Clinicopathological Characteristics Of Idiopathic Membranous Nephropathy And Construction Of Bayesian Prediction Model

Objective:1.The clinical and pathological data of idiopathic membranous nephropathy(IMN)were retrospectively analyzed,and patients were grouped according to age,whether they were accompanied by C3 deposition,and whether they were accompanied by renal interstitial fibrosis(RIF).Clinical data characteristics,pathological types and related risk factors of patients in different groups were discussed.2.To construct a Bayesian network(BNs)prediction model by comparing clinical data of IMN and non-IMN in primary glomerular diseases(PGD).Through Bayesian network analysis,the interdependence between the main risk factors of IMN is revealed to improve the understanding of the disease and provide a new method for early identification of high-risk groups and diagnosis of IMN.Methods:1.The patients were from the Department of Nephrology,Second Hospital of Jilin University and diagnosed with IMN by kidney puncture biopsy from September 2016 to August 2022.The general clinical,laboratory and pathological data of patients were analyzed retrospectively.2.Patients diagnosed with PGD were collected from the Department of Nephrology,the Second Hospital of Jilin University from January 2018 to August2022.All patients underwent kidney puncture biopsy and were divided into IMN group and non-IMN group according to pathological results.IBM SPSS 26.0 software was used for statistical description and multivariate logistic regression analysis to screen the main risk factors for IMN.With the help of the ’ bnlearn ’ package in R studio software,the BNs model is constructed through structural learning and parameter learning.Result:1.Comparison of clinical and pathological data of IMN patients at different ages(1)Comparison of general and clinical data of IMN patients at different agesAccording to the age of the patients,they were divided into three groups.The levels of immunoglobulin Ig G,Ig A,complement C4 and cholesterol,and the proportion of anemia in the middle-aged and elderly group were higher than those in the young group(P < 0.05).There were significant differences in the incidence of hypertension,systolic blood pressure and diastolic blood pressure,serum creatinine,cystatin C,urea nitrogen,plasma albumin,e GFR level and Ig M among the three groups(P < 0.05).It can be seen that with the increase of age,the incidence of hypertension,the proportion of anemia,serum creatinine and Ig A level gradually increased,while plasma albumin,e GFR and immunoglobulin Ig M level gradually decreased.(2)Comparison of pathological data of IMN patients of different agesThere were significant differences in the proportions of glomerulosclerosis,tubule interstitial fibrosis,tubule atrophy,and arteriolar wall thickening among the three groups(P<0.01),indicating that the older the age,the higher the proportions of glomerulosclerosis,tubule interstitial fibrosis,tubule atrophy,and arteriolar wall thickening.The positive deposition rate of C3 and C4 in the middle-aged and elderly group was higher than that in the young group,and the positive deposition rate of C1 q was lower than that in the young group,and the difference was statistically significant(P<0.05).2.Comparison of clinical and pathological data of IMN patients in positive and negative C3 deposition groups(1)Comparison of general and clinical data of IMN patients in positive and negative C3 deposition groupsThe levels of serum creatinine,cystatin C,urea nitrogen,urinary protein,cholesterol and low density lipoprotein in C3 positive group were higher than those in negative group,while the levels of plasma albumin,e GFR,Ig G and C3 were lower than those in negative group,and the differences were statistically significant(P<0.05).(2)Comparison of pathological data of IMN patients in positive and negative C3 deposition groupsThe positive deposition rate of IgA and C1 q in the C3 positive group was higher than that in the negative group,and the difference was statistically significant(P<0.001).There were statistically significant differences in the distribution of pathological stages between the two groups(P<0.05),after multiple comparisons,it could be seen that the proportion of stage I in C3 positive group was lower than that in negative group,and the proportion of stage II was higher than that in negative group,and the difference was statistically significant(P<0.05),and the proportion of stage III in C3 positive group was higher than that in negative group.3.Comparison of clinical and pathological data between the group with and without RIF(1)Comparison of general and clinical data between the group with and without RIFThe levels of serum creatinine,cystatin C,urea nitrogen and triglyceride in the RIF group were prominently higher than those in the non-RIF group,but the levels of albumin and e GFR were lower,and the differences were statistically significant(P<0.05).(2)Comparison of pathological data between the group with and without RIFThe proportions of glomerulosclerosis,tubule atrophy and arteriolar wall thickening were statistically different between the two groups,and the renal interstitial fibrosis group was higher than that of the non-fibrosis group(P<0.001).4.Construction of Bayesian network prediction model(1)According to the variables screened by multivariate logistic regression,a BNs model with 10 nodes and 20 directed edges was constructed for the incidence of IMN and related factors.Directed edges reflect the interaction or influence of IMN influencing factors.As can be seen from the model,node age,hypercholesterolemia,macroalbuminuria,creatinine,hyperuricemia,Ig G,Ig A,and body mass index are associated with IMN through complex network relationships,among which age,macroalbuminuria,creatinine,hyperuricemia,hypercholesterolemia,urea nitrogen,Ig G,and body mass index are directly related to IMN,while serum Ig A is indirectly associated with IMN by Ig G.(2)Bayesian model probabilistic reasoningBayesian network can calculate the probability of an event with given network structure and known evidence.All other conditions being equal,patients with hypercholesterolemia and middle age had the highest probability of diagnosis of IMN(64.33%);Bayesian network was used for risk prediction which constructed by Netica,and the risk of IMN could be judged by clicking the corresponding node to add evidence information.For example,when a middle-aged obese patient was diagnosed with primary glomerulonephritis on admission and had biochemical routine suggestive of hypercholesterolemia,macroalbuminuria,but normal creatinine and urea nitrogen,the estimated risk of an IMN diagnosis was 83.9%.Under the same conditions,the incidence of IMN decreased to 37.6% in patients without hypercholesterolemia and macroalbuminuria.If the patient is middle-aged,overweight or obese,whose biochemical indicators show hypercholesterolemia,massive proteinuria,no hyperuricemia,normal urea nitrogen and serum creatinine,and decreased serum Ig G,the risk probability of diagnosing IMN is up to 86.6 %.(3)Bayesian network evaluationThe accuracy and specificity of the model validation set were 77.2% and 79.9%,respectively,and the AUC value was 0.854(95 % CI : 0.823-0.885).The accuracy and specificity of the prediction set were 78.7% and 79.3%,respectively,and the AUC value was 0.855(95 % CI : 0.804-0.904).Conclusion:1.Compared with young IMN patients,middle-aged and elderly patients are more likely to have poor performance in blood tests,and the renal function decline and pathological injury in elderly patients are more serious than that in young patients.Therefore,timely and effective treatment and proper medical care are essential for elderly IMN patients.2.Compared with patients with negative C3 deposition,renal function decreased more significantly,urinary protein quantification and lipid levels were higher,and renal pathological damage was more serious in IMN patients with C3 deposition.C3 deposition may play an important role in the degree of renal lesions in IMN patients.3.Patients with RIF have higher blood lipid levels,more serious renal function decline,and more serious pathological damage.Therefore,it is necessary to actively control the blood lipid levels of IMN patients and detect the risk factors of renal fibrosis early in order to delay the process.4.In this study,an IMN Bayesian network prediction model is constructed,which can help predict the risk probability of IMN and provide additional selection scheme for the diagnosis of IMN.