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

Carlo Donadio

Abstract

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), β2-microglobulin (β2M), retinol-binding protein (RBP), β-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 (99mTc-DTPA) and serum Cr, Cys, β2M, 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, β2M, and BTP. Urinary albumin was inadequate as an indicator of any level of GFR impairment. Urinary excretion of Cys and β2M 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 99mTc-DTPA and 51Cr-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 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, 3436, 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], β2-microglobulin (β2M; MW 11.8), retinol-binding protein (RBP; MW 21.2), and β-trace protein (BTP; MW 18.5) compared with SCr and urinary excretion of albumin (U-Alb); 2) derive prediction equations for GFR based on SCr, SCys, Sβ2M, 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 <90 ml·min−1·1.73 m−2 by means of these different tests of GFR impairment.

METHODS

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 platinum-based 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 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.

Fig. 1.

Flow diagram of the 435 patients that entered the study. CKD, chronic kidney disease.

View this table:
Table 1.

Main anthropometric and clinical data of patients

Measurement of GFR (reference test).

GFR was measured with a radioisotopic method as the renal clearance of 99mTc-diethylene-triamine-pentaacetic acid (DTPA) (4, 5). The results were adjusted, as usual, to the standard body surface of 1.73 m2. 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, β2M, 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).

β2M was measured with an immune-enzymic method (AxSym β2-Microglobulin, Abbott, Wiesbaden, Germany; mean reference value given for serum β2M is 0.99 ± 0.16, without differences between men and women).

Retinol-binding protein (RBP) was measured with an immune-nephelometric 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 FELMWP=100×(U-LMWP×SCr)/(S-LMWP×UCr)

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 (99mTc-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), β2M (U-β2M), 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 receiver-operating 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×prevalencesensitivity×prevalence+(1specificity)×(1prevalence) NPV=specificity×(1prevalence)(1sensitivity)×(prevalence+specificity)×(1prevalence)

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, β2M, 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, Sβ2M, 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, Sβ2M and SCys showed the best correlation with GFR, while, 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).

View this table:
Table 2.

Correlation coefficients of the logarithmic relationship among GFR and serum creatinine, cystatin C, β2-microglobulin, retinol-binding protein, and β-trace protein

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, Sβ2M, 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, Sβ2M, 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).

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.

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 Sβ2M 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 Sβ2M 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, Sβ2M, 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, Sβ2M, 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 to screen patients with a more advanced impairment in GFR (Table 3).

View this table:
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

U-Alb, U-BTP, U-Cys, U-β2M, 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-β2M 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).

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.

U-Cys, U-β2M, 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.

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.

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).

View this table:
Table 4.

Receiver operating curve analysis of the accuracy of urinary excretion of proteins as indicators of impairment of GFR

Fig. 5.

Serum concentrations (left) and urinary excretions of β-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.

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 anthropometric 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, Sβ2M, and SBTP formulas either in women or in men (Table 5).

View this table:
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)

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, Sβ2M, and SBTP) (Supplemental Table S4).

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.

Fig. 7.

Correlation and agreement between measured GFR and predicted GFR. Serum creatinine GFR (SCr-GFR), SCYS-GFR, serum β2-microglobulin GFR (Sβ2M-GFR), serum RBP (SRBP)-GFR, serum β-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.

View this table:
Table 6.

Correlation coefficients among measured GFR and predicted GFR

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 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).

View this table:
Table 7.

Mean prediction errors of the different estimates of GFR

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 Sβ2M-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).

View this table:
Table 8.

Receiver operating curve analysis of the accuracy of GFR predicted with different equations

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, Sβ2M, SBTP, and body weight in men and of SCr, Sβ2M, SBTP, and SCys in women.

View this table:
Table 9.

Screening of CKD patients with GFR <90 ml · min−1 · 1.73 m−2

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 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, β2M, 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 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-β2M 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, β2M, 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 prostaglandin D2 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 Sβ2M (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 identification of early renal impairment in type 2 diabetes (36), or that BTP may have a place as an alternative endogenous GFR marker in renal transplant patients receiving steroid therapy (31, 53a), and that SBTP is a potential GFR marker in children (9, 18, 70). Finally, an equation based on SBTP, urea, and/or creatinine has been proposed to predict GFR in renal transplant recipients (69).

Urinary excretion of proteins can be expressed in three different ways: 1) integral excretion of proteins (mg/24 h), by means of a timed collection of urine, mainly over 24 h. Unfortunately, this method is too cumbersome and time consuming to be used routinely. 2) Urinary concentration (mg/l) in a spot urine sample. This method is very simple, but the urinary concentration of proteins is affected by urine dilution or concentration. 3) Ratio of protein over creatinine excretion (mg/g urinary creatinine) in a spot urine sample. This is the most commonly used method due to its simplicity: spot urine vs. timed collection of urine. At steady state, when there is no acute variation in GFR, this method is able to correct the misleading effect of urine dilution on the urinary concentration of proteins. Furthermore, different data demonstrate that U-Alb or excretion of total proteins (mg/g urinary creatinine) are strictly correlated with 24-h U-Alb or protein excretion. Unfortunately, this methodological approach does not allow the measurement of the urinary clearance of tested molecules and thus it is impossible to make the direct comparison with measured GFR. Therefore, fractional clearance (and excretion) of LMWP can be estimated only reporting their urinary and serum concentrations over urinary and serum concentrations of creatinine. Since creatinine production and therefore excretion are different, mainly according to body composition and the amount of muscle mass, this may be a critical point. However, we found that U-BTP/creatinine (mg/g creatinine) and U-BTP (mg/l) were significantly correlated (Supplemental Fig. S4; r = 0.77, P < 0.000001). Furthermore, the accuracy of U-BTP/creatinine as an indicator of different degrees of GFR impairment was similar to that of U-BTP (mg/l; as an indicator of a GFR <90, <80, <70, <45 ml·min−1·1.73 m−2), or significantly higher than U-BTP (as an indicator of a GFR <60, <30, <15 ml·min−1·1.73 m−2). Finally, multiple regression analysis of U-BTP/creatinine (dependent variable) demonstrated that this parameter was not significantly correlated with body weight, height, or body cell mass (which is an indicator of muscle mass), while it was directly correlated with age and mainly with SBTP. Possibly, either U-BTP (mg/l) or U-BTP/creatinine (mg/g creatinine) can be used. However, we suggest the U-BTP/creatinine ratio to avoid the misleading effects of urine dilution on U-BTP concentration.

The results of the present study, performed in 295 CKD patients with GFR ranging from normality to advanced renal failure, clearly demonstrated that the urinary excretion of different LMWP increases progressively with the reduction of GFR. In the meantime, U-Alb, which is a well-known marker of renal disease, is not significantly dependent on the level of GFR. Interestingly, the urinary excretion of the different proteins begins from different threshold values of GFR and at different levels of serum concentrations, indicating that proximal tubular cells probably have a different reabsorptive capacity for the different LMWP. No patient was affected by primary tubular diseases, which are quite rare in the general population, or was receiving nephrotoxic drugs. It is important to note that the clinical diagnosis of tubular damage is not robust, and it is probable that some patients with tubular damage were included in the cohort. This possibility should increase the number of false-positive results of U-BTP, reducing its accuracy as a test of GFR impairment. Besides unrecognized tubular damage, biological variability and measurement error may affect the accuracy of U-BTP as an indicator of GFR impairment.

The data on urinary excretion of BTP, which was already present in the urine of CDK patients with normal or slightly reduced GFR and also in patients with normal SBTP, suggest that tubular reabsorptive capacity for BTP is, at least in part, saturated in CKD patients with normal GFR (CKD stage 1). Furthermore, the relationship between serum concentration and urinary excretion of BTP suggests, similarly to data published for other LMWP, that tubular resorption of BTP is a saturable process (2). Thus a further increase in filtered charge to single nephrons, due to the impairment in GFR, causes a progressive and significant increase in U-BTP, already from CKD stage 2. As a consequence, fractional excretion of BTP is significantly increased when GFR is <90 ml·min−1·1.73 m−2. In any case, whatever the causative mechanism, U-BTP was a much better indicator of GFR impairment than any other examined LMWP. However, U-BTP was less accurate than serum markers, except SRBP, and less accurate than old and new formulas, except the SRBP formula, to predict any GFR impairment. Thus the major advantage of U-BTP, as a possible marker of GFR impairment (GFR <90 ml·min−1·1.73 m−2), is its total noninvasiveness, since no blood drawing is required. Note that the value of GFR = 90 ml·min−1·1.73 m−2 is generally assumed as the lower limit of “normal” GFR. According to NKF-KDOQI guidelines, a value of a GFR >90 ml·min−1·1.73 m−2 classifies the examined subject a either a “normal subject” or, in presence of a marker of renal disease, as a CKD patient without impairment in GFR, that, is a CKD patient at stage 1. The prevalence of subjects with a GFR <90 ml·min−1·1.73 m−2 in the general population, is according to a NHANES study (10a) 59.3%. The sensitivity of U-BTP/creatinine to separate CKD patients with normal GFR from those with only slightly impaired GFR (CKD 2) is not excellent (76.1%); however, it is within the range of serum markers (70.0–85.2%). The PPV of U-BTP (85.6%), calculated on the basis of the prevalence of subjects with CKD at a stage >1, was lower than the PPV of most serum markers and prediction formulas (Table 10).

Up to now, very few other data are available on U-BTP. Some preliminary reports in OLETF rats suggest that U-BTP are likely to reflect the underlying increase in glomerular permeability and may be useful in predicting forthcoming glomerular damage following diabetes (50). Data in human diabetic patients indicate that the U-BTP increases in the early stage of kidney injury in patients with type 2 diabetes mellitus (28). Furthermore, ELISA results are suitable for measurement of U-BTP in a routine clinical assay. Results in a small number of CKD and diabetes patients support the hypothesis that U-BTP may reflect the severity of renal disease (49). Finally, the results of the present study in CKD patients indicate that, provided that these patients are not affected by tubular disease or tubular dysfunction, U-BTP is a suitable tool for the screening of a moderate impairment in GFR, like in CKD stage 2 patients. In fact, sensitivity, specificity, PPV, and negative predictive values of U-BTP as an indicator of a GFR <90 ml·min−1·1.73 m−2 were similar to those of serum markers of GFR impairment. New studies are warranted to verify the usefulness of U-BTP as a screening test. There is not general consensus on whether the general population should be screened for CKD. It seems more useful to screen first a selected population at higher risk of CKD, like hypertensive, diabetic, or cardiovascular patients or relatives of CKD patients, and then, eventually the general population. It will be also necessary to assess whether some types of renal disease may affect the diagnostic accuracy of U-BTP. Data from this study indicate that the relationship between U-BTP and GFR was quite similar to that between SBTP and GFR. These data demonstrate, for the first time, that in CKD patients it is possible to estimate GFR by means of a urinary test. Serum concentration of the different serum markers of renal function, combined with some anthropometrical data, and examined with stepwise multiple regression analysis vs. measured values of GFR, allowed us to generate new prediction equations, separately for female and male patients. The equation derived from SRBP had the lowest correlation with GFR and was statistically less accurate than the all other predictions of GFR. The analysis of the other prediction formulas indicated that new formulas, derived from the patients in this study, were more precise and accurate than those already published, derived from other patients. These results support the standardization of laboratory methods to compare the efficiency of prediction formulas in different studies. In any case, the comparison of the performances of the different formulas generated in this study indicate that the results obtained combining different serum markers and anthropometric data (SMarkers-GFR) are the most accurate and precise. However, with the exception of SRBP-GFR, all the other formulas have performances only slightly lower than SMarkers-GFR. Another interesting result of this study is that the accuracies of prediction formulas, to indicate different impairments of GFR, are generally similar to the accuracies of serum concentration of the different markers per se. It resulted that prediction formulas do not add much to the accuracy of serum markers, in particular SLMWP, as indicators of GFR impairment. Only the accuracy of equations based on SCr was better that the accuracy of SCr alone. This particular behavior is probably due to the interference of extrarenal factors (age, body weight, body muscle mass) which is more relevant in the relationship between SCr and GFR than in the relationships between the other serum markers and GFR.

In conclusion, serum levels of LMWP are not more sensitive or accurate than SCr as indicators of GFR impairment. Prediction formulas allow prediction of a definite value of GFR. However, formulas based on the different serum LMWPs are no more accurate to screen the different impairments in GFR than the crude serum concentrations of LMWP.

In our group of CKD patients with different impairments of GFR, the PPV of U-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, even if with a lower accuracy than most serum markers or prediction formulas. Since the results of the present study refer exclusively to a selected population of patients with already diagnosed chronic kidney diseases, further studies are needed to evaluate the accuracy of U-BTP as a screening test for renal impairment.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

ACKNOWLEDGMENTS

Drs. Annalisa Lucchesi, Michela Ardini, Francesca Caprio, Giulia Grassi, and Giada Bernabini are gratefully acknowledged for valuable help in clinical follow-up of patients and in collection of data. Also gratefully acknowledged are Nicola D'Onza for technical assistance in GFR measurements, Giulietta Sbragia for nursing care of patients, and Ida Natarelli for secretarial assistance.

REFERENCES

  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 10a.
  12. 11.
  13. 12.
  14. 13.
  15. 14.
  16. 15.
  17. 16.
  18. 17.
  19. 18.
  20. 19.
  21. 20.
  22. 21.
  23. 22.
  24. 23.
  25. 24.
  26. 25.
  27. 26.
  28. 27.
  29. 28.
  30. 29.
  31. 30.
  32. 31.
  33. 32.
  34. 33.
  35. 34.
  36. 35.
  37. 36.
  38. 37.
  39. 38.
  40. 39.
  41. 40.
  42. 41.
  43. 42.
  44. 43.
  45. 44.
  46. 45.
  47. 46.
  48. 47.
  49. 48.
  50. 49.
  51. 50.
  52. 51.
  53. 52.
  54. 53.
  55. 53a.
  56. 54.
  57. 55.
  58. 56.
  59. 58.
  60. 59.
  61. 60.
  62. 61.
  63. 62.
  64. 63.
  65. 64.
  66. 65.
  67. 66.
  68. 67.
  69. 68.
  70. 69.
  71. 70.
View Abstract