ORIGINAL ARTICLE: Clinical Endoscopy
Multicenter, randomized, controlled trial of confocal laser endomicroscopy assessment of residual metaplasia after mucosal ablation or resection of GI neoplasia in Barrett’s esophagus Michael B. Wallace, MD, MPH,1 Julia E. Crook, PhD,1 Michael Saunders, MD,2 Laurence Lovat, MD,4 Emmanuel Coron, MD,3 Irving Waxman, MD,5 Prateek Sharma, MD,6 Joo H. Hwang, MD,2 Matthew Banks, MD,4 Mathieu DePreville, MD,3 Jean P. Galmiche, MD,3 Vani Konda, MD,5 Nancy N. Diehl, MD,1 Herbert C. Wolfsen, MD1 Jacksonville, Florida, USA
Background: Endoscopic ablation is an accepted standard for neoplasia in Barrett’s esophagus (BE). Eradication of all glandular mucosa in the distal esophagus cannot be reliably determined at endoscopy. Objective: To assess if use of probe-based confocal laser endomicroscopy (pCLE) in addition to high-definition white light (HDWL) could aid in determination of residual BE. Design: Prospective, multicenter, randomized, clinical trial. Setting: Academic medical centers. Patients: Patients with Barrett’s esophagus undergoing ablation. Intervention: After an initial attempt at ablation, patients were followed-up either with HDWL endoscopy or HDWL plus pCLE, with treatment of residual metaplasia or neoplasia based on endoscopic findings and pCLE used to avoid overtreatment. Main Outcome Measurements: The proportion of optimally treated patients, defined as those with residual BE who were treated and had complete ablation plus those without BE who were not treated and had no evidence of disease at follow-up. Results: The study was halted at the planned interim analysis based on a priori criteria. After enrollment was halted, all patients who had been randomized were followed to study completion. Among the 119 patients with follow-up, there was no difference in the proportion of patients achieving optimal outcomes in the two groups (15/57, 26% for HDWL; 17/62, 27% with HDWL ⫹ pCLE). Other outcomes were similar in the two groups. Limitations: The study was closed after the interim analysis due to low conditional power resulting from lack of difference between groups as well as higher-than-expected residual Barrett’s esophagus in both arms. Conclusion: This study yields no evidence that the addition of pCLE to HDWL imaging for detection of residual Barrett’s esophagus or neoplasia can provide improved treatment. (Clinical trial registration number: NCT01032044.) (Gastrointest Endosc 2012;76:539-47.) (footnotes appear on last page of article)
Barrett’s esophagus (BE) is the premalignant lesion for adenocarcinoma of the esophagus. The incidence of esophageal adenocarcinoma has been rapidly rising and has increased 3-fold to 6-fold over the past 2 decades.1
Use your mobile device to scan this QR code and watch the author interview. Download a free QR code scanner by searching ‘QR Scanner’ in your mobile device’s app store. www.giejournal.org
Once BE with high-grade neoplasia (HGN) or early adenocarcinoma is detected, endoscopic therapy is advocated.2,3 Then the patient undergoes periodic surveillance until there is reasonable certainty of complete ablation, documented by biopsies. In this context, it would be valuable to have direct, on-table methods to detect residual disease in order to guide treatment. The aim of the study was to evaluate the impact of using probe-based confocal laser endomicroscopy (pCLE) to guide ablative treatments in BE. In this setting, the primary goal of ablation is eradiation of all BE metaplasia and neoplasia and doing this in the least invasive way Volume 76, No. 3 : 2012 GASTROINTESTINAL ENDOSCOPY 539
GI neoplasia in Barrett’s esophagus
possible. Thus, the primary aim was to assess whether addition of pCLE to confirm findings with high-definition white light (HDWL) would result in a higher proportion of optimally treated patients, defined as those with residual BE who were treated and had complete ablation, plus those without BE who were not treated and had no evidence of disease at follow-up.
METHODS
Wallace et al
Take-home Message ●
●
This study yields no evidence that the addition of probedbased confocal laser endomicroscopy to high-definition white light imaging for detection of residual Barrett’s esophagus or dysplasia can provide improved treatment. Further advances in imaging and tissue assessment technology are needed before these can replace biopsy as a method to assess completeness of Barrett’s esophagus ablation.
The study was approved by the Human Subjects Committee for each institution.
Before-study run-in and standardization Each investigator was provided with a standardized esophagus pCLE and endoscopic images including 10 known high-grade dysplasia (HGD) or cancer cases, 10 known metaplasia cases, and 10 known non-BE cases. This was followed by 20 images of unknown cases, of which at least 5 sites had intraepithelial neoplasia or cancer, and 5 sites had intestinal metaplasia (IM). Each investigator was required to score at least 90% correct diagnoses of the unknown cases before participation in the study. Moreover, each study investigator was required to have experience with at least 10 pCLE cases in the setting of BE endoscopic treatments.
Study population Patients aged 18 years or older who had been previously treated for BE by mucosal ablation therapy and referred for upper endoscopy (EGD) (for the 2-4 months follow-up) were enrolled after providing written informed consent. To be included, patients must have had histologically confirmed BE with IM, low-grade neoplasia (LGN)/ HGN as the original indication for ablative treatment, had to be undergoing any type of endoscopic BE ablation treatment including radiofrequency ablation (RFA), cryotherapy, photodynamic therapy, or EMR, or combinations of these, and must have had ⬍2 cm of circumferential BE and ⬍5 total islands of BE on the most recent ablation. The latter requirement meant that patients were enrolled patients at the time when their endoscopies were expected or near complete ablation of IM/neoplasia. All patients were maintained on at least daily proton pump inhibitor therapy. Patients were excluded if they had a contraindication to biopsies, such as anticoagulation or varices, had allergy to fluorescein, or were pregnant.
Randomization After we obtained the patient’s informed consent, the patients were randomized via a secure Web-based system to one of the following groups: Group 1 : HDWL only Group 2 : HDWL ⫹ pCLE 540 GASTROINTESTINAL ENDOSCOPY Volume 76, No. 3 : 2012
Blinding of endoscopists to randomization group was not feasible in this setting; however, assignment of an endoscopist to a given patient’s procedure was determined independently of the group assignment.
Endoscopic procedures EGD was performed with HDWL (Olympus 180 series; Olympus, Center Valley, PA) only (no red flag techniques such as narrow-band imaging were allowed in the study), on patients who already had previous treatment of a BE lesion. Patients typically underwent ablation every 2 to 4 months until all suspicious disease was confirmed by biopsy to be ablated. The esophagus was defined as the portion of tubular esophagus above the top most extent of gastric folds. The gastroesophageal junction was defined functionally as the 2 cm of tissue immediately below the top of the gastric folds.
Imaging and presumptive diagnosis with HDWL Standard high-definition endoscopy was performed on all patients. The esophagus was examined for any suspicious lesion identified as salmon-colored mucosa with or without a villiform or irregular mucosal pattern consistent with BE metaplasia or neoplasia. Based on the prespecified criteria, each site was diagnosed presumptively as squamous mucosa, gastric metaplasia (GM), IM, or any grade of neoplasia. Within the tubular esophagus, the criteria for Barrett’s esophagus included any columnar epithelium with or without the presence of goblet cells.4 Because of the inherent challenges of precisely defining the exact point of the gastroesophageal junction, we also included as “positive” lesions of IM or neoplasia within 2 cm below the presumed top of the gastric folds. Neosquamous epithelium was classified when there was mucosa with a completely flat mucosal pattern with no pits. GM was classified when there was a circular mucosal pit pattern in the setting of columnar epithelium within the tubular esophagus, but only GM found above the gastroesophageal junction was considered positive.
Imaging and presumptive diagnosis with pCLE For patients assigned to the HDWL ⫹ pCLE group, pCLE was initiated immediately after we recorded a prewww.giejournal.org
Wallace et al
GI neoplasia in Barrett’s esophagus
Figure 1. Flow diagram of patient evaluations and outcomes. Outcomes considered not-optimal are shown in orange boxes.
sumptive diagnosis at each site based on HDWL findings. Sodium fluorescein was injected intravenously 2.5 mL at 10% (Akorn Pharmaceuticals, Lake Forest, Ill), and the pCLE procedure was performed. The endoscopist took at least one image from each quadrant segment that had residual salmon-colored mucosa. All pCLE videos were acquired with a cap on the tip of the endoscope (clear 4-mm mucosectomy cap). The criteria for classification of neoplasia have been previously validated in our multicenter clinical trial.5 A presumptive diagnosis was recorded at the time of the procedure based on pCLE findings for all lesions that had been identified by HDWL. The final presumptive diagnosis for any given lesion was considered to be the lesser of the diagnoses recorded by using HDWL and pCLE. This strategy was anticipated to reduce the likelihood of overtreatment and ultimately result in a higher overall proportion of optimally treated patients.
Patient-level presumptive diagnosis The patient-level presumptive diagnosis was the most extreme diagnosis across all lesions for a patient based on HDWL or HDWL ⫹ pCLE. The effect was that the final presumptive diagnosis with HDWL ⫹ pCLE was the same or lower than would have been obtained with HDWL alone.
Ablation The decision to re-treat or not was made by the endoscopist by using the overall presumptive diagnosis. The intent was to treat a patient if the overall presumptive diagnosis was IM, neoplasia, or GM above the gastroesophageal junction. For all lesions for which the decision was to re-treat, the endoscopist re-ablated the mucosa by www.giejournal.org
using EMR, RFA, or cryotherapy immediately after biopsy during the same endoscopy session.
Biopsy and pathology interpretation In all cases, the endoscopist performed mucosal biopsies in order to confirm the presumed endoscopic diagnosis made by using HDWL or pCLE. In case of EMR, the mucosal specimen was considered the biopsy specimen. Each sample was interpreted by the local GI study pathologist according to standard clinical protocols. All sites with any degree of neoplasia were confirmed by a second expert pathologist. Each site was classified as squamous epithelium, gastric metaplasia, IM without neoplasia, indefinite for neoplasia, LGN, HGN, or invasive carcinoma, according to published standards.4
Three-month follow-up endoscopy All patients were scheduled for a follow-up endoscopic procedure 2 to 4 months after the initial study procedure. Biopsies, using the same algorithm (1 biopsy/segment), were taken during this second procedure in order to confirm the presence of remaining BE and to assess the appropriateness and completeness of the re-treatment (if applicable).
Study flow and definition of primary outcome The study flow and outcome definitions are shown in Figure 1. The primary outcome of optimal treatment was a composite defined for each patient by the absence of all non-optimal individual outcomes indicated in red in the figure and defined as (1) overtreatment—ablation or resection when the final pathology shows no GM in the tubular esophagus, IM, or neoplasia; (2) undertreatment— non-ablation nor resection when the final pathology shows GM in the tubular esophagus, IM, or neoplasia; (3) Volume 76, No. 3 : 2012 GASTROINTESTINAL ENDOSCOPY 541
GI neoplasia in Barrett’s esophagus
incomplete treatment—the presence of IM or neoplasia on any biopsy from a patient at 3-month re-evaluation.
Wallace et al
TABLE 1. Demographic characteristics of 164 randomized participants
Statistical considerations The initially planned sample size was 270 (135 in each group) based on a requirement for the study to have 80% power at the 5% significance level to detect an improvement in the proportion of optimally treated patients with HDWL ⫹ pCLE versus HDWL alone of 15%, assuming a proportion of 70% with HDWL alone. These values were chosen as assumptions for power calculations based on previously published outcomes6 and expert opinion of the study investigators as to the degree of improvement that was plausible. An interim analysis was planned for when 80 patients had completed follow-up visits. The goal of this interim analysis was to re-assess sample size assumptions as well as to assess futility, with a plan to consider stopping the study if conditional power to detect a difference in proportions of 15% was estimated to be less than 20%. Conditional power was assessed by simulation and was defined as the probability of obtaining P ⬍ .05 at the end of the study, taking into account that the data already obtained were fixed and updating assumptions regarding the overall proportion of patients with optimal outcome and of the drop-out rate.
RESULTS Interim analysis The study was halted early because of low conditional power at the planned interim analysis. This low power associated with the originally planned sample size was related to a combination of (1) a higher-than-expected dropout rate, (2) an overall proportion of patients with optimal treatment of around 30% rather than the expected ⱖ70%, and (3) no difference between groups among the first 80 patients to have their follow-up visits. Despite the decision to halt recruitment to the study, at that time 166 patients had been recruited, and they continued until completion of their 3-month follow-up visits.
Patient characteristics, withdrawals, and loss to follow-up A total of 166 patients consented to be in the study across the 6 sites. Two patients withdrew from the study before randomization, and thus 164 patients were randomized, 82 each to the two study arms. Of the 164, 23 were withdrawn before, during, or after the procedure and/or did not have biopsy information. This left 141 patients who completed the initial procedure requirements (68 HDWL only, 73 HDWL ⫹ pCLE). A total of 119 patients completed their 3-month follow-up visits (57 HDWL only, 62 HDWL ⫹ pCLE). Exploratory analysis of the data ac542 GASTROINTESTINAL ENDOSCOPY Volume 76, No. 3 : 2012
Variable Age at randomization*, y
HDWL only (n ⴝ 82)
HDWL ⴙ pCLE (n ⴝ 82)
67.9 (45.4, 87.2) 69.6 (50.2, 86.1)
Sex, male, no. (%)
78 (95)
75 (91)
Intestinal metaplasia
5 (6)
2 (2)
Low-grade neoplasia
3 (4)
9 (11)
High-grade neoplasia
61 (75)
56 (68)
Cancer
11 (14)
15 (18)
Other
1 (1)
0 (0)
Original indication for treatment†, no. (%)
HDWL, High-definition white light; pCLE, probed-based confocal laser endomicroscopy. *Age is reported with median (minimum, maximum). †Original indication was missing for one HDWL participant.
cording to the intent-to-treat principle resulted in no suggestion of systematic bias in the two groups regarding withdrawals or loss to follow-up (data not shown). Only 1 patient was withdrawn because of intolerance to fluorescein. The two groups were comparable with respect to age, sex, and original treatment indication (Table 1).
Presumptive diagnoses Among the 73 patients in the HDWL ⫹ pCLE arm, presumptive diagnoses at the lesion level frequently differed for pCLE compared with HDWL, with disagreement for 88 (35%) of the 254 lesions (Table 2). However, this translated into a downgraded patient-level diagnosis for HDWL ⫹ pCLE compared with that based on HDWL alone in only 8 (11%) of the 73 patients (Table 3). Furthermore, only 2 patients (3%) had a revised presumptive diagnosis with pCLE that affected the treatment decision; one was IM, and the other was neoplasia/cancer, with HDWL, and both were downgraded to a negative diagnosis of GM below the gastroesophageal junction. Both patients had negative biopsy results with respect to IM or neoplasia; 1 had a follow-up visit and remained negative for BE. Given these findings in the HDWL ⫹ pCLE arm, it is no surprise that the patient-level presumptive diagnoses were similarly distributed in the two study arms, with a positive result requiring treatment in 91% (62/68) in the HDWL arm and 86% (63/73) in the HDWL ⫹ pCLE arm (Table 3). The actual proportions of patients treated in the two arms were 90% (61/68) and 86% (63/73). The slight difference was caused by confusion among investigators early in the www.giejournal.org
Wallace et al
GI neoplasia in Barrett’s esophagus
TABLE 2. Presumptive diagnoses for pCLE vs HDWL in those in the HDWL ⴙ pCLE arm, along with biopsy result on a per-lesion basis (254 lesions from 73 patients) pCLE presumptive diagnosis HDWL presumptive diagnosis, no. (%) Negative
GM above GEJ
Intestinal metaplasia
Neoplasia/cancer
Overall
Negative
GM above GEJ
IM
N/C
Overall
31 (12)
0 (0)
28 (11)
1* (⬍1)
60 (24)
(27, 0, 3, 1)
(0, 0, 0, 0)
(24, 0, 4, 0)
(0, 0, 0, 1)
(51, 0, 7, 2)
0 (0)
9 (4)
6 (2)
0 (0)
15 (6)
(0, 0, 0, 0)†
(6, 0, 3, 0)
(0, 2, 4, 0)
(0, 0, 0, 0)
(6, 2, 7, 0)
10 (4)
6 (2)
108 (43)
27 (11)
151 (59)
(7, 0, 0, 3)
(0, 4, 2, 0)
(46, 21, 34, 7)
(10, 7, 5, 5)
(63, 32, 41, 15)
2 (1)
1 (⬍1)
7 (3)
18 (7)
28 (11)
(1, 0, 1, 0)
(0, 1, 0, 0)
(2, 1, 3, 1)
(6, 2, 2, 8)
(9, 4, 6, 9)
43 (17)
16 (6)
149 (59)
46 (18)
254 (100)
(35, 0, 4, 4)
(6, 5, 5, 0)
(72, 24, 45, 8)
(16, 9, 7, 14)
(129, 38, 61, 26)
pCLE, Probed-based confocal laser endomicroscopy; HDWL, high-definition white light; GM, gastric metaplasia; GEJ, gastroesophageal junction; IM, intestinal metaplasia; N/C, neoplasia/cancer. *This lesion had a histologic diagnosis of cancer; the negative result for HDWL actually refers to a missing diagnosis for this lesion; the corresponding patient had 4 lesions and an overall presumptive diagnosis of neoplasia/cancer. †Numbers in parentheses represent histology results (negative, GM above GEJ, IM, neoplasia/cancer). Both HDWL and pCLE agreed with histology for 69 (27%) lesions. HDWL and pCLE agreed, but the result was different from histology for 97 (38%) lesions. pCLE agreed with biopsy but not with HDWL for 30 (12%) lesions. HDWL agreed with biopsy but not with pCLE for 34 (13%) lesions. None agreed for 24 (9%) lesions.
study as to how to treat GM depending on its location: 6 patients with a presumptive diagnosis of GM above the gastroesophageal junction were inadvertently not treated, whereas 5 patients with GM at/below the gastroesophageal junction were treated when they should not have been per protocol (Table 3).
Biopsy results Diagnoses from pathologic interpretation of biopsy specimens were similar in the two groups; this was expected because this was not dependent on the randomization assignment. Agreement between presumptive diagnoses and biopsy results was similar in the two groups (Table 4), at 48% (97/203, HDWL only) and 46% (116/254, HDWL ⫹ pCLE) by using the 4-level diagnosis category, or 71% (144/203) and 67% (171/254), respectively, by using a positive/negative diagnosis. Agreement between HDWL and pCLE with biopsy could be explored in the same patients in the HDWL ⫹ pCLE arm (Table 2). There was agreement of the positive/negative diagnosis between all 3 for 27% (69/254) of lesions, agreement between HDWL and biopsy of 41% (103/254), and agreement between pCLE and biopsy of 39% (99/254). Interestingly, the HDWL ⫹ pCLE diagnosis (defined as the lesser of HDWL and pCLE diagnoses) had the highest agreement with biopsy at 44% (116/254), although they were still very similar. At the patient level, there was greater correspondence between presumptive diagnosis and pathology, with agreement of www.giejournal.org
the positive/negative diagnosis with biopsy in 72% (49/68, HDWL only) and 77% (56/73, HDWL ⫹ pCLE) of patients. If treatment had reflected the protocol guidelines in all cases, these also would represent the proportions of appropriately treated patients (not overtreated or undertreated); however, as noted previously, 11 patients were not treated per protocol guidelines. The actual proportions of appropriately treated patients were 71% (48/68, HDWL only) and 66% (48/73, HDWL ⫹ pCLE), again quite similar between groups.
Three-month follow-up and primary outcomes Among the 141 patients who completed the initial procedure, 119 had follow-up visits. There were no obvious systematic differences between patients who did and did not have follow-up visits; thus the primary results focus on those patients with follow-ups. The overall primary outcome of optimal treatment was achieved for 26% (15/57) of patients in the HDWL arm and for 27% (17/62) in the HDWL ⫹ pCLE arm (Table 5). From an intent-to-treat standpoint, among all 164 patients randomized, optimal treatment was achieved for 18% (15/82) in the HDWL arm and 21% (17/82) in the HDWL ⫹ pCLE arm. Taking account of the 11 patients not being treated per protocol guidelines, supplementary Table S1 (available at www. gieonline.org) shows the range of results that could have been obtained had this not happened. This table also shows the projected results had pCLE alone, or the greater of the HDWL Volume 76, No. 3 : 2012 GASTROINTESTINAL ENDOSCOPY 543
GI neoplasia in Barrett’s esophagus
Wallace et al
HDWL only (n ⴝ 68)
HDWL ⴙ pCLE (n ⴝ 73)
Negative, no. (%)
6† (9)
10 (14)*
Gastric metaplasia above GEJ, no. (%)
3 (4)
3 (4)
Intestinal metaplasia, no. (%)
39 (57)
46 (63)*
Neoplasia/cancer, no. (%)
20 (29)
14 (19)*
Patient treated‡, no. (%)
61 (90)‡
63 (86)‡
Negative
21 (31)
21 (29)
HDWL arm and 35% (22/62) in the HDWL ⫹ pCLE arm (Table 5). Another common reason for not-optimal outcome was overtreatment (25%, 17/25 in HDWL and 25%, 18/73 in HDWL ⫹ pCLE). Interestingly, 7% of patients were treated because of a positive finding on imaging but had a negative initial biopsy (thus were classified per study protocol as overtreatment) but still had residual BE at follow-up. This occurred in 4% (2/57) of HDWL and 10% (6/62) of HDWL ⫹ pCLE patients and likely represents a false negative biopsy during the initial examination, but in fact was a correct endoscopic diagnosis along with correct, but ineffective, treatment. Among the 10 patients (across both arms) who had negative results by both presumptive diagnosis and by initial biopsy, and who also had follow-up, none had positive results at follow-up. Because of concern that we included the very challenging area of the gastroesophageal junction, we separately examined lesions that were above this area; however, results were not qualitatively different (data not shown).
Gastric metaplasia above GEJ
11 (16)
9 (12)
DISCUSSION
Intestinal metaplasia
21 (31)
24 (33)
Neoplasia/cancer
15 (22)
19 (26)
Appropriateness of treatment§, no. (%)
48 (71)
48 (66)
Treated (appropriately)
44 (65)
45 (62)
Not treated (appropriately)
4 (6)
3 (4)
Overtreated
17 (25)
18 (25)
Undertreated
3 (4)
7 (10)
This study did not provide evidence that the combination of HDWL and pCLE is superior to HDWL alone to assess the completeness of ablation of Barrett’s esophagus. There was a slightly higher proportion of patients with residual dysplasia in the HDWL arm and a slightly higher proportion with metaplasia in the HDWL ⫹ pCLE arm; however, with the small numbers these differences were not of statistical significance, and the clinical significance is unknown. Endoscopic resection and ablation of BE, primarily those with HGN and early (T1 mucosal) cancer, is increasingly becoming the standard of care. In a landmark study Shaheen et al6 showed that RFA was superior to sham for ablation of neoplasia in BE. Complete ablation typically required 3 to 4 sessions, with biopsies performed either concurrent with each ablation or in a separate procedure. Some groups have applied RFA to ablation BE before development of intraepithelial neoplasia to eradicate the preneoplastic state, and studies suggest that this is a cost-effectiveness strategy.7 Other methods of ablation include cryotherapy with liquid nitrogen or CO2, photodynamic therapy, or coagulation with argon plasma or multipolar electrocautery. Contrary to other ablative techniques, EMR provides a tissue specimen that is much larger than mucosal pinch biopsy specimens and can be evaluated for staging and histology. Several studies have demonstrated that EMR is a safe and effective alternative to surgery for complete resection of superficial lesions, with the advantage of histopathologic verification.8-11 A challenge of all ablation and resection methods is the ability to determine when treatment is complete. Options include scheduling a separate diagnostic endoscopy with biopsy and treating only those with histologically confirmed disease or performing empiric ablation until all
TABLE 3. Presumptive diagnosis, biopsy, and appropriateness of treatment of 141 patients with completed initial procedure and biopsy result
Final presumptive diagnosis*
Biopsy result, no. (%)
HDWL, High-definition white light; pCLE, probed-based confocal laser endomicroscopy; GEJ, gastroesophageal junction. *Eight patients in the HDWL ⫹ pCLE arm had their presumptive diagnoses by HDWL downgraded with addition of pCLE: 6 were downgraded from neoplasia/cancer to intestinal metaplasia; 1 intestinal metaplasia and 1 neoplasia/cancer were downgraded to a negative result (gastric metaplasia below the GEJ). †Includes 1 patient with no data regarding presumptive diagnosis. ‡Eleven patients were not treated according to protocol guidelines: 6 patients (2 HDWL, 4 HDWL ⫹ pCLE) had presumptive diagnoses of gastric metaplasia above the GEJ but were not treated; 5 patients (1 HDWL, 4 HDWL ⫹ pCLE) had negative presumptive diagnoses (gastric metaplasia below the GEJ) yet were treated. §Appropriateness of treatment refers to correspondence of actual treatment with biopsy result.
and pCLE presumptive diagnoses, been used to guide treatment. There was no suggestion that other strategies would have improved results in the HDWL ⫹ pCLE arm. The most common reason for a patient to have a notoptimal treatment was, in fact, the failure to completely eradicate Barrett’s esophagus despite a correct imaging diagnosis, which occurred in 44% (25/57) of patients in the 544 GASTROINTESTINAL ENDOSCOPY Volume 76, No. 3 : 2012
www.giejournal.org
Wallace et al
GI neoplasia in Barrett’s esophagus
TABLE 4. Lesion-specific presumptive diagnoses for HDWL ⴙ pCLE vs HDWL, with primary biopsy result (n ⴝ 457 lesions from 141 patients) HDWL ⴙ pCLE, biopsy result (254 lesions from 73 patients)
HDWL only, biopsy result (203 lesions from 68 patients) Overall presumptive diagnosis, no. (%)
Negative*
GMA
IM
N/C
Negative*
GMA
IM
N/C
Negative*
27 (13)
1 (⬍1)
3 (1)
0 (0)
59 (23)
0 (0)
8 (3)
5† (2)
GMA
1 (⬍1)
14 (7)
2 (1)
1 (⬍1)
6 (2)
7 (3)
9 (4)
0 (0)
IM
49 (24)
23 (11)
42 (21)
11 (5)
58 (23)
29 (11)
42 (17)
13 (5)
5 (2)
4 (2)
6 (3)
14 (7)
6 (2)
2 (1)
2 (1)
8 (3)
Neoplasia/cancer
HDWL, High-definition white light; pCLE, probed-based confocal laser endomicroscopy; GMA, finding of gastric metaplasia (GM) above the gastroesophageal junction (GEJ); IM, intestinal metaplasia; N/C, neoplasia/cancer. *Negative means the absence of IM, neoplasia, and cancer. Gastric metaplasia (GM) below the GEJ is also considered negative. †One of these 5 lesions had cancer as the biopsy result (diagnosis with HDWL was no data; with pCLE it was neoplasia/cancer). One had GM below the GEJ with HDWL and pCLE and a biopsy diagnosis of low-grade neoplasia (LGN). The remaining 3 (2 from the same patient) had IM with HDWL, GM below the GEJ with pCLE, and LGN with biopsy.
visible disease is treated, then doing a final diagnosis endoscopy. Both options have major shortcomings, with the former requiring extra, costly procedures, and the latter resulting in some overtreatment when the endoscopic assessment is falsely positive. Confocal laser endomicroscopy is a real-time imaging method that may allow precise, in vivo diagnosis of BE and associate neoplasia and, thus, allow an on-table decision to treat or not treat. The technology can be integrated into the endoscope (Pentax Systems, Tokyo, Japan) or into a probebased device through the working channel of the endoscope (Cellvizio; Mauna Kea Technologies, Paris, France).5,12-14 In a feasibility study conducted by Meining et al15 with the pCLE system to detect malignant and premalignant modifications in the GI tract, the investigators were able to demonstrate that this method gives 92% accuracy in the detection of a neoplasia as compared with conventional histopathology. Likewise, after the injection of fluorescein, Kiesslich et al16 examined patients with Barrett’s esophagus by means of an integrated endoscope– based CLE system. The authors were able to achieve a surprisingly high accuracy with this method both in the diagnosis of Barrett’s metaplasia and Barrett’s neoplasia (96.8% and 97.4%, respectively). More recently, Pohl et al12 demonstrated in a multicenter study that pCLE has a high negative predictive value: 98.8% for the diagnosis of endoscopically invisible neoplasia in BE. Finally, the randomized clinical trial that used pCLE in BE demonstrated pCLE to be superior to HDWL endoscopy for detection of BE HGN and allowed a significant reduction in the number of random biopsies.5 The reasons pCLE failed to improve outcomes in our study are unclear. First, the accuracy for assessment of residual BE was much lower in our trial compared with other published literature, including our own.5 Compared with the run-in study by the same investigators, image interpretation in the actual trial was performed in real time, on unselected www.giejournal.org
images, and in the post-ablation setting. Any one of these factors may have resulted in lower accuracy, although individual studies by our group have not shown significant differences in real-time compared to offline interpretation in unselected cases of neoplastic colorectal polyps.17 In a previous, single-center study, we demonstrated accuracy for the interpretation of pCLE images after ablation/resection in colorectal neoplasia, such as after EMR of large colon polyps.18 Another potential factor may be prior knowledge of the state of neoplasia. In the randomized, controlled trial by Sharma et al,5 blinding of the endoscopists was limited because the endoscopists were likely to be aware of the indication for referral. In the current study, although prior neoplasia status was known, all patients had been treated, and thus current status was partially unknown. The ability to diagnose IM or neoplasia in the post-ablation setting is likely to be more challenging because of ongoing inflammation and scar tissue. Finally, the role of pCLE in this study was, at the cardia, to distinguish intestinal metaplasia, with or without neoplasia, which is likely to be more challenging than the distinction between neoplastic and non-neoplastic Barrett’s epithelium. Another limitation of our study was the unexpectedly high proportion of patients who had incomplete treatment despite a correct endoscopic diagnosis. In planning the study, we overestimated the proportion of patients who would have completed ablation at the time of the procedure and thus not require subsequent treatment, although this was based on well-controlled, published trials.6 Our protocol differed, however, and required acquisition of tissue specimens after imaging but before treatment. Endoscopic pinch biopsies are associated with mucosal disruption and bleeding that may have decreased the efficacy of subsequent ablation. In conclusion, we did not find any evidence that pCLE improved diagnostic accuracy nor clinical outcomes in Volume 76, No. 3 : 2012 GASTROINTESTINAL ENDOSCOPY 545
GI neoplasia in Barrett’s esophagus
Wallace et al
ACKNOWLEDGMENT TABLE 5. Final outcomes at follow-up and overall (n ⴝ 119) HDWL only (n ⴝ 57)*
HDWL ⴙ pCLE (n ⴝ 62)*
No. (%)
No. (%)
Follow-up finding
REFERENCES
Absence of BE (negative)
28 (49)
32 (52)
BE metaplasia
25 (44)
22 (35)
BE neoplasia
3 (5)
7 (11)
Cancer
1 (2)
1 (2)
Treated (appropriately); negative
11 (19)
15 (24)
Treated (appropriately); positive
25 (44)
22 (35)
Not treated (appropriately); negative
4 (7)
2 (3)
Not treated (appropriately); positive
0 (0)
0 (0)
Overtreated; negative
12 (21)
10 (16)
Overtreated; positive
2 (4)
6 (10)
Undertreated; negative
1 (2)
5 (8)
Undertreated; positive
2 (4)
2 (3)
15 (26)
17 (27)
Final overall study result
Optimal treatment†
The authors wish to thank Kelly Viola and Victoria Jackson (Academic and Research Support, Mayo Clinic Florida) for editorial assistance.
HDWL, High-definition white light; pCLE, probed-based confocal laser endomicroscopy; BE, Barrett’s esophagus. *A total of 22 patients did not have follow-up visits (11 in each arm). †Optimal treatment means that a patient was not overtreated or undertreated with reference to the biopsy result at the initial procedure and had an absence of Barrett’s esophagus at follow-up.
patients undergoing ablation or resection of BE. Questions for future study include whether it may be useful at earlier or later stages of ablation or if confined to differentiating neoplasia and metaplasia only within the tubular esophagus, well above the gastroesophageal junction. At the current time, it appears that high-resolution white light endoscopy likely is adequate to identify residual BE (with or without neoplasia) in tubular esophagus, but further studies may need to be done to identify whether multimodal imaging may be helpful in specific challenging areas such as the gastroesophageal junction or specifically to identify neoplasia in the esophagus for appropriate treatment (not powered here). 546 GASTROINTESTINAL ENDOSCOPY Volume 76, No. 3 : 2012
1. Surveillance epidemiology and end results: Faststats. Volume 2011, 2011. Available at http://seer.cancer.gov/fastats. Accessed January 1, 2012. 2. Shaheen NJ, Frantz DJ. When to consider endoscopic ablation therapy for Barrett’s esophagus. Curr Opin Gastroenterol 2010;26:361-6. 3. Spechler SJ, Davila R. Endoscopic therapy in Barrett’s esophagus: when and how? Surg Oncol Clin N Am 2009;18:509-21. 4. Riddell RH, Odze RD. Definition of Barrett’s esophagus: time for a rethink—is intestinal metaplasia dead? Am J Gastroenterol 2009;104: 2588-94. 5. Sharma P, Meining AR, Coron E, et al. Real-time increased detection of neoplastic tissue in Barrett’s esophagus with probe-based confocal laser endomicroscopy: final results of an international multicenter, prospective, randomized, controlled trial. Gastrointest Endosc 2011;74: 465-72. 6. Shaheen N, Sharma P, Overholt B, et al. Radiofrequency ablation in Barrett’s esophagus with dysplasia. N Eng J Med 2009;360:2277-88. 7. Inadomi JM, Somsouk M, Madanick RD, et al. A cost-utility analysis of ablative therapy for Barrett’s esophagus. Gastroenterology 2009;136: 2101-14 e1-6. 8. Prasad GA, Wu TT, Wigle DA, et al. Endoscopic and surgical treatment of mucosal (T1a) esophageal adenocarcinoma in Barrett’s esophagus. Gastroenterology 2009;137:815-23. 9. Pouw RE, Wirths K, Eisendrath P, et al. Efficacy of radiofrequency ablation combined with endoscopic resection for Barrett’s esophagus with early neoplasia. Clin Gastroenterol Hepatol 2010;8:23-9. 10. Konda Vani JA, Chennat J, Waxman I. New directions in endoscopic therapy of Barrett’ s esophagus. Minerva Gastroenterol Dietol 2010;56:42135. 11. Prasad GA, Wang KK, Buttar NS, et al. Long-term survival following endoscopic and surgical treatment of high-grade dysplasia in Barrett’s esophagus. Gastroenterology 2007;132:1226-33. 12. Pohl H, Rösch T, Vieth M, et al. Miniprobe confocal laser microscopy for the detection of invisible neoplasia in patients with Barrett’s esophagus. Gut 2008;57:1648-53. 13. Bajbouj M, Vieth M, Rösch T, et al. Probe-based confocal laser endomicroscopy compared with standard four-quadrant biopsy for evaluation of neoplasia in Barrett’s esophagus. Endoscopy 2010;42: 435-40. 14. Dunbar KB, Canto MI. Confocal laser endomicroscopy in Barrett’s esophagus and endoscopically inapparent Barrett’s neoplasia: a prospective, randomized, double-blind, controlled, crossover trial. Gastrointest Endosc 2010;72:668. 15. Meining A, Saur D, Bajbouj M, et al. In vivo histopathology for detection of gastrointestinal neoplasia with a portable, confocal miniprobe: an examiner blinded analysis. Clin Gastroenterol Hepatol 2007;5:1261-7. 16. Kiesslich R, Gossner L, Goetz M, et al. In vivo histology of Barrett’s esophagus and associated neoplasia by confocal laser endomicroscopy. Clin Gastroenterol Hepatol 2006;4:979-87. 17. Shahid MW, Buchner AM, De Melo SW, et al. S1580: Comparison of real time versus offline-blinded accuracy of confocal laser endomicroscopy (pCLE) for diagnosis of neoplasia on colorectal polyps [abstract]. Gastrointest Endosc 2010;71:AB199.
www.giejournal.org
Wallace et al
GI neoplasia in Barrett’s esophagus
18. Shahid MW, Buchner AM, Coron E, et al. Diagnostic accuracy of probebased confocal laser endomicroscopy in detecting residual colorectal neoplasia after EMR: a prospective study. Gastrointest Endosc 2012;75:525-33.
Abbreviations: BE, Barrett’s esophagus; GM, gastric metaplasia; HDWL, high-definition white light; HGD, high-grade dysplasia; HGN, high-grade neoplasia; IM, intestinal metaplasia; LGD, low-grade dysplasia; LGN, low-grade neoplasia; pCLE, probed-based confocal laser endomicroscopy; RFA, radiofrequency ablation. DISCLOSURE: Mauna Kea Technologies provided funding for this study to M. Wallace as a grant, I. Waxman and V. Konda as honoraria, and P. Sharma as research support. I. Waxman also received consulting fees from Olympus, and P. Sharma received research support from Olympus. No other financial relationships relevant to this publication were disclosed.
Copyright © 2012 by the American Society for Gastrointestinal Endoscopy 0016-5107/$36.00 http://dx.doi.org/10.1016/j.gie.2012.05.004 Received January 9, 2012. Accepted May 2, 2012. Current affiliations: Mayo Clinic (1), Jacksonville, Florida; University of Washington (2), Seattle, Washington, USA; Nantes University Hospital (CHU) (3) Nantes, Nantes, France; University College London (4), London, United Kingdom; University of Chicago (5), Chicago, Illinois; University of Kansas (6), Lawrence, Kansas, USA. Reprint requests: Michael B. Wallace, MD, MPH, 4500 San Pablo Road, Jacksonville, FL 32224. If you would like to chat with an author of this article, you may contact Dr Wallace at
[email protected].
GIE on Facebook GIE now has a Facebook page. Fans will receive news, updates, and links to author interviews, podcasts, articles, and tables of contents. Search on Facebook for “GIE: Gastrointestinal Endoscopy” and become a fan.
www.giejournal.org
Volume 76, No. 3 : 2012 GASTROINTESTINAL ENDOSCOPY 547
GI neoplasia in Barrett’s esophagus
Wallace et al
Supplementary Table S1. Results assuming treatment had been done as per protocol guidelines with various scenarios for use of HDWL and pCLE lesion specific findings to obtain the patient-level presumptive diagnosis in HDWL ⴙ pCLE arm HDWL ⴙ pCLEa (n ⴝ 73) HDWL only (n ⴝ 68)
HDWL only
HDWL ⴙ pCLE mina
pCLE only
HDWL ⴙ pCLE maxa
Positive presumptive diagnosis/treateda
62 (91%)
65 (89%)
63 (86%)
69 (95%)
71 (97%)
Appropriateness of treatmenta
49 (72%)
54 (74%)
56 (77%)
54 (74%)
52 (71%)
45 (66%)
49 (67%)
49 (67%)
51 (70%)
51 (70%)
4 (6%)
5 (7%)
7 (10%)
3 (4%)
1 (1%)
17 (25%)
16 (22%)
14 (19%)
18 (25%)
20 (27%)
2 (3%)
3 (4%)
3 (4%)
1 (1%)
1 (1%)
(n ⫽ 57)
(n ⫽ 62)
(n ⫽ 62)
(n ⫽ 62)
(n ⫽ 62)
27-30 (47-53%)
28-34 (45-55%)
28-34 (45-55%)
31-34 (50-55%)
31-34 (50-55%)
14-16 (25-28%)
18-24 (29-39%)
19-25 (31-40%)
20-23 (32-37%)
19-22 (31-35%)
10-12 (18-21%)
17-19 (27-31%)
17-19 (27-31%)
19-21 (31-34%)
19-21 (31-34%)
4 (7%)
1-5 (2-8%)
2-6 (4-10%)
22-24 (35-39%)
0-1 (0-2%)
- Treated (appropriately) - Not treated (appropriately) - Over-treated - Under-treated b
Follow-up finding
- Absence of BE (‘negative’) Optimal treatment overall - Treated (appropriately); negative - Not treated (appropriately); negative
aThe assumption was that a patient with a positive presumptive diagnosis would be treated; in the trial, 11 patients were not treated per protocol. For the
HDWL ⫹ pCLE arm, as well as exploring results using the lesser of the HDWL and pCLE diagnoses (HDWL ⫹ pCLE min) for each lesion as the final diagnosis, we also show results had other strategies been used for comparison: use of HDWL alone, use of pCLE alone, and use of the greater of the HDWL and pCLE diagnoses (HDWL ⫹ pCLE max). bRanges arise because follow-up results may have been different from what was observed for some of the cases where actual treatment did not match up with the presumptive diagnosis. E.g. a patient who had negative presumptive diagnosis, yet who was actually treated and had a negative follow-up result, could have had a positive follow-up if they had not been treated as per the presumptive diagnosis.
547.e1 GASTROINTESTINAL ENDOSCOPY Volume 76, No. 3 : 2012
www.giejournal.org