Introduction
Esophageal cancer is the eighth most common cancer and the sixth leading cause of
cancer-related mortality in the world [1]. Squamous cell carcinoma (SCC) is the most frequent histological subtype accounting
for 90 % of cases of esophageal cancer worldwide due to its high prevalence in Eastern
Asia [1]
[2]
[3].
Detection and characterization of early stage esophageal SCC can be challenging, since
it typically appears as subtle and flat lesions on conventional white light endoscopy
[1]
[4]. Lugol dye chromoendoscopy (LCE) has long been the gold standard for detection of
superficial SCC and it is based on the lack of absorption of the iodine stain by abnormal
squamous tissue [3]
[5]
[6]. However, lugol staining can cause heartburn and chest discomfort, has risk of pulmonary
aspiration and allergic reaction, and increases the duration of the procedure [3]
[6]. Moreover, LCE is a highly sensitive but not very specific technique for detecting
superficial esophageal SCC since both inflamed and dysplastic/neoplastic lesions can
appear as iodine-unstained areas [4]. Therefore, new technologies have been developed to evaluate the esophageal mucosa
of high-risk patients, such as narrow-band imaging (NBI, Olympus) [7]
[8]. NBI is a virtual chromoendoscopy technique that enhances the mucosa surface and
the underlying capillary pattern simply by pressing a button [9]. Early squamous cell lesions can be identified by this technology as brownish, well
demarcated lesions [3]. Disadvantages of this technique include the high cost of the device, which decreases
accessibility, and its’ application requires expertise [3].
Several studies compared the accuracy of NBI and LCE for diagnosing esophageal SCC
[4]
[10]
[11]. A recent meta-analysis showed that NBI has a comparable sensitivity to that of
lugol chromoendoscopy (88 % vs. 92 %), and a superior specificity (88 % vs. 82 % P < 0.001) [12]. However, to our knowledge, there are no data on the effectiveness of NBI in delineating
esophageal SCC margin before endoscopic resection.
Therefore, the aim of this study was to compare the effectiveness of NBI and LCE in
defining resection margins of esophageal squamous cell cancer and dysplasia. Our hypothesis
was that NBI may be equivalent to LCE in determining esophageal SCC limits prior to
endoscopic resection.
Patients and methods
We conducted a retrospective observational cohort study of all patients with esophageal
SCC and dysplasia who underwent en-bloc resection between 1999 and 2017 at the Cliniques
Universitaires Saint-Luc, Brussels, Belgium. The study protocol was approved by our
ethical review board in April 2014 (2014/30AVR/210). The study protocol conforms to
the ethical guidelines of the 1975 Declaration of Helsinki as reflected in the approval
by the Comité d'Ethique Hospitalo-Facultaire Université Catholique de Louvain.
Two groups were defined before endoscopic resection: 1) inspection of esophageal lesions
exclusively with NBI; and 2) inspection with lugol chromoendoscopy (with or without
NBI) ([Fig. 1]). The decision to perform each technique was based on availability and endoscopist
choice. In the first group, abnormal mucosal areas were identified and resection margins
defined by NBI. When NBI was used, brown-stained areas of the mucosa were considered
lesions suspected to be neoplasia (compared to “normal” mucosa, which is green independently
of changes in surface or vascular texture). Delineation was mainly based on differences
between green and brown mucosa, but included analysis of the intrapapillary capillary
loop (IPCLs) either considered as normal (type A) vs abnormal (type B1–3), when magnification
was available [7]. In the other cases, iodine solution (2 %) was sprayed over the entire esophageal
mucosa and areas clearly not stained were suspected to be neoplasia, which is characterized
by a yellow color in contrast with brown “normal” areas. The pink sign was not used
for delineation since it is usually more centrally seen in the tumour. Delineation
of the lesions was always performed immediately before the resection. Markings were
place 2 mm apart from the delineation line and incision a further 2 mm away from markers
([Fig. 2]).
Fig. 1 Esophageal lesion with lugol chromoendoscopy on the left and NBI on the right. The
delineated area identified with NBI is high-grade dysplasia (brown and flat, below
the nodular whitish lesion) and could have been missed with lugol chromoendoscopy
showing normally stained mucosa at the dysplastic site.
Fig. 2 Esophageal lesion with lugol chromoendoscopy on the left (iodine-unstained area)
and NBI on the right (brown). Markings were place 2 mm apart from de delineation line
(on the left).
Procedures were carried out using Olympus scopes (series 140 and 160 without NBI capability,
180 and 190 including HQ190 with near focus) by six experienced endoscopists. Procedures
were performed under deep sedation (propofol) or anesthesia with endotracheal intubation.
After delineation of lesions margins, they were resected by en-bloc endoscopic mucosal
resection (EMR) or endoscopic submucosal dissection (ESD). ESD was the preferred resection
technique whenever there was a suspicion of superficial invasive SCC (intramucosal
or submucosal) aiming at complete lateral and deep resection (R0 resection) ([Fig. 3]) [13]
[14].
Fig. 3 Esophageal lesion with lugol chromoendoscopy on the left (iodine-unstained area)
and NBI in the middle (brown). Stretched resected specimen on the right.
Patient demographic and clinical characteristics were registered including age, gender,
smoking and alcohol habits, and previous treatments (EMR, ESD and/or radiotherapy).
During endoscopy, lesion location (upper, middle and lower third of the esophagus),
size and morphology according to Paris Classification were assessed. We also registered
the endoscopist performing the procedure, the scope model and the resection technique
(en-bloc EMR or ESD). Histology was assessed and registered according to the following
four categories of the World Health Organization (WHO) Classification of tumors of
the digestive system: invasive SCC (tumor invading the lamina propria), high-grade
intraepithelial neoplasia/dysplasia (HGD) (includes all non-invasive intraepithelial
carcinomas formerly called carcinoma in situ), low-grade intraepithelial neoplasia/dysplasia
(LGD), and negative for neoplasia/dysplasia (no atypia) [15]. Lateral margins were evaluated in all lesions. In case of invasive SCC, depth of
invasion (T1a [intramucosal] and T1b [submucosal]), tumor differentiation, and presence
of lymphovascular invasion were also recorded [15]
[16]. Need for additional treatment (EMR, ESD, surgery, radiotherapy or chemotherapy)
and recurrence rate were also evaluated.
Complete lateral resection rate was defined as cancer/dysplasia free margin where
applicable. Complete resection (R0) was defined as resection with lateral and vertical
margin free of cancer and dysplasia. Curative resection was evaluated in cases of
invasive SCC and defined as lesions that did not need additional treatment. Recurrence
was considered local when a new lesion was diagnosed on a scar of a previous resection
and metachronous if it occurred in another location.
The primary endpoint was complete lateral resection rate – cancer and dysplasia-free
margin. Secondary endpoints were R0 resection and recurrence rates.
Statistical analysis
All continuous variables were described as mean and standard deviation or median and
range while categorical variables were expressed as frequency and proportions. In
case of missing data, the denominator was adjusted to calculate the correct proportions.
Mean differences of continuous variables with a normal distribution were analysed
using an independent Student t-test. The other continuous variables were compared
using the Wilcoxon Mann-Whitney test. To explore univariate associations in the distribution
of categorical data, the X2 test or Fisher’s exact test was used as appropriate. For the primary end-point – complete
lateral resection – a multiple logistic regression was used to adjust for potential
confounders. P < 0.05 was considered statistically significant. Statistical analysis was performed
using the software Statistical Package for Social Sciences (SPSS) (version 23.0).
Results
During the study period, 102 patients were included, 65.7 % male, with a mean age
of 64.9 ± 9.4 years. [Table 1] shows patient demographic and clinical characteristics.
Table 1
Patient demographic and clinical characteristics (n = 102).
|
Total (n = 102)
|
LCE (n = 52)
|
NBI (n = 50)
|
P value
|
|
Age – mean ± SD, years
|
64.9 ± 9.4
|
63.4 ± 9.7
|
66.4 ± 8.7
|
0.110
|
|
Male gender – n (%)
|
67 (66 %)
|
39 (75 %)
|
28 (56 %)
|
0.043
|
|
Smoking habits – n (%)[1]
|
|
Never smoker
|
20 (21 %)
|
7 (14 %)
|
13 (28 %)
|
0.107
|
|
Ever Smoker
|
76 (79 %)
|
42 (86 %)
|
34 (72 %)
|
|
Drinking habits – n (%)[2]
|
|
Yes
|
69 (71 %)
|
43 (86 %)
|
26 (55 %)
|
0.001
|
|
No
|
28 (29 %)
|
7 (14 %)
|
21 (45 %)
|
|
Previous treatment – n (%)
|
|
EMR/ESD
|
7 (7 %)
|
5 (10 %)
|
2 (4 %)
|
|
|
Radiotherapy
|
21 (21 %)
|
14 (27 %)
|
7 (14 %)
|
N/A
|
|
EMR and radiotherapy
|
3 (3 %)
|
2 (4 %)
|
1 (2 %)
|
|
|
Number of lesions – n (%)
|
|
1
|
87 (85 %)
|
44 (85 %)
|
43 (86 %)
|
|
|
2–3
|
13 (13 %)
|
7 (13 %)
|
6 (12 %)
|
N/A
|
|
≥ 4
|
2 (2 %)
|
1 (2 %)
|
1 (2 %)
|
|
SD, standard deviation; LCE, Lugol chromoendoscopy; NBI, narrow-band imaging; EMR,
endoscopic mucosal resection; ESD, endoscopic submucosal dissection; N/A, not applicable.
P value comparing LCE group versus NBI group.
1 Missing data from 6 patients.
2 Missing data from 5 patients.
In total, 132 lesions were resected, with a median lesions per patient of 1 (1–6)
([Table 2]). In 52 % (n = 68), resection margins were defined by LCE (with or without NBI)
and in 48 % (n = 64) by NBI exclusively. The endoscopy was performed by the same endoscopist
in 71 % of cases. A H190 or HQ190 scope was used in more than half of patients. About
two-thirds of the lesions were in the middle third of the esophagus, with a median
size of 30 mm (5–100). Lesions were classified as 0-IIa in 29 % of cases and as 0-IIb
in 42 %. In total 92 % (n = 122) of the lesions were resected by ESD and pathological
analysis revealed invasive SCC in 77 % (n = 101). In invasive SCC lesions, 64 % were
T1a and 36 % T1b, 42 % (39/93) and 11 % (10/93) were well and poorly differentiated,
respectively, and lymphovascular invasion was present in 25 % (15/61).
Table 2
Endoscopy and lesion characteristics (n = 132).
|
Total (n = 132)
|
LCE (n = 68)
|
NBI (n = 64)
|
P value
|
|
Endoscopist – n (%)
|
|
P.D.
|
94 (71 %)
|
47 (69 %)
|
47 (73 %)
|
N/A
|
|
H.P.
|
15 (11 %)
|
9 (13 %)
|
6 (9 %)
|
|
R.Y.
|
12 (9 %)
|
11 (16 %)
|
1 (2 %)
|
|
C.S.
|
5 (4 %)
|
0(%)
|
5 (8 %)
|
|
R.D.
|
5 (4 %)
|
0 (0 %)
|
5 (8 %)
|
|
T.A.
|
1 (1 %)
|
1 (2 %)
|
0 (0 %)
|
|
Scope model – n (%)[1]
|
|
140/160
|
6 (5 %)
|
6 (9 %)
|
0 (0 %)
|
|
|
H180
|
51 (39 %)
|
39 (59 %)
|
12 (19 %)
|
< 0.001
|
|
H190/HQ190
|
73 (56 %)
|
21 (32 %)
|
52 (81 %)
|
|
|
Lesion location – n (%)
|
|
Upper third
|
25 (18.9 %)
|
12 (18 %)
|
13 (20 %)
|
|
|
Middle third
|
80 (60.6 %)
|
43 (63 %)
|
37 (58 %)
|
0.816
|
|
Lower third
|
27 (20.5 %)
|
13 (19 %)
|
14 (22 %)
|
|
|
Lesion size – median ± range, mm
|
30 (5–100)
|
22 (5–90)
|
30 (5–100)
|
0.123
|
|
Lesion morphology – n (%)[2]
|
|
0-Is
|
6 (5 %)
|
2 (3 %)
|
4 (7 %)
|
N/A
|
|
0-IIa
|
36 (29 %)
|
13 (21 %)
|
23 (37 %)
|
|
0-IIb
|
53 (42 %)
|
38 (60 %)
|
15 (24 %)
|
|
0-IIc
|
4 (3 %)
|
2 (3 %)
|
2 (3 %)
|
|
0-IIa + IIb
|
6 (5 %)
|
4 (6 %)
|
2 (3 %)
|
|
0-IIa + IIc
|
16 (13 %)
|
3 (5 %)
|
13 (21 %)
|
|
0-IIb + IIc
|
4 (3 %)
|
1 (2 %)
|
3 (5 %)
|
|
Resection technique – n (%)
|
|
ESD
|
122 (92 %)
|
64 (94 %)
|
58 (91 %)
|
0.522
|
|
EMR
|
10 (8 %)
|
4 (6 %)
|
6 (9 %)
|
|
Lesion histology – n (%)
|
|
Invasive SCC
|
101 (76.5 %)
|
51 (75 %)
|
50 (78 %)
|
N/A
|
|
High-grade dysplasia
|
25 (18.9 %)
|
14 (21 %)
|
11 (17 %)
|
|
Low-grade dysplasia
|
6 (4.5 %)
|
3 (4 %)
|
3 (5 %)
|
|
Follow-up – median ± range, months
|
13 (0–154)
|
18 (0–154)
|
8 (0–40)
|
0.006
|
LCE, Lugol chromoendoscopy; NBI, narrow-band imaging; EMR, endoscopic mucosal resection;
ESD, endoscopic submucosal dissection; SCC, squamous cell carcinoma; N/A, not applicable.
P value comparing LCE group versus NBI group.
1 Missing data from two lesions.
2 Paris Classification, missing data from seven lesions.
The rate of complete lateral resection for invasive carcinoma was 92 % (93/101) and
66 % (86/131) for dysplasia. Complete lateral and deep resection (R0 resection) was
achieved in 49 % (64/130). Incomplete R0 resection (66/130) was due to lateral invasion
in 53 % (n = 35), deep invasion in 32 % (n = 21) and both lateral and deep in 15 %
(n = 10). The curative resection rate was 80.5 %.
Of the 102 patients, 21 underwent additional treatment: one with EMR, one with ESD,
eight underwent surgery, one radiotherapy and, 10 chemoradiotherapy.
The median time of follow-up was 13 months (0–154). Local recurrence rate was 3 %
(n = 4) after a median time of follow-up of 16 months (1–17).
Lugol chromoendoscopy versus NBI
There were more male patients in the LCE group than in the NBI group (75 % vs. 56 %,
P = 0.043). There were no differences in patient’s age between groups.
[Table 2] shows endoscopy and lesions characteristics in both groups. In the NBI group lesions
0-IIa were more frequent (37 %) whereas in the LCE group were 0-IIb (60 %). Lesions
location and size and resection technique were similar between groups. The scope model
was H180/H190/HQ190 in all patients in the NBI group (100 % vs. 91 %, P = 0.028). Median follow-up time was longer in the LCE group (18. vs 8 months, 0.006).
The rate of complete lateral resection for invasive carcinoma was 90 % in the LCE
group and 94 % in NBI group (P = 0.715) and 65 % and 67 % (P = 0.813), respectively, for dysplasia complete lateral resection ([Table 3]). On multiple logistic regression, the results were similar after adjusting for
previous local treatment, scope model, lesion size, morphology and histology and resection
technique (Odds ratio [OR] 0.544, 95 % Confidence interval [95 % CI] 0.160–1.847,
P = 0.329 for cancer complete lateral resection; OR 0.590, 95 % CI 0.219–1.592, P = 0.298 for dysplasia complete lateral resection).
Table 3
Outcomes in the LCE and NBI group (n = 132).
|
LCE (N = 68)
|
NBI (N = 64)
|
P value
|
|
Cancer-free lateral margin – n (%)[1]
|
46 (90 %)
|
47 (94 %)
|
0.715
|
|
Dysplasia free lateral margin – n (%)[2]
|
44 (65 %)
|
42 (67 %)
|
0.813
|
|
R0 resection – n (%)[3]
|
30 (45 %)
|
34 (54 %)
|
0.295
|
|
Local recurrence – n (%)
|
3 (4 %)
|
1 (2 %)
|
0.627
|
|
Metachronous lesions – n (%)
|
11 (16 %)
|
6 (9 %)
|
0.244
|
LCE, Lugol chromoendoscopy; NBI, narrow-band imaging.
1 Included only invasive squamous cell carcinoma (n = 101)
2 Missing data from one lesion.
3 Missing data from two lesions.
Complete lateral and deep resection for cancer and dysplasia was 45 % and 54 % in
LCE and NBI group, respectively (P = 0.295)
There were four cases of local recurrence: three (4 %) in the LCE group and one (2 %)
in the NBI group (P = 0.627). Additional treatment was necessary in 15 % of patients in LCE group (one
EMR, one ESD, four surgery and four chemoradiotherapy) and in 17 % in the NBI group
(four surgery, one radiotherapy and six chemoradiotherapy) (P = 0.697).
Predictive factors of complete lateral resection
An exploratory analysis was performed to identify predictive factors for complete
lateral resection for invasive carcinoma and dysplasia, namely previous local treatment,
scope model, lesion location, size, morphology and histology and resection technique.
The scope model (H180, H190 and HQ190 vs. 140 and 160 series) was the only factor
associated with a higher cancer and dysplasia-free lateral margin (68 % vs 17 %, P = 0.017).
Discussion
Herein, for the first time we compared the effectiveness of NBI and LCE in defining
resection margins for esophageal SCC and dysplasia. We found that mucosal inspection
with LCE before endoscopic resection of esophageal squamous cell lesions was not associated
with increased complete lateral resection rate when compared to NBI alone. The rate
of complete lateral and deep resection was also similar between groups. Furthermore,
there were no differences in recurrence rate and need of additional therapies.
Esophageal cancers carry a high mortality mainly due to diagnosis in advanced-stage,
with a five-year survival rate of less than 10 % [17]. Patients with superficial esophageal SCC, confined to the mucosa or submucosa,
have a significantly better prognosis [17], with a 5-year survival rate exceeding 85 % for pT1N0 lesions [18]. However, diagnosis of early stage esophageal SCC and their precursor lesions can
be difficult because of their flat and isochromatic appearance on conventional white-light
imaging endoscopy [1]
[4]. Lugol staining has long been the gold standard for detection of these lesions [5]
[6]. Although it is a low-cost technique, it may have side effects and lead to complications
[3]
[6]. Therefore, new diagnostic strategies based on virtual chromoendoscopy, such as
NBI, have been used for detection of superficial esophageal squamous cells carcinoma
and dysplasia [7]
[8].
Several studies have compared the accuracy of NBI and LCE for diagnosing esophageal
SCC [4]
[10]
[11]. Only one prospective randomized controlled trial (RCT) have been published on this
subject [4]. In this study, Goda et al compared the diagnostic accuracy of NBI magnifying endoscopy
(NBI-ME) and lugol chromoendoscopy with pink-color sign assessment (LCE-PS) for superficial
esophageal SCC in 294 patients, 147 in each arm. In per-patient analysis, there were
no differences between NBI-ME and LCE-PS in sensitivity (82 % vs 81 %), specificity
(95 % vs 94 %), positive predictive value (88 vs 85 %), negative predictive value
(92 % vs 93 %), and overall accuracy (91 vs 91 %) for diagnosing superficial esophageal
SCC. Median examination time with LCE-PS was 1.5 times longer than NBI-ME (P < 0.001). More recently, a meta-analysis compared the diagnostic accuracy of NBI
and LCE in identifying esophageal HGD and/or SCC [12]. Twelve studies were included, 11 cross-sectional and one RCT, and 1911 patients
were analyzed. The authors showed that NBI has a comparable sensitivity to that of
LCE in both per-patient and per-lesion analysis (88 % vs 92 % and 94 % vs. 98 %, respectively),
and a superior specificity (88 % vs. 82 %, P < 0.001 and 65 % vs 37 %, P < 0.001, respectively). Therefore, NBI presented the same rate of detection of esophageal
HGD and SCC when compared to LCE, but was superior in differentiating HGD and SCC
from other esophageal mucosa alterations. However, to our knowledge, there are no
data comparing the effectiveness of NBI and LCE in delineating the esophageal SCC
margin before endoscopic resection. Previous studies on superficial squamous neoplasms
in the orohypopharynx showed that NBI is a useful tool in defining surgical resection
margins of oral and oropharyngeal cancers, since it decreases the rate of positive
superficial margins [19] and can also decrease recurrence and improve survival as compared to traditional
methods (visual examination alone and white-light nasoendoscopy) [20].
In the current study, the complete lateral resection rate was 66 % for dysplasia and
92 % for invasive carcinoma, comparable to that described in two single-center and
one multicenter study on endoscopic resection of early esophageal SCC (lateral margins
positive for carcinoma in 8.3 % to 9.5 % of lesions) [21]
[22]
[23]. It is interesting to note that local recurrence rate was inferior to complete lateral
resection rate for dysplasia which may be due to pathological examination limitations
due to cautery artifacts or due to cautery effect on non-resected adjacent mucosa.
In the subgroup analysis in our study, the rate of complete lateral resection was
similar between NBI and LCE for esophageal SCC (94 % vs 90 %) and dysplasia (67 %
vs 65 %). The results were similar after adjusting for potential confounders. There
were also no differences in the complete lateral and deep resection rate, relapse
rate and need of additional therapies. The lower rates than expected for complete
lateral resection for dysplasia might be due the fact that in the beginning of our
experience we tried to avoid circumferential resections and to leave a small longitudinal
band of mucosa. Other explanations may be due to the multifocal disease observed in
some patients, rendering both LCE (with multiple unstained areas) and NBI (multiple
brown areas) more difficult to assess, and the fact that 41 % and 20 % in LCE and
NBI groups, respectively had undergone previous treatments with scarring and post-radiotherapy
mucosal changes. We therefore suggest in multifocal esophageal disease to start delineation
with NBI and magnification (and marking), and to use LCE only if multiple brown areas
are confounding the global picture. Once LCE is used, the vascular features including
IPCLs are lost. We would also like to emphasize that NBI + Nearfocus was able in certain
cases to give more information on adjacent dysplastic mucosa (brown area) with normal
LCE staining.
Our study has several limitations. First, it is a retrospective study including patients
over an 18 years period. As such, different scope models were used during resection,
including series 140 and 160 in LCE group which can reduce the rate of complete lateral
resection in this group. In fact, we performed an exploratory analysis to detect predictive
factors of complete lateral resection and more recent scope models (H180/H190/HQ190)
were associated with a higher cancer and dysplasia-free lateral margin. Moreover,
the endoscopists were the same over the study period and therefore their skills in
evaluating and resecting lesions improved over time. Second, lesion morphology was
different between groups; for example, lesions 0-IIb were more frequent in the LCE
group and it might be argued that it is more challenging in to assess their lateral
margins. Finally, the follow-up time is limited and was significantly longer in the
LCE group, since NBI was introduced more recently; this can also influence the recurrence
rate and need of additional therapies in the second group.
