Introduction
Sporadic duodenal adenomas are diagnosed in 0.1 %–0.4 % of upper gastrointestinal
endoscopies [1]
[2]. They involve the papilla Vateri (ampullary adenomas) or arise some distance from
the papilla (nonampullary adenomas). Duodenal adenomas are precancerous lesions harboring
the potential for progression to duodenal adenocarcinoma; resection is therefore recommended
[3]
[4]. Treatment options include endoscopic resection and surgical techniques. Endoscopic
resection appears potentially advantageous but complication risks have to be considered.
Endoscopic mucosal resection (EMR) was shown to be effective for nonampullary adenomas,
but a risk of delayed bleeding of up to 23.5 % has been reported for large lesions
[5]
[6]. Endoscopic papillectomy for ampullary adenomas has shown a high rate of curative
resection but also delayed bleeding rates of up to 16.8 % and pancreatitis rates of
up to 20 % [7]
[8].
Published studies mostly describe retrospective series with long inclusion periods
[8]
[9]
[10]. Several attempts have been made to reduce the rate of delayed complications, including
clipping or coagulation of visible vessels, various techniques for closure of the
resection site, and stenting of the pancreatic duct. However, prospective data on
duodenal EMR are scarce. The aim of this study was to prospectively evaluate the complication
rate of duodenal EMR for nonampullary adenomas and ampullary adenomas, and to assess
the efficacy of endoscopic measures that are performed widely in daily clinical practice
to prevent complications.
Methods
The study was conducted as a prospective, uncontrolled, observational, open-label,
single-center study in a German tertiary referral center (Department of Gastroenterology,
University Hospital Augsburg, Germany). The study was approved by the Institutional
Review Board of the University Hospital Augsburg, Germany (IRB number BKF-A-2019-11).
Patients were included from October 2015 to September 2019.
Inclusion criteria were: endoscopic diagnosis of sporadic duodenal adenoma (ampullary
and nonampullary adenomas), age ≥ 18 years, American Society of Anesthesiologists
(ASA) Score I–III, and written informed consent after patients received detailed information
about the EMR procedure (procedure, complication risks).
Exclusion criteria were: polyposis syndrome, biopsies showing adenocarcinoma, pretreated
lesions, EUS showing intraductal extension exceeding 5 mm into the common bile duct
or the pancreatic duct, concomitant malignant disease without curative treatment option.
In circumferential lesions, the additional risk of postinterventional stricture and
the alternative of surgical resection were discussed, and the treatment strategy was
based on the patient’s individual decision.
Outcomes
The primary outcome parameter was the rate of complications (intraprocedural perforation,
delayed bleeding, and delayed perforation for all lesions, as well as acute pancreatitis
after resection of ampullary adenomas). Secondary outcome parameters were procedural
characteristics (en bloc resection, complete resection, procedure time) and recurrences.
Diagnostic work-up
Endoscopy was performed with video gastroscopes (GIF-HQ190; Olympus Medical Systems,
Tokyo, Japan) or video duodenoscopes (TJF-Q180V; Olympus Medical Systems). Lesion
morphology was described according to the Paris classification [11]. Lesions were classified as ampullary adenomas when the papilla Vateri was involved
(ampullary adenomas and laterally spreading tumors involving the papilla [LST-P]).
Lesions located some distance from the papilla were classified as nonampullary adenomas.
Lesion diameter was estimated by reference to a polypectomy snare of known size. In
accordance with the published literature, lesions were classified as small (diameter
< 10 mm), large (10–29 mm) or giant (≥ 30 mm) [6]. In ampullary adenomas, endoscopic ultrasound was performed to rule out intraductal
extension into the pancreatic duct or the common bile duct. Biopsies were taken when
the macroscopic diagnosis of adenoma was doubtful or when malignancy was suspected.
However, most of the patients were referred for EMR and adenoma was confirmed by previous
biopsies. All patients had undergone colonoscopy during the previous 5 years, which
had ruled out polyposis syndrome.
EMR procedure
A video gastroscope (GIF-HQ190 or GIF-1TH190; Olympus) with a transparent hood (D-201-11804
or D-201-12704; Olympus), a video duodenoscope (TJF-Q180V; Olympus) or a combination
was used for EMR, depending on the lesion location and accessibility. EMR was performed
in a standardized way ([Fig. 1]). When lesion margins were unclear, narrow-band imaging or chromoendoscopy with
indigo carmine was used. For nonampullary adenomas and LST-Ps, submucosal injection
was performed routinely with a mixture of saline and epinephrine (1:100 000). For
papillectomy, submucosal injection was performed at the endoscopist’s discretion.
Fig. 1 Endoscopic mucosal resection procedure for a nonampullary adenoma (diameter 40 mm)
in the second part of the duodenum. a White-light endoscopy. b Piecemeal resection. c Coagulation of visible vessels. d Clipping of visible vessels.
Resection was performed with thin-wire snares with a diameter of 10 mm (POL1-B3-10-23-220-OL;
Medwork, Höchstadt, Germany), 15 mm (SD-990-15; Olympus) or 25 mm (SD-990–25; Olympus).
A VIO 300 D electrosurgical generator (ERBE Elektromedizin, Tübingen, Germany) was
used (Endo Cut Q mode, Effect 2 for cutting, and Forced Coag mode 60 W for coagulation).
Insufflation was performed using carbon dioxide. Sedation with midazolam, pethidine,
and propofol was administered by a second physician, under continuous cardiorespiratory
monitoring. Five procedures (lesions > 50 mm) were performed under general anesthesia.
EMR was performed by six endoscopists (A.P., A.E., G.B., S.G., T.W., H.M.) who had
each performed at least 200 colorectal and at least 20 duodenal EMRs prior to the
study.
When the course was uneventful, patients stayed in hospital for 48–120 hours after
EMR depending on the endoscopist’s decision. Anticoagulants except aspirin were stopped
before EMR and were restarted after 5–7 days, depending on the endoscopist’s decision
[12]. Warfarin was stopped 5 days before EMR until an international normalized ratio
of < 1.5 was reached; direct anticoagulants (dabigatran, rivaroxaban, edoxaban, apixaban)
were stopped 48–72 hours before EMR. Post-procedure, all patients received proton
pump inhibitors (pantoprazole 40 mg twice daily for 6 weeks and once daily for a further
6 weeks).
Complications
Complications were defined as bleeding, perforation, pancreatitis, stenosis or death.
Intraprocedural bleeding was noted as a complication when it led to premature termination
of EMR. Delayed bleeding was defined when hematemesis and/or melena were observed
after EMR. Bleeding was classified as major when the hemoglobin drop exceeded 2 g/L
[13]. When clinical bleeding signs occurred, endoscopy and endoscopic treatment were
performed ([Fig. 2]). Intraprocedural perforation was defined as an obvious defect in the muscularis
propria with an endoscopic view into the periduodenal space or the peritoneal cavity.
Delayed perforation was diagnosed when postinterventional imaging showed free air
or extravasation of contrast medium. Pancreatitis was defined as an elevation of the
serum lipase level at least threefold above the upper limit of the normal value in
combination with typical clinical symptoms. Pancreatitis was judged severe when the
Bedside Index of Severity in Acute Pancreatitis (BISAP) score was > 2. Stenosis was
considered as a complication when it was symptomatic. When follow-up endoscopy was
performed in our department, patients were asked for complications occurring after
discharge. Telephone calls were performed for all other patients.
Fig. 2 Complications after duodenal endoscopic mucosal resection (EMR). a Intraprocedural perforation during EMR for a large nonampullary adenoma. b Endoscopic closure with an over-the-scope clip. c Delayed bleeding 24 hours after EMR of a large nonampullary adenoma. d Endoscopic treatment with a hemoclip. e Clipping of visible vessels after uncomplicated EMR of a giant nonampullary adenoma.
f Delayed bleeding and perforation at the same area of the resection ulcer 12 hours
later.
Endoscopic procedures to prevent complications
All visible vessels in the resection ulcer were treated with hemoclips (HX-610-135S
or HX-610-090; Olympus) or with coagulation (Coagrasper FD-410 LR, using Soft Coag
mode, Effect 5, 40 W; Olympus) at the end of the procedure. Topical agents were applied
at the endoscopist’s discretion (Hemospray–Cook Medical, Bloomington, Indiana, USA;
or Purastat–3-DMatrix Europe, Caluire-et-Cuire, France). Hemospray was used during
the whole study period, whereas Purastat was used from January 2017. A scheduled second-look
endoscopy on the day after EMR was performed according to the endoscopist’s recommendation
but not routinely. After endoscopic papillectomy, pancreatic stenting was attempted
in all patients and rectal nonsteroidal anti-inflammatory drugs (indomethacin or diclofenac
100 mg) were administered [4]
[14].
Follow-up
Endoscopy was scheduled 3 months and 12 months after EMR and annually thereafter for
5 years. When residual or recurrent adenoma was detected, endoscopic re-treatment
was performed and 3-month surveillance intervals were recommended until the resection
site was recurrence free.
Statistical analysis
Depending on the number of groups, a Mann–Whitney rank sum test or a Kruskal–Wallis
one-way analysis of variance on ranks was used to compare numeric values. For the
comparison of categorical data, a chi-squared or Fisher’s exact test was employed,
depending on the expected frequency of the observations. P values of < 0.05 were considered statistically significant. Calculations were performed
using the software package Sigma Plot 13.0 (Systat Software, San Jose, California,
USA).
Results
Patient and lesion characteristics
Over a 4-year period, 119 patients were referred. A total of 110 patients presented
with one, 8 patients with two, and 1 patient with three lesions, resulting in a total
number of 129 lesions. After exclusion of four lesions with suspected cancer (two
ulcerated nonampullary adenomas and two ampullary adenomas with biliary obstruction)
and another three lesions with circumferential extension, EMR was performed for 122
lesions in 113 patients. Histopathological analysis did not confirm adenoma in the
EMR specimen in four resection specimens (follicular lymphoma in one and non-neoplastic
duodenal mucosa in three). The remaining 110 patients with 118 resected lesions were
enrolled in the study. Of the 118 lesions, 89 (75.4 %) were nonampullary adenomas
and 103 (87.3 %) were large or giant adenomas ([Table 1], [Fig. 3]).
Table 1
Patient and lesion characteristics.
|
Patient characteristics
|
n = 110
|
|
Age, median (range), years
|
68 (26–95)
|
|
Sex, male/female, n
|
57/53
|
|
ASA grade, 1/2/3, n
|
54/49/7
|
|
Antiplatelet medication, n (%)
|
21 (19.1)
|
|
Direct oral anticoagulation, n (%)
|
12 (10.9)
|
|
Warfarin, n (%)
|
2 (1.8)
|
|
Lesion characteristics
|
n = 118
|
|
Location, n (%)
|
|
|
6 (5.1)
|
|
|
99 (83.9)
|
|
|
13 (11.0)
|
|
Involvement of the papilla Vateri, n (%)
|
29 (24.6)
|
|
|
21 (17.8)
|
|
|
8 (6.8)
|
|
|
89 (75.4)
|
|
Diameter, median (range), mm
|
|
|
15 (4–70)
|
|
|
20 (8–60)
|
|
|
15 (4–70)
|
|
Lesion size by group, n (%)
|
|
|
15 (12.7)
|
|
|
68 (57.6)
|
|
|
35 (29.7)
|
|
Paris classification, n (%)
|
|
|
23 (19.5)
|
|
|
77 (65.3)
|
|
|
11 (9.3)
|
|
|
7 (5.9)
|
|
Histology
|
|
|
94/23/1
|
ASA, American Society of Anesthesiologists; LGIEN, low-grade intraepithelial neoplasia;
HGIEN, high-grade intraepithelial neoplasia.
Fig. 3 Patient flow through study and follow-up.
Procedure characteristics
Complete resection was achieved in 94.1 % (111/118). In six lesions > 50 mm, EMR was
scheduled as a two-stage procedure, and one resection was stopped because of a small
perforation, which was closed using an over-the-scope clip (OTSC; Ovesco, Tübingen,
Germany). Complete resection rates were comparable for small, large, and giant lesions
(P = 0.07). The en bloc resection rate was 39.0 % for all lesions and decreased significantly
in large and giant lesions (P < 0.001). Median procedure time was 41 minutes for all lesions. For giant lesions,
resection time was 105 minutes, which was significantly longer compared with lesions
< 30 mm (P < 0.001) ([Table 2]).
Table 2
Procedure characteristics.
|
Procedure characteristics
|
Lesions size, n (%)
|
|
|
All lesions
(n = 118)
|
Small < 10 mm
(n = 15)
|
Large 10–29 mm
(n = 68)
|
Giant ≥ 30 mm
(n = 35)
|
P
|
|
Resection, n (%)
|
|
|
46 (39.0)
|
11 (73.3)
|
33 (48.5)
|
2 (5.7)
|
< 0.001[1]
|
|
|
72 (61.0)
|
4 (26.7)
|
35 (51.5)
|
33 (94.3)
|
|
|
|
111 (94.1)
|
15 (100)
|
66 (97.1)
|
30 (85.7)
|
0.07
|
|
Procedure time, median (range), minutes
|
41 (9–207)
|
30 (18–70)
|
35 (9–96)
|
105 (35–207)
|
< 0.001[1]
|
|
Major complications[2]
|
18 (15.3)
|
0
|
8 (11.8)
|
10 (28.6)
|
0.02[1]
|
1 Significant difference between lesions –30 mm and < 30 mm.
2 Major bleeding, perforation or pancreatitis (Bedside Index of Severity in Acute Pancreatitis
score > 2).
Complications and mortality
Complications were analyzed according to lesion size (≥ 30 mm vs. < 30 mm) and the
type of adenoma (ampullary vs. nonampullary) ([Table 3], [Table 4]). Ampullary adenomas ≥ 30 mm (n = 10) included 7 LST-Ps and 3 ampullary adenomas,
whereas ampullary adenomas < 30 mm (n = 19) included 1 LST-P and 18 ampullary adenomas.
Endoscopic measures to prevent delayed bleeding were performed in 96/118 (81.4 %)
resections (hemoclips only n = 60, coagulation forceps only n = 11, topical agents
only n = 3). In 22 lesions a combination was used (clipping + coagulation forceps
n = 7, clipping + topical agent n = 5, coagulation forceps + topical agent n = 9,
clipping + coagulation forceps + topical agent n = 1).
Table 3
Complications and preventive measures.
|
Lesion type
|
All lesions (n = 118)
|
Nonampullary adenomas
|
Ampullary adenomas[1]
|
P
|
|
< 30 mm (n = 64)
|
≥ 30 mm (n = 25)
|
< 30 mm (n = 19)
|
≥ 30 mm (n = 10)
|
|
Preventive measures, n (%)
|
|
Prevention of delayed bleeding
|
|
|
73 (61.9)
|
53 (82.8)
|
10 (40.0)
|
5 (26.3)
|
5 (50.0)
|
< 0.001
|
|
|
5 (1–17)
|
5 (1–17)
|
5 (1–10)
|
1 (1–4)
|
8 (1–15)
|
|
|
|
28 (23.7)
|
3 (4.7)
|
12 (48.0)
|
7 (36.8)
|
6 (60.0)
|
< 0.001
|
|
|
18 (15.3)
|
1 (1.6)
|
11 (44.0)
|
3 (15.8)
|
6 (60.0)
|
< 0.001
|
|
|
22 (18.6)
|
3 (4.7)
|
10 (40.0)
|
10 (52.6)
|
6 (60.0)
|
< 0.001
|
|
|
22 (18.6)
|
10 (15.6)
|
2 (8.0)
|
2 (10.5)
|
0
|
< 0.001
|
|
Second-look endoscopy
|
25 (21.2)
|
3 (4.7)
|
13 (52.0)
|
2 (10.5)
|
7 (70.0)
|
< 0.001
|
|
|
15
|
3
|
6
|
2
|
4
|
|
|
|
10
|
0
|
7
|
0
|
3
|
0.22
|
|
Prevention of pancreatico-biliary complications
|
|
|
n.a.
|
n.a.
|
n.a.
|
14 (73.7)
|
4 (40.0)
|
0.11
|
|
|
n.a.
|
n.a.
|
n.a.
|
11 (57.9)
|
3 (30.0)
|
> 0.99
|
|
Delayed bleeding, n (%)
|
22 (18.6)
|
2 (3.1)
|
10 (40.0)
|
6 (31.6)
|
4 (40.0)
|
< 0.001
|
|
Major
|
16 (13.6)
|
1 (1.6)
|
7 (28.0)
|
6 (31.6)
|
2 (20.0)
|
|
|
|
8 (6.8)
|
0
|
4 (16.0)
|
3 (15.8)
|
1 (10.0)
|
0.77
|
|
Minor
|
6 (5.1)
|
1 (1.6)
|
3 (12.0)
|
0
|
2 (20.0)
|
0.28
|
|
Perforation
|
5 (4.2)
|
1 (1.6)
|
2 (8.0)
|
0
|
2 (20.0)
|
0.02
|
|
Intraprocedural
|
3 (2.5)
|
1 (1.6)
|
1 (4.0)
|
0
|
1 (10.0)
|
|
|
Delayed
|
2 (1.7)
|
0
|
1 (4.0)
|
0
|
1 (10.0)
|
0.10
|
|
Pancreatitis
|
n.a.
|
n.a.
|
n.a.
|
3 (15.8)
|
2 (20.0)
|
> 0.99
|
|
Duodenal stricture
|
1 (0.8)
|
0
|
0
|
0
|
1 (10.0)
|
0.01
|
|
Mortality
|
2 (1.7)
|
0
|
1 (4.0)
|
0
|
1 (10.0)
|
0.10
|
1 Ampullary adenomas and laterally spreading tumors involving the papilla.
2 Hemospray (Cook Medical, Bloomington, Indiana, USA) in 12 cases and Purastat (3-DMatrix
Europe, Caluire-et-Cuire, France) in 6.
Table 4
Complications after endoscopic mucosal resection of ampullary and nonampullary adenomas.
|
Lesions type
|
All lesions (n = 118)
|
Nonampullary (n = 89)
|
Ampullary[1] (n = 29)
|
P
|
|
Delayed bleeding, n (%)
|
22 (18.6)
|
12 (13.5)
|
10 (34.5)
|
0.03
|
|
Major
|
16 (13.6)
|
8 (9.0)
|
8 (27.6)
|
0.03
|
|
|
8 (6.8)
|
4 (4.5)
|
4 (13.8)
|
0.19
|
|
Minor
|
6 (5.1)
|
4 (4.5)
|
2 (6.9)
|
0.98
|
|
Perforation
|
5 (4.2)
|
3 (3.4)
|
2 (6.9)
|
0.77
|
|
Intraprocedural
|
3 (2.5)
|
2 (2.2
|
1 (3.4)
|
0.75
|
|
Delayed
|
2 (1.7)
|
1 (1.1)
|
1 (3.4)
|
0.99
|
|
Pancreatitis
|
n.a.
|
n.a.
|
5 (17.2)
|
|
|
Duodenal stricture
|
1 (0.8)
|
0
|
1 (3.4)
|
0.55
|
|
Any complication
|
27 (22.9)
|
14 (15.7)
|
13 (44.8)
|
0.003
|
|
Major complication
[2]
|
18 (15.3)
|
10 (11.2)
|
8 (27.6)
|
0.07
|
|
Mortality
|
2 (1.7)
|
1 (1.1)
|
1 (3.4)
|
0.99
|
1 Ampullary adenomas and laterally spreading tumors involving the papilla.
2 Major bleeding, perforation or severe pancreatitis (Bedside Index of Severity in
Acute Pancreatitis score > 2).
A total of 25 patients underwent a scheduled second-look endoscopy on the day after
EMR. In 10 of them, visible vessels were re-treated with clips or coagulation. In
three additional patients, second look had been planned but delayed bleeding occurred
during the night after EMR. Delayed bleeding was noted in 18.6 % of lesions (22/118).
A total of 16 bleedings were major (13.6 %) and 8 of them needed transfusions. Delayed
bleeding was significantly associated with lesion diameter and with involvement of
the papilla but was not reduced by preventive endoscopic measures ([Table 3], [Table 4]; see also Table 5s in the online-only supplementary material). Endoscopic treatment with hemoclips or
coagulation forceps was successful in all minor bleeding episodes. Four of the 16
lesions with major bleeding needed additional embolization of the gastroduodenal artery.
Clinical characteristics of the delayed bleeding cases are summarized in Table 6s.
Intraprocedural perforation was noted in three patients and was managed successfully
with OTSC. EMR was completed at the same time in two of these patients, and the remaining
patient underwent removal of the clip and completion of the resection with EMR 2 months
later. Pancreatic duct stenting was possible in 18/29 patients after endoscopic papillectomy
(62.1 %). Pancreatitis was diagnosed in 5/29 (17.2 %). The course was mild in all
of these patients (BISAP 1 n = 1, BISAP 2 n = 4). Pancreatitis was not associated
with stenting of the pancreatic duct (Table 5s).
Delayed perforation was seen in a 68-year-old patient who presented with major bleeding
12 hours after EMR of a 60-mm nonampullary adenoma in the third part of the duodenum.
Urgent endoscopy showed massive bleeding and a 3-mm perforation in the resection ulcer
([Fig. 2]). Perforation was closed with clips but bleeding could not be stopped. Surgery was
discussed but not performed because of severe cardiopulmonary comorbidity. Radiologic
embolization of the gastroduodenal artery was performed. However, progressive multiorgan
failure developed and the patient died 2 days later.
Duodenal stricture was observed 3 weeks after circumferential EMR and endoscopic papillectomy
of a 60-mm LST-P in a 72-year-old woman. Over a 4-week period, six sessions of endoscopic
balloon dilation were performed. During the final dilation, perforation in the second
part of the duodenum was seen. Computed tomography (CT) scan showed a small extravasation
of contrast medium. Abdominal examination was unremarkable and a conservative approach
with antibiotics and nil-by-mouth was started after interdisciplinary discussion.
CT scan the following day showed progressive periduodenal fluid collections and the
patient underwent surgery. However, she died because of intractable multiorgan failure
2 days later.
In summary, complications were noted in 22.9 % and major complications occurred in
15.3 % of all lesions. Complication rates differed significantly between ampullary
and nonampullary adenomas (P = 0.003). Major complications increased significantly with lesion diameter (P = 0.02) ([Table 2], [Table 4]).
Two out of 118 resections showed a fatal outcome, resulting in a procedure-related
mortality of 1.7 %.
We performed a multivariate analysis with any complication as a composite end point.
Patient characteristics (sex, age, ASA score), lesion characteristics (ampullary vs.
nonampullary, diameter), and procedure characteristics (en bloc vs. piecemeal, procedure
time) were analyzed. Lesion diameter (P = 0.002) and piecemeal resection (P < 0.001) were shown to be statistically predictive. A further regression analysis
confirmed these parameters as being independent (P = 0.29).
Follow-up and recurrences
First follow-up endoscopy was performed for 93/118 lesions (78.8 %). Residual or recurrent
adenoma was diagnosed in 19/93 (20.4 %) and underwent endoscopic re-treatment ([Fig. 3]). At first follow-up, recurrence rate for ampullary and nonampullary adenomas were
37.5 % (9/24) and 14.5 % (10/69), respectively (P = 0.04).
Discussion
EMR of duodenal adenomas has been shown to be effective in several studies over the
past decade [5]
[6]
[7]
[8]
[9]
[10]
[15]
[16]
[17]. Substantial complication rates have been described especially after resection of
large and giant lesions. Fanning et al. reported a 26.3 % rate of major complications
after EMR of giant nonampullary adenomas; the most frequent complication was delayed
bleeding [16]. Another study reported delayed bleeding in 25 % when LST-Ps with a median diameter
of 35 mm were resected [6]. Lepilliez et al. described a potential decrease in delayed bleeding in a retrospective
and nonrandomized study when clipping and argon plasma coagulation were used after
EMR; the delayed bleeding rate was 22 % in lesions without preventive measures compared
with 0 % in lesions that were treated with clips or argon plasma coagulation [5].
Most of the studies on duodenal EMR are retrospective. Endoscopic measures to prevent
delayed complications (especially delayed bleeding) are used widely in daily clinical
practice but prospective data regarding their efficacy are lacking. Therefore, we
initiated a prospective study and included all ampullary and nonampullary adenomas
referred for EMR in a German tertiary referral center. Systematic endoscopic measures
for bleeding prevention were performed including clipping or coagulation of visible
vessels, with the additional option of applying topical agents (i. e. Hemospray or
Purastat). Lesions were predominantly large or giant adenomas. EMR was highly effective
regarding complete resection (94.1 %). As expected, en bloc resection decreased significantly
with increasing lesion diameter. For lesions ≥ 30 mm, en bloc resection was possible
in two polypoid ampullary adenomas but was impossible for nonampullary adenomas. Surprisingly,
even in small lesions < 10 mm, EMR had to be performed piecemeal in 26.3 %. Retrospective
series have reported R0 resections rates of 68 %–82.2 % in lesions with a mean diameter
of 10 mm [10]
[18].
Endoscopic measures to prevent delayed bleeding were performed in 81.4 % after EMR.
In giant lesions, the rate was 94.3 % and only two lesions were not treated prophylactically.
Clipping or coagulation of visible vessels was performed in 87.6 % of nonampullary
adenomas and in 79.3 % of ampullary adenomas. In giant lesions clipping of all visible
vessels was more difficult or even impossible compared with smaller resection ulcers.
This might explain why less clipping and more coagulation was performed in giant lesions.
In smaller resection ulcers, clipping was predominantly used. Despite these preventive
measures, we observed delayed bleeding in 18.6 % of all resections, with a significant
increase in giant lesions. For giant lesions, the total rate of delayed bleeding and
the rate of major bleeding was 40.0 % and 25.7 %, respectively. Delayed bleeding after
EMR of lesions < 30 mm could be controlled endoscopically in all cases, whereas 25.0 %
of delayed bleeding in giant lesions required additional radiological embolization
of the gastroduodenal artery. The delayed bleeding rate is similar to or even higher
than that reported in the retrospective studies mentioned above [6]
[16]. As the resection procedure was comparable, the preventive measures performed in
our study were obviously insufficient to reduce delayed bleeding.
Delayed perforation is another hazardous and potentially fatal complication occurring
in up to 1 % after duodenal EMR [6]
[19]. The suspected reason for delayed perforation is intraprocedural coagulation of
the muscle layer. A case report has described another potential mechanism whereby
delayed perforation was caused by an endoscopic clip in the resection ulcer [20]. In our study, we observed two delayed perforations with fatal outcomes after EMR
of giant duodenal adenomas.
The risk of residual or recurrent neoplasia is a known disadvantage after piecemeal
resection of gastrointestinal neoplasia. In our study 20.4 % of lesions showed residual
adenoma at the first follow-up endoscopy. The recurrence rate is similar to published
data, which report rates ranging from 14.4 % up to 29 % [6]
[8]
[21]
[22]. Over recent years, endoscopic submucosal dissection (ESD) has gained acceptance
in the treatment of esophageal, gastric, and colorectal neoplasia; however, the role
of ESD for duodenal neoplasia remains controversial. A high perforation rate has been
reported in initial studies. A small study from Korea described perforations in 5/14
patients, two of whom required surgery [23]. Hoteya et al. reported a 39 % rate of intraprocedural perforations after ESD in
41 large nonampullary adenomas (mean diameter 26 mm) [24]. Three patients required conversion to surgery. The rate of delayed bleeding was
18.4 %, which is similar to the rate for EMR. A large retrospective study on 146 EMRs
(mean lesions diameter 9.8 mm) and 174 ESDs (mean lesions diameter 27.4 mm) demonstrated
similar R0 resection rates (82.2 % vs. 85.1 %), but a significant increased perforation
rate after ESD (15.5 % vs. 0.68 %) [18]. Data on duodenal ESD are restricted to Asian expert centers and its role for Western
endoscopists is not yet defined.
New techniques for the prevention of delayed complications after duodenal endoscopic
resection have been introduced recently. Covering the resection ulcer with polyglycolic
acid sheets in combination with fibrin glue was reported to prevent delayed perforation
[25]. Asian authors have proposed different strategies to close the resection ulcer especially
after ESD. Analysis of delayed bleeding after duodenal ESD identified closure of the
resection ulcer as an independent risk factor [26]. Kato et al. reported on 173 ESDs and confirmed the efficacy of endoscopic closure
using different techniques (clips, endoloop/clips technique, string clip suturing
technique) [27]. The rate of delayed adverse events was significantly reduced to 1.7 % after complete
closure of the defect compared with 25 % after incomplete closure, and 15.6 % without
closure. Another Japanese group described the use of OTSCs after duodenal ESD, and
reported delayed bleeding in 6.4 %, delayed perforation in 2.4 %, and conversion to
surgery in 4 % [28]. Underwater EMR and additional closure of the resection ulcer with clips or line-assisted
clipping showed promising results without delayed complications in a recent small
study in nonampullary adenomas; however, included lesions were small with a mean diameter
of 12 mm [29]. Laparoscopic-endoscopic cooperative surgery has also been reported as another treatment
option to close the resection site [30].
Endoscopic or laparoscopic closure techniques have shown promising results, but data
are widely restricted to duodenal ESD. It appears obvious to transfer these techniques
to EMR but a 20 % risk of residual adenoma has to be taken into account. Line-assisted
closure has been reported for wide-field EMR but long-term follow-up and larger series
are lacking [31]. Further limitations of the closure techniques are the reduced success rate in very
large defects and the impossibility to close defects after endoscopic papillectomy.
Further data are needed for endoscopic closure after duodenal EMR.
EMR is offered as a minimally invasive procedure with less morbidity and mortality
compared with surgical approaches. Klein et al. described a shorter hospital stay
and lower costs for endoscopic resection techniques when comparing 102 EMRs of large
adenomas (ampullary and nonampullary) with alternative hypothetical surgical resections
[32].
Traditionally, pancreaticoduodenectomy, with its substantial complication risk, was
the surgical alternative treatment option. Over past decades, less invasive surgical
techniques have been introduced, such as local ampullary resections for ampullary
adenomas or pancreas-preserving partial duodenectomies (PPPD) for nonampullary adenomas
[33]. A retrospective comparative study included 91 EMRs and 30 PPPDs [22]. En bloc resection was achieved in 53 % and 100 %, respectively. Early complications
including five intraprocedural perforations were observed after EMR, while the surgical
group showed no early but eight delayed complications, including gastroparesis, pulmonary
embolism, and pancreatic fistula. Recurrence rates were 32 % after EMR and 0 % after
PPPD [22].
Our study demonstrates substantial complication risks after EMR of giant duodenal
adenomas despite intensive preventive measures. To our knowledge, this study is one
of the largest prospective series on duodenal EMR. The data represent the efficacy
and the complication risk of duodenal EMR in daily clinical practice in a tertiary
referral center. Limitations of the study are the nonrandomized design and no control
group, the single-center setting, and the inclusion of both ampullary and nonampullary
adenomas. Another limitation is that preventive measures after EMR and recommendations
on second-look endoscopy were not fully standardized.
In conclusion, EMR shows a high success rate but also a substantial complication risk
especially in large and giant lesions and in lesions involving the papilla. Measures
to prevent delayed major complications need to be improved urgently. ESD with closure
of the resection ulcer seems to be promising in expert hands. However, widespread
use outside Asia cannot be expected in the near future. The treatment of high-risk
duodenal adenomas is challenging and requires endoscopic, surgical, and radiological
expertise in tertiary referral centers. Treatment decisions should follow an interdisciplinary
approach, and treatment options (endoscopic vs. surgical) should be discussed with
the patient, including efficacy and complication risk but also the risk of recurrence.
