Review of the studies
Endosonography versus surgical staging
The ASTER study (Assessment of Surgical sTaging versus Endobronchial and endoscopic
ultrasound in lung cancer: a Randomized controlled trial) by Annema et al. [18 ] compared immediate surgical mediastinal staging versus combined endosonography staging
(endobronchial ultrasound [EBUS] and endoscopic esophageal ultrasound [EUS] combined)
followed by surgical staging if no mediastinal nodal metastases were detected. In
detail, 241 patients with enlarged or FDG-avid mediastinal lymph nodes, enlarged or
FDG-avid hilar lymph nodes, or a central lung lesion were randomized. The reference
standard was surgical pathological staging including mediastinal nodal dissection.
The sensitivity for mediastinal lymph node metastasis was 79 % for surgical staging
versus 94 % for endosonography followed by surgical staging (P = 0.04), with corresponding negative predictive values (NPVs) of 86 % and 93 % (P = 0.26), respectively. The sensitivity of the combination of EUS and EBUS alone – without
subsequent surgical staging – was 85 %; this was not significantly different from
immediate surgical staging. Among patients with (suspected) NSCLC, a staging strategy
combining endosonography and surgical staging versus immediate surgical staging reduced
the percentage of unnecessary thoracotomies from 18 % to 7 % (P = 0.02) [18 ].
In the ASTER study [18 ], following a negative endosonography, 65 patients underwent mediastinoscopy which
detected 6 additional cases of N2 /N3 disease. In the subgroup of patients with an
abnormal mediastinum shown by radiological imaging, after a negative endosonography
the post-test probability for lymph node metastasis was 20 % (95 % confidence interval
[95 %CI] 12 % – 32 %), and adding a confirmatory mediastinoscopy in these patients
with negative endosonography decreased the post-test probability for missed nodal
metastases to 5 % (95 %CI 2 % – 20 %) [27 ]. Therefore, additional surgical staging, especially in this specific subset of patients,
is indicated. If negative endosonography results are not followed by confirmatory
surgical staging, careful follow-up is mandatory.
EBUS-TBNA or EUS with fine needle aspiration (FNA) alone
The accuracy of EBUS-TBNA and EUS-FNA separately for assessing mediastinal lymph node
metastases has been described in several studies. In a meta-analysis by Gu et al.
[29 ], involving 11 studies and 1299 patients, the pooled sensitivity of EBUS-TBNA in
mediastinal staging for lung cancer was 93 % (95 %CI 91 % – 94 %). The reference standard
was histopathology in 5 studies, and histopathology or clinical follow-up in 6. In
the subgroup of patients with an abnormal mediastinum on the basis of CT or PET, pooled
sensitivity was 94 % (95 %CI 93 % – 96 %), which was significantly higher than for
the subgroup of patients who were included regardless of CT or PET abnormalities (76 %,
95 %CI 65 % – 85 %).
Concerning EUS-FNA, a meta-analysis by Micames et al. (18 studies, 1201 patients)
reported a pooled sensitivity of 83 % (95 %CI 78 % – 87 %) [30 ]. The reference standard was histopathology in 10 studies, and histopathology or
clinical follow-up in 8. The sensitivity was 90 % (95 %CI 84 % – 94 %) in the subgroup
of patients with abnormal mediastinal lymph nodes at radiological imaging, and 58 %
(95 %CI 39 % – 75 %) among patients without abnormal mediastinal lymph nodes. There
was risk of bias in many of the studies included in these meta-analyses. This may
have led to overestimations of the sensitivity of the tests.
EBUS-TBNA and EUS-(B)-FNA combination versus either technique alone
To date, no RCTs have been performed comparing the EBUS plus EUS-(B) combination versus
either EBUS-TBNA or EUS-(B)-FNA alone.
EBUS-TBNA and EUS-(B)-FNA combination studies
We found 11 studies that assessed the accuracy of systematically performing both EBUS
and EUS for mediastinal staging in (suspected) lung cancer patients ([Table 2 ]) [18 ]
[31 ]
[32 ]
[33 ]
[34 ]
[35 ]
[36 ]
[37 ]
[39 ]
[40 ]
[41 ]. In the prospective comparative study by Wallace et al. [31 ], TBNA, EBUS-TBNA, and EUS-FNA for mediastinal staging of lung cancer were performed
in 138 patients against a reference standard of surgery or clinical follow-up. The
overall sensitivity of the combination of EBUS-TBNA and EUS-FNA was 93 %. This was
significantly higher than the sensitivities of EBUS-TBNA (69 %), EUS-FNA (69 %), and
conventional TBNA (36 %) alone. Vilmann et al. [32 ] found that the accuracy of EUS-FNA and EBUS-TBNA in combination for the diagnosis
of mediastinal cancer was 100 % in 28 patients, against a reference standard of surgery
or clinical follow-up.
Table 2
Studies that systematically assessed the accuracy of endobronchial ultrasound (EBUS)
and endoscopic (esophageal) ultrasound (EUS) for mediastinal staging in patients with
(suspected) lung cancer.
Author
Reference standard
Test order
Patients, n
Prevalence N2 /N3, %
EBUS
EUS
EBUS + EUS (95 %CI)
Sensitivity (95 %CI)
NPV (95 %CI)
Sensitivity (95 %CI)
NPV (95 %CI)
Sensitivity (95 %CI)
NPV (95 %CI)
Vilmann
2005 [32 ]
Surgery:
Clinical follow-up
EUS – EBUS
28
71 %
0.85 (0.62 – 0.97)
0.72 (0.39 – 0.94)
0.80 (0.56 – 0.94)
0.67 (0.35 – 0.90)
1.00 (0.83 – 1.00)
1.00 (0.63 – 1.00)
Wallace
2008 [31 ]
Surgery:
Clinical follow-up
EBUS – EUS
138
30 %
0.69 (0.53 – 0.82)
0.88 (0.80 – 0.93)
0.69 (0.53 – 0.82)
0.88 (0.80 – 0.93)
0.93 (0.81 – 0.99)
0.97 (0.91 – 0.99)
Annema
2010 [18 ]
EUS – EBUS
123
54 %
–
–
–
–
0.85 (0.74 – 0.92)
0.85 (0.74 – 0.93)
Herth
2010 [34 ]
Surgery:
Clinical follow-up
EBUS – EUS-B
139
52 %
0.92 (0.83 – 0.97)
0.92 (0.83 – 0.97)
0.89 (0.79 – 0.95)
0.89 (0.80 – 0.95)
0.96 (0.88 – 0.99)
0.96 (0.88 – 0.99)
Hwangbo
2010 [36 ]
EBUS – EUS-B
143
31 %
0.84 (0.71 – 0.94)
0.93 (0.87 – 0.97)
–
–
0.91 (0.79 – 0.98)
0.96 (0.90 – 0.99)
Szlubowski
2010 [41 ]
EUS – EBUS
120
23 %
0.46 (0.28 – 0.66)
0.86 (0.78 – 0.92)
0.50 (0.31 – 0.69)
0.87 (0.79 – 0.93)
0.68 (0.48 – 0.84)
0.91 (0.83 – 0.96)
Ohnishi
2011 [33 ]
EBUS – EUS
110
28 %
–
–
–
–
0.84 (0.71 – 0.97)
0.94 (0.89 – 0.99)
Kang (1)
2014 [35 ]
EBUS – EUS-B
74
46 %
–
–
–
–
0.84 (0.66 – 0.95)
0.94 (0.87 – 0.98)
Kang (2)
2014 [35 ]
EUS-B – EBUS
74
34 %
0.82 (0.65 – 0.93)
0.87 (0.74 – 0.95)
–
–
0.85 (0.69 – 0.95)
0.89 (0.76 – 0.96)
Lee
2014 [39 ]
EBUS – EUS-B
37
78 %
0.79 (0.60 – 0.92)
0.57 (0.29 – 0.82)
–
–
1.00 (0.88 – 1.00)
1.00 (0.63 – 1.00)
Liberman
2014 [40 ]
EBUS – EUS
166
32 %
0.72 (0.58 – 0.83)
0.88 (0.81 – 0.93)
0.62 (0.48 – 0.75)
0.85 (0.78 – 0.91)
0.91 (0.79 – 0.97)
0.96 (0.90 – 0.99)
Oki
2014 [37 ]
Surgery:
Clinical follow-up
EBUS – EUS-B
146
23 %
0.52 (0.34 – 0.69)
0.88 (0.81 – 0.93)
0.45 (0.28 – 0.64)
0.86 (0.79 – 0.92)
0.73 (0.54 – 0.87)
0.93 (0.86 – 0.97)
NPV, negative predictive value; 95 %CI, 95 % confidence interval; EUS-B, endoscopic
(esophageal) ultrasound using the EBUS scope; TEMLA, transcervical extended bilateral
mediastinal lymph adenectomy
The diagnostic value of the combined endosonography approach has recently been compared
with that of CT-PET for mediastinal nodal staging of lung cancer [33 ]. Overall, 120 consecutive patients with suspected resectable lung cancer on CT findings
(with and without enlarged mediastinal lymph nodes) underwent CT-PET and combined
EUS-FNA plus EBUS-TBNA. A final pathological N stage was established in 110 patients.
The accuracy of the combination of EUS-FNA plus EBUS-TBNA was significantly higher
than that of CT-PET (90.0 % vs. 73.6 %).
Herth et al. [34 ] analyzed 139 patients who underwent combined EBUS and EUS-B endosonographic staging.
The reference standard was surgical confirmation or clinical follow-up. Sensitivity
was 89 % for EUS-FNA and 92 % for EBUS-TBNA. The combined approach had a sensitivity
of 96 % and an NPV of 96 %.
In a recent RCT [35 ], 160 patients were randomized to either EBUS-TBNA followed by EUS-B-FNA (group A)
or to receive EUS-B-FNA followed by EBUS-TBNA (group B). In both arms, the second
procedure was performed on mediastinal nodes inaccessible or difficult to access by
the first procedure. No significant differences in final accuracy emerged between
groups A and B. However, while in group A, adding EUS-FNA to EBUS-TBNA did not significantly
increase the accuracy or sensitivity, in group B, adding EBUS-TBNA to EUS-FNA did
significantly increase the accuracy and sensitivity.
Hwangbo et al. [36 ] evaluated the role of EUS-B-FNA for mediastinal lymph nodes that were inaccessible
or difficult to access by EBUS-TBNA in 143 patients, with a reference standard of
surgical confirmation. The sensitivity, NPV, and diagnostic accuracy of EBUS-TBNA
alone for the detection of mediastinal metastasis were 84.4 %, 93.3 %, and 95.1 %,
respectively. The corresponding values for the combination of EBUS-TBNA plus EUS-B-FNA
increased to 91.1 %, 96.1 %, and 97.2 %, respectively. The proportion of mediastinal
nodal stations accessible by EBUS-TBNA was 78.6 %, and the proportion increased to
84.8 % by combining EUS-B-FNA with EBUS-TBNA (P = 0.015). EUS-B-FNA identified mediastinal metastasis in 3 additional patients.
In a recent prospective NSCLC staging trial in 146 patients, by Oki et al., EBUS was
routinely followed by EUS-B. The prevalence of mediastinal nodal metastases was 23 %.
The sensitivities of EBUS, EUS-B, and the combination were 52 %, 45 %, and 73 %, respectively,
with NPVs of 88 %, 86 %, and 93 %, when using a surgical procedure (or clinical follow-up
in a minority of patients) as the reference standard [37 ]. The subcentimeter size of the lymph nodes in combination with the low prevalence
of malignancy might account for the low sensitivity of EBUS. Often small lymph nodes,
especially in the left paratracheal station 4 L, are more easily aspirated from the
esophagus. In coughing patients, getting a good sample out of these small lymph nodes
with EBUS can be troublesome. In this study, patients were turned on their left side
for EUS-B; it is questionable whether this is needed as EUS-B is mostly performed
with patients in supine position [38 ].
Lee et al. [39 ] retrospectively analyzed 37 cases in which EUS-B was performed in addition to EBUS
when nodes were inaccessible by EBUS or when tissue sampling by EBUS alone was unsatisfactory.
A reference standard of mediastinoscopy or mediastinal lymph node dissection was used.
The sensitivity of EBUS compared with the combination was 79 % vs. 100 % (P = 0.008), and in 6 patients (13 %) their disease was upstaged based on EUS-B findings.
In a study by Liberman et al. [40 ], 166 patients with (suspected) NSCLC underwent EBUS, EUS, and mediastinoscopy in
the same setting. The prevalence of mediastinal metastases was 32 %. Against a reference
standard of mediastinoscopy, the sensitivity and NPVs were: for EBUS, 72 % and 88 %;
for EUS, 62 % and 85 %; and for combined EBUS/EUS, 91 % and 96 %. Endosonography was
diagnostic for N2/N3/M1 disease in 24 patients in whom mediastinoscopy findings were
negative, preventing futile thoracotomy in 14 % of patients [40 ].
The combination of EBUS-TBNA and EUS-FNA showed a pooled sensitivity of 86 % (95 %CI
82 % – 90 %) with a 100 % specificity for mediastinal nodal staging in a meta-analysis
based on 8 studies (821 patients) [25 ]. The sensitivity of the combined EBUS and EUS investigations appeared to be higher
in the subgroup with mediastinal abnormalities, but pooled data were not provided.
Although the authors concluded that “the current evidence suggests that the combined
technique is more sensitive than EBUS-TBNA or EUS-FNA alone,” they did not statistically
compare results from individual tests with the combined approach.
Random-effects meta-analysis was performed to evaluate the increase in sensitivity
provided by the combined approach. Adding EUS-(B)-FNA to EBUS-TBNA for mediastinal
nodal staging in a series of patients with established or suspected lung cancer, showed
an increase in sensitivity of 13 % (95 %CI 8 % – 20 %) for the combined approach compared
with EBUS-TBNA alone (9 studies; [Fig. 2 a ]). Adding EBUS-TBNA to EUS-(B)-FNA showed an increase in sensitivity of 21 % (95 %CI
13 % – 30 %) for the combined approach compared with EUS-(B)-FNA alone (7 studies;
[Fig. 2 b ]). Assuming a prevalence of nodal metastasis of 50 %, these numbers would indicate
that in 100 patients, adding EUS-(B)-FNA would avoid further surgical staging in an
additional 6.5 cases not identified by initial EBUS-TBNA; conversely, adding EBUS-TBNA
would avoid further surgical staging in an additional 10.5 cases not identified by
initial EUS-(B)-FNA. Considering that the studies included in these meta-analyses
are highly variable regarding quality and study population [25 ], that some studies included only patients with mediastinal lymph nodes that were
not accessible by EBUS-TBNA, that a “complete” EBUS-TBNA or EUS-FNA was not always
performed, that the reference standard included imperfect tests in some cases, and
in the absence of randomized trials comparing complete staging in single tests with
the combined approach, the results of this pooled analysis should be interpreted with
caution.
Fig. 2 Endobronchial ultrasound with transbronchial needle aspiration (EBUS-TBNA) combined
with endoscopic (esophageal) ultrasound with real-time guided fine needle aspiration
either using the conventional EUS endoscope or using the EBUS scope (that is, EUS-(B)-FNA)
for mediastinal nodal staging: comparison of the sensitivity of a single test with
that of the combined approach. a Increase in sensitivity of the combined approach compared with EBUS-TBNA alone. b Increase in sensitivity of the combined approach compared with EUS-(B)-FNA alone.
Random-effects meta-analysis was performed to evaluate the increase in sensitivity
of the combined approach versus a single test. The “Events” columns show the numbers
of cases that were detected by the combined approach, but not by a single test. The
“Total” columns show the total number of cases, as determined by the reference standard.
The “Proportion” column shows the increase in sensitivity of the combined approach
versus the single test. Fig. 2a suggests a 13 % (95 % confidence limits [CL] 8 % – 20 %) increase in sensitivity
from the combined approach over EBUS-TBNA or alone. Fig. 2b suggests a 21 % (95 %CL 13 % – 30 %) increase in sensitivity from the combined approach
over EUS-(B)-FNA alone. Study quality, especially the quality of the reference standard,
and the patient populations of included studies vary considerably so the risk of bias
may be substantial.
Safety
Complications of endosonographic procedures are rare. In a 2014 systematic review
on adverse events in 16 181 patients undergoing endosonography for mediastinal, hilar,
or primary lung tumor analysis, 23 serious adverse events (0.14 %) were reported:
0.3 % for EUS and 0.05 % for EBUS [42 ]. A systematic review of 13 studies (1536 patients) that reported on the safety of
EBUS-TBNA in lung cancer was published in 2009 [43 ]; no complications were reported in 11 studies, while one study reported no “major
complication,” and one study reported rare side-effects, notably cough. In a systematic
review [25 ] of combined EUS-FNA and EUS-TBNA for the staging of mediastinal lymph nodes in lung
cancer, severe complications were reported in 2 patients (0.3 %), consisting of pneumothorax
and lymph node abscess [25 ]. A nationwide survey, by the Japan Society for Respiratory Endoscopy, of complications
associated with EBUS-TBNA [44 ] found that, among 7345 procedures performed in 210 facilities, 90 complications
occurred (complication rate 1.23 %, 95 %CI 0.97 % – 1.48 %), resulting in one patient
death (mortality rate 0.01 %). The most frequent complications were hemorrhage (55 %)
and infection (16 %). The reported death was related to cerebral infarction during
withdrawal of antiplatelet drugs (replaced by heparin).
Cost – effectiveness
In the only study that measured actual costs and took health care utilization into
account [18 ], endosonography followed by surgical staging in those with negative test findings
proved to be cost-effective over surgical staging alone [18 ]
[24 ]. The cost – effectiveness gain at 6 months was mainly related to a statistically
significant reduction of the post-staging utility with the surgical compared with
the combined endoscopic approach, and with a reduction in the overall costs associated
with the nonsurgical staging procedure. The higher costs in the “surgical” arm were
due to the higher number of thoracotomies that had to be performed in this arm, and
not due to mediastinoscopy itself.
In a simulated model of evaluation of lung cancer patients with different prevalences
of mediastinal disease, a cost-minimization analysis showed that the combination of
EBUS-FNA/EUS-FNA would appear as the most cost-effective approach, compared with bronchoscopy
and mediastinoscopy, when the expected prevalence of lymph node metastasis is higher
than 32.9 %. This occurs in patients with a finding of abnormal mediastinum at radiological
staging [45 ]. In that model, EUS-FNA alone appeared to be the most cost-effective approach if
the prevalence of lymph node metastasis is lower than mentioned above as well as in
patients without abnormal lymph nodes on CT [45 ].
2. For mediastinal nodal staging in patients with suspected or proven non-small-cell
peripheral lung cancer without mediastinal involvement at CT or CT-PET , we suggest
that EBUS-TBNA and/or EUS-(B)-FNA should be performed before therapy, provided that
one or more of the following conditions is present: (i) enlarged or fluorodeoxyglucose
(FDG)-PET-avid ipsilateral hilar nodes; (ii) primary tumor without FDG uptake; (iii)
tumor size ≥ 3 cm; ([Fig. 3a – c ]) (Recommendation grade C).
3. In patients with suspected or proven < 3 cm peripheral NSCLC with normal mediastinal
and hilar nodes at CT and/or PET, we suggest initiation of therapy without further
mediastinal staging (Recommendation grade C).
Fig. 3 Schematic representation of peripheral lung cancer with normal mediastinum and with:
a ipsilateral hilar node, and tumor < 3 cm; b no fluorodeoxyglucose (FDG) uptake in the tumor, and tumor < 3 cm; c with or without FDG uptake in the tumor, and tumor ≥ 3 cm.
Background
Patients with small mediastinal lymph nodes without increased FDG uptake present a
6 % – 30 % risk of having mediastinal metastases in the following cases: (i) enlarged
or FDG-avid hilar lymph nodes, or small and FDG-avid hilar lymph nodes; (ii) any FDG-cold
lung tumor (i. e., pulmonary carcinoid, pulmonary adenocarcinoma in situ; (iii) lung
tumor > 3 cm (mainly in the case of adenocarcinoma with high FDG uptake) without any
lymph node involvement at CT or PET [9 ]
[10 ]
[11 ]
[21 ]
[46 ]. Mediastinal staging in those cases should be performed for accurate mediastinal
nodal assessment in order to allocate patients appropriately for curative-intent therapy.
Mediastinal lymph node metastases are present in less than 6 % of patients with small
peripheral tumors that present with neither enlarged nor FDG-avid hilar or mediastinal
lymph nodes [27 ].
Review of the studies
Data on the accuracy of endosonography for staging in patients without mediastinal
involvement on PET and/or CT are scarce. In two abovementioned meta-analyses [29 ]
[30 ], the sensitivity for mediastinal nodal staging in the subgroup of patients regardless
of, or without suspicious lymph nodes at CT or PET was 76 % (95 %CI 65 % – 85 %) for
EBUS-TBNA (3 studies, 263 patients) and 58 % (95 %CI 39 % – 75 %) for EUS-FNA (4 studies,
175 patients). Assuming a prevalence of 20 %, these numbers would indicate that 100
patients need to undergo endosonography to detect, respectively, 15.2 and 11.6 cases
of mediastinal involvement in whom further surgical staging can be prevented. However,
given the wide confidence intervals, especially for EUS, and the varying prevalence,
these numbers should be interpreted with caution.
Dooms et al. [47 ]
[48 ] prospectively evaluated 100 consecutive patients with suspected resectable clinical
N1 (cN1) disease, and a normal mediastinum, based on CT-PET with EBUS. The primary
outcome was the sensitivity of endosonography to detect N2 disease, against a reference
standard of histopathology. A total of 24 patients were diagnosed with N2 disease.
The sensitivity from endosonography alone was 38 % and this was increased to 73 %
by adding mediastinoscopy. So, in this population, 10 underwent mediastinoscopy to
detect a single case with N2 disease missed by endosonography. In this study, EBUS
was performed in all patients, while EUS was only added in patients with inaccessible
or difficult-to-reach lymph nodes. However, in 8 of the 14 false-negative cases where
no EUS was performed, the affected nodes were well within the reach of EUS-(B), being
stations 4 L, 7, and 8. Should EUS-(B) have been routinely performed, the sensitivity
of endosonography could have been above 70 % [38 ]
[47 ].
According to a post hoc subgroup analysis of the ASTER trial [27 ], the prevalence of mediastinal metastases in patients without a suspicious mediastinum
at CT-PET imaging was 26 % and the sensitivity of combined EBUS and EUS staging was
71 %, although confidence intervals were wide (36 % – 92 %) because of the small number
of patients in this subgroup. In this subgroup of patients, the post-test probability
for lymph node metastasis after a negative endosonography was 9 % (95 %CI 4 % – 24 %).
After the addition of mediastinoscopy, the post-test probability remained unaffected
[27 ]. In the surgical staging arm of the study, in patients with a non-suspicious mediastinum,
the prevalence of mediastinal metastases was 17 % and the sensitivity of surgical
staging was 60 % (23 % – 88 %), with a post-test probability of 8 % (95 %CI 3 % – 19 %)
after a negative test.
Wallace et al. [31 ] described a subgroup of 60 patients with negative mediastinal findings at CT and
PET who underwent both EBUS and EUS. The sensitivity and NPV were 17 % (95 %CI 2 % – 48 %)
and 83 % for TBNA, 50 % (95 %CI 21 % – 79 %) and 89 % for EBUS-TBNA, 67 % (95 %CI
35 % – 90 %) and 92 % for EUS-FNA, and 75 % (95 %CI 43 % – 95 %) and 94 % for combined
EBUS – EUS.
We found only one prospective study [41 ] that aimed to assess the diagnostic yield of the combined endosonographic approach
in patients with NSCLC and a normal mediastinum on CT alone (stage lA – llB). A total
of 120 patients underwent the combined approach with both EBUS-TBNA and EUS-FNA followed
by transcervical extended bilateral mediastinal lymphadenectomy (TEMLA) and, if negative,
pulmonary resection with dissection of the mediastinum as a confirmatory test. The
overall sensitivity of the combined approach was 68 %, the NPV was 91 %, and the positive
predictive value (PPV) was 91 %, at a prevalence of N2/N3 disease of 22 %. In this
study, 120 patients needed to undergo endosonography to detect 19 cases (16 %) in
which further surgical staging could be prevented. Additional surgical staging in
the remaining 101 patients identified another 9 cases. The overall sensitivity of
the combined technique was significantly higher than the sensitivity with EBUS alone
(46 %, 95 %CI 28 % – 65 %) and also higher and close to the level of significance
when compared with the sensitivity of EUS alone (50 %, 95 %CI 31 % – 69 %).
4. For mediastinal staging in patients with centrally located suspected or proven NSCLC
without mediastinal or hilar involvement at CT and/or CT-PET, we suggest performance
of EBUS-TBNA, with or without EUS-(B)-FNA, in preference to surgical staging ([Fig. 4 ]) (Recommendation grade D).
Fig. 4 Schematic representation of centrally located lung cancer with normal mediastinum.
Background
According to the ESTS guidelines, for centrally located lung tumors exploration of
mediastinal lymph nodes is indicated [21 ]. The false-negative rates of CT and PET imaging for mediastinal staging are high
for patients with a centrally located lung tumor (20 % – 25 % and 24 % – 83 %, respectively)
[13 ]
[49 ].
Review of the studies
There are no diagnostic accuracy studies specifically focusing on the EBUS and EUS-(B)
combination for patients with a centrally located lung tumor and a normal mediastinum/hilum.
Therefore recommendations are based on the evidence level of expert opinion.
The combination of EBUS-TBNA and EUS-(B)-FNA has been shown to have a high sensitivity
and high NPV in the staging of the mediastinal nodes. There are few studies in the
literature about the role of endosonography for mediastinal staging of patients with
a centrally located tumor [50 ]
[51 ]. Moreover, it must be noted that there is no agreement in the studies concerning
the definition of centrally located lung tumors. In a retrospective cohort of 16 patients
who had EUS-FNA of lung mass lesions adjacent to or abutting the esophagus, 10 patients
had invasion of the mediastinum by the tumor as shown by EUS, defined as loss of interface
between the tumor and the mediastinum, with an irregular border. Out of those 10 patients,
6 had mediastinal lymph nodes. EUS-FNA of the lymph nodes in 3 of those 6 patients
did not yield a preliminary diagnosis after 3 needle passes. It was technically difficult
to assess the mediastinal lymph nodes in the other 3 patients, because the lung mass
was in close proximity and precluded lymph node access [50 ]. In another study, out of 17 patients undergoing EUS-FNA of a centrally located
primary lung neoplasm, EUS identified metastatic lymph node involvement in 3 [51 ]. The accuracy of EBUS-only for mediastinal staging has already been addressed above
[29 ].
5. For mediastinal nodal restaging following neoadjuvant therapy, EBUS-TBNA and/or EUS-(B)-FNA
is suggested for detection of persistent nodal disease, but, if this is negative,
subsequent surgical staging is indicated (Recommendation grade C).
Background
According to current guidelines, stage III NSCLC (N2 /N3), that is, with metastatic
involvement of the ipsilateral (stage IIIA-N2) or contralateral (stage IIIB-N3) mediastinal
lymph nodes, should be treated with chemoradiation therapy [22 ]
[52 ]. The role of surgery in stage III (N2/N3) disease is under debate. It has been shown
that patients whose disease is downstaged to N0 with chemoradiation therapy, and who
subsequently undergo complete surgical resection of the lung tumor, have improved
survival in comparison to those patients who undergo surgery with persistent nodal
disease [53 ]
[54 ]. Therefore, if surgery is being considered following chemoradiation therapy, adequate
nodal restaging is essential to identify those patients whose disease has downstaged
to N0.
Review of the studies
EUS studies
In 2003, Annema et al. published the first case study with EUS-FNA for mediastinal
restaging in 19 NSCLC patients with N2 disease who had been treated with induction
chemotherapy. In the absence of regional lymph node metastasis (N0) at EUS-FNA, surgical
resection of the tumor with lymph node sampling or dissection was performed. A sensitivity,
NPV, and diagnostic accuracy of 75 %, 67 %, and 83 %, respectively, were found [55 ].
In a retrospective study that included 14 patients with NSCLC and biopsy-proven N2
disease, restaging by EUS-FNA following chemoradiation therapy had a 86 % diagnostic
accuracy for predicting mediastinal response [56 ].
In a prospective study in 28 patients, Stigt et al. re-evaluated the mediastinum after
induction therapy, and found a NPV of 91.6 % and accuracy of 92.3 % [57 ].
Von Bartheld et al. retrospectively analyzed 58 consecutive patients with tissue-proven
stage IIIA-N2 or IIIB-N3 NSCLC who underwent EUS-FNA for mediastinal restaging after
neoadjuvant chemoradiation therapy. Sensitivity, NPV, false-negative rate, and accuracy
of EUS-FNA for mediastinal restaging were 44 %, 42 %, 58 %, and 60 %, respectively.
A large percentage (22 %) of nodal metastases found at surgery were in locations beyond
the reach of EUS [58 ].
In a recent retrospective restaging study, EUS and/or EBUS was performed in 88 patients
followed by TEMLA if results were negative (n = 78). Significant differences were
found between EBUS or EUS and TEMLA for sensitivity (64.3 % and 100 %; P < 0.01) and NPV (82.1 % and 100 %; P < 0.01), in favor of TEMLA [59 ].
EBUS studies
Herth et al. retrospectively investigated EBUS-TBNA for restaging the mediastinum
following induction chemotherapy in 124 patients with NSCLC. Overall sensitivity,
specificity, PPV, NPV, and diagnostic accuracy of EBUS-TBNA for mediastinal restaging
after induction chemotherapy were 76 %, 100 %, 100 %, 20 %, and 77 %, respectively
[60 ].
Szlubowski et al. retrospectively analyzed a group of 61 consecutive NSCLC patients
with pathologically confirmed N2 disease who underwent neoadjuvant chemotherapy, and
in whom EBUS-TBNA was performed for restaging. The sensitivity and negative NPV of
the restaging EBUS-TBNA were 67 % and 78 %, respectively [61 ]. Recently, Szlubowski et al. prospectively assessed the diagnostic utility of combined
EBUS-TBNA and EUS-B-FNA for NSCLC restaging after induction therapy in 106 patients
with pathologically proven N2 disease. The prevalence of persistent mediastinal lymph
node metastases was 51.9 % and the sensitivity, specificity, total accuracy, PPV,
and NPV values of the combined approach were 67.3 %, 96.0 %, 81.0 %, 95.0 %, and 73.0 %,
respectively. The overall accuracy of the combined approach was higher as compared
with EBUS-TBNA and EUS-FNA alone [62 ].
6. A complete assessment of mediastinal and hilar nodal stations, and sampling of at
least three different mediastinal nodal stations (4 R, 4 L, 7) ([Fig. 1 ], [Fig. 5 ]) is suggested in patients with NSCLC and an abnormal mediastinum by CT or CT-PET
(Recommendation grade D).
Fig. 5 Schematic representation of sampling of at least three different mediastinal nodal
stations.
Background
For surgical nodal staging by mediastinoscopy, clear recommendations have been made
regarding the number and nodes to be sampled [21 ]
[22 ]. For endosonography, there is no agreement about how many and which lymph node stations
should be sampled and which level of thoroughness is necessary for different situations.
Some advise a thorough evaluation of all lymph nodes detectable by EBUS and EUS followed
by sampling. In many centers, however, the so-called “hit and run” approach is followed,
where only the lymph nodes that are suspicious at CT-PET imaging are sampled.
In the recent Guidelines from the American College of Chest Physicians (ACCP) [22 ], a classification of levels of thoroughness has been developed and could serve as
a guide. Four approaches were proposed: A, complete sampling of each node in each
major mediastinal node station (2 R, 4 R, 2 L, 4 L, 7, and possibly 5 or 6); B, systematic
sampling of each node station; C, selective sampling of suspicious nodes only; D,
very limited or no sampling, with only visual assessment.
In line with the ESTS guidelines [21 ], we recommend that at least three stations should be assessed (subcarinal, left
paratracheal, and right paratracheal) and biopsy samples should be taken if possible
with EBUS, EUS, or mediastinoscopy. Furthermore, all other abnormal lymph nodes, identified
by size or FDG avidity, should be sampled. This “complete” mediastinal staging is
based on the concept that identification of one malignant lymph node does not mean
that mediastinal staging was optimal.
7. For diagnostic purposes, in patients with a centrally located lung tumor that is
not visible at conventional bronchoscopy, endosonography is suggested, provided the
tumor is located immediately adjacent to the larger airways (EBUS) or esophagus (EUS-(B))
(Recommendation grade D).
To date, there are limited studies regarding the role of EUS-FNA and EBUS-TBNA in
the diagnosis of lung parenchymal masses. A recent retrospective study [63 ] assessed the diagnostic yield and safety of EUS-FNA of central mediastinal lung
masses. In 11 out of 73 patients, the lung mass could not be visualized by EUS. The
sensitivity of EUS was 96.7 % when only the visualized masses were considered, but
this value dropped to 80.8 % when the 11 nonvisualized masses were also taken into
account. Annema et al. [64 ] conducted a prospective study with 32 patients to assess the feasibility and diagnostic
yield of EUS-FNA for the diagnosis of centrally located lung tumors following a nondiagnostic
bronchoscopy. EUS-FNA provided a diagnosis of malignancy in 97 % of patients. In 39 %
of the patients, EUS-FNA not only established the diagnosis of lung cancer, but also
staged patients as having T4 disease, based on tumor invasion; however, the latter
was not verified surgically. None of the included patients had mediastinal lymph node
involvement at CT scan [64 ]. Varadarajulu et al. [50 ] conducted a retrospective study including 18 patients who had undergone EUS-FNA
of a lung mass abutting the esophageal wall. A diagnosis was obtained in all patients.
Hernandez et al. [51 ] retrospectively described their experience with EUS-FNA of centrally located primary
lung cancers; 17 patients had FNA of both the lung mass and the mediastinal lymph
nodes, and all procedures provided an accurate diagnosis of the primary lung lesion.
In a retrospective noncomparative study including 60 patients with a central parenchymal
lung lesion suspected to be lung cancer (82 % with a prior nondiagnostic flexible
bronchoscopy), Tournoy et al. [65 ] demonstrated that the sensitivity of EBUS-TBNA was 82 % with a NPV of 23 %. An exploratory
analysis showed that the sensitivity for small versus large lesions, when a short-axis
cutoff was arbitrarily set at 25 mm, was 78 % (95 %CI 57 % – 91 %) vs. 86 % (95 %CI
68 % – 96 %), respectively (P = 0.50). Verma et al. [15 ] also recently demonstrated in 37 patients that EBUS-TBNA is an effective way (overall
sensitivity 91.4 %) to diagnose parenchymal lesions located centrally close to the
airways.
8. In patients with a left adrenal gland suspected for distant metastasis we suggest
performance of EUS-FNA (Recommendation grade C), while the use of EUS-B with a transgastric
approach is at present experimental (Recommendation grade D).
Background
The adrenal glands are a predilection site for lung cancer metastases. Distant metastases
have significant impact on prognosis and treatment. Adrenal metastases originating
from NSCLC have been found in approximately 10 % – 59 % of patients in autopsy series
[66 ].
FDG-PET-CT has a high accuracy (sensitivity of 94 % and specificity of 85 %) for adrenal
metastases in patients with lung cancer [67 ]. However, adrenal glands that are suspicious at FDG-PET-CT can be false-positive
[68 ] and therefore tissue verification is indicated to either confirm or rule out metastatic
spread in order to prevent PET/CT-based upstaging in patients.
Traditionally, adrenal masses have been sampled by percutaneous biopsy. A small study
involving only 15 patients reported sensitivity and NPV for adrenal biopsy of 73 %
and 60 %, respectively [69 ]. A study involving 79 patients reported an overall complication rate for percutaneous
adrenal biopsies of 8.4 % [70 ], including hemorrhage, pneumothorax, pancreatitis, adrenal abscesses, bacteremia,
and needle-tract metastases. Transgastric EUS-guided FNA can be performed during the
same session as a mediastinal staging procedure, using the same endoscope.
Review of the studies
In 1996, Chang reported the first application of EUS and EUS-FNA for left adrenal
gland analysis and found that the left adrenal gland was visualized in 30 out of 31
patients (97 %) [71 ].
In 31 patients with suspected thoracic or gastrointestinal malignancies and enlarged
left adrenal gland on abdominal imaging (including 15 patients with lung cancer),
Eloubeidi et al. reported that EUS-FNA obtained adequate tissue from the left adrenal
gland in all patients [72 ].
In a mixed series of 119 patients with gastrointestinal or pulmonary disease, who
underwent EUS with or without FNA, the left adrenal gland was routinely examined.
The overall prevalence of a left adrenal mass was 4/119 (3.4 %), all detected in the
cohort of patients (n = 12) with lung cancer [73 ]. In a retrospective analysis of 40 patients, with established or suspected lung
cancer and an enlarged left adrenal gland shown at EUS, the diagnostic yield of EUS-FNA
for detecting left adrenal metastases was 95 % [74 ].
In a retrospective analysis by Schuurbiers et al. of 85 patients with (suspected)
lung cancer and a left adrenal gland suspicious for metastasis identified by CT and/or
FDG-PET, EUS-FNA demonstrated left adrenal metastases in 62 % and benign adrenal tissue
in 29 %. Sensitivity and NPV for EUS-FNA of the left adrenal gland were at least 86 %
(95 %CI 74 – 93 %) and 70 % (95 %CI 50 – 85 %), respectively. No complications occurred
[75 ].
Eloubeidi et al. evaluated 59 patients with enlarged adrenal gland(s) on abdominal
CT, magnetic resonance imaging (MRI), and/or PET, and known or suspected malignancy.
All patients underwent EUS-guided FNA (54 left adrenal gland and 5 right adrenal gland),
and adrenal tissue adequate for interpretation was obtained in all patients. On multivariable
analysis, altered adrenal gland shape (loss of seagull configuration) was a significant
predictor of malignancy [76 ].
Most literature about EUS of the adrenal gland concerns the left adrenal gland. However
there are some reports about transduodenal EUS-guided FNA of the right adrenal gland.
It seems feasible and safe in experienced hands [77 ]
[78 ]
[79 ].
Recently, Uemura and colleagues retrospectively analyzed a consecutive series of 150
patients with potentially resectable lung cancer who were undergoing EUS/EUS-FNA for
mediastinal staging of lung cancer. Routinely, both the left and right adrenal glands
were assessed. The left adrenal gland was visualized in all patients (100 %) and the
right adrenal gland in 87.3 % of patients [79 ]. Transgastric analysis and FNA of the left adrenal gland using an EBUS scope has
been described [77 ], but its feasibility and safety are under investigation.
Complications of EUS-guided FNA of adrenal glands are rare; an adrenal hemorrhage
has been described [80 ]. However, it should be emphasized, that in the case of signs of a pheochromocytoma,
endocrinologic evaluation must be done prior to endosonography.
9. For optimal endosonographic staging of lung cancer, we suggest that individual endoscopists
should be trained in both EBUS and EUS-B, in order to perform complete endoscopic
staging in one session (Recommendation grade D).
Background
The quality and safety of endosonography is very dependent on the skills and experience
of the operator. Diagnostic yield improves with practice [81 ], and the number of complications is also associated with operator experience [82 ]. Despite this, there is a paucity of evidence-based structured training programs,
and the important decision about when a trainee is considered competent is often based
on an arbitrary number of performed procedures or on subjective impressions.
As the combined staging by EBUS and EUS is superior to staging by a single technique
[25 ], it seems logical that the skills should be present in a single operator [83 ]. For practical and economic reasons, the majority of procedures will be performed
with EBUS scopes for both the endobronchial and esophageal route.
10. We suggest that new trainees in endosonography should follow a structured training
curriculum consisting of simulation-based training followed by supervised practice
on patients (Recommendation grade D).
Background
Increased focus on patient safety has put pressure on the traditional apprenticeship
model where trainees under supervision practice on patients. Simulation-based training,
on phantoms and virtual reality devices, has been suggested for helping trainees surmount
the initial, steep part of the learning curve.
Review of the studies
A systematic review and meta-analysis regarding technology-enhanced simulation, based
on 609 papers, found “large effects for outcomes of knowledge, skills, and behaviors
and moderate effects for patient-related outcomes” [84 ]. There are two virtual reality simulators commercially available for EBUS: the GI
Bronch Mentor (Simbionix, Cleveland, Ohio, USA) and the AccuTouch Flexible Bronchoscopy
Simulator (CAE Healthcare, Montreal, Quebec, Canada). Both simulators can discriminate
between novices and experienced operators (indicating construct validity) [85 ]
[86 ], but there are no published studies exploring the effect of EBUS simulator training
on patient care. No software exists for mediastinal sampling using EUS, but EUS-FNA
as well as EBUS-TBNA can be practiced on rubber models, animal organs, or live anesthetized
animals. A study regarding EBUS-TBNA training found both computer simulation and wet
lab simulation to be effective and complementary [87 ].
However, despite the positive effects of simulation-based training, it is important
to remember that no existing simulators are 100 % realistic and not all aspects of
a procedure can be practiced. Supervised performance during initial patient encounters
is essential, even after a thorough simulation-based training program – self-learning
of endosonography should be discouraged [88 ].
11. We suggest that competency in EBUS-TBNA and EUS-(B)-FNA for staging lung cancer be
assessed using available validated assessment tools (Recommendation grade D).
Background
Thorough knowledge of (endosonographic) anatomy and its relation to the TNM lung cancer
staging system is crucial for the performance of an endosonographic evaluation. Upstaging
could prevent the patient from receiving potentially curative therapy, and downstaging
may cause the patient to undergo unnecessary surgery and treatments without therapeutic
benefit [89 ]. To avoid this, basic competency must be ensured before trainees are allowed to
perform procedures independently.
Review of the studies
Early guidelines for gastrointestinal EUS recommend a minimum of 150 total supervised
procedures [90 ], but a more recent study on learning curves showed “substantial variability in achieving
competency and a consistent need for more supervision than current recommendations”
[91 ]. It is generally agreed that sampling in the mediastinum is technically easier than
in other locations [92 ] and a study showed that chest physicians achieved satisfactory results after participating
in an EUS implementation program for staging lung cancer patients [93 ]. The only study exploring learning curves for EUS-FNA for lung cancer staging found
that acquisition of skills varied between individuals and that 20 procedures were
not enough to ensure basic competency [94 ].
Early guidelines on training requirements for EBUS from the American Thoracic Society/European
Respiratory Society and the ACCP respectively recommend minimum numbers of 40 and
50 procedures for initial acquisition of competence [95 ]
[96 ]. These numbers are based on expert opinions, are arbitrary, and are debated [89 ]
[97 ]. Studies on EBUS learning curves have shown that performance of 50 procedures does
not ensure basic competency [60 ]
[95 ], and the latest Guidelines from the British Thoracic Society recognise that “Individuals
have different learning curves and hence focus should be towards monitoring an individual’s
performance and outcomes” [98 ]. Specific tools for assessment of performance in endosonography [99 ]
[100 ] could be used for monitoring trainees’ progression, and all programs should continuously
monitor their outcomes.
These guidelines from ESGE, ERS, and ESTS represent a consensus of best practice based
on the available evidence at the time of preparation. They may not apply in all situations
and should be interpreted in the light of specific clinical situations and resource
availability. Further controlled clinical studies may be needed to clarify aspects
of the statements, and revision may be necessary as new data appear. Clinical consideration
may justify a course of action at variance to these recommendations. ESGE guidelines
are intended to be an educational device to provide information that may assist endoscopists
in providing care to patients. They are not rules and should not be construed as establishing
a legal standard of care or as encouraging, advocating, requiring, or discouraging
any particular treatment.
