ot_bib_bo_ebusperipherallesions

G Model LUNG-3790; No. of Pages 7

ARTICLE IN PRESS Lung Cancer xxx (2011) xxx–xxx

Contents lists available at ScienceDirect

Lung Cancer journal homepage: www.elsevier.com/locate/lungcan

A randomised trial of endobronchial ultrasound guided sampling in peripheral lung lesions夽 Kjetil Roth a,∗ , Tomas Mikal Eagan b , Alf Henrik Andreassen b , Friedemann Leh c , Jon Andrew Hardie d a

Department of internal medicine, Helse Sunnmøre, Aalesund, Norway and Department of Thoracic Medicine, Haukeland University Hospital, Bergen, Norway Department of Thoracic Medicine, Haukeland University Hospital, Bergen, Norway Department of Pathology, Haukeland University Hospital, Bergen, Norway d Institute of Medicine, University of Bergen, Norway b c

a r t i c l e

i n f o

Article history: Received 10 September 2010 Received in revised form 18 January 2011 Accepted 27 February 2011 Keywords: Diagnostic techniques Respiratory System Endobronchial ultrasound Bronchoscopy Lung cancer Cost-effectiveness analyis Biopsy

a b s t r a c t Aim: The aim of the study was to evaluate endobronchial ultrasound (EBUS) for peripheral lung lesions and to find the most cost effective combination of sampling techniques.

Materials: 264 patients with lesions suspicious of malignancy were recruited in Bergen and Aalesund, Norway from 2005 to 2008.

Methods: The study was a prospective randomised cohort study. EBUS was performed with a 1.7 mm rotating probe. X-ray fluoroscopy was used in both arms. The different sampling techniques were evaluated in a cost-effectiveness analysis.

Results: The detection rate for cancer was 36% in the EBUS group, 44% in the non-EBUS group (ns). Lesions below 3 cm and lesions assumed difficult to reach had significant lower detection rates in the EBUS group.

Lesions visualised by EBUS had a higher detection rate for cancer than lesions not visualised by EBUS (62% vs. 19%, p < 0.01). The cost of one additional positive sample was 1211 euro when brushing was added to biopsy. It was not cost effective to add washing or TBNA. Conclusion: EBUS did not increase the detection rate for cancer in peripheral lung lesions when bronchoscopy was performed by bronchoscopists at all levels of expertise. Biopsy and brushing was the most cost effective combination of sampling techniques. © 2011 Elsevier Ireland Ltd. All rights reserved.

1. Introduction There is an ongoing discussion about the optimal diagnostic approach for obtaining a definite diagnosis from lesions in the peripheral areas of the lungs. In a previous study from our centre, endobronchial visibility and lesion size predicted a high detection rate [1], in line with other previous studies [2–8]. The detection rate for cancer by the initial bronchoscopy in our department was 16.7% for non-visible lesions, 4.8% without X-ray fluoroscopy, and 35.4% when X-ray fluoroscopy guided the samplings [1]. This was comparable to the results of a Scottish multicentre study [9], but lower than that reported in Schreiber’s summary of published reports [10] and Rivera’s clinical practice guideline [11].

Endobronchial ultrasound (EBUS) is a potentially valuable method to increase the detection rate, since one can use EBUS to

夽 ClinicalTrials.gov NCT00398970. ∗ Corresponding author. Tel.: +47 70105000.

E-mail addresses: Kjetil.Roth@helse-sunnmore.no, kjetil.roth@med.uib.no (K. Roth), Tomas.Eagan@med.uib.no (T.M. Eagan), Alf.Andreassen@helse-bergen.no (A.H. Andreassen), Friedemann.Leh@helse-bergen.no (F. Leh), jon.hardie@med.uib.no (J.A. Hardie).

visualise lesions that are too distal to be visible in the bronchoscopic camera. The first studies of EBUS showed a detection rate for cancer between 60% and 90% in diagnosing non-visible lesions [12–16].

However, these studies have been carried out in specialised centres by a limited number of highly experienced bronchoscopists. Also, patients were in some cases excluded based on low compliance in the screening bronchoscopy [17]. Herth et al. has shown high yields for EBUS in small lesions not visible by fluoroscopy (detection rate for cancer 47–71%) [18,19] In a randomised crossover study with 50 patients performed mostly under general anaesthesia, Herth did not find an increase in the detection rate compared to non-EBUS [20]. Paone et al. found, in a prospective randomised trial, a higher detection rate with EBUS. 799 patients with peripheral lung lesions were screened, only 293 patients included [17]. Subgroup analyses revealed that EBUS was significantly better for lesions smaller than 3 cm, but there was no difference in detection rates for lesions larger than 3 cm [17]. All procedures were performed by highly experienced pulmonologists in both studies. It is not clear as to whether less experienced bronchoscopists or low-volume centres will have any advantage of using EBUS compared to fluoroscopic guidance alone.

The aim of the current study was to evaluate the use of endobronchial ultrasound for peripheral lesions in a clinical practice

0169-5002/$ – see front matter © 2011 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.lungcan.2011.02.013

Please cite this article in press as: Roth K, et al. A randomised trial of endobronchial ultrasound guided sampling in peripheral lung lesions. Lung Cancer (2011), doi:10.1016/j.lungcan.2011.02.013

G Model LUNG-3790; No. of Pages 7 2

ARTICLE IN PRESS K. Roth et al. / Lung Cancer xxx (2011) xxx–xxx

where the bronchoscopies are performed by several pulmonologists with various levels of experience. The study was also designed to reveal the most cost-effective combination of sampling techniques for non-visible lesions.

2. Materials and methods 2.1. Study sample The patients were examined at the Department of Thoracic Medicine, Haukeland University Hospital, Bergen and the Department of Medicine, Aalesund Hospital, both in Western Norway, between June 2005 and January 2008. All patients attending for investigation of lesions suspicious of malignancy in the lungs were eligible for inclusion. Patients were not included if a computer tomography (CT) scan indicated that the lesion was visible by bronchoscopy.

Diagnostic yield in the non-EBUS group was predicted to be 40% and in the EBUS group to be 60% [1,13–16,20]. Standard power calculation (˛ = 0.05 and power = 0.9) required 120 in each study arm. A simple randomisation was performed without stratification. When inclusion was closed, 289 patients were randomised, though 25 patients were subsequently excluded due to finding

of endobronchial visible lesions. Retrospectively all bronchoscopic procedures performed at the two labs in the study period were reviewed and 130 additional patients were found. These patients had lesions suspicious of malignancy and no obvious endobronchial findings (Fig. 1). The main reasons for non-inclusion were periods with equipment failure, patients not willing to participate, and an incorrect assumption that the lesion was visible based on the CT scan.

2.2. Study procedure All physicians in the two bronchoscopy labs underwent a training session prior to participation for use of EBUS with guide sheath and curette. Following a theoretical training session including EBUS operation and image interpretation, a member of the study team was present during the first EBUS procedures for each bronchoscopist. The level of the bronchoscopists’ previous experience varied from more than 30 years to less than one year. Few of the bronchoscopists had significant previous experience with EBUS. Altogether, 29 physicians performed the bronchoscopies in the study. 25% of the physicians contributed with less than 3 bronchoscopies and 25% contributed with more than 14 (median = 8). The bronchoscopy was performed with Olympus BF 1T 160 broncho-

Fig. 1. Consort statement flow diagram.

Please cite this article in press as: Roth K, et al. A randomised trial of endobronchial ultrasound guided sampling in peripheral lung lesions. Lung Cancer (2011), doi:10.1016/j.lungcan.2011.02.013

G Model LUNG-3790; No. of Pages 7

ARTICLE IN PRESS K. Roth et al. / Lung Cancer xxx (2011) xxx–xxx

scopes. If the bronchoscopy revealed a visible lesion, the patient was excluded from the study. Patients with constriction or compression but no certain visible endobronchial lesion were retained in the study. In both study arms Boston 1.8 mm “Radial Jaw 3” was used for biopsies, Boston 21 gauche “stifcor” or “eXcelon” transbronchial needles for TBNA and Boston “Cellebrity” for brushings.

The washings were performed as a small volume lavage following the other samplings. The ultrasound was performed with an Olympus 20 MHz 1.7 mm rotating probe. All procedures were performed transorally without an endotracheal tube. Patients were semi-sedated with pethidine hydrochloride 25–75 mg or midazolam 2.5–5 mg. X-ray fluoroscopy was used for guidance in both arms of the study.

A miniprobe inside a guide sheath was directed towards the lesion in patients randomised to EBUS. If the ultrasound signal indicated that the probe was inside the lesion, the probe was removed and the samples were taken through the guide sheath. Two TBNA punctures were regularly performed for all patients. A cytotechnician was present to evaluate the TBNA smears. The cytotechnician evaluated the smears while four biopsies were taken. If the smears were representative, the investigation was ended with brushing and small volume lavage. The TBNA was repeated one or two times if the smears were not representative. The end point of the study was the final pathological result of the bronchoscopy.

Two hours after the bronchoscopy the patients filled in a visual analogue scale (VAS) where zero was no discomfort and 10 extremely high grade of discomfort during the procedure.

3

2.3. Evaluation of variables Lesions were divided in two groups – larger than or smaller than 3 cm – based on axial CT scans. Based on the endobronchial findings lesions were categorised as non-visible, concentric bronchial lumen constriction, or bronchial wall compression. Based on the CT scans Kjetil Roth (blinded for the randomisation and pathological results) reviewed and categorised all lesions retrospectively: I: a bronchial branch straight to the lesion. II: no direct path to the lesion, but one or two divisions to pass beyond the visible divisions. III: no direct path to the lesion, but three or more divisions to pass beyond the visible divisions. IV: no bronchial branch leading to the lesion.

All patients were followed to September 2009 unless the diagnosis was confirmed by operation or autopsy prior to this time.

2.4. Statistical analyses All data analyses were performed as “intention-to-treat” regardless of whether EBUS was practically possible to perform or not. The statistical analyses were performed with SPSS [21] and Stata [22].

The cost-effectiveness analysis was performed with TreeAge Pro Healthcare [23]. All costs were measured in Euro 2007 value. Equipment and personnel costs in both the bronchoscopy and pathology labs were included in the cost of each sampling technique. The effectiveness was the average detection rate for malignant and benign disease. The willingness to pay for one additional positive sample was estimated to be 2800 euro, based on the diagnosis

Table 1 Baseline characteristics of 264 cases.

Sex Female Male Age <59 years 59–67 years 68–74 years >70 years Sizeb <3 cm ≥3 cm Location Upper lobe Middle lobe Lower lobe Endobronchial difficultyc I: a bronchial branch straight to the lesion II: one or two divisions to pass III: three or more divisions to pass IV: no bronchial branch leading to the lesion Endobronchial visibility No visible lesions Compression Constriction Uncertain findings Methodse Did use X-ray fluoroscopy Did use biopsy Did use brushing Did use TBNAf Did use washing

Randomised to EBUSa (n = 124)

Randomised to non EBUSa (n = 140)

n

n

%

p

% 0.07

37 87

29.8 70.2

57 83

40.7 59.3

27 41 28 28

21.8 33.1 22.6 22.6

25 32 40 43

17.9 22.9 28.6 30.7

50 74

40.3 59.7

58 82

41.4 58.6

65 17 42

52.4 13.7 33.9

74 19 47

52.9 13.6 33.6

21 32 46 25

16.9 25.8 37.1 20.2

34 36 33 37

24.3 25.7 23.6 26.4

105 10 7 2

84.7 8.1 5.6 1.6

114 6 13 7

81.4 4.3 9.3 5.0

121 102 93 105 124

97.6 82.3 75.0 84.7 100.0

137 118 107 129 140

98.6 84.3 76.4 92.1 100.0

0.14

0.86

1.00

0.08

0.19d

0.67d 0.66 0.79 0.06 NA

Statistical analysis: ␹2 except for d. a EBUS: endobronchial ultrasound. b Size was measured with CT thorax. c The difficulty to reach the lesions was estimated based on the CT scans. d Fisher’s exact test. e Methods: the numbers displayed are positive cases; the negative cases are not displayed. f TBNA: transbronchial needle aspiration.

Please cite this article in press as: Roth K, et al. A randomised trial of endobronchial ultrasound guided sampling in peripheral lung lesions. Lung Cancer (2011), doi:10.1016/j.lungcan.2011.02.013

G Model LUNG-3790; No. of Pages 7

ARTICLE IN PRESS

4

K. Roth et al. / Lung Cancer xxx (2011) xxx–xxx

related group (DRG) cost for 5 days in day ward and the cost of an additional investigation.

All patients received written and oral information, and signed informed consent prior to participation. The Regional Norwegian Ethical Committee and the Norwegian Social Science Data Service approved the study. The study was registered in ClinicalTrials.gov (NCT00398970).

3. Results The baseline characteristics are displayed in Table 1. The mean size of the lesions were 3.7 cm (Standard Deviation (SD):2.0 cm) in the EBUS group and 3.8 cm (SD:2.3 cm) in the non-EBUS group.

The first bronchoscopy obtained a malignant diagnosis in 40.2% of malignant cases, but was not able to get a final result for any of the benign cases (Table 2). The biopsy result in the benign cases was indicative of normal morphology (66.7%), inflammation (19.0%), represented necrosis, bleeding or too little material for a definite diagnosis (11.1%), fibrosis (1.6%) or false positive carcinoma (1.6%). The final diagnosis of the malignant cases was adenocarcinoma (38.6%), squamous cell carcinoma (18.0%), non-classifiable non small cell (20.1%) and small cell carcinoma (3.7%). Otherwise, 8.5% had metastasis from cancer in other organs.

Table 3 displays the intention to treat analysis. The detection rate for malignant disease was 36.0% in the EBUS group compared to 43.7% in the non-EBUS group (p = 0.29). The negative predictive value was about 40% in both groups. There were no significant differences in the detection rates for brushing, biopsy, TBNA or washing. The non-EBUS group had a significantly higher detection rate for cancer for lesions smaller than three centimetre (p = 0.04), and if there were one or two divisions to pass beyond the visible Table 2 The diagnostic method and final diagnosis of 264 cases.

Malignant disease (n = 189) Diagnostic method First bronchoscopy Repeated bronchoscopy Transthoracic sampling Operation or autopsy Sampling from other organs than the lung Clinical diagnosis of cancer Final diagnosis Small cell lung cancer Adenocarcinoma Squamous cell carcinoma Large cell carcinoma Non classifiable non small cell lung cancer Metastasis of cancer from other organs to the lung Only clinical diagnosis Other malignant disease in the lunga Benign disease (n = 75) Diagnostic method First bronchoscopy Repeated bronchoscopy Transthoracic sampling Operation Liver biopsy Follow up Final diagnosis Sarcoidosis Hamartoma Tuberculosis Atypical mycobacterial disease Removed by operation without malignancy The lesion decreased or vanished.

The lesion was unchanged by follow up The lesion was uncertain, but benign by follow up

n

%

76 5 76 15 9 8

40.2 2.6 40.2 7.9 4.8 4.2

7 73 34 14 38 16 3 4

3.7 38.6 18.0 7.4 20.1 8.5 1.6 2.1

0 1 2 7 1 64

0 1.3 2.7 9.3 1.3 85.3

2 3 2 2 3 36 21 6

2.7 4.0 2.7 2.7 4.0 48.0 28.0 8.0

a Other malignant disease in the lung: carcinoid tumour: 1, lymphoma: 1, mesothelioma: 1, bronchioloalveolar carcinoma: 1.

divisions (grade II difficulty to reach the lesions) (p = 0.02). There was no significant difference in the pneumothorax rate (Table 3).

The average procedure time was 41 min in the EBUS group compared with 36 min in the non-EBUS group (p = 0.01). The median patient reported discomfort was 1.6 in the EBUS group compared to 2.5 in the non-EBUS group (VAS scale, p = 0.10). The combined mean detection rate for cancer in both groups was 44.5% if the investigator reported that X-ray fluoroscopy showed the sampling device in the lesion, compared to 19.4% if not (p = 0.01). Fluoroscopy visualised 24/34 (70.6%) of the malignant lesions below 3 cm in the EBUS group and 25/35 (71.4%) of the lesions below 3 cm in the non-EBUS group. Malignant lesions above 3 cm were visualised by fluoroscopy in 86.0% (EBUS group) and 94.0% (non-EBUS group) of the patients. Only one of the lesions missed by fluoroscopy was visualised by EBUS. The bronchoscopists reported that EBUS visualised 37 of 80 cases with final malignant disease. Among the 37 positively visualised cases, the detection rate for cancer was 62.2% compared to 18.6% in 43 cases when the bronchoscopists found no certain visualisation (p < 0.01). The number of procedures performed by each bronchoscopist in the study did not affect the result.

The detection rate was 28.6% in the EBUS group and 47.9% in the non EBUS group for physicians with more than 15 procedures (NS) and 41.2% vs. 40.0% for those with 15 procedures or fewer (NS).

Table 4 displays the bivariate and multivariate analyses of the potential predictors for a higher detection rate for cancer. There was a significant interaction between EBUS and size. When the size of the lesion was below 3 cm, randomisation to EBUS predicted a lower detection rate for cancer; odds ratio (OR) 0.25 (Table 4). Endobronchial difficulty assessed by CT scans was a strong predictor for the detection rate. Constriction, compression, or localisation did not predict a higher detection rate for cancer.

The cost-effectiveness analysis was based on the 178 cases where all sampling techniques were performed, regardless of randomisation (Table 5). Brushing in addition to biopsy increased the detection rate for cancer to 43.8% compared to 36.7% for biopsy alone (p < 0.01). The cost for the detection of one additional positive sample was 1211 euro when brushing was added to biopsy. This was below the estimated willingness to pay of 2800 euro. Threshold analyses showed that the cost of brushing had to be below 142 euro and the detection rate for cancer had to increase with 3.1% for biopsy and brushing to be cost effective compared to biopsy alone.

Incremental cost-effectiveness ratio (ICER) was 4761 euro when washing was added to biopsy and brushing and 8262 euro when TBNA was added to biopsy, brushing and washing, both above the willingness to pay. Biopsy and TBNA had a slightly higher diagnostic yield than biopsy and brushing, but the ICER for biopsy and TBNA was 12,430 euro compared to biopsy and brushing.

4. Discussion EBUS did not result in an increase in the detection rate for cancer in the intention to treat analysis. On the contrary, randomisation to EBUS predicted a lower detection rate for lesions below 3 cm compared to sampling guided by X-ray fluoroscopy alone. However, when the physicians reported that the lesions were visualised with EBUS the detection rate for malignant disease was 62.2%, compared with 18.6% when the physicians did not report certain tumour visualisation by EBUS. The detection rate for cancer was 45.5% when fluoroscopy showed the optimal position for sampling in the non-EBUS group. A CT scan showing a bronchial branch leading straight into the lesion predicted a fivefold higher detection rate for malignant disease. Biopsy and brushing was the most cost effective combination of sampling techniques.

The main strength of this study was its design; a prospective randomised trial, and its high inclusion rate; 289 out of 419 eligi-

Please cite this article in press as: Roth K, et al. A randomised trial of endobronchial ultrasound guided sampling in peripheral lung lesions. Lung Cancer (2011), doi:10.1016/j.lungcan.2011.02.013

G Model LUNG-3790; No. of Pages 7

ARTICLE IN PRESS K. Roth et al. / Lung Cancer xxx (2011) xxx–xxx

5

Table 3 Intention to treat analysis of EBUS.

Detection rate for cancer (sensitivity) False positive rate Positive predictive value Negative predictive value Subgroup analyses, detection rates for cancer Brushing Biopsy TBNAc Washing Lesions < 3 cm Lesions ≥ 3 cm I: a bronchial branch straight to the lesion II: one or two divisions to pass III: three or more divisions to pass IV: no bronchial branch leading to the lesion Upper lobe Middle lobe and lingula Lower lobe No visible lesions Endobronchial compression Endobronchial constriction Uncertain endobronchial lesions Complications Pneumothoraxd

Randomised to EBUSa (benign:38, malignant:86)

Randomised to non-EBUSa (benign:37, malignant:103)

n

%

p

n

%

31/86 1/38 31/32 37/92

36.0 2.6 96.9 40.2

45/103 1/37 45/46 36/94

43.7 2.7 97.8 38.3

0.29 1.00b 1.00b 0.79

13/67 22/70 20/71 4/86 4/35 27/51 16/18 3/19 8/32 4/17 21/53 3/8 7/25 22/68 4/9 3/7 2/2

19.4 31.4 28.2 4.7 11.4 52.9 88.9 15.8 25.0 23.5 39.6 37.5 28.0 32.4 44.4 42.9 100

20/77 34/86 25/97 10/102 11/35 34/68 21/29 15/30 8/25 1/19 22/57 7/14 16/32 29/79 3/5 9/12 4/7

26.0 39.5 25.8 9.8 31.4 50.0 72.4 50.0 32.0 5.3 38.6 50.0 50.0 36.7 60.0 75.0 57.1

0.35 0.29 0.73 0.18 0.04 0.75 0.18 0.02 0.56 0.11 0.91 0.68b 0.09 0.58 1.0b 0.33b 0.50b

10/99

10.1

6/111

5.4

0.20

Statistical analysis: ␹2 except for b. a EBUS: Endobronchial ultrasound. b Fisher’s exact test. c TBNA: Transbronchial needle aspiration. d 210 patients with X-ray pictures within 3 days were included in the analysis.

ble. Further, the study was performed at two centres that perform a fair number of bronchoscopies per year, however, organised so that a large number of physicians partake in the procedures. Thus, this study offered the ability to evaluate EBUS in a real life setting where bronchoscopy is performed by investigators at various levels of expertise. The results of this study should be comparable to other centres starting with EBUS, where each doctor performs approximately thirty bronchoscopies per year. The study included

a cost-effectiveness analysis of the sampling methods, which most studies tend not to do.

A weakness of the study was that a significant number of participating physicians performed only a few procedures with EBUS. Thus it did not allow evaluation of individual skills nor evaluation of EBUS when in expertly skilled hands. The detection rate for cancer in the EBUS group increased from 29.0% during the first year to 40.0% thereafter, (p = 0.31). The level of training might likely explain

Table 4 Multivariate analysis of detection rates for cancer in the first bronchoscopy. Combined result of all localising and sampling techniques (n = 189).

Unadjusted odds ratio OR (95% CIb )

Adjusted odds ratioa OR (95%CIb )

EBUS × size <3 cm non-EBUS <3 cm with EBUS >3 cm non-EBUS >3 cm with EBUS Endobronchial difficultye I: a bronchial branch straight to the lesion II: one or two divisions to pass III: three or more divisions to pass IV: no bronchial branch leading to the lesion Visibility No visible lesions Constriction Compression Uncertain endobronchial lesions Lobe Upper lobe Middle lobe/lingula Lower lobe d

a b c d e

LRc 0.04

1 0.28 (0.08–1.00) 1 1.13 (0.54–2.33)

1 0.25 (0.06–1.03) 1 1.32 (0.57–3.04)

1 0.16 (0.06–0.39) 0.11 (0.04–0.26) 0.04 (0.01–0.14)

1 0.17 (0.07–0.44) 0.14 (0.05–0.36) 0.05 (0.02–0.19)

1 1.88 (0.63–5.66) 3.23 (1.20–8.70) 3.77 (0.90–15.68)

1 2.61 (0.65–10.5) 1.59 (0.50–5.13) 1.43 (0.25–8.25)

1 1.30 (0.52–3.27) 1.05 (0.55–2.03)

1 0.80 (0.26–2.47) 0.60 (0.27–1.33)

<0.01

0.51

0.44

All variables included in the multivariate analysis are displayed in the table.

95%CI: 95% confidence interval.

LR: likelihood ratio.

EBUS: endobronchial ultrasound.

The difficulty to reach the lesions was estimated based on the CT scans.

Please cite this article in press as: Roth K, et al. A randomised trial of endobronchial ultrasound guided sampling in peripheral lung lesions. Lung Cancer (2011), doi:10.1016/j.lungcan.2011.02.013

G Model LUNG-3790; No. of Pages 7

ARTICLE IN PRESS

6

K. Roth et al. / Lung Cancer xxx (2011) xxx–xxx

Table 5 Cost-effectiveness analysis of different combinations of sampling techniques (only cases where all sampling techniques were performed were included).

Cost for the combination (euro)

Biopsy 49.5 Biopsy and brushing 111.7 Biopsy, brushing and washing 160.8 Biopsy, brushing, TBNAc and washing 356.2

Detection rate for cancer (n = 128)

36.7% 43.8% 45.3% 48.4%

pa

0.004 0.500 0.125

Final diagnosis obtained for benign disease (n = 50)

0% 0% 0% 0%

ICERb (euro)

1211 4761 8262

Sensitivity analyses

Upper threshold value for cost (euro)

Minimum increase in the sensitivity for cancer

Brushing <142.3 Washing <30.1 TBNA <62.1

>3.1% >2.4% >9.8%

Willingness to pay: 2800 euro. a McNemars test compares the detection rate to the result in the line above. b ICER: incremental cost-effectiveness ratio: the increase in cost divided on the average diagnostic yield for benign and malignant disease. c TBNA: transbronchial needle aspiration.

a part of the difference in detection rate compared to other studies [17,20].

Rivera and Mehta found sensitivity for cancer between 20% and 77% for results of biopsy in 36 studies from 1971 to 2003 (up to 88% with all sampling techniques combined) [11]. The variation in the diagnostic yields between the studies might have been due to different skill levels, different inclusion criteria, or different choice of sampling techniques. The current study reflects the realistic yield in a teaching hospital where physicians at various levels of experience perform the bronchoscopies and is comparable with the Scottish multicenter study (diagnostic yield 9%) [9]. Studies like Kawaraya’s with few but highly specialised pulmonologists have higher yields (diagnostic yield 88%) [24]. All procedures in the current study were done with fluoroscopy guidance. In the retrospective study from the current study centre 48/131 (36.6%) of the bronchoscopies for peripheral lesions were done under fluoroscopic guidance (diagnostic yield 16.7%) [1]. The diagnostic yield in the non-EBUS group in the current study has improved compared to the previous study, but likely would be higher with further training.

The diagnosis of peripheral lesions by bronchoscopy depends on two factors: First, the physician’s ability to accurately locate the lesions, with or without radiological guidance, and second, the ability to harvest representative samples. The lesions were visualised by EBUS (fluoroscopy guided) in 51 of 117 (46.3%) of the patients with malignant or benign disease. Previous studies have identified the lesions with EBUS in 66% to 93% of the cases [12–14,18–20,25–28]. Some studies performed fluoroscopy guided EBUS [12,14,29], while other performed EBUS without fluoroscopy [18–20,26–28]. The ability to reach the lesion will be dependent on where the lesion is located and whether the local or general anaesthesia makes it possible for the patients to remain calm during the procedure. Thus studies performed under general anaesthesia will likely have higher detection rates due to better working conditions. The Heidelberg group has performed most procedures under general anaesthesia with detection rates for cancer as high as 80% [20,30]. Although many studies performed with local anaesthesia have detection rates around 60% [12,25–28,31], some have detection rates up to 80% [13,17]. The size of the lesion will influence on the ability to localise it [28]. In the current study 28.3% of lesions below 3 cm were localised with EBUS compared to 54.3% above 3 cm (p = 0.01). Small lesions can be misinterpreted as vessels and missed if the EBUS probe is retracted too fast. Larger lesions can have multiple bronchial branches leading into the lesion, simplifying navigation.

The current study introduced a new classification system for estimation of endobronchial difficulty based on the CT scans. The estimated number of bronchial divisions the sampling probe had to pass beyond the visual field defined the scale. Although this system is likely to be affected by inter-observer variability, the classifi-

cation was a strong predictor for a higher detection rate. Future studies should consider evaluating and adjusting detection rates for endobronchial difficulty.

For the purpose of power calculation, the detection rate in the EBUS group was predicted to be 60% based on previous studies [12,13,15,20]. However, the detection rate for cancer reached only 36.0% and no benign cases achieved a definite diagnosis during the first bronchoscopy. The detection rate for cancer was 62.2% when the lesions were visualised with EBUS, comparable to previous studies [14,18,30], but the overall results were lower than in the previous studies. One explanation could be the use of the guide sheath. The main advantage of the guide sheath is to guide the samplings directly from the position visualised by EBUS, but when EBUS failed to visualise the lesions, the physician will experience only the disadvantages of the guide sheath. A bend might obstruct the lumen. Coughing, deep respiration or movements on the bronchoscope can displace the guide sheath and the TBNA can slide into the guide sheath instead of penetrating the lesion.

The interaction between size and use of EBUS (randomisation) suggest that the difficulties regarding locating and sampling from lesions were probability stronger for EBUS in small lesions. Previous studies of EBUS from high volume centres have found increased detection rates for small lesions [17]. Thus, EBUS might be most useful in the diagnosis of small lesions for highly trained staff, possibly in combination with virtual navigation [25,30,32]. Based on the results of the current study future studies with endobronchial ultrasound should be designed with few operators performing a large number of bronchoscopies each. The case selection based on the CT scan should always be reported, especially regarding tumour size.

A cost-effectiveness analysis compares the increase in cost to the increase in effectiveness [33]. All costs were seen from a societal perspective concerning equipment costs and wages for personnel. Quality adjusted life years (QALY) is the most common effectiveness measurement. Previous cost-effectiveness studies of bronchoscopy have used reduced-quality days [34] as effectiveness measurement, or performed cost minimisation analyses [35,36].

The current study used the combined detection rate for malignant and benign lesions as the effectiveness measurement. Incremental cost-effectiveness ratio (ICER) is the increase in cost divided by the increase in effectiveness. It represents the cost of an additional positive sample. The willingness to pay for one additional positive sample is dependent on how difficult it is to get the sample by a repeated procedure. Although the estimate might be too low, the diagnosis related group (DRG) for five days in day ward and the cost for a repeated bronchoscopy was set as the willingness to pay for an additional positive sample. The detection rate for cancer increased from 36.7% for biopsy to 43.8% (p < 0.01) for biopsy and brushing. The cost of one additional positive sample was 1211 euro

Please cite this article in press as: Roth K, et al. A randomised trial of endobronchial ultrasound guided sampling in peripheral lung lesions. Lung Cancer (2011), doi:10.1016/j.lungcan.2011.02.013

G Model LUNG-3790; No. of Pages 7

ARTICLE IN PRESS K. Roth et al. / Lung Cancer xxx (2011) xxx–xxx

when brushing was added to biopsy. This was below the willingness to pay of 2800 euro. The ICER of additional washing or TBNA was above the willingness to pay. Biopsy and brushing was the most cost effective combinations of sampling techniques for non visible lesions. Bronchoscopies with use of all sampling techniques were performed in 178 patients. Only these patients were included in the cost-effectiveness analysis. Some procedures were ended because of patient discomfort, others because the lesions were too close to the chest wall to warrant the risk of pneumothorax. Except for lesions close to the chest wall, the authors believe this error was random and unlikely to represent large selection bias.

5. Conclusion EBUS did not increase the overall detection rate for cancer in the current study when pulmonologists at various levels of expertise performed the bronchoscopies. The detection rate was adequate if EBUS visualised the lesion. The guide sheath should probably be removed before taking samples from lesions not visualised by EBUS as it then has no function. The detection rate was lower with EBUS than with simple fluoroscopy alone in small lesions and in lesions that were hard to reach. Biopsy and brushing was the most cost effective combination of sampling techniques for non-visible lesions.

Conflicts of interest None of the authors had any interests to declare.

Acknowledgements We would like to thank Gunvor Mo Norstein, Bjørg Guldbrandsen, Gerd Eli Dale Askeland, Mona Grundeland, Lise Østgård Monsen, Sissel Eide-Olsen, Ragny Skuseth and the other nurses at the bronchoscopy lab in Bergen and Ålesund for their help and cooperation during the study. The study was funded by grants from Helse Vest, Helse Sunnmore and the Connie Gulborg Jansen foundation. Guide sheaths were not commercial available and were provided from Olympus.

References [1] Roth K, Hardie JA, Andreassen AH, Leh F, Eagan TM. Predictors of diagnostic yield in bronchoscopy: a retrospective cohort study comparing different combinations of sampling techniques. BMC Pulm Med 2008;8(2).

[2] Baaklini WA, Reinoso MA, Gorin AB, Sharafkanch A, Manian P. Diagnostic yield of fiberoptic bronchoscopy in evaluating solitary pulmonary nodules. Chest 2000;117(4):1049–54.

[3] Castella J, Buj J, Puzo C, Anton PA, Burgues C. Diagnosis and staging of bronchogenic carcinoma by transtracheal and transbronchial needle aspiration. Ann Oncol 1995;6:21–4.

[4] Estarriol MH, Goday MR, Sanchez MV, Padro XB, Sot MTC, Quetglas FS. Bronchoscopic lung biopsy with fluoroscopy to study 164 localized pulmonary lesions.

Arch Bronconeumol 2004;40(11):483–8.

[5] Pilotti S, Rilke F, Gribaudi G, Spinelli P. Cytologic diagnosis of pulmonarycarcinoma on bronchoscopic brushing material. Acta Cytol 1982;26(5):655– 60.

[6] Radke JR, Conway WA, Eyler WR, Kvale PA. Diagnostic-accuracy in peripheral lung lesions – factors predicting success with flexible fiberoptic bronchoscopy.

Chest 1979;76(2):176–9.

[7] Stringfield JT, Markowitz DJ, Bentz RR, Welch MH, Weg JG. Effect of tumor size and location on diagnosis by fiberoptic bronchoscopy. Chest 1977;72(4):474– 6.

[8] Wallace JM, Deutsch AL. Flexible fiberoptic bronchoscopy and percutaneous needle lung aspiration for evaluating the solitary pulmonary nodule. Chest 1982;81(6):665–71.

7

[9] Mclean AN, Semple PD, Franklin DH, Petrie G, Millar EA, Douglas JG. The Scottish multi-centre prospective study of bronchoscopy for bronchial carcinoma and suggested audit standards. Respir Med 1998;92(9):1110–5.

[10] Schreiber G, McCrory DC. Performance characteristics of different modalities for diagnosis of suspected lung cancer – summary of published evidence. Chest 2003;123(1):115S–28S. [11] Rivera MP, Mehta AC. Initial diagnosis of lung cancer: ACCP evidence-based clinical practice guidelines (2nd edition). Chest 2007;132(Suppl (3)):131S–48S. [12] Kikuchi E, Yamazaki K, Sukoh N, Kikuchi J, Asahina H, Imura M, et al.

Endobronchial ultrasonography with guide-sheath for peripheral pulmonary lesions. Eur Respir J 2004;24(4):533–7.

[13] Kurimoto N, Miyazawa T, Okimasa S, Maeda A, Oiwa H, Miyazu Y, et al. Endobronchial ultrasonography using a guide sheath increases the ability to diagnose peripheral pulmonary lesions endoscopically. Chest 2004;126(3):959–65.

[14] Shirakawa T, Imamura F, Hamamoto J, Honda I, Fukushima K, Sugimoto M, et al.

Usefulness of endobronchial ultrasonography for transbronchial lung biopsies of peripheral lung lesions. Respiration 2004;71(3):260–8.

[15] Yang MC, Liu WT, Wang CH, Lin HC, Chen HC, Chou CL, et al. Diagnostic value of endobronchial ultrasound-guided transbronchial lung biopsy peripheral lung cancers. J Formos Med Assoc 2004;103(2):124–9.

[16] Yoshikawa M, Sukoh N, Yamazaki K, Kanazawa K, Fukumoto SI, Harada M, et al. Diagnostic value of endobronchial ultrasonography with a guide sheath for peripheral pulmonary lesions without X-ray fluoroscopy. Chest 2007;131(6):1788–93.

[17] Paone G, Nicastri E, Lucantoni G, Iacono RD, Battistoni P, D’Angeli AL, et al.

Endobronchial ultrasound-driven biopsy in the diagnosis of peripheral lung lesions. Chest 2005;128(5):3551–7.

[18] Eberhardt R, Ernst A, Herth FJF. Ultrasound-guided transbronchial biopsy of solitary pulmonary nodules less than 20 mm. Eur Respir J 2009;34(6):1284–7.

[19] Herth FJF, Eberhardt R, Becker HD, Ernst A. Endobronchial ultrasound-guided transbronchial lung biopsy in fluoroscopically invisible solitary pulmonary nodules – a prospective trial. Chest 2006;129(1):147–50.

[20] Herth FJF, Ernst A, Becker HD. Endobronchial ultrasound-guided transbronchial lung biopsy in solitary pulmonary nodules and peripheral lesions. Eur Respir J 2002;20(4):972–4.

[21] SPSS for Windows Rel. 15.0.1.1. Chicago: SPSS Inc; 2007.

[22] Stata Statistical Software: Relase 11. StataCorp. College Station. TX: StataCorp LP; 2009.

[23] TreeAgePro Healthcare 1.4.1. Williamstown: TreeAge Software Inc; 2008.

[24] Kawaraya M, Gemba K, Ueoka H, Nishii K, Kiura K, Kodani T, et al. Evaluation of various cytological examinations by bronchoscopy in the diagnosis of peripheral lung cancer. Br J Cancer 2003;89:1885–8.

[25] Asahina H, Yamazaki K, Onodera Y, Kikuchi E, Shinagawa N, Asano F, et al. Transbronchial biopsy using endobronchial ultrasonography with a guide sheath and virtual bronchoscopic navigation. Chest 2005;128(3):1761–5.

[26] Chung YH, Lie CH, Chao TY, Wang YH, Lin AS, Wang JL, et al. Endobronchial ultrasonography with distance for peripheral pulmonary lesions. Respir Med 2007;101(4):738–45.

[27] Dooms CA, Verbeken EK, Becker HD, Demedts MG, Vansteenkiste JF. Endobronchial ultrasonography in bronchoscopic occult pulmonary lesions. J Thorac Oncol 2007;2(2):121–4.

[28] Huang CT, Ho CC, Tsai YJ, Yu CJ, Yang PC. Factors influencing visibility and diagnostic yield of transbronchial biopsy using endobronchial ultrasound in peripheral pulmonary lesions. Respirology 2009;14(6):859–64.

[29] Kurimoto N, Murayama M, Yoshioka S, Nishisaka T. Analysis of the internal structure of peripheral pulmonary lesions using endobronchial ultrasonography. Chest 2002;122(6):1887–94.

[30] Eberhardt R, Anantham D, Ernst A, Feller-Kopman D, Herth F. Multimodality bronchoscopic diagnosis of peripheral lung lesions – a randomized controlled trial. Am J Respir Crit Care Med 2007;176(1):36–41.

[31] Fielding DIK, Robinson PJ, Kurimoto N. Biopsy site selection for endobronchial ultrasound guide-sheath transbronchial biopsy of peripheral lung lesions.

Intern Med J 2008;38(2):77–84.

[32] Asano F, Matsuno Y, Tsuzuku A, Anzai M, Shinagawa N, Yamazaki K, et al. Diagnosis of peripheral pulmonary lesions using a bronchoscope insertion guidance system combined with endobronchial ultrasonography with a guide sheath.

Lung Cancer 2008;60(3):366–73.

[33] Gold M, Siegel J, Russel L, Weinstein M. Cost-effectiveness in health and medicine. Oxford university press; 1996. pp. 261–262.

[34] Govert JA, Kopita JM, Matchar D, Kussin PS, Samuelson WM. Cost-effectiveness of collecting routine cytologic specimens during fiberoptic bronchoscopy for endoscopically visible lung tumor. Chest 1996;109(2):451–6.

[35] Roth K, Hardie JA, Andreassen AH, Leh F, Eagan TML. Cost minimization analysis for combinations of sampling techniques in bronchoscopy of endobronchial lesions. Respir Med 2009;103(6):888–94.

[36] van der Drift MA, van der Wilt GJ, Thunnissen FBJM, Janssen JP. A prospective study of the timing and cost-effectiveness of bronchial washing during bronchoscopy for pulmonary malignant tumors. Chest 2005;128(1):394–400.

Please cite this article in press as: Roth K, et al. A randomised trial of endobronchial ultrasound guided sampling in peripheral lung lesions. Lung Cancer (2011), doi:10.1016/j.lungcan.2011.02.013