Endobronchial ultrasound in real life: primary diagnosis and mediastinal staging of lung cancer in patients submitted to thoracic surgery

Background Since the first articles published for over 10 years ago, endobronchial ultrasound (EBUS) has gained a strong scientific backing and has been incorporated into routine medical practice in pulmonology and thoracic surgery centers. How is EBUS performing outside the scientific environment, as a diagnostic and mediastinal staging tool in a subset of patients that undergo thoracic surgery, is an interesting question. Methods This study evaluated consecutive patients who, during the period from January 2010 to August 2012, were submitted to EBUS and later to thoracic surgery. The samples obtained by endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA) were compared to surgical samples. The primary endpoint was the proportion of patients with a final diagnosis of non-small cell lung cancer (NSCLC) by EBUS-TBNA correctly subtyped. The secondary endpoint was the negative predictive value (NPV) of EBUS-TBNA for mediastinal staging of lung cancer. Results Two hundred eighty seven patients were studied. Considering 84 patients with a final diagnosis of NSCLC by EBUS-TBNA, 79 % (CI 95 % 70.1–87.3) were correctly subclassified. The NPV of EBUS-TBNA for mediastinal staging was 89 % (IC 95 % 84.9–92.7). From a total of 21 false negative cases of mediastinal staging, 16 (76 %) did not undergo positron emission tomography-computed tomography (PET-CT) before the EBUS and in 15 (71 %) the affected lymph node chain was not punctured by EBUS-TBNA. Ten (47 %) patients had only lymph node metastases not directly accessible by the EBUS. Conclusions Performed in hospital routine and in patients submitted to thoracic surgery, EBUS-TBNA proved to be a good tool for proper pathological diagnosis of lung cancer. The negative predictive value of 89 % for mediastinal staging of lung cancer is comparable to that reported in previous studies, but the relatively high number of 21 false negative cases points to the need for standardization of routine strategies before, during and after EBUS.


Background
Lung cancer is the malignant tumor with the highest mortality rate among men and women worldwide, with more than 1,4 million deaths a year [1]. The emergence of endobronchial ultrasound-guided transbronchial needle aspiration (EBUS-TBNA), a minimally invasive technique able to provide valuable information for a primary tumor diagnosis and mediastinal staging, significantly changed the approach to lung cancer [2,3]. Since the first articles published for over 10 years ago [4], endobronchial ultrasound (EBUS) has gained strong scientific backing [5][6][7] and has been incorporated into routine medical practice of pulmonology and thoracic surgery centers. Guidelines of respiratory societies reinforce the importance of ultrasound-guided needle techniques in the primary diagnosis and mediastinal staging of lung cancer [8][9][10][11]. The success of the technique and the spread of its use around the world make it important to revaluate its performance in "real life", outside the scientific environment, especially in patients undergoing thoracic surgery, in which the sharpest staging tools are required.
The aim of our study is to determine EBUS performance in patients undergoing thoracic surgery in hospital routine. Using surgical pathology as the gold standard, we calculated the proportion of patients with a final diagnosis of non-small cell lung cancer (NSCLC) by EBUS-TBNA correctly subtyped and the performance of EBUS-TBNA for the mediastinal staging of lung cancer.

Study design
The present work is a single-center, retrospective, observational study. All patients who underwent EBUS and thoracic surgery during the period from January 2010 to August 2012 were selected.

Procedures
The examinations and surgeries took place in the Lungenklinik Hemer, a traditional center for respiratory diseases in Germany. EBUS procedures were performed on hospitalized patients, under general anesthesia. The examination could be accompanied by the presence of a pathologist in the procedure room for rapid on-site evaluation (ROSE) or not. EBUS-TBNA samples were collected using a 22-gauge needle. At least 2 aspirates were obtained from each target lesion. When a pathologist was present, 2 pairs of smears were prepared for ROSE and cytologic examinations. The resulting material (tissue cores, shreds of tissue, cellular components, fluids) was processed using a cell-block technique. If there was no pathologist in the room, all the samples were processed as cell-blocks. Indications for a surgical exploration or resection due to suspicion or confirmation of lung cancer were discussed and had to be approved by a local interdisciplinary clinical session with the mandatory participation of pulmonologists, oncologists, chest surgeons, radiologists and radiotherapists. EBUS-TBNA and surgical samples were analyzed by two different pathology services.

Data collection
Data collection and statistical analysis were conducted at the Federal University of Rio de Janeiro, Brazil. Through the computer program Teamviewer ® (TeamViewer GmbH, Germany) the authors had remote access to a database provided by the Lungenklinik Hemer to access patient data from local electronic medical records.

Inclusion criteria
Patients who underwent EBUS for any indication in the Lungenklinik Hemer from January 2010 to August 2012 and were subsequently subjected to surgical procedures.

Exclusion criteria
Patients with no lymph node sampling by EBUS-TBNA during an EBUS procedure or patients who did not have histopathologic sampling of mediastinal lymph nodes during surgery.

Study population
Patients were identified in the electronic system of the Lungenklinik Hemer by crossing the specific code representing the EBUS with codes representing thoracic surgical procedures such as thoracotomy, thoracoscopy or mediastinoscopy. Patients whose electronic records showed both codes in the same hospital stay were selected. Four hundred thirty-nine patients underwent both an EBUS and thoracic surgery between January 2010 and August 2012 in the Lungenklinik Hemer. One hundred fifty-two cases were excluded for not having mediastinal lymph node sampling from both EBUS and surgery. The remaining 287 patients were studied (Fig. 1).

Endpoints
The primary endpoint was the proportion of patients with a final diagnosis of NSCLC by EBUS-TBNA correctly subtyped as compared to the surgical samples. The secondary endpoint was the negative predictive value (NPV) of EBUS-TBNA for mediastinal staging of lung cancer, according to the 7 th edition of the lung cancer staging system of the American Joint Committee on Cancer [12].

Results
The demographic data, tumor characteristics and details of the procedures performed are summarized in Table 1 (Table 3) did not undergo positron emission tomography-computed tomography (PET-CT) or it was performed after the EBUS. In 15 (71 %) patients the affected lymph node chain was not punctured by EBUS-TBNA. Ten (47 %) patients had only lymph node metastases not directly accessible by the EBUS (lymph node chains 5,6,8 and 9). In 11 (52 %) patients the tumor was located in the left superior lobe. The more often affected lymph node chain was station 5, with 7 false negative cases.
Considering the 4 false positive cases, 2 were surgically classified as N1 and 2 were classified as N0 (all confirmed surgically as NSCLC). In just one of the cases there was a disagreement of the pathological findings of EBUS-TBNA and surgery (EBUS-TBNA suggested adenocarcinoma and surgery confirmed large cell carcinoma). All the 4 cases were submitted to surgical lobectomy and lymph node dissection.
Twenty patients had a mediastinoscopy after the EBUS (4 as the main surgical procedure and 16 as part of the mediastinal staging before surgery). There were 15 true negative, 1 true positive, 4 false negative and no false positive findings comparing mediastinoscopy to final surgical mediastinal staging. Mediastinoscopy did not contribute to a better mediastinal staging than the EBUS in any of the patients. All the 15 true negative cases were also negative by the EBUS. The true positive case was also positive by the EBUS, and one of the 4 false negative mediastinoscopies cases was positive by the EBUS.

Discussion
Our study evaluated the EBUS-TBNA performance in a key subset of patients with lung cancer: those undergoing thoracic surgery. Despite not representing the majority of patients diagnosed with lung cancer, this is the subgroup in which we need to have the sharpest diagnostic and staging tools to ensure an accurate referral to surgery and expectation of cure. To our knowledge, this is the work with the highest number of patients submitted to surgery who had their EBUS-TBNA results directly compared with the surgical sampling. The use of EBUS was evaluated in a hospital routine, without adherence to study protocols influencing the exam.
The primary endpoint of the study was the proportion of patients with a final diagnosis of NSCLC by EBUS-TBNA correctly subtyped. Our results show that NSCLC was correctly subclassified in 79 % of cases. One of the limitations of existing evidence on EBUS diagnostic performance is that many studies included results of the index test and clinical follow-up in a reference standard test and did not account for the surgical sample being the only possible gold standard. This may have overestimated the result of some studies [13]. Esterbrook et al. showed that EBUS-TBNA samples when made into cellblocks and subjected to a panel of immunohistochemical    stains returned adequate tissue for NSCLC subtyping in 79 %, with a NSCLC-NOS reate of 21 % [14]. In a large, multicenter study, Navani et al. demonstrated that samples from EBUS-TBNA provide sufficient information for subtyping NSCLC in 77 % of the cases [15]. We chose the NPV of EBUS-TBNA for the mediastinal staging of lung cancer as the secondary endpoint of our study because we consider it to be the most clinically relevant measure in the subset of patients who undergo thoracic surgery. From a total of 238 patients with NSCLC surgically evaluated, 53 had mediastinal metastatic involvement ipsilateral to the target tumor lesion (N2 disease) and 2 had contralateral mediastinal involvement (N3 disease), representing a prevalence of mediastinal nodal involvement of 23 %. The results of our study showed an NPV of 89 %. Probably some of the most important publications of EBUS and mediastinal lymph node staging in patients with potentially resectable NSCLC are the ASTER trial [16], published in 2010, and the work of Yasufuku et al. [17], published in 2011. The ASTER trial found an NPV for endosonography staging alone without additional surgical staging of 85 % (prevalence of N2/N3 54 %). Yasufuku showed an NPV of 91 % for EBUS-TBNA (prevalence of N2/N3 disease 35 %).
The sensitivity of 61 % in our work was lower than expected and previously reported. This fact is due to the relatively high number of 21 false negative cases. In reviewing such cases we realize that 16 (76 %) did not undergo PET-CT or it was performed after EBUS and in 15 (71 %) the affected lymph node chain was not punctured by EBUS-TBNA. In the ASTER trial [16] all patients underwent PET-CT before EBUS. In the discussion of the Lung-BOOST trial [18], the authors suggest that a PET-CT may not be needed before EBUS-TBNA. In that trial EBUS-TBNA was performed using a systematic aspiration of all visible lymph node stations. Unfortunately, in our study, we are unable to determine if most of the procedures adopted a systematic or selective approach.
In 11 (52 %) patients from the false negative cases, the tumor was located in the left superior lobe and 10 (47 %) had only lymph node metastases not directly accessible by EBUS (lymph node stations 5, 6, 8 and 9). Endoscopic ultrasound (EUS), using the same scope as EBUS, was done in only 6 of our 287 cases. None of the false negative cases underwent also EUS in the same procedure as EBUS. The current consensus is that for a more complete needle-guided ultrasound evaluation of the mediastinum we should associate EBUS and EUS whenever necessary [19,20]. Perhaps with the more frequent use of EUS in routine practice, some false negative results could have been avoided. However, even though recently showed to be feasible and safe [21], it would still be difficult to access the lymph nodes in chains 5 (the most often affected) and 6. This points to the need for more cautious strategies in patients susceptible to metastases in these chains, such as patients with tumors in the left superior lobe.
The 4 false positive cases represent patients with NSCLC. Unfortunately we do not have follow-up data, so is difficult to affirm that these are really false positive cases or maybe incorrectly surgically staged patients The small number of 20 mediastinoscopies already reflects the lower use of this technique in our hospital routine. All mediastinoscopies were performed after EBUS to clarify questionable situations before the final surgical decision. Mediastinoscopy did not contribute to better mediastinal staging than EBUS-TBNA in any of those patients. These findings do not reflect the results of the ASTER trial [16] or corroborate the current recommendations of the main guidelines of American and European respiratory societies.

Limitations
Some limitations apply to this study. First, it is a singlecenter study. We recognize that our results cannot be easily generalized. EBUS was performed in hospitalized patients under general anesthesia. Although this is a standard practice in our service, many procedures, probably the majority, performed in other bronchoscopy services around the world are performed in conscious or moderate sedation. Even tough the World Association for Bronchology and Interventional Pneumology (WABIP) guidelines [22] state that there is not enough evidence to recommend for or against any type of anesthesia, ideally, we should also have data in patients undergoing the procedure under conscious or moderate sedation. Second; two distinct pathology services have analyzed the samples obtained by EBUS-TBNA and surgery, which may have prevented results of one of the tests influencing the analysis of the other test. But it is difficult to clarify if the differences in pathological classification are due to the quality of the samples or a distinct interpretation by the pathologists. Third, at least 2 aspirates were routinely obtained from each target lesion per EBUS. Lee and colleagues [23] showed that, in the absence of ROSE, at least 3 aspirates should be obtained from each target lesion in order to provide optimal results from the test. Unfortunately we are not able to identify the number of aspirates that were used in each examination.

Conclusions
In hospital routine and in the subgroup of patients eligible for surgical resection EBUS-TBNA has been proven to be a good tool for the primary diagnosis of lung cancer. The negative predictive value of 89 % for mediastinal staging of lung cancer is comparable to that reported in previous studies, but the relatively high number of false negative cases points to the need for standardization of routine strategies before, during and after EBUS.