The Establishment Of Prediction Models For Lymph Node Metastasis In Male And Female With Early Gastric Cancer From A Single Center Database

Background and Objective:Gastric cancer(GC),as one of the most common malignant tumors,still ranks top among all tumor diseases in terms of incidence and mortality.China is one of the countries with the largest number of GC patients,therefore,it has a huge amount of GC data resources.In the era of big data,the realization of the digital integration of medical resources has become a development trend.The special disease database based on clinical data can help to analyze the development law of diseases and build relevant models.Compared with the large-scale clinical data databases that have been built for many years in foreign countries,in China,there is still a lack of large-scale GC specific disease database for use.The First Affiliated Hospital of Anhui Medical University is one of the largest third-class hospitals in Anhui Province.In the past 10 years,more than 10,000 GC patients have been treated,and sufficient clinical data have been accumulated for research.In order to provide data support for GC occurrence and development law and other related studies,this study planned to establish a single-center GC patient database based on the data of GC patients in the First Affiliated Hospital of Anhui Medical University.Early gastric cancer(EGC)is classified as gastric cancer infiltrating into mucosa or submucosa without considering lymph node metastasis(LNM)status.Accurate preoperative assessment of the risk of LNM in EGC patients is significant for the formulation of appropriate treatment plans,and ultimately affects the prognosis of EGC patients.There are differences between male and female EGC patients in LNM.In addition,estrogen levels in female patients vary before and after menopause,and relevant studies have found that estrogen and its receptors are involved in the occurrence and development of GC and LNM.Therefore,it is necessary to explore the relationship between menopause status and LNM in EGC patients.In order to achieve more accurate preoperative assessment of LNM risk in EGC patients,it is necessary to construct risk prediction models for LNM in EGC patients in different subgroups according to gender.Methods:This study aims to collect clinicopathologic data of GC patients who underwent surgical treatment in the First Affiliated Hospital of Anhui Medical University from November 2011 to December 2021.A number of patients’ data will be imported into the electronic database established earlier,including the patients’ age,gender,location of the tumor,size of the lesion,postoperative pathological type of tumor,total number of lymph nodes,and lymph node invasion(LVI),perineural invasion(PNI)and TNM stage.In this study,SPSS26.0 software was used to analyze the clinical and pathological characteristics of GC patients.Combined with the data collected to construct the LNM prediction model for male and female EGC patients in my previous published papers,this study extended the year of data collection based on GC database platform,and added 150 data of EGC patients meeting the requirements to further optimize the model.In this study,EGC patients selected from GC database of our center were selected as the training set,and independent risk factors of LNM in male and female EGC patients were screened out by univariate analysis and multivariate logistics regression analysis.By using R language software(version 4.0.3),independent risk factors screened for EGC patients of different genders were used to establish the nomogram prediction model.In addition,we totally collected 246 EGC patients from the Second and the Fourth Affiliated Hospital of Anhui Medical University from January 2017 to December 2017 as a validation set.The receiver operator characteristic curve(ROC curve)and calibration curve were drawn using each data set to verify the prediction model internally and externally.The "plot ROC","pROC" and "rms" function packages of R software were used to draw and verify the above models.Results:After screening,the data of 11512 GC patients meeting the requirements were imported into the database,among which 8564 were male patients,accounting for74.4%.Besides,there were 2948 female patients,accounting for 25.6%.The mean age of GC patients was 63.0±10.4 years old,and the mean age of male patients was greater than that of female patients(63.8±9.7 VS 60.7±10.8).The mean tumor size was 4.7±2.6 cm,and the highest proportion(51.9%)was in the range of 2.1 to 5cm.LVI(+)accounted for 30.7% and PNI(+)accounted for 27.5% in GC patients.The most common tumor location was the gastric cardia(41.9%),and the most common pathological type was tubular adenocarcinoma(78.1%).In the degree of GC tumor differentiation,the low differentiation type was the most common(69.0%).In all patients,EGC accounted for 17.0%,and stage Ⅲ GC patients accounted for 55.5%.From 2011 to 2021,the proportion of EGC patients in GC patients admitted to our center increased,and the total number of patients showed an overall trend of decline.1646 EGC patients meeting the requirements were collected for the training set totally,of which 1171 were males and 475 were females.In the verification set,there were 187 males and 59 females.Multivariate analysis indicated that tumor size,tumor invasion depth,Lauren classification,LVI and menopause status were independent risk factors for LNM in female EGC patients.For male patients,tumor size,LVI,invasion depth,WHO pathological classification,Lauren’s pathological classification and the location of tumor were independent risk factors for LNM.Based on these risk factors,nomogram prediction models were established for male and female patients,after internal verification,the AUCs were 94.0%(95%CI: 0.920-0.960)and 88.2%(95%CI:0.847-0.917),respectively.The AUCs of external validation were 92.5%(95%CI:0.875-0.975)and 92.4%(95%CI: 0.798-1.000),respectively.The calibration curves of the training set and the validation set in both male and female EGC patients indicated that the LNM risk predicted by the model was in good agreement with the actual risk.Conclusion:In this study,a specific disease database of GC patients was established based on the clinicopathological data of GC patients in a single center,providing a substantial data basis for the study of the disease characteristics,occurrence and the development of GC patients.The size of tumor,Lauren pathological classification,LVI and invasion depth were both independent risk factors for LNM in male and female EGC patients.For female EGC patients,menopausal status was an independent risk factor for LNM;as for male EGC patients,tumor location and WHO pathological classification were independent risk factors for LNM.Gender based LNM risk prediction models for EGC patients will provide help for accurate preoperative assessment of LNM risk and formulation of appropriate treatment plan for EGC patients,and ultimately bring benefits to patients.