Effectiveness and safety of endobronchial ultrasound–transbronchial needle aspiration: a systematic review

Bibliographic search

For study purposes, a bibliographic search of the medical literature was conducted in November 2007 and updated in April 2008. No time limits were specified and the following biomedical databases were searched: MEDLINE; EMBASE (Elsevier); ISI Web of Knowledge; Cochrane Library Plus; and National Health Service Centre for Reviews and Dissemination. A common search strategy was used for all databases, employing the terms (“endobronchial ultrasound” OR “endobronchial ultrasonography” OR “EBUS” OR “endobronchial ultrasound-guided”) AND (“TBNA” OR “transbronchial needle aspiration”). To complete the search, general information was located via the Google search engine, and a manual search was made of all bibliographic references cited in the original papers included.

Selection of papers and classification of scientific evidence

Potentially relevant papers were selected by perusing the abstracts yielded by the search and definitive inclusion was decided based on pre-established selection criteria. Only original papers in English, Spanish, French, Italian or Portuguese published in journals with a peer review process were included. Studies with <10 patients, studies that did not use at least one comparison and/or reference technique and studies that failed to furnish results on effectiveness, safety, or clinical management of patients were excluded. The results of radial endobronchial ultrasonography or EBUS applications other than CP–EBUS were not contemplated. The quality of the evidence was evaluated using the scientific evidence classification scale of the UK National Institute for Health and Clinical Excellence (NICE) 11 for studies of diagnostic validity.

Effectiveness in bronchopulmonary carcinoma

Two evidence-based clinical practice guidelines on initial diagnosis and mediastinal staging of lung cancer and 14 original studies that examined the usefulness of EBUS–TBNA in the diagnosis and staging of bronchopulmonary cancer were retrieved. Study design was prospective in 12 studies and retrospective in two studies. Sample size ranged 18–502 patients. Seven of the largest-sized prospective studies were conducted by the Japanese group of Yasufuku and co-workers 9, 12–14 and the multicentre group of Herth and co-workers 15–17 with main contributions from Germany, Denmark and the USA. Table 1⇓ summarises the characteristics of all original studies along with the main results.

The largest-scale prospective original study published on lung cancer was undertaken by the working group Herth et al. 15 in 2006 and covered 502 patients with mediastinal and hilar adenopathies. The sensitivity and negative predictive value (NPV) of EBUS–TBNA for detecting malignancy were 94% and 11% respectively. This same group carried out another investigation in the same year to investigate the performance of EBUS–TBNA in the staging of patients with suspected tumours of pulmonary origin (tumour stage T1 to T4) as evidenced by CT, but without enlargement of lymph nodes (nodes measuring <1 cm) 16. In this second study, which included 100 consecutive patients, sensitivity and NPV were 92.3% and 96.3% respectively. In another study undertaken very recently, Herth et al. 17 determined the results of EBUS–TBNA in sampling mediastinal lymph nodes in 100 patients with NSCLC, radiologically normal mediastinum and no PET activity. Comparing all results with those based on surgical staging, the sensitivity for detecting malignancy was 89% and NPV was 98.9%.

Wallace et al. 19 are responsible for the second-largest prospective study published on lung cancer diagnosis and staging. It included 150 patients with known or suspected lung cancer based on a lung or mediastinal abnormality on CT but with no proven extrathoracic metastases. EBUS–TBNA and transoesophageal ultrasonography (EUS)-guided fine needle aspiration (FNA) showed the same sensitivity (69%) and NPV (88%). TBNA alone had lower values (36% and 78%, respectively). The combination of EBUS and EUS obtained a sensitivity of 93% and an NPV 97%. In the subgroup of 60 patients with negative results on CT and PET, the estimated sensitivities of EUS–FNA, EBUS–FNA and EUS plus EBUS were 67%, 50% and 75%, respectively. NPVs were 92%, 89% and 94%, respectively. Another recent study carried out by Lee et al. 18 included patients with strongly suspected or histologically confirmed potentially operable NSCLC with lymph nodes of 5–20 mm on chest CT accessible by EBUS. Considering surgical–pathological staging as the gold standard, the sensitivity and NPV of EBUS were 93.8% and 96.9%, respectively.

Recently, Bauwens et al. 20 investigated the value of EBUS–TBNA in 106 patients with suspected or proven lung cancer and fluorodeoxyglucose (FDG)–PET-positive mediastinal adenopathy. In this subset of patients, the sensitivity and NPV for staging mediastinal N1, N2 or N3 lesions were 95% and 91%, respectively. Vincent et al. 21 also carried out a retrospective analysis of 152 patients to determine the additional value of EBUS–TBNA in comparison to staging with CT, PET and CT/PET. The results showed that EBUS–TBNA resulted in nodal status downstaging of 16% of the patients and upstaging in 9.75% of the patients.

The first study conducted by the group of Yasufuku et al. 9 included 70 patients with hilar and/or mediastinal adenopathies measuring >1 cm on CT. EBUS–TBNA avoided thoracotomy in six patients and other invasive procedures, such as mediastinoscopy and thoracotomy, in 17 patients. The second study 12 furnished results on 108 consecutive patients with enlarged mediastinal lymph nodes (measuring >1 cm) or suspected mediastinal malignancy (N2 or N3) detected on CT. The sensitivity and NPV observed were 94.6% and 89.5% respectively. In the opinion of a group of chest surgeons, EBUS–TBNA avoided 29 mediastinoscopies, eight thoracotomies, four thoracoscopies and nine CT-guided percutaneous biopsies. A third study by this group 13 compared the efficacy of EBUS–TBNA against that of CT and FDG–PET. This study included 102 patients with suspected or anatomopathologically confirmed lung cancer who were considered candidates for curative thoracic surgery. The sensitivity and NPV of EBUS–TBNA for predicting the stage of mediastinal lymphatic nodes were 92.3 and 97.4%, respectively. Using CT, these same parameters were 76.9% and 87.5% respectively; FDG–PET produced values of 80%, and 91.5% respectively. The last study of this group 14 was a retrospective analysis of 106 patients with metastasis of lung cancer, in whom CT had shown nodes measuring ≥5 mm. The sensitivity and NPV in 43 patients who underwent EBUS–TBNA were 92% and 95.3%.

The other two studies that assessed diagnosis or staging of lung cancer presented sensitivity and NPV above 85% and 70%, respectively 23, 24. Vilmann et al. 24 compared the performance of EBUS–TBNA with that of EUS–FNA for assessment of mediastinal lesions. EBUS–TBNA diagnosed 11 cancers unidentified by EUS–FNA, and EUS–FNA diagnosed 12 cancers undetected by EBUS–TBNA. With the combination of the two techniques, diagnostic accuracy was 100%. Rintoul et al. 23 reported the results of 18 patients evaluated by EBUS–TBNA with nodular enlargement or presence of paratracheal or parabronchial masses on the CT scan. EUS–FNA was used in six patients with enlarged lymph nodes in the posterior or inferior mediastinum and provided additional findings of nodular invasion in all cases.

The study of Monsó et al. 22 assessed the value of EBUS–TBNA exclusively for staging patients with mediastinal or lobular lymph nodes >5 mm found on ultrasound, obtaining a diagnostic yield of 92.5%.

Evidence-based clinical practice guidelines on initial diagnosis of lung cancer, recently published by the American College of Chest Physicians 25 recommend that diagnosis of patients with suspected small cell lung cancer based on clinical and radiological findings must be confirmed by the simplest method (sputum cytology, thoracocentesis, FNA, bronchoscopy including TBNA, EBUS–TBNA or oesophageal EUS–TBNA), as determined by patient presentation (grade of recommendation 1C). The clinical practice guidelines on mediastinal staging 26 conclude that staging by CT or PET is not sufficiently accurate for patients with discrete mediastinal lymph node enlargement. In patients with N2 or N3 node status, invasive techniques are needed for confirmation. The guidelines suggesting as reasonable transthoracic needle aspiration, TBNA, EBUS–TBNA or EUS–TBNA, given the appropriate experience and skills (grade of recommendation 1B). In general, mediastinoscopy is recommended for patients with a central tumour or N1 lymph node enlargement even though EUS–TBNA and EBUS–TBNA can be reasonable alternatives. A nonmalignant result from a needle technique should be further confirmed by mediastinoscopy (grade of recommendation 1B).

Effectiveness in other pathologies

Three prospective studies considered in the present review assessed the usefulness of EBUS–TBNA in patients with suspicion of sarcoidosis 27–29 (table 2⇓). The largest-scale study included 65 patients 27. EBUS–TBNA was diagnostic in 85–91.8% of patients with a final diagnosis of sarcoidosis. The NPVs found in two of these studies 27, 29 were 11% and 12.5%, respectively. In the investigation carried out by Wong et al. 27, 21.6% of the patients (11 out of 51) demonstrated with EBUS–TBNA to have sarcoidosis revealed benign transbronchial lung biopsy results. Oki et al. 28 obtained the same performance with EBUS–TBNA and TBNA.

One study has evaluated the usefulness of EBUS–TBNA in the diagnosis of lymphoma: this was a retrospective study that included 25 patients with idiopathic mediastinal adenopathies and suspected lymphomas 30. EBUS–TBNA enabled a sample of lymph tissue to be obtained in 96% of the patients (24 out of 25). The sensitivity, specificity, positive predictive value (PPV) and NPV observed were 90.0%, 100%, 100% and 92.6%, respectively.

Safety

None of the studies reported serious complications. Only three studies reported having observed agitation 31, cough 20, 31, and presence of blood 9 at the puncture site.

Quality of the evidence

According to the NICE scale, the level of evidence was Ib in two studies carried out by Herth et al. 16, 17 and III in all the remaining studies.

Diagnostic performance

EBUS with real-time needle aspiration has shown itself to be extremely useful in assessing nodular metastases. In three of the studies reviewed 13, 19, 20, imaging techniques, fundamentally CT and FDG-PET, which are frequently used as initial staging methods, displayed a lower sensitivity and specificity than that observed for EBUS–TBNA. The findings of a recent meta-analysis 32 are also in line with the inferiority of imaging techniques in detecting mediastinal involvement: the estimated sensitivity and specificity of FDG–PET and CT were 83% and 92%, and 59% and 78%, respectively. Unlike imaging techniques, EBUS–TBNA enables the identification and sampling of lymph nodes ≤1 cm in size. Two studies carried out by Herth et al. 16, 17 report sensitivities of 89% and 92%, respectively, among patients with no enlargement of lymph nodes on CT (measuring <1 cm), suggesting that this technique might have a special interest in such patients.

The sensitivity of real-time EBUS–TBNA for assessment of mediastinal and hilar metastasis seems to be equivalent or even superior to that of mediastinoscopy, the reference technique currently used to assess nodular metastases. According to the results of a recent pooled analysis, the sensitivity of mediastinoscopy ranges from 72–89% 33. EBUS–TBNA could replace mediastinoscopy in an important percentage of cases, though, due to the low NPV observed in some studies, surgical techniques cannot be ruled out in negative cases.

The main limitation of EBUS–TBNA is its inability to visualise posterior nodes (stations 5, 7, 8 and 9) 13. EUS–FNA is a complementary technique that enables visualisation of posterior nodes not visualised by EBUS–TBNA but does not allow the visualisation of the anterior mediastinum 13, 19. Two studies 19, 24 support the theory that the combination of EBUS–TBNA and EUS–FNA could dispense with surgical techniques in the great majority of cases, but one of these studies includes only 20 patients 20 and the validity of both is diminished due to the lack of confirmation of all negative results using reference techniques such as mediastinoscopy and/or thoracoscopy. Additionally, Wallace et al. 19 used lymph nodes, instead of patients, as the unit of analysis, so the possibility that study results are inflated cannot be dismissed 34.

Methodological shortcomings of the available literature

The lack of verification of the totality of results using a gold-standard test (surgery procedures) and the absence of adequate follow-up studies to assess the change in the therapeutic management of patients are important drawbacks when it comes to reaching definitive conclusions about the true usefulness of the EBUS–TBNA technique. With the exception of the two studies published by Herth and co-workers 16, 17, the studies included used the gold standard solely to check negative cases. Although the present authors acknowledge that in usual clinical practice a positive result does not need to be confirmed by additional surgical procedures, which are very invasive, in initial studies assessing the diagnostic performance of EBUS–TBNA all results should have been verified by surgical pathological staging in order to really assess the diagnostic yield of the new procedure. Due to this absence of verification of positive EBUS–TBNA results in these studies, specificities and PPVs have not been reported, assuming that these values are 100%. To confirm negative cases, the studies used a range of reference tests (thoracotomy, thoracoscopy, mediastinoscopy, mediastinotomy or clinical follow-up), such that verification bias can also not be ruled out.

Among the studies presented in the current review, there is great heterogeneity in the size and type of nodes punctured, the method used for puncture and the number of aspirations made to obtain cytology/histology samples, and this too might influence diagnostic accuracy. According to Lee et al. 18, sensitivity ranges from 69.8–95.3% when the number of aspirations increases from 1 to 3, but remains constant afterwards.

Generalisability of the published results

There are other important sources of variations that limit the generalisability of these results. Firstly, studies are very heterogeneous in respect to tumour characteristics (suspicion of lung cancer or previously diagnosed patients referred for staging), the types of lung cancer included (small-cell lung cancer or NSCLC) and the location of the lymphadenopathy as evidenced by imaging techniques (N1, N2/N3). The majority of studies do not differentiate between these subgroups and this is an important drawback. Even though the current authors could find no observable differences in studies that exclusively enrolled patients with NSCLC 12, 13, 15, 17 or that primarily analyse mediastinal and hilar metastasis 9, 13, 17, 20 in comparison to those that include only patients with mediastinal metastases 12, 18, 24, additional studies are required to clarify the effectiveness of this test in different clinical situations.

The high prevalence of mediastinal involvement in the reviewed studies and the disease severity are other factors that may limit the extrapolation of results. With the exception of four studies 16–19, prevalence of lymph node malignancy was >50% (table 1⇑) and the patients recruited displayed enlargement of lymph nodes (>1 cm). This might not represent the typical patient population. In a meta-analysis comparing PET with CT scanning for mediastinal staging, the prevalence of malignant lymph nodes was 37% 33. In another recent meta-analysis comparing PET and CT scanning, the post-test probability for N2 disease was found to be 5% for lymph nodes measuring 10–15 mm on CT and 21% for lymph nodes measuring ≥16 mm, when patients had a negative FDG–PET 35. Even though prevalence can be highly dependent on the pre-tests carried out in combination with CT (post-test probability of 62% with FDG–PET), it must be acknowledged that recruiting of patients with a high probability of suffering from the disease, of severe cases or of cases with appreciable enlargement of the lymph glands (>2 cm) can lead to an increase in sensitivity and PPV.

Result generalisability is also hampered by the lack of multicentre studies. Six out of the 14 studies retrieved that assessed the usefulness of EBUS–TBNA in the diagnosis and staging of bronchopulmonary carcinoma were conducted by the same research groups, suggesting that the latter have extensive practical experience in performing the procedure. It is readily acknowledged that EBUS–TBNA calls for in-depth, practical training in the interpretation of ultrasonographic images 36, and a lower degree of effectiveness and a greater proportion of adverse effects may thus be obtained by other, less-experienced teams. In fact, 45% of all the data available comes from the Heidelberg group of Herth and co-workers 15–17, increasing to 60% if the patients of Yasufuku and co-workers 9, 12–14 are included.

Conclusion

In summary, real-time endobronchial ultrasound-guided transbronchial needle aspiration is a technique of great interest for the identification and staging of mediastinal and hilar lymph nodes in patients with suspected and/or known bronchopulmonary neoplasm but its real value in clinical practice is still partly unknown. The present authors cite as an important limitation of the present work the absence of a meta-analysis, but felt that this task was at present practically impossible due to the great variability regarding selection criteria (cancer histology, severity of disease, location of the lymphadenopathy), procedure protocol or result interpretation. It is suggested that additional stratified analysis be carried out to establish diagnostic accuracy for different subgroups, especially for patients with nonsmall cell carcinoma and lymph node stations N2/N3, where staging is clinically relevant. Despite this device's apparent superiority over other existing procedures used for the same purpose, the present authors did not find suitably designed studies that compare these techniques against other alternative new techniques such as positron emission tomography–computed tomography, and this is needed in order to clarify the place that endobronchial ultrasound-guided transbronchial needle aspiration can occupy in the diagnostic algorithm. Appropriate follow-up studies are also called for to ascertain its usefulness in patient prognosis and therapeutic management. There is insufficient evidence, in terms of both -quality and quantity, to determine its usefulness in clinical indications other than that of lung cancer, although the published studies on sarcoidosis point to the effectiveness of endobronchial ultrasonography in diagnosing this disease.