Cornwell LD, Bakaeen FG, Lan CKW, Omer S, Preventza O, Pickrell B, Nguyen A, Casal RF. Endobronchial Ultrasonography-Guided Transbronchial Needle Aspiration Biopsy for Preoperative Nodal Staging of Lung Cancer in a Veteran Population. JAMA Surg. 2013;148(11):1024-1029. doi:10.1001/jamasurg.2013.3776
Recently, preoperative lung cancer staging has evolved to include endobronchial ultrasonography-guided transbronchial needle aspiration (EBUS-TBNA) biopsies of the hilar and mediastinal lymph nodes, but the feasibility and usefulness of the procedure have not been well studied in the veteran population.
To determine the safety and effectiveness of EBUS-TBNA as a key component of a preoperative staging algorithm for lung cancer in veterans.
Design, Setting, and Participants
Review of a prospectively maintained thoracic surgery database that includes patients who underwent lung resection for lung cancer between January 1, 2009, and December 31, 2012, at a single Veterans Affairs medical center among a consecutive cohort of 166 patients with clinically early-stage (I or II) lung cancer who underwent lobectomy with nodal dissection.
Endobronchial ultrasonography-guided transbronchial needle aspiration mediastinal staging (EBUS group) in 62 patients (37.3%) was compared with noninvasive nodal staging plus integrated positron emission tomography–computed tomography only (PET/CT-only group) in 104 patients (62.7%). The accuracy of nodal staging was assessed by comparison with the final pathological staging after complete nodal dissection (the gold standard).
Main Outcomes and Measures
Primary outcomes were feasibility, safety, accuracy, and negative predictive value of EBUS-TBNA for preoperative nodal staging. A secondary outcome was the rate of nontherapeutic lung resection for occult N2 disease, with comparison between the EBUS group and the PET/CT-only group.
No significant complications were attributable to the EBUS-TBNA procedure. In the EBUS group, 258 lymph node stations were sampled. N1 hilar metastases were diagnosed in 8 patients (12.9%) before surgery, and the remainder were staged N0. Accuracy and negative predictive value of EBUS-TBNA were 93.5% (58 of 62) and 92.6% (50 of 54), respectively. The overall rate of nontherapeutic lung resection performed in patients with occult N2 disease was 10.8% (18 of 166) (8.1% in the EBUS group and 12.5% in the PET/CT-only group) (P = .37).
Conclusion and Relevance
A preoperative lung cancer staging strategy that includes EBUS-TBNA seems to be safe and effective in a veteran population, resulting in a low rate of nontherapeutic operations because of occult N2 nodal disease.
Surgical resection of early-stage lung cancer offers the best chance for cure. Accurate preoperative nodal staging is imperative not only for making prognoses but also for developing optimal treatment strategies and limiting the number of nontherapeutic surgical resections. Nodal staging protocols routinely include integrated positron emission tomography–computed tomography (PET/CT), with reported accuracy rates ranging from 62% to 84%,1- 3 but frequent false-negative and false-positive results often necessitate mediastinal tissue biopsy. Endobronchial ultrasonography-guided transbronchial needle aspiration (EBUS-TBNA) biopsies of mediastinal and hilar lymph nodes, with real-time images provided by a convex probe, have revolutionized the preoperative workup since the initial 2004 study.3 Such biopsies improve staging accuracy beyond PET/CT,4,5 and the diagnostic yield of real-time EBUS-TBNA has been comparable to that of surgical mediastinoscopy in randomized clinical trials.6- 8 The advantages of EBUS over mediastinoscopy include its rare morbidity, ease of restaging, patient acceptance, reduced invasiveness, ability to reach N1 nodal stations, and lack of requirement for general anesthesia.9- 11
The veteran population with lung cancer differs from patient populations previously studied. Veterans are mostly male, often have significant comorbidities, and frequently continue to smoke after receiving the diagnosis of lung cancer.12- 14 Endobronchial ultrasonography-guided transbronchial needle aspiration has not been well studied in the veteran population but is a key element of the preoperative lung cancer staging algorithm at our Veterans Affairs (VA) medical center. This study was undertaken to assess the feasibility, safety, and accuracy of preoperative EBUS-TBNA nodal staging for patients with clinical stage I and II lung cancer who underwent surgery at a VA medical center compared with the gold standard of nodal dissection. Our hypothesis was that EBUS-TBNA nodal staging would be safe and accurate and would have a high negative predictive value (NPV) comparable with data from previous clinical trial settings, resulting in a low rate of futile resection.
A retrospective medical record review was performed on a prospectively maintained database of patients who underwent lobectomy for treatment of clinical stage I and II lung cancer at the Michael E. DeBakey Veterans Affairs Medical Center between January 1, 2009, and December 31, 2012. These operations were performed by 4 attending cardiothoracic surgeons (L.D.C. and 3 others). This study was approved by the Baylor College of Medicine Institutional Review Board, and waiver of consent was granted. The thoracic surgery database that was used included prospectively collected variables similar to those in the Society of Thoracic Surgeons database. Medical records were reviewed retrospectively for additional information. The Charlson Comorbidity Index was used to summarize the patients’ comorbid preoperative conditions.15 For this study, operative mortality was defined as death within 30 days or during the index hospitalization, and 180-day mortality was used as a mid-term outcome measure.
Patients with lung cancer being considered for treatment at our institution were discussed during a multidisciplinary tumor board meeting. Preoperative clinical staging with integrated PET/CT was performed in all patients with lung cancer considered for surgical resection. If PET/CT was performed by the referring institution, it was not usually repeated at the Michael E. DeBakey Veterans Affairs Medical Center. During this study, EBUS-TBNA use for preoperative staging became increasingly routine. Depending on the findings of PET/CT, EBUS mediastinal and hilar nodal biopsy was recommended for selected patients, particularly if PET results showed a central tumor location, a tumor 3 cm or larger, or positivity in the hilum or mediastinum or if the patient was deemed a high-risk surgical candidate. We were also more likely to consider EBUS for adenocarcinoma and for those patients whose primary tumors had low fludeoxyglucose F 18 avidity on PET/CT. Patients were divided into 2 groups (the EBUS group and the PET/CT-only group) based on their preoperative staging.
Real-time EBUS-TBNA with rapid on-site cytopathological evaluation16 was performed before surgery as a separate procedure using conscious sedation or general anesthesia. After airway examination with a conventional fiber-optic bronchoscope, EBUS-TBNA was performed with a convex real-time ultrasonography biopsy scope (XBF-UC260F-OL8; Olympus Ltd). All EBUS-TBNA procedures were performed by the same interventional pulmonologist (R.F.C.). All stations were systematically evaluated, starting with N3 and ending at N1, and were sampled if the minimum lymph node size was 5 mm on the short axis. Lymph node diameter and location, following the Mountain-Dressler map, were documented in all EBUS records.17 Each station was biopsied with a dedicated 22-gauge needle, with at least 3 passes per lymph node. Adequacy of the sample, defined as diagnostic or consisting of lymphatic tissue, was confirmed immediately with cytopathology. If N3 or N2 metastatic mediastinal nodal disease was diagnosed, further sampling of N1 was not performed.
Patients who had N0 or N1 clinical staging after the completion of all preoperative staging studies, including EBUS-TNBA, underwent surgical resection. Patients who underwent at least a lobectomy, with complete mediastinal and hilar lymph node dissection, were included in this study. Patients who had less extensive operations, such as wedge resections and segmentectomy, were excluded. Both open thoracotomy and video-assisted thoracoscopy (VATS) lung resection were performed.
Primary outcome measures were the feasibility, safety, accuracy, and NPV of EBUS-TBNA nodal biopsies for diagnosing N1 and N2 disease. A secondary outcome measure was the rate of nontherapeutic lung resection because of unrecognized N2 disease. We also compared the NPV and rate of nontherapeutic resection between the EBUS group and the PET/CT-only group.
The final pathological findings on complete mediastinal and hilar nodal dissection were used as the gold standard to analyze the accuracy of the preoperative staging with EBUS and PET/CT only. Standard definitions were used to calculate agreement between test results, diagnostic yield, accuracy, sensitivity, specificity, NPV, and positive predictive value (PPV). These were calculated per patient rather than per lymph node station, combining both N1 and N2 findings.
Statistical analysis was performed with a commercially available program (SAS/STAT software, version 9.1.3; SAS Institute Inc). The analysis compared the demographics and preoperative characteristics and the surgical variables of the EBUS group with those of the PET/CT-only group. For the categorical variables, statistical significance between the 2 groups’ frequencies was determined with the χ2 test; whenever the results were invalid or questionable because of small numbers, the Fisher exact test was used instead. For the continuous variables, the median values were considered, and the median 2-sample test was used. The interquartile ranges were also calculated for the continuous variables. The accuracy and NPV were compared using the χ2 test. P < .05 was considered statistically significant.
Of 166 patients who underwent lung resection and nodal dissection during the study period, 62 had EBUS-TBNA (EBUS group), and 104 had PET/CT only (PET/CT-only group). The 2 groups of patients were fairly similar in terms of their preoperative characteristics (Table 1).
The diagnostic results of 62 patients who underwent EBUS-TBNA are summarized in Table 2. The mean (SD) lymph node size sampled was 8.7 (3.8) mm. A total of 258 lymph node stations were sampled. The mean (SD) number of nodal stations sampled in each patient was 4.2 (1.2), with a mean (SD) of 3.5 (0.9) passes per station. For those lymph node stations accessible by EBUS-TBNA, the accuracy was 93.5% (58 of 62), the sensitivity was 66.7% (8 of 12), the specificity was 100.0% (50 of 50), the PPV was 100.0% (8 of 8), and the NPV was 92.6% (50 of 54). Of 54 patients who were staged N0 on the basis of preoperative EBUS-TBNA, 8 patients (3 with N1 disease and 5 with N2 disease) were subsequently upstaged by the nodal dissection (Table 2). However, of these 8 patients, only 4 had nodal disease in stations accessible and sampled by EBUS. These 4 authentic false-negative results showed only partial lymph node involvement by malignancy on final pathological examination. One of the patients upstaged to N2 was ultimately found to have unrecognized small-cell carcinoma, and the patient’s disease had progressed considerably between EBUS and surgical resection. Four patients were upstaged because of lymph node stations inaccessible to EBUS; 2 had positive N2 nodes in station 5, the aortopulmonary window region, which is beyond the reach of EBUS. Two of the patients who were upstaged to N1 had positive nodes in stations 12 and above, which are deeper in the lung tissue and inaccessible by EBUS.
Among 104 patients in the PET/CT-only group who were staged N0 before surgery, 27 results (26.0%) were false negative. Those upstaged by surgical nodal dissection included 14 patients with unsuspected N1 disease and 13 patients with unsuspected N2 disease. The accuracy was 74.0% (77 of 104), and the NPV was 74.0% (77 of 104). Tumor size, central tumor location, adenocarcinoma histologic findings, and status as an interfacility consultation were not significantly associated with likelihood for false-negative PET/CT results.
The NPVs were 92.6% (50 of 54) for the EBUS group and 74.0% (77 of 104) for the PET/CT-only group (P = .004). The nontherapeutic surgical resection rate in patients who were ultimately proved to have N2 disease was 18 of 166 (10.8%) overall, 5 of 62 (8.1%) in the EBUS group, and 13 of 104 (12.5%) in the PET/CT-only group (P = .37).
Table 3 summarizes the surgical variables, pathological results, and outcomes. The VATS lobectomy technique was used in 64.5% (107 of 166 patients). The median number of N2 stations dissected (n = 3) was the same for the EBUS and PET/CT groups (P = .07) (Table 3) and for patients who had open lobectomy and those who underwent the VATS lobectomy technique (P = .26) (data not shown). The EBUS group had more central tumors and more lymph nodes on final pathological examination than the PET/CT-only group.
No significant complications were attributable to the EBUS-TBNA procedure. The overall complication rate was lower in the EBUS group than in the PET/CT-only group. Two operative deaths occurred among 166 patients (1.2%), 1 in each group; both occurred after 30 days but during the index hospitalization and involved respiratory complications. The 180-day mortality was 7 of 166 (4.2%).
Endobronchial ultrasonography-guided transbronchial needle aspiration has quickly become important in lung cancer staging.11,18 Cancer centers have adopted the technique so rapidly because it is minimally invasive and easy to use,19 and it is accurate, with results comparable to those of mediastinoscopy in clinical trials.6- 8 Reports of any serious complication associated with its use are rare. As in our study, the specificity and PPV of EBUS-TBNA have often been reported as 100%; infrequent false-positive results are usually due to primary tumor location.6- 8,18 Therefore, mediastinal metastases can be definitively diagnosed by EBUS-TBNA, with no need for confirmatory mediastinoscopy.
Other advantages of EBUS-TBNA over mediastinoscopy include the ability to accurately diagnose N1 disease in addition to N2 disease. Diagnosis of hilar N1 disease is becoming increasingly important because of the burgeoning use of stereotactic body radiation therapy, which focuses on treating only the primary tumor for patients with stage I lung cancer. Any N1 disease would rule out such radiation therapy as an appropriate option, so EBUS is an important tool if stereotactic body radiation therapy is considered.
Studies of EBUS-TBNA nodal staging have generally been performed in selected tertiary care referral centers with specialized experts and close monitoring of results, but the usefulness of the technique in a typical real-world setting has not been often described. In particular, the literature lacks data on EBUS-TBNA for lung cancer staging in a VA population, and we believe our study is the first to report the use of the technique in a veteran population. Maximizing the accuracy of staging is of utmost importance in the veteran population because the patients are higher risk: many actively smoke or have chronic obstructive pulmonary disease or other comorbidities.
In our VA medical center, we use a general algorithm for lung cancer staging, with EBUS-TBNA as a key component. Properly diagnosing advanced disease before surgery is crucial to avoid nontherapeutic lung resection, and improved surgical selection should also result in improved overall outcomes. In our study, the overall rate of nontherapeutic resection because of unrecognized N2 disease was 10.8%, which is at the lower end of those rates reported in the literature (ranging from approximately 9% to more than 30%)20- 22; we believe this rate validates the effectiveness of our algorithm. The EBUS group had a less than 5% chance (3 of 62 [4.8%]) that EBUS missed N2 disease in a station accessible by the technique. Adding surgical mediastinoscopy would probably not improve this result much because surgical mediastinoscopy samples the same N2 stations as EBUS. The high NPV of EBUS-TBNA in our study was comparable to previous clinical trial results for both EBUS-TBNA and mediastinoscopy.6- 8
However, when the surgical population is high risk, such as the veteran population in this study, minimizing the chance for occult N2 disease is paramount. Therefore, we must consider the potential effect of unrecognized nodal disease in lymph node stations not accessible by EBUS or mediastinoscopy. Selective tissue biopsy of those inaccessible N2 stations (eg, stations 5, 6, 8, or 9) may help to further decrease the rate of nontherapeutic resection.
In our study, occult N2 disease was found in EBUS-inaccessible N2 regions in only 2 additional EBUS-group patients (2 of 62 [3.2%]): both were station-5 aortopulmonary window nodes. Therefore, the potential yield for additional procedures in other regions is probably fairly low when both the PET/CT and the EBUS findings are negative. Additional tissue staging techniques could include VATS biopsy, parasternal anterior mediastinotomy (or Chamberlain procedure), endoscopic ultrasonography–TBNA, transcervical extended mediastinal lymphadenectomy, or extended mediastinoscopy such as video-assisted mediastinal lymphadenectomy. We have used endoscopic ultrasonography–TBNA infrequently; although it can access stations 8 and 9 well, those nodal regions seem to be rarely involved as single-station disease, and we have not found endoscopic ultrasonography to be useful for stations 5 and 6 because these regions are usually a considerable distance from the esophageal endoscopic approach. Mediastinotomy (Chamberlain procedure), video-assisted mediastinal lymphadenectomy, and transcervical extended mediastinal lymphadenectomy are involved surgical procedures with inherent risks.
We have preferred the VATS biopsy approach for station 5 aortopulmonary window and station 6 para-aortic nodes. However, if the patient is a reasonable operative candidate and no mediastinal disease is suspected after thorough staging, including EBUS-TBNA, we generally proceed with lung resection in the same setting. Some evidence suggests that primary surgical resection in the presence of single-station occult N2 disease, especially in the aortopulmonary window, may have a better prognosis than is usual for N2 disease.22 Because we can complete the lobectomy with VATS, no additional invasive incisions are required; if occult N2 disease is found on pathological examination, the patient will probably be able to begin postoperative adjuvant chemotherapy and radiation therapy quickly.
Although the NPV for EBUS-TBNA was good, the NPV for the PET/CT-only group was surprisingly low (74.0%). This is somewhat lower than other reported results,1,2,18 but studies often report only N2 disease rather than both N1 and N2 disease, as we have done herein. Reviewing our data, we were concerned that the low NPV for PET/CT may have been related to studies performed at outside referral centers, where we have no quality control, but we found no statistically significant difference between outside and in-house results. Previously identified risk factors for false-negative PET results were not verified in our study, perhaps because of small sample sizes. The reduced complication rate in the EBUS group may reflect better overall preoperative selection of appropriate surgical candidates in that group than in the PET/CT-only group. Because of the results of this study, we may become increasingly likely to recommend EBUS-TBNA staging, despite negative PET/CT results.
Limitations of our study include those inherent in any retrospective review, as well as small sample size and inherent selection bias. A complete evaluation of the accuracy of the EBUS technique would also require inclusion of those nonsurgical patients with positive mediastinal disease, especially relevant to the calculation of the sensitivity of the test. Although these patients are commonly seen in our practice, we have excluded them from this analysis because they had not undergone nodal dissection to confirm the diagnosis. Our comparison of PET/CT results with EBUS results is inherently biased by preoperative selection according to our algorithm. Therefore, the 2 groups are heterogeneous. Randomization of EBUS vs noninvasive staging with PET/CT only for all patients would be the most reliable indicator of the superiority of EBUS in ruling out nodal disease. However, such a trial is unnecessary because clinicians accept that PET/CT is merely an imaging test; therefore, results should often be supplemented by tissue sampling, including EBUS-TBNA, as deemed appropriate by clinical judgment.
The strengths of our study include having a control group that was concurrent rather than historical, studying a novel patient population by looking specifically at a veteran population, and showing real-world results of both EBUS-TBNA and PET/CT in a VA referral center setting compared with the most reliable gold standard of full surgical nodal dissection. Future studies should include a wider distribution of patients.
In conclusion, the use of EBUS-TBNA for preoperative nodal staging in veterans with lung cancer can be feasible and safe and may be associated with potential benefits, including improved NPV over PET/CT only. The results using this less invasive approach are comparable to those previously reported with mediastinoscopy.
Accepted for Publication: July 9, 2013.
Corresponding Author: Lorraine D. Cornwell, MD, Division of Cardiothoracic Surgery, Michael E. DeBakey Department of Surgery, Baylor College of Medicine, 2002 Holcombe Blvd, Houston, TX 77030 (firstname.lastname@example.org).
Published Online: September 18, 2013. doi:10.1001/jamasurg.2013.3776.
Author Contributions:Study concept and design: Cornwell, Preventza, Casal.
Acquisition of data: Cornwell, Pickrell, Nguyen, Casal.
Analysis and interpretation of data: Cornwell, Bakaeen, Lan, Omer, Pickrell, Nguyen, Casal.
Drafting of the manuscript: Cornwell, Pickrell, Nguyen, Casal.
Critical revision of the manuscript for important intellectual content: Cornwell, Bakaeen, Omer, Preventza, Pickrell, Nguyen, Casal.
Statistical analysis: Cornwell, Pickrell, Nguyen.
Administrative, technical, or material support: Cornwell, Lan, Omer.
Study supervision: Cornwell, Bakaeen, Preventza, Casal.
Conflict of Interest Disclosures: None reported.
Previous Presentation: Presented at the 37th Annual Surgical Symposium of the Association of VA Surgeons; April 22, 2013; Milwaukee, Wisconsin.
Additional Contributions: Katherine H. Simpson, MS, assisted with the statistical analysis.