Introduction
Endoscopic submucosal dissection (ESD) has been described to achieve higher en-bloc
resection and lower recurrence rates in esophageal, gastric, and colorectal tumors
[1]
[2] when compared to endoscopic mucosal resection (EMR). This technique using knives
can lead to complete resections in lesions with presumed submucosal invasion in preoperative
diagnosis and avoid major surgery.
However, the duodenum represents a differentiated location with its own characteristics
that can influence the outcomes of ESD in non-ampullary superficial duodenal tumors
(NASDTs). The thin wall, narrow lumen, limited maneuverability, and stability of the
scope and increased risk of complications may lead to poor outcomes even in experienced
referral centers. Additionally, the malignancy of NASDTs may be extremely difficult
to determine based on the macroscopic characteristics [3]. Indeed, poor mucosal lifting may be due to the presence of Brunner’s glands in
the submucosal layer and not to fibrosis of invasive patterns.
Epithelial duodenal tumors are rare, and EMR has been proposed as the acceptable treatment
in these cases, including large lesions requiring piecemeal approach [4]. Actually, there are no randomized controlled trials comparing ESD to EMR, and the
retrospective published studies to date performing both techniques are based on different
criteria [5]
[6]
[7]. Thus, the choice of ESD or EMR in these lesions is challenging and the benefits
of ESD remain unclear. The aim of the present study was to comparatively analyze the
feasibility, outcomes, and safety of these procedures in NASDTs.
Methods
Patients
This is an observational and retrospective study. All consecutive patients presenting
with NASDTs who underwent ESD or EMR in our center between June 2005 and March 2017
were included. Pedunculated lesions or neuroendocrine tumors and lesions resected
by polypectomy (without injection into submucosal layer) were not considered. Patients
with ampullary tumors as well as those with a history of familial adenomatous polyposis
were also excluded. Age, sex, tumor size, number of lesions, procedure time, endoscopic
follow-up duration, and histological diagnosis were collected. The study protocol
was approved by the local ethics committee and was registered as 2017/16AOU/405.
EMR and ESD procedures
Endoscopic resection approach was decided based on tumor size, macroscopic morphology,
suspected type of lesion, location, and maneuverability. Previous endoscopic ultrasonography
was not performed systematically. Basically, ESD was carried out when R0 was not supposed
to be achieved by EMR, in case of local recurrence after previous endoscopic treatment,
non-lifting sign, or suspicion of malignancy. Large superficial tumors were also intended
to be resected by ESD. EMR was performed using a single-use symmetric (Captivator,
Boston Scientific, Zaventem, Belgium) or asymmetric snare (Olympus, Berchem, Belgium)
after submucosal injection by a mixture of epinephrine (dilution 1:50 000), fluid
gelatin and methylene blue, or normal saline injection. A forward-viewing gastroscope
or side-viewing duodenoscope were used at the discretion of the endoscopist. ESD was
performed by DualKnife (Olympus) as the main device, using monopolar current (VIO3/VIO300 D
generators; ERBE, Zaventem, Belgium) by endocut I/Q (effect 1 – 2) and forced/swift
coagulation modes. Previous mucosal marking was carried out in large tumors or when
the limits where unclear. A transparent hood was placed on the tip of the endoscope
in all cases.
Hemostatic forceps (Coagrasper, Olympus) in soft coagulation mode (output: 80 W/5.8)
was employed in cases of significant bleeding when DualKnife coagulation was insufficient.
At the end of the excision, preventive hemostasis was decided on an individual basis
and at the choice of the endoscopist. Prophylactic endoscopic closure to reduce the
risk of delayed complications was performed using endoscopic clipping and/or endoloop
(Olympus) if possible. The size of the excised lesion was measured after spreading
onto a white cork surface, and tumor size was defined as the largest diameter. Radiological
imaging was performed during the procedure in case of suspected perforation. All procedures
were performed under general anesthesia by an experienced endoscopist (PD, HP, RY,
TM) after written informed consent. A treatment by proton-pump inhibitors was proposed
after resection in all cases, and a progressive liquid diet was indicated 12 – 48
hours after the procedure if there were no complications. Finally, a follow-up gastroscopy
was indicated in 1 – 3 months.
Histopathological assessment
The specimens were assessed by an expert pathologist (AJM). The resected specimen
was cut into 2 – 3-mm slices after fixation in formalin. Histological type, size,
depth of invasion, lateral and vertical margins, and lymphovascular invasion were
evaluated. Adenomatous tumors were classified into low-grade dysplasia, high-grade
dysplasia, and adenocarcinoma. Histopathological grade was assigned according to the
highest grade present in the tumor. Malignancy was defined as high-grade dysplasia
or adenocarcinoma.
Outcomes and definitions
The following main outcomes were comparatively evaluated: en-bloc and complete (R0)
resection rates and local recurrence. Secondary outcomes were delayed bleeding and
perforation. En-bloc resection was defined as endoscopic resection of the tumor in
1 piece. Complete resection rate was only calculated in adenomatous/dysplastic NASDTs
or carcinoma and was defined as resection of the tumor with free lateral and vertical
margins. Delayed bleeding was defined as hematemesis or melena requiring endoscopic
hemostasis, readmission, blood transfusion, or surgical intervention after the completion
of the procedure. Perforation was classified as intraoperative or delayed as well
as major or minor, based on whether intra-abdominal space was directly visualized.
Local recurrence was defined as regrowth of adenomatous tissue at the site of the
previous resection, confirmed by a positive histology during the follow-up. Follow-up
endoscopy was routinely performed at 3 and 6 months following the resection and annually
thereafter.
Statistical analysis
The main and secondary outcomes (categorical variables) were compared using χ² or
Fisher’s exact tests. A subgroup analysis including only tumors ≥ 20 mm resected by
ESD or EMR was also carried out to highlight these outcomes in more comparable populations.
Normally and non-normally distributed continuous variables were analyzed by Student’s
t-test and the Mann-Whitney U-test. They were presented as mean (SD) or median (range),
respectively. A 2-sided P-value of < 0.05 was considered statistically significant. SPSS version 23 was used
(IBM, SPSS Inc, Chicago, IL).
Results
Patients
Between 2005 and 2017, there were 203 patients presenting with non-ampullary neoplasms
who underwent ESD or EMR. Those with familial adenomatous polyposis (n = 39) and subepithelial
lesions (n = 14) were excluded. Thus, 166 NASDTs in 150 patients (age: 66 years, range:
31 – 83, 42.7 % males) were resected by ESD (n = 37) or EMR (n = 129) and were finally
included. There were 2 patients with 3 tumors and 12 presenting with 2 lesions. Among
these cases with multiple lesions, the endoscopic resection of all the tumors was
performed during the same procedure in most of them (n = 11, 78.6 %).
Overall, most of the resected lesions (95.2 %) were located in D2 or distal with no
differences between both groups (91.9 % vs. 96.1 %, P = 0.289). The median procedure time was higher in ESD cases (81 vs. 50 min, P = 0.007). The median tumor size was larger in lesions resected by ESD compared to
EMR (25 vs. 20 mm, P = 0.01), while the overall endoscopic median size was 20 mm (range: 5 – 50). Indeed,
98 NASDTs (59 %) were ≥ 20 mm in size. Additionally, the 21.1 % of tumors involved ≥ 50 %
of the duodenal circumference. The hybrid technique was used in most of ESD procedures
(n = 29, 78.4 %) as shown in [Fig. 1], and the cap-aspiration technique was used in 8 patients who underwent EMR.
Fig. 1 A 15-mm adenoma with high-grade dysplasia located in the bulb. Mucosal marking using
a transparent hood (a). Circumferential endoscopic submucosal dissection by DualKnife (b). En-bloc hybrid resection using an asymmetric snare (c). Duodenal mucosal defect after resection (d).
Comparative outcomes
Concerning ESD/EMR outcomes, there were no differences in en-bloc and complete resection
rates as shown in [Table 1]. Incomplete resection in en-bloc resected lesions was due to positive lateral margins
in all cases. Closure of the mucosal defect was carried out in 68.1 % of patients
with statistically significant differences between the first (2005 – 2011) and second
periods (2012 – 2017) of the study (27.5 % vs. 83.6 %, P < 0.001). Nine perforations were confirmed in 6 lesions resected by ESD and 3 by
EMR (P = 0.001) ([Table 2]). All of them occurred during the first period of the study in tumors larger or
equal to 20 mm (77.8 %) without prophylactic closure (11.3 % vs. 2.7 %, P = 0.022). Most of the perforations in the ESD group occurred during full ESD technique
(n = 2/8, 25 %) followed by hybrid ESD approach (n = 4/39, 13.8 %). The median length
of the hospitalization was 3 days (range: 2 – 15). Notably, en-bloc resection was
achieved in 3 lesions resected by ESD in spite of the perforation.
Table 1
Outcomes of endoscopic resection procedures for 166 lesions in 150 patients with NASDTs.
|
ESD (n = 37)
|
EMR (n = 129)
|
P-value
|
Age (median, range, years)
|
64 (44 – 83)
|
66 (31 – 82)
|
0.187
|
Tumor size (median, range, mm)
Size ≥ 20 mm
|
25 (12 – 50)
29 (78.4 %)
|
20 (5 – 50)
69 (53.5 %)
|
0.010[1]
0.007[1]
|
Procedure time
|
81 (5 – 236)
|
50 (14 – 186)
|
0.007[1]
|
En-bloc resection
|
11 (29.7 %)
|
57 (44.2 %)
|
0.115
|
R0 resection[2]
|
7 (19.4 %)
|
43 (35.5 %)
|
0.069
|
Histology
|
|
22
|
57
|
|
|
12
|
56
|
|
|
2 (5.6 %)
|
10 (8.1 %)
|
0.607
|
|
1
|
6
|
|
Closing of mucosal defect
|
25 (67.6 %)
|
88 (68.2 %)
|
0.940
|
Perforation
|
6 (16.2 %)
|
3 (2.3 %)
|
0.001[1]
|
|
5 (2/3)
|
3 (1/2)
|
|
|
1 (1/0)
|
0
|
|
Delayed bleeding
|
3 (9.3 %)
|
12 (8.1 %)
|
0.823
|
Follow-up (median, range, months)
|
4.5 (2 – 84)
|
10 (2 – 125)
|
0.278
|
Local recurrence[3]
|
5 (14.7 %)
|
17 (16.7 %)
|
0.788
|
1 Statistically significant.
2 En-bloc and R0 resection in 157 patients with adenoma or adenocarcinoma.
3 Calculated for 136 patients with follow-up.
Table 2
Characteristics of patients presenting with a duodenal perforation.
Spreading
|
Location
|
Technique
|
Type of perforation
|
Hospitalization (days)
|
< 25 %
|
D2
|
Full ESD
|
Major
|
4
|
50 %
|
D2
|
Full ESD
|
Minor
|
15
|
50 %
|
D3
|
Hybrid ESD
|
Delayed
|
10
|
50 %
|
D2
|
Cap-EMR
|
Major
|
7
|
25 %
|
D2
|
Hybrid ESD[1]
|
Major
|
2
|
66 %
|
D2
|
EMR
|
Minor
|
2
|
50 %
|
D2
|
Hybrid ESD[2]
|
Minor
|
2
|
75 %
|
D2
|
EMR
|
Minor
|
3
|
< 25 %
|
Bulb
|
Hybrid ESD[2]
|
Minor
|
3
|
1 The perforation occurred during snare resection.
2 The perforation occurred during the ESD procedure.
Overall, they were successfully treated during the procedure by standard (n = 7) clips,
over-the-scope clips (Ovesco, Geetbets, Belgium) (n = 4), and endoloop (Olympus) (n = 2)
in all but 1 patient who underwent emergency surgery after hybrid ESD. A 59-year-old
woman underwent a piecemeal hybrid ESD for a 40-mm tumor with high-grade dysplasia
located in D3. The mucosal defect was not closed due to the location of the lesion
and limited scope maneuverability, but a contrast opacification performed after the
procedure was normal. However, 12 hours later she presented with abdominal pain, and
a delayed perforation was confirmed by CT abdominal scan, needing emergency open surgery.
The patient was discharged after 14 days, and a feeding jejunostomy was placed for
40 days. Duodenal contrast opacification under radiological control was performed
to ensure the closure of the intraoperative perforation in all cases and antibiotics
were indicated.
The delayed bleeding rate was 9 %, with no differences between the ESD and EMR groups.
Interestingly, two-thirds (n = 10) of patients presenting with delayed bleeding did
not have a mucosal defect closure following resection. Therefore, delayed bleeding
was much more frequent in lesions without prophylactic closure (18.9 % vs. 4.4 %,
P = 0.002), and there was no bleeding in lesions located in the bulb.
Histology and follow-up
One hundred thirty-six patients (34 ESD, 102 EMR) underwent a median follow-up of
6.5 months (range: 2 – 125). Local recurrence occurred in 5 (14.7 %) and 17 (16.7 %)
tumors in the ESD and EMR groups, respectively (P = 0.788), and it was managed endoscopically in 19 patients and surgically in 3 cases.
The global malignancy rate was 50.3 % with comparable rates in the ESD and EMR groups
(38.9 % vs. 56.7 %, P = 0.119). Similarly, the adenocarcinoma rate (median tumor size: 30 mm) was comparable
(5.6 % vs. 8.1 %, P = 0.607). The median tumor size was higher in malignant lesions (25 mm, range: 8 – 50)
compared to benign tumors (20 mm, range: 5 – 50) (P = 0.001). Although malignancy was not statistically associated with local recurrence,
adenocarcinomas presented a higher recurrence rate compared to adenomas with low-/high-grade
dysplasia (36.4 % vs. 15 %, P = 0.07). Notably, all recurrent adenocarcinomas were resected by piecemeal EMR. Considering
only patients with ≥ 12 months follow-up (n = 60), the local recurrence rate was 23.3 %,
with no statistically significant differences between the ESD (n = 5/11) and EMR groups
(n = 9/49) (45.5 % vs. 18.4 %, P = 0.055). However, overall tumor size was associated with recurrence (28 vs. 20 mm,
P = 0.008). All the long-term recurrences in the ESD group underwent the hybrid technique,
and most of tumors in the EMR group (88.8 %) were resected in piecemeal fashion. En-bloc
(10.6 % vs. 19.1 %, P = 0.202) and complete resection (12.1 % vs. 18.6 %, P = 0.394) were not statistically associated with lower recurrence rates in global
analysis or subgroup analysis.
Subgroup analysis
A subgroup analysis of 98 lesions larger or equal than 20 mm (29 ESD, 69 EMR) was
carried out. Thus, there were no differences in tumor size for both techniques (P = 0.711). The en-bloc (27.6 % vs. 17.4 %, P = 0.253) and complete resection rates (13.8 % vs. 11.6 %, P = 0.762) were higher in the ESD group and the recurrence rate was lower in the ESD
group (15.4 % vs. 23.4 %, P = 0.396), but these differences were not statistically significant. Notably, the
en-bloc resection rate decreased drastically in tumors ≥ 20 mm for EMR (17.4 % vs.
75 %, P < 0.001) but not for ESD (27.6 % vs. 37.5 %, P = 0.587). Similarly, the differences in perforation (13.8 % vs. 4.3 %, P = 0.097) and bleeding rates (10.3 % vs. 15.9 %, P = 0.470) were not significant. Tumor size ≥ 20 mm was associated with a higher local
recurrence (21.1 % vs. 6.5 %, P = 0.029).
Discussion
In the present series, we report on 166 NASDTs who underwent EMR or ESD with no differences
in en-bloc and complete resection rates as well as local recurrence between both techniques.
The en-bloc resection rates in the ESD group were comparable in large (≥ 20 mm) and
small lesions, while this outcome decreased significantly in lesions resected by EMR.
Overall, the en-bloc (10.6 % vs. 19.1 %) and R0 resection rates (12.1 % vs. 18.6 %)
were not statistically associated with lower recurrence, but the differences might
be clinically relevant. This study demonstrated that ESD may have a higher risk of
perforation compared to EMR.
There are few retrospective studies in the literature comparing ESD and EMR in duodenal
lesions, and most of them have been reported by Japanese authors. In many cases, they
included subepithelial and epithelial tumors all confounded [7]
[8]
[9] and considered patients with pedunculated lesions [10] or familial adenomatous polyposis [11]. Additionally, the criteria to choose ESD or EMR relies on the local experience
and tumor characteristics; therefore, the populations are quite heterogeneous within
different series. Furthermore, both groups may not be fully comparable, and the validity
of the published results might be difficult to interpret and apply in a Western setting.
In our series, we have performed ESD in large duodenal tumors, lesions suspected to
be malignant, or with non-lifting sign. Indeed, half of the overall resected tumors
were malignant and adenocarcinomas were confirmed to have a larger size and higher
long-term recurrence rates. Thus, it is important to point out the malignant potential
of these lesions, where complete endoscopic resection might avoid major surgery.
This study represents the second series of duodenal ESD in a Western setting where
the ESD is not as developed as in the Asian world; however, this is one of the largest
reports published to date. Naturally, en-bloc resection is mainly related to tumor
size. According to Basford et al. [5], we have performed a hybrid ESD in patients with larger lesions compared to those
resected by piecemeal EMR. The few studies to date regarding ESD in duodenal tumors
included selected patients with heterogeneous mean sizes (10 – 22 mm) [11]
[12] and locations. Indeed, Hoteya et al. [6] reported a higher complete resection rate in large NASDTs (mean size: 31.3 mm) compared
to small lesions (mean size: 11.6 mm) resected by ESD. Despite this procedure being
technically challenging, the en-bloc resection rates in Asian series have been reported
as above 90 % [13]. However, the included number of lesions is less than 10 in most of these reports
[7]
[9]
[11]
[12]
[14]
[15], and the impact of en-bloc resection in local recurrence is still unclear. In our
series, we considered 37 NASDTs who underwent ESD. The en-bloc resection rate was
lower (29.7 %) than previously described and lower compared to the EMR group, probably
because of the higher tumor size (median: 25 mm) and the hybrid ESD approach in 78.4 %
of cases. Similarly, the complete resection rate was higher in the EMR group and not
associated with recurrence.
Although overall recurrence following EMR has been described up to 37 % and associated
with tumor size [16], endoscopic resection of residual adenoma has been described to be successful [17]
[18]. Navaneethan et al. [19], in a systematic review and meta-analysis on duodenal EMR of 440 patients from 14
studies, concluded a recurrence rate of 15 %, with no association based on the initial
type of resection (en-bloc or piecemeal). Similarly, in our series, the recurrence
was 16.7 % in the EMR group, and no association of recurrence was observed with the
endoscopic technique, but only with a larger tumor size. We have also considered a
20-mm threshold as proposed by other authors [8]
[11], and the results were similar to the overall population for the main outcomes. Thus,
these outcomes were probably higher in the EMR group because of the smaller tumor
size and ESD learning curve, but this difference was much lower in local recurrence
rates.
Safety is a main concern in ESD and particularly in duodenal ESD where the benefits
of the technique remain unknown. Although one may argue that the higher risk of perforation
in this location is also because of the poor development of the technique at the early
beginning of the learning curve, most studies reporting on duodenal ESD have been
carried out by experienced endoscopists in referral centers. Indeed, the risk of perforation
is a major issue and occurs in 7 – 39 % of cases [6]
[20]
[21]. Conversely, other authors [19]
[22]
[23] reported a 0 – 2.8 % perforation rate by using the EMR technique, concluding that
this approach is safe even in large spreading lesions [24]. Recently, underwater EMR has been also described as a successful technique in small
adenomas with an en-bloc resection rate of 87 % [25]. Our overall perforation rate in lesions resected using a DualKnife was 16.2 %.
There were probably other factors that led to this complication, as the limited scope
maneuverability, fibrosis, non-lifting, and previous polypectomy, but they were difficult
to assess retrospectively. However, the perforation was identified during the procedure
and managed endoscopically in all but 1 patient who presented with a delayed perforation.
This very low rate of delayed perforations may be due to the careful inspection of
the mucosal defect after the procedure and systematic closure, sometimes using different
hoods and contrast under radiological control. Similarly, delayed bleeding was much
more frequent in lesions without prophylactic closure (18.9 % vs. 4.4 %).
This study has several limitations. Different factors such as tumor morphological
type and scope maneuverability may have influenced the choice between EMR and ESD.
The different operators during a wide inclusion period and the consequent learning
curve may have influenced the outcomes. Additionally, the retrospective design and
the lost to follow-up of 18 % may have underestimated the overall local recurrence.
In conclusion, ESD may be an alternative to EMR and surgery in selected NASDTs, such
as large duodenal tumors where EMR achieves lower en-bloc resection rates and the
local recurrence rate may be higher. However, this technique may have a higher risk
of perforations. Moreover, high-grade dysplasia and invasive cancer are frequent in
NASDTs and should be suspected in large lesions. Prospective studies including a higher
number of patients with similar tumor characteristics are needed to clarify the association
between the endoscopic features and local recurrence in this location, where it is
very difficult to assess the malignancy preoperatively.