Estimating glomerular filtration rate in the general population: the second Health Survey of Nord-Trondelag (HUNT II)

Nephrol Dial Transplant (6) 21: 1525 1533 doi:1.193/ndt/gfl35 Advance Access publication 28 February 6 Original Article Estimating glomerular filtration rate in the general population: the second Health Survey of Nord-Trondelag (HUNT II) Stein Hallan 1,2, Brad Astor 3,4 and Stian Lydersen 1 1 Faculty of Medicine, Department of Cancer Research and Molecular Biology, Norwegian University of Science and Technology, Trondheim, 2 Department of Medicine, Division of Nephrology, St Olav Universtiy Hospital, Trondheim, Norway, 3 Welch Center for Prevention, Epidemiology, and Clinical Research, The Johns Hopkins University, Baltimore, Maryland and 4 Department of Epidemiology, The Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA Abstract Background. Guidelines recommend the modification of diet in the renal disease (MDRD) formula or the Cockcroft Gault formula for estimating the glomerular filtration rate (GFR). However, there is an ongoing discussion whether the MDRD formula should be used in the general population as several studies have found a large underestimation of its GFR estimates. Methods. In this study, 129 low-risk subjects, eligible for kidney donation according to internationally accepted criteria were selected from the population-based second Health Survey of Nord-Trondelag (HUNT II). Serum creatinine values traceable to isotope dilution mass spectrometry were used with the re-expressed MDRD formula recently published. The 2.5th, 5th and 97.5th percentiles of GFR by age were calculated and compared to reference values from the literature, which are based on GFR measured with gold standard methods in potential kidney donors. Results. The difference between the 5th percentiles for MDRD estimates and measured GFR in the literature was small and constant over age: þ.5 ml/min/1.73 m 2 at age and 2. ml/min/ 1.73 m 2 at age 8. Bias for Cockcroft Gault estimates varied from. ml/min/1.73 m 2 to 21.4 ml/min/ 1.73 m 2. Other formulae also had a too steep age correction, and bias among the elderly varied from 1 to 3 ml/min/1.73 m 2. Hence, 3 8% of the general population above age 6 had GFR estimates below their age-specific 2.5th percentile of normal kidney function, while the MDRD formula was much more conservative (13.3%). Conclusion. The MDRD formula gave nearly unbiased estimates for normal GFR. All other formulae tested Correspondence and offprint requests to: Stein Hallan, IKM, Kreftbygget 5.etg, St Olav University Hospital, N-76 Trondheim, Norway. Email: stein.hallan@ntnu.no had, especially in the elderly, a much larger negative bias and cannot be recommended for use in the general population. Keywords: chronic kidney disease; Cockcroft Gault; general population; glomerular filtration rate; MDRD formula Introduction The US K/DOQI guidelines and European Best Practice guidelines state that kidney function should be assessed with glomerular filtration rate (GFR) estimating formulae, such as Cockcroft Gault or the modification of diet in renal disease (MDRD) study formula, instead of relying on serum creatinine alone [1,2]. The MDRD formula was derived from a large group of Chronic Kidney Disease (CKD) patients using renal clearance of iothalamate as reference method [3]. The methodological foundation should therefore be very strong, and the formula is generally accepted to perform well at GFR levels below 6 ml/min/1.73 m 2. However, there is an ongoing discussion whether the formula should be used in the general population where most subjects have a GFR above this level [4]. Calibration of serum creatinine measurements is an important aspect that often has been overlooked in previous studies. Most laboratories report values that deviate substantially from reference methods like isotope dilution mass spectrometry (IDMS). In the MDRD study, serum creatinine was analysed with a Jaffe method giving unusually low values, and thereby a potential for underestimation of GFR in most other laboratories [5,6]. Rule et al. [7] studied 58 kidney donors and found that the MDRD formula underestimated GFR with 29 ml/min/1.73 m 2. Poggio et al.[8] ß The Author [6]. Published by Oxford University Press on behalf of ERA-EDTA. All rights reserved. For Permissions, please email: journals.permissions@oxfordjournals.org

1526 S. Hallan et al. also studied the performance of the formula in potential kidney donors and found a bias of 9 ml/min/1.73 m 2. Interestingly, neither of the two studies paid sufficient attention to the problems arising from inter-laboratory calibration differences. A new re-expressed version of the four-variable MDRD formula for use with IDMS traceable serum creatinine values was recently published to overcome these problems [9]. GFR measured with reference methods in potential kidney donors can be considered to show the normal GFR range and its change with age. GFR estimated with a creatinine-based formula in a similar group from the general population should give similar results. The second Health Survey of Nord-Trondelag County (HUNT II) is a large population-based study conducted in Norway with a high participation rate. Information is available for proper recalibration of serum creatinine values and also for accurately defining a subpopulation eligible for kidney donation based on repeated measurements of urine albumin, blood pressure, general health, medical history, etc. The aim of our study was to evaluate if GFR, estimated with the new MDRD formula, as well as other frequently used formulae, in such a population-based low-risk group, is in accordance with established reference range of normal kidney function. Material and methods Study sample and design for HUNT II study During the years 1995 97, a large-scale general health survey was conducted in Nord-Trondelag County, Norway. Inhabitants of the county (n ¼ 92 73) aged > were invited to participate in the study, and 7.1% responded. The survey comprised an extensive questionnaire and a brief clinical examination in addition to several clinical chemistry analyses. The population in Nord-Trondelag is stable (net out-migration of.3% per year) and ethnically homogenous (97% Caucasians). Nord-Trondelag County is located in the middle of Norway, and it is a representative for Norway regarding geography, economy, industry, age distribution, morbidity and mortality [1]. All participants signed an informed consent, and additional permissions were obtained from the Regional Ethics Review Committee and the National Data Inspectorate. A more detailed description on objectives, contents, methods and participation in HUNT II has been given elsewhere [11]. Clinical and laboratory investigations The clinical examination included measurement of height, weight and blood pressure. Three consecutive standardized blood pressure measurements were recorded in the sitting position at 1 min intervals using an automatic oscillometric method (Dinamap 845XT; Criticon). Blood was drawn, often in the non-fasting state, from all the participants, and fresh serum and urine samples were analysed at the Central Laboratory of Levanger Hospital on a Hitachi 911 autoanalyser within 2 days. Glucose was measured by means of an enzymatic hexokinase method, total and HDL cholesterol by means of an enzymatic calorimetric cholesterolesterase method. Day-to-day coefficients of variation were 1 2% for all analyses. Urine samples were from three consecutive mornings, and those reporting urine infection during the previous week or menstruation at the time of collection were excluded. Urine albumin was measured by an immunoturbidimetric method (anti-human serum albumin from Dako AS, Denmark) and urine creatinine was measured with the Jaffe method. The albumin-to-creatinine ratio (ACR) was used as an expression for urine albumin excretion. Two or three ACR determinations ranging 2 3 mg/mmol for men and 3 mg/mmol for women were classified as persistent microalbuminuria. Subjects with one or more samples above the microalbuminuric range were classified as macroalbuminuric. GFR estimation Serum creatinine was analysed on fresh serum in 1995 97 using a Jaffé method with sample blank (Roche Diagnostics, Mannheim, Germany). This was a kinetic Jaffe method without deproteinization based on the test principle that creatinine forms a coloured complex with picrate in alkaline solution. The same method was used until 3 when samples from the HUNT II study were thawed and reanalysed to ensure stability of methods over time. Mean difference was.18 mg/dl (1.6 mmol/l) and r 2 was.994 (Pearson correlation coefficient). The laboratory then switched to an enzymatic creatinine test (CREA Plus, Roche Diagnostics, Mannheim, Germany), and a thorough correlation between the old and the new method was established. Formulae used for estimating GFR are given in Table 1. For the MDRD formula, we used two different versions, and serum creatinine values were recalibrated to the same level as used when developing the MDRD formula: (i) New re-expressed four-variable formula for IDMS traceable serum creatinine values: GFR ¼175 ½s-creatinine ðmg=dlþš 1:154 age :3 ð:742 for womenþ ð1:21 for blacksþ ½9Š: Serum creatinine ðmg=dlþ ¼ :31 þ 1:11 serum creatinine This recalibration of Jaffé based creatinine values to the Roche enzymatic method provided IDMS-traceable values. This enzymatic method was the same as used for recalibrating the original MDRD creatinine values to the IDMS level [9], and the method was also found to be comparable to the IDMS level in the Nordic Reference Interval Project [13] (ii) Original four-variable formula: GFR ¼186:3 ðs-creatinineþ 1:154 age :3 ð:742 for womenþð1:21 for blacksþ ½12Š: Serum creatinine ðmg=dlþ ¼ :21 þ 1:8 serum creatinine ½6Š: This recalibration previously provided unbiased estimates of measured GFR by EDTA plasma clearance in a sample of

Estimating GFR in the general population 1527 Table 1. Formulae used for estimating kidney functions in females and males New MDRD [9] F: 175 scr 1.154 age.3.742 M: 175 scr 1.154 age.3 Original MDRD [12] F: 186.3 scr 1.154 age.3.742 M: 186.3 scr 1.154 age.3 Cockcroft and Gault [21] a F: (1 age)/scr (weight/72).85 M: (1 age)/scr (weight/72) Jelliffe and Jelliffe [22] F: {98 [.8 (age )]}/scr.9 M: {98 [.8 (age )]}/scr Salazar and Corcoran [23] a F: (146 age) [(.287 weight) þ (9.74 height 2 )]/(6 scr) M: (137 age) [(.285 weight) þ (12.1 height 2 )]/(51 scr) Mawer et al. [24] a F: weight [25.3 (.175 age)] [1 (.3 scr)]/[(14.4 scr) (7/weight)] M: weight [29.3 (.3 age)] [1 (.3 scr)]/[(14.4 scr) (7/weight)] Bjornsson [25] a F: [25 (.175 age)] weight.7/scr M: [27 (.173 age)] weight.7/scr Gates [26] F: (6 scr 1.1 ) þ (56 age) (.3 scr 1.1 ) M: (89.4 scr 1.2 ) þ (55 age) (.447 scr 1.1 ) Wright et al. [27] a F: (658 38.8 age) BSA.832/(sCr 88.4) M: (658 38.8 age) BSA/(sCr 88.4) Serum creatinine (scr) is measured as mg/dl in all the formulae. a Formulae with results expressed as millilitre per minute were adjusted to a body surface area (BSA) of 1.73 m 2 by multiplying the estimate by 1.73/BSA. 215 kidney patients and potential kidney donors at St Olav Hospital [6]. We also checked for calibration differences between Levanger Hospital and St Olav University Hospital: both the laboratories used the same kinetic Jaffe method with sample blank on Hitachi analysers with reagents from Roche during the entire period 1995 3. Results from 64 monthly external quality controls (Lab Quality, Finland) were available, and there were no significant differences between the laboratories (mean difference 1.1 mmol/l, 95% CI 1. 3.2 mmol/l). Defining a subgroup eligible for kidney donation Subjects from a 5% random sample of HUNT II participants (n ¼ 327) were evaluated, as these had undergone extensive testing for albuminuria. Subjects were excluded, according to internationally accepted criteria for kidney donation, excluded if any of the following were found [14]: (i) macroalbuminuria defined as albumin/creatinine ratio >3 mg/mmol for men and > mg/mmol for women; (ii) diabetes mellitus defined as previously known diabetes or blood glucose 11.2 mmol/l, two or more hours after the last meal; (iii) hypertension defined as 1 mmhg systolic or 9 mmhg diastolic or taking antihypertensive medication; (iv) prior cardiovascular disease (CVD) defined as angina pectoris, myocardial infarction or stroke; (v) serum creatinine above normal range (115 mmol/l for women and 133 mmol/l for men); (vi) malignancy; (vii) body mass index (BMI) >35 kg/m 2 ; (viii) not reporting good or excellent general health. Defining normal GFR level using published data from potential kidney donors We searched the literature for studies reporting individually measured GFR values and age in subjects evaluated for live kidney donation. Reference methods for measuring GFR based on renal or plasma clearance of injected radioisotopes or contrast agents (EDTA, iothalamate, inulin, etc.) were considered. We recorded age and measured GFR from these studies into a common database. Similar methods for estimating percentiles of GFR in this pooled database as well as for the HUNT data were used: a linear regression model was fit to the log-transformed GFR values, estimating the mean and standard deviation of log (GFR) as b þ b 1 age and s, respectively. The appropriateness of log transforming was confirmed by Q Q plots. The 2.5, 5 (median) and 97.5th percentile of GFR for a given age were estimated as exp(b þ b 1 age þ k s) with k as 1.96,, and þ1.96, respectively. Results We searched the literature and found four previous studies reporting individual GFR values by age: Grewal and Blake [15], Rule et al. [16], Hamilton et al. [17] and Davies and Shock [18]. We were able to regenerate 97.5, 99.2, 95. and % of the reported data, respectively, from figures and tables in the articles. Correspondingly, the differences between mean GFR reported in the studies and mean GFR in the regenerated data used in our analysis were 1.,.,.1 and. ml/min/1.73 m 2, respectively. Figure 1 shows that GFR values measured with reference methods in these 1 potential kidney donors ranged from 31 to 162 ml/min/1.73 m 2. Median measured GFR fell from 112.5 ml/min/1.73 m 2 at age to 78.5 ml/min/1.73 m 2 at age 8. Correspondingly, the 2.5th percentile fell from 82.5 ml/min/1.73 m 2 to 57.5 ml/min/1.73 m 2. From the 65 186 participants in the HUNT II study, 3 27 were selected randomly for urine collection. Of these, we identified 1 29 low risk subjects aged 83 who would have been eligible for kidney donation (Figure 2). By definition, none of them had risk factors such as hypertension, diabetes mellitus or prior CVD, while the prevalence of these risk factors in the

1528 S. Hallan et al. 18 16 Measured GFR (ml/min/1.73m 2 ) 1 1 8 6 6 8 Age (years) Fig. 1. The GFR measured with reference methods in 1 subjects eligible for kidney donation from the four previously published studies (2.5th, 5th, 97.5th percentiles). Data are from Grewal and Blake [15], Rule et al. [16], Hamilton et al. [17] and Davies and Shock [18]. Invited: n = 9273 Attended: n = 65186 5% random sample with urine testing: n = 327 3 urine samples returned: n = 2528 Excluded due to: Macroalbuminuria: n = 21 Diabetes mellitus: n = 11 Hypertension: n = 143 Prior CVD: n = 38 Not good/excellent health: n = 258 Abnormals-Creatinine: n = 1 Cancer: n = 21 BMI >35: n = 16 Included as eligible for kidney donation: n = 129 Fig. 2. Selection of low risk subjects, in the HUNT II study, eligible for kidney donation. Many excluded subjects had several reasons for exclusion, but subjects were counted only once. general population was 43.1, 3.4, and 7.9%, respectively. Other characteristics of the participants are given in Table 2. Figure 3 shows individual GFR values estimated with the new re-expressed MDRD formula in the 129 potential kidney donors from the HUNT II study. Estimated GFR ranged from 49 to 182 ml/min/1.73 m 2. Calculated 2.5th, 5th, and 97.5th percentiles are also shown. The 5th percentile for estimated GFR fell from 113. ml/min/1.73 m 2 at age to 76.5 ml/min/1.73 m 2 at age 8. Correspondingly, the 2.5th percentile fell from 79.5 ml/min/1.73 m 2 to 54. ml/min/1.73 m 2. Figure 4A shows that the difference between GFR, estimated with the new re-expressed MDRD formula using directly calibrated creatinine values and measured GFR, was small at all GFR levels and constant throughout the age range: for the 5th percentile, bias ranged from þ.5 ml/min/1.73 m 2 at age to 2. ml/ min/1.73 m 2 at age 8. For the 2.5th percentile, bias ranged from 3. to 3.5 ml/min/1.73 m 2. Using similar analysis, Figure 4B shows that GFR estimated with Cockcroft Gault formula was correct among the young, but the age correction was too steep. For the 5th percentile, bias varied from at age to 21 ml/min/1.73 m 2 at age 8. For the 2.5th percentile, bias varied from þ5 to 13 ml/min/1.73 m 2. Corresponding results using other well-known formulae for estimating GFR are given in Table 3. The original four-variable MDRD formula with indirectly calibrated GFR values had a small and constant negative bias ( 4.5 ml/min/1.73 m 2 for the 5th percentile). The Gates formula had a large negative bias at all GFR levels and all ages. The remaining formulae had an age correction that was too steep, so that most overestimated GFR in the young and underestimated GFR in the elderly. We also report the results if unadjusted serum creatinine values were used with the original four-variable MDRD formula. This gave a large bias of aproximately ml/min/1.73 m 2 at all ages.

Estimating GFR in the general population 1529 Table 2. Characteristics of the participants in HUNT II and the selected subgroup of low risk participants eligible for kidney donation HUNT II (n ¼ 65 186) Low risk subgroup (n ¼ 129) Female gender (%) 53.2 58.6 Age (years) 5.3 (17.4) 41.3 (12.7) Higher education (%) 29.1.9 Never smoked (%) 44.7 46.6 Moderate or high physical activity (%) 26.7 35.7 Systolic blood pressure (mmhg) 138.1 (21.9) 123.1 (1.2) Diastolic blood pressure (mmhg) 8.4 (12.3) 74.1 (7.9) Total cholesterol (mmol/l) 5.9 (1.3) 5.5 (1.1) HDL-cholesterol (mmol/l) 1.4 (.4) 1.4 (.4) Glucose (mmol/l) 5.5 (1.6) 5.1 (.8) Weight (kg) 76.4 (14.) 74.1 (12.4) BMI (kg/m 2 ) 26.4 (4.1) 25.1 (3.1) Waist circumference (cm) 86.5 (11.8) 82.2 (1.1) Estimated GFR (ml/min/1.73 m 2 ) 94.3 (21.9) 95.7 (18.7) Results are expressed as percentage or mean (SD). To convert the conventional units to SI units, multiply cholesterol with.2586 and glucose with.5551. GFR was estimated with the new MDRD formula. 18 16 Estimated GFR (ml/min/1.73m 2 ) 1 1 8 6 6 8 Age (years) Fig. 3. The GFR estimated with the new re-expressed MDRD formula with IDMS traceable serum creatinine values in 129 subjects eligible for kidney donation from the HUNT II study (2.5th, 5th, 97.5th percentiles). Bias for the 2.5th percentile ranged from 2 to 17 ml/min/1.73 m 2 at age 8 using various formulae. As a consequence, many subjects would have been classified as having an abnormally low GFR for their age, i.e. below the 2.5th percentile for reference methods, when applying these formulae on the general population. Figure 5 shows that the new re-expressed MDRD formula estimated 7.1% of the total adult population to have an abnormally low GFR, and the Cockcroft Gault formula estimated 18.6%. Among subjects over age 6, the corresponding results were 13.3 and 48.8%. The other formulae estimated that 3 8% of the latter group have low GFR. The correct prevalence of low GFR is not known, but the new MDRD formula clearly gave the most conservative estimate. The prevalence of CKD stage 3 5 (GFR <6 ml/min/ 1.73 m 2 ) was also calculated using various formulae: 4.7% when using the new MDRD formula (364 cases among 65 193 HUNT II participants), 14.6% with Cockcroft Gault, 25.1% with Jelliffe, and 1 15% with the other formulae. Discussion We compared the estimated GFR in healthy low-risk subjects eligible for kidney donation from the HUNT II study to the measured GFR in potential kidney donors pooled from the literature. The new re-expressed MDRD formula gave unbiased estimates of normal

153 S. Hallan et al. A 18 16 1 GFR (ml/min/1.73m 2 ) 1 8 6 6 8 Age (years) Estimated with MDRD (potential kidney donors in HUNT II) Measured with reference method (potential kidney donors from literature) B 18 16 1 GFR (ml/min/1.73m 2 ) 1 8 6 6 8 Age (years) Estimated with Cockcroft Gault (potential kidney donors in HUNT II) Measured with reference method (potential kidney donors from literature) Fig. 4. Comparing 2.5th, 5th, and 97.5th percentiles for normal GFR measured with reference methods to GFR estimated with (A) new re-expressed MDRD formula and (B) Cockcroft Gault formula. GFR (þ.5 ml/min/1.73 m 2 at age and 2 ml/min/ 1.73 m 2 at age 8). Other formulae tested were much more biased, and all had a large negative bias among the elderly. Median GFR estimated with the Cockcroft Gault formula was unbiased among subjects years old, but the bias increased to 21 ml/min/ 1.73 m 2 at age 8. We found that 49% of the general population above the age of 6 would have an estimated

Estimating GFR in the general population 1531 Table 3. Bias for 2.5th, 5th and 97.5th percentile when estimating GFR with various formulae in subjects aged and 8 years Age Age 8 2.5th 5th 97.5th 2.5th 5th 97.5th New MDRD 3..5 6.8 3.5 2. 1.5 Original MDRD 3.8 4.7 5.8 3.4 4.3 5.3 Original MDRD a 14.5 23.9 38. 3.4 15.7 25.3 Cockcroft and Gault [21] 5.3. 9.4 13. 21.4 33.9 Jelliffe and Jelliffe [22] 1.1 11.2 27.3 16.9 27.9 44.1 Salazar and Corcoran [23] 8.5 13. 19.6 12.4 16.3 21.3 Bjornsson [25] 4.8 11.8 23.7 11.4 12.8 13.5 Gates [26] 19.8 16.7 6.9 12.4 9.4 1.4 Mawer et al. [24] 13.9 7.5 4. 8.6 17.6 31.3 Wright et al. [27] 7.1 1.8 16. 5.5 14.2 27.6 Bias is calculated as the difference between estimated GFR and measured GFR (ml/min/1.73 m 2 ). Negative numbers represent an underestimation of GFR. Positive numbers represent an overestimation of GFR. a Using scr values not calibrated to the same level as in the MDRD study. Percentage with GFR below 2.5th percentile 8 6 General population > years General population >6 years MDRD Cockcroft Jelliffe Salazar Bjornsson Gates Mawer Wright Fig. 5. Percentage of the general population found to have an estimated GFR below their age-specific 2.5th percentile of normal kidney function. GFR below their age-specific 2.5th percentile using the Cockcroft Gault formula, while the new MDRD formula gave a much more conservative estimate of 13.3%. Several large studies have recently concluded that using the MDRD formula in healthy subjects is problematic. Rule et al. [7] found that the formula underestimated GFR by 29 ml/min/1.73 m 2 in kidney donors, and Poggio et al. [8] found a bias of 9 ml/min/ 1.73 m 2. Both were aware of the serum creatinine calibration problems but did not solve the problem sufficiently. Rule et al. mentioned in the discussion that a recalibration similar to Coresh et al. [5] would have reduced the bias to 4 ml/min/1.73 m 2. Poggio restricted his study to the years after 1996 as the bias in the preceding years was even larger ( to 3 ml/min/ 1.73 m 2 ). Froissart et al. [19] solved the problem elegantly by ultrafiltrating plasma through a kda cut-off membrane before measurement and thereby, eliminated the problem of non-creatinine chromogens. In their study of 1 44 persons with GFR above 6 ml/min/1.73 m 2, they found a bias of only 3.3 ml/min/1.73 m 2. The relationship between serum creatinine, muscle mass and age is problematic and has enforced the need for formulae to estimate GFR. However, the accuracy of different formulae in the young vs the old have seldom been reported. Froissart et al. [19] found that the Cockcroft Gault formula changed from a slight positive bias at age under 65 (þ3 ml/min/1.73 m 2 )toa significant negative bias at age over 65 ( 13 ml/min/ 1.73 m 2 ). The MDRD formula had a bias that was similar among younger and older subjects. Cirillo et al. [] also found a linear trend for relative error in

1532 S. Hallan et al. Cockcroft Gault estimates stratified by age (þ5% at ages 18 24, 36% at ages 65 88). The MDRD formula had, without recalibration of serum creatinine values, a constant relative error of 11% in their study. This is well in accordance with our results, and most other formulae tested in our study also showed a bias that changed with age. When the problems with interlaboratory calibration differences for serum creatinine are properly taken care of, the MDRD formula seems to be nearly unbiased and without the problem of underestimating GFR among the elderly hampering all other formulae. The re-expressed four-variable MDRD equation, which was published very recently, has coefficients appropriate for use with a zero-biased creatinine method [9]. This will facilitate correct calibration of serum creatinine values and thereby, bring implementation of formula based GFR estimation several steps forward. However, the accuracy of the MDRD is only moderate among subjects with normal or near normal GFR. Errors of ±15 3% are not uncommon [6,8,19], so direct measurement of GFR with a reference technique should be considered if high accuracy is needed. The large underestimation of GFR in older subjects using the Cockcroft Gault formula is now increasingly documented [19,]. As CKD is most prevalent among the elderly, using this formula will introduce large errors in epidemiological research, and the underestimation will also be of clinical importance for many patients. The recommendation of the Cockcroft Gault formula found in most international CKD guidelines should therefore be reconsidered. The weakness of the current study is the indirect comparison of estimated and measured GFR. However, the optimal alternative, which means measuring GFR with reference technique in a large number of subjects from the general population, is unlikely to be done. The strength or our study is the large number of subjects with an age ranging from to 83 years. Due to the very comprehensive HUNT II database, we were able to ensure that all of them really were eligible for kidney donation according to internationally accepted criteria. Our indirect calibration of serum creatinine for the original four-variable MDRD formula has been well described and has previously been shown to give unbiased estimates when using the MDRD formula [6]. However, correct calibration of the creatinine values is now more straightforward with the new MDRD formula. Our gold standard, i.e. the reference limits for normal GFR by age, should also be optimal as we have pooled individual results from available studies. We therefore, feel that our results have a solid methodological basis and can be generalized. In summary, when using IDMS traceable serum creatinine values and the re-expressed MDRD formula, GFR estimates from the general population were nearly unbiased at all ages (þ.5 to 2. ml/min/ 1.73 m 2 ). The Cockcroft Gault formula had a large negative bias among the elderly ( 21 ml/min/1.73 m 2 ), and similar results were found in all other formulae tested ( 1 to 3 ml/min/1.73 m 2 ). Cockcroft Gault classified 49% of the general population above age 6 as having an estimated GFR below their age-specific 2.5th percentile for normal kidney function as compared to 13.3% with the new MDRD formula. We therefore conclude that the MDRD formula was the best formula available for estimating GFR in the general population, and that the recommendation of the Cockcroft Gault formula in guidelines should be reconsidered. Acknowledgements. The HUNT Study is a collaboration between the HUNT Research Center, Faculty of Medicine, Norwegian University of Science and Technology, Verdal; The Norwegian Institute of Public Health, Oslo; Nord-Trondelag County Council and Central Norway Regional Health Authority. The authors thank the health service and the people of Nord-Trondelag for their endurance and participation. Conflict of interest statement. None of the authors have a conflict of interest according to the COPE guidelines. References 1. K/DOQI clinical practice guidelines for chronic kidney disease: evaluation, classification, and stratification. Kidney Disease Outcome Quality Initiative. Am J Kidney Dis 2; 39: S1 S246 2. European Best Practice guidelines: Measurement of renal function, when to refer and when to start dialysis. Nephrol Dial Transplant 2; 17 [Suppl 7]: 7 15 3. Levey AS, Bosch JP, Lewis JB et al. A more accurate method to estimate glomerular filtration rate from serum creatinine: a new prediction equation. Modification of Diet in Renal Disease Study Group. Ann Intern Med 1999; 13: 461 47 4. Stevens LA, Levey AS. Clinical implications of estimating equations for glomerular filtration rate. Ann Intern Med 4; 141: 959 961 5. 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 2; 39: 9 929 6. Hallan S, Aasberg A, Lindberg M, Johnsen H. Validation of the Modification of Diet in Renal Disease formula for estimating GFR with special emphasis on calibration of the serum creatinine assay. Am J Kidney Dis 4; 44: 85 93 7. Rule AD, Larson TS, Bergstralh EJ et al. Using serum creatinine to estimate glomerular filtration rate: accuracy in good health and in chronic kidney disease. Ann Intern Med 4; 141: 929 937 8. Poggio ED, Wang X, Greene T, Van Lente F, Hall PM. Performance of the Modification of Diet in Renal Disease and Cockcroft-Gault Equations in the Estimation of GFR in health and in chronic kidney disease. J Am Soc Nephrol 5; 16: 459 466 9. Levey AS, Coresh J, Greene T et al. Expressing the MDRD study equation for estimating GFR with IDMS traceable (Gold Standard) serum creatinine values (Abstract). J Am Soc Nephrol 5; 11: 16:69A 1. Statistics Norway. (Accessed May 4 at http:/www.ssb.no/ english/subjects/). 3 11. Holmen J, Midthjell K, Kruger O et al. The Nord-Trondelag Health Study 1995 97 (HUNT 2): objectives, contents, methods and participation. Norsk Epidemiologi 3; 13: 19 32 12. Levey AS, Greene T, Kusek J, Beck GJ, Group MS. A simplified equation to predict glomerular filtration rate

Estimating GFR in the general population 1533 from serum creatinine (Abstract). J Am Soc Nephrol ; 11: A828 13. Rustad P. Reference intervals for 25 of the most frequently used proerties in clincial chemistry; proposal by Nordic Reference Interval Project (NORIP). Klinisk Biokemi i Norden 3; 15: 1 17 14. Delmonico F. A report of the Amsterdam Forum on the care of the live kidney donor: data and medical guidelines. Transplantation 5; 79: S53 S66 15. Grewal GS, Blake GM. Reference data for 51Cr-EDTA measurements of the glomerular filtration rate derived from live kidney donors. Nucl Med Commun 5; 26: 61 65 16. Rule AD, Gussak HM, Pond GR et al. Measured and estimated GFR in healthy potential kidney donors. Am J Kidney Dis 4; 43: 112 119 17. Hamilton D, Riley P, Miola U et al. Total plasma clearance of 51Cr-EDTA: variation with age and sex in normal adults. Nucl Med Commun ; 21: 187 192 18. Davies DF, Shock NW. Age changes in glomerular filtration rate, effective renal plasma flow, and tubular excretory capacity in adult males. J Clin Invest 195; 29: 496 57 19. Froissart M, Rossert J, Jacquot C, Paillard M, Houillier P. Predictive performance of the Modification of Diet in Renal Disease and Cockcroft-Gault equations for estimating renal function. J Am Soc Nephrol 5; 16: 763 773. Cirillo M, Anastasio P, De Santo NG. Relationship of gender, age, and body mass index to errors in predicted kidney function. Nephrol Dial Transplant 5; : 1791 1798 21. Cockcroft DW, Gault MH. Prediction of creatinine clearance from serum creatinine. Nephron 1976; 16: 31 41 22. Jelliffe RW, Jelliffe SM. A computer program for estimation of creatinine clearance from unstable serum creatinine levels, age, sex and weight. Math Biosci 1972; 14: 17 24 23. Salazar DE, Corcoran GB. Predicting creatinine clearance and renal drug clearance in obese patients from estimated fat-free body mass. Am J Med 1988; 84: 153 16 24. Mawer GE, Lucas SB, Knowles BR, Stirland RM. Computerassisted prescribing of kanamycin for patients with renal insufficiency. Lancet 1972; 1: 12 15 25. Bjornsson TD. Use of serum creatinine concentrations to determine renal function. Clin Pharmacokinet 1979; 4: 222 26. Gates GF. Creatinine clearance estimation from serum creatinine values: an analysis of three mathematical models of glomerular function. Am J Kidney Dis 1985; 5: 199 5 27. Wright JG, Boddy AV, Highley M et al. Estimation of glomerular filtration rate in cancer patients. Br J Cancer 1; 84: 452 459 Received for publication: 7.12.5 Accepted in revised form: 23.1.6