Differences in Glomerular Filtration Rate Estimates by 2 Cystatin C–Based Equations

Glomerular filtration rate (GFR) represents the best overall index of kidney function (1), and the National Kidney Foundation has recommended that clinical laboratories routinely report an estimate of GFR(1)(2). Several cystatin C–based equations for calculation of GFR have been reported(3)(4). The most recent equations have been published by Larsson et al. (GFRLarsson = 99.43 × cystatin C−1.5837)(4) and Grubb et al. [GFRGrubb = 84.69 × cystatin C−1.680 (× 0.948 if female)](5). Both groups report on GFR estimates obtained from particle-enhanced turbidimetric immunoassay measurements of cystatin C (DakoCytomation). After correcting the Larsson estimate for body surface according to the Du Bois and Du Bois formula (6), we compared both equations, using data from 29 adult renal transplant patients who also had undergone 125I-iothalamate clearance determination as a reference measurement of GFR(7)(8). For cystatin C measurements, we used a particle-enhanced turbidimetric immunoassay (Dako) run on a Cobas Mira instrument (Roche Diagnostics), as described earlier(7)(8). The correlation of the Larsson and Grubb cystatin C-based GFR estimates was highly significant (r = 0.98; P <0.001); however, evaluation of these methods by linear regression showed a slope substantially different from 1, indicating 23% higher GFR values obtained with the Larsson estimate (Fig. 11 ). The Larsson estimates were substantially higher than the Grubb estimates even when adjustment for body surface area was not performed (data not shown). When we compared the slopes and intercepts of these 2 linear regression lines by the method described by Zar (9), a method equivalent to analysis of covariance, the 2 regression lines showed significantly different intercepts (P <0.001), whereas we found no significant difference between the slopes (P = 0.23; GraphPad Prism 4 Software). These data are consistent with 2 distinct but parallel regression lines. The reasons for these differences remain unclear. A possible explanation may be a reformulation of the cystatin C assay (10). Other reasons might include the different procedures to evaluate the equations and the fact that our patients are different from the patient populations in which the equations were evaluated. Standardization of cystatin C assays will lead to better comparability of cystatin C–based GFR estimates. Efforts to establish an IFCC working group focused on standardization of cystatin C measurements have already been initiated. However, standardization of measurements should also evoke consensus recommendations for the use of cystatin C–based prediction equations for GFR. In the meantime, clinical laboratories will need to carefully evaluate cystatin C– and creatinine-based equations for the estimation of GFR before introducing them into routine clinical use. Cystatin C–based estimates of GFR calculated according to the equations of Larsson et al. (4) and Grubb et al.(5). The equation of the linear regression line (dashed) is GFRLarsson [mL · min−1 · (1.73m2)−1] = 1.23 (0.046) × GFRGrubb + 2.19 (1.59). The solid diagonal line indicates equal GFR estimates.