Risk Factors Analysis And Prediction Model Construction And Evaluation Of Chronic Kidney Disease

Part ?:Monocyte lymphocyte ratio is an independent predictor of new-onset chronic kidney diseaseObjectives:Chronic kidney disease(CKD),an inflammation-related disease,is a serious threat to human health.The purpose of this study was to investigate whether monocyte lymphocyte ratio(MLR)was an independent predictor of new-onset CKD.Methods:This study was a retrospective cohort study,in which data were continuously collected from 14,033 annual physical examinations(at least two physical examinations during the study period,with intervals of at least 1 year),and data missing and juvenile participants were excluded.The primary outcome was the new-onset CKD defined as an estimated glomerular filtration rate(e GFR)<60m L/min/1.73m~2 or the presence of proteinuria after follow-up.Firstly,a descriptive analysis of baseline data was performed to compare differences in baseline data between participants with and without CKD during follow-up.Secondly,univariate Cox proportional risk regression model was used to screen out the baseline indicators associated with new-onset CKD.Then MLR was compared with new-onset CKD and other variables.Finally,multivariate Cox proportional risk regression models were used to determine whether MLR could predict the risk of new-onset CKD.In addition,the predictive ability of MLR for new-onset CKD among different genders and participants with previous medical history were compared.Results:12868 participants were included in the final analysis,and 7712(59.9%)of them were male.The mean age was 45.97±13.54 years.After a median follow-up of 1.78 years,414(3.22%)of participants developed new-onset CKD.There were significant differences in baseline data between participants diagnosed with or without new-onset CKD after follow-up.Compared with participants without CKD,those developed CKD after follow-up were elder,fatter,more likely to be male and had higher blood pressures,higher inflammation marker(higher leukocyte,neutrophil,monocyte,neutrophil lymphocyte ratio(NLR)and MLR),higher fasting blood sugar,higher low density lipoprotein cholesterol,higher triglycerides,higher blood uric acid,lower e GFR and lower high density lipoprotein cholesterol.Univariate Cox proportional risk regression model showed the number of leukocytes(HR=1.09,95%CI=1.03-1.16,P=0.0023),the number of neutrophils(HR=1.12,95%CI=1.04-1.20,P=0.0017),the number of monocytes(HR=4.24,95%CI=2.27-7.91,P<0.0001),NLR(HR=1.18,95%CI=1.06 1.30,p=0.0024)and MLR(HR=2.84,95%CI=1.90 4.24,p<0.0001)were positively correlated with the risk of CKD.After adjusting for age,body mass index,systolic blood pressure,mean corpuscular haemoglobin concentration,high-density lipoprotein cholesterol,globulin,blood urea nitrogen,estimated glomerular filtration rate,and fasting blood glucose,MLR(HR=2.29,95%CI=1.03-5.10,P=0.043)weas still independent predictors of new-onset CKD.The number of leukocytes,neutrophils and monocytes were also independent risk factors for new-onset CKD.Subgroup analysis showed that MLR showed a better predictive power for new-onset CKD in men without diabetes,hypertension,cardio-cerebrovascular disease,or non-alcoholic fatty liver disease.Conclusion:MLR is an independent predictor of the risk of new-onset CKD.Number of leukocytes,neutrophils and monocytes are also independent risk factors of new-onset CKD in people with normal or near-normal kidney function at baseline.This suggests that reducing the level of persistent low-grade inflammation in the body may help prevent or delay the occurrence and development of CKD.Part ?: Establishment and evaluation of a 3-year risk prediction model for chronic kidney disease in community populationObjective: Chronic kidney disease(CKD)is a major global public health problem.There is a lack of models for predicting new-onset CKD in the community population in mainland China.The purpose of this study was to establish a 3-year risk prediction model for CKD in community population and evaluate the effect of the model.Methods: Participants in part I of this study whose follow-up time was no less than 3 years were selected to construct the training cohort and validation cohort.(1)Compare the differences in baseline of all variables between participants with and without new-onset CKD after 3-year follow-up between the training cohort and the validation cohort.(2)The importance of each variable in predicting new-onset CKD was evaluated and ranked by XGBoost algorithm based on machine learning.(3)Univariate logistic regression analysis was used to screen out variables associated with new-onset CKD.(4)Based on the results of machine learning and logistic regression analysis,some variables were selected and used to construct the prediction model.A total of five models,the full variable model,the stepwise model based on Akaike information criterion(AIC)and backward step-down selection,the MFP model based on multivariate fractional polynomial,the full variable model(bootstrap resampling =500 times)and the stepwise model(bootstrap resampling =500 times)were established respectively.The receiver operating characteristic curves(ROC)of each model were drawn,and the area under curve(AUC),threshold,sensitivity,specificity,positive predictive value,negative predictive value of the five models were described.Delong test was used to compare the differences among AUCs.(5)Decision curve analysis and calibration curve were carried out for the prediction model.(6)The prediction model establishted by the training cohort was used to analyse the data of validation cohort to evaluate the discrimination and stability of the prediction model.(7)Draw a nomogram of the prediction model.Results: 2406 participants were included in the final analysis,and 1476(61.35%)of them were male.The mean age was 49.96 ± 14.31 years.After screening,16 variables were finally used to construct the prediction model.They included age,body mass index,systolic blood pressure,diastolic blood pressure,number of monocytes,MLR,high density lipoprotein,triglyceride,total cholesterol,e GFR,fasting glucose,aspartate aminotransferase,alkaline phosphatase,platelet distribution width,mean platelet volume and mean corpuscular haemoglobin concentration.The five prediction models all showed good discriminative ability,with AUC greater than 0.8,and there was no statistical difference between AUCs.Therefore,the stepwise model(Bootstrap resampling = 500 times)was selected as the main model.Eleven variables were included in this model,including age,body mass index,systolic blood pressure,MLR,high density lipoprotein cholesterol,total cholesterol,e GFR,fasting blood glucose,alkaline phosphatase,platelet distribution width and mean platelet volume.The AUC of the model was 0.82(0.79-0.85),the sensitivity was 76.85%,the specificity was 72.27%,the positive predictive value was 34.37%,and the negative predictive value was 94.29%.In the validation cohort,the model also showed good stability: the AUC was 0.81(0.75-0.87),sensitivity was 76.81%,specificity was 77.97%,positive predictive value was 44.92%,and negative predictive value was 93.5%.The calibration curve showed that the model had good accuracy and decision curve analysis showed that the model had good net benefit.Conclusion: This study established a 3-year prediction model for the new-onset of CKD in community adults,and the model showed good discriminative ability and stability.Visualizing the prediction model as a nomogram can be directly used for personalized quantitative assessment of the risk of CKD in 3 years,which has clinical practical value.