Serum and urinary markers of early impairment of GFR in chronic kidney disease patients: diagnostic accuracy of urinary -trace protein

Am J Physiol Renal Physiol 299: F1407 F1423, 2010. First published September 15, 2010; doi:10.1152/ajprenal.00507.2009. Serum and urinary markers of early impairment of GFR in chronic kidney disease patients: diagnostic accuracy of urinary -trace protein Carlo Donadio Department of Internal Medicine, Division of Nephrology, University of Pisa, Pisa, Italy Submitted 28 August 2009; accepted in final form 8 September 2010 Donadio C. Serum and urinary markers of early impairment of GFR in chronic kidney disease patients: diagnostic accuracy of urinary -trace protein. Am J Physiol Renal Physiol 299: F1407 F1423, 2010. First published September 15, 2010; doi:10.1152/ajprenal.00507.2009. The screening for chronic kidney diseases (CKD) patients with impaired GFR needs the measurement of serum creatinine (SCr) or cystatin C (SCys). GFR can also be predicted from SCr or SCys with different formulas. The aim of this study, performed in a group of CKD patients with different levels of GFR, was to evaluate the possibility to select the patients with a GFR 90 ml min 1 1.73 m 2 by means of serum levels and urinary excretion of different low-molecular-weight proteins (LMWP), cystatin C (Cys), 2-microglobulin ( 2M), retinolbinding protein (RBP), -trace protein (BTP), and derived prediction equations for GFR. In the 295 CKD patients (137 women), at all stages of GFR impairment a very high correlation was found between GFR ( 99m Tc-DTPA) and serum Cr, Cys, 2M, and BTP. All these serum markers showed a similar accuracy as indicators of different GFR impairments. RBP had the lowest correlation with GFR and was also significantly less accurate. The different prediction formulas derived from gender, anthropometric data and SCr or S-LMWP had a diagnostic accuracy similar to that of serum Cr, Cys, 2M, and BTP. Urinary albumin was inadequate as an indicator of any level of GFR impairment. Urinary excretion of Cys and 2M increased significantly only in patients with a GFR 30 ml min 1 1.73 m 2, while urinary BTP increased already at GFR 90 ml min 1 1.73 m 2. In this selected group of CKD patients, the positive predictive value of urinary BTP for a GFR 90 ml min 1 1.73 m 2 was 85%, indicating that, in CKD patients, a urine-based test can predict a slight GFR impairment. glomerular filtration rate; screening for GFR impairment; low-molecular-weight proteins; sensitivity and specificity EARLY DETECTION OF RENAL DISEASE and screening for early impairment of renal function could allow for slowing of the rate of progression of the impairment of renal function in chronic kidney disease (CKD) patients. The measurement of the glomerular filtration rate (GFR) is the gold standard for the assessment of renal function. GFR can be measured as the clearance of inulin or other suitable tracers like 99m Tc-DTPA and 51 Cr-EDTA (5, 56). None of these methods is adequate for routine clinical use or for screening purposes. Twenty-four-hour creatinine clearance (24h-CCr) is frequently used for the evaluation of renal function. However, 24h-CCr lacks both precision and accuracy (13, 19, 21, 68). Furthermore, due to the necessity of a 24-h urine collection, 24h-CCr is inadequate for screening studies. The major limitation of serum creatinine (SCr) is its low sensitivity as an indicator of early impairment of GFR, since Address for reprint requests and other correspondence: C. Donadio, Dept. of Internal Medicine-Nephrology, Univ. of Pisa, I-56100 Pisa, Italy (e-mail: c.donadio@med.unipi.it). SCr overcomes the upper limit of reference range only in patients at stages 3b (GFR 45 ml min 1 1.73 m 2 ). Furthermore, SCr, besides the level of GFR, is influenced by the amount of muscle mass and, as a consequence, by age, gender, and nutritional status of patients (14). To overcome some of the problems linked to the use of 24h-CCr or SCr as indicators of GFR, it has been proposed to predict GFR by different formulas based on SCr and anthropometric data (11, 34 36, 41, 43). Their competence to evaluate an early impairment of GFR is debatable (63). Furthermore, the standardization of SCr measurements becomes mandatory to avoid differences in the results of prediction formulas due to interlaboratory variability in the measurement of SCr (40, 59, 64, 67). Finally, SCr-based prediction formulas should be validated in the different ethnic groups and in patients with different body composition, such as obese and malnourished patients (12, 15, 20, 34, 58, 61). Different low-molecular-weight proteins (LMWP), with a molecular weight in the range 10 25 kda, have renal handling compatible with that of an ideal marker of GFR. In fact, they are cleared by the plasma through free glomerular filtration, subsequent complete tubular resorption, and degradation inside tubular cells (6, 42). As a consequence, their serum concentrations increase progressively with the reduction of GFR. Furthermore, age, gender, and body composition have a low influence on serum concentrations of LMWP. Due to this behavior, the measurement of serum concentrations of various LMWP has been proposed as a useful tool for evaluating an impairment of GFR, possibly more sensitive than SCr (8, 22, 33, 38, 47, 53, 62, 66). In normal subjects, the urinary excretion of LMWP, due to their extensive tubular reabsorption after glomerular filtration, is almost undetectable. On the contrary, an increased urinary excretion of some LMWP occurs when proximal tubules are damaged or when the filtered load to single nephrons overcomes the tubular resorptive capacity (2, 42). Either a relevant increase in the production of LMWP or a severe reduction in renal filtration of LMWP may increase their serum concentrations and hence their filtered load to single nephrons. Indeed, it is now known that urinary excretion of some LMWP increases in patients with end-stage renal disease. However, the precise relationship between urinary excretion of LMWP and the level of GFR has not yet been examined. The aims of this study, performed in a group of CKD patients with different impairment of GFR, were to 1) assess the precise relationship between the level of GFR of serum levels and urinary excretions of cystatin C [Cys; molecular weight (MW) 13.3], 2-microglobulin ( 2M; MW 11.8), retinol-binding protein (RBP; MW 21.2), and -trace protein (BTP; MW 18.5) compared with SCr and urinary excretion of albumin (U-Alb); 2) derive prediction equations for GFR based http://www.ajprenal.org 0363-6127/10 Copyright 2010 the American Physiological Society F1407

F1408 URINARY BTP TO SCREEN EARLY GFR IMPAIRMENT 90 ml min 1 1.73 m 2 by means of these different tests of GFR impairment. METHODS Fig. 1. Flow diagram of the 435 patients that entered the study. CKD, chronic kidney disease. on SCr, SCys, S 2M, SRBP, SBTP, and their combination; 3) validate these new prediction equations compared with published equations for GFR based on SCr and SCys; and 4) evaluate the possibility to select CKD patients with a GFR Patient recruitment and selection. The setting of the study was the laboratory for the functional evaluation of kidney disease of the Division of Nephrology at the University of Pisa. Patients were referred to our laboratory from nephrology and internal medicine clinics at the Pisa Teaching Hospital for functional evaluation of chronic nephropathies diagnosed on the basis of history of renal disease, presence of morphological or laboratory markers of kidney disease, and level of predicted GFR, according to National Kidney Foundation Kidney-Kidney Disease Outcomes Quality Initiative (NKF K-DOQI) guidelines (46). Inclusion criterion were age 18 yr and diagnosis of CKD at any stage. Exclusion criteria were the diagnosis of primary tubular disease, in particular Dent s disease and related syndromes, and recent or concurrent administration of potentially nephrotoxic drugs, like aminoglycosides, iodinated contrast media, and platinumbased chemotherapy. The flow diagram of the 435 eligible patients is reported in Fig. 1 (9). Seventy-three of 435 were excluded from the study: 62 patients did not fulfill criteria for the diagnosis of CKD, while 11 patients were in treatment with platinum-based combination chemotherapy for the treatment of ovarian cancer. No patient was excluded due to primary tubular disease or to contrast media or aminoglycoside administration. Sixty-seven other patients had an inadequate collection of blood or urine samples for the determination of all index tests. The data of the remaining 295 adult CKD patients, affected by different kidney diseases with various degree of impairment of renal function (SCr 0.40 12.1 mg/dl), are analyzed in the present study. The main anthropometric and clinical data of the Table 1. Main anthropometric and clinical data of patients All CKD Stages Stage 1 Stage 2 Stage 3a Stage 3b Stage 4 Stage 5 All patients, n 295 31 80 42 50 48 44 GFR, ml min 1 1.73 m 2 50.5 31.0 109.6 14.6 73.2 8.3 53.4 4.1 37.8 4.5 22.9 4.6 9.5 3.3 Age, yr 52.4 15.6 37.5 13.7 47.8 14.5 57.4 13.9 56.1 15.2 58.0 12.9 56.4 14.7 Body weight, kg 71.6 15.3 68.8 16.5 70.6 14.3 70.8 14.9 70.8 14.7 77.7 17.4 70.8 14.0 Height, cm 163.8 10.0 162.7 10.1 163.8 10.0 163.7 9.1 161.4 10.5 166.0 9.7 164.6 10.7 Body mass index, kg/m 2 26.6 4.6 25.1 5.1 26.2 4.5 26.3 4.7 27.1 4.6 28.1 4.5 26.0 4.15 Serum creatinine, mg/dl 2.03 1.83 0.70 0.15 0.96 0.26 1.10 0.26 1.7 0.41 2.75 0.92 5.49 2.05 Women, n 137 20 37 24 25 13 18 GFR, ml min 1 1.73 m 2 55.2 31.8 107.6 15.8 74.3 7.6 53.7 3.7 37.0 4.9 22.0 5.0 9.0 2.7 Age, yr 53.5 15.8 41.3 12.3 45.8 14.9 60.5 13.0 58.9 15.4 61.8 11.4 59.9 13.5 Body weight, kg 65.2 13.3 63.4 12.6 65.0 13.1 66.5 15.6 66.5 11.3 66.5 16.5 63.0 12.8 Height, cm 156.7 7.0 156.8 5.4 157.8 7.2 158.9 8.2 154.8 6.6 155.6 6.8 154.8 6.7 Body mass index, kg/m 2 26.6 5.2 25.9 5.3 26.3 5.5 26.3 5.5 27.8 5.0 27.3 5.4 26.3 5.1 Serum creatinine, mg/dl 1.52 1.33 0.64 0.12 0.78 0.15 1.00 0.21 1.38 0.26 2.18 0.56 4.44 1.34 Men, n 158 11 43 18 25 35 26 GFR, ml min 1 1.73 m 2 46.5 29.9 113.4 11.8 72.2 8.9 53.1 4.6 38.7 4.0 23.2 4.4 9.9 3.6 Age, yr 51.5 15.4 30.6 14.1 49.5 14.1 53.2 14.3 53.4 14.9 56.6 13.3 53.9 15.2 Body weight, kg 77.2 14.8 78.6 18.8 75.3 13.7 76.5 12.1 75.0 16.7 81.8 16.0 76.1 12.3 Height, cm 169.9 8.1 173.5 7.1 169.0 9.1 170.1 5.7 168.0 9.6 169.8 7.6 171.4 7.0 Body mass index, kg/m 2 26.7 4.0 25.9 4.9 26.3 3.7 26.4 3.4 26.4 4.2 28.5 4.2 25.9 3.5 Serum creatinine, mg/dl 2.48 2.07 0.81 0.13 1.12 0.23 1.22 0.27 1.88 0.38 2.96 0.93 6.22 2.16 Primary kidney disease n Chronic renal failure 81 Primary and secondary 71 glomerulonephritis Ischemic nephropathy 43 Interstitial nephropathy 28 Diabetic nephropathy 26 Adult polycystic kidney disease 20 and cystic kidney disease Renal transplant recipients 15 Kidney donors 11 Values are means SD. n, No. of patients; CKD, chronic kidney disease; GFR, glomerular filtration rate.

URINARY BTP TO SCREEN EARLY GFR IMPAIRMENT F1409 examined patients are reported in Table 1. No patient needed replacement of renal function by dialysis. The study was approved by the Institutional Ethical Committee and was conducted in accordance with ethical guidelines of the Helsinki Declaration. All patients gave their informed consent. Measurement of GFR (reference test). GFR was measured with a radioisotopic method as the renal clearance of 99m Tc-diethylenetriamine-pentaacetic acid (DTPA) (4, 5). The results were adjusted, as usual, to the standard body surface of 1.73 m 2. On the basis of the value of the GFR measurement, patients were classified in the five different stages of CKD. We used the modified classification of CKD, which subdivides the stage 3 (GFR 30 60 ml min 1 1.73 m 2 ) into 3a (GFR 45 60 ml min 1 1.73 m 2 ) and 3b (GFR 30 45 ml min 1 1.73 m 2 ) (46). Measurement of SCr and urinary concentration of creatinine, BTP, Cys, 2M, retinol-binding protein, and of U-alb (index tests). Blood and urine samples were drawn at the time of GFR measurement, in the morning before breakfast. Serum and urinary samples were divided into Eppendorf tubes, which were hermetically closed and stored at 20 C, thus avoiding sample concentration. Creatinine was measured with a rate-blanked creatinine/jaffé method (CREA Roche/Hitachi automated analysis for Hitachi 917, Roche Diagnostics, Mannheim, Germany; reference intervals for serum concentration are 0.50 0.90 mg/dl in women and 0.70 1.20 mg/dl in men). Cys was measured with a particle-enhanced immune-nephelometric method (N Latex Cystatin C, Dade Behring, Marburg, Germany; reference intervals given for SCys are 0.53 0.95 mg/l, without differences between men and women). 2M was measured with an immune-enzymic method (AxSym 2-Microglobulin, Abbott, Wiesbaden, Germany; mean reference value given for serum 2M is 0.99 0.16, without differences between men and women). Retinol-binding protein (RBP) was measured with an immunenephelometric method (N antiserum to human retinol-binding protein, Dade Behring, Marburg, Germany; reference intervals given for serum RBP are 3 6 mg/dl). BTP, which is also known as lipocalin-type prostaglandin D synthase, was measured with a particle-enhanced immune-nephelometric method (N Antiserum to human BTP, Dade Behring). Reference intervals (2.5 97.5%) for BTP have been calculated in our laboratory from 120 normal subjects (60 men and 60 women), aged 18 59 yr, mean 32.2. In men, the serum reference intervals were 0.37 0.77 mg/l, mean 0.58 mg/l, and median 0.57 mg/l, while in women the reference intervals were 0.40 0.70 mg/l, mean 0.54 mg/l, and median 0.53 mg/l. The slight difference between mean values of BTP in men and women resulted significantly different (P 0.043) (16). U-Alb was measured with an immune-nephelometric method (N antiserum to human albumin, Dade Behring). Urinary excretion of proteins was measured on spot urine samples at the time of GFR measurement. Urinary concentration of proteins was then reported, as usual, to urinary concentration of creatinine. These laboratory data do not allow the measurement of the urinary clearance of proteins. Thus we calculated fractional excretion of LMWP from serum and urinary concentrations of creatinine and proteins in the same spot urine sample, as is commonly done to measure fractional excretion of sodium. Fractional excretion (that is fractional clearance) of the different LMWP was calculated as FE LMWP 100 U-LMWP SCr S-LMWP U Cr Prediction of GFR. GFR was predicted with the most commonly used prediction equations based on SCr and SCys: Cockcroft and Gault formula (CG-CCr) (11), Modification of Diet in Renal Disease study simplified formula (MDRD-GFR) (41), and Cys-based formula for adults (Cys-GFR) (23); CG-CCr (ml/min) (140 age years) body wt/(scr mg/dl 72) 0.85 (if a woman); MDRD-GFR (ml min 1 1.73 m 2 ) 186 SCr 1.56 age years 0.203 0.742 (if a woman); and Cys-GFR (ml min 1 1.73 m 2 ) 86.49 SCys 1.686 (mg/l) 0.948 (if a woman). Derivation of new prediction equations for GFR based on SCr and serum LMWP. New prediction equations for GFR were derived on the basis of stepwise multiple regression analysis among GFR ( 99m Tc- DTPA) and anthropometric data (age, body weight, height) and SCr or of the different LMWP alone or in combination. Men and women were analyzed separately. Logarithmic transformation of all data was used. The stepwise method enters in the model the different variables if the significance of the correlation with GFR was 0.05, while it removes variables with P 0.1. On the basis of stepwise multiple regression analysis, after logarithmic resolution, different formulas have been derived to predict GFR in men or women from the combination of age, anthropometric data, and serum concentration of each marker, and from the combination of the different serum markers. Statistical analysis. The correlations between GFR and serum or urinary concentrations of creatinine (U-Cr), Cys (U-Cys), 2M (U- 2M), RBP (U-RBP), BTP (U-BTP), and U-Alb were tested from the values of correlation coefficient r. The best correlations were found using logarithmic transformation of data. The diagnostic accuracy (sensitivity and specificity) of all parameters, tested as indicators of different degree of GFR impairment, was assessed using receiveroperating characteristic (ROC) analysis. Positive predictive values (PPV) and negative predictive values were calculated from sensitivity and specificity and from the prevalence of different CKD stages in the general population PPV sensitivity prevalence sensitivity prevalence (1 specificity) (1 prevalence) NPV specificity (1 prevalence) 1 sensitivity) (prevalence specificity) (1 prevalence) The prevalence of egfr 90 ml min 1 1.73 m 2 in the general population (59.3%) was calculated on the basis of data from the National Health and Nutrition Examination Survey (NHANES) 1999 2004 (10a). The significance of the differences among the mean values of serum or urinary levels of Cr, Cys, 2M, RBP, BTP, and U-Alb in different groups of patients was assessed using a nonparametric Mann-Whitney test. The Wilcoxon test for paired samples was used to evaluate the statistical significance of the differences between the different predictions of GFR and measured GFR. Statistical analysis was performed using MedCalc, version 9.3.9.0 (Mariakerke, Belgium). A P value 0.05 was considered statistically significant. Multiple regression analysis, after logarithmic transformation of all data, was used to derive, separately in women and men, new formulas to predict GFR from serum markers and anthropometric data. Passing and Bablok regression was used to test the correlation between predicted and measured values of GFR (51). Bland and Altman plots (7) were used to evaluate the agreement between measured and predicted values of GFR. Mountain plots were used to evaluate the distribution of the differences between measured and predicted values of GFR and to identify the central 95% difference (25). The mean prediction error of the different estimates of GFR vs. measured GFR was calculated as the root mean squared error (37). RESULTS SCr, SBTP, SCysS, S 2M, and SRBP vs. GFR. A high logarithmic correlation was found between GFR and the serum concentrations of the different markers of renal function (Table 2 and Supplemental Fig. S1). When patients were examined together, S 2M and SCys showed the best correlation with GFR, while,

F1410 URINARY BTP TO SCREEN EARLY GFR IMPAIRMENT Table 2. Correlation coefficients of the logarithmic relationship among GFR and serum creatinine, cystatin C, 2-microglobulin, retinol-binding protein, and -trace protein All Patients Women Men n 295 137 158 Serum creatinine 0.9253 0.9463 0.9467 Serum cystatin C 0.9361 0.9393 0.9343 Serum 2-microglobulin 0.9380 0.9345 0.9393 Serum retinol-binding protein 0.6919* 0.6953* 0.6848* Serum -trace Protein 0.9250 0.9203 0.9368 n, No. of patients. Significance of the differences among the different correlation coefficients is reported. *P 0.0001 retinol-binding protein vs. all the other serum markers. when data were analyzed separately for male and female patients, SCr showed the best correlation with GFR. In any case, the worst correlation was that between SRBP and GFR (Table 2 and Supplemental Fig. S1). Serum concentration of all markers progressively increased with the reduction of GFR (Fig. 2, Supplemental Fig. S1, and Supplemental Table S1). Keeping male and female patients together, the slight increases found in patients at CKD stage 2 (GFR 60 90 ml min 1 1.73 m 2 ) were already statistically significant vs. the values found in the group of CKD at stage 1 (GFR 90 ml min 1 1.73 m 2 ). Statistical significance was higher for SCr, S 2M, and SBTP than for SCys and SRBP (Fig. 2). Some differences among the different markers were found according to the gender of patients. In fact, the increase in SCr, S 2M, and SBTP was already significant in both men and women at CKD stage 2. However, the increase in SCys became significant in men at CKD stage 2 and in women at CKD stage 3a (GFR 45 60 ml min 1 1.73 m 2 ), and the increase in SRBP became significant in both men and women at CKD stage 3a (Fig. 2, Supplemental Table S1). The area under the curve (AUC) of ROC plots results were highly statistically significant for all serum markers in all functional groups of CKD patients (Table 3). However, the AUC, and hence the accuracy, of SRBP, as an indicator of a GFR 90, 80, 70, and 60 ml min 1 1.73 m 2 were significantly lower than that of the other serum markers. The lower accuracy of SRBP was confirmed when men and women were examined separately. In men, SBTP had a similar diagnostic accuracy of SCr, SCys, and S 2M as an indicator of a GFR 60 or 70 ml min 1 1.73 m 2. In the group of women, SBTP was significantly less accurate than SCr, SCys, and S 2M as an indicator of a GFR 60 or 70 ml min 1 1.73 m 2. No statistically significant difference was found among SCr, SCys, S 2M, and SBTP as indicators of a GFR 80 or 90 ml min 1 1.73 m 2, either in men or in women. Furthermore, the accuracy of SCr, as an indicator of a GFR 80 or 90 ml min 1 1.73 m 2, was higher in men than in women. The accuracy of SCr, SCys, S 2M, and SBTP as indicators of more pronounced impairments in renal function (GFR 45, 30, and 15 ml min 1 1.73 m 2 ) was similar in both men and women. For all serum markers, the criterion values to screen the same GFR impairment were higher for male than for female patients (Table 3). Furthermore, independently of gender, to screen accurately patients with mild impairment of GFR, one must use criterion values lower than those necessary Fig. 2. Serum concentration of creatinine and low-molecular weight proteins (LMWP) in groups of patients clustered according to the stage of CKD on the basis of glomerular filtration rate (GFR; ml min 1 1.73 m 2 ). Grey bars, all patients; white bars, women; black bars, men. The height of the bars represents means, and the length of the line over the bars represents SD. Significance of the differences vs. CKD stage 1 (Mann-Whitney test): P 0.05; P 0.01; xp 0.001; P 0.0005; P 0.0001.

URINARY BTP TO SCREEN EARLY GFR IMPAIRMENT Table 3. Receiver operating curve analysis of the accuracy of serum concentration of creatinine and low-molecular weight proteins as indicators of different impairment of GFR F1411 Creatinine, mg/dl Cystatin C, mg/l 2-Microglobulin, mg/l Retinol-Binding Protein, mg/dl -Trace Protein, mg/l GFR, ml min 1 1.73 m 2 n AUC Criterion AUC Criterion AUC Criterion AUC Criterion AUC Criterion All patients 295 90 264 0.932 0.94 0.939 0.94 0.921 1.52 0.808 (x) 4.5 0.912 0.89 80 246 0.904 1.0 0.926 1.0 0.925 1.60 0.817 (x) 4.5 0.909 0.92 70 212 0.939 1.0 0.950 1.0 0.937 1.52 0.835 (x) 4.6 0.919 0.92 60 184 0.924 1.25 0.964 (x) 1.30 0.948 ( ) 2.27 0.839 (x) 4.4 0.908 1.11 45 142 0.973 1.35 0.990 ( ) 1.4 0.979 2.45 0.881 (x) 5.5 0.970 1.42 30 92 0.984 1.82 0.986 1.99 0.985 3.93 0.898 (x) 5.5 0.983 2.08 15 44 0.983 2.96 0.988 2.59 0.987 5.29 0.881 (x) 5.8 0.992 3.11 Women 137 90 117 0.918 0.78 0.930 0.8 0.893 1.52 0.783 ( ) 4.0 0.882 0.87 80 109 0.902 0.78 0.917 0.94 0.906 1.52 0.799 ( ) 4.3 0.881 0.91 70 88 0.942 0.91 0.949 1.0 0.936 1.73 0.815 (x) 4.3 0.885 ( ) 0.91 60 57 0.954 0.92 0.970 1.0 0.949 2.04 0.856 (x) 4.3 0.888 ( ) 0.93 45 56 0.982 1.06 0.989 1.45 0.931 2.45 0.888 (x) 4.5 0.962 1.30 30 31 0.993 1.48 0.986 1.90 0.995 4.28 0.929 ( ) 5.5 0.990 2.19 15 18 0.993 2.46 0.991 2.58 0.993 5.29 0.911 5.5 0.994 2.85 Men 158 90 147 0.963 0.94 0.941 0.9 0.952 1.46 0.813 ( ) 4.8 0.931 0.89 80 137 0.934 1.23 0.937 1.0 0.942 1.6 0.819 (x) 5.1 0.935 1.0 70 124 0.946 1.25 0.946 1.15 0.934 1.49 0.837 (x) 5.4 0.943 1.0 60 104 0.938 1.56 0.966 ( ) 1.3 0.944 2.17 0.826 (x) 5.0 0.938 1.46 45 86 0.986 1.62 0.991 1.40 0.963 2.17 0.866 (x) 5.5 0.980 1.54 30 61 0.980 2.10 0.987 1.99 0.980 3.70 0.858 (x) 5.7 0.974 2.08 15 26 0.987 3.29 0.987 2.80 0.982 5.60 0.853 ( ) 6.9 0.992 3.11 Statistical significance of the area under the curve (AUC) ( P 0.0001) and criterion values are reported. n, No. of patients. Statistical significance of the difference of AUC of the different markers vs. AUC of serum creatinine is reported: ( )P 0.05; ( )P 0.01; (x)p 0.001; ( )P 0.0001. to screen patients with a more advanced impairment in GFR (Table 3). U-Alb, U-BTP, U-Cys, U- 2M, and U-RBP vs. GFR. U-Alb was widely variable in the whole range of GFR impairments, without reaching, at any stage of CKD, a statistically significant difference compared with CKD patients at stage 1 (Fig. 3, Supplemental Fig. S2, Supplemental Table S2). U-Cys and U- 2M were almost undetectable in patients with GFR 30 ml min 1 1.73 m 2, while their excretion increased significantly in patients at CKD stages 4 and 5 (GFR 30 and 15 ml min 1 1.73 m 2, respectively) (Fig. 3, Supplemental Fig. S2, Supplemental Table S2). Also, U-RBP increased significantly only in patients at CKD stages 4 and 5 (Fig. 3, Supplemental Fig. S2, Supplemental Table S2). On the contrary, urinary BTP, which was measurable even in patients with normal GFR, increased significantly already in patients at CKD stage 2 (Fig. 3, Supplemental Fig. S2, Supplemental Table S2). U-Cys, U- 2M, and U-RBP became significant at a threshold value of GFR of 30 ml min 1 1.73 m 2, while the GFR threshold value was 90 ml min 1 1.73 m 2 for U-BTP (Fig. 4, left). The fractional excretion, that is its urinary clearance as percentage of GFR, was much higher for BTP than for the other LMWP (Fig. 4, middle). The U-BTP increased progressively, according to the increase in SBTP, up to a maximal value of U-BTP, which was reached when SBTP was 2 mg/l (Fig. 4, right). This behavior suggests that the complete saturation of tubular resorption of BTP is reached when SBTP, and hence BTP concentration in tubular fluid, is 2 3 mg/l. The accuracy of the urinary excretion of the different proteins, as indicators of different impairments of GFR, was tested, keeping together men and women. Indeed, results for the accuracy of the different urinary markers are quite different (Table 4). In particular, the accuracy of U-BTP, as an indicator of moderate impairments in GFR (GFR 90, 80, 70, 60 ml min 1 1.73 m 2 ), was significantly higher than that of any other protein, while the result for U-Alb showed inadequate accuracy to indicate any GFR impairment. U-BTP had also the highest accuracy as an indicator of more pronounced impairments in renal function (GFR 45, 30 ml min 1 1.73 m 2 ). Only in very advanced renal failure patients (CKD stage 5, i.e., GFR 15 ml min 1 1.73 m 2 ) was there any significant difference in the accuracy of the different LMWP proteins. Finally, the shape of the relationship between U-BTP and GFR was quite similar to that of SBTP with GFR (Fig. 5). When patients were clustered according to underlying renal disease, U-BTP was significantly higher in renal transplant recipients than in glomerulonephritis, ischemic renal disease, and interstitial nephritis. U-BTP was also higher in chronic renal failure than in glomerulonephritis and in ischemic renal disease patients (Supplemental Table S3). At CKD stage 5, U-BTP was higher in chronic renal failure patients (23.4 12.5 mg/g creatinine) than in ischemic renal disease patients (13.5 9.3) and glomerulonephritis patients (18.1 10.8) (Supplemental Fig. S3). These differences were not statistically significant. Also, at the other CKD stages no significant differences were found in U-BTP among these three kinds of renal disease. Due to the low number of patients, this analysis was not possible for the other kinds of renal diseases. New prediction equations for GFR based on SCr and SLMWP. The different prediction equations for GFR, derived from stepwise multiple regression analysis of GFR (dependent variable) and the combination of gender, age, and anthropo-

F1412 URINARY BTP TO SCREEN EARLY GFR IMPAIRMENT Fig. 3. Urinary excretion of albumin and LMWP in groups of patients clustered according to the stage of CKD on the basis of GFR. RBP, retinol-binding protein. Results are expressed per milligram of excreted creatinine. Grey bars, all patients; white bars, women; black bars, men. The height of the bars represents the mean value, and the length of the line over the bars represents SD. Significance of the differences vs. CKD stage 1 (Mann-Whitney test): P 0.05; P 0.01; xp 0.001; P 0.0005; P 0.0001. metric data and serum concentration of the different markers, are reported in Table 5, together with the values of multiple correlation coefficients (MCC). The MCC of SRBP-GFR, for either women or men, were significantly lower than all the other MCC. The MCC of the serum markers (SMarkers) formula was significantly higher than MCC of SCys, S 2M, and SBTP formulas either in women or in men (Table 5). A high linear correlation was found between measured GFR and the values of GFR predicted with the different equations. The correlation, agreement, and mountain plots obtained with already published formulas, that is CG-CCr (11), MDRD- GFR(41), and Cys-GFR (23) (Fig. 6) and with the new equations derived in the present study (Fig. 7) indicate some differences. Passing and Bablok regression between predicted and measured values of GFR (Figs. 5 and 6, left) indicated that a 95% confidence interval for the intercept was significantly different from 0 for Cys-GFR and for SRBP-GFR. In the meantime, a 95% confidence interval for the slope of regression was significantly different from 1 for CG-CCr, MDRD- GFR, Cys-GFR, and SRBP-GFR, demonstrating a proportional difference between the prediction formulas with GFR. The closest correlation with measured GFR was obtained using the Serum Markers formula, which was significantly better correlated with GFR than all other prediction formulas, except for MDRD-GFR and SCr-GFR (Table 6). On the contrary, SBTP- GFR had the lowest correlation with measured GFR than all the other prediction formulas. Correlation coefficients with GFR of the different prediction formulas were generally lower or, in some cases, similar to the correlation coefficients with GFR of the serum concentration of the different markers (SCr, SCys, S 2M, and SBTP) (Supplemental Table S4). The mean difference in GFR found for CG-CCr ( 12.1 ml min 1 1.73 m 2 ), MDRD-GFR( 7.3 ml min 1 1.73 m 2 ), Cys-GFR( 8.8 ml min 1 1.73 m 2 ), and SRBP-GFR ( 4.1 ml min 1 1.73 m 2 ) was statistically significant (Fig. 6, middle), while the differences in GFR were insignificant for the other prediction formulas (Fig. 7, middle). The best agreement with GFR was obtained with the prediction formula based on a combination of different serum markers (SMarkers-GFR; range of agreement 47.8 ml min 1 1.73 m 2 ), and with SCr-GFR (range of agreement 51.0 ml min 1 1.73 m 2 ). Mountain plots indicated that the differences between prediction formulas and measured GFR were symmetrically distributed around the 0 value, except for CG-CCr, MDRD-GFR, Cys-GFR, and SRBP-GFR (Figs. 6 and 7, right). The central 95% of the difference between GFR and prediction formulas was asymmetrically distributed for CG-CCr, MDRD-GFR, Cys-GFR, and SRBP-GFR, while it was symmetrically distributed for the other prediction formulas (Supplemental Table S5). Results of the different prediction equations in groups of patients clustered according to CKD stage are reported in Supplemental Table S6. SRBP-GFR was less effective in discriminating between the different CKD stages in particular when women and men were examined separately. The prediction errors of the different estimates of GFR were generally lower in men than in women (Table 7). In particular, SMarkers-GFR had the lowest prediction errors: 9.8 ml min 1 1.73 m 2 in men and 13.2 ml min 1 1.73 m 2 in women. Absolute prediction errors were lower in CKD patients at stage 5 than in CKD patients with better preserved renal function. In CKD patients at stage 5, prediction errors of SMarkers-GFR were in range 2.7 3.4

URINARY BTP TO SCREEN EARLY GFR IMPAIRMENT F1413 Fig. 4. Urinary excretion of LMWP in all patients. Left: urinary excretion of LMWP in groups of patients clustered according to the stage of CKD on the basis of GFR. Results are expressed per milligram of excreted creatinine. The height of the bars represents the mean value, and the length of the line over the bars represents SD. Middle: fractional excretion of the LMWP in groups of patients clustered according to GFR. Results are expressed as a percentage. Right: log of urinary excretion of LMWP are plotted vs. serum concentration of LMWP. Men and women are considered together. Significance of the differences vs. CKD stage 1 (Mann-Whitney test): P 0.01; xp 0.001; P 0.0001.

F1414 URINARY BTP TO SCREEN EARLY GFR IMPAIRMENT Table 4. Receiver operating curve analysis of the accuracy of urinary excretion of proteins as indicators of impairment of GFR Albumin, mg/g creatinine Cystatin C, mg/g creatinine 2-Microglobulin, mg/g creatinine Retinol-Binding Protein, mg/g creatinine -Trace Protein, mg/g creatinine GFR, ml min 1 1.73 m 2 n AUC Criterion AUC Criterion AUC Criterion AUC Criterion AUC Criterion 90 264 0.545 (x) 14.3 0.619 (x) 0.10 0.600 (x) 0.11 0.588 (x) 4.6 0.824 2.24 80 246 0.528 (x) 195.5 0.654 (x) 0.10 0.673 (x) 0.11 0.608x (x) 4.6 0.822 3.19 70 212 0.516 (x) 215.7 0.690 (x) 0.10 0.678 (x) 0.13 0.605 (x) 2.38 0.814 3.23 60 184 0.568 (x) 54.7 0.716 (x) 0.13 0.704 (x) 0.12 0.664 (x) 2.38 0.812 4.49 45 142 0.565 (x) 54.7 0.794 0.07 0.763 (x) 0.19 0.694 (x) 2.30 0.850 3.23 30 92 0.587 (x) 31.7 0.885 0.11 ( ) 0.863 0.42 0.799 ( ) 2.47 0.887 8.0 15 44 0.567 (x) 60.2 0.954 0.35 0.921 ( ) 1.62 0.895 4.26 0.858 13.79 Statistical significance of AUC ( P 0.05; P 0.02; P 0.01; xp 0.001; P 0.0001) and criterion values are reported. n, No. of patients. Statistical significance of the difference of AUC of the different markers vs. AUC of urinary BTP is reported in parentheses: ( )P 0.05; ( )P 0.01; (x)p 0.001. ml min 1 1.73 m 2, while in CKD stage 1 patients the range was 14.6 22.2 ml min 1 1.73 m 2 (Table 7). The AUC of ROC plots was highly statistically significant for all prediction formulas in all functional groups of CKD patients (Table 8). However, the AUC, and hence the accuracy, of SRBP-GFR, as an indicator of a GFR 90, 80, 70, and 60 ml min 1 1.73 m 2, was significantly lower than that for the other prediction formulas. The lower accuracy of SRBP- GFR was confirmed when men and women were examined separately. When men and women were considered together, the accuracy of SCr was slightly but significantly lower than that of CG-CCr as an indicator of a GFR 80 and 60 ml min 1 1.73 m 2. The accuracy of SCr, as an indicator of a GFR 60 and 45 ml min 1 1.73 m 2 was also lower than those of MDRD-GFR, Cys-GFR, SCr-GFR, and SCys-GFR, while it was lower than that of SMarkers-GFR as an indicator of a GFR 90, 70, and 60 ml min 1 1.73 m 2. No significant differences were found compared with S 2M-GFR and with SBTP-GFR. SCr was always more accurate than SRBP- GFR. When men and women were considered separately, very few differences were found in the accuracy of SCr vs. all prediction formulas, except a better accuracy vs. SRBP-GFR (Table 7). SMarkers-GFR showed the highest accuracy in predicting GFR. Its accuracy was always significantly better than that of SRBP-GFR, while it was similar to the accuracy of the other prediction formulas (Table 8). As indicated by the values of the AUC, the accuracy of serum concentration of each marker per se, as an indicator of the different impairments of GFR, was not significantly different from the accuracy of the corresponding prediction formulas, obtained from the combination of each serum markers with anthropometric data of patients with some exceptions, mainly regarding SCr. In fact, SCr was less accurate as an indicator of a GFR 80 ml min 1 1.73 m 2 than CG-CCr, and less accurate than CG-CCr, MDRD-GFR and SCr-GFR as an indicator of a GFR 60 ml min 1 1.73 m 2 ; finally, SCr was also less accurate than MDRD-GFR and SCr-GFR as an indicator of a GFR 5 ml min 1 1.73 m 2. SBTP was significantly less accurate than SBTP-GFR as an indicator of a GFR 60 ml min 1 1.73 m 2. SRBP was less accurate than SRBP-GFR as an indicator of a GFR 70 and 60 ml min 1 1.73 m 2. As indicated by the comparison of AUC values, urinary BTP was significantly less accurate, as an indicator of a GFR impairment 90 ml min 1 1.73 m 2 (CKD stage 2), than serum tests, except SRBP, and also less accurate than prediction formulas, except SRBP-GFR. However, sensitivity, specificity, negative predictive value, and particularly the PPV (85.6%) of U-BTP appears adequate to identify CDK patients with such a slight impairment in GFR (Table 9). The accuracy of SCr to screen CKD stage 2 was similar to that of the different serum markers and prediction formulas. A significantly higher accuracy was obtained only with SMarkers formula, which needs the values of SCr, S 2M, SBTP, and body weight in men and of SCr, S 2M, SBTP, and SCys in women. DISCUSSION Early screening of CKD, which may allow intervention strategies aimed to reduce or stop the progression of renal disease, is often difficult to achieve. In fact, after the initial kidney damage, subjective symptoms of renal impairment may Fig. 5. Serum concentrations (left) and urinary excretions of -trace protein vs. GFR (right). Patients are clustered in groups of CKD stages according to GFR (mean values and SD are reported). Men and women are considered together. Significance of the differences vs. CKD stage 1 (Mann-Whitney test): P 0.01; xp 0.001; P 0.0001.

URINARY BTP TO SCREEN EARLY GFR IMPAIRMENT Table 5. Equations estimating GFR derived on the basis of multiple regression analysis of GFR (dependent variable) vs. anthropometric data and serum concentration of creatinine and of the different low-molecular-weight proteins (independent variables) F1415 SCr GFR ( ) 135.5 age 0.2391 Scr 1.1289 (MCC 0.9512) SCr GFR ( ) 40.5 age 0.2096 BW 0.3252 SCr 1.1025 (MCC 0.9533) SCys GFR ( ) 66.4 SCys 1.2887 (MCC 0.9394) SCys GFR ( ) 73.0 SCys 1.3112 (MCC 0.9343) S 2M GFR ( ) 104.8 S 2M 0.9686 (MCC 0.9394) S 2M GFR ( ) 107.1 S 2M 0.9815 (MCC 0.9345) SRBP GFR ( ) 3,450.9 age 0.5456 SRBP 1.3813 (MCC 0.7371) SRBP GFR ( ) 2,320.6 age 0.3439 SRBP 1.6277 (MCC 0.7009) SBTP GFR ( ) 124.8 age 0.2191 SBTP 0.9233 (MCC 0.9245) SBTP GFR ( ) 107.8 age 0.1371 SBTP 1.03 (MCC 0.9245) SMarkers GFR ( ) 64.9 SCr 0.5224 SCys 0.2993 S 2M 0.1944 x SBTP 0.1629 (MCC 0.9623) SMarkers GFR ( ) 57.4 BW 0.193 SCr 0.54 S 2M 0.2669 SBTP 0.2914 (MCC 0.9656) SCr, serum creatinine, mg/dl); SCys, serum cystatin C, mg/l; S 2M, serum 2-microglobulin (mg/l); SRBP, serum retinol-binding protein, mg/dl; serum SBTP, -trace protein (mg/l. Age, yr; BW, body wt, kg; Women ( ) and men ( ) were examined separately. The value of the multiple correlation coefficient (MCC) is reported. The statistical significance of the different correlation coefficients were as follows: SRBP GFR, P 0.001 vs. all other predictions. SMarkers GFR, men: P 0.005 vs. SCys GFR and vs. S 2M GFR, P 0.0005 vs. SBTP GFR; SMarkers GFR, women: P 0.05 vs. SCys GFR and S 2M GFR, P 0.005 vs. SBTP GFR. be completely lacking in early stages of CKD. In many patients, even a history of primary kidney disease is absent. This is particularly frequent for ischemic kidney disease and for diabetic nephropathy (45). Recent data indicate that overall awareness of CKD in a high-risk population is quite low (60). Since clinical symptoms are very poor, the screening for CKD is necessarily based on laboratory tests. Urinary findings (proteinuria, albuminuria, or erythrocytes in urinary sediment) are useful markers of renal disease, while they cannot give information about an eventual impairment in GFR. Serum parameters, like SCr, are commonly used to evaluate the impairment in renal function and the progression of CKD. The major disadvantage of SCr is its poor sensitivity, in particular when the upper limit of the reference range is used as a screening value. In fact, patients at CKD stages 1 and 2 are usually missed if one takes a criterion value of 1.2 mg/dl for women and 1.4 mg/dl for men, which are the upper limits of reference ranges for SCr in many laboratories. In the last several years, the measurement of serum levels of different LMWP has been proposed as a more sensitive marker of GFR impairment compared with SCr. Indeed, different studies did not confirm the superiority of LMWP vs. creatinine. An important point for comparing the results of the different studies is the need for standardization of their assays (24). In particular, different analytic methods have been proposed to measure Cys, 2M, and RBP, while only one method can be used to measure BTP. A major advantage of SCr is the availability of standardized assays. Since 24h-CCr is too complex to be used to screen large populations and is also inaccurate due to the high variability of its measurement, different formulas have been proposed to predict creatinine clearance or GFR from the measurement of SCr and some anthropometric data. Most commonly, Cockcroft and Gault and the simplified MDRD formula are used (11, 41). However, their reliability to assess renal function in CKD stages 3, that is when GFR is 60 ml min 1 1.73 m 2, has not yet been demonstrated. In any case, even the simpler methods, up to now available to screen for an impairment in renal function, need a blood sample, which represents a complication compared with a urine-based method. Many LMWP are handled by the kidneys with similar pathways. The first step is their filtration through the glomeruli with a sieving coefficient which depends on MW, shape, and electrical charge of the different molecules (48). After filtration into the preurine, LMWP undergo proximal tubular reabsorption, which in normal subjects is almost complete; thus their excretion into the final urine is minimal or undetectable with common laboratory methods. Tubular damage, even a slight tubular dysfunction, or the competition for the same transporter, may increase urinary excretion of LMWP by reducing the reabsorptive capacity of tubular cells. This phenomenon is well known to happen in primary tubular diseases or as a consequence of ischemic or toxic tubular damage (3). Competition between some LMWP and albumin has been reported in nephrotic syndrome (65). Besides tubular damage and competition for the transporter, urinary excretion of LMWP may also increase as a consequence of a marked impairment in GFR. The conceptual background for use of urinary concentrations of LMWP to detect decreased GFR is based mainly on the following two premises. First, serum concentrations of LMWP increase progressively with the reduction in GFR. As a consequence, the filtered load of LMWP to single nephrons increases. However, the increase in filtered load is probably dishomogeneous according to the different value of single nephrons GFR. Second, LMWP are reabsorbed from tubular urine by proximal tubular cells via a saturable pathway. Thus, in patients with normal GFR and normal serum concentration of LMWP, U-LMWP is almost null due to their extensive tubular reabsorption. To the contrary, when the increased filtered load of LMWP to single nephrons overcomes their maximal tubular reabsorptive capacity, U-LMWP becomes measurable and progressively increases with the increase in single-nephron filtered charge, that is, with the decrease in total GFR. However, maximal tubular resorptive capacity is probably inhomogeneous among the different nephrons, which possibly affects the relationship between GFR and U-LMWP. Different data are in agreement with this hypothesis. In fact, in patients at CKD stage 5 serum concentrations of LMWP are definitely increased (5 times or more the normal values) (14, 16). As a consequence, the filtered load of these proteins to the

F1416 URINARY BTP TO SCREEN EARLY GFR IMPAIRMENT Fig. 6. Correlation and agreement between measured GFR and predicted GFR. CG-CCr, creatinine clearance by Cockcroft and Gault Formula (11); MDRD-GFR, Simplified Modification of Diet in Renal Disease study equation formula (41); cystatin C (CYS)-GFR, Grubb formula for adults (23). Inside the Passing and Bablok regression plots (left) are drawn identity lines and the regression lines with their 95% confidence intervals. In the agreement plots (middle) are drawn the mean differences between measured and predicted values of GFR and the ranges of agreement ( 1.96 SD) which encompasses 95% of the population. Mountain plots (right) represent the frequency distribution of the differences between measured and predicted values of GFR. residual nephrons may become higher than single-nephron maximal tubular reabsorptive capacity, and U-LMWP should increase significantly. Recently, we found that U-CysC and U- 2M significantly increase in CKD stages 4 and 5 (17). Furthermore, the ratio of U-Cys to U-Cr has been reported as a reliable screening tool for detecting decreased GFR in pediatric CKD patients (26). However, conflicting results have been reported by a recent study performed in children and adults, which indicated a poor accuracy of U-Cys/U-Cr as an indicator of an egfr 60 ml min 1 1.73 m 2 (27). The present study, performed in a relevant number of CKD patients affected by various kidney disease at different functional stages, confirms that serum concentrations of some LMWP, namely, Cys, 2M, and BTP, are useful markers for GFR impairment. However, their sensitivity as indicators for an early impairment of GFR is not higher than that of SCr, while their analytic procedure is more complex and expensive than that for SCr. BTP is a glycosylated LMWP highly concentrated in human cerebrospinal fluid. Its molecular weight ranges between 23 and 29 according to the degree of N-glycosylation. The brain type of BTP has a lower molecular weight than the SBTP, while the two forms have the same number of amino acids (30). More recently, BTP has been identified as a lipocalin-type Fig. 7. Correlation and agreement between measured GFR and predicted GFR. Serum creatinine GFR (SCr-GFR), SCYS-GFR, serum 2-microglobulin GFR (S 2M-GFR), serum RBP (SRBP)-GFR, serum -trace protein GFR (SBTP-GFR), and serum markers GFR (SMarkers-GFR) are the new prediction equations for GFR (this study, Table 5). Inside the Passing and Bablok regression plots (left) are drawn the identity lines and the regression lines with their 95% confidence intervals. In the agreement plots (middle) are drawn the mean differences between measured and predicted values of GFR and the ranges of agreement ( 1.96 SD) which encompasses 95% of population. Mountain plots (right) represent the frequency distribution of the differences between measured and predicted values of GFR.

URINARY BTP TO SCREEN EARLY GFR IMPAIRMENT F1417

F1418 URINARY BTP TO SCREEN EARLY GFR IMPAIRMENT Table 6. Correlation coefficients among measured GFR and predicted GFR All Patients, n 295 Women, n 137 Men, n 158 CG Ccr 0.8946 0.8611 0.9277 g ; h g g : h i i MDRD GFR 0.9064 0.8831 0.9254 g ; h ; c g g ; h Cystatin C GFR 0.8713 0.8623 0.8773 g g g i ; d, b i i ;i ; a,b Serum creatinine GFR 0.9112 0.8835 0.9343 g ; hx; c g, h g ; c, h ; e Serum cystatin C GFR 0.8935 0.8841 0.8988 g g g i ; d i i ; d Serum 2-microglobulin GFR 0.8821 0.8564 0.9050 g g g ix i Serum retinol-binding Protein GFR 0.6488 0.6150 0.6618 a,b,c,d,e f,h,i a,b,c,d,e,f,i ; hx a,b,c,d,e,f,h,i Serum -trace protein GFR 0.8478 0.8070 0.8815 g gx g i ; dx: a,b,e ix, b,d,e i ; d ; a,b Serum markers GFR 0.9312 0.9127 0.9465 c,g,h ; fx; a,e g ; hx; a,c,f g ; hx; c ; a,f CG Ccr, creatinine clearance by Cockcroft and Gault formula (11); MDRD GFR, simplified equation from Modification of Diet in Renal Disease study (41); cystatin C GFR, Grubb formula for adults (23); Serum creatinine GFR, serum cystatin C GFR, serum 2-microglobulin GFR, serum retinol-binding protein GFR, serum -trace protein GFR, and serum markers GFR are based on the new prediction equations for GFR (this study, Table 5). Statistical significance of the difference is reported: P 0.05, P 0.01, xp 0.001, P 0.0005, P 0.0001. a, Vs. CG Ccr; b, vs. MDRD GFR; c, vs. cystatin C GFR; d, vs. serum creatinine GFR; e, vs. serum cystatin C GFR; f, vs. serum 2-microglobulin GFR; g, vs. serum retinol-binding protein GFR; h, vs. serum -trace protein GFR; i, vs. serum markers GFR. prostaglandin D 2 synthase (29, 44), an enzyme with different vascular actions. Preliminary reports on a possible use of BTP as a diagnostic protein in renal disease were based on the finding of highly elevated concentrations of BTP in the serum of hemodialysis or peritoneal dialysis patients (16, 44). Other data support the view that BTP may be suitable as an indicator of reduced GFR even in the creatinine-blind range (54). Our previous data demonstrated that serum levels of BTP increase progressively with the reduction of GFR, and its accuracy as a marker of GFR impairment is similar to that of SCr, SCys, and S 2M (16). Other papers suggest that BTP is not better than CysC as an indicator of reduced GFR either in the general population (55) or in children with spina bifida, who have a reduced muscle mass (52), and indicate that BTP, like CysC, are poor markers of GFR during pregnancy (1). On the other hand, other authors believe that SBTP is a good marker for the Table 7. Mean prediction errors of the different estimates of GFR CKD Stage n GFR CG CCr MDRD GFR Cys GFR SCr GFR SCys GFR S 2M GFR SRBP GFR SBTP GFR SMarkers GFR All 295 50.5 31.0 21.5 17.2 27.0 13.0 14.6 15.2 25.7 17.7 11.5 1 31 109.6 14.6 35.4 29.5 51.9 25.6 27.7 27.4 45.2 31.2 19.7 2 80 73.2 8.3 27.4 20.7 35.4 14.9 17.8 19.0 29.9 21.1 14.2 3a 42 53.4 4.1 22.7 21.2 24.6 14.0 14.7 14.9 26.0 22.0 12.8 3b 50 37.9 4.5 13.1 9.4 11.0 6.6 8.7 9.3 20.2 9.9 7.3 4 48 22.9 4.6 11.9 7.8 7.0 6.2 5.1 7.1 8.3 6.7 5.6 5 44 9.5 3.3 6.6 4.7 3.4 3.8 4.1 3.9 14.7 3.6 3.0 Women 137 55.2 31.8 26.1 19.6 31.1 15.3 15.6 17.5 27.9 20.2 13.2 1 20 107.6 15.8 37.1 30.8 54.4 28.6 27.4 29.4 45.1 30.2 22.2 2 37 74.3 7.6 35.8 24.8 41.0 17.3 19.2 22.6 33.0 24.2 16.6 3a 24 53.7 3.7 21.6 18.7 20.3 12.8 11.1 13.3 27.7 23.5 10.7 3b 25 37.0 4.9 15.0 10.1 11.6 6.3 9.2 8.7 13.4 11.8 7.2 4 13 22.0 5.0 12.5 7.1 5.8 5.0 4.3 4.1 6.0 5.3 3.7 5 18 9.0 2.7 8.1 5.9 2.7 4.5 4.0 4.5 15.6 3.9 3.4 Men 158 46.5 29.9 16.6 14.0 23.0 10.7 13.7 12.9 23.7 15.2 9.8 1 11 113.4 11.8 33.3 28.3 48.9 20.0 29.5 29.9 46.5 34.3 14.6 2 43 72.2 8.9 17.4 16.6 30.1 12.7 16.6 15.5 27.3 18.3 11.9 3a 18 53.1 4.6 24.5 24.4 29.9 15.7 18.9 17.0 24.4 20.4 15.3 3b 25 38.7.0 11.0 8.7 10.5 7.1 8.4 10.0 25.6 7.8 7.4 4 35 23.2 4.4 11.8 8.1 7.4 6.6 5.5 8.0 9.1 7.2 6.2 5 26 9.9 3.6 5.4 3.8 3.8 3.3 4.2 3.5 14.2 3.4 2.7 Values are means SD. expressed as ml min 1 1.73 m 2. n, No. of patients. Patients were clustered in groups according to the stage of CKD. SCr GFR, SCys GFR, S 2M GFR, SRBP GFR, SBTP GFR, and SMarkers GFR are based on the new prediction equations for GFR (this study, Table 5).