Optimization Of Glomerular Filtration Rate Measurement And Equation Estimation In Non-Dialysis Patients With End-Stage Renal Disease And Establishment Of A Calculation Model For 24-Hour Urinary Potassium Excretion In Hyperkalemia

Background:Chronic kidney disease（CKD）affects 10%of adults worldwide and causes 1.2 million deaths each year.End stage renal disease（ESRD）is the terminal stage of the progression of CKD,which is often accompanied by multiple complications such as hyperkalemia,endangering patients’ lives.Glomerular filtration rate（GFR）is the basis of renal function staging,and accurate evaluation of GFR was the basis of clinical diagnosis and treatment in CKD patients.At present,there are many equations for estimating GFR in CKD patients reported in the literature.How to accurately estimate GFR in ESRD patients without dialysis is an urgent clinical problem to be solved.In recent years,the technique of transdermal detection of GFR has been applied to animal experiments,the equation for estimating GFR in rats were published in 2021,but the accuracy of the two methods in animals with chronic renal failure was rarely reported.Hyperkalemia is one of the common serious complications in non-dialysis ESRD patients,which significantly increases the risk of sudden death due to fatal arrhythmia.Guiding dietary potassium intake according to daily urinary potassium excretion is the main method to prevent hyperkalemia in non-dialysis ESRD patients.24-hour urine potassium detection is an accurate method to measure daily urine potassium excretion,but the time-consuming and tedious urine collection process greatly limits its clinical application.How to accurately and conveniently estimate the daily urinary potassium excretion of non-dialysis ESRD patients is an urgent clinical problem to be solved.Objective:1.To analyze and compare 23 equations for estimating GFR and screen out a relatively accurate equation for estimating GFR in non-dialysis ESRD patients.2.To investigate the accuracy of the transdermal GFR measurement technique and the rat equation for estimating GFR in normal rats and rats with chronic renal failure.3.To establish a calculation model of 24-hour urinary potassium excretion in non-dialysis ESRD patients with hyperkalemia,evaluate its accuracy,and compare it with previous estimation models of urinary potassium excretion.Method:1.The subjects included in this study were a total of 132 non-dialysis ESRD patients who were admitted to China-Japan Friendship Hospital from January 2012 to March 2022 and were≥18 years old and underwent renal dynamic imaging.ESRD is defined as measuring GFR（Gates method corrected by the dual plasma method）<15ml/min/1.73m2.The general data of the patients were collected and compared 23 GFR estimation equations（CKD-EPIScr,MDRDII,FASScr,EKFC,Lund-Malm? Revised,Mayo,XiangYa,XiangYa-s,Vilar,Shafiβ2MG,CKD-EPISCcysC,FASScysC,CAPA,Hoek,Yang,CKD-EPIScr-SCysC,FASScr-SCysC,Adachi,ShafiBTP,ShafiBTP-β2MG,Wong,CKD-EPI₃M,CKD-EPI₄M）with the Gates method corrected by the dual plasma method in terms of bias,accuracy,precision,and consistency with the gold standard.2.Six weeks old Wistar rats were used as the research subjects.Chronic renal failure was induced by 5/6 nephrectomy（n=11）,and normal control group（n=5）was set up.In the two groups,GFR was measured using a transdermal GFR detection device,blood samples were collected from the tail vein to detect creatinine and urea nitrogen,inulin was injected continuously through femoral vein catheterization,blood and urine samples were collected at regular intervals to measure inulin clearance rate.The accuracy of transdermal GFR detection and the GFR estimation equation in rats were compared from four aspects of bias,accuracy,precision and consistency with inulin clearance rate.3.A total of 145 non-dialysis ESRD patients with hyperkalemia aged≥18 years who were admitted to China-Japan Friendship Hospital from January 2012 to February 2023 were enrolled.Hyperkalemia was defined as serum potassium>5.0mmol/l.The patients were randomly divided into training group（n=102,70%）and validation group（n=43,30%）using Python language in the Anaconda environment.The demographic and laboratory data were collected,and the 24-hour urinary potassium excretion calculation model was established by multi-layer perceptron of neural network and multiple linear regression.Using the 24-hour urine potassium measurement as the gold standard,the bias,accuracy,consistency and correlation between the newly established model and the previously published equations（Kawasaki and Tanaka equations）were compared.Results:1.A total of 132 patients were enrolled.The age was 58（48.25,68.00）years old,including 78 males（59.09%）and 54 females（40.91%）.The calibration values of revised Gates method and dual plasma method were 10.61（7.39,12.80）ml/min/1.73m2 and 13.10（11.40,14.30）ml/min/1.73m2,respectively.The FASScr,XangYa,XiangYa-s,CKD-EPISCysC,FASSCysC,FASScr-ScysC,Adachi and CKD-EPI₃M equations overestimated GFR when compared with the Gates method revised by dual plasma method.The remaining 15 equations underestimated GFR.The bias of Lund-Malm?Revised equation was low（-0.67,P<0.001）,the precision（interquartile range of bias=4.39）,accuracy（percentage of estimated GFR±30%of corrected Gates method=67.42%）and consistency with the revised Gates method（95%limits of agreement of Bland-Altman plot=12.54 ml/min/1.73m2）were high.2.The levels of serum cretinine and blood urea nitrogen in 5/6 nephrectomy group were significantly higher than those in normal group（serum cretinine:40.90 μmol/l and 104.10 μmol/l,blood urea nitrogen:6.77 mmol/l and 20.28 mmol/l,all P<0.001）.In the normal group,transdermal GFR（2.69 ml/min）and estimated GFR（0.68 ml/min）were lower than inulin clearance rate（2.76 ml/min）in varying degrees.Transdermal GFR measurement had low bias（0.10）,high accuracy（the percentage of transdermal GFR in the range of 70%-130%inulin clearance=100%）and high consistency with inulin clearance（95%limits of agreement of Bland-Altman plot--0.48 ml/min）.Transdermal GFR（0.64 ml/min）and estimated GFR（0.52 ml/min）in 5/6 nephrectomy group were higher than inulin clearance（0.43 ml/min）in varying degrees.The GFR equation had low bias（-0.14）,high accuracy（the proportion of GFR estimated by the equation within the inulin clearance rate of 70%-130%=81.82%）and high consistency with inulin clearance rate（95%limits of agreement of Bland-Altman plot-0.61 ml/min）.3.A total of 145 patients were included in this study.There were 95 males（66.15%）and 50 females（33.85%）,aged 54.72±14.76 years.The multiple linear regression equation is as follows:24h urine potassium（mmol/24h）=0.05 × age（y）-3.92 × height（m）+0.11 × weight（kg）+3.33 × GFR（ml/min/1.73m2）0.5+2.25 × urine potassium（mmol/l）0.5+3.49（for glomerular disease）;24h urine potassium（mmol/24h）=0.11 ×age（y）+29.48 × height（m）+0.27 × weight（kg）+1.82 × GFR（ml/min/1.73m2）0.5+1.50 × urine potassium（mmol/1）0.5-49.83（for renal tubular disease）.The 24-hour urine potassium calculated by multiple linear regression equation and neural network model was lower than the measured value,and the difference was not statistically significant（P>0.05）.The two new models had low bias（-0.02 and-0.57）,high accuracy（percentage of calculated values within±30%of measured values=83.45%and 84.14%）,correlation with measured values（Spearman correlation coefficient=0.72,intraclass correlation coefficient=0.67 and 0.70）and agreement with 24-hour urine potassium measurements（95%limits of agreement of Bland-Altman plot=13.70 mmol/24h and 13.20 mmol/24h）.Conclusion:1.In non-dialysis ESRD patients,the GFR measured by revised Gates method is in good agreement with that measured by dual plasma method.The GFR calculated by 23 GFR estimation equations was compared with the GFR measured by Gates method corrected by dual plasma method,and the accuracy was not very satisfactory.In comparison,the Lund-Malm? Revised equation had low bias,high precision and accuracy,and good consistency with the rrevised Gates method.Is the optimal equation for estimating GFR in non-dialysis ESRD patients.2.Compared with inulin clearance,there were some deviations in transdermal GFR measurement and estimated GFR equation in normal rats and 5/6 nephrectomy rats.In normal rats,compared with inulin clearance rate,transdermal GFR measurement has lower bias,higher accuracy and consistency with inulin clearance rate,and better accuracy than GFR estimation equation.In 5/6 nephrectomized rats,compared with inulin clearance,transdermal GFR measurement has large bias,low accuracy and consistency with inulin clearance,and lower accuracy than GFR estimation equation.The application value of transdermal GFR measurement in rats with severe renal failure needs to be further verified.The reason for the poor accuracy of transdermal GFR measurement in 5/6 nephrectomized rats may be due to the reduction of FITC-sinistrin fluorescence readings by the in vivo uremic toxin environment.3.In non-dialysis ESRD patients with hyperkalemia,the newly constructed neural network model and multiple linear regression equation for calculating 24-hour urinary potassium excretion developed by neural network multi-layer perceptron,Python language and multiple linear regression have lower bias,higher accuracy and correlation with 24-hour urine potassium measurement than the previously published equations,and have high clinical application value for the estimation of 24-hour urine potassium excretion in non-dialysis ESRD patients with hyperkalemia.