Patients
A total of 2517 patients treated with endoscopic resection at Chungnam National University Hospital (CNUH) from January 2011 to December 2015 were screened (Fig. 1). Among these, 857 patients were excluded from the study because they were not diagnosed with gastric dysplasia at the initial endoscopic biopsy (e.g., adenocarcinoma, neuroendocrine tumor, hyperplastic polyp, lipoma, intestinal metaplasia). In addition, 755 patients were excluded because they were not followed up for more than 1 year after endoscopic resection. As a result, a total of 905 patients diagnosed with LGD or HGD by initial endoscopic biopsy were included in the study. All patients underwent endoscopic resection through EMR or ESD, regardless of the lesion size. EMR and ESD were performed in 66.2% (n=599) and 33.8% (n=306) of cases, respectively. Data regarding comorbidities (i.e., hypertension, diabetes mellitus, chronic obstructive pulmonary disease, or chronic kidney disease), smoking history, and alcohol consumption history were collected by medical record review.
Evaluation of endoscopic features
The surface gross type (i.e., elevated, flat, depressed, or nodularity), color change (i.e., whitish or redness), size, location, atrophic gastritis, and intestinal metaplasia were determined by a review of endoscopic recordings and photographs. Lesion location was classified by dividing the stomach in three equal sections: upper (fundus and upper body), middle (middle and lower body), and lower (angle and antrum). Lesion size was classified as 2.0 cm or >2.0 cm.
EMR/ESD techniques
The approach to endoscopic resection for gastric dysplasia was determined by the endoscopists, in consideration of the lesion characteristics, such as size, shape, and location.
During the procedure, midazolam or diazepam was administered intravenously for sedation, and cardiorespiratory functions were monitored. A dual-channel gastroduodenoscope (GIF-ITQ 260M; Olympus, Japan) was used for EMR and a single-channel gastroduodenoscope (GIF-H260; Olympus) was used for ESD.
Before endoscopic resection, 0.1% indigo carmine solution was applied to the lesion to identify its location and margins. After confirming the lesion, areas of the normal mucosa 1-2 mm away from the margin of the lesion were marked with a fixed flexible snare (KachuTechnology Co., Ltd. Korea) or electrosurgical generator (ERBE VIO300D VIO 300D; Erbe, Tübingen, Germany). Then, a saline solution containing diluted epinephrine (1:10,000) was injected into the submucosal layer of the lesion using needle forceps, and the mucosal layer was completely floated from the muscular layer of the lesion. These procedures were the same for both EMR and ESD, but the subsequent steps differed. For EMR, the lesion was pulled using grasping forceps, and en bloc resection was performed by using a snare loop to cover all marked regions at once; if en bloc resection was impossible, a piecemeal resection was performed (Fig. 2). For ESD, an incision was made with a fixed flexed snare outside the marker, and a circumferential incision was performed using an IT knife (single-use electrosurgical knife KD-61 1L; Olympus). Then, the submucosal layer was dissected until the lesion was completely resected using an IT knife.
During or after endoscopic resection, endoscopic hemostasis was performed for any bleeding or exposed vessels using an IT knife or hemostatic forceps (FD-410LR; Olympus) (Fig. 3). Fig. 2 shows the process of en bloc resection of a gastric dysplasia by EMR, and Fig. 3 shows the process of en bloc resection of a gastric dysplasia (approximately 2 cm in size) by ESD.
Definitions
En bloc resection was defined as resection of a lesion in one piece (as opposed to piecemeal resection) [3]. Complete resection was defined as R0 resection in which the resected lesion was pathologically free of dysplasia in the lateral and deep margins. After the procedure, patients were evaluated for complications such as bleeding, perforation, and pyloric stenosis. Delayed bleeding was defined as hematemesis or melena, with a decrease in the hemoglobin level of more than 2 g/dl, requiring endoscopic hemostasis after endoscopic resection [3,4]. Perforation was defined as direct perforation of the mesenteric fat during endoscopic procedures or free air on abdominal x-ray examination after endoscopic resection [3]. Pyloric stenosis was defined as the occurrence of symptoms such as dyspepsia due to a narrowing of the pyloric ring precluding passage of an endoscopic fiber after endoscopic resection [5].
Histological analysis
All specimens collected for histological analysis were immediately fixed into paraffin blocks using a neutral buffer with 10% formalin. Paraffin blocks were cut at 2-mm intervals and stained with hematoxylin and eosin to confirm complete resection [2]. The presence of Helicobacter pylori was evaluated by the rapid urease test (CLO1 test; Kimberly-Clark, UT) and histologic examination (Wright-Giemsa stain). If any test results were positive, H. pylori infection was considered as present. Histologic diagnosis was made by experienced pathologists, in accordance with the Vienna classification of gastric epithelial dysplasia [6].
Follow-up and confirmation of local recurrence
All patients had their first follow-up endoscopy at 3 or 6 months after endoscopic resection, and annually thereafter. During the follow-up endoscopy, a biopsy was performed when an abnormality was found, such as an overgrowth of the mucosa at the scar of the previous endoscopic resection or a change in color. The results of the histopathologic examination were defined as local recurrence in cases under categories 3-5 of the Vienna classification (i.e., LGD, HGD, or adenocarcinoma).
Statistical analysis
All statistical analyses were performed using IBM SPSS Statistics version 21 (IBM Corp., Armonk, NY). Statistical significance was set at p<0.05. Group differences (EMR vs. ESD) in baseline characteristics were evaluated using Pearson’s chi-squared test for categorical variables. A logistic regression model was used to analyze factors affecting local recurrence. Significant univariate factors (p<0.05) were examined using a multivariate Cox proportional hazards regression model to identify the independent factors associated with local recurrence. Odds ratios (OR) and 95% confidence intervals (CI) were calculated to estimate the relative risks of local recurrence. The Kaplan-Meier method was used to analyze the cumulative rate of local recurrence.