The Performance Of Combining A Functional Marker(Serum Cystatin C)with A Tubular Damage Marker(Urinary N-acetyl-β-D-glucosaminidase)for Acute Kidney Injury Prediction And Development Of A Risk Model For Acute Kidney Injury In Critically Ill

Background and ObjectionAcute kidney injury(AKI)is frequent and associated with adverse outcomes,espaciallly in critically ill.Unfortunately,early diagnosis of AKI remains a challenge.Combining functional and tubular damage biomarkers may provide better precision for AKI prediction.However,the predictive accuracy of this combination for AKI in critically ill is unclear.Serum cystatin C(sCysC)is considered as a biomarker indicating damage for glomerular function,while urinary N-acetyl-β-D-glucosaminidase(uNAG)represents tubular damage.We aimed to assess the performances of these two clinical available biomarkers and their combination for AKI prediction in critically ill,including patients after resection of intracranial space-occupying lesion and patients with sepsis.In addition,we found that glycemic status had no impact on the accuracy of uNAG in detecting AKI in critically ill.Similar to uNAG,sCysC was reported to be independently associated with hemoglobin(HbAlc)levels,diabetes,and prediabetes.We aimed to assess the influence of glycemic status on the performance of sCysC for detecting AKI in the critically ill.MethodsSection 1:A prospective study was conducted,enrolling adults undergoing resection of intracranial space-occupying lesion and admitted to the intensive care unit(ICU).The discriminative abilities of postoperative sCysC,uNAG and their combination in predicting AKI were compared using the area under the receiver operator characteristic curve(AUC),continuous net reclassification index(cNRI),and incremental discrimination improvement(IDI).A multivariate logistic regression analysis was performed to constructed a predcitive model,and nomogram wased then builted based on it.The calibration and clinical utility of this nomogram were assessed.Section 2:A prospective study was conducted,enrolling septic adults admitted to the ICU,including development cohort and validation cohort.The discriminative abilities of sCysC,uNAG and their combinations in predicting AKI were compared using the AUC,cNRI,and IDI.A multivariate logistic regression analysis was firstly performed in the development cohort to constructed the clinical model,and nomogram wased then builted based on it.The calibration and clinical utility of this nomogram were assessed.The accuracy of the risk model was validated in the validation cohort.Section 3:A prospective observational study was conducted in the ICU.Patients were divided into four quartiles based on levels of HbAlc or serum glucose at ICU admission,respectively.Additionally,patients were stratified into four subgroups according to HbAlc levels and history of diabetes,namely recognised diabetes(previous diagnosis of diabetes),unrecognised diabetes,prediabetes,and normal glycemic status.Comparisons were made using the AUC for AKI detection,and reassessed after patient stratification by abovementioned glycemic status.ResultSection 1:Of 605 enrolled patients,AKI occurred in 67 patients.The cut-off values of sCysC and uNAG to predict postoperative AKI were 0.72 mg/L and 19.98 U/g creatinine,respectively.For predicting AKI,the composite of sCysC and uNAG(AUC=0.785)outperformed either individual biomarkers.Compared to the the clinical model without sCysC and uNAG,the predictive model containing these two bomarkers improved the AUC,IDI,and cNRI.Nomogram wased then builted based on the risk model including sCysC and uNAG and yielded good calibration.Section 2:Of 358 enrolled patients,232 patients were in the development cohort(69 AKI patients)and 126 patients were in the validation cohort(52 AKI patients).Two models for predicting AKI in development cohort were built.The first model included APACHE II score,Sequential organ failure assessment score,serum creatinine,and vasopressor used at ICU admission.The second model included sCysC,uNAG,Acute Physiology and Chronic Health Evaluation II(APACHE II)score,serum creatinine,and vasopressor used at ICU admission.Compared to the first model,the second one improved the AUC to 0.831 and significantly improved risk reclassification,with cNRI(0.489)and IDI(0.053).Nomogram wased then builted based on the risk model including sCysC and uNAG.Application of the nomogram in the validation cohort yielded fair discrimination with an ROC-AUC of 0.784(95%CI,0.703-0.865)and good calibration.Section 3:Of 1317 enrolled patients,379(28.8%)patients were diagnosed as AKI.Multivariable linear regression revealed that HbAlc levels and history of diabetes were positively related with sCysC(all P<0.05).Although stratification for abovementioned glycemic status displayed no significant difference between AUC of sCysC(all P>0.05),sCysC yielded the highest AUCs for detecting AKI in diabetic patients.Moreover,higher optimal cut-off values of sCysC to detect AKI were observed in patients with versus without diabetes.ConclusionsCombination of functional and tubular damage biomarkers(sCysC and uNAG)increases the predictive accuracy for AKI and improves the performance of risk model for AKI predicting in critically ill.Based on the model containing sCysC and uNAG,a nomogram which can predict the risk of AKI in critically ill patients was then developed.The nomogram yielded good discrimination and calibration,showing that this nomogram have the potential to assist doctors in prevention and treatment of AKI in clinical practice.Similar to uNAG,the accuracy of sCysC for AKI detection in critically ill was not significantly affected by glycemic status and a higher optimal cut-off value of sCysC in detecting AKI should be considered in diabetic patients.