10,11,12,26]. In our study, among the patients with lymph nodes yielding negative diagnostic impact of PET, 64.7% had various inflammatory lung diseases. Previous pulmonary TB was the leading cause of inflammatory lung disease, followed by pneumoconiosis. These parenchymal lung diseases develop from focal inflammation and spread via lymphatic drainage [27]. Hilar lymph nodes are the sentinel gates that prevent infectious or proinflammatory substances from reaching the mediastinum. This may be attributed to the significantly high rate of false-positive PET results for hilar lymph nodes. EBUS-TBNA could balance the weakness of PET in such condition. In our study, a negative    EBUS-TBNA finding from a hilar node provided a 94.1% negative predictive value. On the other hand, a same finding from mediastinal lymph nodes showed an 80% negative predictive value. A large proportion of false-positive PET hilar lymph nodes in patients with no previous history of TB may be attributed to primary TB infection since a high percentage of adults in Taiwan are purified protein derivative (PPD)-positive [28,29]. In our study, due to a high rate of false-positive results, the accuracy of PET for nodal diagnosis in cases of lung cancer patients with non-enlarged mediastinal and hilar lymph nodes did not increase. However, our results do not dispute the value of PET in detecting malignant lymph nodes on the basis of 18 F-FDG uptake by the lymph nodes. The high sensitivity (85.7%) suggests that PET potentially can be used as a screening method to detect lymph nodes that are ,1 cm. However, the high false-positive rate observed in hilar nodes warrants further investigation. A dualtime-point evaluation of FDG uptake, which combines standard uptake value with a retention index, has been proposed to improve lymph node diagnosis in NSCLC [30,31]. However, other studies have reported controversial results of the superiority of dual-timepoint evaluation technique over conventional VAS [32,33]. On the other hand, our results show that the sensitivity and specificity of PET for distant metastasis is high; therefore, PET is an important staging tool to exclude distant metastasis.
In current study, we included lung cancer patients with nonenlarged lymph nodes to investigate the frequent clinical discrepancy between CT and PET results, but this was at the expense of a larger study population. However, the power of primary aim was not hampered by sample size because of the large effect size, i.e. the diagnostic accuracy difference of EBUS-TBNA and PET. In addition, PET imaging was performed on an older PET-only scanner, without CT correlation to intra-thoracic structures. This might be difficult for PET readers to determine whether the activity seen on PET correlates to lymph node or other structures. Whether use of integrated PET/CT might improve the accuracy deserves further investigation. However, given the dated technology of the scanner, it may not be enough to overcome the limited nature of PET arm of this study.
In conclusion, a high proportion of lung cancer patients with non-enlarged mediastinal and hilar lymph nodes on contrastenhanced CT were found to have mediastinal or hilar lymph node metastasis. In a TB-endemic area, a high rate of negative diagnostic impact limits the clinical usefulness of PET for nodal staging; therefore, we recommend direct tissue sampling using EBUS-TBNA for nodal staging.