Original Article. Susceptibility of glomerular filtration rate estimations to variations in creatinine methodology: a study in older patients

Transcription

1 Original Article Susceptibility of glomerular filtration rate estimations to variations in creatinine methodology: a study in older patients Edmund J Lamb 1, Joanna Wood 1, Helen J Stowe 1, Shelagh E O Riordan 2, Michelle C Webb 3 and R Neil Dalton 4 Abstract Addresses Departments of 1 Clinical Biochemistry 2 Health Care of the Older Person and 3 Renal Medicine, East Kent Hospitals NHS Trust, Kent and Canterbury Hospital, Canterbury, Kent, UK 4 Department of Paediatrics, Guy s Hospital, London, UK Correspondence Dr Edmund Lamb edmund.lamb@ekht.nhs.uk Background It is recommended that measurement of serum creatinine should be supplemented with a creatinine-based estimation of glomerular filtration rate (GFR). The influence of creatinine methodology on these estimates is not always appreciated. We have studied differences in creatinine methods and their influence on GFR estimation specifically in older people. Methods In all, 46 older patients (mean age 80 y, range y) with predominantly mild or moderate kidney disease were studied. Serum creatinine was measured using a rate Jaffe method and two different enzymatic methods. Isotope dilution mass spectrometry served as the reference creatinine method. GFR was estimated using both the Modification of Diet in Renal Disease (MDRD) and Cockcroft and Gault formulae: a 51 Cr-EDTA GFR estimation served as the reference GFR method. Results Both enzymatic methods produced creatinine results that were significantly different (Po0.001) from the reference method. The Jaffe method over- and underestimated creatinine at low and high concentrations, respectively. The most likely explanation for these differences relates to standardization of the assays. Irrespective of creatinine method, the Cockroft and Gault formula tended to underestimate GFR, and the MDRD formula to overestimate GFR. Use of the differing creatinine methods to estimate GFR produced predictable biases of the estimate, with mean GFR estimates varying by 14% across the creatinine methods. Conclusion Estimates of GFR depend critically upon the accuracy and precision of the creatinine measurement used in their calculation. Introduction Measurement of serum creatinine is an imperfect test of kidney function. Creatinine assays are susceptible to a variety of interferences from non-creati nine chromogens and drugs. Although enzymatic and rate approaches to measurement have reduced these problems to an extent, interferences, particularly from bilirubin, remain an important issue. There is large variation in reported creatinine concentrations using di ering methods, 1 in part re ecting calibration di erences. Further, use of serum creatinine alone results in failure to appreciate the presence of kidney disease in many patients, particularly among the older population. 2 It has been recommended 3 that measurement of serum creatinine should be supplemented with an estimate of the glomerular ltration rate (GFR) using either the Cockcroft and Gault 4 or Modi cation of Diet in Renal Disease (MDRD) study 5 formulae. It is known that these formulae are susceptible to changes in the standardization of creatinine assays, particularly at creatinine concentrations in the normal to high--normal range In the present study we have explored di erences in creatinine results produced by three standard laboratory approaches, speci cally in samples from older people. Veri cation of true creatinine concentration was achieved using an isotope dilution mass spectrometry (ID-MS) approach, which has been proposed as the reference method. 10 The importance of analytical diversity was illustrated by application of the Cockcroft and Gault and MDRD formulae to these results. r 2005 The Association of Clinical Biochemists 11

2 12 Lamb et al. Subjects and methods Patients The patients included in this study have been described previously. 11 Of the 53 patients originally recruited, su cient serum was available in 48 for further analysis. Two further patients were excluded on the basis that their serum creatinine concentrations were 4300 mmol/l and were unrepresentative of the cohort. Physical and biochemical characteristics of the remaining 46 patients (23 men, 23 women) are described in Table 1. In all, 28 patients had moderate kidney disease (GFR o60 ml/min/1.73 m 2 ) and most (n¼43) had GFRs below 80 ml/min/1.73 m. 2 All patients signed a written consent form and were free to leave the study at any time. This study was approved by the East Kent Local Research Ethics Committee (LREC 0019/02/00). Laboratory analytical methods GFR GFR was estimated from a single 51 Cr-EDTA injection and three blood samples (90,180 and 270 min after injection) using mono-exponential analysis with Brochner--Mortenson s correction. 12 GFR measurements were undertaken by quali ed physicists in a nuclear medicine department accredited by the Institute of Physical Sciences in Medicine. All GFR calculations were checked by a specialist accredited in nuclear medicine. Serum creatinine Serum for creatinine measurement was stored at --801C prior to analysis. Creatinine was measured using four di erent methods, all of which were calibrated prior to analysis to ensure optimal performance: 1. Serum creatinine was measured by ID-MS, a modi ed chromatographic isotope dilution electrospray mass spectrometry-mass spectrometry method. 13 Serum (5 ml) was diluted with 250 mlof deionized water containing 500 pmol d 3 -creatinine and 250mL acetonitrile with 0.05% formic acid. Following mixing and centrifuging, supernatant (2 ml) was automatically injected using an HTS PAL autosampler (CTC Analytics AG, Switzerland) into a 200 ml/min solvent stream of acetonitrile:water (50:50) with 0.025% formic acid. Chromatography was performed on a Symmetry- C8,3.5 mm, mm 2 column (Waters Corporation, UK) and precursor/product ion pairs (m/z 114.2/44.2, 117.2/47.2) were acquired in positive ion multiple reaction monitoring mode using an API4000 (Applied Biosystems, UK). Results were calculated using Analyst version Calibrators were prepared by dissolving creatinine (BDH Chemicals Ltd., Poole, Dorset, UK) in 0.1mmol/L HCl. The between-day coe cients of variation (CVs) were o5% at concentrations of 112 and 256 mmol/l. Mean (median, range) serum creatinine measured by ID-MS was mmol/l (94.6 mmol/l, mmol/l). 2. An enzymatic dry slide method on avitros 950 analyser (Ortho-Clinical Diagnostics, Rochester, USA). The assay was calibrated using the manufacturer s recommended material with values traceable to a high-performance liquid chromatography (HPLC) reference procedure. The assay principle utilizes a creatininase/creatinase/sarcosine oxidase system. Bilirubin is reported not to interfere with this method. The between-day CVs were o3% at concentrations of 83 mmol/l and 525 mmol/l. 3. An enzymatic wet chemistry method on an Integra 800 analyser (creatinine plus ver.2, cat. no , Roche Diagnostics Ltd., Lewes, East Sussex, UK). The assay principle utilizes a creatininase/creatinase/sarcosine oxidase system with detection at 552 nm and absorbance blanking at 659 nm. Bilirubin interference is reported to be insigni cant up to 340 mmol/l. The assay is calibrated with a lyophilized human serum cali- Table 1. Physical and biochemical characteristics of patients Mean SD Median Observed range Age Height (m) Weight (kg) BSA (m 2 ) BMI (weight [kg]/height [m 2 ]) Cr-EDTA clearance (ml/min/1.73 m 2 ) Serum creatinine (mmol/l) a Serum bilirubin (mmol/l) a Derived using ID-MS method.

3 Estimation to variations in GFR creatinine methodology 13 brator (C.f.a.s., Roche Diagnostics) in which the creatinine concentration (normally approximately 320 mmol/l) is traceable to an ID-MS determination. Water is used as the zero standard. The between-day CVs were o2.5% at concentrations of 109 and 371 mmol/l. 4. A kinetic Ja e assay on an Integra 800 analyser (cat. no , Roche Diagnostics Ltd.). This is a rate assay without deproteinization, measuring the increase in absorbance at 512 nm between 55 s and 70 s after initiation of the reaction. Absorbance blankingat583nmisused.bilirubininterference is reported to be insigni cant up to 85 mmol/l. The assay is calibrated with C.f.a.s. (Roche Diagnostics) as above. The between-day CVs were o3% at concentrations of 113 mmol/l and 333 mmol/l. Other analytes Serum urea was measured using an enzymatic method on avitros 950 analyser (Ortho-Clinical Diagnostics). Serum albumin was measured using a bromocresol green method and serum bilirubin using a diazotized sulphanilic acid method on an Integra 800 analyser (Roche Diagnostics). The between-day CVs were o3% in all cases. Formulae for estimation of GFR The Cockcroft and Gault 4 clearance estimate was calculated using the formula: GFR (ml/min) ¼ ([140 age [years]] weight [kg]/ [0.814 plasma creatinine [mmol/l]]). A correction factor of 0.85 was used for females. The MDRD 5 estimate of GFR was calculated as: GFR (ml/min/1.73 m 2 ) ¼ 170 (plasma creatinine [mmol/l] ) (age) (0.762 if patient is female) (1.180 if patient is black) (plasma urea [mmol/l]2.801) (plasma albumin [g/l] 0.1). þ Cr-EDTA and the Cockcroft and Gault estimated GFRs were adjusted to an average body surface area (BSA) 14 of 1.73 m 2. The MDRD formula produces GFR estimates already thus adjusted. Laboratory measurements were undertaken by state registered biomedical scientists in a CPA accredited laboratory; sta were blinded to the 51 Cr-EDTA results. Data analysis Which of the creatinine methods gives the closest agreement with ID-MS and are there explanations for any differences observed? ID-MS served as the reference method against which other methods were compared by studying bias (plots of di erence between test method and ID-MS plotted against ID-MS) and agreement (regression analysis). Each of the four serum creatinine data-sets had a nonparametric distribution, as con rmed using the Shapiro-- Wilk W test. Log transformations produced a normal distribution in all cases and these were subsequently used for regression analyses. The signi cance of bias compared with the ID-MS method was tested using the Wilcoxon s matched pairs signed ranks test. The possibility that age, body mass index (BMI: weight [kg]/height [m] 2 ), BSA, serum albumin and bilirubin were contributing to observed di erences between the test methods and ID-MS was tested with regression analysis. What is the contribution of creatinine method differences to variation in estimates of GFR? Each of the four estimations of serum creatinine was used in both the Cockcroft and Gault and MDRD formulae. 51 Cr-EDTA GFR estimation served as the reference method against which formulaic estimates of GFR were compared by studying bias (plots of di erence between test method and 51 Cr-EDTA plotted against 51 Cr-EDTA) and agreement (regression analysis). Each of the nine GFR data-sets had a normal distribution, as con rmed using the Shapiro--WilkW test. The signi cance of bias compared with the 51 Cr-EDTA method was tested using one-wayanova and the paired t-test. Results Both of the enzymatic methods tested produced creatinine results that were signi cantly (Po0.0001) di erent from the ID-MS reference method. Conversely, mean bias with the Ja e method (2.8%) was not signi cantly di erent from the ID-MS method. Overall, however, the enzymatic methods produced a more precise estimate of true creatinine concentration across a range of concentrations, whereas the Ja e method tended to overestimate serum creatinine at low concentrations and underestimate it at higher concentrations (Table 2, Figure 1). There was no relationship between di erence (test method -- ID-MS) and creatinine concentration for the Ortho enzymatic method, but a signi cant negative correlation was observed for the Roche enzymatic and Roche Ja e methods (Figure1). Some relatively weak relationships (R 2 o0.14, 0.054P40.01, in all cases) were observed for some of the methods between di erence (test method -- ID-MS) and bilirubin, albumin, BMI and BSA. No relationships between di erence and age were observed for any of the methods. Given the low signi cance levels

4 14 Lamb et al. Table 2. Relationship of three serum creatinine methods (y) to an ID-MS (x) Creatinine (mmol/l) Regression analysis Difference plot analysis (y x) Limits of agreement (95% CI of estimate) Mean bias (95% CI of mean bias) Lower Upper Range (%) Creatinine method Median Range R 2 Slope Intercept ID-MS Enzymatic (Ortho) * 5.2% (3.8 to 6.7) z 4.2% ( 6.7 to 1.9) 15.4% (12.7 to 18.0) 19.6 Enzymatic (Roche) % ( 10.4 to 7.9) z 17.8% ( 20.0 to 15.6) 0.9% ( 3.0 to 0.9) 16.9 Jaffe rate (Roche) w 2.8% (0.5 to 5.4) 14.6% ( 19.4 to 9.9) 21.1% (16.4 to 26.2) 35.7 All four serum creatinine data-sets demonstrated a non-parametric distribution and regression and difference plot analyses shown were therefore calculated following log transformation. * Significantly (Po0.0005) different from zero. w Significantly (Po0.0001) different from zero. z Significantly (P o0.0001) different from ID-MS Difference between methods (Ortho enzymatic ID-MS, μmol/l) Difference between methods (Roche enzymatic ID-MS, μmol/l) Difference between methods (Jaffe ID-MS, μmol/l) Serum creatinine (ID-MS, μmol/l) Serum creatinine (ID-MS, μmol/l) Zero bias Zero bias Serum creatinine (ID-MS, μmol/l) Zero bias Figure 1. Bias plots showing difference between Ortho enzymatic (top panel), Roche enzymatic (middle panel) and Roche Jaffe (lower panel) creatinine methods and the ID-MS reference method across a range of concentrations. The solid horizontal line indicates zero bias and the broken horizontal lines indicate the mean bias and the 95% limits of agreement. A significant negative correlation was observed between difference (test method ID-MS) and creatinine concentration for the Roche enzymatic (R 2 ¼ 0.63, Po0.0001; difference ¼ ID-MS creatinine þ 2.3) and Jaffe (R , Po0.0001; difference ¼ ID-MS creatinine þ 18.0) methods, but not for the Ortho enzymatic method (R 2 ¼ 0.00). observed, it is reasonable to conclude that these factors were not important confounders of the data. Overall, the Cockcroft and Gault formulae produced an estimate of GFR which was slightly negatively biased compared with 51 Cr-EDTA, whereas the MDRD formulae estimates were positively biased, irrespective

5 Estimation to variations in GFR creatinine methodology 15 of the creatinine method (Table 3). The imprecision of the estimate of GFR (limits of agreement) was slightly smaller with the Cockcroft and Gault formula.with respect to creatinine methods, the Roche enzymatic method produced the highest GFR estimate with both formulae. Mean GFR estimated by both the Cockcroft and Gault and MDRD formulae varied by 14% across the creatinine methods (Table 3). Discussion Di erent serum creatinine methods have varying degrees of accuracy, imprecision and susceptibility to interferences. Spectral interferences remain a signi cant problem in clinical practice, but for the vast majority of samples such interferences exert a minimal e ect in modern assays. Although imprecision remains out with desirable performance criteria de ned in terms of biological variation, 15 accuracy remains a more critical issue. As a result of reaction with non-creatinine chromogens, conventional endpoint Ja e methods were typically judged to overestimate true serum creatinine concentration by approximately 20% at physiological concentrations. Consequently, automated kinetic, enzymatic and chromatographic methods could produce creatinine measurements signi cantly lower than the early Ja e methods. However, since this would result in a relative overestimation of GFR (for example, when measuring or calculating clearance), some reagent and instrument manufacturers have calibrated their assays to produce higher serum creatinine results. Consequently, commercially available creatinine methods may demonstrate a positive bias compared with HPLC or ID-MS methods, particularly at concentrations within the reference range. 1,16,17 A national pro ciency study in the USA demonstrated that laboratories, on average, overestimate serum creatinine by11to15 mmol/l ( mg/dl) at a concentration of 91 mmol/l (1.00 mg/dl). 18 In 2003, the International Measurement Evaluation Programme (IMEP)-17 survey of approximately 1000 laboratories in 35 countries demonstrated almost universal overestimation of serum creatinine in a pool with a certi ed concentration of 75 mmol/l (0.83 mg/dl). Of the fourteen method groups, 13 demonstrated mean positive bias compared with the reference value, typically varying between10 and15%. (The exception was the Roche enzymatic method.) A pool at pathological concentration (169 mmol/l, 1.86 mg/dl) showed an approximately equivalent number of laboratories with positive and negative bias (see for further information). Both these studies used ID-MS as the reference method. Calibration of commercial methods is not standardized, leading to variation between laboratories. This is illustrated in the present study where the two Roche methods which utilize di erent reaction principles but the same calibrator both show increasing negative bias at higher concentrations, whereas the Ortho method, using the same reaction principle as the enzymatic Roche method, shows a small but consistent positive bias across the range studied. Indeed, systematic di erences in the calibration of serum creatinine assays reportedly account for 85% of the observed di erences between laboratories. 18 Pro ciency studies demonstrate that while between-laboratory CVs of o3% are achievable within method groups, overall betweenlaboratoryagreement across methods is much poorer. 1,17 Systematic variation between laboratories of 18 to 36 mmol/l ( mg/dl) is common. 1,3,19 Further, inter- (and within-) laboratory agreement deteriorates as serum creatinine concentration nears the reference interval: the exponential relationship between serum creatinine and GFR means that imprecision at lower creatinine concentrations contributes to greater error in GFR estimation than imprecision at higher creatinine concentrations does. Clearances calculated from formulae will clearly vary depending on how accurate and precise the creatinine measurement is that is used in their calculation. The more a method overestimates true creatinine, the greater will be the underestimation of GFR, and vice versa. For example, in the present study, the Roche enzymatic method gave the highest estimate of GFR with both formulae tested, consistent with the negative bias of the method compared with ID-MS. Conversely, the Ortho method, which showed a consistent slight positive bias across the range studied, produced the lowest estimates of GFR. Overall, GFR estimated using the Cockcroft and Gault formula produced a slight negative bias compared with the reference method, whereas the MDRD formula produced signi cantly positively biased estimates of GFR irrespective of the creatinine method used. Further, GFR estimates calculated using the MDRD method appeared less precise (larger 95% limits of agreement). Our observations are in keeping with the study of Wuyts et al. 9 in younger patients, where enzymatic and Ja e assays also produced higher and lower GFR estimates, respectively, compared with an HPLC method. Furthermore, their MDRD estimates of GFR also tended to be slightly higher and more variable than those calculated using the Cockcroft and Gault formula. The main aim of our study was not to compare the relative performance of these formulae but to illustrate that both are critically susceptible, in a predictable manner, to the de ciencies of the various creatinine assays. Following the recommendations of the National Kidney Foundation (NKF) Kidney Disease Outcomes Quality Initiative (K/DOQI), a burgeoning literature

6 Table 3. Effect of creatinine method on estimates of GFR using the Cockcroft and Gault and MDRD formulae Regression analysis Difference plot analysis Limits of agreement (95% CI), ml/min/1.73 m 2 Mean, ml/ SD, ml/ Intercept, Mean bias (95% CI), min/1.73 m 2 min/1.73 m 2 R 2 Slope min/1.73 m 2 ml/min/1.73 m 2 Lower Upper Range GFR ( 51 Cr-EDTA) Cockcroft and Gault ID-MS ( 6.2 to 1.3) w 19.8 ( 23.9 to 15.8) 12.3 (8.3 to 16.4) 32.1 Enzymatic (Ortho) ( 9.0 to 4.0) z 22.7 ( 26.8 to 18.6) 9.8 (5.7 to 13.9) 32.5 Enzymatic (Roche) ( 1.9 to 3.5) 17.1 ( 21.6 to 12.6) 18.6 (14.1 to 23.1) 35.7 Jaffe rate (Roche) * 5.9 ( 8.7 to 3.1) z 24.4 ( 29.1 to 19.8) 12.7 (8.0 to 17.4) 37.1 MDRD ID-MS (5.5 to 11.6) z 11.8 ( 16.9 to 6.7) 28.9 (23.8 to 34.0) 40.7 Enzymatic (Ortho) (2.5 to 7.8) w 12.4 ( 16.9 to 8.0) 22.7 (18.2 to 27.1) 35.1 Enzymatic (Roche) (10.6 to 17.7) z 9.3 ( 15.2 to 3.4) 37.5 (31.6 to 43.4) 46.8 Jaffe rate (Roche) (3.1 to 8.2) z 11.1 ( 15.3 to 6.9) 22.5 (18.3 to 26.7) 33.6 In all cases, GFR estimates were compared against 51 Cr-EDTA clearance, which served as the reference method. SD, standard deviation; CI, confidence interval. * Significantly (Po0.01) different from zero. w Significantly (Po0.005) different from 51 Cr-EDTA clearance. z Significantly (Po0.0001) different from 51 Cr-EDTA clearance. 16 Lamb et al.

7 Estimation to variations in GFR creatinine methodology 17 has appeared evaluating the relative merits and validity of the MDRD and Cockcroft and Gault formulae The true accuracy of the creatinine methods underlying these formulae has not been an overriding concern in most studies. A major national study, the National Health and Nutrition Examination Survey III (NHANES III), 27 recently adjusted its creatinine results to mimic those generated by the Cleveland Clinic laboratory used for the MDRD study (which used a kinetic Ja e assay on an Astra 8 analyser [Beckman- Coulter Ltd.]); these authors have encouraged others to follow suit 7 and at least one major laboratory provider is known to have done so (see discussion at report.pdf). This is a development that is causing slight concern by mimicking the alignment of HbA 1c (a form of glycosylated haemoglobin) methods to the assay used in the Diabetes Control and Complications Trial, 28 which has resulted in laboratories internationally perpetuating the use of an imperfect standard. 29 The NKF K/DOQI classi cation of kidney disease represents a major opportunity for the assessment and management of CKD to be improved and for better epidemiological data to be provided to facilitate healthcare planning. In practice, however population data and individual patients will be strati ed on the basis of formulae derived using creatinine methods which are critically erroneous. It is essential that the clinical and laboratory communities begin to address problems of creatinine assay standardization and this should not be via alignment to an inaccurate method. Solutions will involve (i) the calibration of commercial methods against certi ed reference material, (ii) the circulation of ID-MS-authenticated serum-based materials in national and international pro ciency survey schemes to raise awareness of the accuracy issue, (iii) revalidation of formulae using creatinine methods traceable to ID- MS methodology (or indeed by using an ID-MS method) and (iv) education of physicians with respect to the changes in the reference range which may be observed. A recent discussion forum led by the National Kidney Disease Education Program ( gov/about/meetings/ gfr_report.pdf) reached similar conclusions. There have been earlier calls for the use of an international standard when calibrating serum creatinine assays, 3 but there remain practical barriers to implementation. Matrix problems have been identi ed when lyophilized preparations with ID-MS-authenticated creatinine concentrations have been used. 1,17 It is clear that serum or protein-based human material must be used to calibrate creatinine assays but this introduces its own inherent problems (i.e. di ering reactivities of the non-creatinine chromogens in the di erent Ja e variations). This is clearly an ambitious programme of work but one which must be embraced sooner rather than later. Acknowledgements We would like to thank Dr AJ Coakley, Dr DE Simpon, Ms K Bridger and Ms R Gawler of the Nuclear Medicine Department for the 51 Cr-EDTA measurements. Dr M Kandarpa assisted with patient recruitment. We would also like to thank Dr D Bullock of the UK National Quality Assessment Scheme for helpful discussions. The study was partly supported by the South East Regional NHS Project Grant Scheme (grant reference SEO 150) and the East Kent Hospitals NHS Trust Internal Project Grant Scheme. Addendum Following the completion of this study, Roche Diagnostics Ltd. modi ed their kinetic Ja e assay. According to the manufacturer, the new, so-called compensated assay (cat. no ) has been standardized against ID-MS and, for the USA, against Standard Reference Material 914. An automatic mathematical correction is now applied to all results to adjust for the reaction of non-creatinine chromogens, which account for an average positive bias of 18 mmol/l in the uncompensated method compared with HPLC, ID-MS and enzymatic methods. Insu cient sample remained for us to compare this method in our patients. However, in 183 patients with chronic kidney disease (median serum creatinine 298 mmol/l, range mmol/l), the compensated method was related to the uncompensated method by the Passing and Bablok regression equation (compensated creatinine ¼ uncompensated creatinine [Dr Susan Vickery, personal communication]). This should yield lower results within the physiological range and is more closely aligned to the same manufacturer s enzymatic assay: Passing and Bablok regression equation (compensated creatinine ¼ 1.03 enzymatic creatinine [manufacturer s data]). It will therefore almost certainly be signi cantly negatively biased compared with the ID-MS method used in our study and consequently will yield higher GFR estimates. It is desirable that methods should be traceable to an ID-MS reference method and that di erent methods by the same manufacturer should agree with each other. The adjustment introduced by Roche Diagnostics Ltd. is similar to that ( 20.3 mmol/l) used to align the NHANES III to the Cleveland Clinic laboratory, 6 alluded to above. However, it should be noted that the in uence of non-creatinine chromogens is not a constant between samples. The purpose of our study was not to criticize particular manufacturers methods but to illustrate the e ects of analytical diversity: this general principle remains unchanged by the recent realignment of the Roche Diagnostics method.

8 18 Lamb et al. References 1 Lawson N, Lang T, Broughton A, Prinsloo P, Turner C, Marenah C. Creatinine assays: time for action? Ann Clin Biochem 2002; 39: Swedko PJ, Clark HD, Paramsothy K, Akbari A. Serum creatinine is an inadequate screening test for renal failure in elderly patients. Arch Intern Med 2003; 163: National Kidney Foundation-K/DOQI. Clinical practice guidelines for chronic kidney disease: evaluation, classification, and stratification. Am J Kidney Dis 2002; 39(Suppl 1): S Cockcroft DW, Gault MH. Prediction of creatinine clearance from serum creatinine. Nephron 1976; 16: Levey AS, Bosch JP, Lewis JB, Greene T, Rogers N, Roth D. A more accurate method to estimate glomerular filtration rate from serum creatinine. Ann Intern Med 1999; 130: Coresh J, Astor BC, McQuillan G, et al. Calibration and random variation of the serum creatinine assay as critical elements of using equations to estimate glomerular filtration rate. Am J Kidney Dis 2002; 39: Coresh J, Eknoynan G, Levey AS. Estimating the prevalence of low glomerular filtration rate requires attention to the creatinine assay calibration. J Am Soc Nephrol 2002; 13: Clase CM, Garg AX, Kiberd BA. Estimating the prevalence of low glomerular filtration rate requires attention to the creatinine assay calibration. J Am Soc Nephrol 2002; 13: Wuyts B, Bernard D, Van Den Noortgate N, et al. Reevaluation of formulas for predicting creatinine clearance in adults and children, using compensated creatinine methods. Clin Chem 2003; 49: Welch MJ, Cohen A, Hertz HS, Ng KJ, Schaffer R, Van der Lijn P. Determination of serum creatinine by isotope dilution mass spectrometry as a candidate definitive method. Anal Chem 1986; 58: O Riordan SE, Webb MC, Stowe HJ, et al. Cystatin C improves the detection of mild renal dysfunction in older patients. Ann Clin Biochem 2003; 40: Blaufox MD, Aurell M, Bubeck B, et al. Report of radionuclides in nephrourology committee on renal clearance. J Nucl Med 1996; 37: Gillingwater SD, Turner C, Swaminathan R, Dalton RN. Measurement of plasma creatinine using low-injection electrospray isotope-dilution tandem mass spectrometry In: Martin S, ed. Proceedings of Pathology London: Association of Clinical Biochemists, 2000; DuBois D, DuBois EF. A formula to estimate the approximate surface area if height and weight are known. AnnInternMed1916; 17: Blijenberg BG, Brouwer HJ, Kuller TJ, Leeneman R, van Leeuwen CJM. Improvements in creatinine methodology: a critical assessment. Eur J Clin Chem Clin Biochem 1994; 32: Blijenberg BG, Brouwer HJ. The accuracy of creatinine methods based on the Jaffe reaction: a questionable matter. Eur J Clin Chem Clin Biochem 1994; 32: Carobene A, Ferrero C, Ceriotti F, et al. Creatinine measurement proficiency testing: assignment of matrix-adjusted ID GC-MS target values. Clin Chem 1997; 43: Ross JW, Miller WG, Myers GL, Praestgaard J. The accuracy of laboratory measurements in clinical chemistry: a study of 11 routine chemistry analytes in the College of American Pathologists Chemistry Survey with fresh frozen serum, definitive methods, and reference methods. Arch Pathol Lab Med 1998; 122: Hsu CY, Chertow GM, Curhan GC. Methodological issues in studying the epidemiology of mild to moderate chronic renal insufficiency. Kidney Int 2002; 61: Bostom AG, Kronenberg F, Ritz E. Predictive performance of renal function equations for patients with chronic kidney disease and normal serum creatinine levels. J Am Soc Nephrol 2002; 12: Vervoort G, Willems HL, Wetzels JFM. Assessment of glomerular filtration rate in healthy subjects and normoalbuminuric diabetic patients: validity of a new (MDRD) prediction equation. Nephrol Dial Transplant 2002; 17: Van Den Noortgate NJ, Janssens WH, Delanghe JR, Afschrift MB, Lameire NH. Serum cystatin C concentration compared with other markers of glomerular filtration rate in the old. J Am Geriatr Soc 2002; 50: Lamb EJ, Webb MC, Simpson DE, Coakley AJ, Newman DJ, O Riordan SE. Estimation of glomerular filtration rate in older patients with chronic renal insufficiency: is the Modification of Diet in Renal Disease formula an improvement? J Am Geriatr Soc 2003; 51: Lin J, Knight EL, Hogan ML, Singh AK. A comparison of prediction equations for estimating glomerular filtration rate in adults without kidney disease. J Am Soc Nephrol 2003; 14: Pierrat A, Gravier E, Saunders C, et al. Predicting GFR in children and adults: a comparison of the Cockcroft Gault, Schwartz, and Modification of Diet in Renal Disease formulas. Kidney Int 2003; 64: Rodrigo E, Fernandez-Fresnedo G, Ruiz JC, et al. Assessment of glomerular filtration rate in transplant recipients with severe renal insufficiency by Nankivell, Modification of Diet in Renal Disease (MDRD), and Cockcroft Gault equations. Transplant Proc 2003; 35: Coresh J, Astor BC, Greene T, Eknoyan G, Levey AS. Prevalence of chronic kidney disease and decreased kidney function in the adult US population: Third National Health and Nutrition Examination Survey. Am J Kidney Dis 2003; 41: Diabetes Control and Complications Trial Research Group. The effect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus. N Engl J Med 1993; 329: Middle J. Standardization of glycated haemoglobin. Ann Clin Biochem 2002; 39: Accepted for publication 7 September 2004