This was a single-center, retrospective, cohort study, analyzing outpatient probe-based CLE database and histopathology reports from the Department of Endoscopy Unit, the First Affiliated hospital, College of Medicine, Zhejiang University between January 1, 2015 and December 31, 2017. The research protocol was reviewed and accepted by the research ethics committee of the First Affiliated Hospital, College of Medicine, Zhejiang University on June 13, 2018 (Reference Number: 750/2017). Patients enrolled were allocated into two groups according to their anesthesia type. Patients in sedated group received a combination of propofol-based sedation and lidocaine-based pharyngeal anesthesia, while patients in un-sedated group received lidocaine-based pharyngeal anesthesia only. Reports on juvenile patients (18 years old), or patients with advanced gastric cancer were excluded.
Anesthetic procedure
For all patients, after arrival, the electrocardiogram, non-invasive blood pressure and oxygen saturation were monitored. Lidocaine Hydrochloride mucilage (0.2 g/10 ml; Jiangsu Jichuan Pharmaceutical Co., Jiangsu, China) was administered orally 5 minutes before the beginning of CLE for pharyngeal anesthesia. For patients with sedated CLE, a 20-gauge cannula was placed in a vein in the forearm, and sedation was induced with propofol (0.5 g/50 ml; Xian Libang Pharmaceutical Co., Shanxi, China) 1.5-2 mg/kg intravenously. Sedation maintained was using 5–8 mg/kg/h propofol. Patients in un-sedated group received pharyngeal anesthesia only.
CLE procedure
CLE was conducted by two endoscopists who were with at least 5 years’ experience in performing diagnostic CLE. The procedure involved the use of a cellvizio confocal miniprobe (CM-4880, GastroFlex™ UHD, Mauna Kea, Paris, France), and a contrast agent (5 ml of 10% fluorescein sodium; Alcon Laboratories,Inc. Fort Worth, USA).
CLE for all participants were performed according to the standard protocol. After a mucosal lesion was visualized by WLE, a total of 5 ml 10% fluorescein sodium was administered intravenously. To obtain controlled CLE images, the probe was first gently contacted to normal mucosa around the lesion, ideally showing regular round or oval glands with homogeneous epithelial cells. The probe was subsequently moved to suspicious lesion to obtain CLE image, and following it, biopsies were obtained from the area. If the lesion was considered as high-grade intraepithelial neoplasia (HGIN) or EGC by the two endoscopists under CLE, the endoscopists would not get biopsy during the CLE procedure, but suggest patients for a endoscopic submucosal dissection (ESD) directly (6).
CLE criteria and histopathological criteria
The CLE criteria were based on the 2011 Miami classification (11). Four CLE diagnoses were given through evaluating architecture of glands, cells and micro-vessels as follows: (1) normal mucosa or benign inflammatory lesions; (2) atrophy and/or intestinal metaplasia (IM); (3) intraepithelial neoplasia (IN), including low-grade intraepithelial neoplasia (LGIN) and HGIN; (4) cancer.
The histopathological diagnostic criteria were based on the Updated Sydney System for classification and grading of gastritis (12) and the WHO classification of tumors (7). EGC is defined as carcinoma confined to the mucosa or submucosa, regardless of lymph node metastatic status. Surgical or endoscopic ablation was recommended for neoplastic lesions. Endoscopic resection was selected for HGIN or mucosal carcinoma. Surgery was used for incomplete endoscopic resection (6).
Data Collection
Two trained research assistants collected the following data: patients’ characteristics, CLE reports by endoscopists, pathological reports by pathologists. They also collected the information of patients who were strongly suspected for HGIN or EGC and further received ESD. These patients did not get biopsy during the CLE procedure, but had pathological results of ESD specimens. So the data of CLE reports and pathological diagnoses of ESD specimens were collected.
These data were entered into a Microsoft Excel 2013 (Microsoft Corporation, Washington, United States). Another two trained research assistants randomly extracted 10% of them to check the completeness, accuracy, and relevance of the information.
Statistical analysis
No specific power calculation was performed, and the sample size of this study was determined by the number of patients recruited across site. Statistical analysis was performed by using the software STATA/MP 15.0 (Stata Inc., TX, USA). Numerical data were presented as numbers (percentage). The Pearsonχ2 test was used to examine the significance of the association between two variables in a contingency table. Variables with a normal distribution were presented as mean ± standard deviation (SD), and compared using analysis of t test. A P value of 0.05 (two-sided) was considered statistically significant.
The primary outcome was the area under receiver operating characteristic (ROC) curve (AUC) of EGC and precancerous lesions. The nonparametric analysis was used for two-group comparison in AUC.
The second outcome was the misdiagnosis rate of CLE. For patients without biopsy during CLE and received an ESD subsequently, the misdiagnosis rate of CLE was analyzed according to their pathological results of ESD specimens.
Another second outcome was the comparison of sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV) and kappa (κ) value along with bionormal 95% confidence intervals(95% CI) between the two groups. Agreement was regarded as poor withκvalue below 0.4, good withκ value between 0.4 and 0.75, and excellent withκ value over 0.75.