Introduction
Endoscopic submucosal dissection (ESD) is the preferred endoscopic treatment for colorectal
tumors for which en bloc resection cannot be achieved by endoscopic mucosal resection
(EMR) [1]. When compared with the stomach, the colon and rectum have a narrower tubular lumen,
greater angulation at the flexures, and a rather thinner muscular layer. These unavoidable
features make endoscopic control and maneuverability difficult. Hence, colorectal
ESD is considered one of the most challenging endoscopic procedures for novice endoscopists.
Difficult ESD procedures result in undesirable outcomes including incomplete resection,
long procedural times, and complications. Several contributing factors for poor outcomes
have been evaluated [2]
[3]
[4] and include large tumor size [2], tumor located in a non-rectal [3] or flexure location [5], poor endoscopic operability, deep submucosal invasive cancer, and submucosal fibrosis
[2]
[3]
[6]
[7]
[8]
[9]. Severe fibrosis is associated with longer procedure time, higher risk of perforation
[2]
[3]
[4]
[9]
[10], and incomplete resection [6]
[11]
[12]
[13]. These adverse sequelae make it important to correctly determine predictive factors
for technical difficulty before performing ESD. Tumor size, location, and submucosal
invasion depth can be properly evaluated by preoperative endoscopic examination [14]
[15]. In contrast, preoperative endoscopic images do not allow for precise prediction
of presence and stratification of severity of submucosal fibrosis.
Matsumoto et al. [11] revealed that incidence of severe fibrosis in nodular-mixed type granular lateral
spreading tumors (LSTs) was significantly higher than in homogeneous type LSTs. Protruding
lesions [16], deep submucosal invasion, and a tumor size > 30 mm [17] were also considered to be risk factors for submucosal fibrosis. Appropriate diagnostic
measures for addressing severity of submucosal fibrosis in colorectal tumors before
ESD are currently lacking. Endoscopic ultrasonography (EUS) was recently proposed
for predicting depth of gastric and colorectal neoplasms. However, EUS in LSTs showed
only moderate sensitivity and low specificity for prediction of fibrosis [18]. Therefore, the current study was performed to identify macroscopic predictors of
severe submucosal fibrosis in colorectal tumors before ESD and to compare clinical
outcomes of ESD between patients with severe and non-severe fibrosis.
Patients and methods
Patients
Patients who underwent colorectal ESD at Nagoya University Hospital from January 2006
to September 2017 were included. ESD indications for superficial colorectal neoplasms
were based on the principle that for large lesions (> 20 mm in diameter) for which
endoscopic treatment was indicated, en bloc resection by snare EMR would be difficult
and on criteria proposed by the Colorectal ESD Standardization Implementation Working
Group [19]. Patients with lesions which recurred on a previous scar that failed to achieve
endoscopic removal and patients with a history of longstanding inflammatory bowel
disease were excluded from the analysis. All patients were informed of the potential
risks and benefits of ESD, and each patient provided written informed consent to undergo
the endoscopic procedure.
ESD procedure
ESD was performed under conscious sedation by experienced and novice endoscopists
in the university hospital setting. Novice endoscopists referred to colonoscopists
who had performed fewer than 100 colorectal ESD procedures. A standard colonoscope
(PCF-Q260J; Olympus, Tokyo, Japan) or a gastroscope (GIF-Q260J; Olympus, Tokyo, Japan)
were used with appropriate distal attachment. The injection solution was a mixture
of normal saline and 0.4 % sodium hyaluronate (MucoUp; Johnson & Johnson, New Brunswick,
New Jersey, United States) with a small amount of indigo carmine. A FlushKnife (Fujinon-Toshiba
ES System Co., Omiya, Japan), DualKnife (KD-650L; Olympus, Tokyo, Japan), and/or SB
Knife Jr (Sumitomo Bakelite, Tokyo, Japan) and hemostatic forceps (Coagrasper; Olympus,
Tokyo, Japan) were used as appropriate. The ESD procedure was classified as one of
two methods: conventional ESD or hybrid ESD. Conventional ESD involved submucosal
dissection with a knife, and hybrid ESD involved snaring following circumferential
incision and sufficient submucosal dissection [20].
Probable risk factors and definitions
Patient-related, tumor-related, and procedure-related variables were investigated.
Degree of submucosal fibrosis was judged by two reviewers who retrospectively evaluated
the still images. The reviewers evaluated without known patient-related and tumor-related
variables but with known procedure-related information. The reference to the classification
was suggested by Hiroshima University as follows: F0, no fibrosis, manifesting as
a blue transparent layer; F1, mild fibrosis, appearing as a white web-like structure
in the blue submucosal layer; and F2, severe fibrosis, appearing as a white muscular
structure without a blue transparent layer in the submucosal layer [11] ([Fig. 1]). F0 and F1 were combined into non-severe fibrosis group for comparison with F2
as the severe fibrosis group.
Fig. 1 Fibrosis grading. a F0 – no fibrosis; manifests as a blue transparent layer. b F1 – mild fibrosis; appears as a white web-like structure in the blue submucosal
layers. c F2 – severe fibrosis; appears as a white muscular structure without blue transparent
layer.
Tumor location was stratified into right-sided (from the ileocecal valve to the transverse
colon), left-sided (from the splenic flexure to the sigmoid colon), or rectal. Protruding
tumors included sessile and subpedunculated types, and superficial tumors included
elevated, flat, and depressed types in accordance with the Japanese Society for Cancer
of the Colon and Rectum (JSCCR) criteria [21]. LSTs were subclassified into four types: LST granular homogeneous, LST granular
nodular-mixed, LST non-granular flat-elevated, and LST non-granular pseudo-depressed
[22]. Tumors that extended across at least one fold and the oral-side margin and were
difficult to detect in the forward view were classified as beyond-fold lesions ([Fig. 2]). Fold convergence was recognized when more than three concentrating folds were
visible after substantial distention of the colonic wall ([Fig. 3]). Lifting conditions were assessed as good or poor according to expansion of submucosal
space following submucosal injection.
Fig. 2 Beyond fold tumors extend across at least one mucosal fold and oral-side margin was
difficult to detect in the forward view.
Fig. 3 Fold convergence recognized as more than three concentrating folds visualized after
distention of the colonic wall.
Tumor size and height were measured histologically from excised specimens. To analyze
the correlation between fibrosis and tumor size, the tumor was divided by two cut
points (≥ 30 or ≥ 40 mm) according to the long axis. En bloc resection refers to resection
of a specimen in a single piece. Procedure time was defined as the time from mucosal
incision to complete tumor removal. Resection speed was calculated from the surface
area (specimen diameter in long axis × specimen diameter in short axis × π × 0.25)
divided by the procedure time in minutes. Tumor height of protruding lesions was measured,
including Is, Isp, and Is + IIa lesions. Tumor volume was then calculated by estimating
the tumor as an elliptical cone shape using the following formula: surface area × height × 1/3.
With respect to complications, perforation was diagnosed either when the muscle layer
was injured and the peritoneal cavity was observed endoscopically or when free air
was found on a plain abdominal radiograph or computed tomography image. If the muscle
layer was injured but it was not penetrated through the peritoneal cavity, it was
defined as damaged muscularis propria. Pain in the abdomen with localized tenderness
that occurred after ESD was also recorded. When abdominal pain and localized tenderness
accompanied with fever (≥ 37.8 °C) and/or leukocytosis (≥ 10,000 cells/μL) and/or
C-reactive protein (CRP) > 0.5 without evidence of perforation on the radiologic images,
post-ESD coagulation syndrome was diagnosed [23]. Delayed bleeding was defined as hematochezia, melena, hypotension, or a hemoglobin
level that had decreased by ≥ 2 g/dL.
Histopathological assessment
All specimens were fixed in 10 % formalin, cut into 2-mm sections, and examined microscopically.
Complete resection was defined as horizontal margin-negative and vertical margin-negative.
A positive or undetermined margin was considered as incomplete resection. Curative
resection was identified using the JSCCR guideline criteria [21], which required all four of the following characteristics: a well/moderately differentiated
or papillary carcinoma, no vascular invasion, a submucosal invasion depth < 1,000
μm, and grade 1 budding. Tumors were classified as adenoma, intramucosal adenocarcinoma,
carcinoma with superficial submucosal invasion (< 1,000 mm), or carcinoma with deep
submucosal invasion (≥ 1,000 mm) according to the World Health Organization classification
system [24].
Statistical analysis
Continuous variables were summarized as the mean ± standard deviation or median and
range. Categorical variables were presented as frequency (%). Significance of the
differences of variables between groups was determined using a chi-square test for
categorical variables, Student’s t-test, and Wilcoxon Rank-Sum test for continuous variables, as appropriate. Logistic
regression analysis was performed to identify independent risk factors for severe
submucosal fibrosis. Odds ratios (ORs) and 95 % confidence intervals (CIs) were calculated
with univariate and multivariate models. Data were analyzed by SPSS for Windows (version
19.0; IBM Corp., Armonk, New York, United States), and P values < 0.05 were considered statistically significant.
This study was approved by the Institutional Review Board of Nagoya University Hospital
and performed according to the Declaration of Helsinki.
Results
Patients and procedures
From all 551 patients with 581 colorectal lesions, 57 lesions were initially excluded.
Neuroendocrine tumor, unable to complete ESD, previous biopsy or recurrent tumor and
concomitant ulcerative colitis accounted for exclusion in 24, 15, 10 and 8 patients,
respectively. Thus, 524 lesions in 497 patients were analyzed. Sixty percent of the
patients were male, and their mean age was 67.8 years (31 – 92 years). Mean size of
lesions was 43.9 mm (13 – 175 mm). Severity of submucosal fibrosis was defined as
F0 in 203 lesions, F1 in 241 lesions, and F2 in 80 lesions. Baseline characteristics
are shown in [Table 1] and [Table 2].
Table 1
Baseline characteristics of included patients.
|
Baseline
|
Total
|
Fibrosis group
|
P value
|
|
(n = 524)
|
Severe
(n = 80)
|
Non-severe
(n = 444)
|
|
Sex (male)
|
315 (60.1)
|
52 (65.0)
|
263 (59.2)
|
0.332
|
|
Age (years), mean ± SD
|
67.8 ± 10.7
|
66.8 ± 11.4
|
68.0 ± 10.5
|
0.367
|
|
Comorbidity
|
|
|
22 (4.2)
|
5 (6.2)
|
17 (3.8)
|
0.358
|
|
|
57 (10.9)
|
12 (15.0)
|
45 (10.1)
|
0.198
|
|
|
135 (25.8)
|
20 (25.0)
|
115 (25.9)
|
0.865
|
|
|
41 (7.8)
|
8 (10.0)
|
33 (7.4)
|
0.431
|
|
|
5 (1.0)
|
0 (0)
|
5 (1.1)
|
1.0
|
|
|
11 (2.1)
|
1 (1.2)
|
10 (2.3)
|
1.0
|
|
|
10 (1.9)
|
0 (0)
|
10 (2.3)
|
0.373
|
|
|
6 (1.1)
|
0 (0)
|
6 (1.4)
|
0.597
|
|
Use of antithrombotics
|
73 (13.9)
|
7 (8.8)
|
66 (14.9)
|
0.146
|
SD, standard deviation; CAD, coronary artery disease; AF, atrial fibrillation; CKD,
chronic kidney disease; ESRD, end-stage renal disease.
Table 2
Baseline characteristics of colorectal tumors undergoing ESD.
|
Baseline
|
Total
|
Fibrosis group
|
P value
|
|
(n = 524)
|
Severe
(n = 80)
|
Non-severe
(n = 444)
|
|
Size (mm), mean ± SD
(range)
|
43.9 ± 19.1
(13 – 175)
|
48.8 ± 20.7
|
43.0 ± 18.7
|
0.011
|
|
Diameter ≥ 30 mm
|
423 (80.7)
|
68 (85.0)
|
355 (80.0)
|
0.292
|
|
Diameter ≥ 40 mm
|
279 (53.2)
|
52 (65.0)
|
227 (51.1)
|
0.022
|
|
Height (mm), mean ± SD
|
12.1 ± 6.6
(n = 89)
|
16.0 ± 7.6
(n = 28)
|
10.3 ± 5.2
(n = 61)
|
0.0008
|
|
Tumor volume (mm3), median (range)
|
2,473
(75 – 30,615)
|
5,332
(75 – 30,615)
|
1,773
(144 – 30,521)
|
0.0003
|
|
Pathology
|
|
|
|
< 0.001
|
|
|
117 (22.3)
|
12 (15.0)
|
105 (23.6)
|
|
|
|
320 (61.1)
|
39 (48.7)
|
281 (63.3)
|
|
|
|
55 (10.5)
|
16 (20.0)
|
39 (8.8)
|
|
|
|
32 (6.1)
|
13 (16.3)
|
19 (4.3)
|
|
|
Location
|
|
|
|
0.623
|
|
|
247 (47.1)
|
34 (42.5)
|
213 (48.0)
|
|
|
|
133 (25.4)
|
21 (26.2)
|
112 (25.2)
|
|
|
|
144 (27.5)
|
25 (31.2)
|
119 (26.8)
|
|
|
Morphology
|
|
|
|
< 0.001
|
|
|
255 (48.7)
|
26 (32.5)
|
229 (51.6)
|
|
|
|
199 (38.0)
|
30 (37.5)
|
169 (38.0)
|
|
|
|
70 (13.3)
|
24 (30.0)
|
46 (10.4)
|
|
|
Beyond fold
|
340 (64.9)
|
60 (75.0)
|
280 (63.1)
|
0.039
|
|
Fold convergence
|
51 (9.7)
|
20 (25.0)
|
31 (7.0)
|
< 0.001
|
SD, standard deviation; SM, submucosal; LST-G, lateral spreading tumor granular type;
LST-NG, lateral spreading tumor non-granular type.
The overall en bloc resection and curative resection rates were 94.3 % and 77.7 %,
respectively. For tumors considered to have non-curative resection (n = 117), 32 had
deep submucosal invasion, 10 of which had a positive vertical margin ([Supplementary Table 1]). The major population in the non-curative resection group comprised adenomatous
lesions with an undeterminable horizontal margin. Complications were observed in 73
patients (13.9 %). No procedure-related deaths occurred. Additional operations were
needed in 40 patients (7.6 %) who underwent non-curative resection (positive vertical
margin, n = 23; lymphovascular and deep submucosal invasion, n = 17). Emergency operations
for delayed perforation were performed in two patients ([Table 3]).
Supplementary Table 1
Non-curative resection tumors: association of margin and invasion depth.
|
Margin positive
|
N = 91
|
Deep SM
(n = 12)
|
Superficial SM (n = 18)
|
Intramucosal carcinoma
(n = 44)
|
Adenoma
(n = 17)
|
|
Positive or undetermined horizontal margin
|
68
|
2
|
10
|
41
|
15
|
|
Positive or undetermined vertical margin
|
14
|
6
|
6
|
2
|
0
|
|
Positive or undetermined both margin
|
9
|
4
|
2
|
1
|
2
|
SM, submucosa.
Table 3
Outcome of colorectal ESD according to degree of fibrosis.
|
Parameters
|
Total
|
Fibrosis group
|
P value
|
|
(n = 524)
|
Severe
(n = 80)
|
Non-severe
(n = 444)
|
|
En bloc resection
|
494 (94.3)
|
73 (91.2)
|
421 (94.8)
|
0.197
|
|
Curative resection
|
407 (77.7)
|
50 (62.5)
|
357 (80.4)
|
< 0.001
|
|
Poor lifting
|
89 (17.0)
|
36 (45.0)
|
53 (11.9)
|
< 0.001
|
|
Hybrid method
|
87 (16.6)
|
13 (16.2)
|
74 (16.7)
|
0.927
|
|
Procedure time (min), median(range)
|
99
(63 – 146)
|
136.5
(96 – 194)
|
91.5
(61 – 141)
|
< 0.001
|
|
Resection speed (mm2/min), median(range)
|
11.0
(6.5 – 18.5)
|
9.3
(5.9 – 13.7)
|
11.5
(6.9 – 19.5)
|
0.005
|
|
Complication
|
73 (13.9)
|
13 (16.2)
|
60 (13.5)
|
0.515
|
|
|
20 (3.8)
|
3 (3.8)
|
17 (3.8)
|
1.0
|
|
|
84 (16.0)
|
27 (33.8)
|
57 (12.8)
|
< 0.001
|
|
|
31 (5.9)
|
6 (7.5)
|
25 (5.6)
|
0.450
|
|
|
55 (10.5)
|
9 (11.2)
|
46 (10.4)
|
0.811
|
|
|
66 (12.6)
|
14 (17.5)
|
52 (11.7)
|
0.151
|
|
|
286 (57.8)
|
55 (69.6)
|
231 (55.5)
|
0.020
|
|
|
48 (9.2)
|
8 (10.0)
|
40 (9.0)
|
0.778
|
|
Emergency operation
|
2 (0.4)
|
0 (0)
|
2 (0.5)
|
1.0
|
MP, muscularis propria; T, temperature; WBC, white blood cell; CRP, C-reactive protein;
PECS, post-ESD coagulation syndrome.
Severity of submucosal fibrosis and procedure results
When compared with non-severe fibrosis, patients with severe fibrosis encountered
a longer procedural time (137 vs. 92 min, P < 0.001) and slower resection speed (9.3 vs. 11.5 mm2/min, P = 0.005). A significantly lower rate of curative resection was observed in the severe
fibrosis group (62.5 % vs. 80.4 %, P < 0.001). Forty-three percent of non-curative tumors in the severe fibrosis group
(13/30) invaded the deep submucosal layer. Severe fibrosis was also associated with
a higher incidence of complications (16.2 % vs. 13.5 %, P = 0.52) and a lower incidence of en bloc resection (91.2 % vs. 94.8 %, P = 0.20), but these differences did not reach statistical significance.
Clinical endoscopic characteristics with respect to degree of submucosal fibrosis
Severity of submucosal fibrosis was associated with large tumor size (≥ 40 mm), tumor
morphology, tumor volume, and histology. Probable risk factors for submucosal fibrosis
are shown in [Table 4]. Deep submucosal invasive cancer and protruding lesions were associated with a higher
risk of severe fibrosis. The finding of a lesion across a semilunar fold and presence
of fold convergence were independent predictors of severe fibrosis. No significant
differences were found among the subtypes of LSTs. Among 89 lesions with protruding
morphology, mean height in the severe fibrosis group was significantly higher than
that in the non-severe fibrosis group (16.0 ± 7.6 vs. 10.3 ± 5.2 mm, P = 0.0008); the same findings were obtained for tumor volume ([Table 2]). Given the limited number of protruding lesions, we did not perform a regression
analysis for tumor height and volume. Multivariate analysis showed that histologic
and morphologic factors were different between the two groups. Tumors with deep submucosal
invasion (OR, 4.80; 95 % CI, 1.75 – 13.2; P = 0.001) and protruding morphology (OR, 4.43; 95 % CI, 2.17 – 9.06; P = 0.001) carried a higher possibility of severe fibrosis. Endoscopic features of
beyond-fold lesions (OR, 2.07; 95 % CI, 1.10 – 3.89; P = 0.024), fold convergence (OR, 5.20; 95 % CI, 2.47 – 10.91; P < 0.001), and large tumor size (≥ 40 mm) (OR, 2.22; 95 % CI, 1.23 – 4.02; P = 0.008) were identified as independent predictors of severe fibrosis ([Table 4]).
Table 4
Univariate and multivariate analyses of risk factors for severe submucosal fibrosis.
|
Factors
|
Univariate
|
Multivariate
|
|
OR (95 %CI)
|
P value
|
OR (95 %CI)
|
P value
|
|
Pathology
|
|
|
1
|
< 0.0001
|
1
|
0.0012
|
|
|
1.21 (0.61 – 2.41)
|
|
0.96 (0.46 – 2.00)
|
|
|
|
3.59 (1.56 – 8.26)
|
|
2.04 (0.81 – 5.11)
|
|
|
|
5.99 (2.38 – 15.09)
|
|
4.80 (1.75 – 13.23)
|
|
|
Morphology
|
|
|
1
|
< 0.001
|
1
|
0.001
|
|
|
1.56 (0.89 – 2.74)
|
|
1.21 (0.61 – 2.42)
|
|
|
|
4.60 (2.43 – 8.71)
|
|
4.43 (2.17 – 9.06)
|
|
|
Beyond fold
|
1.76 (1.02 – 3.02)
|
0.041
|
2.07 (1.10 – 3.89)
|
0.024
|
|
Fold convergence
|
4.44 (2.38 – 8.29)
|
< 0.001
|
5.20 (2.47 – 10.91)
|
< 0.001
|
|
Diameter ≥ 40 mm
|
1.78 (1.08 – 2.91)
|
0.023
|
2.22 (1.23 – 4.02)
|
0.008
|
SM, submucosal; LST-G, lateral spreading tumor granular type; LST-NG, lateral spreading
tumor non-granular type.
Discussion
This study demonstrated the association of severity of submucosal fibrosis with preoperative
endoscopic findings and clinical outcomes. Severe submucosal fibrosis was more commonly
found in association with protruding lesions, larger tumor size, higher tumor volume,
and tumors with deep submucosal invasion. When compared with non-severe fibrosis,
severe fibrosis resulted in prolonged procedure time and lower curative resection
rate. Complication and en bloc resection rates did not differ between the two groups.
Large tumor size (≥ 40 mm), lesions across the fold, and presence of fold convergence
were identified as independent predictors of severe submucosal fibrosis.
Submucosal fibrosis is triggered by multiple stimuli [25]. Injuries from chronic inflammation (e. g., ulcerative colitis) [26], prior tattooing [27], prior mucosal biopsy [28]
[29], or resection [10]
[30] are well-known risk factors. Direct tumor invasion or a desmoplastic reaction from
carcinoma with submucosal invasion also appears to be associated with severe submucosal
fibrosis [6]
[17]. Nevertheless, mucosal carcinomas and adenomas, although in a lower proportion,
also exhibit severe fibrosis [6]. These make preoperative diagnosis of submucosal fibrosis difficult.
As a tumor increases in size, the accompanying fibrosis is more likely to worsen.
Large tumor size was generally accepted as a significant factor for poor ESD outcome
[31]. Whether this is caused by more frequent deep invasion [32] or severity of fibrosis is uncertain. However, recent studies have shown that large
tumor size (> 50 mm) is not significantly associated with these outcomes [7]
[33]. Hence, fibrosis might be responsible for these poor outcomes. Several tumor sizes
have been specified to categorize severe versus non-severe fibrosis. In the current
study, tumors > 40 mm contained significantly more severe fibrosis, while Lee et al.
[17] found that a tumor size of 30 mm was the cut-off point. This different cut-off point
could have resulted from the different mean sizes of tumors included in each study
(31.8 ± 11 mm in the study by Lee et al. [17] vs. 43.9 ± 19.1 mm in the current study). However, size was not an independent predictor
of severe fibrosis in another retrospective study [6] with a smaller mean tumor size (25.9 ± 10.7 mm). Thus, the optimal cut-off point
for colorectal tumors with severe submucosal fibrosis should be evaluated in a larger-scale
study.
Tumor configuration also influenced ESD outcome. Imai et al. [7] revealed that regardless of size, a protruding lesion was an indicator of en bloc
resection failure or perforation (OR, 3.58; 95 % CI, 1.81 – 7.07]. Unfavorable outcomes
with protruding tumors resulted from a higher rate of submucosal cancer invasion and
the higher likelihood of severe fibrosis when compared with LSTs [16]. Similarly, a large retrospective study by Inada et al. [10] showed that protruding tumors, particularly those > 40 mm, contained a higher proportion
of severe fibrosis than superficial tumors. Our study emphasizes that protruding morphology
is a strong predictor of severe fibrosis. Nevertheless, why protruding tumors carried
a higher incidence of severe fibrosis remains unclear. Despite the limited number
of protruding lesions, our study has shown that severe tumor fibrosis is associated
with greater tumor height and volume. During peristaltic movement of the colonic wall,
a taller tumor might induce greater physical stress on the base of the lesion than
a shorter tumor, and this stress may lead to repetitive trauma to the submucosal tissue,
with resultant chronic inflammation and fibrosis [16]. Genetic factors probably also play a major role in development of polypoid-type
carcinoma [34].
A recent large retrospective colorectal ESD study showed that presence of an underlying
semilunar fold and fold convergence were independent risk factors for en bloc resection
failure and perforation [7]. Lesions across more than one fold are also reportedly associated with longer procedure
time [8]. Fold convergence is well recognized as an endoscopic feature of submucosal deeply
invasive colorectal carcinoma [35]. Severe fibrosis and newly apparent fold convergence on a flat lesion that had not
been present prior to mucosal biopsy was also demonstrated in a case report [28]. In brief, the endoscopic finding of fold convergence could be representative of
either submucosal fibrosis or submucosal invasive cancer. Our study demonstrated an
association between severe fibrosis and these endoscopic findings.
Our study had some limitations. It was a retrospective study of ESD performed by endoscopists
with heterogeneous skills and performance. However, en bloc resection rate, curative
resection rate, and complications in our study were comparable with those in large
multicenter and worldwide studies [36]
[37]. In addition, the reviewers evaluated the images without known patient-related and
tumor-related variables but with the known procedure-related information. Moreover,
we did not stratify severity of submucosal fibrosis by pathological evaluation findings,
but rather, by endoscopic diagnosis. To overcome these limitations, we had all endoscopic
images reviewed by two reviewers, and experienced ESD endoscopists examined those
with uncertain fibrosis staging. We assumed that degree of submucosal fibrosis based
on endoscopic findings was correlated with histologic fibrosis [38]. Finally, we excluded lesions that could not be removed endoscopically, which could
be a substantial selection bias. A barrier to complete endoscopic resection might
be severe fibrosis, indicating that the impact of severe fibrosis might be greater
than shown by our study results.
The main strength of the current investigation was our large study population, which
allowed us to assess only preoperatively available factors that greatly affect accurate
stratification of lesions before ESD. In addition, the procedures were not performed
by only experienced endoscopists. This enabled us to evaluate procedures performed
by novice endoscopists, increasing the generalizability of our findings. Finally,
we excluded lesions with known risk factors for severe fibrosis. This allowed us to
identify precise predictors that can be applied in general practice. To the best of
our knowledge, this is the first study to identify preoperative endoscopic features
other than tumor size that predict the severity of fibrosis.
Conclusion
In conclusion, preoperative predictors of severe submucosal fibrosis were identified
in this study. Recognizing these factors could help to accurately stratify lesions
that tend to have poor outcomes. Thus, when size > 40 mm, protruding morphology, presence
of fold convergence, or underlying semilunar folds are observed during preoperative
endoscopy, endoscopists performing ESD should suspect the presence of severe submucosal
fibrosis. Further studies are expected to prove the relationship between tumor morphology
and severity of fibrosis.
