- Email: [email protected]
EUS-guided FNA for the diagnosis of advanced lung cancer
ORIGINAL ARTICLE
EUS-guided FNA for the diagnosis of advanced lung cancer Mandeep S. Sawhney, MBBS, Robert A. Kratzke, MD, Frank A. Lederle, MD, Amy M. Holmstrom, RN, Douglas B. Nelson, MD, Rosemary F. Kelly, MD Minneapolis, Minnesota, USA
Background: A majority of patients with lung cancer are incurable but are symptomatic and may benefit from palliative therapy. Currently available diagnostic methods are either too risky or unsuccessful in obtaining a tissue diagnosis in up to 30% of patients. Objective: To evaluate the role of EUS-guided FNA in obtaining a tissue diagnosis in patients with advanced lung cancer. Design: Prospective, uncontrolled. Setting: Veterans Administration Medical Center. Subjects and Methods: Patients with suspected lung cancer who were not candidates for curative therapy were prospectively identified. CT scans were reviewed, and patients with lesions considered suitable for sampling by EUS were enrolled. Outcomes were analyzed by a final tissue diagnosis or by serial imaging. Results: Sixty-nine patients met inclusion criteria, of which 3 refused participation. The remaining 66 patients constituted the study population. EUS was technically successful in 95% of patients. A final diagnosis was based on tissue in 63 of 66 patients, serial imaging in 1 of 66 patients, and was unavailable in 2 of 66 patients. A lung mass was sampled in 21 patients, and a metastatic lesion was sampled in 45 patients. EUS made a correct diagnosis in 55 of 64 patients (86%, 95% confidence interval [CI] 77%-93%), including 24% that had undergone a failed prior attempt at diagnosis. The sensitivity of EUS was 86%, and the specificity was 100%. Sampling a metastasis was more likely to yield a correct diagnosis than sampling a lung mass (P Z .02). Two self-limited complications were noted during the study. Conclusions: EUS was an accurate and a safe method for obtaining a tissue diagnosis in patients with advanced incurable lung cancer. (Gastrointest Endosc 2006;63:959-65.)
Approximately 175,000 new cases of lung cancer occur every year in North America alone. Surgical resection is offered with the intent to cure. However, more than 70% of patients are not candidates for a curative resection because of the presence of advanced disease or comorbid illness.1 It is often still important to obtain a tissue diagnosis, because a vast majority of these patients are symptomatic and may benefit from palliative therapy.2 Most experts recommend that the diagnosis of lung cancer be confirmed with a biopsy before patients are subjected to radiation and chemotherapy.3 Also, because of a difference in therapy and prognosis, it is important to
Copyright ª 2006 by the American Society for Gastrointestinal Endoscopy 0016-5107/$32.00 doi:10.1016/j.gie.2005.11.061
differentiate small-cell from non-small-cell lung cancer before commencing treatment. Bronchoscopy is commonly used for the diagnosis of lung cancer and provides adequate tissue in up to 70% of patients.4 In the remaining patients, other methods are required. Transthoracic needle aspiration is an alternative to bronchoscopy but carries a 30% risk of a pneumothorax.5 The risk of a pneumothorax is even higher in patients with chronic obstructive pulmonary disease (COPD), a frequent comorbidity associated with lung cancer.6 Furthermore, lesions in close proximity to mediastinal vascular structures cannot be sampled. Mediastinoscopy and thoracoscopy can also be used to diagnose lung cancer but need to be performed in the operating room, require general anesthesia, are expensive, are invasive, and are associated with a small but finite mortality.7 The development of EUS has made it possible to image and sample, with high accuracy, lesions in the
www.giejournal.org
Volume 63, No. 7 : 2006 GASTROINTESTINAL ENDOSCOPY 959
EUS-FNA for lung cancer diagnosis
mediastinum, the left lobe of the liver, and the adrenal glands, among other locations. The role of EUS in staging mediastinal lymph nodes in lung cancer has been well established. EUS has been shown to be superior to both CT and positron emission tomography in this regard.8-15 However, the value of EUS for making a tissue diagnosis of lung cancer has not been fully explored. Recently published guidelines by the American College of Chest Physicians and the American Society of Clinical Oncology do not mention EUS as a potential option for obtaining tissue to confirm the diagnosis of lung cancer.3,16 Few published studies have made the role of EUSguided FNA (EUS-FNA) in obtaining a pathologic diagnosis of cancer their primary objective. Furthermore, these studies have been small, retrospective, or have limited their analysis to patients with enlarged mediastinal lymph nodes. The ability of EUS to obtain a tissue diagnosis of lung cancer by sampling malignant lung masses that invade the mediastinum and lung cancer metastasis to the liver and adrenals has not been systematically studied. The aim of our study was to evaluate the role of EUSFNA in the pathologic diagnosis of lung cancer.
Sawhney et al
Capsule Summary What is already known on this topic d
d
Many patients with incurable lung cancer may benefit from palliative therapy, but tissue diagnostic methods are either too risky or unsuccessful in up to 30% of patients. EUS is valuable in staging mediastinal lymph nodes in lung cancer but has an unclear role in making a lung-tissue diagnosis.
What this study adds to our knowledge d
d
In a prospective, uncontrolled study, EUS-FNA made a correct diagnosis in 55 of 64 patients (86%), including 24% who had undergone a failed prior attempt. The sensitivity and the specificity of EUS was 86% and 100%, respectively.
FNA was accepted as the final diagnosis. Patients with nonmalignant cytology underwent further diagnostic studies at the discretion of the treating physicians.
EUS-FNA PATIENTS AND METHODS Patients Patients with suspected lung cancer evaluated at our institution from March 2003 to May 2005 were prospectively considered for enrollment in the study. All patients had undergone a complete history, physical examination, routine blood tests, chest radiograph, and a CT of the chest and the upper abdomen. The suspicion for lung cancer was based on the presence and characteristic appearance of lesions in the lung parenchyma and mediastinum on a CT of the chest. Mediastinal lymph nodes larger than 1 cm in short-axis diameter were considered malignant. Only patients who were not suitable candidates for a curative surgical resection in the opinion of a staff thoracic surgeon were considered for the study. Patients who were suitable for surgical resection are subjects of another report. Tumor location, presence of metastasis, comorbid illness, and subject’s function status were considered while determining resectability. The choice of further diagnostic studies was individualized. Based on CT imaging, patients with malignant-appearing lesions located in areas accessible by EUS-FNA were offered participation in the study. This included patients with primary lung tumors that abutted or invaded the mediastinum; mediastinal nodes located in the upper or lower paratracheal mediastinum, aortopulmonary window, subcarina, and paraesophageal mediastinum; or metastasis located in the left lobe of the liver or the adrenal glands. Patients with a prior tissue diagnosis of lung cancer were excluded. The Minneapolis Veterans Administration (VA) Institutional Review Board approved the study. Malignant cytology obtained by EUS960 GASTROINTESTINAL ENDOSCOPY Volume 63, No. 7 : 2006
EUS and EUS-FNA were performed by using the standard technique.17 Mediastinal lymph nodes were characterized according to stations as defined by American Thoracic Society Guidelines.18 A pathologist was available for immediate preliminary interpretation of specimens. Needle passes for aspiration were continued until a preliminary diagnosis of malignancy was made or until the pathologist was satisfied that an adequate sample representative of the lesion had been obtained. Lesions that would confer the highest stage of cancer were sampled first.
Criterion standard The final diagnosis was based on cytology (obtained by EUS-FNA) or histology (biopsy specimen obtained at surgery, bronchoscopy, or via CT guidance). Patients were also included in the final analysis if follow-up CT, done at least 1 year after EUS, was available for comparison. Patients with lesions that showed no change or a decrease in size were designated as having a benign disorder. An increase in size or number of lesions was considered evidence of malignancy for the purpose of study reporting.
Data analysis The purpose of EUS in this study population was to obtain a tissue diagnosis; therefore, the unit of analysis was a patient (and not a lesion). A nonmalignant EUS-FNA was considered to be a false negative if a subsequent biopsy of any lesion showed the presence of cancer. Furthermore, results were analyzed as ‘‘intent to diagnose.’’ For example, if a patient underwent EUS with the intent to sample a lung mass, the patient was considered as www.giejournal.org
Sawhney et al
having a false-negative EUS if the lung mass could not be imaged (and hence sampled) by EUS. Cytology that was reported as ‘‘suspicious for malignancy,’’ or ‘‘suggestive but not diagnostic of malignancy’’ was considered nonmalignant, because it did not prevent further diagnostic testing. Sensitivity was calculated as the proportion of patients with malignancy with a positive test, specificity was the proportion of patients without malignancy with a negative test, accuracy was the proportion of patients with correct test results, positive predictive value was the proportion of patients with a positive test who had malignancy, and negative predictive value was the proportion of patients with a negative test without malignancy. Likelihood ratios were also calculated: a positive likelihood ratio was defined as sensitivity/(1 specificity) and expressed the increase in the odds of having the disease when the test was positive, and a negative likelihood ratio was defined as (1 sensitivity)/specificity and expressed the decrease in the odds of having the disease when the test was negative. To avoid division by 0, values for true positives, false positives, true negatives, and false negatives were increased by 0.5 when computing likelihood ratios.19 An exact binomial distribution was used to calculate confidence intervals (CI). Because cytology is accepted as a criterion standard for the diagnosis of cancer, confidence intervals for the specificity and the positive predictive value were not reported. The Fisher test was used to perform significance test for categorical variables. Our previous experience showed a significant difference between metastasis and primary lung masses in regard to technical aspects of EUS and cytologic yield. We, therefore, a priori, decided to analyze data from lung masses separately from metastatic lesions. Based on CT imaging, mass lesions involving the lung parenchyma were considered primary lung masses, while discrete lesions located in the mediastinum that did not disrupt the pleura were considered mediastinal lymph nodes. Mediastinal lymph nodes, and liver and adrenal gland lesions were considered metastatic lesions for the purpose of data analysis.
EUS-FNA for lung cancer diagnosis
During the specified study period, 69 patients met study criteria; 3 of the 69 declined further diagnostic procedures (Fig. 1). The remaining 66 patients constituted the study population. The baseline characteristics of these patients are given in Table 1. The final diagnosis was based on tissue in 63 of 66 patients and on serial imaging in 1 patient. Two patients did not undergo further evaluation after EUS-FNA found no evidence of malignancy. One patient opted for hospice care, and the other received palliative therapy without a tissue diagnosis. These 2 patients were excluded from the final data analysis, because no
conclusive diagnosis was available (therefore, the final data analysis was conducted on 64 subjects). EUS was attempted in 66 patients and was technically successful in 63 patients (95%, 95% CI 87%-99%). In 21 patients (32%), the intended target of EUS-FNA was a primary lung mass abutting or invading the mediastinum (Table 2). The median short-axis diameter of the lung mass measured on a CT was 4 cm. A median of 5 (range, 3-8) EUS-guided needle passes were made. All 3 cases of EUS technical failure occurred in this group. The echoendoscope could not be advanced into the esophagus of 1 patient because of the presence of a radiation-induced cricopharyngeal stricture. In 2 patients, primary lung masses, located in the azygous lobe and the right lower lobe, could not be imaged for sampling by EUS (despite the appearance to the contrary on CT). Data from these 3 patients were included in the final analysis. In the remaining 45 patients (68%), EUS-FNA was performed for a metastatic lesion and was technically successful in all cases (Table 2). The median short-axis and long-axis diameters for mediastinal lymph nodes sampled were 15 mm (range, 6-35 mm) and 25 mm (range, 12-50 mm), respectively. A median of 5 (range, 3-8) EUS-guided needle passes were made. The median short-axis and long-axis diameters for liver and adrenal metastasis sampled were 15 mm (range, 6-35 mm) and 25 mm (range, 12-50 mm), respectively. A median of 4.5 (range, 3-7) EUS-guided needle passes were made. The prevalence of cancer in the study sample was 97%. EUS-FNA made a correct diagnosis in 55 of 64 patients (86%, 95% CI 74%-93%). Of the 53 patients with cancer, 44 patients were diagnosed with non-small-cell lung cancer and 9 with small-cell lung cancer. In all cases, EUSFNA obtained adequate tissue for a conclusive diagnosis and cytologic subclassification of cancer. Before referral, 24% of these patients had undergone at least one unsuccessful attempt at tissue diagnosis before EUS was performed. A brochoscopy had been attempted in 12 patients, a diagnostic thoracentesis in 2 patients, and a CT-guided biopsy in 1 patient. Overall, EUS-FNA had a sensitivity of 86% (95% CI 74%-93%), a specificity of 100%, a positive predictive value of 100%, a negative predictive value of 18% (95% CI 2%-52%), a positive likelihood ratio 5.1 (95% CI 1.2-48.2), and a negative likelihood ratio of 0.2 (95% CI 0.1-0.5). In 2 of 64 patients, a correct diagnosis of a benign lesion was made by EUS-FNA. One patient, with a large lung mass, was scheduled for a CT-guided biopsy after a nonmalignant EUS-FNA. This was deferred, because marked improvement of the lesion was noted on repeat imaging. The patient was treated with oral antibiotics, with complete resolution of the lesion on follow-up CT. The final (presumptive) diagnosis in this patient was pneumonia, with an atypical radiographic appearance. The other patient was confirmed to have nonspecific granulomatous inflammation by a surgical biopsy.
www.giejournal.org
Volume 63, No. 7 : 2006 GASTROINTESTINAL ENDOSCOPY 961
RESULTS
EUS-FNA for lung cancer diagnosis
Sawhney et al
Figure 1. Study flow sheet.
In 9 of 64 patients (14%, 95% CI 7%-25%), EUS failed to obtain a correct diagnosis. This included the 3 patients in whom EUS was technically unsuccessful. A final diagnosis of small-cell lung cancer (n Z 2), bronchoalveolar cancer (n Z 2), and non-small-cell lung cancer (n Z 5) was made in these patients by a video-assisted thoracoscopy (n Z 1), mediastinoscopy (n Z 3), CT-guided biopsy (n Z 3), and bronchoscopy (n Z 2). In the subgroup analysis, EUS-FNA performance was compared for sampling metastasis with primary lung masses (Table 3). EUS-FNA made a correct diagnosis in 42 of 45 patients where metastases were sampled, compared with 13 of 19 patients where a primary lung mass was sampled (accuracy, 93% vs 68%; P Z.02). Also, metastases were associated with a higher chance of technical success when compared with lung masses (100% vs 84%, P Z.02). Two complications occurred after sampling of a lung mass. One patient developed severe chest pain and required admission to the hospital for further evaluation. No evidence of a pneumothorax, injury to mediastinal
organs, or an acute cardiac event was found. The pain resolved spontaneously, and the patient was discharged within 24 hours. Self-limited hemoptysis was noted in the other patient. No complications were associated with FNA of metastatic lesions.
962 GASTROINTESTINAL ENDOSCOPY Volume 63, No. 7 : 2006
www.giejournal.org
DISCUSSION We found EUS to be a safe and accurate method for diagnosing lung cancer in patients with advanced, inoperable disease. A correct diagnosis was made in more than 85% of patients referred for EUS. The accuracy and rate of technical success of EUS was significantly higher for metastatic lesions than for primary lung masses. Although several studies have reported on the diagnostic value of EUS-FNA in lung cancer, few have made it the primary end point of their analysis. Fritscher-Ravens et al20 performed EUS on 35 patients with suspected lung cancer in whom bronchoscopy had failed to establish a diagnosis.
Sawhney et al
EUS-FNA for lung cancer diagnosis
TABLE 1. Baseline characteristics
TABLE 2. Accuracy of EUS-FNA by location of lesions
No. subjects
66
Age, y (range)
67 (43-82)
Sex, % of men
97
Location of lung mass, %* Upper lobe
63
Middle lobe
8
Lower lobe
15
Azygous lobe
2
Bilateral
8
Indeterminable or absent
5
No. patients,* (n Z 64)
Accuracy, %
Lung mass
19 (30%)
68
Mediastinal lymph nodes
39 (61%)
92
25
96
Paratracheal
6
83
Aortopulmonary window
6
83
Paraesophageal
2
100
Liver
4 (6%)
100
Adrenal gland
2 (3%)
100
Location
Subcarinal
*Two patients without a conclusive final diagnosis were excluded.
Inoperable, %*,y Locally advanced disease
17
Metastatic disease
20
Comorbid illnesses
47
Other causes
18
Only mediastinal lymph nodes were sampled, and a tissue diagnosis of cancer was obtained in 25 patients. EUS had a sensitivity of 96%, a specificity of 100%, and an accuracy of 97%. In another study by the same investigators, EUS made a correct diagnosis of cancer in 16 of 17 patients.21 In a study of 27 patients with enlarged mediastinal nodes, Silvestri et al22 found that EUS had a sensitivity of 89% and a specificity of 100% for the diagnosis of cancer. Varadarajulu et al23 published a retrospective case series of 18 patients with lung masses located adjacent to the mediastinum. In all 18 patients, a tissue diagnosis of malignancy was obtained by EUS. No complications were noted in this study. Our study is not directly comparable with any of these but does confirm some of the earlier findings. EUS was highly accurate and safe in patients with suspected malignant mediastinal lymph nodes. In all cases of malignancy, adequate material to enable a pathologist to distinguish between small-cell and non-small-cell cancer was obtained. The accuracy of EUS-guided sampling of lung masses in our study, however, was lower than that reported by Varadarajulu et al.23 If we exclude from analysis patients in whom EUS was technically unsuccessful, our results are more in agreement (accuracy, 81%; 95% CI 56%-94%). For patients with advanced lung cancer, EUS appears to be complementary to bronchoscopy in making a tissue diagnosis. Although bronchoscopy is very accurate for large, central lung lesions (sensitivity, 88%), it is less so for
smaller peripheral lesions (sensitivity, 33%).24 For lesions such as mediastinal nodes that do not have an endobronchial component, the sensitivity of bronchoscopy can be improved by performing transbronchial needle aspiration but remains low (sensitivity, 56%).24 EUS, on the other hand, is highly accurate and safe for mediastinal lymph nodes and peripheral lung masses adjacent to the mediastinum. EUS also has the advantage of providing a tissue diagnosis and confirming advanced cancer stage at the same time. Furthermore, EUS can sample lesions located in the aortopulmonary window, the paraesophageal mediastinum, the liver, and the adrenal glands, locations that cannot be accessed with either bronchoscopy or transbronchial needle aspiration. Transthoracic needle aspiration is the other most often used modality to diagnose lung cancer. The main advantage of EUS over transthoracic needle aspiration is fewer procedure-related complications. On average, 5% of patients undergoing transthoracic needle aspiration develop hemoptysis.25 Even though this is usually mild and selflimited, 2 cases of death from hemorrhage have been reported.25 The rate of pneumothorax after transthoracic needle aspiration is 30%. This is higher for patients with COPD (50%) or when multiple needle passes are made (40%).6 A chest tube is required in 5% to 25% of patients and usually necessitates admission to the hospital.26 EUSguided sampling has a significantly lower rate of complications. In our study, 2 self-limited complications were noted, and only 1 of 66 patients required admission to the hospital. In a large, prospective, multicenter study of 457 patients, 5 complications were noted as a result of EUS-guided sampling (0.5%, 95% CI 0.1%-0.8%).27 All complications were nonfatal and involved mainly pancreas cysts. A recently published study showed that EUS could also be safely performed in critically ill patients.28 Several limitations of our study warrant further discussion. First, for 1 patient, we did not have tissue confirmation
www.giejournal.org
Volume 63, No. 7 : 2006 GASTROINTESTINAL ENDOSCOPY 963
*Does not equal 100% because of rounding error. yReason patients were not considered for curative surgery.
EUS-FNA for lung cancer diagnosis
Sawhney et al
TABLE 3. Comparison of lung masses with metastatic lesions Lung mass
Metastasis
19
45
Technical success, % (95% CI)
84 (60-97)
100 (93-100)
Accuracy, % (95% CI)
68 (43-87)
93 (81-99)
Sensitivity, % (95% CI)
67 (40-87)
93 (81-99)
Specificity, % y
100
100
PPV, % y
100
100
No. subjects*
ple size but also allowed us to evaluate and report on the role of EUS in these difficult-to-diagnose patients. In conclusion, EUS-FNA is a safe and effective means of obtaining tissue in patients with advanced lung cancer. If available, it should be included along with bronchoscopy and transthoracic needle biopsy in the armamentarium of diagnostic procedure. Studies directly comparing EUS to brochoscopy and transthoracic needle aspiration are needed to further clarify its role.
REFERENCES
of the final diagnosis. We categorized the patient as having benign disease based upon complete radiologic resolution of the abnormal lesion. The strategy of using radiologic follow-up as a criterion standard for the diagnosis of cancer has been used by others and is unlikely to have misclassified this patient.27,29 Second, nearly all subjects in our study were men, reflecting the VA population seen at our institution. Third, the prevalence of lung cancer in our study sample was high (97%), limiting somewhat the generalizability of our results. The high prevalence of cancer was not unexpected, because all patients included in the study were suspected of having advanced inoperable disease. The fact that 3% of patients did not have cancer and that 14% had smallcell lung cancer underscores the importance of obtaining a tissue diagnosis before starting palliative therapy. Fourth, patients were enrolled in our study only if they were not candidates for a curative surgical resection and if malignant-appearing lesions on a CT were accessible by EUS. Thus, our results are applicable only to this select group of patients. Fifth, our study included patients who were referred to us by other centers, some of whom had already undergone failed attempts at tissue diagnosis. Because these patients did not represent a random sample of all patients with lung cancer, we were not able to report a ‘‘denominator’’ from which our patients were selected, nor were we able to specify a hierarchy of diagnostic tests that subjects should have undergone upon entry into the study. However, including these patients not only increased our sam-
1. Fry W, Menck H, Winchester D. The National Cancer Data Base report on lung cancer. Cancer 1996;77:1947-55. 2. Kvale P, Simoff M, Prakash U. Lung cancer. Palliative care. Chest 2003;123:284S-311S. 3. Pfister D, Johnson D, Azzoli C, et al. American Society of Clinical Oncology treatment of unresectable non-small-cell lung cancer guideline: update 2003. J Clin Oncol 2004;22:330-53. 4. Arroliga A, Matthay R. The role of bronchoscopy in lung cancer. Clin Chest Med 1993;14:87-98. 5. Zarbo R, Fenoglio-Preiser C. Interinstitutional database for comparison of performance in lung fine-needle aspiration cytology. A College of American Pathologists Q-Probe Study of 5264 cases with histologic correlation. Arch Pathol Lab Med 1992;116:463-70. 6. Sanders C. Transthoracic needle aspiration. Clin Chest Med 1992;13: 11-6. 7. Detterbeck F, DeCamp MJ, Kohman L, et al. Lung cancer. Invasive staging: the guidelines. Chest 2003;123:167S-75S. 8. Wallace M, Ravenel J, Block M, et al. Endoscopic ultrasound in lung cancer patients with a normal mediastinum on computed tomography. Ann Thorac Surg 2004;77:1763-8. 9. Wallace M, Fritscher-Ravens A, Savides T. Endoscopic ultrasound for the staging of non-small-cell lung cancer. Endoscopy 2003;35:606-10. 10. Gress F, Savides T, Sandler A, et al. Endoscopic ultrasonography, fineneedle aspiration biopsy guided by endoscopic ultrasonography, and computed tomography in the preoperative staging of non-small-cell lung cancer: a comparison study. Ann Intern Med 1997;127:604-12. 11. Fritscher-Ravens A, Davidson B, Hauber H, et al. Endoscopic ultrasound, positron emission tomography, and computerized tomography for lung cancer. Am J Respir Crit Care Med 2003;168:1293-7. 12. Fritscher-Ravens A, Bohuslavizki K, Brandt L, et al. Mediastinal lymph node involvement in potentially resectable lung cancer: comparison of CT, positron emission tomography, and endoscopic ultrasonography with and without fine-needle aspiration. Chest 2003;123:442-51. 13. Annema J, Hoekstra O, Smit E, et al. Towards a minimally invasive staging strategy in NSCLC: analysis of PET positive mediastinal lesions by EUS-FNA. Lung Cancer 2004;44:53-60. 14. Eloubeidi M, Cerfolio R, Chen V, et al. Endoscopic ultrasound-guided fine needle aspiration of mediastinal lymph node in patients with suspected lung cancer after positron emission tomography and computed tomography scans. Ann Thorac Surg 2005;79:263-8. 15. Kramer H, van Putten J, Post W, et al. Oesophageal endoscopic ultrasound with fine needle aspiration improves and simplifies the staging of lung cancer. Thorax 2004;59:596-601. 16. Rivera M, Detterbeck F, Mehta A. Diagnosis of lung cancer: the guidelines. Chest 2003;123:129S-36S. 17. Wiersema M, Wiersema L, Khusro Q, et al. Combined endosonography and fine-needle aspiration cytology in the evaluation of gastrointestinal lesions. Gastrointest Endosc 1994;40:199-206. 18. Mountain C, Dresler C. Regional lymph node classification for lung cancer staging. Chest 1997;111:1718-23. 19. Hasselblad V, Hedges L. Meta-analysis of screening and diagnostic tests. Psychol Bull 1995;117:167-78.
964 GASTROINTESTINAL ENDOSCOPY Volume 63, No. 7 : 2006
www.giejournal.org
NPV, % (95% CI)
14 (0-58)
25 (1-80)
LR C, % (95% CI)
1.7 (0.4-17)
3.7 (1.1-34.4)
LR , % (95% CI)
0.8 (0.5-3.0)
0.1 (0.1-0.5)
Misclassification rate, % (95% CI)
31 (12-57)
7 (0-18)
11
0
Complications, %
PPV, Positive predictive value; NPV, negative predictive value; LR C, likelihood ratio for a positive test; LR , likelihood ratio for a negative test. *Two patients without a final diagnosis are excluded from analysis. yBecause cytology is accepted as a criterion standard for the diagnosis of lung cancer, CIs are not reported.
Sawhney et al
EUS-FNA for lung cancer diagnosis
20. Fritscher-Ravens A, Soehendra N, Schirrow L, et al. Role of transesophageal endosonography-guided fine-needle aspiration in the diagnosis of lung cancer. Chest 2000;117:339-45. 21. Fritscher-Ravens A, Petrasch S, Reinacher-Schick A, et al. Diagnostic value of endoscopic ultrasonography-guided fine-needle aspiration cytology of mediastinal masses in patients with intrapulmonary lesions and nondiagnostic bronchoscopy. Respiration 1999;66:150-5. 22. Silvestri G, Hoffman B, Bhutani M, et al. Endoscopic ultrasound with fine-needle aspiration in the diagnosis and staging of lung cancer. Ann Thorac Surg 1996;61:1441-5; discussion 1445-6. 23. Varadarajulu S, Hoffman B, Hawes R, et al. EUS-guided FNA of lung masses adjacent to or abutting the esophagus after unrevealing CTguided biopsy or bronchoscopy. Gastrointest Endosc 2004;60:293-7. 24. Schreiber G, McCrory D. Performance characteristics of different modalities for diagnosis of suspected lung cancer: summary of published evidence. Chest 2003;123:115S-28S. 25. Detterbeck F, Rivera M. Clinical presentation and diagnosis. In: Detterbeck F, Socinski M, Rivera M, et al, editors. Diagnosis and Treatment of Lung Cancer: An Evidence Based Approach. 1st ed. Philadelphia: W.B. Saunders; 2001. p. 45-72. 26. Zwischenberger J, vanSonnenberger E, Alpard SK. Interventional radiology in the chest. In: Hazelrigg SR, Bashar M, Landreneau RJ, et al, editors. Minimal Access Cardiothoracic Surgery. 1st ed. Philadelphia: W.B. Saunders; 2000. p. 388-97.
27. Wiersema MJ, Vilmann P, Giovannini M, et al. Endosonography-guided fine-needle aspiration biopsy: diagnostic accuracy and complication assessment. Gastroenterology 1997;112:1087-95. 28. Fritscher-Ravens A, Sriram P, Pothman W, et al. Bedside endosonography and endosonography-guided fine-needle aspiration in critically ill patients: a way out of the deadlock? Endoscopy 2000;32:425-7. 29. Larsen S, Krasnik M, Vilmann P, et al. Endoscopic ultrasound guided biopsy of mediastinal lesions has a major impact on patient management. Thorax 2002;57:98-103.
www.giejournal.org
Volume 63, No. 7 : 2006 GASTROINTESTINAL ENDOSCOPY 965
Received August 2, 2005. Accepted November 8, 2005. Current affiliations: Section of Gastroenterology, (Drs Sawhney and Nelson), Center for Epidemiological and Clinical Research (Drs Sawhney and Lederle), Minneapolis VA Medical Center; Department of Medicine (Drs Sawhney, Kratzke, Lederle, and Nelson), Section of Hematologyd Oncology and Transplantation (Dr Kratzke); and Section of Cardiothoracic Surgery (Drs Holmstrom and Kelly), University of Minnesota, Minneapolis, Minnesota, USA. M. S. Sawhney is supported, in part, by VA Clinical Science R&D Service (Grant no. 04S-CRCOE-001). Reprint requests: Mandeep S. Sawhney, MBBS, Section of Gastroenterology (111 D), One Veterans Dr, Minneapolis, MN 55417.
EUS-guided FNA for the diagnosis of advanced lung cancer Mandeep S. Sawhney, MBBS, Robert A. Kratzke, MD, Frank A. Lederle, MD, Amy M. Holmstrom, RN, Douglas B. Nelson, MD, Rosemary F. Kelly, MD Minneapolis, Minnesota, USA
Background: A majority of patients with lung cancer are incurable but are symptomatic and may benefit from palliative therapy. Currently available diagnostic methods are either too risky or unsuccessful in obtaining a tissue diagnosis in up to 30% of patients. Objective: To evaluate the role of EUS-guided FNA in obtaining a tissue diagnosis in patients with advanced lung cancer. Design: Prospective, uncontrolled. Setting: Veterans Administration Medical Center. Subjects and Methods: Patients with suspected lung cancer who were not candidates for curative therapy were prospectively identified. CT scans were reviewed, and patients with lesions considered suitable for sampling by EUS were enrolled. Outcomes were analyzed by a final tissue diagnosis or by serial imaging. Results: Sixty-nine patients met inclusion criteria, of which 3 refused participation. The remaining 66 patients constituted the study population. EUS was technically successful in 95% of patients. A final diagnosis was based on tissue in 63 of 66 patients, serial imaging in 1 of 66 patients, and was unavailable in 2 of 66 patients. A lung mass was sampled in 21 patients, and a metastatic lesion was sampled in 45 patients. EUS made a correct diagnosis in 55 of 64 patients (86%, 95% confidence interval [CI] 77%-93%), including 24% that had undergone a failed prior attempt at diagnosis. The sensitivity of EUS was 86%, and the specificity was 100%. Sampling a metastasis was more likely to yield a correct diagnosis than sampling a lung mass (P Z .02). Two self-limited complications were noted during the study. Conclusions: EUS was an accurate and a safe method for obtaining a tissue diagnosis in patients with advanced incurable lung cancer. (Gastrointest Endosc 2006;63:959-65.)
Approximately 175,000 new cases of lung cancer occur every year in North America alone. Surgical resection is offered with the intent to cure. However, more than 70% of patients are not candidates for a curative resection because of the presence of advanced disease or comorbid illness.1 It is often still important to obtain a tissue diagnosis, because a vast majority of these patients are symptomatic and may benefit from palliative therapy.2 Most experts recommend that the diagnosis of lung cancer be confirmed with a biopsy before patients are subjected to radiation and chemotherapy.3 Also, because of a difference in therapy and prognosis, it is important to
Copyright ª 2006 by the American Society for Gastrointestinal Endoscopy 0016-5107/$32.00 doi:10.1016/j.gie.2005.11.061
differentiate small-cell from non-small-cell lung cancer before commencing treatment. Bronchoscopy is commonly used for the diagnosis of lung cancer and provides adequate tissue in up to 70% of patients.4 In the remaining patients, other methods are required. Transthoracic needle aspiration is an alternative to bronchoscopy but carries a 30% risk of a pneumothorax.5 The risk of a pneumothorax is even higher in patients with chronic obstructive pulmonary disease (COPD), a frequent comorbidity associated with lung cancer.6 Furthermore, lesions in close proximity to mediastinal vascular structures cannot be sampled. Mediastinoscopy and thoracoscopy can also be used to diagnose lung cancer but need to be performed in the operating room, require general anesthesia, are expensive, are invasive, and are associated with a small but finite mortality.7 The development of EUS has made it possible to image and sample, with high accuracy, lesions in the
www.giejournal.org
Volume 63, No. 7 : 2006 GASTROINTESTINAL ENDOSCOPY 959
EUS-FNA for lung cancer diagnosis
mediastinum, the left lobe of the liver, and the adrenal glands, among other locations. The role of EUS in staging mediastinal lymph nodes in lung cancer has been well established. EUS has been shown to be superior to both CT and positron emission tomography in this regard.8-15 However, the value of EUS for making a tissue diagnosis of lung cancer has not been fully explored. Recently published guidelines by the American College of Chest Physicians and the American Society of Clinical Oncology do not mention EUS as a potential option for obtaining tissue to confirm the diagnosis of lung cancer.3,16 Few published studies have made the role of EUSguided FNA (EUS-FNA) in obtaining a pathologic diagnosis of cancer their primary objective. Furthermore, these studies have been small, retrospective, or have limited their analysis to patients with enlarged mediastinal lymph nodes. The ability of EUS to obtain a tissue diagnosis of lung cancer by sampling malignant lung masses that invade the mediastinum and lung cancer metastasis to the liver and adrenals has not been systematically studied. The aim of our study was to evaluate the role of EUSFNA in the pathologic diagnosis of lung cancer.
Sawhney et al
Capsule Summary What is already known on this topic d
d
Many patients with incurable lung cancer may benefit from palliative therapy, but tissue diagnostic methods are either too risky or unsuccessful in up to 30% of patients. EUS is valuable in staging mediastinal lymph nodes in lung cancer but has an unclear role in making a lung-tissue diagnosis.
What this study adds to our knowledge d
d
In a prospective, uncontrolled study, EUS-FNA made a correct diagnosis in 55 of 64 patients (86%), including 24% who had undergone a failed prior attempt. The sensitivity and the specificity of EUS was 86% and 100%, respectively.
FNA was accepted as the final diagnosis. Patients with nonmalignant cytology underwent further diagnostic studies at the discretion of the treating physicians.
EUS-FNA PATIENTS AND METHODS Patients Patients with suspected lung cancer evaluated at our institution from March 2003 to May 2005 were prospectively considered for enrollment in the study. All patients had undergone a complete history, physical examination, routine blood tests, chest radiograph, and a CT of the chest and the upper abdomen. The suspicion for lung cancer was based on the presence and characteristic appearance of lesions in the lung parenchyma and mediastinum on a CT of the chest. Mediastinal lymph nodes larger than 1 cm in short-axis diameter were considered malignant. Only patients who were not suitable candidates for a curative surgical resection in the opinion of a staff thoracic surgeon were considered for the study. Patients who were suitable for surgical resection are subjects of another report. Tumor location, presence of metastasis, comorbid illness, and subject’s function status were considered while determining resectability. The choice of further diagnostic studies was individualized. Based on CT imaging, patients with malignant-appearing lesions located in areas accessible by EUS-FNA were offered participation in the study. This included patients with primary lung tumors that abutted or invaded the mediastinum; mediastinal nodes located in the upper or lower paratracheal mediastinum, aortopulmonary window, subcarina, and paraesophageal mediastinum; or metastasis located in the left lobe of the liver or the adrenal glands. Patients with a prior tissue diagnosis of lung cancer were excluded. The Minneapolis Veterans Administration (VA) Institutional Review Board approved the study. Malignant cytology obtained by EUS960 GASTROINTESTINAL ENDOSCOPY Volume 63, No. 7 : 2006
EUS and EUS-FNA were performed by using the standard technique.17 Mediastinal lymph nodes were characterized according to stations as defined by American Thoracic Society Guidelines.18 A pathologist was available for immediate preliminary interpretation of specimens. Needle passes for aspiration were continued until a preliminary diagnosis of malignancy was made or until the pathologist was satisfied that an adequate sample representative of the lesion had been obtained. Lesions that would confer the highest stage of cancer were sampled first.
Criterion standard The final diagnosis was based on cytology (obtained by EUS-FNA) or histology (biopsy specimen obtained at surgery, bronchoscopy, or via CT guidance). Patients were also included in the final analysis if follow-up CT, done at least 1 year after EUS, was available for comparison. Patients with lesions that showed no change or a decrease in size were designated as having a benign disorder. An increase in size or number of lesions was considered evidence of malignancy for the purpose of study reporting.
Data analysis The purpose of EUS in this study population was to obtain a tissue diagnosis; therefore, the unit of analysis was a patient (and not a lesion). A nonmalignant EUS-FNA was considered to be a false negative if a subsequent biopsy of any lesion showed the presence of cancer. Furthermore, results were analyzed as ‘‘intent to diagnose.’’ For example, if a patient underwent EUS with the intent to sample a lung mass, the patient was considered as www.giejournal.org
Sawhney et al
having a false-negative EUS if the lung mass could not be imaged (and hence sampled) by EUS. Cytology that was reported as ‘‘suspicious for malignancy,’’ or ‘‘suggestive but not diagnostic of malignancy’’ was considered nonmalignant, because it did not prevent further diagnostic testing. Sensitivity was calculated as the proportion of patients with malignancy with a positive test, specificity was the proportion of patients without malignancy with a negative test, accuracy was the proportion of patients with correct test results, positive predictive value was the proportion of patients with a positive test who had malignancy, and negative predictive value was the proportion of patients with a negative test without malignancy. Likelihood ratios were also calculated: a positive likelihood ratio was defined as sensitivity/(1 specificity) and expressed the increase in the odds of having the disease when the test was positive, and a negative likelihood ratio was defined as (1 sensitivity)/specificity and expressed the decrease in the odds of having the disease when the test was negative. To avoid division by 0, values for true positives, false positives, true negatives, and false negatives were increased by 0.5 when computing likelihood ratios.19 An exact binomial distribution was used to calculate confidence intervals (CI). Because cytology is accepted as a criterion standard for the diagnosis of cancer, confidence intervals for the specificity and the positive predictive value were not reported. The Fisher test was used to perform significance test for categorical variables. Our previous experience showed a significant difference between metastasis and primary lung masses in regard to technical aspects of EUS and cytologic yield. We, therefore, a priori, decided to analyze data from lung masses separately from metastatic lesions. Based on CT imaging, mass lesions involving the lung parenchyma were considered primary lung masses, while discrete lesions located in the mediastinum that did not disrupt the pleura were considered mediastinal lymph nodes. Mediastinal lymph nodes, and liver and adrenal gland lesions were considered metastatic lesions for the purpose of data analysis.
EUS-FNA for lung cancer diagnosis
During the specified study period, 69 patients met study criteria; 3 of the 69 declined further diagnostic procedures (Fig. 1). The remaining 66 patients constituted the study population. The baseline characteristics of these patients are given in Table 1. The final diagnosis was based on tissue in 63 of 66 patients and on serial imaging in 1 patient. Two patients did not undergo further evaluation after EUS-FNA found no evidence of malignancy. One patient opted for hospice care, and the other received palliative therapy without a tissue diagnosis. These 2 patients were excluded from the final data analysis, because no
conclusive diagnosis was available (therefore, the final data analysis was conducted on 64 subjects). EUS was attempted in 66 patients and was technically successful in 63 patients (95%, 95% CI 87%-99%). In 21 patients (32%), the intended target of EUS-FNA was a primary lung mass abutting or invading the mediastinum (Table 2). The median short-axis diameter of the lung mass measured on a CT was 4 cm. A median of 5 (range, 3-8) EUS-guided needle passes were made. All 3 cases of EUS technical failure occurred in this group. The echoendoscope could not be advanced into the esophagus of 1 patient because of the presence of a radiation-induced cricopharyngeal stricture. In 2 patients, primary lung masses, located in the azygous lobe and the right lower lobe, could not be imaged for sampling by EUS (despite the appearance to the contrary on CT). Data from these 3 patients were included in the final analysis. In the remaining 45 patients (68%), EUS-FNA was performed for a metastatic lesion and was technically successful in all cases (Table 2). The median short-axis and long-axis diameters for mediastinal lymph nodes sampled were 15 mm (range, 6-35 mm) and 25 mm (range, 12-50 mm), respectively. A median of 5 (range, 3-8) EUS-guided needle passes were made. The median short-axis and long-axis diameters for liver and adrenal metastasis sampled were 15 mm (range, 6-35 mm) and 25 mm (range, 12-50 mm), respectively. A median of 4.5 (range, 3-7) EUS-guided needle passes were made. The prevalence of cancer in the study sample was 97%. EUS-FNA made a correct diagnosis in 55 of 64 patients (86%, 95% CI 74%-93%). Of the 53 patients with cancer, 44 patients were diagnosed with non-small-cell lung cancer and 9 with small-cell lung cancer. In all cases, EUSFNA obtained adequate tissue for a conclusive diagnosis and cytologic subclassification of cancer. Before referral, 24% of these patients had undergone at least one unsuccessful attempt at tissue diagnosis before EUS was performed. A brochoscopy had been attempted in 12 patients, a diagnostic thoracentesis in 2 patients, and a CT-guided biopsy in 1 patient. Overall, EUS-FNA had a sensitivity of 86% (95% CI 74%-93%), a specificity of 100%, a positive predictive value of 100%, a negative predictive value of 18% (95% CI 2%-52%), a positive likelihood ratio 5.1 (95% CI 1.2-48.2), and a negative likelihood ratio of 0.2 (95% CI 0.1-0.5). In 2 of 64 patients, a correct diagnosis of a benign lesion was made by EUS-FNA. One patient, with a large lung mass, was scheduled for a CT-guided biopsy after a nonmalignant EUS-FNA. This was deferred, because marked improvement of the lesion was noted on repeat imaging. The patient was treated with oral antibiotics, with complete resolution of the lesion on follow-up CT. The final (presumptive) diagnosis in this patient was pneumonia, with an atypical radiographic appearance. The other patient was confirmed to have nonspecific granulomatous inflammation by a surgical biopsy.
www.giejournal.org
Volume 63, No. 7 : 2006 GASTROINTESTINAL ENDOSCOPY 961
RESULTS
EUS-FNA for lung cancer diagnosis
Sawhney et al
Figure 1. Study flow sheet.
In 9 of 64 patients (14%, 95% CI 7%-25%), EUS failed to obtain a correct diagnosis. This included the 3 patients in whom EUS was technically unsuccessful. A final diagnosis of small-cell lung cancer (n Z 2), bronchoalveolar cancer (n Z 2), and non-small-cell lung cancer (n Z 5) was made in these patients by a video-assisted thoracoscopy (n Z 1), mediastinoscopy (n Z 3), CT-guided biopsy (n Z 3), and bronchoscopy (n Z 2). In the subgroup analysis, EUS-FNA performance was compared for sampling metastasis with primary lung masses (Table 3). EUS-FNA made a correct diagnosis in 42 of 45 patients where metastases were sampled, compared with 13 of 19 patients where a primary lung mass was sampled (accuracy, 93% vs 68%; P Z.02). Also, metastases were associated with a higher chance of technical success when compared with lung masses (100% vs 84%, P Z.02). Two complications occurred after sampling of a lung mass. One patient developed severe chest pain and required admission to the hospital for further evaluation. No evidence of a pneumothorax, injury to mediastinal
organs, or an acute cardiac event was found. The pain resolved spontaneously, and the patient was discharged within 24 hours. Self-limited hemoptysis was noted in the other patient. No complications were associated with FNA of metastatic lesions.
962 GASTROINTESTINAL ENDOSCOPY Volume 63, No. 7 : 2006
www.giejournal.org
DISCUSSION We found EUS to be a safe and accurate method for diagnosing lung cancer in patients with advanced, inoperable disease. A correct diagnosis was made in more than 85% of patients referred for EUS. The accuracy and rate of technical success of EUS was significantly higher for metastatic lesions than for primary lung masses. Although several studies have reported on the diagnostic value of EUS-FNA in lung cancer, few have made it the primary end point of their analysis. Fritscher-Ravens et al20 performed EUS on 35 patients with suspected lung cancer in whom bronchoscopy had failed to establish a diagnosis.
Sawhney et al
EUS-FNA for lung cancer diagnosis
TABLE 1. Baseline characteristics
TABLE 2. Accuracy of EUS-FNA by location of lesions
No. subjects
66
Age, y (range)
67 (43-82)
Sex, % of men
97
Location of lung mass, %* Upper lobe
63
Middle lobe
8
Lower lobe
15
Azygous lobe
2
Bilateral
8
Indeterminable or absent
5
No. patients,* (n Z 64)
Accuracy, %
Lung mass
19 (30%)
68
Mediastinal lymph nodes
39 (61%)
92
25
96
Paratracheal
6
83
Aortopulmonary window
6
83
Paraesophageal
2
100
Liver
4 (6%)
100
Adrenal gland
2 (3%)
100
Location
Subcarinal
*Two patients without a conclusive final diagnosis were excluded.
Inoperable, %*,y Locally advanced disease
17
Metastatic disease
20
Comorbid illnesses
47
Other causes
18
Only mediastinal lymph nodes were sampled, and a tissue diagnosis of cancer was obtained in 25 patients. EUS had a sensitivity of 96%, a specificity of 100%, and an accuracy of 97%. In another study by the same investigators, EUS made a correct diagnosis of cancer in 16 of 17 patients.21 In a study of 27 patients with enlarged mediastinal nodes, Silvestri et al22 found that EUS had a sensitivity of 89% and a specificity of 100% for the diagnosis of cancer. Varadarajulu et al23 published a retrospective case series of 18 patients with lung masses located adjacent to the mediastinum. In all 18 patients, a tissue diagnosis of malignancy was obtained by EUS. No complications were noted in this study. Our study is not directly comparable with any of these but does confirm some of the earlier findings. EUS was highly accurate and safe in patients with suspected malignant mediastinal lymph nodes. In all cases of malignancy, adequate material to enable a pathologist to distinguish between small-cell and non-small-cell cancer was obtained. The accuracy of EUS-guided sampling of lung masses in our study, however, was lower than that reported by Varadarajulu et al.23 If we exclude from analysis patients in whom EUS was technically unsuccessful, our results are more in agreement (accuracy, 81%; 95% CI 56%-94%). For patients with advanced lung cancer, EUS appears to be complementary to bronchoscopy in making a tissue diagnosis. Although bronchoscopy is very accurate for large, central lung lesions (sensitivity, 88%), it is less so for
smaller peripheral lesions (sensitivity, 33%).24 For lesions such as mediastinal nodes that do not have an endobronchial component, the sensitivity of bronchoscopy can be improved by performing transbronchial needle aspiration but remains low (sensitivity, 56%).24 EUS, on the other hand, is highly accurate and safe for mediastinal lymph nodes and peripheral lung masses adjacent to the mediastinum. EUS also has the advantage of providing a tissue diagnosis and confirming advanced cancer stage at the same time. Furthermore, EUS can sample lesions located in the aortopulmonary window, the paraesophageal mediastinum, the liver, and the adrenal glands, locations that cannot be accessed with either bronchoscopy or transbronchial needle aspiration. Transthoracic needle aspiration is the other most often used modality to diagnose lung cancer. The main advantage of EUS over transthoracic needle aspiration is fewer procedure-related complications. On average, 5% of patients undergoing transthoracic needle aspiration develop hemoptysis.25 Even though this is usually mild and selflimited, 2 cases of death from hemorrhage have been reported.25 The rate of pneumothorax after transthoracic needle aspiration is 30%. This is higher for patients with COPD (50%) or when multiple needle passes are made (40%).6 A chest tube is required in 5% to 25% of patients and usually necessitates admission to the hospital.26 EUSguided sampling has a significantly lower rate of complications. In our study, 2 self-limited complications were noted, and only 1 of 66 patients required admission to the hospital. In a large, prospective, multicenter study of 457 patients, 5 complications were noted as a result of EUS-guided sampling (0.5%, 95% CI 0.1%-0.8%).27 All complications were nonfatal and involved mainly pancreas cysts. A recently published study showed that EUS could also be safely performed in critically ill patients.28 Several limitations of our study warrant further discussion. First, for 1 patient, we did not have tissue confirmation
www.giejournal.org
Volume 63, No. 7 : 2006 GASTROINTESTINAL ENDOSCOPY 963
*Does not equal 100% because of rounding error. yReason patients were not considered for curative surgery.
EUS-FNA for lung cancer diagnosis
Sawhney et al
TABLE 3. Comparison of lung masses with metastatic lesions Lung mass
Metastasis
19
45
Technical success, % (95% CI)
84 (60-97)
100 (93-100)
Accuracy, % (95% CI)
68 (43-87)
93 (81-99)
Sensitivity, % (95% CI)
67 (40-87)
93 (81-99)
Specificity, % y
100
100
PPV, % y
100
100
No. subjects*
ple size but also allowed us to evaluate and report on the role of EUS in these difficult-to-diagnose patients. In conclusion, EUS-FNA is a safe and effective means of obtaining tissue in patients with advanced lung cancer. If available, it should be included along with bronchoscopy and transthoracic needle biopsy in the armamentarium of diagnostic procedure. Studies directly comparing EUS to brochoscopy and transthoracic needle aspiration are needed to further clarify its role.
REFERENCES
of the final diagnosis. We categorized the patient as having benign disease based upon complete radiologic resolution of the abnormal lesion. The strategy of using radiologic follow-up as a criterion standard for the diagnosis of cancer has been used by others and is unlikely to have misclassified this patient.27,29 Second, nearly all subjects in our study were men, reflecting the VA population seen at our institution. Third, the prevalence of lung cancer in our study sample was high (97%), limiting somewhat the generalizability of our results. The high prevalence of cancer was not unexpected, because all patients included in the study were suspected of having advanced inoperable disease. The fact that 3% of patients did not have cancer and that 14% had smallcell lung cancer underscores the importance of obtaining a tissue diagnosis before starting palliative therapy. Fourth, patients were enrolled in our study only if they were not candidates for a curative surgical resection and if malignant-appearing lesions on a CT were accessible by EUS. Thus, our results are applicable only to this select group of patients. Fifth, our study included patients who were referred to us by other centers, some of whom had already undergone failed attempts at tissue diagnosis. Because these patients did not represent a random sample of all patients with lung cancer, we were not able to report a ‘‘denominator’’ from which our patients were selected, nor were we able to specify a hierarchy of diagnostic tests that subjects should have undergone upon entry into the study. However, including these patients not only increased our sam-
1. Fry W, Menck H, Winchester D. The National Cancer Data Base report on lung cancer. Cancer 1996;77:1947-55. 2. Kvale P, Simoff M, Prakash U. Lung cancer. Palliative care. Chest 2003;123:284S-311S. 3. Pfister D, Johnson D, Azzoli C, et al. American Society of Clinical Oncology treatment of unresectable non-small-cell lung cancer guideline: update 2003. J Clin Oncol 2004;22:330-53. 4. Arroliga A, Matthay R. The role of bronchoscopy in lung cancer. Clin Chest Med 1993;14:87-98. 5. Zarbo R, Fenoglio-Preiser C. Interinstitutional database for comparison of performance in lung fine-needle aspiration cytology. A College of American Pathologists Q-Probe Study of 5264 cases with histologic correlation. Arch Pathol Lab Med 1992;116:463-70. 6. Sanders C. Transthoracic needle aspiration. Clin Chest Med 1992;13: 11-6. 7. Detterbeck F, DeCamp MJ, Kohman L, et al. Lung cancer. Invasive staging: the guidelines. Chest 2003;123:167S-75S. 8. Wallace M, Ravenel J, Block M, et al. Endoscopic ultrasound in lung cancer patients with a normal mediastinum on computed tomography. Ann Thorac Surg 2004;77:1763-8. 9. Wallace M, Fritscher-Ravens A, Savides T. Endoscopic ultrasound for the staging of non-small-cell lung cancer. Endoscopy 2003;35:606-10. 10. Gress F, Savides T, Sandler A, et al. Endoscopic ultrasonography, fineneedle aspiration biopsy guided by endoscopic ultrasonography, and computed tomography in the preoperative staging of non-small-cell lung cancer: a comparison study. Ann Intern Med 1997;127:604-12. 11. Fritscher-Ravens A, Davidson B, Hauber H, et al. Endoscopic ultrasound, positron emission tomography, and computerized tomography for lung cancer. Am J Respir Crit Care Med 2003;168:1293-7. 12. Fritscher-Ravens A, Bohuslavizki K, Brandt L, et al. Mediastinal lymph node involvement in potentially resectable lung cancer: comparison of CT, positron emission tomography, and endoscopic ultrasonography with and without fine-needle aspiration. Chest 2003;123:442-51. 13. Annema J, Hoekstra O, Smit E, et al. Towards a minimally invasive staging strategy in NSCLC: analysis of PET positive mediastinal lesions by EUS-FNA. Lung Cancer 2004;44:53-60. 14. Eloubeidi M, Cerfolio R, Chen V, et al. Endoscopic ultrasound-guided fine needle aspiration of mediastinal lymph node in patients with suspected lung cancer after positron emission tomography and computed tomography scans. Ann Thorac Surg 2005;79:263-8. 15. Kramer H, van Putten J, Post W, et al. Oesophageal endoscopic ultrasound with fine needle aspiration improves and simplifies the staging of lung cancer. Thorax 2004;59:596-601. 16. Rivera M, Detterbeck F, Mehta A. Diagnosis of lung cancer: the guidelines. Chest 2003;123:129S-36S. 17. Wiersema M, Wiersema L, Khusro Q, et al. Combined endosonography and fine-needle aspiration cytology in the evaluation of gastrointestinal lesions. Gastrointest Endosc 1994;40:199-206. 18. Mountain C, Dresler C. Regional lymph node classification for lung cancer staging. Chest 1997;111:1718-23. 19. Hasselblad V, Hedges L. Meta-analysis of screening and diagnostic tests. Psychol Bull 1995;117:167-78.
964 GASTROINTESTINAL ENDOSCOPY Volume 63, No. 7 : 2006
www.giejournal.org
NPV, % (95% CI)
14 (0-58)
25 (1-80)
LR C, % (95% CI)
1.7 (0.4-17)
3.7 (1.1-34.4)
LR , % (95% CI)
0.8 (0.5-3.0)
0.1 (0.1-0.5)
Misclassification rate, % (95% CI)
31 (12-57)
7 (0-18)
11
0
Complications, %
PPV, Positive predictive value; NPV, negative predictive value; LR C, likelihood ratio for a positive test; LR , likelihood ratio for a negative test. *Two patients without a final diagnosis are excluded from analysis. yBecause cytology is accepted as a criterion standard for the diagnosis of lung cancer, CIs are not reported.
Sawhney et al
EUS-FNA for lung cancer diagnosis
20. Fritscher-Ravens A, Soehendra N, Schirrow L, et al. Role of transesophageal endosonography-guided fine-needle aspiration in the diagnosis of lung cancer. Chest 2000;117:339-45. 21. Fritscher-Ravens A, Petrasch S, Reinacher-Schick A, et al. Diagnostic value of endoscopic ultrasonography-guided fine-needle aspiration cytology of mediastinal masses in patients with intrapulmonary lesions and nondiagnostic bronchoscopy. Respiration 1999;66:150-5. 22. Silvestri G, Hoffman B, Bhutani M, et al. Endoscopic ultrasound with fine-needle aspiration in the diagnosis and staging of lung cancer. Ann Thorac Surg 1996;61:1441-5; discussion 1445-6. 23. Varadarajulu S, Hoffman B, Hawes R, et al. EUS-guided FNA of lung masses adjacent to or abutting the esophagus after unrevealing CTguided biopsy or bronchoscopy. Gastrointest Endosc 2004;60:293-7. 24. Schreiber G, McCrory D. Performance characteristics of different modalities for diagnosis of suspected lung cancer: summary of published evidence. Chest 2003;123:115S-28S. 25. Detterbeck F, Rivera M. Clinical presentation and diagnosis. In: Detterbeck F, Socinski M, Rivera M, et al, editors. Diagnosis and Treatment of Lung Cancer: An Evidence Based Approach. 1st ed. Philadelphia: W.B. Saunders; 2001. p. 45-72. 26. Zwischenberger J, vanSonnenberger E, Alpard SK. Interventional radiology in the chest. In: Hazelrigg SR, Bashar M, Landreneau RJ, et al, editors. Minimal Access Cardiothoracic Surgery. 1st ed. Philadelphia: W.B. Saunders; 2000. p. 388-97.
27. Wiersema MJ, Vilmann P, Giovannini M, et al. Endosonography-guided fine-needle aspiration biopsy: diagnostic accuracy and complication assessment. Gastroenterology 1997;112:1087-95. 28. Fritscher-Ravens A, Sriram P, Pothman W, et al. Bedside endosonography and endosonography-guided fine-needle aspiration in critically ill patients: a way out of the deadlock? Endoscopy 2000;32:425-7. 29. Larsen S, Krasnik M, Vilmann P, et al. Endoscopic ultrasound guided biopsy of mediastinal lesions has a major impact on patient management. Thorax 2002;57:98-103.
www.giejournal.org
Volume 63, No. 7 : 2006 GASTROINTESTINAL ENDOSCOPY 965
Received August 2, 2005. Accepted November 8, 2005. Current affiliations: Section of Gastroenterology, (Drs Sawhney and Nelson), Center for Epidemiological and Clinical Research (Drs Sawhney and Lederle), Minneapolis VA Medical Center; Department of Medicine (Drs Sawhney, Kratzke, Lederle, and Nelson), Section of Hematologyd Oncology and Transplantation (Dr Kratzke); and Section of Cardiothoracic Surgery (Drs Holmstrom and Kelly), University of Minnesota, Minneapolis, Minnesota, USA. M. S. Sawhney is supported, in part, by VA Clinical Science R&D Service (Grant no. 04S-CRCOE-001). Reprint requests: Mandeep S. Sawhney, MBBS, Section of Gastroenterology (111 D), One Veterans Dr, Minneapolis, MN 55417.