Endoscopic Tri-Modal Imaging Is More Effective Than Standard Endoscopy in Identifying Early-Stage Neoplasia in Barrett's Esophagus

Imaging and Advanced Technology

Endoscopic Tri-Modal Imaging Is More Effective Than Standard Endoscopy in Identifying Early-Stage Neoplasia in Barrett’s Esophagus WOUTER L. CURVERS,* LORENZA ALVAREZ HERRERO,*,‡ MICHAEL B. WALLACE,§ LOUIS–MICHEL WONG KEE SONG,储 KRISH RAGUNATH,¶ HERBERT C. WOLFSEN,§ GANAPATHY A. PRASAD,储 KENNETH K. WANG,储 VENKATARAMAN SUBRAMANIAN,¶ BAS L. A. M. WEUSTEN,‡ FIEBO J. TEN KATE,# and JACQUES J. G. H. M. BERGMAN* *Department of Gastroenterology and Hepatology, Academic Medical Center, Amsterdam, Netherlands; ‡Department of Gastroenterology and Hepatology, St Antonius Hospital, Nieuwegein, Netherlands; §Division of Gastroenterology and Hepatology, Mayo Clinic, Jacksonville, Florida; 储Division of Gastroenterology and Hepatology, Mayo Clinic, Rochester, Minnesota; ¶Nottingham Digestive Diseases Center and NIHR Biomedical Research Unit, Queens Medical Center, Nottingham, United Kingdom; and #Department of Pathology, Academic Medical Center, Amsterdam, Netherlands

See related article, Longcroft-Wheaton G et al, on page 843 in CGH.

BACKGROUND & AIMS: Endoscopic tri-modal imaging (ETMI) incorporates high-resolution endoscopy (HRE), autofluorescence imaging (AFI), and narrow band imaging (NBI). A recent uncontrolled study found that ETMI improved the detection of high-grade dysplasia (HGD) and early carcinoma (Ca) in Barrett’s esophagus (BE). The aim was to compare ETMI with standard video endoscopy (SVE) for the detection of HGD/Ca with the use of a randomized cross-over design. METHODS: Patients referred for work-up of inconspicuous HGD/Ca were eligible and underwent both SVE and ETMI in randomized order within an interval of 6 –12 weeks. During ETMI, inspection with HRE was followed by AFI. Detected lesions were inspected in detail with NBI and biopsied, followed by random biopsies. During SVE, any visible lesion was biopsied followed by random biopsies. RESULTS: Eighty-seven patients with BE underwent ETMI and SVE. No significant difference was observed in overall histologic yield between ETMI and SVE. ETMI had a significantly higher targeted yield compared with SVE because of AFI. However, the yield of targeted biopsies of ETMI was significantly inferior to the overall yield of SVE. Detailed inspection with NBI reduced the falsepositive rate of HRE ⫹ AFI from 71% to 48% but misclassified 17% of HGD/Ca lesions as not suspicious. CONCLUSIONS: ETMI statistically significant improves the targeted detection of HGD/Ca compared with SVE. Subsequent characterization of lesions with NBI appears to be of limited value. At this stage, ETMI cannot replace random biopsies for detection of lesions or targeted biopsies for characterization of lesions in a high-risk population. GASTROENTEROLOGY 2010;139:1106 –1114

Keywords: Esophageal Adenocarcinoma; Endoscopy; Autofluorescence Imaging; Narrow Band Imaging.

I

n Barrett’s esophagus (BE), the mucosal lining of the distal esophagus undergoes metaplastic change to columnar-type mucosa with the presence of specialized intestinal metaplasia.1 BE is associated with an increased risk of esophageal adenocarcinoma, a disease that is witnessing a steadily rising incidence.2 Endoscopic surveillance of patients with BE is recommended to detect dysplasia or malignancy at an early and curable stage because the prognosis of advanced adenocarcinoma is dismal.3 However, with standard white-light endoscopy, it is difficult to visually detect early neoplastic lesions in BE. In the absence of visible abnormalities, detection of early neoplasia depends on random biopsies that are obtained from the Barrett’s segment, with practice guidelines recommending 4 quadrant biopsies for every 2-cm length of the BE.4,5 Obtaining random biopsies, however, can be labor intensive, and early neoplastic lesion may still be missed because only approximately 5% of the surface area is sampled by this method.6 New endoscopic imaging techniques may improve the detection of early neoplasia in BE. Autofluorescence imaging (AFI) is a promising technique in this respect.7 Normal and neoplastic tissue have different autofluorescence characteristics that may enable their distinction.8 AFI has been incorporated into a real-time videoAbbreviations used in this paper: AFI, autofluorescence imaging; Ca, carcinoma; CCD, charge-coupled device; ETMI, endoscopic tri-modal imaging; HGD, high-grade dysplasia; HRE, high-resolution white light endoscopy; ID, indefinite for dysplasia; IQR, interquartile range; LGD, low-grade dysplasia; NBI, narrow band imaging; RGB, red, green, and blue; SVE, standard video endoscopy. © 2010 by the AGA Institute 0016-5085/$36.00 doi:10.1053/j.gastro.2010.06.045

Imaging and Advanced Technology continued

endoscopy system, and feasibility studies have suggested that its use may improve the detection of early BE neoplasia (eg, high-grade dysplasia [HGD] and carcinoma [Ca]).7,9 AFI, however, is associated with a high falsepositive rate (⬎50%).7,9 This false-positive rate can be reduced by detailed inspection of suspicious lesions with narrow band imaging (NBI), a technique that enhances the visibility of superficial mucosal structures similar to chromoendoscopy with the use of optical filters.9 –11 Recently, a novel endoscopy system was introduced that incorporates high-resolution white light endoscopy (HRE), AFI, and NBI: endoscopic tri-modal imaging (ETMI). In a multicenter study, the AFI component of ETMI increased the targeted detection of HGD/Ca from 53% to 90%, and subsequent characterization of suspicious lesions with NBI reduced the false-positive rate of AFI from 81% to 26%.12 This uncontrolled study, however, lacked a direct comparison with standard video endoscopy (SVE) and random surveillance biopsies. The aim of the current study, therefore, was to compare the diagnostic accuracy of ETMI and SVE for the detection of early neoplasia in BE, in a multicenter, randomized, cross-over trial.

Materials and Methods Setting This multicenter randomized cross-over study was performed in 5 centers with a tertiary referral function for the detection and treatment of patients with early BE neoplasia: Academic Medical Center, Amsterdam, Netherlands; St Antonius Hospital, Nieuwegein, Netherlands; Mayo Clinic, Jacksonville, Florida; Mayo Clinic, Rochester, Minnesota; and Queens Medical Center, Nottingham, United Kingdom. The study was approved by the institutional review boards of all participating centers (ISRCTN68328077).

Patients All patients with BE referred to the participating centers for work-up of endoscopically inconspicuous HGD/Ca were eligible. Patients had to meet the following inclusion criteria: (1) age ⬎ 18 years; (2) prior diagnosis of BE, defined as the presence of columnar-lined epithelium with specialized intestinal metaplasia on histologic investigation; (3) prior diagnosis of HGD/Ca with no endoscopically visible abnormalities according to the referring physician; (4) a minimum Barrett’s length of Cⱖ2, Mⱖ2, or C⬍2, Mⱖ4, according to the Prague C&M classification; and (5) written informed consent. Patients were excluded for the following reasons: (1) presence of active erosive esophagitis grade B or worse according to the Los Angeles classification of erosive esophagitis13; (2) description of an endoscopically visible suspicious lesion in the Barrett’s segment in the referring

center; (3) at first endoscopy: the presence of a type 0 –I or type 0 –III lesion or a lesion that, according to the discretion of the endoscopist, did not allow a delay in intervention for a period of 6 weeks; (4) presence of conditions that precluded safe histologic sampling of the esophagus (eg, esophageal varices, coagulation disorders, anticoagulant therapy).

Technical Background of the ETMI Endoscopy System The ETMI system consists of a high-resolution white-light endoscope with optical zoom (magnification 100⫻; XGIF-Q240/260FZ; Olympus Inc, Tokyo, Japan) equipped with an AFI and NBI mode. This endoscope has 2 separate monochromatic charge-coupled devices (CCDs); one for white-light imaging and NBI, and one for AFI. The light source (Lucera; Olympus Inc) contains red, green, and blue (RGB) band-pass filters placed in a rotating disk in front of a xenon lamp, resulting in sequential RGB illumination. For NBI, the red light band-pass filter is removed, and the band-pass ranges of the green and blue filters are narrowed to 530 –550 nm (green) and 390 – 445 nm (blue), with a relative increase in the intensity of blue light excitation, enhancing imaging of the superficial mucosal and vascular structures.14 In the white-light and NBI mode, the reflected RGB light is detected by a monochromatic CCD and converted into an electrical signal that is transmitted to the video processor (Lucera, Olympus Inc). The processor is synchronized with the rotary filter of the light source and electronically overlays the reflected RGB light to produce a white-light or NBI image. In the AFI mode, the image is composed of total emitted autofluorescence after blue light excitation (390 – 470 nm) and green reflectance (540 –560 nm). A barrier filter is placed in front of the AFI-CCD to allow passage of fluorescent light with a wavelength between 500 and 630 nm and to block the blue light excitation. In the pseudo-colored image unsuspicious AFI areas appear green, whereas suspicious areas appear dark purplish in color. All 3 imaging modalities of the ETMI system provide real-time endoscopic images. The endoscopist can switch from one modality to another in 1–2 seconds by pushing control buttons on the handle of the endoscope.

Study Design, Randomization, and Experience of the Endoscopists All patients underwent 2 consecutive endoscopies. One procedure was performed with an SVE system, and the other with the ETMI system. In this way every patient acted as his or her own control. Between both procedures an interval of 6 –12 weeks allowed biopsy sites to heal sufficiently. At each center all procedures were performed by 2 endoscopists with extensive experience in the detec1107

Imaging and Advanced Technology continued

tion and treatment of early BE neoplasia. One of the endoscopists was assigned to the first procedure before randomization to SVE or ETMI on the basis of availability. The second procedure was automatically allocated to the other endoscopist. Randomization of the technique was done at the first endoscopy by sealed opaque envelopes to ensure concealed allocation. Randomization envelopes were provided to the participating centers by the study coordinator. Before the procedures both endoscopists had the same clinical information, and the endoscopist assigned to the second procedure was blinded to the results of the first procedure. At each center, a study coordinator attended all endoscopic procedures for uniform data collection. Endoscopists participating in this study had extensive experience with the ETMI system before this study. All centers participated in the aforementioned uncontrolled feasibility study and had the ETMI system available long before the start of the current study.12 In preparation of the ETMI feasibility study and the current study the endoscopists were provided with an instructional DVD that showed examples of AFI-positive areas, HRE and NBI images of regular and irregular mucosal, and vascular patterns, as well as instructions on the study protocol. They also participated in the aforementioned uncontrolled feasibility study on ETMI in BE and several interobserver studies on AFI and NBI images in BE. During this process all centers were visited twice by the principal study coordinator to supervise ETMI procedures, to ensure uniform data collection, and to optimize the practicality of the study design.

Endoscopic Procedures Patients were sedated according to the standard protocol of the participating centers, mostly intravenous midazolam (2.5–15 mg) supplemented with fentanyl (0.1– 0.2 mg) or pethidine (50 –100 mg) if necessary. Total procedure time was recorded and, at 2 centers (Amsterdam and Nieuwegein) each part of the procedure was separately scored for time. ETMI procedure. The esophagus was first examined with HRE without magnification. Presence and length of the Barrett’s segment and/or hiatal hernia were recorded according to the Prague C&M classification.15 The Barrett’s segment was inspected for the presence of visible abnormalities. For each suspicious lesion, the location was recorded. Subsequently, the Barrett’s segment was inspected with AFI for the presence of additional lesions. Inspection with HRE and AFI was performed in the antegrade as well as in retroflexed position. For all lesions the technique that primarily led to their detection was noted, followed by detailed inspection of all suspicious lesions with NBI. For this purpose a black flexible cap (MB-046; Olympus Inc) was mounted on the tip of the endoscope before starting the procedure to 1108

facilitate magnification endoscopy. The following NBI characteristics were documented: mucosal pattern (regular, irregular, or flat), vascular pattern (regular or irregular), and presence of abnormal blood vessels. Finally, the overall NBI appearance was classified in 3 categories: suspicious for neoplasia, not suspicious for neoplasia, or indeterminate. SVE procedure. SVE was performed (Olympus GIF-140, GIF-160). The esophagus was inspected, and the presence and length of the Barrett’s segment and/or hiatal hernia were recorded according to the Prague C&M classification.15 The Barrett’s segment was inspected in the antegrade as well as in retroflexed position for the presence of visible abnormalities. For all suspicious lesions the location was recorded.

Histologic Assessment All biopsy specimens were routinely processed and evaluated in the participating centers. After routine histologic assessment all specimens were subsequently reviewed by one expert gastrointestinal pathologist at each center. The pathologists were blinded to the allocated endoscopic modality. On standardized forms the histologic outcome was recorded according to the revised Vienna classification of gastrointestinal neoplasia in the following categories: nondysplastic BE, indefinite for dysplasia (ID), low-grade dysplasia (LGD), HGD, or Ca.16

Outcome Parameters The overall histologic yield of ETMI and SVE was defined as the highest grade of neoplasia diagnosed in any biopsy specimen (ie, targeted or random) obtained during ETMI and SVE, respectively. The targeted histologic yield of ETMI and SVE was defined as the highest grade of neoplasia diagnosed in any targeted biopsy specimen obtained from visible abnormalities identified during ETMI or SVE. The histologic diagnosis of a visible abnormality was defined as the highest grade of neoplasia diagnosed in any targeted biopsy obtained from the corresponding abnormality for either ETMI or SVE. Given inherent uncertainties in matching lesion locations across separate procedures, no attempt was made to correlate size, appearance, or location of abnormalities detected in ETMI and SVE procedures. Primary outcomes. Primary outcomes included the following: (1) the overall histologic yield of ETMI and SVE, (2) the targeted histologic yield of ETMI and SVE, and (3) the number of patients diagnosed with HGD/Ca by ETMI and SVE. Secondary outcomes. The secondary outcomes included the following: (1) the number of visible abnormalities with HGD/Ca detected by ETMI and SVE, (2) accuracy of NBI for detailed inspection of visible abnor-

Imaging and Advanced Technology continued Table 1. Overall Histologic Yield of ETMI and SVE in 87 Patients (P ⫽ .150) ETMI (n)

SVE (n) NDBE ID/LGD HGD/CA

NDBE

ID/LGD

HGD/Ca

2 2 2

5 23 7

7 8 31

Ca, carcinoma; ETMI, endoscopic tri-modal imaging; HGD, high-grade dysplasia; ID, indefinite for dysplasia; LGD, low-grade dysplasia; NDBE, nondysplastic Barrett’s esophagus; SVE, standard video endoscopy.

malities identified with HRE and/or AFI, and (3) procedure times and number of biopsy specimens obtained.

Sample Size and Statistical Analysis On the basis of the findings of our studies on endoscopic video autofluorescence endoscopy and previous randomized cross-over imaging studies that used the same selection criteria by our group, 50% of included patients were expected to have HGD and/or Ca.7,9,17,18 SVE with targeted biopsies was expected to detect 50% of these patients. We calculated that 84 patients would be required to detect 25% identifying early neoplasia between SVE and ETMI with a power of 90% and a significance level of 5%. All statistical analyses were performed with the use of a statistical software package (Statistical Package for the Social Sciences 12.0.1; SPSS Inc, Chicago, IL). Continuous variables with normal distribution were summarized by the mean and standard deviation (SD), whereas those with a skewed distribution were summarized by the median and the interquartile range (IQR). When appropriate the McNemar’s test, Wilcoxon signed-rank test, or the paired t test were used. For comparing the primary outcome measures as paired evaluations with multiple outcome categories the generalized McNemar–Bowker test of symmetry was used. For all pairwise comparisons of the primary outcome measures diagnosis of ID was grouped with LGD in one category (6% of targeted biopsies and 2% of random biopsies showed ID).

did not meet the inclusion criteria, and 3 patients withdrew consent after the first procedure. Eighty-seven patients underwent both procedures and were analyzed (mean age, 67.1 ⫾ 9.1 years; 71 males [82%]; median circumferential Barrett’s extend, 4.0 cm [IQR, 2.0 – 8.0]; median maximal Barrett’s extend, 7.0 cm [IQR, 4.0 – 10.0]). The median time interval between both procedures was 8.1 weeks (SD, 2.6 weeks). On the basis of the combined histologic outcome of the 2 endoscopic procedures, the overall histologic diagnosis in these 87 patients was as follows: nondysplastic BE in 3, ID in 1, LGD in 28, and HGD/Ca in 55. Thirty-eight patients were included in Amsterdam and Nieuwegein, 19 in Jacksonville, 15 in Rochester, and 15 in Nottingham. We found no significant differences in detection between the participating centers.

Overall Histologic Yield of ETMI and SVE When combining targeted and random biopsies, ETMI detected more patients with higher grades of neoplasia than SVE (Table 1). In 56 patients ETMI and SVE yielded the same histologic diagnosis. In 20 patients, ETMI yielded a higher histologic grade than SVE, whereas in 11 patients SVE had a higher histologic yield (Table 1; P ⫽ .15; McNemar–Bowker test of symmetry).

Targeted Histologic Yield of ETMI and SVE When considering targeted biopsies alone, ETMI significantly increased the detection of neoplasia compared with SVE (Table 2). In 43 patients, targeted sampling during ETMI yielded a higher histologic grade than SVE in contrast to 7 patients in whom SVE yielded a higher grade than ETMI (Table 2; P ⬍ .001; McNemar– Bowker test of symmetry). Table 3 shows the relative contributions of HRE and AFI in the targeted detection of ETMI. SVE detected visible abnormalities with HGD/Ca in 24 patients. ETMI detected 36 patients with visible abnormalities containing HGD/Ca: 24 patients Table 2. Targeted Histologic Yield of Endoscopically Visible Abnormalities Detected With ETMI and SVE in 87 Patients (P ⬍ .001) ETMI (n)

Results Patients Between January 2007 and September 2009, 111 eligible patients (age mean ⫾ SD age, 68.0 ⫾ 9.0 yeas; 92 males [83%]; median circumferential Barrett’s extent, 4.0 cm [IQR. 2.0 – 8.0]; median maximal Barrett’s extent, 7.0 cm [IQR, 4.0 –9.0]) were included in the participating centers. Twenty-four patients were excluded after the first procedure (Supplementary Figure 1): 17 patients exhibited a type 0 –I or type 0 –III lesion that did not allow a delay in intervention, 4 patients had Barrett’s length that

SVE (n) No lesiona NDBE ID/LGD HGD/Ca

No lesion

NDBE

ID/LGD

HGD/Ca

8 1 0 1

12 0 2 1

14 1 9 2

11 3 2 20

Ca, carcinoma; ETMI, endoscopic tri-modal imaging; HGD, high-grade dysplasia; ID, indefinite for dysplasia; LGD, low-grade dysplasia; NDBE, nondysplastic Barrett’s esophagus; SVE, standard video endoscopy. aNo visible abnormalities were detected and no targeted biopsies were obtained. 1109

Imaging and Advanced Technology continued

Table 3. Detection of Patients With HGD/Ca and LGD/HGD/Ca by Random Plus Targeted Biopsies and by Targeted Sampling Only in 87 Patients SVE

ETMI

Total

No. of patients detected with HGD/Ca (random and targeted biopsies) No. of patients with HGD/Ca detected with targeted biopsies

40 24

55

No. of patients detected with LGD/HGD/Ca (random and targeted biopsies) No. of patients with LGD/HGD/Ca detected with targeted biopsies

71 34

46 HRE: 24 HRE ⫹ AFI: 36 76 HRE: 36 HRE ⫹ AFI: 57

83

AFI, autofluorescence imaging; Ca, carcinoma; ETMI, endoscopic tri-modal imaging; HGD, high-grade dysplasia; HRE, high-resolution endoscopy; LGD, low-grade dysplasia; NDBE, nondysplastic Barrett’s esophagus; SVE, standard video endoscopy.

were detected with HRE, whereas, in 12 patients, visible abnormalities were only detected with AFI and not seen during initial inspection with HRE. AFI, therefore, significantly increased the targeted detection of HGD/Ca of ETMI compared with SVE (P ⫽ .012; McNemar’s test). The results for targeted detection of LGD/HGD/Ca (instead of diagnosing just HGD/Ca) were found to be comparable (Table 3).

Number of Patients Diagnosed With HGD/Ca by ETMI and SVE In total 55 patients were diagnosed with HGD/Ca after ETMI (n ⫽ 46) and/or SVE (n ⫽ 40) (Table 3).

In 15 patients HGD/Ca was diagnosed by ETMI but missed by SVE. In 9 of these cases (60%), ETMI detected visible lesions with HGD/Ca: 5 had lesions detected with both HRE and AFI, and in 4 patients the lesions were detected with AFI only (Figure 1). In the remaining 6 patients HGD/Ca was diagnosed by the random sampling during ETMI. In 9 patients HGD/Ca was diagnosed with SVE but missed by ETMI. In one patient SVE detected a visible lesion containing HGD/Ca that was not detected during ETMI. In the remaining 8 patients HGD/Ca was diagnosed in random biopsies obtained during SVE.

Figure 1. (A–C) A lesion containing HGD/Ca is depicted at the 11 o’clock position. (A) A more prominent lesion located at the 5 o’clock position is shown. Both lesions were detected with HRE and AFI. NBI showed irregular mucosal and vascular patterns suspicious for dysplasia. (D and E) Two lesions with HGD/Ca located at the 2 and the 5 o’clock positions are shown. These lesions were missed during HRE but were subsequently detected with AFI (E). NBI (F) showed irregular mucosal and vascular patterns and abnormal blood vessels suspicious for dysplasia. AFI, autofluorescence imaging; Ca, Carcinoma; HGD, high-grade dysplasia; HRE, high-resolution endoscopy; NBI, narrow band imaging. 1110

Imaging and Advanced Technology continued

Figure 2. Images G and H show an AFI false-positive lesion. Detailed inspection with NBI (Image I) showed regular mucosal and regular vascular patterns not suspicious for dysplasia. Histopathology revealed non-dysplastic BE. AFI: autofluoresence imaging; NBI: narrow band imaging; BE: Barrett’s eosphagus.

The results for diagnosing LGD/HGD/Ca (instead of diagnosing just HGD/Ca) were found to be comparable (Table 3).

Diagnosis of Visible Abnormalities With HGD/Ca by ETMI and SVE SVE detected a total of 68 visible abnormalities. Thirty-two areas contained HGD/Ca and 17 contained LGD. The false-positive rate of SVE for HGD/Ca was therefore 53% (36/68). During ETMI, HRE detected a total of 94 lesions: biopsies showed HGD/Ca in 38 lesions, whereas LGD was diagnosed in 24. HRE, therefore, had a false-positive rate of 60% for the targeted detection of HGD/Ca (56/94). After inspection with HRE, AFI detected an additional 135 areas with an abnormal autofluorescence appearance (28 HGD/Ca and 35 LGD). The false-positive rate of AFI for detecting additional lesions with HGD/Ca after HRE therefore was 79% (107/ 135).

NBI Evaluation During ETMI NBI imaging of suspicious lesions was able to reduce the false-positive rate of AFI but did so at the expense of misclassifying some neoplastic lesions as nonneoplastic. AFI and HRE detected a total of 229 visible abnormalities. Detailed inspection of the mu-

cosal and vascular patterns of these detected lesions with NBI was successfully performed in 228 areas (99.6%). Of the 65 lesions with HGD/Ca, 54 were abnormal on NBI evaluation, but NBI misclassified 11 HGD/Ca lesions as unsuspicious (ie, false-negative rate, 17%) (Figure 1). Of the 163 lesions that did not contain HGD/Ca, 110 were classified as suspicious on NBI, whereas 53 were scored as unsuspicious/indeterminate (Figure 2). NBI therefore reduced the falsepositive rate of ETMI (HRE ⫹ AFI) from 71% (163/ 229) to 48% (110/229).

Procedure Time and Number of Biopsies Mean (⫾SD) procedure time of SVE was 15:03 ⫾ 6:42 minutes compared with 25:55 ⫾ 9:33 minutes for ETMI (P ⬍ .001). Detailed analysis of the 38 patients enrolled in 2 of the participating centers showed that the differences in procedure times between both techniques were due to additional inspection time for AFI and NBI (Table 4). The total number of biopsies obtained during ETMI and SVE was comparable, yet more random biopsy samples were obtained during SVE than during ETMI (14 vs 11.5; P ⬍ .001), whereas more targeted biopsy samples were obtained during ETMI (2 vs 5; P ⬍ .001).

Table 4. Procedure Times of SVE, HRE, AFI, NBI, and Tissue Sampling in 38 Patients

Inspection time, minutes:seconds (IQR) HRE AFI NBI Targeted biopsy sampling time, minutes:seconds (IQR) Random biopsy sampling time, minutes:seconds (IQR) Total procedure time, minutes:seconds (IQR)

SVE

ETMI

P

5:10 (4:10–7:04) — — — 1:22 (0:00–2:50) 6:18 (4:03–8:30) 13:30 (9:45–17:27)

13:44 (10:56–16:08) 5:57 (5:24–7:03) 4:14 (2:48–5:59) 3:07 (1:20–4:37) 3:13 (1:47–6:37) 4:15 (3:44–6:11) 22:07 (15:52–30:42)

⬍.001

⬍.001 .005 ⬍.001

AFI, autofluorescence imaging; ETMI, endoscopic tri-modal imaging; HRE, high-resolution endoscopy; IQR, interquartile range; NBI, narrow band imaging; SVE, standard video endoscopy. 1111

Imaging and Advanced Technology continued

Table 5. Yield of ETMI-Targeted Biopsies Only Compared With the Overall Yield of SVE (Targeted Plus Random Biopsies) in 87 patients (P ⫽ .02)

grade than ETMI, whereas in 11 patients ETMI had a higher histologic yield (Table 5; P ⫽ .02; McNemar– Bowker test of symmetry).

ETMI targeted NDBE SVE (targeted ⫹ random) NDBE ID/LGD HGD/Ca

8 13 4

ID/LGD 2 15 9

HGD/Ca 4 5 27

NOTE. Patients with no lesions detected during ETMI were grouped in the NDBE group. Ca, carcinoma; ETMI, endoscopic tri-modal imaging; HGD, high-grade dysplasia; ID, indefinite for dysplasia; LGD, low-grade dysplasia; NDBE, nondysplastic Barrett’s esophagus; SVE, standard video endoscopy.

Random Sampling Effect of Targeted Biopsies We performed a post hoc analysis to evaluate if part of the increase in targeted detection rate of AFI might be explained by a random sampling effect (ie, the possibility that the “gain” seen by AFI is simply due to taking more biopsies, not necessarily targeted biopsies). AFI identified 12 additional patients with HGD/Ca by detecting 135 lesions with an abnormal autofluorescence of which 28 contained HGD/Ca compared with HRE alone. AFI-targeted biopsies, therefore, had a detection rate of 21% for HGD/Ca. Random biopsies during SVE diagnosed 16 additional patients with HGD/Ca. HGD/Ca was detected in 66 of 1344 random biopsy samples which corresponds to a HGD/Ca detection rate of 5% per random biopsy, significantly less than that of AFI (P ⬍ .001). In other words, AFI detected 12 additional patients with HGD/Ca in 135 biopsy sample, whereas random sampling during SVE with an equal number of biopsy samples would have detected 1.6 additional patients with HGD/Ca (16/1344 ⫻ 135).

Targeted Histologic Yield of ETMI Versus Overall Histologic Yield of SVE To test the hypothesis that ETMI-targeted biopsies alone are sufficient, we compared the yield of this approach with the current standard of SVE-targeted plus random biopsy. Overall, we found the ETMI-targeted biopsies alone were less accurate than the current SVE approach. The additional 28 HGD/Ca lesions detected by AFI were diagnosed in 21 patients. In 13 of these patients, the diagnosis of HGD/Ca was also made in biopsies obtained from lesions identified with HRE (n ⫽ 9) and/or by random biopsies (n ⫽ 9). Eight patients with HGD/Ca in the ETMI group were solely diagnosed by AFI. The targeted histologic yield of ETMI versus the overall histologic yield of SVE is presented in Table 5. In 50 patients ETMI and SVE yielded the same histologic diagnosis. In 26 patients SVE yielded a higher histologic 1112

Discussion This is the first randomized cross-over study to compare ETMI with SVE in patients referred for endoscopic evaluation of early BE neoplasia. The randomized cross-over design ensured that each patient acted as his or her own control, allowing pairwise comparison of the 2 techniques investigated. Although expensive and labor intensive, this is considered the optimal method for endoscopic imaging studies. The study was performed in 5 centers with a tertiary referral function for early Barrett’s neoplasia. Participating endoscopists were experienced in the endoscopic recognition of early neoplastic lesions and participated in several studies on AFI and/or NBI in BE before the current study.12,19,20 It is, therefore, unlikely that a learning effect influenced the findings of the study. We found no significant difference in detection of dysplasia by ETMI and SVE when the histologic yield of random and targeted biopsies was combined. ETMI did, however, result in a significant increase in the targeted detection of dysplasia compared with SVE, because of the additional detection by AFI with comparable yields of SVE and HRE. The increase in targeted detection rate by AFI was not explained by a random sampling effect due to simply obtaining more targeted biopsies with AFI. This suggests that multimodality imaging may increase the targeted detection of dysplasia in BE by providing information on other unique features of tissue-light interaction that are not apparent with standard white-light imaging. There may be some bias with regard to this conclusion. First, HRE and AFI inspections were performed sequentially by the same endoscopist. The AFI assessment may, therefore, have been biased by the HRE findings. Second, the inspection time with HRE and AFI was doubled compared with inspection time with SVE. In the current design the SVE procedure was not controlled for double inspection time and may, therefore, have led to an overestimation of the detection rate of AFI. NBI was used to characterize suspicious lesions detected by HRE and AFI in an attempt to reduce the false-positive rate. The value of NBI for detailed inspection of suspicious lesions is questionable. Magnification endoscopy with NBI is operator dependent, and, despite the experience of the endoscopists in the current study, it increased the duration of the ETMI procedure almost to the same extent as inspection with AFI. Just taking targeted biopsy samples of suspicious areas is probably easier and faster. More importantly, the accuracy of NBI in correctly characterizing lesions with HGD/Ca was dis-

Imaging and Advanced Technology continued

appointing: 17% of HGD/Ca lesions were falsely misclassified by NBI, and the false-positive rate of HRE ⫹ AFI was only marginally reduced from 71% to 48%. A subanalysis showed that 10 of the 11 HGD/Ca lesions that were misclassified with NBI as unsuspicious or indeterminate were found in 2 centers that performed each other’s cross-over procedures (Amsterdam and Nieuwegein). We compared the results of these 2 centers with the other 3. False-positive rates of ETMI before NBI evaluation were the same in both groups (68% vs 74%). The 2 centers with the highest reduction in false-positive rate (from 68% to 36%) were also responsible for most of the misclassified lesions (10 lesions; 9.9%), whereas centers with a low misclassification rate (0.8%) only had a marginal reduction in false-positive rate (from 74% to 58%). We postulate that a high threshold for calling lesions suspicious with NBI leads to a higher rate of misclassification of lesions containing HGD/Ca as unsuspicious. A lower threshold results in only a marginal reduction of the ETMI false-positive rate. Other reasons for misclassification of HGD/Ca lesions as unsuspicious with NBI may be endoscopic sampling error during NBI evaluation or biopsy sampling error. The NBI results of the current study are in accordance with recent interobserver studies and suggest that detailed inspection with NBI cannot replace histologic sampling of suspicious areas in BE.19,21,22 The ETMI system (Lucera; Olympus Inc) used in the current study is not commercially available in the United States. In addition, the NBI system we used has some distinct features (eg, optical magnification) than the NBI system (Exera; Olympus Inc) that is available in the United States. A recent study by Wolfsen et al23 suggested that the NBI system available in the United States may improve the detection of dysplasia in BE. This study, however, used a tandem endoscopy design in which standard endoscopy was followed by NBI endoscopy in the same session. The study, therefore, had significant methodologic limitations that led us to conduct the current project with the use of a randomized cross-over design.24 Note, however, that we performed our study with a different NBI system and only used NBI to characterize lesions detected with HRE and AFI and did not study NBI as a primary detection tool as in the study by Wolfsen et al.23 The study findings imply that AFI improves the targeted detection of HGD/Ca in high-risk patients with BE. AFI may be of additional value in the work-up and treatment of patients with early Barrett’s neoplasia. Endoscopic treatment in these patients requires precise localization of neoplastic lesions within the Barrett’s segment to allow for accurate endoscopic resection that is imperative for optimal staging and patient selection. Our study, however, does not allow us to assess the effect of

AFI-detected lesions on decisions about the therapeutic management on a per patient level. Although AFI may improve the targeted detection of early neoplasia, the results of our study suggest that multimodality imaging cannot replace random biopsy sampling in BE. When we compared the overall histologic yield of SVE with the targeted histologic yield of ETMI, the overall histologic yield of SVE was significantly higher than the targeted histologic yield of ETMI. In addition, both ETMI and SVE missed a significant number of patients with HGD/Ca, and with both techniques a considerable number of patients were solely detected with random sampling. Our study lacked a central pathology review. All participating centers, however, are tertiary referral centers for the detection and treatment of Barrett’s neoplasia with extensive histopathologic expertise in the field of BE. All pathology specimens were evaluated by ⱖ2 pathologists, including dedicated review for the purpose of this study by an expert gastrointestinal pathologist at each center. Furthermore, given the randomized crossover design patients acted as their own control; therefore, individual outcomes are not likely to be influenced by interobserver variations in the histopathologic evaluation between different centers. Our study population (patients referred for the workup of inconspicuous HGD/Ca), had a high pretest likelihood for early neoplasia, as reflected by the high rate of dysplasia that was diagnosed. In addition, participating endoscopists were highly experienced in the recognition of subtle abnormalities suspicious for early neoplasia, resulting in high detection rates of dysplasia even with standard endoscopy equipment. In a BE population with a lower rate of neoplasia and endoscopists without particular expertise in BE, the additional value of AFI and NBI may be different. A trial comparing ETMI and SVE in a randomized cross-over fashion in which procedures are performed by community gastroenterologists in patients with an intermediate risk profile (ie, confirmed LGD) is currently under way. In conclusion, ETMI significantly improved the targeted detection of HGD/Ca compared with SVE. This improved detection of ETMI is due to the additional detection of lesions containing HGD/Ca by AFI that are not apparent with HRE. However, the targeted histologic yield of ETMI is still inferior to the overall histologic yield (random ⫹ targeted biopsies) of SVE. Detailed inspection with NBI to characterize lesions identified by HRE and AFI only marginally reduced the false-positive rate but misclassified a significant number of HGD/Ca lesions as not suspicious. At this stage, therefore, multimodality imaging with the current ETMI system cannot replace random biopsies for the detection of 1113

Imaging and Advanced Technology continued

early neoplasia or the use of targeted biopsies for characterization of lesions in a high-risk population.

Supplementary Material Note: To access the supplementary material accompanying this article, visit the online version of Gastroenterology at www.gastrojournal.org, and at doi: 10.1053/j.gastro.2010.06.045. References 1. Spechler SJ. Clinical practice. Barrett’s esophagus. N Engl J Med 2002;346:836 – 842. 2. Devesa SS, Blot WJ, Fraumeni JF Jr. Changing patterns in the incidence of esophageal and gastric carcinoma in the United States. Cancer 1998;83:2049 –2053. 3. van Sandick JW, van Lanschot J, Kuiken BW, Tytgat GN, Offerhaus GJ, Obertop H. Impact of endoscopic biopsy surveillance of Barrett’s oesophagus on pathological stage and clinical outcome of Barrett’s carcinoma. Gut 1998;43:216 –222. 4. Sampliner RE. Updated guidelines for the diagnosis, surveillance, and therapy of Barrett’s esophagus. Am J Gastroenterol 2002; 97:1888 –1895. 5. Hirota WK, Zuckerman MJ, Adler DG, et al. ASGE guideline: the role of endoscopy in the surveillance of premalignant conditions of the upper GI tract. Gastrointest Endosc 2006;63:570 –580. 6. Tschanz ER. Do 40% of patients resected for Barrett esophagus with high-grade dysplasia have unsuspected adenocarcinoma? Arch Pathol Lab Med 2005;129:177–180. 7. Kara MA, Peters FP, ten Kate FJ, van Deventer SJ, Fockens P, Bergman JJ. Endoscopic video autofluorescence imaging may improve the detection of early neoplasia in patients with Barrett’s esophagus. Gastrointest Endosc 2005;61:679 – 685. 8. Kara M, Dacosta RS, Wilson BC, Marcon NE, Bergman J. Autofluorescence-based detection of early neoplasia in patients with Barrett’s esophagus. Dig Dis 2004;22:134 –141. 9. Kara MA, Peters FP, Fockens P, ten Kate FJ, Bergman JJ. Endoscopic video-autofluorescence imaging followed by narrow band imaging for detecting early neoplasia in Barrett’s esophagus. Gastrointest Endosc 2006;64:176 –185. 10. Kara MA, Bergman JJ. Autofluorescence imaging and narrowband imaging for the detection of early neoplasia in patients with Barrett’s esophagus. Endoscopy 2006;38:627– 631. 11. Kara MA, Ennahachi M, Fockens P, ten Kate FJ, Bergman JJ. Detection and classification of the mucosal and vascular patterns (mucosal morphology) in Barrett’s esophagus by using narrow band imaging. Gastrointest Endosc 2006;64:155–166. 12. Curvers WL, Singh R, Song LM, et al. Endoscopic tri-modal imaging for detection of early neoplasia in Barrett’s oesophagus: a multi-centre feasibility study using high-resolution endoscopy, autofluorescence imaging and narrow band imaging incorporated in one endoscopy system. Gut 2008;57:167–172. 13. Lundell LR, Dent J, Bennett JR, et al. Endoscopic assessment of oesophagitis: clinical and functional correlates and further validation of the Los Angeles classification. Gut 1999;45:172–180. 14. Gono K, Obi T, Yamaguchi M, et al. Appearance of enhanced tissue features in narrow-band endoscopic imaging. J Biomed Opt 2004;9:568 –577. 15. Sharma P, Dent J, Armstrong D, et al. The development and validation of an endoscopic grading system for Barrett’s esophagus: the Prague C & M criteria. Gastroenterology 2006;131: 1392–1399.

1114

16. Schlemper RJ, Riddell RH, Kato Y, et al. The Vienna classification of gastrointestinal epithelial neoplasia. Gut 2000;47:251–255. 17. Kara MA, Smits ME, Rosmolen WD, et al. A randomized crossover study comparing light-induced fluorescence endoscopy with standard video endoscopy for the detection of early neoplasia in Barrett’s esophagus. Gastrointest Endosc 2005;61:671– 678. 18. Kara MA, Peters FP, Rosmolen WD, et al. High-resolution endoscopy plus chromoendoscopy or narrow-band imaging in Barrett’s esophagus: a prospective randomized crossover study. Endoscopy 2005;37:929 –936. 19. Curvers WL, Bohmer CJ, Mallant-Hent RC, et al. Mucosal morphology in Barrett’s esophagus: interobserver agreement and role of narrow band imaging. Endoscopy 2008;40:799 – 805. 20. Curvers WL, Singh R, Wallace MB, et al. Identification of predictive factors for early neoplasia in Barrett’s esophagus after autofluorescence imaging: a stepwise multicenter structured assessment. Gastrointest Endosc 2009;70:9 –17. 21. Curvers W, Baak L, Kiesslich R, et al. Chromoendoscopy and narrow-band imaging compared with high-resolution magnification endoscopy in Barrett’s esophagus. Gastroenterology 2008; 134:670 – 679. 22. Herrero LA, Curvers WL, Bansal A, et al. Zooming in on Barrett oesophagus using narrow-band imaging: an international observer agreement study. Eur J Gastroenterol Hepatol 2009;21: 1068 –1075. 23. Wolfsen HC, Crook JE, Krishna M, et al. Prospective, controlled tandem endoscopy study of narrow band imaging for dysplasia detection in Barrett’s esophagus. Gastroenterology 2008;135: 24 –31. 24. Curvers WL, Bergman JJ. Multimodality imaging in Barrett’s esophagus: looking longer, seeing better, and recognizing more. Gastroenterology 2008;135:297–299.

Received April 6, 2010. Accepted June 9, 2010. Reprint requests Address requests for reprints to: Jacques Bergman, MD, PhD, Department of Gastroenterology and Hepatology, Academic Medical Center, Meibergdreef 9, 1105 AZ, Amsterdam, The Netherlands. e-mail: [email protected]; fax: (31) 20 691 7033. Conflicts of interest The authors disclose the following: Louis-Michel Wong Kee Song has received research support from Olympus and Fujinon. Krish Ragunath has received research support, educational grants, and speaker honorarium from Olympus-Keymed, United Kingdom. Ganapathy Prasad has received funding from Takeda Pharmaceuticals, NIH/NCI, and the American College of Gastroenterology. Wouter Curvers, Lorenza Alvarez Herrero, Michael B. Wallace, Herbert Wolfsen, Kenneth Wang, Venkataraman Subramanian, Bas Weusten, Fiebo ten Kate, and Jacques Bergman have no conflict of interest to disclose. Funding This study was supported by an unrestricted research grant from Olympus Inc, Tokyo, Japan. The work of Wouter Curvers and Lorenza Alvarez Herrero is supported by an unrestricted research grant from Astra-Zeneca Netherlands. This trial was registered at www.trialregister.nl as ISRCTN 68328077; NTR945.

October 2010

Imaging and Advanced Technology

1114.e1

Supplementary Figure 1. Flow chart of eligible patients. ETMI, endoscopic tri-modal imaging; pt, patient; SVE, standard video endoscopy.