| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Other Areas of Clinical Chemistry |
1 Division of Nephrology, Tufts-New England Medical Center, Boston, MA.
2 Johns Hopkins Medical Institutions, Baltimore, MD.
3 Department of Quantitative Health Sciences, Cleveland Clinic Foundation, Cleveland, OH.
4 National Institute of Diabetes, Digestive and Kidney Diseases, Bethesda, MD.
5 Department of Clinical Pathology, Cleveland Clinic Foundation, Cleveland, OH.
aAddress correspondence to this author at: Division of Nephrology, Tufts-New England Medical Center, 750 Washington St., Box 391, Boston, MA 02111. Fax 617-636-8329; e-mail lstevens1{at}tufts-nemc.org.
 |
Abstract |
|---|
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Methods: Serum specimens included creatinine reference materials prepared by the College of American Pathologists (CAP), traceable to primary reference material at the NIST, with assigned values traceable to isotope dilution mass spectrometry (IDMS), a calibration panel prepared by the Cleveland Clinic Research Laboratory (CCRL), and frozen samples from the MDRD Study. Split specimens were measured at the CCRL using the Roche enzymatic and Beckman CX3 kinetic alkaline picrate assays.
Results: Roche enzymatic assay results on CAP samples were comparable to IDMS-assigned values. Beckman CX3 assay results in 2004–2005 were significantly higher than but highly correlated with simultaneous Roche enzymatic assay results (r2 = 0.9994 on 40 CCRL samples) and showed minimal but significant upward drift from Beckman CX3 assay results during the MDRD Study in 1989–1991 (r2 = 0.9987 in 253 samples). Combining these factors, standardized Scr = 0.95 x original MDRD Study Scr. The reexpressed 4-variable MDRD Study equation for Scr (mg/dL) is GFR = 175 x standardized Scr–1.154 x age–0.203 x 1.212 (if black) x 0.742 (if female), and for Scr (µmol/L) is GFR = 30849x standardized Scr–1.154 x age–0.203 x 1.212 (if black) x 0.742 (if female) [GFR in mL · min–1 · (1.73 m2)–1].
Conclusion: When the calibration of Scr methods is traceable to the Scr reference system, GFR should be estimated using the MDRD Study equation that has been reexpressed for standardized Scr.
|
Introduction |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
To date, the relationship of the MDRD Study creatinine assay to standardized creatinine was unknown, although results of smaller studies suggested the CX3 rate Jaffe assay had a small positive bias compared with an IDMS reference method (10)(11). The purposes of this report are to describe procedures for calibrating the Scr assay at the Cleveland Clinic Research Laboratory (CCRL) where the MDRD Study samples were measured and to reexpress the 4-variable (modified) MDRD Study equation (6) for use with standardized Scr.
|
Materials and Methods |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
). First, CAP Creatinine Accuracy Calibration Verification/Linearity Survey LN24 (LN24) samples were used as trueness controls to verify that the calibration of the Roche enzymatic method was traceable to LC-IDMS and correctly implemented in the CCRL. Second, values measured with the Roche enzymatic assay on a calibration panel of 40 native pooled sera prepared by the CCRL were compared with values assigned using the Beckman CX3 assay, the assay that was used during the MDRD Study and is currently in use in the CCRL. Third, with frozen samples from the MDRD Study, the Beckman CX3 assay was adjusted for drift over the past decade. The study was approved by the institutional review boards of all participating institutions.
|
reference samples from cap
CAP 2004 LN24 samples 01 through 07 were prepared from a female-only donor pool, so the creatinine value would be slightly lower than a mixed-sex pool (10)(11)(13). Sample 02 is the base female serum pool. Sample 07 had reagent-grade creatinine added to bring the creatinine to 
352 µmol/L (4 mg/dL). Samples 03 through 06 were prepared by gravimetric admixing of samples 02 and 07. Sample 01 was prepared from sample 02 by gravimetric dilution with 0.01 mol/L phosphate buffered saline [prepared by adding one packet of Sigma P-3813 phosphate-buffered saline, pH 7.4, to 1 L of deionized water (personal communication, Mary Zimmer, January 22, 2007)].
Creatinine concentrations in samples LN24-02 and LN24-07 were value-assigned by the NIST using an LC-IDMS method. NIST-assigned values for samples 02 and 07 were 65.032 and 353.32 µmol/L (0.7390 and 4.0150 mg/dL), respectively. The creatinine values in the other pools were computed based on the known admixtures.
calibration panel
We developed a calibration panel of pooled human sera to establish the calibration relationship between the CX3 (2004) and Roche enzymatic methods across the range of Scr for use in patient samples. The calibration panel was developed at the CCRL from pooled patient sera from the Cleveland Clinic. The calibration panel included 40 reference sera (20 aliquots of 1.0 mL each frozen at –70 °C) pooled from at least 10 mixed-sex donors known to have Scr values covering the full range of 177 to 442 µmol/L (0.5 to 5.0 mg/dL).
Briefly, serum pools were constructed by combining within 2 h of collection excess clear serum obtained from apparently healthy patients and patients with CKD as soon as routine testing was completed. Sera were refrigerated and combined according to the creatinine concentration to achieve a final pool volume of 
25 mL. The serum pools were mixed by gentle inversion and filtered; their combined creatinine concentrations were measured, and then serum was apportioned into separate 1.0 mL aliquots and frozen at –70 °C. A set of 20 aliquots was thawed and analyzed in triplicate in 3 separate runs on the same day. This process was repeated on a 2nd set of 20 aliquots on a subsequent day. Each sample was analyzed with the Roche enzymatic and CX3 assays after verification that the methods were and remained within internal quality control limits. The runs also included a CAP sample C-02 from the 2003 C mailing (11) prepared according to Clinical and Laboratory Standards Institute 37A (14) as a validated reference material for each 20 pooled specimens. Runs were deemed acceptable if the concentration value for this material remained within 1 SD of the peer method mean.
Assigned values for the calibration panel were based on the Roche enzymatic assay. This method was selected because it is free from most interferences and its calibration is traceable to IDMS (15). In addition, as shown later, this method was verified to recover values assigned by NIST to the CAP trueness controls.
stored samples from the mdrd study
Samples from the MDRD Study were fresh frozen at –70 °C without thawing until 2004. The MDRD Study equation was derived using samples from the 1st baseline visit (B0) during 1989–1991. The 253 samples used in the current study were from the 3rd baseline visit (B3, 3 months later) and selected sequentially from the serum repository.
creatinine assays at the cleveland clinic foundation research laboratory
The Roche enzymatic assays were performed on a Roche/Hitachi P module automated analyzer with Creatinine Plus enzymatic assay reagents. The enzymatic method is based on the determination of sarcosine after conversion of creatinine with creatininase, creatinase, and sarcosine oxidase. This assay demonstrated CVs of 1.8% and 2.0% at creatinine values of 518 and 78.7 µmol/L (5.86 and 0.89 mg/dL), respectively, in 2004 (n = 194) and 1.1% and 1.6% at creatinine values of 340 and 88.4 µmol/L (3.84 and 1.00 mg/dL), respectively, in 2005 (n = 409).
The Beckman modified kinetic rate alkaline picrate (Jaffe) reaction was performed with the Beckman Synchron CX3 method during the MDRD Study and in 2004. Measurement of the picrate-creatinine complex formation was taken at 520 and 560 nm at 25.6 s after sample introduction. The Beckman CX3 demonstrated CVs of 2.6% and 5.6% at creatinine values of 571 and 78.7 µmol/L (6.46 and 0.89 mg/dL), respectively, in 2004 (n = 390) and 2.4% and 3.8% at creatinine values of 359 and 93.7 µmol/L (4.06 and 1.06 mg/dL), respectively, in 2005 (n = 165).
CAP proficiency test results (n = 25, 2004 C01 to 2005 C10) for these 2 methods at the CCRL demonstrated a mean percentage bias to the peer method target mean of 1.01% (range, 3.46%–5.11%) and 1.79% (range, 2.09%–9.49%) for the Modular P and CX3, respectively. All results were within acceptance criteria. The range of creatinine values for these challenges was 61.9 to 663 µmol/L (0.7 to 7.5 mg/dL).
computations
The mean values of repeated measurements were used for all computations. Linear regression slopes and intercepts were obtained for each comparison of assays. Intercepts that were very small and nonsignificant (P >0.05) were dropped from the regression. The final calibration relationship was obtained by combining point estimates for slopes and intercepts into a single new equation. The final creatinine calibration factor was rounded to 2 significant digits. The SE for the final calibration factor was computed with the delta method (16). Ordinary least-squares regressions were used instead of errors-in-variables regressions because correlations were >0.993. These extremely high correlations indicated the measurement error variance was a very small fraction of the total variance in this calibration setting, at which samples spanned a wide range of creatinine values and assays were averaged across repeated measurements. The final creatinine calibration factor was then incorporated into the constant in the MDRD Study GFR estimating equation.
|
Results |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
. The Roche assay recovered the NIST-assigned values for the LN24 trueness controls, confirming that the Roche method was calibrated to be traceable to IDMS. Assays of the LN24 survey materials in 2004 and 2005 showed nearly identical regression parameters (Fig. 2
). For all samples combined, the intercept (SE) was –3.01 (0.88) µmol/L [–0.035 (0.01) mg/dL; P = 0.02], and the slope was 1.00 (0.004), r2 = 0.9999. Assay results for samples 01 to 07 were all within 6% of the LN24 survey assigned values [with all values >88 µmol/L (1.0 mg/dL) assaying within 2%]; Statistically, the data suggest that 3.01 µmol/L (0.035 mg/dL) could be subtracted from the Roche enzymatic assay to get the best comparability to NIST-assigned values. Because this difference is small and its inclusion would complicate calculations and the final equation, however, this difference was omitted, and the Roche enzymatic values were used without correction to assign reference values.
|
|
Standardization of Beckman Synchron CX3 method to Roche enzymatic assay using the CCRL calibration panel.
The Beckman Synchron CX3 and Roche enzymatic methods were highly correlated (Fig. 3
). The regression of the Roche enzymatic on the Beckman Synchron CX3 method showed a slope (SE) of 0.915 (0.009) and intercept of –2.30 (2.21) µmol/L [–0.026 (0.025) mg/dL]. The P value for the intercept was 0.31. After the small and nonsignificant intercept was dropped, the slope (SE) was 0.906 (0.004; r2 = 0.996).
|
Adjustment of Beckman Synchron CX3 for drift over time by re-assay of MDRD Study samples.
Analysis of 253 frozen MDRD Study samples in 2004 revealed a change over the past decade (Fig. 4
). The regression showed a slope of 1.037 (0.008) and intercept of 2.53 (2.12) µmol/L [0.029 (0.024) mg/dL], with an intercept P value of 0.23. After the small and nonsignificant intercept value was dropped, slope was 1.046 (0.0024), r2 = 0.9986.
|
computation of the idms-traceable calibration factors and reexpression of the mdrd study equation
The general method for calibrating the CCRL assays to the IDMS assay at NIST is given in Table 2
, equation 1. For all 3 comparisons, intercepts for the regression were taken to be zero, so the final calibration was derived by multiplication of regression slopes (1.0 · 0.906 · 1.046 = 0.95; Fig. 1
). The SE of this correction factor, calculated with the delta method and assuming the Roche enzymatic method to be equivalent to the gold standard method, was 0.005. The published MDRD Study equation (Table 2
, equation 2) was reexpressed by substitution of equation 1 for the term Scr (Table 2
, equation 3). Sensitivity analyses retaining intercepts and retaining additional significant digits to the final calibration factors to compute estimated GFR did not show clinically meaningful differences (see the Data Supplement that accompanies the online version of this article at http://www.clinchem.org/content/vol53/issue4).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The National Kidney Disease Education Program (NKDEP) has initiated a creatinine standardization program to improve and normalize Scr results used in estimating equations (8). After creatinine reference materials that are traceable to higher order reference standards are developed, a proficiency testing system will be used to enable ongoing monitoring of calibration among clinical laboratories. This process is expected to be completed by 2008. Reexpression of the MDRD Study equation based on standardized assays will enable implementation of reporting estimated GFR in clinical practice using calibrated Scr assays, thereby overcoming this limitation to the current use of GFR estimating equations.
Using the 2004 CAP LN24-A survey samples, we found the Roche enzymatic assay is comparable to IDMS across a range of Scr values from 0.5 to 4.0 mg/dL. Note that CAP LN24 survey samples were used as recommended for trueness control, validating the Roche assay, rather than as calibration materials (8). Using the CCRL calibration panel and adjusting the Beckman CX3 assay for drift since the MDRD Study, we calibrated the Scr assay at the MDRD Study laboratory and reexpressed the 4-variable MDRD Study equation for use with creatinine methods traceable to an IDMS reference measurement procedure (8).
Based on the Roche method’s performance in the CAP LN24, creatinine results from the Roche enzymatic method were considered to be traceable to IDMS values. These results have consistently showed the Roche enzymatic method to give results in agreement with IDMS target values. The LN24 samples were verified by the NKDEP Laboratory Working Group to have results that were commutable with those for native clinical samples for the Roche enzymatic creatinine method (personal communication, Greg Miller, December 12, 2006). From a practical perspective, the CAP LN24 is the only material currently available for use as a trueness control and is reasonable to use for a clinical laboratory verification.
These methods may not fully account for measurement error. Our study design includes a 2-step approach to IDMS-traceable calibration of the MDRD Study samples, using combined results from 2 regressions based on split samples rather than results of a single regression based on reassay of original MDRD Study specimens on the Roche enzymatic assay. Nonetheless, our approach is robust. First, the CCRL calibration panel was prepared with rigorous techniques, and the MDRD Study stored samples were collected at the same time as the samples used to develop the equation. Second, several factors make multiplication of the regression slopes a maximum likelihood estimator with high efficiency. The 2 regressions have extremely high correlations (r2 = 0.987 and 0.994), making the loss of efficiency very small. These extremely high correlations over a wide range of creatinine show the linear relationship between the different assays. The SEs for all comparisons performed in this study were minute. Omission of the intercepts is justified by their small magnitude and absence of a statistically or clinically meaningful effect on GFR estimates in sensitivity analyses. The final IDMS-traceable calibration factor of 0.95, relating original MDRD Study Scr measures to standardized creatinine, is our best approximation. Regardless of study design, storage effects on specimens are possible, but the small difference between results on thawed MDRD Study specimens assayed a decade later and CAP samples assayed a year later suggests that any storage effects are small.
The 2003 CAP survey of 5624 clinical laboratories, using a fresh-frozen serum reference material with an assigned value of 79.38 µmol/L (0.902 mg/dL) by IDMS, showed a range of method-dependent bias across laboratories from an underestimate of 5.28 µmol/L (0.06 mg/dL) to an overestimate of 27.28 µmol/L (0.31 mg/dL) (10). This finding suggests that in many laboratories, standardization of Scr assays will lead to decreased reported Scr concentrations, requiring redefinition of the reference range. Without reexpression of estimating equations, lower reported values for Scr would increase GFR estimates. Reexpression of the MDRD Study equation according to the standardized assay enables consistent interpretation of estimated GFR by use of this equation. Use of other GFR estimating equations will require reexpression of the equations with standardized Scr. The effect of standardization of Scr assays on urine creatinine results has not been studied. Thus, it is not possible at this time to determine the effect of standardization on measurements and estimates of creatinine clearance.
Creatinine calibration is only 1 limitation of the current estimating equations. Despite calibration, performance of the MDRD Study equation in populations with higher GFRs appears worse than in populations with lower GFRs. Possible reasons for decreased accuracy include reduced creatinine generation attributable to loss of muscle mass or decreased protein intake, especially in elderly and chronically ill persons, greater measurement error and biological variation in GFR at higher GFR levels, and limitations of generalizing equations developed in populations with CKD to populations without CKD (38). The Chronic Kidney Disease Epidemiology Collaboration is addressing these issues by use of pooled analysis of individual patient data and IDMS-traceable calibration of stored specimens by standardized methods.
NKDEP currently suggests that laboratories should report estimated GFR using the original 4-variable MDRD Study equation (Table 2
, equation 2), even without IDMS-traceable calibration of Scr assays, recognizing lesser accuracy, especially at levels of GFR >60 mL · min–1 · (1.73 m2)–1 (17). With the availability of appropriate IDMS-traceable calibrator materials from NIST (SRM 967) and trueness control materials from the CAP (LN24 Survey), clinical laboratories can establish and maintain IDMS-traceable calibrated Scr assays and use reexpressed estimating equations, such as the 4-variable MDRD Study equation, to report GFR estimates (Table 2
, equation 3). Data presented here and in the 2003 CAP survey suggest the Roche enzymatic method meets these criteria. Until improved equations are developed, it may be appropriate to report a specific numeric result only for estimated GFR <60 mL · min–1 · (1.73 m2)–1, as is recommended by current guidelines (4). Although differentiating higher estimates may be useful in research studies, higher values can be reported as ">60 mL · min–1 · (1.73 m2)–1" for clinical reports.
|
Acknowledgments |
|---|
|
Footnotes |
|---|
|
References |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The following articles in journals at HighWire Press have cited this article:
|
|
 |
|
  M. Chonchol, G. Lippi, M. Montagnana, M. Muggeo, and G. Targher Association of inflammation with anaemia in patients with chronic kidney disease not requiring chronic dialysis Nephrol. Dial. Transplant., September 1, 2008; 23(9): 2879 - 2883. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Stavroulopoulos, O. Sahota, I. Qamar, C. J. Porter, M. J.D. Cassidy, and S. D. Roe Presence of chronic kidney disease stages 3 and 4 in patients with low trauma fracture Nephrol. Dial. Transplant., August 1, 2008; 23(8): 2710 - 2711. [Full Text] [PDF]
|
|
|
|
 |
|
 H. E. van Deventer, J. A. George, J. E. Paiker, P. J. Becker, and I. J. Katz Estimating Glomerular Filtration Rate in Black South Africans by Use of the Modification of Diet in Renal Disease and Cockcroft-Gault Equations Clin. Chem., July 1, 2008; 54(7): 1197 - 1202. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. R. Delanghe How to estimate GFR in children Nephrol. Dial. Transplant., May 29, 2008; (2008) gfn306v1. [Full Text] [PDF]
|
|
|
|
|
|
J. Coresh, L. A. Stevens, and A. S. Levey Chronic kidney disease is common: What do we do next? Nephrol. Dial. Transplant., April 1, 2008; 23(4): 1122 - 1125. [Full Text] [PDF]
|
|
|
|
 |
|
 B. G. Jaar, R. Khatib, L. Plantinga, L. E. Boulware, and N. R. Powe Principles of Screening for Chronic Kidney Disease Clin. J. Am. Soc. Nephrol., March 1, 2008; 3(2): 601 - 609. [Full Text] [PDF]
|
|
|
|
|
|
F. Ceriotti, J. C. Boyd, G. Klein, J. Henny, J. Queralto, V. Kairisto, M. Panteghini, and on behalf of the IFCC Committee on Reference Inter Reference Intervals for Serum Creatinine Concentrations: Assessment of Available Data for Global Application Clin. Chem., March 1, 2008; 54(3): 559 - 566. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. Delanaye, E. Cavalier, and J. M. Krzesinski Determining Prevalence of Chronic Kidney Disease Using Estimated Glomerular Filtration Rate JAMA, February 13, 2008; 299(6): 631 - 631. [Full Text] [PDF]
|
|
|
|
|
|
J. Coresh, L. A. Stevens, and A. S. Levey Determining Prevalence of Chronic Kidney Disease Using Estimated Glomerular Filtration Rate--Reply JAMA, February 13, 2008; 299(6): 631 - 632. [Full Text] [PDF]
|
|
|
|
|
|
M. P. Quinn, A. Rainey, K. J. Cairns, A. H. Marshall, G. Savage, F. Kee, A. Peter Maxwell, E. Reaney, and D. G. Fogarty The practical implications of using standardized estimation equations in calculating the prevalence of chronic kidney disease Nephrol. Dial. Transplant., February 1, 2008; 23(2): 542 - 548. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P.D. Giles, P.B. Rylance, and D.C. Crothers New results from the Modification of Diet in Renal Disease study: the importance of clinical outcomes in test strategies for early chronic kidney disease QJM, February 1, 2008; 101(2): 155 - 158. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. M. Parry Use of Single-Value Protein Compensation of the Jaffe Creatinine Assay Contributes to Clinically Significant Inaccuracy in Results Clin. Chem., January 1, 2008; 54(1): 215 - 216. [Full Text] [PDF]
|
|
|
|
|
|
J. Coresh, E. Selvin, L. A. Stevens, J. Manzi, J. W. Kusek, P. Eggers, F. Van Lente, and A. S. Levey Prevalence of Chronic Kidney Disease in the United States JAMA, November 7, 2007; 298(17): 2038 - 2047. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
F. Artunc and T. Risler Serum erythropoietin concentrations and responses to anaemia in patients with or without chronic kidney disease Nephrol. Dial. Transplant., October 1, 2007; 22(10): 2900 - 2908. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. A. Stevens, J. Coresh, H. I. Feldman, T. Greene, J. P. Lash, R. G. Nelson, M. Rahman, A. E. Deysher, Y. Zhang, C. H. Schmid, et al. Evaluation of the Modification of Diet in Renal Disease Study Equation in a Large Diverse Population J. Am. Soc. Nephrol., October 1, 2007; 18(10): 2749 - 2757. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
