Biomarkers And Prediction Models For Chronic Kidney Disease

objective:Chronic kidney disease(CKD)refers to the dysfunction of renal structure and/or damage of renal.CKD is defined as the decrease of glomerular filtration rate(<60ml/min·1.73m~2)or accompanying proteinuria.Recently,the incidence and fatality rate of CKD increased dramatically.The prevalence of CKD has exceeded 10%in worldwide.Epidemiology studies on CKD have been carried out in different countries or areas around the world.The prevalence of CKD is high in worldwide,but people knows little about it.When people is diagnosed as CKD,it is hard to cure.So,it is important to explore marker to diagnose early CKD.Besides,As the occurrence of CKD is closely related to race,region,and economic development level,it is a great practical significance to carry out epidemiological studies on CKD and to establish prediction models in certain regions or populations.Our research is based on the“Tianjin Medical University CKD Cohort”.We adopt a nested case-control study to explore biomarkers and gene loci related to the pathogenesis of CKD and to create a disease risk prediction model for CKD.Methods 1.Participants in“Tianjin Medical University CKD Cohort”were selected from“Tianjin Medical University Chronic Disease Cohort”who met the criteria of not having CKD at the time of the first physical examination and had complete baseline information in August 2006 and later.Up to August 2018,all the subjects were followed up twice for at least two physical examinations.We performed a Cox regression model to analyze risk factors for CKD,we used Kaplan-Meier survival curve to analyze the prevalence rate of CKD.2.A nested case-control study was performed from“Tianjin Medical University CKD Cohort”to establish a 5-year risk prediction model for CKD.All the subjects have completed blood samples.A total of 1804 people met the screening criteria,we selected 116 subjects from 156 people who was newly diagnosed CKD as the case group.Meanwhile,232 subjects were selected as the control group according to gender and age±3 years.Finally,348 participants were included in the nested case-control study.We detected the concentrations of serum Cystatin C(Cys C),transforming growth factor beta(TGF-beta),granulocyte gelatinases people center related lipid carrier protein(NGAL),and asymmetric dimethyl arginine(ADMA)by enzyme-linked immunosorbent assay(ELISA).Multivariate Cox regression model was performed to analyze the correlation among biomarkers and CKD.And a risk prediction model for non-genetic factors of CKD was constructed by using natural logarithmic OR value weighting.GWAS was performed in three CKD-related phenotypes(estimated Glomerular Filtration Rate,Serum Creatinine,Cystatin C)in over 480,000 people in UK-Biobank database to screen candidate genes and SNPs that associated to CKD.We used logistic regression analysis model and natural logarithm OR value weighting to establish CKD genetic/non-genetic risk prediction models.Results 1.Epidemiological analyses of CKD(1)A total of 7462(male 5710,female 1752)patients were included in“Tianjin Medical University CKD Cohort”,average follow-up time in the cohort was 5.7years.By the end of follow-up,665 cases of CKD were newly added,the cumulative incidence of CKD is 8.9%(male 8.4%,female 10.4%,?~2=2.0 P=0.157).(2)The results of Cox proportional risk regression showed that female,age,normal high value of BUN,normal high value of creatinine,hyperuricemia,and diabetes were associated with CKD.They were independent risk factors for CKD in our cohort.2.Establishing risk prediction model for CKD(1)We conducted a nested case-control(NCC)study of CKD based on“Tianjin Medical University CKD Cohort”.A total of 348 participants(116 cases and 232controls)who were followed up for 5 years.Blood sample were collected for nested case-control subjects at the baseline and 5 years later.Average age in the study was63.27±10.09 years(male 63.19±10.03 years,female 63.53±10.32 years).Cox proportional risk regression model showed that age,diabetes mellitus,normal high value of urea,normal high value of TGF-β,and ADMA were risk factors for CKD.(2)CKD non-genetic prediction modelA total of 5 predictors including age,diabetes mellitus,normal high value of BUN,normal high value of TGF-β,and ADMA were included in the non-genetic prediction model for CKD.The prediction equation was logit P=1.84×S1+1.137×S2+0.84×S3+0.497×S4+0.603×S5,the area under the ROC curve(AUC)was 0.889,the sensitivity of the model was 0.851,while the specificity was0.770.The non-genetic prediction model was internally verified by Bootstrap resampling and 5-fold cross-validation,and its AUC value was 0.786,which still had high predictive power.(3)CKD genetic prediction modelBy integrating the results of CKD-related genetic loci in the UK-Biobank subjects and previous studies,17 SNP sites were selected in CKD genetic prediction model,including 7 SNPs derived from UK-Biobank.The genetic risk prediction equation for CKD was logit P=0.577×rs17319721Gi+(-0.183)×rs700233+(-0.362)×rs671Gi+(-0.286)×rs11864909Gi+1.099×rs653178Gi+0.255×rs3752462Gi+0.228×rs13146355Gi+0.253×rs881858Gi+(-0.24)×rs1153849Gi+(-0.234)×3770636Gi+(-0.178)×rs504915Gi+0.149×rs16853722Gi+0.683×rs12917707Gi+(-0.133)×rs1731274Gi.The area under the ROC curve of CKD genetic prediction model was 0.643,the sensitivity of the model was 0.794,while the specificity was 0.838.After internal verification,the model still had high predictive power,and its AUC was 0.692.(4)CKD comprehensive prediction modelAfter combining the predictors in both non-genetic and genetic prediction models of CKD,the prediction equation of the comprehensive predictionmodel for CKD was established,predictors including age,diabetes mellitus,normal high value of BUN,normal high value of TGF-β,ADMA,and 17CKD-related SNPs.The logit P=0.577×rs17319721Gi+(-0.183)×rs700233+(-0.362)×rs671Gi+(-0.286)×rs11864909Gi+1.099×rs653179Gi+0.255×rs3752462Gi+0.228×rs13146355Gi+0.253×rs881858Gi+(-0.24)×rs1153849Gi+(-0.234)×3770636Gi+(-0.178)×rs504915Gi+0.149×rs16853722Gi+0.683×rs12917707Gi+(-0.133)×rs1731274Gi+1.84×S1+1.137×S2+0.84×S3+0.497×S4+0.603×S5.The predictive power of CKD comprehensive prediction model is high,the AUC was 0.894,and the sensitivity and specificity were both at high levels,while the sensitivity was 0.827 and the specificity was 0.801.The AUC value was 0.820 after the internal verification of the 5-fold cross-validation in the nested case-control study.Conclusions 1.Age,diabetes,normal high values of urea,normal high values of TGF-β,and ADMA were independent risk factors for CKD.2.After including renal function related biomarkers,the predictive power of CKD non-genetic prediction model was at a high level.At the same time,the genetic risk model which included 5 SNPs from UK-Biobank also had a good predictive power.The predictive power of the comprehensive prediction model further improved.Sensitivities and specificities of all CKD prediction models remained high after internal validations.