| Objective:1.The clinical data of patients with multicenter Ig A nephropathy(Ig AN)were retrospectively analyzed.They were divided into groups according to pathological Oxford classification(MEST-C)and stratified according to estimated glomerular filtration rate(e GFR).The clinical features and risk factors of pathological progression were analyzed.2.Bayesian network(BN)and other machine learning(ML)models were applied to construct the non-invasive diagnosis model of Ig AN.And the effectiveness of the models was evaluated to provide a new model for the diagnosis of Ig AN.Methods:1.From July 2014 to July 2022,the clinical laboratory data and renal pathological data of patients first diagnosed as Ig AN after renal biopsy in the second Hospital of Jilin University,Jilin Provincial people’s Hospital and Jilin City people’s Hospital were collected.Oxford classification is the evaluation standard of renal pathological data.The clinical characteristics and risk factors of pathological damage progression in different pathological groups were compared.2.The clinical and pathological data of patients with primary glomerular disease who underwent renal biopsy in the second Hospital of Jilin University from January2018 to August 2022 were collected and divided into Ig AN group and non-Ig AN group.R4.1.3 software was used to analyze and model the data and LASSO regression method was used to screen variables.Tabu search algorithm was selected to establish BN structure,and the probability of each node waas estimated by maximum likelihood method according to the BN structure.The BN model was compared with the other ML models and their performance was evaluated.Result:1.Comparison of Clinicopathological Data of Different Oxford classification of Ig AN(1)Basic characteristics of patients with different Oxford classification of Ig ANA total of 456 patients were enrolled in the study,with an average age of 38.93±12.98 years,a male-female ratio of 1.2: 1,and a female ratio of 45.4%,with an average BMI of 24.35 ±3.8,and 9.6% of them had symptoms of precursor infection.(2)Comparison of clinicopathological data between Oxford classification M and Ig AN patientsIn M group,diastolic blood pressure,urinary red blood cell and albumin in M1 group were higher than those in M0 group,while edema of lower extremities,total cholesterol,LDL-C,24-hour urinary protein and C3 in M1 group were lower than those in M0 group.Multivariate Logistic regression found that albumin was an independent risk factor for the progression of Oxford M in Ig AN patients,while LDL-C and C3 were independent protective factors.(3)Comparison of clinicopathological data between Oxford classification E and Ig AN patientsIn E group,urinary red blood cell,SCys-C,drinking,lower limb edema,blood urea nitrogen,serum creatinine,urinary α1-microglobulin,leukocyte,and 24-hour urinary protein were significantly higher in group E0 than in group E1,while hemoglobin,albumin,e GFR and blood Ig G decreased significantly in group E0.The results of multivariate Logistic regression showed that alcohol consumption,urine red blood cells and SCys-C were independent risk factors for the progression of Oxford type E in Ig AN patients,while albumin was an independent protective factor for the progression of this pathological type.For Ig AN patients with renal insufficiency,alcohol consumption,urine red blood cells and urine α1-microglobulin were independent risk factors for the progression of Oxford type E.(4)Comparison of clinicopathological data between Oxford classification S and Ig AN patientsIn S group,urea nitrogen and SCys-C in S1 group were higher than those in S0 group,while blood C3 in S1 group was lower than that in S0 group.Multivariate Logistic regression showed that platelet and SCys-C were independent risk factors for the progression of Oxford classification of S in patients with Ig AN,which was consistent with the stratification of renal insufficiency.Edema of both lower extremities was an independent protective factor for the progression of Oxford type S in patients with Ig AN.In patients with normal renal function in Ig AN,LDL-C was an independent protective factor for the progress of Oxford classification S.(5)Comparison of clinicopathological data between Oxford classification T and Ig AN patientsIn T group,systolic blood pressure,diastolic blood pressure,blood urea nitrogen,uric acid,serum creatinine,SCys-C,urinary α1-microglobulin,24-hour urinary protein and serum C4 were increased,while glutamic oxaloacetic transaminase,hemoglobin,albumin and e GFR were decreased in T0 group compared with T1/2group.Multivariate Logistic regression showed that glutamic oxaloacetic transaminase and e GFR were independent protective factors for the progression of Oxford type T in patients with Ig AN,which was consistent with the stratification of renal insufficiency.In Ig AN patients with normal renal function,hemoglobin was an independent protective factor for the progression of Oxford type T,and serum creatinine was an independent risk factor for the progression of Oxford type T.(6)Comparison of clinicopathological data between Oxford classification C and Ig AN patientsIn C group,urinary red blood cells,serum creatinine,SCys-C,urinaryα1-microglobulin and LDL-C in group C1/2 were significantly higher than group C0,while albumin,hemoglobin and e GFR were significantly lower than group C0.Multivariate Logistic regression showed that albumin,e GFR and HDL-C were independent protective factors for the progression of Oxford classification C in Ig AN patients,which was consistent with renal function stratification.Urinary red blood cells were an independent risk factor for the progression of Oxford type C.2.Construction and evaluation of Ig AN diagnosis model(1)Demographic data and clinical characteristics of Ig AN and non-Ig ANThere was no significant difference in sex between the Ig AN and non-Ig AN groups,and the female ratio was 44%(117/265)and 40%(214/531)respectively.There were significant differences in age,hemoglobin,albumin,serum uric acid,serum creatinine,e GFR,LDL-C,total cholesterol,24-hour urine protein,Ig G,Ig A,Ig M,C3 and C4 between the two groups.(2)LASSO regression selection variableThe eight variables selected by LASSO regression parameter log(λ)1se=-3.62 were albumin,elevated serum creatinine,LDL-C,total cholesterol,24-hour urinary protein,Ig G,Ig A,Ig M.The main factor of Ig AN was the increase of serum creatinine,followed by cholesterol and serum Ig A.(3)Establishment and Evaluation of BNFour models were established: Bayesian network,decision trees(DT),random forest(RF)and support vector machine(SVM).In the training set,the AUC values from high to low are BN(0.948),DT(0.934),SVM(0.905)and RF(0.872).The accuracy of BN is 85.8%,the precision rate is 77.1%,the recall rate is 75.0%,and the F1 score is 0.761.Conclusion:1.There were some differences and correlations between the pathological changes of MEST-C in Oxford classification and clinical laboratory indexes of Ig AN patients.Clinical indexes such as renal function,albumin and proteinuria were common risk factors for the progress of Oxford classification of Ig AN patients.Some clinical risk factors had certain differences in e GFR stratification.2.The diagnosis and prediction model of Ig AN was established by using BN and other ML methods.After evaluation,the AUC value of BN and the score of comprehensive evaluation index F1 were the highest among the four models,and it can be considered that BN has the best clinical efficacy. |
PDF Full Text Request