This study was a multicenter retrospective cohort study. Medical records dating from January 2007 to April 2018 from Incheon St. Mary’s Hospital, Seoul St. Mary’s Hospital, and Bucheon St. Mary’s Hospital, The Catholic University of Korea were reviewed. Approval for the study was obtained from The Institutional Review Board of The Catholic University of Korea.

SNADTs that were removed endoscopically during this period were included in the study. Ampullary duodenal tumors and SNADTs other than adenomas and adenocarcinomas were excluded. Furthermore, patients who underwent additional endoscopic and/or surgical procedures for the management of complications, such as bleeding or perforation, during and/or immediately after the procedure were also excluded. In order to compare the data regarding these complications, patients were divided into two: the immediate clipping group (ICG) and the no clipping group (NCG).

Endoscopic mucosal resection (EMR) and endoscopic submucosal dissection (ESD) were performed using either twin-channel multi-bending gastroscopes (GIF-2TQ260M; Olympus, Tokyo, Japan) or conventional gastroscopes (GIF-H260 or GIF-HQ290; Olympus) along with a CO₂ insufflation system under conscious sedation. For EMR, a mixture of normal saline and diluted epinephrine (1:200,000) was injected into the submucosal layer. The lesions were then removed using snares and an electrosurgical unit (VIO300D, Endocut Q mode, effect 2; ERBE, Tübingen, Germany). After removal of the lesion, proton pump inhibitors (PPIs) were administered intravenously for 2 days followed by oral administration for 4 weeks.

For ESD, chromoendoscopy was performed using acetic acid (1.5%) and indigo carmine (0.2%). Dot markings were made 5 mm from the tumor margin using argon plasma coagulation. The same mixture used in EMR was also injected into the submucosal layer. Afterwards, an initial incision was made outside the marks using either a dual knife (KD-655; Olympus) or a hook knife (KD-620LR; Olympus). A knife was then inserted into the initial incision and an electrical current was applied using an electrosurgical unit (VIO300D, Endocut I mode, effect 2; ERBE) to complete a circumferential mucosal incision around the lesion. The submucosal layer was then dissected using either a dual knife or a hook knife. After removal, PPIs were administered in the same way described in the EMR procedure.

Whether clipping was used to close the defect was decided by the endoscopists’ preferences. After ER, the mucosal defect was closed using clips (HX-610-090L; Olympus Medical Systems Co., Tokyo, Japan) delivered by a clip-fixing device (HX110UR; Olympus Medical Systems). Complete closure was defined as an invisible mucosal defect after clipping.

Delayed complications included delayed bleeding and delayed perforations; the incidence rates of these complications were considered as the primary outcomes. Delayed bleeding was defined as the presence of melena and/or hematemesis occurring at least 24 hours after procedure, or by duodenoscopic documentation of visible hemorrhages from the mucosal defect that eventually required endoscopic intervention. Delayed perforation was defined as the presence of intra-peritoneal free air on plain chest X-ray and/or on abdominal computed tomography at least 24 hours after the procedure. Factors associated with delayed complications, such as age, sex, comorbidity, anticoagulant use, and tumor characteristics (tumor location, tumor size, and tumor pathology), along with the treatment modalities used (EMR vs. ESD) were then analyzed. Secondary outcomes, which include the duration of hospitalization, transfusion requirements, the need for surgery, and local recurrences, were compared between ICG and NCG.

Dependent variables were determined using descriptive data. Continuous data were expressed as means±standard deviation. The statistical differences between the baseline characteristics of patients from ICG and NCG were assessed using the student t-test for continuous variables, and the chi-square test or the Fisher exact test for categorical variables. Factors associated with the occurrence of delayed complications were also analyzed. SPSS version 13.0 (SPSS Inc., Chicago, IL, USA) was used for statistical analyses.

ER was performed by a total of 20 endoscopists with more than 3 years of clinical experience each. A total of 270 patients underwent ER (EMR or ESD) for SNADT from January 2007 to April 2018. Forty six patients with Brunner’s gland hyperplasia, 36 patients with neuroendocrine tumors, 25 patients with inflammatory polyps, 24 patients with hyperplastic polyps, 12 patients with hamartomatous polyps, 5 patients with lipoma, 4 patients with heterotopic gastric mucosa, 3 patients with Brunner’s gland adenoma, 3 patients with pyloric gland adenoma, 2 patients with ectopic pancreas, 2 patients with ganglioneuroma, 1 patient with leiomyoma, 1 patient with duodenal cyst, and 1 patient with gastrointestinal stromal tumor were excluded. Additionally, 1 patient that developed perforation during the procedure, and 5 patients that experienced hemorrhage immediately after the procedure, were also excluded. Because Brunner’s glands originate from the submucosal layer, and that pyloric gland adenomas differ in nature from adenomas and adenocarcinomas of the duodenum, patients with these conditions were excluded. After the process of exclusion, 99 of the 270 original patients were analyzed in this study (Fig. 1). There were no differences in baseline characteristics such as age, male to female ratio, presence of comorbidities, social histories such as alcohol consumption and smoking, and current drug histories between the 2 groups (Table 1). An example of endoscopic clipping after EMR is presented in Fig. 2.

Complete closure was attained in every case. However, closure was not successful in 1 case where immediate perforation occurred and in 1 case where bleeding occurred immediately after closure (Fig. 1). Endoscopic clipping was more commonly performed on lesions located in the 2nd portion of the duodenum (p=0.01; Table 2) and on adenomas presenting with low grade dysplasia histologically (p=0.01; Table 2). There were no differences in the removal technique used, tumor size, circumferential location, histologic type, and the type of resection performed, whether en bloc resection or complete, between the 2 groups (Table 2).

Among the 99 patients who underwent ER, delayed complications occurred in 13 patients (13.1%). Delayed bleeding occurred in 9 patients (9.1%) and delayed perforation occurred in 4 patients (4.0%). Delayed bleeding occurred in 1 patient from ICG and in 8 patients from NCG (1.9% vs. 17.0%, p=0.01; Table 3). However, delayed perforation occurred in 1 patient from ICG and in 3 patients from NCG (1.9% vs. 6.4%, p=0.34; Table 3). Patients from ICG showed a tendency to have shorter lengths of hospital stay than those from NCG (3.6±2.5 vs. 4.7±3.3, p=0.06; Table 3). There were no differences in the need for transfusion and in the need for surgery between the 2 groups (Table 3).

Univariate analysis showed that the delayed complications experienced by the patients were related to ER methods, endoscopic clipping, and histologic tumor types (Table 4). In multivariate analysis, ER methods (odds ratio [OR], 13.0; 95% confidence interval [CI], 1.7 to 102.1; p=0.025) and endoscopic clipping (OR, 8.7; 95% CI, 1.6 to 48.7; p=0.014) were independently associated with the occurrence of delayed complications (Table 5).

All 9 patients with delayed bleeding were managed endoscopically and all 4 patients with delayed perforation were managed by laparoscopic simple closure successfully. There were no deaths among patients with delayed complications and the longest length of hospitalization was 19 days.