Estimated Glomerular Filtration Rate from Serum Cystatin C: Significant Differences among Several Equations Results

Abstract

Objectives: Several equations for the estimation of glomerular filtration rate (GFR) from serum cystatin C have been reported. We compared the results obtained using these equations to test the homogeneity of their results as well as their usefulness in clinical practice. Design and methods: Seven hundred and twenty-seven outpatients were studied. Of these, 439 were male and 288 were female, and their mean age was 60.8 ± 24.1 years. GFR was estimated from serum creatinine using the abbreviated Modification of Diet in Renal Disease (MDRD-4) equation. GFR was estimated from serum cystatin C levels using five different equations. Results: The simplest (100/cystatin C) formula rendered the highest estimated GFR and the Hoek’s equation rendered the lowest GFR, even significantly lower than the MDRD-4 equation (p < 0.001, Student’s t-test). From the simplest formula to the Hoek equation the mean difference calculated was 25.1 ± 8.7 mL/min (p < 0.001, Student’s t-test). No differences by gender were found among the results of different equations. All cystatin C-derived equations reduced the number of patients diagnosed of chronic renal failure when compared with MDRD-4 formula. No patient with normal renal function was shifted to the renal disease group. Conclusions: A higher value could be expected when GFR is estimated from cystatin C. Nevertheless, vast differences were found in the results when tested using several equations. Physicians should be aware of this problem to avoid a wrong clinical diagnosis of renal function.

• cystatin C
• glomerular filtration rate
• renal function
• chronic kidney disease
• renal failure

INTRODUCTION

Over the past decades several different markers for the estimation of glomerular filtration rate (GFR) have been proposed. The ideal marker for the estimation of GFR should be an endogenous molecule produced at a constant rate, which is cleared solely by the kidneys via free glomerular filtration, without being secreted either by tubular cells or reabsorbed into peritubular circulation.¹ The “gold standard” for the estimation of GFR is a clearance of exogenous substances such as inulin, iohexol, 51CrEDTA, 99mTcDTPA, or 125I-iothalamate. These techniques are time-consuming, labor-intensive, expensive, and require administration of substances that make them incompatible with routine monitoring.

Despite all known disadvantages, serum creatinine concentration has become the most commonly used marker to estimate GFR in clinical practice as in the case of most studies.² Nevertheless, creatinine-based estimations seem to be low-sensitive markers of renal disease in persons with GFR ≤60 mL/min³; therefore, cystatin C has been recommended as a marker for evaluating renal function. Cystatin C appears to be more sensitive for detecting reduced GFR than either creatinine or estimated GFR.^4,5 Although both creatinine and cystatin C are freely filtered in the glomerulus, a major difference is that creatinine is secreted by renal tubules, whereas cystatin C is metabolized by the proximal tubule and only a small fraction appears in the urine.⁶ Cystatin C has also been shown to be a stronger predictor of adverse outcomes than serum creatinine, so it could be used as a cardiovascular risk marker.^7,8

Although serum cystatin C level is used as renal disease progression marker, several equations have been developed to estimate GFR from serum cystatin C.^9–15 We compared the results obtained by these equations to test the homogeneity of their results as well as their usefulness in clinical practice.

DESIGN AND METHODS

Seven hundred and twenty-seven Caucasian outpatients were studied. Of these, 439 were male and 288 were female, and their mean age was 60.8 ± 24.1 years. Two hundred and thirty-seven patients were older than 65 years. Kidney function was evaluated by measuring serum cystatin C using a BNII nephelometer (Siemens Diagnostics, Deerfield, IL, USA) through a particle-enhanced immunonephelometric assay (N Latex Cystatin C). The range for the assay was 0.195–7.330 mg/L, with the reference range for young healthy individuals reported as 0.53–0.95 mg/L. The cut off for the highest quartile of serum cystatin distribution was 1.33 mg/L. Mean creatinine value was 1.26 ± 0.75 mg (men 1.31 ± 0.76 vs. women 1.19 ± 0.74, p = 0.032, Student’s t-test).

GFR was estimated from uncalibrated serum creatinine using the abbreviated Modification of Diet in Renal Disease (MDRD-4) equation.^16,17 Because Caucasian outpatients were included in this study, they were not included in calculation. Patients were classified according to K/DOQI stages of chronic renal disease following the results of MDRD formula¹⁸: 1.5% were in stage V, 12.1% were in stage IV, 23.3% were in stage III, 18.6% were in stage II, and the remaining patients had GFR higher than 90 mL/min (44.5%). GFR was estimated from serum cystatin C levels using five different equations (all equations are described in Table 1).

Table 1. Equations used to estimate GFR from serum cystatin.

Author
Hoek et al.^10	GFR = −4.32 + (80.35 × 1/cystatin C)
Filler and Lepage^13	GFR = e^[1.962 + (1.123 × log(1/cystatin C))]
Larsson et al.^14	GFR = 77.239 + cystatin C^−1.2623
Hojs et al.^9	GFR = 90.63 × cystatin C^−1.192
Simple^9	GFR = 100/cystatin C

Note: GFR, glomerular filtration rate.

Results are expressed as mean ± 1 SD. All statistical tests were two-sided. p-Values lower than 0.05 were considered significant. Student’s t-test for paired samples was used for continuous variables, and χ² test was used for categorical variables for comparison between groups. To evaluate concordance among the measures the Pearson Coefficient correlation test was used and receiver operating characteristic (ROC) curves were plotted. Analysis of the test was done using the statistical package PASW 17.0 (Armonk, NY, USA).

RESULTS

Table 2 shows the differences of estimated GFR among the serum cystatin-derived equations and the classical MDRD-4 equations. The simplest formula rendered the highest estimated GFR and the Hoek equation rendered the lowest GFR, even significantly lower than the MDRD-4 (p < 0.001 Student’s t-test). From the simplest formula (106.0 ± 44.6 mL/min/1.73 m²) to the Hoek equation (81.0 ± 35.7 mL/min/1.73 m²) mean difference was calculated as 25.1 ± 8.7 mL/min (p < 0.001, Student’s t-test).

Table 2. Differences by gender.

	Cystatin C (mg/L)	Hojs S^a	Hojs	Larrson*	Filler	Hoek*	MDRD*
Men	1.16 ± 0.7	106.8 ± 42.4	99.8 ± 46.0	86.4 ± 41.5	100.0 ± 43.5	81.7 ± 33.9	84.6 ± 42.7
Women	1.23 ± 0.7	104.9 ± 47.7	98.3 ± 52.1	84.9 ± 47.3	98.1 ± 49.4	80.0 ± 38.3	83.5 ± 48.7
Global	1.19 ± 0.7	106.0 ± 44.6	99.2 ± 48.5	85.8 ± 43.9	99.3 ± 45.9	81.0 ± 35.7	84.2 ± 45.2

Notes: ^aSimplified Hojs formula, see Table 1. GFR is expressed in mL/min/1.73 m² of body surface. Differences between men and women are not significant. *p < 0.001 versus Hojs equations (Student’s t-test). Difference of Hoek versus MDRD is significant (p < 0.001 vs. Hojs equations, Student’s t-test).

No differences by gender were found among the results of different equations. When elderly people (older than 65 years) were separately analyzed, the results were similar but Larrson and Hoek equations had closer results, very near to those estimated using the MDRD-4 formula (see the Table 3).

Table 3. Elderly patients.

	Cystatin C (mg/L)	Hojs S^a	Hojs	Larrson*	Filler*	Hoek*	MDRD*
Men	1.42 ± 0.7	87.1 ± 39.0	78.6 ± 41.8	67.0 ± 37.7	79.5 ± 40.0	65.7 ± 31.3	65.0 ± 38.5
Women	1.42 ± 0.7	88.8 ± 44.0	80.0 ± 47.8	69.0 ± 43.3	81.5 ± 45.4	67.0 ± 35.3	67.2 ± 44.5
Global	1.42 ± 0.7	88.0 ± 41.3	79.7 ± 44.9	68.0 ± 40.6	80.5 ± 42.7	66.4 ± 33.4	66.1 ± 41.6

Notes: ^aSimplified Hojs formula, see Table 1. GFR is expressed in mL/min/1.73 m² of body surface. Differences between men and women are not significant. *p < 0.001 versus Hojs equations (Student’s t-test). Differences between Hoek and MDRD are not significant.

Classification of patients with estimated GFR <60 mL/min also showed wide variations from one formula to another. Nevertheless, all cystatin C-derived equations reduced the number of patients diagnosed with chronic renal failure compared with the MDRD-4 formula. The Larsson (n = 11) and Hoek (n = 16) equations had a minor frequency of misclassifications. Any patient diagnosed as not having renal failure by the MDRD-4 formula was reclassified by the other equations, but many patients were excluded from this diagnosis when cystatin C equations were used. Table 4 shows the differences by frequency of changed diagnosis.

Table 4. Change in the diagnosis of CRF versus MDRD.

	CRF	Normal function	Agreed
Hojs	0	69	665
Hojs S^a	0	122	612
Larsson	0	11	723
Filler	0	88	646
Hoek	0	16	718

Notes: ^aSimplified. Differences are significant for all equations (p < 0.001, Student’s t-test).

Pearson’s correlation of the MDRD-estimated GFR was highly significant for all the equations: Hojs, 0.999; simplest, 0.999; Hoek, 0.994; Filler, 0.995; and Larsson, 1. ROC curves were plotted for an estimated GFR <60 mL/min using the MDRD-4 equation (Figure 1). Again, Larsson formula formed a closer relationship with MDRD-4 and the simplest equation had the lowest correlation.

Figure 1. ROC curve comparison.

DISCUSSION

Wide differences were observed from one formula to another when GFR was estimated from serum cystatin C. Nevertheless, all cystatin C-derived equations reduced the diagnosis of kidney failure when compared with the usual MDRD-4 equation.

The K/DOQI guidelines recommended estimation of GFR by using prediction equations based on serum creatinine determinations.⁴ The abbreviated MDRD¹⁹ equation was advocated in these guidelines because it correlated well with GFR measured from iothalamate clearance.¹⁶ It also performed as a more complicated MDRD equation that required serum urea nitrogen and albumin determinations.¹⁷ In fact it is currently the most used equation all over the world. Current guidelines and recommendations for patients with chronic kidney disease (CKD) also call for reducing the dosage of drugs excreted by the kidney, avoiding contrast media for imaging procedures and phosphate-based enemas in preparation for colonoscopy, and setting lower targets for cardiovascular risk factors in patients with decreased GFR.^20–22 GFRs falsely estimated as low could therefore lead to insufficient drug dosing, withholding of important diagnostic tests, and overaggressive cardiovascular risk factor reduction in patients without CKD. Therefore, the effect of different estimates of GFRs on clinical decision making should be cautiously evaluated. Our results suggest that MDRD-4 underestimated the GFR, so it should be used cautiously. This problem has been previously reported, and recently, a new formula, the CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration),²³ has been proposed to improve the estimation of GFR in patients without CKD.

The most common measure used to assess overall kidney function is the plasmatic creatinine concentration. Interpretation of this index is complicated, as it is inversely proportional to the GFR and varies between individuals based on differences in age, sex, and muscle mass. Furthermore, creatinine-based measures seem to be low-sensitive markers of renal disease in persons with GFR ≤60 mL/min/1.73 m³. Cystatin C has been used as an alternative to evaluate renal function. Cystatin C has also been shown to be a stronger predictor of adverse outcomes than serum creatinine, so that it could be a cardiovascular risk marker.^7,8

Differences in renal tubular handling of creatinine and cystatin C may have affected their different ability to predict GFR. Creatinine is secreted by tubular cells, increasing with decreasing renal function,²⁴ whereas at least 99% of filtered cystatin C is reabsorbed and degraded by tubular cells.²⁵ Furthermore, both cystatin C and creatinine levels are influenced by changes in thyroid function.^26,27 Cystatin C levels are decreased in the hypothyroid state and increased in the hyperthyroid state compared with the euthyroid state, whereas the opposite prevails for creatinine. Large doses of glucocorticoids increase cystatin C levels in contrast to creatinine levels.²⁷

Apart from distinctly different sources of errors between creatinine and cystatin C in estimating GFR, there may still be a number of unrevealed factors affecting the analyses. The marker used to measure GFR may have had a strong influence in the results. Filler and Lepage¹³ validated the (99m)Tc-DTPA single-injection technique to estimate GFR from cystatin C by using nuclear medicine GFR clearance studies. Hojs et al.⁹ also used a nuclear medicine technique but with a different marker, (51) Cr-EDTA. Both equations rendered close results for estimated GFR in our sample. In fact, these isotopic methods rendered similar results when compared.²⁸ In the Larsson study plasma creatinine and cystatin C were compared with iohexol clearance and two equations were described accordingly in a way that cystatin was calibrated (Dade Behring or DakoZytomation; we selected the first because it was the method used in our laboratory).¹⁴ GFR was determined with [125I]iothalamate in the study of Hoek et al.¹⁰ This was the same method used to create the MDRD-4 formula, which explains why these two equations (MDRD-4 and Hoek) gave the closest results for the estimation of GFR. It could be concluded that every formula agreed with the method directly used to measure GFR and, most likely, this is the reason for the differences detected.

Sample size and selection criteria could also have some influence. Hoek et al. used plasma samples obtained from 93 consecutive patients (from 11 to 73 years old) with several kinds of CKD and from 30 patients with type 2 diabetes mellitus. Hojs studied 764 adult (older than 18 years) patients with CKD. Filler et al. selected 536 children with various renal pathologies. Larrson et al.²⁹ recruited 94 consecutive patients (ranged 32–94 years old) with iohexol-calculated GFR <30 mL/min/1.73 m². The etiology of the GFR reduction was not recorded; the homogeneity of samples in relationship to the origin of CKD is unknown. In this regard, most patients in our sample were not CKD patients. On the contrary, big differences could be found in the age of recruited patients, from children to very old people. Moreover, the sample included a small group of advanced stage CKD patients (stages IV and V); these differences could be lesser in this group of population as in the comparison of MDRD and CKD-EPI formulations.²³

Recent reports indicate that the arithmetic^30,31 or geometric mean^12,32 of plasma creatinine and cystatin C-based GFR equations or deriving GFR equations including both plasma creatinine and cystatin C^27–33 improves estimated GFR compared with using these parameters separately. Nevertheless, this improvement is small but it could still be valuable and this could be an alternative for getting an assured result while estimating GFR when very big differences are found between both estimations.

Wide differences were found among the equations studied and no recommendation could be made about the best equation to be used for estimating GFR from cystatin C. Higher GFR could be expected than when MDRD-4 is used to estimate this parameter, although Hoek and Larson equations rendered closer values when compared with the other cystatin-derived equations. Nephrologists should be aware of this problem when testing renal function using cystatin C.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.