Objective Study on Tongue Diagnosis of Chronic Atrophic Gastritis

Overview

Chronic atrophic gastritis with intestinal metaplasia or dysplasia is called precancerous lesions of gastric cancer ( PLGC ). It is an important stage in the transformation of normal mucosa to gastric cancer and is significantly associated with the risk of gastric cancer. Early identification of PLGC high-risk groups is the focus of prevention and treatment of gastric cancer. Gastroscopy and pathological examination are the key means to diagnose PLGC. However, due to the invasiveness, high cost, strong professional operability and traumatic pathological sampling, the application of gastroscopy is greatly limited, and the predictive significance of serum markers for the occurrence and prognosis of PLGC is limited. The current PLGC monitoring methods are not targeted enough, and there is still a lack of effective risk prediction models for PLGC. New screening methods are needed to improve the early diagnosis rate of PLGC. Tongue diagnosis is simple, convenient, easy and inexpensive. It can dynamically monitor the risk of PLGC in patients with chronic atrophic gastritis in early real time, and timely formulate individualized intervention measures, which is of great significance for improving the prognosis of patients, controlling medical expenses and truncating disease progression. Based on the objectification of tongue diagnosis, it may become an important method for screening PLGC from the perspective of macroscopic tongue image characteristics and microscopic tongue coating flora.

1.1 Research design Study type : Prospective cross-sectional study and clinical prediction model study. Research hypothesis : Taking PLGC as the observation object, the tongue image and tongue coating flora of PLGC patients were collected objectively and standardized, and the PLGC tongue image research database was established. Based on the characteristics of tongue image and tongue coating flora, the clinical prediction model of PLGC was constructed, so as to realize the real-time dynamic prediction of the risk of PLGC in CAG patients through tongue image, reduce invalid endoscopic screening and reduce the cost of repeated screening. Research Methods : ( 1 ) Clinical data survey : The questionnaire was used to collect patient information. The baseline data included personal basic information, lifestyle, history of HP infection ( previous HP positive and current HP positive ), history of digestive tract disease and family history of cancer. The medical records also include current medical history, past medical history, personal life history, gastrointestinal adverse symptoms, etc. ; ( 2 ) Electronic endoscopy iodine staining : Olympus electronic endoscopy was used for iodine staining. If any positive or suspicious lesion was found in the gastric mucosa, the corresponding area was bitten and sent for pathological examination. If no abnormality was found by endoscopy, biopsy was not taken. Gastroscopic manifestations ( erosion, ulcer, edema, congestion, erythema, bile reflux, HP infection ) were recorded. ( 3 ) Pathological examination : After the biopsy specimens were processed, the pathological HE staining was performed, and the pathological diagnosis was observed and performed. The histopathological lesions ( chronic inflammation, active inflammation, intestinal metaplasia, atrophy, degree and range of dysplasia ) were recorded. ( 4 ) Tongue image information collection : 1 Standardized acquisition equipment : The tongue image was collected and automatically analyzed by the tongue surface diagnosis information acquisition system ( DS01-B, Shanghai Daosheng Medical Technology Co., Ltd., Shanghai Machinery Injection 20202200062 ), and the image information was digitized and stored. 2 tongue extension method : Before shooting, the subjects were trained to extend their tongues. The subjects were seated at the end, and the mouth was opened as much as possible when the tongue was extended. The tongue body was relaxed, the tongue surface was flattened, and the tip of the tongue was naturally drooping. It is not appropriate to extend the tongue too hard, and the time of extending the tongue should be minimized to avoid the fatigue of the tongue body affecting the stability of the tongue. 3 Note : Avoid the influence of diet and drugs. ( 5 ) Collection of tongue coating flora : Patients gargle with water 3-5 times before the start of the sampling to rinse out the possible food residue pollution in the mouth. Then, a proper amount of tongue coating samples were scraped on the surface of the tongue coating with a sterile swab, scraped back and forth for 20 times, placed in a sterile tube containing the preservation solution, and frozen in a refrigerator at - 80 ° C as soon as possible. The 16s rDNA analysis method was used to analyze the flora. 1.2 Subjects ( including inclusion and exclusion criteria ) A cross-sectional study was conducted in Peking University First Hospital.CAG patients who underwent upper gastrointestinal endoscopy and were pathologically diagnosed as PLGC were selected as the study subjects. According to gender and age matching, the control group was confirmed by gastroscopy as non-PLGC CAG patients. All participants were informed of the objectives, procedures, potential risks and benefits of the study and then signed a consent form. Diagnostic criteria : Gastroscopy and pathological diagnostic criteria for chronic gastritis were established according to the \' Chinese Consensus on Chronic Gastritis ( 2017, Shanghai ) \' and \' Consensus on Integrated Traditional Chinese and Western Medicine Diagnosis and Treatment of Chronic Atrophic Gastritis ( 2017 ) \'. Inclusion criteria : ( 1 ) Upper gastrointestinal endoscopy was performed within 3 months ; ( 2 ) Patients with chronic atrophic gastritis diagnosed by gastroscopy and pathological examination ; ( 3 ) Age 40-70 years old, gender unlimited ; ( 4 ) agreed to collect tongue image and accept follow-up ; ( 5 ) Voluntary to participate in this study and signed informed consent. Exclusion criteria : ( 1 ) those who cannot cooperate with standardized tongue image collection and data collection ; ( 2 ) patients with tongue scraping, tongue staining or abnormal tongue extension ; ( 3 ) Suspected gastric malignant tendency or gastric tumor and history of gastric surgery ; ( 4 ) Have hepatitis, syphilis, HIV, schistosomiasis and other known infectious diseases ; ( 5 ) patients with severe heart, liver, kidney, blood diseases and malignant tumors ; ( 6 ) Combined with acute respiratory tract infection, active peptic ulcer and other acute diseases ; ( 7 ) Patients who used antibiotics, acid inhibitors and microecological regulators within 1 month ; ( 8 ) long-term use of immunosuppressive agents, hormones and other drugs ; ( 9 ) pregnant or lactating women. 1.3 Observation indicators and follow-up plan 1.Observation indicators 1.1 Correlation between tongue image characteristics and PLGC Analysis of tongue color ( pale tongue, pink tongue, crimson tongue, lavender tongue, ecchymosis tongue ) and tongue shape ( tooth-printed tongue, pricking tongue, ecchymosis tongue, fissured tongue, fat and thin tongue ) in PLGC patients and PLGC correlation ; 2. 2 Analysis of tongue coating characteristics : The correlation between tongue coating quality ( thick and thin coating, greasy coating, rotten coating, peeling coating ) and coating color ( white and yellow ) and PLGC in PLGC patients was analyzed. Analysis of sublingual collaterals : According to the diagnostic criteria of sublingual collaterals grading, the sublingual collaterals were described according to the length, width ( thickness ), tortuosity and color of sublingual collaterals, and the correlation with the pathological changes of PLGC was analyzed. Analysis of tongue color and tongue coating color parameters : RGB, HIS and Lab color space types were used to analyze the RGB, Lab and HSV color space indexes of the whole tongue and the middle part ( corresponding to the spleen and stomach partition ) of PLGC patients. Among them, R represents the red value, G represents the green value, B represents the blue value, H represents the hue, S represents the color saturation, I represents the brightness, L represents the brightness, a represents the red-green axis, b represents the yellow-blue axis. 1.2 Correlation between tongue coating flora and PLGC 1. α diversity analysis : The microbial community diversity among different groups was compared by α diversity analysis. 2. β diversity analysis : Through β analysis, the differences in community structure between different groups were explored, and the key flora was explored. 3. Species composition analysis : The differences in flora composition between groups were compared at the phylum level and genus level to find characteristic flora. Through species and functional contribution analysis, regression analysis and correspondence analysis, the function of microbial composition was predicted to find the function of characteristic flora. 1.3 Construction of PLGC risk prediction model based on tongue image features Univariate conditional Logistic regression analysis was performed on the factors with statistical significance by t test and χ2 test. PLGC was used as the outcome variable to calculate the odds ratio ( OR ) and 95 % confidence interval of each factor. With αin = 0.05, αout = 0.1, the above variables were introduced into Logistic multivariate analysis to screen out the risk factors related to PLGC, and a risk prediction model was established. The receiver operating characteristic ( ROC ) curve was used to represent the ability of the prediction model to distinguish. The goodness of fit and prediction accuracy of the model were evaluated by Hosmer-Lemeshow goodness of fit, Pseudo R-square and area under the curve ( AUC ). 2.Follow-up plan Follow-up : patients diagnosed with severe atrophic gastritis, or severe intestinal metaplasia, or low-grade intraepithelial neoplasia, should be followed up at least once a year ; patients with high-grade intraepithelial neoplasia who refused treatment should be followed up every six months. The severity of CAG was judged by OLGA system ( atrophy ) or OLGIM system ( intestinal metaplasia ). Follow-up was performed by endoscopic iodine staining, indicative biopsy and pathological diagnosis. The follow-up rate was \> 70 %. 1.4 Basis for determining the sample size According to the requirements of expert consultation, literature review and deep learning modeling, the sample size was calculated by using the empirical method that the sample size was more than 10 times the number of independent variables ( 20 independent variables ). In addition, the previous pre-survey found that the incidence of PLGC in CAG patients in the department of this study was expected to be 50 %. Considering the shedding rate of 10 %, the minimum sample size of CAG in this study was 500 cases. This study is expected to include 500 CAG patients. In the Logistic regression model, the number ratio of the training group and the test group is 7 : 3, and the sample size of the training group is set to 350 cases, so the sample size of the test group should not be less than 150 cases. 1.5 Statistical analysis method SPSS 25.0 software was used to analyze the data of this study. The measurement data conforming to the normal distribution were expressed as mean ± standard deviation ( x ± s ), and the t test was used for comparison between groups. The measurement data that did not conform to the normal distribution were expressed as M ( P25, P75 ), and the non-parametric test was used for comparison between the two groups. Enumeration data were expressed as ( % ), and chi-square test ( χ2 ) was used for comparison between groups. P \< 0.05 indicated that the difference was statistically significant. The methods used in model construction include linear regression, Logistic regression ( nomogram ), Cox regression ( nomogram ), and Lasso regression ( screening variables ). Prediction model evaluation : 1 Discrimination : The ROC curve was drawn using the fitted Logistic regression model to obtain the AUC value of the ROC curve and evaluate the discrimination. 2 Calibration : The calibration was evaluated by the Hosmer-Lemeshow goodness-of-fit test. If P \< 0.05, the model was considered to be poorly fitted. If P \> 0.05, it is suggested that the model fits well in the validation of the new data set.3 Clinical effectiveness : The clinical effectiveness of the model was evaluated by Decision Curve Decision Curve Analysis ( DCA ). 1.6 Data acquisition and management Data entry and management is the responsibility of the data administrator. The data administrator compiles the database and uses Epidata software for data entry and management. Two data administrators independently perform double entry and proofreading. After the established database is determined, the database is locked by the main researchers, statistical analysts and data managers, and finally the analysis database is determined.