Development and validation of a more accurate estimating equation for glomerular filtration rate in a Chinese population

Previously published equations to estimate glomerular filtration rate (GFR) have limited accuracy in Asian populations. We aimed to develop and validate a more accurate equation for estimated GFR (eGFR) in the Chinese population, using data from 8571 adults who were referred for direct measurement of GFR by renal dynamic imaging (mGFR) at 3 representative hospitals in China. Patients from the Third Xiangya Hospital were included in our development (n=1730) and internal validation sets (n=1042) and patients from the other hospitals comprised the external validation set (n=5799). We excluded patients who were prescribed medications known to influence the tubular secretion of creatinine, patients on dialysis, kidney transplant recipients, and patients with missing creatinine values or with creatinine >700 μmol/l. We derived a novel eGFR equation by linear regression analysis and compared the performance to 12 creatinine-based eGFR equations, including previously published equations for use in Chinese or Asian populations. In the development and internal validation sets, the novel Xiangya equation had high accuracy (accuracy within 30% [P30], 79.21% and 84.33%, respectively), low bias (mean difference between mGFR and eGFR, -1.97 and -1.85 ml/min per 1.73 m2, respectively), and high precision (interquartile range of the differences, 21.13 and 18.88 ml/min per 1.73 m2, respectively). In external validation, the Xiangya equation had the highest P30 among all eGFR equations, with P30 ≤ 75% for the other 12 equations. This novel equation provides more accurate GFR estimates in Chinese adults and could replace existing eGFR equations for use in the Chinese population. Previously published equations to estimate glomerular filtration rate (GFR) have limited accuracy in Asian populations. We aimed to develop and validate a more accurate equation for estimated GFR (eGFR) in the Chinese population, using data from 8571 adults who were referred for direct measurement of GFR by renal dynamic imaging (mGFR) at 3 representative hospitals in China. Patients from the Third Xiangya Hospital were included in our development (n=1730) and internal validation sets (n=1042) and patients from the other hospitals comprised the external validation set (n=5799). We excluded patients who were prescribed medications known to influence the tubular secretion of creatinine, patients on dialysis, kidney transplant recipients, and patients with missing creatinine values or with creatinine >700 μmol/l. We derived a novel eGFR equation by linear regression analysis and compared the performance to 12 creatinine-based eGFR equations, including previously published equations for use in Chinese or Asian populations. In the development and internal validation sets, the novel Xiangya equation had high accuracy (accuracy within 30% [P30], 79.21% and 84.33%, respectively), low bias (mean difference between mGFR and eGFR, -1.97 and -1.85 ml/min per 1.73 m2, respectively), and high precision (interquartile range of the differences, 21.13 and 18.88 ml/min per 1.73 m2, respectively). In external validation, the Xiangya equation had the highest P30 among all eGFR equations, with P30 ≤ 75% for the other 12 equations. This novel equation provides more accurate GFR estimates in Chinese adults and could replace existing eGFR equations for use in the Chinese population. see commentary on page 494 see commentary on page 494 Kidney disease is a major medical and worldwide public health problem that has a high incidence and poses a heavy burden on the population.1Eckardt K.U. Coresh J. Devuyst O. et al.Evolving importance of kidney disease: from subspecialty to global health burden.Lancet. 2013; 382: 158-169Abstract Full Text Full Text PDF PubMed Scopus (746) Google Scholar The prevalence of chronic kidney disease (CKD, defined2National Kidney FoundationK/DOQI clinical practice guidelines for chronic kidney disease: evaluation, classification, and stratification.Am J Kidney Dis. 2002; 39: S1-S266PubMed Google Scholar by a reduction in glomerular filtration rate (GFR) o <60 ml/min per 1.73 m2) was reported to be 10.8% in 2012, equivalent to 119.5 million CKD patients in China.3Ning L. Lin W. Hu X. et al.Prevalence of chronic kidney disease in patients with chronic hepatitis B: a cross-sectional survey.J Viral Hepat. 2017; 24: 1043-1051Crossref PubMed Scopus (10) Google Scholar Acute kidney injury (AKI) was a complicating factor in 2.4%–8.1% of all hospital adult admissions for this population of patients, with an associated mortality rate4Yang L. Xing G. Wang L. et al.Acute kidney injury in China: a cross-sectional survey.Lancet. 2015; 386: 1465-1471Abstract Full Text Full Text PDF PubMed Scopus (284) Google Scholar of 18.6%–28.5%. Thus, quick and accurate assessment of changes in renal function is important, as is early detection of kidney damage, in time to provide successful treatment. The GFR is regarded as the most valuable indicator of overall renal function,2National Kidney FoundationK/DOQI clinical practice guidelines for chronic kidney disease: evaluation, classification, and stratification.Am J Kidney Dis. 2002; 39: S1-S266PubMed Google Scholar, 5Kilbride H.S. Stevens P.E. Eaglestone G. et al.Accuracy of the MDRD (Modification of diet in renal disease) study and CKD-EPI (CKD epidemiology collaboration) equations for estimation of GFR in the elderly.Am J Kidney Dis. 2013; 61: 57-66Abstract Full Text Full Text PDF PubMed Scopus (195) Google Scholar, 6Lamb E.J. Stevens P.E. Deeks J.J. What is the best glomerular filtration marker to identify people with chronic kidney disease most likely to have poor outcomes?.BMJ. 2015; 350: g7667Crossref PubMed Scopus (9) Google Scholar and accurate estimation of GFR facilitates better identification, prognostication, and management of kidney disease. To date, measuring GFR typically has relied on renal clearance of either exogenous filtration markers, such as inulin,7Macedo E. Mehta R.L. Measuring renal function in critically ill patients: tools and strategies for assessing glomerular filtration rate.Curr Opin Crit Care. 2013; 19: 560-566PubMed Google Scholar iohexol,5Kilbride H.S. Stevens P.E. Eaglestone G. et al.Accuracy of the MDRD (Modification of diet in renal disease) study and CKD-EPI (CKD epidemiology collaboration) equations for estimation of GFR in the elderly.Am J Kidney Dis. 2013; 61: 57-66Abstract Full Text Full Text PDF PubMed Scopus (195) Google Scholar and 125I-iothalamate,8Inker L.A. Tighiouart H. Coresh J. et al.GFR estimation using β-trace protein and β2-microglobulin in CKD.Am J Kidney Dis. 2016; 67: 40-48Abstract Full Text Full Text PDF PubMed Scopus (97) Google Scholar or endogenous markers including creatinine9Levey A.S. Coresh J. Greene T. et al.Using standardized serum creatinine values in the modification of diet in renal disease study equation for estimating glomerular filtration rate.Ann Intern Med. 2006; 145: 247-254Crossref PubMed Scopus (4112) Google Scholar and cystatin C.10Lamb E.J. Cystatin C: why clinical laboratories should be measuring it.Ann Clin Biochem. 2015; 52: 709-711Crossref PubMed Scopus (5) Google Scholar Using inulin as a marker is the “gold standard” method, but it is used only in research work because it is expensive and difficult to assay and requires continuous infusion and repeated blood sampling.8Inker L.A. Tighiouart H. Coresh J. et al.GFR estimation using β-trace protein and β2-microglobulin in CKD.Am J Kidney Dis. 2016; 67: 40-48Abstract Full Text Full Text PDF PubMed Scopus (97) Google Scholar Technetium-99m-diethylenetriaminepentaacetic acid (99mTc-DTPA) renal dynamic imaging (Gate’s method11Gates G.F. Split renal function testing using Tc-99m DTPA. A rapid technique for determining differential glomerular filtration.Clin Nucl Med. 1983; 8: 400-407Crossref PubMed Scopus (175) Google Scholar) is recommended for measurement of GFR by the Nephrology Committee of the Society of Nuclear Medicine12Blaufox M.D. Aurell M. Bubeck B. et al.Report of the radionuclides in Nephrourology Committee on renal clearance.J Nucl Med. 1996; 37: 1883-1890PubMed Google Scholar and has been widely accepted in clinical practice.13Claudon M. Durand E. Grenier N. et al.Chronic urinary obstruction: evaluation of dynamic contrast-enhanced MR urography for measurement of split renal function.Radiology. 2014; 273: 801-812Crossref PubMed Scopus (28) Google Scholar, 14Inoue Y. Itoh H. Tagami H. et al.Measurement of renal depth in dynamic renal scintigraphy using ultralow-dose CT.Clin Nucl Med. 2016; 41: 434-441Crossref PubMed Scopus (8) Google Scholar In addition, 99mTc-DTPA measurement is the standard value of GFR used in most Chinese estimated GFR (eGFR) equations15Pei X. Yang W. Wang S. et al.Using mathematical algorithms to modify glomerular filtration rate estimation equations.Plos One. 2013; 8: e57852Crossref PubMed Scopus (18) Google Scholar, 16Liu X. Gan X. Chen J. et al.A new modified CKD-EPI equation for Chinese patients with type 2 diabetes.Plos One. 2014; 9: e109743Crossref PubMed Scopus (23) Google Scholar, 17Chang H. Ye S.D. Application of modified glomerular filtration rate estimation equations in Chinese diabetic patients with chronic kidney diseases.West Indian Med J. 2015; 64: 209-212PubMed Google Scholar; it has been found to have a renal clearance rate close to that of inulin, with a degree of consistency of 0.99 over wide ranges of GFR, confirming that the renal system handling of these is similar.18Itoh K. Comparison of methods for determination of glomerular filtration rate: Tc-99m-DTPA renography, predicted creatinine clearance method and plasma sample method.Ann Nucl Med. 2003; 17: 561-565Crossref PubMed Scopus (85) Google Scholar However, 99mTc-DTPA is costly and cumbersome to use, and it is not readily available in most clinical settings in China.19Mazza A. Rampin L. Montemurro D. et al.Renal scintigraphy predicts global cardiovascular risk in hypertensive subjects with normal serum creatinine levels.Blood Press. 2011; 20: 387-393Crossref PubMed Scopus (3) Google Scholar Thus, assessing GFR values indirectly by using eGFR equations has become more common. A search of PubMed articles related to eGFR equations indicates that about 84 such equations have been developed to date. Of these, 69 were developed in whites and blacks,20Soveri I. Berg U.B. Björk J. et al.Measuring GFR: a systematic review.Am J Kidney Dis. 2014; 64: 411-424Abstract Full Text Full Text PDF PubMed Scopus (308) Google Scholar, 21Shaffi K. Uhlig K. Perrone R.D. et al.Performance of creatinine-based GFR estimating equations in solid-organ transplant recipients.Am J Kidney Dis. 2014; 63: 1007-1018Abstract Full Text Full Text PDF PubMed Scopus (97) Google Scholar, 22Cockcroft D.W. Gault M.H. Prediction of creatinine clearance from serum creatinine.Nephron. 1976; 16: 31-41Crossref PubMed Scopus (13131) Google Scholar, 23Levey A.S. Bosch J.P. Lewis J.B. et al.A more accurate method to estimate glomerular filtration rate from serum creatinine: a new prediction equation. Modification of Diet in Renal Disease Study Group.Ann Intern Med. 1999; 130: 461-470Crossref PubMed Scopus (13031) Google Scholar, 24Levey A.S. Stevens L.A. Schmid C.H. et al.A new equation to estimate glomerular filtration rate.Ann Intern Med. 2009; 150: 604-612Crossref PubMed Scopus (16096) Google Scholar, 25Inker L.A. Schmid C.H. Tighiouart H. et al.Estimating glomerular filtration rate from serum creatinine and cystatin c.N Engl J Med. 2012; 367: 20-29Crossref PubMed Scopus (2507) Google Scholar, 26MacIsaac R.J. Tsalamandris C. Thomas M.C. et al.Estimating glomerular filtration rate in diabetes: a comparison of cystatin-C- and creatinine-based methods.Diabetologia. 2006; 49: 1686-1689Crossref PubMed Scopus (132) Google Scholar, 27Schaeffner E.S. Ebert N. Delanaye P. et al.Two novel equations to estimate kidney function in persons aged 70 years or older.Ann Intern Med. 2012; 157: 471-481Crossref PubMed Scopus (425) Google Scholar, 28Levey A.S. Greene T. Kusek J.W. et al.A simplified equation to predict glomerular filtration rate from serum creatinine.J Am Soc Nephrol. 2000; 11 (abstr A0828): 155AGoogle Scholar whereas 15 were established15Pei X. Yang W. Wang S. et al.Using mathematical algorithms to modify glomerular filtration rate estimation equations.Plos One. 2013; 8: e57852Crossref PubMed Scopus (18) Google Scholar, 16Liu X. Gan X. Chen J. et al.A new modified CKD-EPI equation for Chinese patients with type 2 diabetes.Plos One. 2014; 9: e109743Crossref PubMed Scopus (23) Google Scholar, 17Chang H. Ye S.D. Application of modified glomerular filtration rate estimation equations in Chinese diabetic patients with chronic kidney diseases.West Indian Med J. 2015; 64: 209-212PubMed Google Scholar, 29Ma Y.C. Zuo L. Chen J.H. et al.Modified glomerular filtration rate estimating equation for Chinese patients with chronic kidney disease.J Am Soc Nephrol. 2006; 17: 2937-2944Crossref PubMed Scopus (1444) Google Scholar, 30Wang J. Xie P. Huang J. et al.The new Asian modified CKD-EPI equation leads to more accurate GFR estimation in Chinese patients with CKD.Int Urol Nephrol. 2016; 48: 2077-2081Crossref PubMed Scopus (28) Google Scholar, 31Leung T.K. Luk A.O. So W.Y. et al.Development and validation of equations estimating glomerular filtration rates in Chinese patients with type 2 diabetes.Kidney Int. 2010; 77: 729-735Abstract Full Text Full Text PDF PubMed Scopus (11) Google Scholar, 32Liao Y. Liao W. Liu J. et al.Assessment of the CKD-EPI equation to estimate glomerular filtration rate in adults from a Chinese CKD population.J Int Med Res. 2011; 39: 2273-2280Crossref PubMed Scopus (25) Google Scholar, 33Teo B.W. Xu H. Wang D. et al.GFR estimating equations in a multiethnic Asian population.Am J Kidney Dis. 2011; 58: 56-63Abstract Full Text Full Text PDF PubMed Scopus (196) Google Scholar in Asia. Particularly, the Cockcroft–Gault (C-G) and the modification of diet in renal disease (MDRD) study equations are those most routinely applied to GFR estimation in clinical practice, as endorsed by the Kidney Diseases Outcomes Quality Initiative (K/DOQI).34Michels W.M. Grootendorst D.C. Verduijn M. et al.Performance of the Cockcroft-Gault, MDRD, and new CKD-EPI formulas in relation to GFR, age, and body size.Clin J Am Soc Nephrol. 2010; 5: 1003-1009Crossref PubMed Scopus (377) Google Scholar A later, simplified Chinese version of MDRD, known as the c-MDRD29Ma Y.C. Zuo L. Chen J.H. et al.Modified glomerular filtration rate estimating equation for Chinese patients with chronic kidney disease.J Am Soc Nephrol. 2006; 17: 2937-2944Crossref PubMed Scopus (1444) Google Scholar equation, performed better than the original version in the Chinese population,32Liao Y. Liao W. Liu J. et al.Assessment of the CKD-EPI equation to estimate glomerular filtration rate in adults from a Chinese CKD population.J Int Med Res. 2011; 39: 2273-2280Crossref PubMed Scopus (25) Google Scholar especially for CKD.29Ma Y.C. Zuo L. Chen J.H. et al.Modified glomerular filtration rate estimating equation for Chinese patients with chronic kidney disease.J Am Soc Nephrol. 2006; 17: 2937-2944Crossref PubMed Scopus (1444) Google Scholar Recently, the Chronic Kidney Disease Epidemiology Collaboration creatinine (CKD-EPIcr) equation24Levey A.S. Stevens L.A. Schmid C.H. et al.A new equation to estimate glomerular filtration rate.Ann Intern Med. 2009; 150: 604-612Crossref PubMed Scopus (16096) Google Scholar was also recommended in the 2012 clinical practice guideline.35Stevens P.E. Levin A. Evaluation and management of chronic kidney disease: synopsis of the kidney disease: improving global outcomes 2012 clinical practice guideline.Ann Intern Med. 2013; 158: 825-830Crossref PubMed Scopus (1711) Google Scholar However, accuracy of the above equations was limited when applied to clinical practice. A study36Melloni C. Peterson E.D. Chen A.Y. et al.Cockcroft-Gault versus modification of diet in renal disease.J Am Coll Cardiol. 2008; 51: 991-996Crossref PubMed Scopus (114) Google Scholar from the United States that included 46,942 patients who had acute coronary syndrome revealed that their GFR levels were partly overestimated by the C-G and MDRD equations (median eGFR was 53.2 and 65.8 ml/min per 1.73 m2 as determined by the C-G and MDRD equations, respectively). The dosage of antithrombotic agents was increased based on these equations, which ultimately increased the risk of major bleeding, which occurred in 17.5% who received an excess dose increase (compared with 8.5% who received no excess increase) based on the C-G equation, and 16.1% who received an excess increase (compared with 9.3% who received no excess increase) based on the MDRD equation. Chinese renal physicians choose renal dynamic imaging as the preferred method of accurately assessing renal function.13Claudon M. Durand E. Grenier N. et al.Chronic urinary obstruction: evaluation of dynamic contrast-enhanced MR urography for measurement of split renal function.Radiology. 2014; 273: 801-812Crossref PubMed Scopus (28) Google Scholar, 14Inoue Y. Itoh H. Tagami H. et al.Measurement of renal depth in dynamic renal scintigraphy using ultralow-dose CT.Clin Nucl Med. 2016; 41: 434-441Crossref PubMed Scopus (8) Google Scholar This may be because the eGFR equations based on whites and blacks are not suitable for Chinese patients, due to racial differences, and the equations developed from an Asian population are based on a small sample and have not gotten wide clinical approval. But renal dynamic imaging is not accessible to many clinicians, so we aimed to develop a novel and optimal eGFR equation based on multicenter clinical practice among a large-scale Chinese population. Among the initial 4946 inpatients and 746 outpatients referred to 99mTc-DTPA renal dynamic imaging measurement in the Third Xiangya Hospital of Central South University (TXH), 79 inpatients and 16 outpatients were aged <18 years. For inpatients and outpatients, respectively, 615 and 14 had undergone dialysis previously, 19 and 7 had incomplete data for mGFR (including those with unilateral renal agenesis, such as transplant recipients), 25 and 392 lacked creatinine values, 1692 and 3 were taking drugs that reduced creatinine levels within 10 days prior to GFR measurement, and 44 and 14 had serum creatinine levels >700 μmol/l. After the above-mentioned participants were excluded (further details regarding excluded patients are available in Supplementary Table S1), 2472 inpatients and 300 outpatients were enrolled in the final analysis. Inpatients were randomly divided into 2 data sets, with a ratio of 7:3. A random sample of 1730 patients was included in the development sample, and the remaining participants were included in the internal validation data set (Supplementary Figure S1). Outpatients were also used for the internal validation. Characteristics of these individuals from this center are summarized in Table 1. In the TXH center, for the inpatient cohort, mean mGFR was 71.03 ± 23.99 ml/min per 1.73 m2, mean age was 52.71 ± 13.25 years, and 51.86% were male. For outpatients at this center, the mean mGFR was 75.49 ± 23.04 ml/min per 1.73 m2, mean age was 49.52 ± 9.47 years, and 41.67% were male. Inpatients from the Second Xiangya Hospital of Central South University (SXH; n = 532; mean mGFR 71.86 ± 36.49 ml/min per 1.73 m2; mean age 52.10 ± 12.85 years; 51.80% male) and the First Affiliated Hospital of Xinjiang Medical University (FXH; n = 5004; mean mGFR 87.44 ± 27.31 ml/min per 1.73 m2; mean age 52.09 ± 14.27 years; 60.15% male), and outpatients from FXH (n = 263; mean mGFR 69.50 ± 28.64 ml/min per 1.73 m2; mean age 52.54 ± 15.41 years; 55.51% male) were used for external validation (Supplementary Table S2).Table 1Characteristics of the study population in the Third Xiangya Hospital of Central South University, ChinaaValues are mean of means (SD), unless otherwise indicated.CharacteristicEntire cohortDevelopment sampleInternal validation sampleInpatients (n = 2472)Outpatients (n = 300)Inpatients (n = 1730)Inpatients (n = 742)Outpatients (n = 300)Age (yr)52.71 (13.25)49.52 (9.47)52.34 (13.23)53.57 (13.26)49.52 (9.47)Male (n [%])1282 (51.86)125 (41.67)892 (51.56)391 (52.70)125 (41.67)Height (m)1.59 (7.5)1.60 (5.00)1.60 (7.59)1.58 (7.28)1.60 (5.00)Weight (kg)60.28 (13.33)58.79 (4.50)60.69 (13.43)58.77 (13.08)58.79 (4.50)CKD (n [%])775 (31.35)32 (10.67)533 (30.81)242 (32.61)32 (10.67)Malignant tumor (n [%])362 (14.64)3 (1.00)265 (15.32)97 (13.07)3 (1.00)AKI (n [%])1 (0.04)0 (0)1 (0.06)0 (0)0 (0)Nephrotic syndrome (n [%])268 (10.84)0 (0)183 (10.58)85 (11.46)0 (0)Mean Scr level (μmol/l)119.25 (97.25)100.71 (89.37)119.56 (98.57)118.49 (94.08)100.71 (89.37)Mean mGFR (ml/min per 1.73 m2)bTo convert Scr from μmol/l to mg/dl, divide by 88.4.71.03 (23.99)75.49 (23.04)71.32 (23.96)70.40 (24.05)75.49 (23.04)mGFR, ml/min per 1.73 m2 (n [%]) <30126 (5.10)17 (5.67)89 (5.14)37 (4.99)17 (5.67) 31–44225 (9.10)17 (5.67)146 (8.44)79 (10.65)17 (5.67) 45–59424 (17.15)32 (10.66)298 (17.23)126 (16.98)32 (10.66) 60–891174 (47.49)162 (54.00)827 (47.80)347 (46.77)162 (54.00) ≥90523 (21.16)72 (24.00)370 (21.39)154 (20.75)72 (24.00)Carbon dioxide binding (mmol/l)22.89 (5.14)—22.82 (5.22)23.05 (4.96)—Anion gap (mmol/l)16.2 (4.57)—16.14 (4.61)16.35 (4.47)—Alanine aminotransferase (U/l)23.48 (26.1)—23.29 (27.04)23.95 (23.77)—Aspartate aminotransferase (U/l)22.91 (20.4)—22.81 (22.85)23.16 (12.98)—Total bilirubin (μmol/l)12.52 (5.69)—12.62 (5.9)12.3 (5.15)—Total protein (g/l)67.55 (8.22)—67.37 (8.34)67.97 (7.93)—White globulin ratio1.46 (0.35)—1.46 (0.35)1.46 (0.35)—Direct bilirubin (μmol/l)3.61 (2.48)—3.67 (2.61)3.48 (2.15)—Globulin (g/l)28 (5.65)—27.9 (5.7)28.22 (5.54)—Total bile acid (μmol/l)4.92 (6.49)—4.93 (6.83)4.9 (5.62)—High-density lipoprotein cholesterol (mmol/l)1.01 (0.64)—1.01 (0.65)1.01 (0.64)—Low-density lipoprotein cholesterol (mmol/l)1.97 (1.28)—1.96 (1.27)2 (1.29)—Serum triglycerides (mmol/l)1.21 (1.2)—1.2 (1.2)1.23 (1.22)—High-density cholesterol and total cholesterol ratio0.22 (0.14)—0.22 (0.14)0.22 (0.14)—Serum total cholesterol (mmol/l)3.63 (2.18)—3.62 (2.18)3.66 (2.17)—Prothrombin time (s)9.18 (4.78)—9.24 (4.77)9.04 (4.8)—Activated partial thromboplastin time (s)23.32 (13.46)—23.51 (13.54)22.89 (13.26)—Thrombin time (s)13.39 (7.19)—13.41 (7.15)13.37 (7.3)—Albumin (g/l)39.66 (6.48)—39.56 (6.58)39.91 (6.23)—White blood cells (×109/l)7.69 (15.26)—7.91 (17.5)7.17 (7.81)—Monocytes (%)6.69 (2.22)—6.72 (2.27)6.64 (2.1)—Monocytes (absolute value, ×109/l)0.46 (0.21)—0.46 (0.21)0.44 (0.19)—Red blood cells (×1012/l)4.17 (0.76)—4.16 (0.77)4.2 (0.74)—Red blood cell distribution width (%)10.08 (5.94)—10.05 (5.98)10.14 (5.84)—Hematocrit (%)37.85 (6.88)—37.69 (6.98)38.24 (6.63)—Lymphocytes (%)26.06 (9.68)—25.73 (9.78)26.84 (9.45)—Lymphocytes (absolute value, ×109/l)1.69 (0.63)—1.68 (0.64)1.72 (0.61)—Potassium (mmol/l)4 (0.73)—4 (0.74)4.01 (0.71)—Calcium (mmol/l)2.25 (0.37)—2.25 (0.38)2.27 (0.35)—Sodium (mmol/l)138.06 (20.43)—137.88 (20.85)138.47 (19.39)—Urea (mmol/l)6.47 (3.87)—6.48 (3.96)6.45 (3.66)—Uric acid (mmol/l)343.3 (111.38)—340.88 (111.71)349.46 (111.29)—Mean erythrocyte hemoglobin content (pg)29.61 (3.4)—29.54 (3.54)29.76 (3.04)—Mean erythrocyte hemoglobin concentration (g/l)325.44 (27.88)—325.19 (28.89)326.01 (25.37)—Basophils (%)0.46 (0.37)—0.46 (0.39)0.46 (0.32)—Basophils (absolute value, ×109/l)0.03 (0.03)—0.03 (0.03)0.03 (0.03)—Eosinophils (%)3.11 (2.81)—2.99 (2.66)3.4 (3.11)—Eosinophils (absolute value, ×109/l)0.21 (0.2)—0.2 (0.2)0.22 (0.22)—Hemoglobin (g/l)124.02 (23.9)—123.47 (24.23)125.34 (23.09)—Blood glucose (mmol/l)4.43 (2.3)—4.43 (2.37)4.43 (2.13)—Platelet hematocrit (%)0.19 (0.12)—0.19 (0.12)0.19 (0.12)—Platelet distribution width (fL)11.61 (6.5)—11.61 (6.52)11.61 (6.47)—Mean platelet volume (fL)10.17 (2.15)—10.16 (2.15)10.2 (2.15)—Neutrophils (%)63.01 (11.71)—63.37 (11.86)62.13 (11.32)—Neutrophils (absolute value, ×109/l)4.56 (2.61)—4.63 (2.72)4.39 (2.31)—AKI, acute kidney injury; CKD, chronic kidney disease; GFR, glomerular filtration rate; mGFR, measured GFR; Scr, serum creatinine.a Values are mean of means (SD), unless otherwise indicated.b To convert Scr from μmol/l to mg/dl, divide by 88.4. Open table in a new tab AKI, acute kidney injury; CKD, chronic kidney disease; GFR, glomerular filtration rate; mGFR, measured GFR; Scr, serum creatinine. To evaluate the performance of the model, we focused on R2, adjusted R2, Akaike’s information criterion (AIC), and especially, clinical applicability. In general, the model with the smallest AIC value is preferred.37Shipley B. The AIC model selection method applied to path analytic models compared using a d-separation test.Ecology. 2013; 94: 560-564Crossref PubMed Scopus (308) Google Scholar However, we found that the model with 3 variables had an adjusted R2 similar to that of the original model which had 15 variables (0.5720 vs. 0.5985), suggesting that they have similar prediction accuracy, although the AIC of the model with 3 variables was much larger. In addition, the model with 3 variables has greater clinical applicability than other models. Therefore, the simplest model, which includes serum creatinine level, gender, and age, was selected as the best-performing equation for general use (Supplementary Table S3 and Supplementary Figure S2). Notably, these 3 variables were also included in most equations, to reflect the impact of the associated factors. Because the equation was developed based on data from the TXH, it is called “the Xiangya equation.” Table 2 shows the Xiangya equation in a form that could be implemented in clinical laboratories. No further variables were included in the equation (further details are available in Supplementary Appendix S1).Table 2The new equation for estimating GFR on the natural scale in a Chinese populationaExpressed for specified mGFR level and sex. Serum creatinine was measured in μmol/l. To convert GFR from ml/min per 1.73 m2 to ml/s per 1.73 m2, multiply by 0.0167. To convert Scr from μmol/l to mg/dl, divide by 88.4.NameApplicable populationSexEquationXiangyaChineseMaleeGFR = 2374.78 × Scr –0.54753 × Age –0.25011FemaleeGFR = 2374.78 × Scr –0.54753 × Age –0.25011 × 0.8526126eGFR, estimated glomerular filtration rate; GFR, glomerular filtration rate; mGFR, measured GFR; Scr, serum creatinine.a Expressed for specified mGFR level and sex. Serum creatinine was measured in μmol/l. To convert GFR from ml/min per 1.73 m2 to ml/s per 1.73 m2, multiply by 0.0167. To convert Scr from μmol/l to mg/dl, divide by 88.4. Open table in a new tab eGFR, estimated glomerular filtration rate; GFR, glomerular filtration rate; mGFR, measured GFR; Scr, serum creatinine. In the TXH center (Supplementary Table S4), the mean mGFRs of the inpatients in the entire cohort, development sample, and internal-validation data set were71.03 ± 23.99, 71.32 ± 23.96, and 70.40 ± 24.05 ml/min per 1.73 m2, respectively. The percentage of eGFR deviation of <30% from mGFR (P30) was 79.21%, 79.42%, 78.42%, and 84.33%, respectively. Overall, the new Xiangya equation met the criterion of a P30 ≥75%, which is considered sufficient for good clinical decision-making according to the 2002 K/DOQI benchmark2National Kidney FoundationK/DOQI clinical practice guidelines for chronic kidney disease: evaluation, classification, and stratification.Am J Kidney Dis. 2002; 39: S1-S266PubMed Google Scholar (Table 3). This indicator reached 91.74% and 83.37% in individuals with 60 ≤ mGFR < 90 ml/min per 1.73 m2 and mGFR ≥ 90 ml/min per 1.73 m2, respectively (Supplementary Table S5). The Xiangya equation also worked well for both genders (P30 of 77.69% for males and 80.84% for females, respectively; Supplementary Table S6). In addition, the equation produced almost no bias (bias –0.20 ml/min per 1.73 m2) in elderly patients (Supplementary Table S7). For outpatients (Table 4), the mean mGFR in the internal-validation data set was 75.49 ± 23.04 ml/min per 1.73 m2, and the Xiangya equation still had good accuracy (mainly, a P30 of 86.55%) in this center. In general, the development and internal validation results showed that the Xiangya equation performed very well.Table 3Performance of eGFR equations for inpatients of Third Xiangya Hospital of Central South University, ChinaaeGFR, bias, and IQR are given as ml/min per 1.73 m2. To convert GFR from ml/min per 1.73 m2 to ml/s per 1.73 m2, multiply by 0.0167. To convert Scr from μmol/l to mg/dl, divide by 88.4.EquationPerformance evaluation (mGFR = 71.03 ml/min per 1.73 m2)eGFRBiasbBias refers to mGFR minus eGFR.IQRcIQR refers to the 25th–75th percentile.P30 (%)dP30 refers to percentage of GFR estimates that are within 30% of mGFR.RMSEC-G72.131.0928.3464.8924.26MDRD68.60–2.4326.2166.4622.93a-MDRD72.911.8827.8965.5324.25c-MDRD89.9018.8736.4647.4930.15MDRD(CN)78.417.3832.5058.9428.34CKD-EPIcr72.141.1127.9366.1021.51CKD-EPI(CN)79.368.3231.1458.2123.55Asian modified CKD-EPI83.6112.5735.1852.4324.94New modified CKD-EPI76.885.8426.7965.9820.54Chinese MDRD 677.296.2532.2659.4328.43New modified MDRD64.13–6.9021.3175.0817.98Ruijin70.83–0.2024.3872.6120.79Xiangya69.06–1.9721.1379.2116.84a-MDRD, abbreviated Modification of Diet in Renal Disease; C-G, Cockcroft–Gault; CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration; CKD-EPI(CN), Chinese improved Chronic Kidney Disease Epidemiology Collaboration equation; CKD-EPIcr, Chronic Kidney Disease Epidemiology Collaboration creatinine; c-MDRD, Chinese–Modification of Diet in Renal Disease; eGFR, estimated glomerular filtration rate; GFR, glomerular filtration rate; IQR, interquartile range; MDRD, Modification of Diet in Renal Disease; MDRD(CN), Chinese improved Modification of Diet in Renal Disease equation; mGFR, measured glomerular filtration rate; RMS