The Construction And Establishment Of Risk Prediction Model On Precancerous Lesions Of Gastric Cancer Based On Logistic Regression Analysis

BackgroundGastric cancer is one of the most common malignancies in the world.The incidence of gastric cancer in the world ranks fifth,and the mortality rate is third.According to the WHO/International Cancer Research Center(IARC)released statistics show that in 2012 the global gastric cancer 989 thousand new cases,about 50%occur in the East Asian region,especially in the China most,accounted for the global incidence of gastric cancer 46.8%;gastric cancer deaths worldwide in 2012737 thousand,Chinese deaths was 352 thousand,accounted for 47.8%of cancer deaths in the world.According to statistics,in 2015,the incidence of gastric cancer in China was 22.7/10 million,second only to lung cancer,ranking second;the mortality rate of gastric cancer was 17.9/10 million,second only to lung cancer and liver cancer,ranking third.The incidence of gastric cancer has undergone a series of evolution by Correa et al.Proposed the evolution model of gastric adenocarcinoma of intestinal type:normal gastric mucosa of chronic superficial gastritis,chronic atrophic gastritis,intestinal metaplasia,dysplasia and intestinal type of gastric cancer.The atrophy of the gastric mucosa lesion background of intestinal metaplasia(Intestinal Meteplasia,IM)and dysplasia(Dysplasia,DYS)is regarded as the precancerous lesions of gastric cancer(precancerous lesions of gastric cancer,PLGC),is also the focus and difficulty in the field of prevention and treatment of gastric cancer.Severe dysplasia has obvious canceration tendency,to actively carry out endoscopic resection,moderate to severe intestinal metaplasia and mild to moderate dysplasia at home and abroad have unanimously recommended regular endoscopic surveillance.The biopsy of pathological tissue under electronic gastroscope belongs to interventional operation,and it has great trauma and high cost.It is difficult to carry out large-scale investigation as a census method under the existing medical conditions in our country.Serum ABC method combined with G-17 detection can improve the detection rate of gastric cancer and atrophic,and has the advantages of non-invasive,simple,fast and inexpensive,but whether these indicators on the incidence of PLGC and prognosis prediction still can make nothing of it.At present,the research tool to develop screening less on PLGC at home and abroad,the existing model is limited in terms of environmental factors,promotion of the poor,and the lack of relevant TCM syndromes in clinical practice,there are certain limitations.Objective1.Screening out the important risk factors of PLGC and TCM syndrome elements,and making clear the important factors of PLGC.2.Based on Logistic regression analysis,PLGC risk prediction model was established to provide data model for PLGC screening.3.To evaluate the PLGC risk prediction model,and to provide a scientific basis for screening high-risk groups in PLGC.Methods1.According to the diagnostic criteria of chronic atrophic gastritis and dysplasia,set inclusion and exclusion criteria.According to the Chinese medicine dialectical syndromes diagnosis is divided into two aspects:one is the consensus,country,association of published materials,guidelines and expert opinion;the two is empirical syndrome diagnosis in two chronic atrophic gastritis of TCM diagnosis and treatment work for more than 20 years or more senior practitioners.The patients with chronic atrophic gastritis(CAG)were selected from the Department of spleen and stomach clinic of Xiyuan Hospital,Chinese Academy of traditional Chinese medicine and gastroscope room,and were screened.2 Methods of relevant indicators and questionnaire survey by serological detection,collection of PLGC related clinical data including clinical risk factors(general information,life behavior,dietary habits,gastroscopy and pathological emotional factors,family history,performance)and TCM syndrome elements and etc.;3.Logistic single factor and multiple factor regression analysis were used to screen PLGC related risk factors;4.Establishing PLGC risk for ecasting model based on Logistic regression;5.Using Hosmer-Lemeshow goodness of fit,Pesudo,R-square and AUC to evaluate the PLGC risk prediction model,and using ROC curve to analyze the ability of the model to enrich high-risk population.Results1.PLGC related factors?Life behavior:smoking in the PLGC group and the non PLGC group were exposed to:31%vs24.8%(P>0.05);alcohol exposure in the PLGC group and non PLGC group were:40.5%vs27.7%(P>0.05).?The diet:eating vegetables were exposed in the PLGC group and non PLGC group:14.0%vs26.3%(P<0.05),OR 0.236(95%CI=0.058-0.952);bean products were in the PLGC group and the non exposure group:34.9%vs17.5%,PLGC(P>0.05);drink tea respectively in the PLGC group and non PLGC exposure group 16.3%vs4.4%(P<0.05),OR 6.288(95%CI=1.465-26.989);eating pickled food were exposed in the PLGC group and non PLGC group:37.2%vs16.1%(P<0.05),OR 3.198(95%CI=1.289-7.930).?The history of Hp infection:a history of Hp infection were exposed in the PLGC group and non PLGC group:7.0%vs16.8%(P>0.05);?The social psychological factors of anxiety in the PLGC group were exposed and the non PLGC group:41.9%vs19.7%(P<0.05),OR 2.404(95%CI=1.058-5.511);genetic factors:family history.Gastric cancer respectively in PLGC group and non PLGC group:14.0%vs10.2%exposure(P>0.05);?Family history of digestive tract tumor were exposed in the PLGC group and non PLGC group(P>0.05):16.3%vsl 1.7%.?The disease:gastroesophageal reflux disease were exposed in the PLGC group and non PLGC group:39.8%vs23.3%(P>0.05);duodenal ulcer were exposed in the PLGC group and non PLGC group:79.3%vs7.3%(P>0.05);history of cholecystectomy were exposed in the PLGC group and non PLGC group(vs 1.5%:97%P<0.05),OR 7.351(95%CI=1.824-55.384).?The serological examination:PLGC serum PGI,serum PGII,G-17 levels were in group PLGC:55.01 + 35.85ug/L,4.50 + 2.82 ug/L,4.21 +7.69ng/L,respectively in non PLGC group:62.40 + 48.59 ug/L 5.40 + 5.33 ug/L 4.13+ 9.77 ng/L(P>0.05);Hp IgG the anti body respectively in the PLGC group and non PLGC group:20.9%vs16.1%exposure(P>0.05);?TCM syndrome:symptoms:the total symptom score respectively in PLGC group and non PLGC group:14.19 +11.21vs17.97 + 13.56(P>0.05);the main symptoms were in PLGC group and non PLGC group.10.7 + 6.11vs11.31 + 6.21(P>0.05);aspects:blood stasis syndrome factors were exposed in the PLGC group and non PLGC group:44.2%vs28.5%(P<0.05),OR 2.420(95CI%=0.998-5.851)respectively in PLGC group;damp heat exposure and non PLGC group(P>0.05):41.9%vs27%.2.Screening of PLGC risk factors based on Logistic regression analysisThrough the use of Logistic single factor and multi factor regression analysis,the final screening of the pickled,anxiety,cholecystectomy history,blood stasis,strong tea 5 important risk factors.The history of cholecystectomy and PLGC risk most closely,the value of OR was 7.351(95%CI=1.824-55.384);the second is tea,the value of OR was 5.351(95%CI=1.824-55.384);salting,blood stasis,OR values were 3.198(95%CI=1.289-7.930),2.420(95%CI=0.998-5.851);anxiety again,the value of OR was 2.404(95%CI=1.058-5.511).3.Establishment of PLGC risk prediction model based on Logistic regression analysisLn(p/(1-p)=-2.5072.153X1+1.995X2+1.162X3+0.884X4+0.877X5)(X1= cholecystectomy,X2= strong tea,X3= pickled,X4= blood stasis,X5=anxiety)4.PLGC risk prediction model evaluationThe goodness of fit:Hosmer-Lemeshow(H-L)test,2 = 3.997,P=0.550>0.05,not very good fit to the data model to reject the hypothesis,namely the model better fit.Model accuracy:about Pesudo,R-square,Cox,and,Snell is 0.209,Nagelkerke is 0.314,AUC=0.812(95%,CI=0.738-0.885),P<0.01,which indicates that the prediction accuracy of this model is acceptable.5.PLGC risk prediction model,enrichment,high risk,capability analysisAccording to the model ROC curve,the sensitivity is set at 100%,corresponding to the lowest 1-specificity is 66.7%.Under this condition,the lowest risk population is 87%,and the corresponding predictive probability P is 0.00673502.Endoscopic screening of the high-risk group revealed that 1 of the 4 high-risk patients required endoscopic monitoring of the true PLGC,with a rate of 1.15.Conclusions1 Pickled,cholecystectomy,anxiety and strong tea for PLGC related risk factors.2 Blood stasis syndrome is of great significance in the evolution of PLGC.3 PLGC risk prediction model for:Ln(p/(1-p)=-2.5072.153X1+1.995X2+1.162X3+0.884X4+0.877X5)(X1= cholecystectomy,X2= strong tea,X3= pickled,X4= blood stasis,X5=anxiety)4 The PLGC risk prediction model has good fitting and prediction accuracy,and further validation and calibration of the model need long-term follow-up.