Cystatin C is not a better estimator of GFR than plasma creatinine in the general population

http://www.kidney-international.org & 2010 International Society of Nephrology Cystatin C is not a better estimator of GFR than plasma creatinine in the general population Bjørn O. Eriksen 1,2, Ulla D. Mathisen 1,2, Toralf Melsom 1,2, Ole C. Ingebretsen 3,4, Trond G. Jenssen 2,5, Inger Njølstad 6, Marit D. Solbu 1 and Ingrid Toft 1,2 1 Clinic of Internal Medicine, Section of Nephrology, University Hospital of North Norway, Tromsø, Norway; 2 Department of Clinical Medicine, University of Tromsø, Tromsø, Norway; 3 Department of Medical Biochemistry, University Hospital of North Norway, Tromsø, Norway; 4 Department of Medical Biology, University of Tromsø, Tromsø, Norway; 5 Department of Internal Medicine, Section of Medicine, Oslo University Hospital, Rikshospitalet, Oslo, Norway and 6 Department of Community Medicine, University of Tromsø, Tromsø, Norway Accurate measurement of glomerular filtration rate (GFR) is complicated and costly; therefore, GFR is commonly estimated by assessing creatinine or cystatin C concentrations. Because estimates based on cystatin C predict cardiovascular disease better than creatinine, these estimates have been hypothesized to be superior to those based on creatinine, when the GFR is near the normal range. To test this, we measured GFR by iohexol clearance in a representative sample of middle-aged (50 62 years) individuals in the general population, excluding those with coronary heart or kidney disease, stroke or diabetes mellitus. Bias, precision (median and interquartile range of estimated minus measured GFR (mgfr)), and accuracy (percentage of estimates within 30% of mgfr) of published cystatin C and creatinine-based GFR equations were compared in a total of 1621 patients. The cystatin C-based equation with the highest accuracy (94%) had a bias of 3.5 and precision of 18 ml/min per 1.73 m 2, whereas the most accurate (95%) creatinine-based equation had a bias of 2.9 and precision of 15 ml/min per 1.73 m 2. The best equation, based on both cystatin C and creatinine, had a bias of 7.6 ml/min per 1.73 m 2, precision of 15 ml/min per 1.73 m 2, and accuracy of 92%. Thus, estimates of GFR based on cystatin C were not superior to those based on creatinine in the general population. Hence, the better prediction of cardiovascular disease by cystatin C than creatinine measurements, found by others, may be due to factors other than GFR. Kidney International (2010) 78, 1305 1311; doi:10.1038/ki.2010.321; published online 15 September 2010 KEYWORDS: cardiovascular disease; clinical nephrology; Cockcroft Gault; creatinine; glomerular filtration rate Correspondence: Bjørn O. Eriksen, Section of Nephrology, University Hospital of North Norway, Tromsø 9038, Norway. E-mail: bjorn.odvar.eriksen@unn.no Received 13 February 2010; revised 30 May 2010; accepted 13 July 2010; published online 15 September 2010 Chronic kidney disease has been established as a risk factor for cardiovascular disease. 1 Reduction in the glomerular filtration rate (GFR) near the normal range is also being studied as a potential risk factor of cardiovascular disease in the general population. Because measuring GFR accurately in surveys is complicated and costly, GFR has been estimated from creatinine or cystatin C. In community-based and patient studies, cystatin C-based estimates have been superior to creatinine in predicting cardiovascular disease. 2 4 The putative explanation is that cystatin C may be more sensitive to small reductions in GFR and better for estimating GFR near the normal range than creatinine. 5 Creatinine-based GFR estimates have been found to be biased and imprecise for GFR values in the normal range. 6 11 One reason for this bias is that, in contrast to cystatin C, the production rate of creatinine depends on muscle mass, which varies significantly in both health and disease. Also, the creatinine-based equations that are currently used were developed from creatinine measurements using Jaffe s method, which is unspecific and has been replaced by improved enzymatic methods in many laboratories. 12 However, cystatin C may be a less ideal filtration marker than initially thought. Some studies have reported that cystatin C is influenced by confounders not related to GFR, such as age, gender, C-reactive protein, and albumin. 13,14 Thus, a superior ability to predict cardiovascular disease does not necessarily imply that cystatin C produces better GFR estimates. In fact, cystatin C- and creatinine-based GFR estimates have never been compared with precise GFR measurements in the general population. Similarly, the question of whether cystatin C offers better assessments of GFR than creatinine in this setting is not yet answered. 15 In the Renal Iohexol Clearance Survey in Tromsø 6 (RENIS- T6), we studied a representative sample of middle-aged persons from the general population using measurements of GFR obtained by iohexol clearance, plasma creatinine measured using an enzymatic method, and cystatin C. Our aim was to compare published cystatin C-based equations with the most commonly used creatinine-based equations and to validate both against iohexol clearance. Kidney International (2010) 78, 1305 1311 1305

BO Eriksen et al.: Estimating GFR with cystatin C in the general population RESULTS Of the 3564 individuals aged 50 62 years who completed the main part of the sixth Tromsø population survey (Tromsø 6), 739 reported a previous myocardial infarction, angina pectoris, stroke, diabetes mellitus, or renal disease. The remaining 2825 eligible subjects were invited to participate in RENIS-T6. Of the 2107 individuals who responded positively, 12 were excluded owing to an allergy to contrast media, iodine, or latex; 65 owing to other reasons; and 48 did not appear at their appointments. A total of 1982 subjects remained for potential inclusion, of which 1632 were investigated according to a predefined target. Five participants were excluded because their iohexol clearance measurements were technical failures. Accordingly, the RENIS-T6 cohort consists of 1627 subjects. None of the subjects had measured GFR (mgfr) lower than 15 ml/min per 1.73 m 2 (0%, 95% confidence interval 0 0.21%). Two persons (0.12%, 95% confidence interval 0.02 0.43%) had mgfr in the interval 15 29 ml/min per 1.73 m 2, and 32 persons (2.0%, 95% confidence interval 1.4 2.7%) in the interval 30 59 ml/min per 1.73 m 2. For the present analysis, six subjects were excluded because they used prednisolone. The characteristics of the 1621 subjects included in the study are compared with the total group of eligible subjects (n ¼ 2825) in Table 1. There were statistically significant, though small, differences in age and body mass index (Po0.01). Mean (s.d.) mgfr for women was 87.8 (14) and for men 95.7 (13.7) ml/min per 1.73 m 2. The bias, precision, and accuracy of the investigated equations in Table 2 are shown in Table 3. For GFR estimates based on cystatin C (estimated GFR (egfr cys )), percentage bias varied between 1 and 37, and percentage precision between 21 and 37. Calibrated cystatin C improved bias and precision for Grubb s equation, which is based on the Dako particle-enhanced turbidimetric immunoassay (Dakocytomation, Glostrup, Denmark), but had the opposite effect on those based on the Dade-Behring particle-enhanced nephelometric immunoassay (Siemens, Marburg, Germany). Of the egfr cys equations, Rule s and Stevens A equations with uncalibrated cystatin C had percentage precision and P 30 Table 1 Study population characteristics compared with all eligible subjects Included (n=1621) All eligible persons (n=2825) a Female (%) 50.7 50.7 Mean age (s.d.), years 56.9 b (3.9) 56.8 (3.9) Mean weight (s.d.), kg 78.7 (14.4) 78.3 (14.5) Mean height (s.d.), cm 170.6 (8.7) 170.7 (9.0) Mean body mass index (s.d.), kg/m 2 26.9 b (4.0) 26.8 (4.0) Mean plasma creatinine (s.d.), mmol/l 68.5 (12.8) 68.8 (12.5) Mean plasma cystatin C (s.d.), mg/l 0.78 (0.12) 0.78 (0.11) Mean measured GFR (s.d.), ml/min per 1.73 m 2 91.7 (14.4) a Means and s.d. weighted according to the age stratification and gender stratification of RENIS-T6. b Po0.01 for difference between included and eligible. comparable to GFR estimated from creatinine (egfr cre ). Both the Modification of Diet in Renal Disease (MDRD) and Chronic Kidney Disease Epidemiology Collaboration (CKD- EPI) equations were nearly unbiased with percentage biases of 1 and 3, respectively. The Cockcroft Gault (CG) and CKD- EPI equations were significantly more precise than all egfr cys equations. Accuracy, expressed as P 30, was highest for the CKD-EPI equation (95%). The two equations based on both cystatin C and creatinine (egfr comb ) performed at a level between egfr cys and egfr cre (Table 3). Calibration markedly worsened the performance of Stevens B equation. Figure 1 shows the relationships between mgfr and egfr for the egfr cre, egfr cys, and egfr comb equations with the highest P 30. Local regression of mgfr on egfr (dashed gray lines) demonstrates that Stevens A egfr cys equation tended to overestimate mgfr for higher egfrs (Figure 1a). The same was true for Stevens B egfr comb equation (Figure 1c). The CKD-EPI egfr cre equation was essentially unbiased across the range of egfr (Figure 1b). Because the CKD-EPI equation has different parameters for low and high creatinine values, the shape of the distribution of observation in Figure 1b differed from the cigar-shaped distributions in Figure 1a and c. The correlation coefficients between egfr cys and mgfr were between 0.50 and 0.58. The coefficients for the MDRD and CKD-EPI equations were 0.56 and 0.55, respectively. The correlation coefficient for Stevens combined equations reached 0.65, which was significantly higher than all of the other equations. All the analyses in Table 2 were repeated for each gender separately (not shown). We found no evidence of superior performance of any of the cystatin C-based equations for either men or women. Except for the CG and Rule s equation with uncalibrated cystatin C, all of the equations had o0.50 sensitivity and 40.85 specificity for discrimination at GFR 60 and 80 ml/ min per 1.73 m 2 (Table 4). The corresponding areas under the receiver operating characteristic curves were similar for both egfr cys and egfr cre (0.73 0.81; Table 4), but were slightly higher for the combined equations. DISCUSSION To the best of our knowledge, this is the first study in which equation-based GFR estimates are compared with mgfr in a population survey. A recent study demonstrated that singlesample iohexol clearance is as reliable as the multi-sample method when validated against Cr-EDTA clearance. 16 We found no evidence that equations based on cystatin C alone or in combination with creatinine provide better GFR estimates in middle-aged members of the general population than the commonly used MDRD and CKD-EPI equations. Two reviews on GFR estimates among persons with reduced GFR have pointed out that some studies show a better performance of cystatin C than creatinine, whereas others have found the two markers to be equivalent. 17,18 Louvar et al. 19 found the MDRD equation to be better than egfr cys in a follow-up of former kidney donors. 1306 Kidney International (2010) 78, 1305 1311

BO Eriksen et al.: Estimating GFR with cystatin C in the general population Table 2 Investigated cystatin C-based equations Equation Author ref. Assay Manufacturer (80.35/cystatin C) 4.32 Hoek, 2003 37 PENIA Dade-Behring/Siemens (100/cystatin C) 14 Tidman, 2008 38 A PETIA Gentian 79.901 cystatin C 1.4389 Flodin, 2007 39 PETIA Gentian 77.24 cystatin C 1.2623 Larsson, 2004 40 PENIA Dade-Behring/Siemens 66.8 cystatin C 1.30 Rule, 2006 41 PENIA Dade-Behring/Siemens 86.49 cystatin C 1.686 (0.948 if female) Grubb, 2005 42 PETIA Dako 127.7 cystatin C 1.17 age 0.13 (0.91 if female; 1.06 if black) Stevens, 2008 43 A PENIA Dade-Behring/Siemens 177.6 creatinine 0.65 cystatin C 0.57 age 0.20 (0.82 if female; 1.11 if black) Stevens, 2008 43 B PENIA Dade-Behring/Siemens Mean of MDRD equation and (100/cystatin C) 14 Tidman, 2008 38 B PETIA Gentian Abbreviations: MDRD, modification of diet in renal disease; PENIA, particle-enhanced nephelometric immunoassay; PETIA, particle-enhanced turbidimetric immunoassay. Table 3 Comparison of GFR-estimating equations based on cystatin C, creatinine, and both Difference from mgfr a, ml/min per 1.73 m 2 Median Interquartile range Median Percent difference from mgfr b Interquartile range P 30 c,% Correlation with mgfr, coefficient Cystatin C-based equations (uncalibrated cystatin C) Flodin, 2007 39 21.0 (19.8 to 22.3) 26.3 (24.8 27.8) 23 (22 to 25) 30 (28 31) 62 (60 64) 0.50 (0.47 0.54) Grubb, 2005 42 34.4 (32.8 to 36.1) 32.7 (30.9 34.6) 37 (35 to 39) 37 (35 39) 39 (36 41) 0.53 (0.50 0.57) Hoek, 2003 37 5.9 (4.9 to 6.9) 18.6 (17.4 19.8) 7 (6 to 8) 21 (19 22) 91 (89 92) 0.51 (0.47 0.55) Larsson, 2004 40 12.8 (11.7 to 13.7) 21.9 (20.6 23.1) 14 (13 to 15) 25 (24 27) 78 (76 80) 0.51 (0.47 0.54) Tidman, 2008 38 A 21.1 (20.0 to 21.9) 21.4 (20.2 22.6) 23 (22 to 24) 25 (24 27) 63 (60 65) 0.51 (0.47 0.55) Rule, 2006 41 0.7 ( 1.4 to 0.2) 20.2 (19.1 21.4) 1 ( 1 to 0) 23 (21 24) 93 (91 94) 0.51 (0.47 0.54) Stevens, 2008 43 A 3.5 (2.7 to 4.6) 18.4 (17.3 19.6) 4 (3 to 5) 21 (20 22) 94 (92 95) 0.58 (0.55 0.61) Cystatin C-based equations (calibrated cystatin C) Grubb, 2005 42 16.5 (15.4 to 17.7) 23.3 (22.1 24.6) 18 (17 to 20) 27 (25 28) 72 (69 74) 0.55 (0.51 0.58) Hoek, 2003 37 20.7 (19.5 to 21.7) 22.4 (21.1 23.8) 22 (21 to 24) 27 (25 28) 64 (62 67) 0.51 (0.47 0.54) Larsson, 2004 40 31.8 (30.5 to 32.9) 29.4 (27.6 31.2) 35 (33 to 37) 33 (31 35) 42 (40 45) 0.50 (0.46 0.54) Rule, 2006 41 16.7 (15.6 to 18.1) 26.6 (25.1 28.3) 19 (17 to 20) 30 (28 32) 68 (66 71) 0.50 (0.46 0.54) Stevens, 2008 43 A 19.2 (18.2 to 20.4) 23.9 (22.7 25.3) 21 (20 to 23) 27 (26 29) 66 (64 69) 0.57 (0.53 0.60) Cystatin C- and creatinine-based equations (uncalibrated cystatin C) Stevens, 2008 43 B 7.6 (7.1 to 8.4) 14.9 (14.1 15.9) 8 (8 to 9) 17 (16 18) 92 (90 93) 0.65 (0.62 0.68) Tidman, 2008 38 B 11.8 (11.2 to 12.5) 15.7 (14.8 16.6) 13 (12 to 14) 18 (17 19) 87 (85 89) 0.62 (0.59 0.66) Cystatin C- and creatinine-based equation (calibrated cystatin C) Stevens, 2008 43 B 15.6 (14.5 to 16.4) 16.8 (15.9 17.9) 17 (16 to 18) 19 (18 20) 81 (79 82) 0.65 (0.62 0.68) Creatinine-based equations Cockcroft Gault 14.5 ( 15.4 to 13.9) 16.3 (15.2 17.3) 16 ( 17 to 15) 16 (15 17) 91 (90 92) 0.48 (0.45 0.52) MDRD 1.3 (0.4 to 2.1) 18.2 (17.0 19.5) 1 (0.4 to 2.3) 20 (19 22) 93 (91 94) 0.56 (0.52 0.60) CKD-EPI 2.9 (2.2 to 3.5) 15.4 (14.5 16.3) 3 (2.3 to 3.9) 17 (16 19) 95 (94 96) 0.55 (0.52 0.59) Abbreviations: CI, confidence interval; CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration; egfr, estimated glomerular filtration rate; MDRD, modification of diet in renal disease; mgfr, measured GFR. a egfr mgfr b 100 (egfr mgfr)/mgfr. c P 30 is defined as the proportion of egfr within 30% of mgfr. The performance of the MDRD and CKD-EPI equations was good. Bias and accuracy were better for the MDRD equation than that reported in two other external validations. 7,20 The bias was substantially smaller than in several investigations of potential kidney donors 9 11 and chronic kidney disease patients with near normal GFR, 6,8 but similar to the investigation of former kidney donors by Sebasky et al. 21 and potential kidney donors by Froissart et al. 22 The two egfr comb equations had bias and accuracy values midway between those of the equations based on either marker alone, and a precision similar to egfr cre (Table 3). Several factors may have contributed to the large bias of most of the egfr cys equations. The most important factor was probably that they were all developed in populations with chronic kidney disease and low GFR. The influence of non-gfr factors on plasma cystatin C may differ between Kidney International (2010) 78, 1305 1311 1307

BO Eriksen et al.: Estimating GFR with cystatin C in the general population Measured GFR, ml/min per 1.73 m 2 160 120 100 80 60 40 20 Uncalibrated cystatin C (Stevens A) Creatinine (CKD-EPI) 20 60 100 140 20 60 100 140 Estimated GFR, ml/min per 1.73 m 2 Measured GFR, ml/min 1.73 m 2 160 120 100 80 60 40 20 Estimated GFR, ml/min per 1.73 m 2 Measured GFR, ml/min per 1.73 m 2 160 120 100 80 60 40 20 Uncalibrated cystatin C and creatinine (Stevens B) 20 60 100 140 Estimated GFR, ml/min per 1.73 m 2 Figure 1 Comparison of measured and estimated GFR. Comparison of measured and estimated GFR (egfr) for Stevens A (a), CKD-EPI (b), and Stevens B equations (c) on the log scale. The straight line represents identity, and the gray area indicates where the egfr is within 30% of the measured GFR. Local regression of measured on estimated GFR by the loess method in SAS is indicated by the gray dashed lines. Abbreviations: CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration; GFR, glomerular filtration rate. these patients and the general population. 13,14 Also, standardization between assays and laboratories, where substantial progress has been made for creatinine, is lacking for cystatin C. 23 We used published results to calibrate our measurements of cystatin C to the other methods, but this resulted in a better performance of the Dako-based Grubb equation and worse results for equations developed with Dade-Behring s assay. The lack of standardization is a serious obstacle for the routine use of egfr cys. The sensitivity of any of the equations for detecting an mgfr of o60 ml/min per 1.73 m 2 was p30% (Table 4). The confidence intervals of the sensitivities were wide because there were few persons in this group, but their upper limits were also well below 50%. Specificities at 100% indicate poor calibration, but areas under the receiver operating characteristic curve of 0.72 0.80 demonstrate that none of the equations discriminated well at this cutoff point. The results for detecting a GFR of o80 ml/min per 1.73 m 2 were also poor. However, as most of the low mgfrs to be detected were clustered in the range immediately below each cutoff, any equation would have difficulty in obtaining a high sensitivity because anything short of nearly perfect precision would give many false negatives by chance. Better results have been found in populations with a higher proportion of very low GFR values. 20 This is an example of the so-called spectrum bias. 24 Our results indicate that screening for low GFR with creatinine or cystatin C in a low-risk general population is probably not worthwhile. Although they did not perform better than egfr cre, the combined equations had a significantly higher correlation with mgfr than both egfr cre and egfr cys, which would, in principle, make, for example, Stevens B equation a better proxy variable for mgfr when predicting cardiovascular disease. However, a recent population-based study found that there was a weaker association between overall and cardiovascular mortality and Stevens B (egfr comb ) than Stevens A (egfr cys ). 3 Overestimation of GFR in persons with chronic disease and low muscle mass owing to the creatinine term in the equation was thought to be the reason. Our results suggest that the dependence of cystatin C on non- GFR factors should be considered as an alternative explanation. 13,14 One limitation of the present study was that it was restricted to persons between 50 and 62 years of age. However, studies with a wider age range have not reported interactions between age and cystatin C or creatinine that would lead us to expect better results for egfr cys in other age groups. Because our study population included Caucasians only, we were not able to study the effect of ethnicity. This study did not include persons with self-reported cardiovascular disease or diabetes mellitus. It is possible that cystatin C may perform better than creatinine in these individuals, but there is also the risk that the performance may be confounded by factors influencing both cardiovascular risk and the production rate of cystatin C or creatinine. 13,14 Estimating the performance of cystatin C- and 1308 Kidney International (2010) 78, 1305 1311

BO Eriksen et al.: Estimating GFR with cystatin C in the general population Table 4 Diagnosis of low GFR by cystatin C- and creatinine-based equations for estimating GFR Sensitivity, % GFR o60 ml/min per 1.73 m 2 GFRo80 ml/min per 1.73 m 2 Specificity, % Area under ROC-curve Sensitivity, % Specificity, % Area under ROC-curve Cystatin C-based equations (uncalibrated cystatin C) Flodin, 2007 39 21 (8 35) 100 (100 100) 0.769 (0.679 0.850) 20 (15 25) 98 (97 99) 0.768 (0.738 0.796) Grubb, 2005 42 21 (8 35) 100 (100 100) 0.785 (0.698 0.864) 15 (11 19) 99 (98 99) 0.784 (0.756 0.811) Hoek, 2003 37 21 (8 35) 100 (100 100) 0.770 (0.679 0.850) 26 (21 31) 96 (95 97) 0.768 (0.738 0.796) Larsson, 2004 40 21 (8 35) 100 (100 100) 0.770 (0.679 0.851) 22 (18 27) 97 (96 98) 0.768 (0.738 0.796) Tidman, 2008 38 A 9 (3 20) 100 (100 100) 0.770 (0.679 0.850) 10 (7 14) 99 (99 100) 0.768 (0.738 0.796) Rule, 2006 41 30 (15 46) 99 (98 99) 0.770 (0.679 0.851) 51 (45 57) 83 (81 85) 0.768 (0.738 0.796) Stevens, 2008 43 A 27 (13 42) 99 (99 100) 0.809 (0.727 0.883) 43 (38 49) 91 (89 92) 0.806 (0.779 0.832) Cystatin C-based equations (calibrated cystatin C) Grubb, 2005 42 21 (8 35) 100 (99 100) 0.785 (0.698 0.864) 24 (19 29) 98 (97 98) 0.784 (0.756 0.811) Hoek, 2003 37 9 (3 20) 100 (100 100) 0.770 (0.679 0.850) 11 (8 15) 99 (99 100) 0.768 (0.738 0.796) Larsson, 2004 40 15 (4 28) 100 (100 100) 0.770 (0.679 0.851) 10 (7 14) 99 (99 100) 0.768 (0.738 0.796) Rule, 2006 41 24 (10 39) 100 (99 100) 0.770 (0.679 0.851) 25 (20 30) 97 (96 98) 0.768 (0.738 0.796) Stevens, 2008 43 A 21 (8 35) 100 (100 100) 0.804 (0.721 0.880) 20 (16 25) 98 (98 99) 0.802 (0.775 0.829) Cystatin C- and creatinine-based equations (uncalibrated cystatin C) Stevens, 2008 43 B 15 (4 28) 100 (100 100) 0.802 (0.709 0.881) 30 (25 35) 97 (96 98) 0.835 (0.810 0.858) Tidman, 2008 38 B 12 (3 24) 100 (100 100) 0.789 (0.697 0.869) 17 (13 21) 99 (99 100) 0.822 (0.796 0.847) Cystatin C- and creatinine-based equation (calibrated cystatin C) Stevens, 2008 43 B 15 (4 28) 100 (100 100) 0.802 (0.709 0.881) 16 (12 20) 99 (98 99) 0.835 (0.810 0.858) Creatinine-based equations Cockcroft Gault 24 (10 39) 97 (96 98) 0.728 (0.628 0.816) 87 (83 90) 44 (42 47) 0.741 (0.711 0.770) MDRD 18 (6 32) 100 (99 100) 0.763 (0.666 0.851) 49 (43 55) 87 (86 89) 0.783 (0.755 0.810) CKD-EPI 18 (6 32) 100 (100 100) 0.764 (0.667 0.848) 27 (22 32) 96 (95 97) 0.785 (0.760 0.811) Abbreviations: CI, confidence interval; CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration; GFR, glomerular filtration rate; MDRD, Modification of Diet in Renal Disease; ROC, receiver operating characteristic. creatinine-based equations without this source of confounding was one of the strengths of the present study. We conclude that published equations for estimating GFR from cystatin C alone or combined with creatinine do not estimate GFR better in the general population than the creatinine-based MDRD and CKD-EPI equations. This finding suggests that the superior prediction of cardiovascular disease by cystatin C-based GFR estimates reported by others may be the result of non-gfr determinants of plasma cystatin C concentration. METHODS Subjects The Tromsø Study is a population-based prospective survey in the municipality of Tromsø, North Norway (current population 65,000). Since 1974, the Tromsø Study has carried out six health surveys that invited total birth cohorts and random population samples to participate. The main emphasis of the study is on cardiovascular and other chronic diseases. RENIS-T6 is an ancillary part of Tromsø 6. In the main part of the Tromsø 6 survey, a representative sample of 12,984 adults participated between October 2007 and December 2008. The participation rate was 66%. Those invited to Tromsø 6 included a 40% random sample of individuals aged 50 59 years and all individuals aged 60 62 years (5464 total subjects). Of these subjects, 3564 individuals between 50 and 62 years of age completed the main part of Tromsø 6 (65%). From this group of 3564 subjects, those who did not report a previous myocardial infarction, angina pectoris, stroke, diabetes mellitus, or any renal disease (except urinary tract infection) were invited to participate in RENIS-T6 and were included in the study, stratified according to gender and age in a consecutive order until a predetermined target of 1600 subjects was met. Individuals using anti-diabetic medication, pregnant women, and individuals reporting allergies to contrast media, iodine, or latex were excluded. In the present analysis, individuals taking medication that influences creatinine or cystatin C pharmacokinetics independently of GFR (cimetidine, trimethoprim, pyrimethamine, calcitriol, alphacalcidol, and steroids) were also excluded. 18,25 This study was approved by the Norwegian Data Inspectorate and the Regional Ethics Committee of North Norway. All subjects provided written consent. Measurements Iohexol clearance. mgfr was measured using single-sample plasma clearance of iohexol at the Clinical Research Unit at the University Hospital of North Norway. This method has been validated against gold standard methods for measuring GFR. 16,26 31 The subjects were instructed to avoid large meals with meat and non-steroid anti-inflammatory drugs 2 days before the investigation, which was performed after fasting, including abstinence from nicotine. The investigation was re-scheduled if the patient suffered from any acute illness or had a radiological examination using contrast within a week before. The subjects were reminded to not restrict the intake of water. A Teflon catheter was placed in an antecubital vein, and a null sample and blood samples for creatinine and cystatin C were drawn. A total of 5 ml of iohexol (Omnipaque, 300 mgi/ml; Amersham Health, London, UK) was injected, and the syringe was weighed before and after the injection. The catheter was flushed with 30 ml isotonic saline and used for iohexol analysis Kidney International (2010) 78, 1305 1311 1309

BO Eriksen et al.: Estimating GFR with cystatin C in the general population samples. 30 The subjects were observed for allergic reactions for 30 min and then allowed to walk about freely and eat a light breakfast, but they were not allowed to eat meat or to smoke. The optimal time for measuring iohexol concentration after injection was calculated by Jacobsson s method based on the GFR estimated from creatinine. 32 To ensure complete distribution of iohexol in the extracellular fluid volume, the shortest sampling time was set at 180 min. The exact time from injection to sampling was measured in minutes using a separate stopwatch for each person. Omnipaque from one batch purchased from Amersham was used. The serum iohexol concentration was measured by high performance liquid chromatography as previously described by Nilsson-Ehle. 33 Proteins were precipitated by adding 800 ml 0.33 mol/l perchloric acid to 200 ml serum. After centrifugation, 10 ml supernatant was injected on a 4 200 mm Nucleosil 5 mm 100 C18 column (Hichrom, Theale, Reading, UK). The mobile phase contained 95% (v/v) citric acid/citrate buffer, 20 mol/l, ph 4.5 and 5% acetonitrile and was maintained at a flow of 0.9 ml/min. After elution of both iohexol isomers (10 min), the column was washed with 100% methanol during 1.5 min at a flow of 0.9 ml/min. The concentration of iohexol was calculated from the area under the largest iohexol peak as compared with suitable external standards of iohexol. The coefficient of variation for the analysis during the study period was 3%. The external quality control was provided by Equalis (Equalis AB, Uppsala, Sweden). The GFR was calculated using the formulae described by Jacobsson. 32 The following system of three equations for correction for non-immediate mixing (r), correction for non-uniform distribution (m), and iohexol clearance in ml/min (Cl) was solved with a numeric method from the extracellular volume in ml (V), time in minutes from injection to sampling (t), concentration of iohexol in mg/l (C(t)) at time t, and amount of injected iohexol in mg (Q(0)): p r ¼0:495 1 þ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 0:0049Cl m ¼0:991 0:00122Cl 1 Cl ¼ ðtm=vþ lnðrþ=cl ln Qð0Þ CðtÞV=m Extracellular volume (V) was estimated by Granerus equation. 34 Extra-renal iohexol clearance was ignored in accordance with the practice of other authors. GFR was normalized to 1.73 m 2 body surface area, which was estimated using Dubois equation. 35 Creatinine. Plasma creatinine analyses were performed on the Hitachi Modular model using an enzymatic method that has been standardized against isotope dilution mass spectroscopy (CREA Plus, RocheDiagnostics,GmbH,Mannheim,Germany).Thelaboratory participated in an external quality assessment program provided by Labquality, Helsinki, Finland. The analysis at our laboratory was calibrated with serum X in the Nordic Reference Interval Project, which has been validated against isotope dilution mass spectroscopy. 36 Inter-assay coefficient of variation in the study period was 2.3%. Cystatin C. Cystatin was measured by particle-enhanced turbidimetric immunoassay using reagents from Gentian (Gentian, Moss, Norway) on a Modular E analyzer (Roche Diagnostics). Interassay coefficient of variation in the study period was 3.1%. Estimating equations Equations for egfr cys and egfr comb are presented in Table 2. 37 43 Equations for adults with a kidney transplant or those developed for specific diagnostic groups were excluded. There are three different commercially available cystatin C assays: Dako (Dakocytomation), Dade-Behring (Siemens), and Gentian (Gentian). The Gentian and Dako assays are particle-enhanced turbidimetric immunoassays, whereas the Dade-Behring assay is a particle-enhanced nephelometric immunoassay. Each of the estimating equations was developed for one of these assays, which are known to give somewhat different results. Delanaye et al. 23 compared the three methods and established the linear relationships between them. We studied both uncalibrated cystatin C measurements and measurements calibrated to Dade-Behring or Dako, as appropriate, according to the regression equations from Delanaye s work (Table 5 of Delanaye et al. 23 ): cystatin C Dako ¼ (cystatin C Gentian 0.8908) þ 0.1593; and cystatin C Dade-Behring ¼ (cystatin C Gentian 1.0247) 0.1173. The creatinine-based equations investigated in the present study were the recalibrated four-variable MDRD, CKD-EPI, and CG equations. The MDRD equation is 175 (creatinine/ 88.4) 1.154 age 0.203 1.212 (if black) 0.742 (if female). 44 The CKD-EPI equation is 141 min(creatinine/k, 1) a max(creatinine/ k, 1) 1.209 0.993 age 1.018 (if female) 1.159 (if black), where k is 62 for females and 80 for males, a is 0.329 for females and 0.411 for males, min indicates the minimum of creatinine/k or 1, and max indicates the maximum of creatinine/k or 1. 45 The CG equation normalized to 1.73 m 2 body surface area is ((140 age) body weight 1.73/(creatinine 0.8 body surface area)) 0.85 (if female). 46 Creatinine was expressed in mmol/l, body weight in kg, and body surface area in m 2. When used in the CG equation, creatinine measurements were calibrated back to the Jaffe s method based on the comparison between the two methods performed when the Jaffe s method was replaced at our hospital. All the estimating equations give egfr as ml/min per 1.73 m 2. Statistical methods Comparison of differences between included and eligible persons in Table 1 were done with analysis of covariance adjusted for age and gender. The means and s.d. of characteristics of all eligible persons were weighted according to the age stratification and gender stratification of RENIS-T6. Bias and precision were calculated as the median and interquartile range of egfr mgfr, respectively; percentage bias and precision as the median and interquartile range of 100 (egfr mgfr)/mgfr, respectively; and accuracy as the percentage of subjects in which absolute percentage bias was o30% (P 30 ). Pearson s correlation coefficient with mgfr was estimated for each equation. The sensitivity and specificity for detecting a GFR o60 and o80 ml/min per 1.73 m 2 were calculated for each equation. The area under the receiver operating characteristic curve for each of the cutoffs was calculated by estimating the c-statistic for a logistic model with the GFR estimate in question as the only independent variable. The bootstrap method was used to determine 95% confidence intervals from 2000 re-samples of the original observations. 47 To explore variation in bias across the range of egfr, mgfr was regressed on egfr with local regression and displayed in a scatterplot of mgfr vs egfr for the three equations with the highest P 30 among the equations based on cystatin C, creatinine, and both, respectively. PROC SGPLOT in SAS (SAS Institute, Cary, NC, USA) was used. All analyses were performed using SAS version 9.2 (SAS Institute). Significance was set at Po0.01. DISCLOSURE All the authors declared no competing interests. 1310 Kidney International (2010) 78, 1305 1311

BO Eriksen et al.: Estimating GFR with cystatin C in the general population ACKNOWLEDGMENTS We thank Britt-Ann Winther Eilertsen, Bjørg Skog Høgset, Saskia van Heusden, and the rest of the staff at the Clinical Research Unit (University Hospital of North Norway) for performing the study; Harald Strand and the staff at the Department of Medical Biochemistry (University Hospital of North Norway) for HPLC analyses of iohexol; Inger Sperstad and Ingrid Dorthea Sandstad (Clinical Research Centre, University Hospital of North Norway) for database support; and Tom Wilsgaard, Sriharan Sivasingarajah and Kurt Jøran Nyland (Department of Community Medicine, University of Tromsø) for identifying eligible subjects from the Tromsø 6 cohort. REFERENCES 1. Sarnak MJ, Levey AS, Schoolwerth AC et al. Kidney disease as a risk factor for development of cardiovascular disease: a statement from the American Heart Association Councils on Kidney in cardiovascular disease, high blood pressure research, clinical cardiology, and epidemiology and prevention. Circulation 2003; 108: 2154 2169. 2. Shlipak MG, Sarnak MJ, Katz R et al. Cystatin C and the risk of death and cardiovascular events among elderly persons. N Engl J Med 2005; 352: 2049 2060. 3. Astor BC, Levey AS, Stevens LA et al. Method of glomerular filtration rate estimation affects prediction of mortality risk. J Am Soc Nephrol 2009; 20: 2214 2222. 4. Madero M, Sarnak MJ. Association of cystatin C with adverse outcomes. Curr Opin Nephrol Hypertens 2009; 18: 258 263. 5. Reese PP, Feldman HI. More evidence that cystatin C predicts mortality better than creatinine. J Am Soc Nephrol 2009; 20: 2088 2090. 6. Verhave JC, Fesler P, Ribstein J et al. Estimation of renal function in subjects with normal serum creatinine levels: influence of age and body mass index. Am J Kidney Dis 2005; 46: 233 241. 7. Stevens LA, Coresh J, Feldman HI et al. Evaluation of the modification of diet in renal disease study equation in a large diverse population. J Am Soc Nephrol 2007; 18: 2749 2757. 8. 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; 13: 2140 2144. 9. Lin J, Knight EL, Hogan ML et al. A comparison of prediction equations for estimating glomerular filtration rate in adults without kidney disease. J Am Soc Nephrol 2003; 14: 2573 2580. 10. Poggio ED, Wang X, Greene T et al. 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 2005; 16: 459 466. 11. 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 2004; 141: 929 937. 12. Panteghini M. Enzymatic assays for creatinine: time for action. Clin Chem Lab Med 2008; 46: 567 572. 13. Stevens LA, Schmid CH, Greene T et al. Factors other than glomerular filtration rate affect serum cystatin C levels. Kidney Int 2009; 75: 652 660. 14. Knight EL, Verhave JC, Spiegelman D et al. Factors influencing serum cystatin C levels other than renal function and the impact on renal function measurement. Kidney Int 2004; 65: 1416 1421. 15. Taglieri N, Koenig W, Kaski JC. Cystatin C and cardiovascular risk. Clin Chem 2009; 55: 1932 1943. 16. Bird NJ, Peters C, Michell AR et al. Comparison of GFR measurements assessed from single versus multiple samples. Am J Kidney Dis 2009; 54: 278 288. 17. Zahran A, El Husseini A, Shoker A. Can cystatin C replace creatinine to estimate glomerular filtration rate? A literature review. Am J Nephrol 2007; 27: 197 205. 18. Madero M, Sarnak MJ, Stevens LA. Serum cystatin C as a marker of glomerular filtration rate. Curr Opin Nephrol Hypertens 2006; 15: 610 616. 19. Louvar DW, Rogers TB, Bailey RF et al. Cystatin C is not superior to creatinine-based models in estimating glomerular filtration rate in former kidney donors. Transplantation 2007; 84: 1112 1117. 20. Botev R, Mallie JP, Couchoud C et al. Estimating glomerular filtration rate: Cockcroft-Gault and modification of diet in renal disease formulas compared to renal inulin clearance. Clin J Am Soc Nephrol 2009; 4: 899 906. 21. Sebasky M, Kukla A, Leister E et al. Appraisal of GFR-estimating equations following kidney donation. Am J Kidney Dis 2009; 53: 1050 1058. 22. Froissart M, Rossert J, Jacquot C et al. Predictive performance of the modification of diet in renal disease and Cockcroft-Gault equations for estimating renal function. J Am Soc Nephrol 2005; 16: 763 773. 23. Delanaye P, Pieroni L, Abshoff C et al. Analytical study of three cystatin C assays and their impact on cystatin C-based GFR-prediction equations. Clin Chim Acta 2008; 398: 118 124. 24. Ransohoff DF, Feinstein AR. Problems of spectrum and bias in evaluating the efficacy of diagnostic tests. N Engl J Med 1978; 299: 926 930. 25. Andreev E, Koopman M, Arisz L. A rise in plasma creatinine that is not a sign of renal failure: which drugs can be responsible? J Intern Med 1999; 246: 247 252. 26. Brown SC, O Reilly PH. Iohexol clearance for the determination of glomerular filtration rate in clinical practice: evidence for a new gold standard. J Urol 1991; 146: 675 679. 27. Sterner G, Frennby B, Mansson S et al. Determining true glomerular filtration rate in healthy adults using infusion of inulin and comparing it with values obtained using other clearance techniques or prediction equations. Scand J Urol Nephrol 2008; 42: 278 285. 28. Eriksson CG, Kallner A. Glomerular filtration rate: a comparison between Cr-EDTA clearance and a single sample technique with a non-ionic contrast agent. Clin Biochem 1991; 24: 261 264. 29. Rydstrom M, Tengstrom B, Cederquist I et al. Measurement of glomerular filtration rate by single-injection, single-sample techniques, using 51Cr-EDTA or iohexol. Scand J Urol Nephrol 1995; 29: 135 139. 30. Brandstrom E, Grzegorczyk A, Jacobsson L et al. GFR measurement with iohexol and 51Cr-EDTA. A comparison of the two favoured GFR markers in Europe. Nephrol Dial Transplant 1998; 13: 1176 1182. 31. Stevens LA, Levey AS. Measured GFR as a confirmatory test for estimated GFR. J Am Soc Nephrol 2009; 20: 2305 2313. 32. Jacobsson L. A method for the calculation of renal clearance based on a single plasma sample. Clin Physiol 1983; 3: 297 305. 33. Nilsson-Ehle P. Iohexol clearance for the determination of glomerular filtration rate: 15 years experience in clinical practice. ejifcc 2006; 13. 34. Boijsen M, Jacobsson L, Tylen U. Glomerular filtration rate estimated from a single plasma sample after contrast-enhanced radiological examinations. Clin Physiol 1988; 8: 309 316. 35. DuBois D, Dubois EF. The measurement of the surface area of man. Arch Intern Med 1915; 15: 868 881. 36. Rustad P, Felding P, Franzson L et al. The Nordic Reference Interval Project 2000: recommended reference intervals for 25 common biochemical properties. Scand J Clin Lab Invest 2004; 64: 271 284. 37. Hoek FJ, Kemperman FA, Krediet RT. A comparison between cystatin C, plasma creatinine and the Cockcroft and Gault formula for the estimation of glomerular filtration rate. Nephrol Dial Transplant 2003; 18: 2024 2031. 38. Tidman M, Sjostrom P, Jones I. A comparison of GFR estimating formulae based upon s-cystatin C and s-creatinine and a combination of the two. Nephrol Dial Transplant 2008; 23: 154 160. 39. Flodin M, Jonsson AS, Hansson LO et al. Evaluation of Gentian cystatin C reagent on Abbott Ci8200 and calculation of glomerular filtration rate expressed in ml/min per 1.73 m(2) from the cystatin C values in mg/l. Scand J Clin Lab Invest 2007; 67: 560 567. 40. Larsson A, Malm J, Grubb A et al. Calculation of glomerular filtration rate expressed in ml/min from plasma cystatin C values in mg/l. Scand J Clin Lab Invest 2004; 64: 25 30. 41. Rule AD, Bergstralh EJ, Slezak JM et al. Glomerular filtration rate estimated by cystatin C among different clinical presentations. Kidney Int 2006; 69: 399 405. 42. Grubb A, Nyman U, Bjork J et al. Simple cystatin C-based prediction equations for glomerular filtration rate compared with the modification of diet in renal disease prediction equation for adults and the Schwartz and the Counahan-Barratt prediction equations for children. Clin Chem 2005; 51: 1420 1431. 43. Stevens LA, Coresh J, Schmid CH et al. Estimating GFR using serum cystatin C alone and in combination with serum creatinine: a pooled analysis of 3,418 individuals with CKD. Am J Kidney Dis 2008; 51: 395 406. 44. Levey AS, Coresh J, Greene T et al. Using standardized serum creatinine values in the modification of diet in renal disease study equation for estimating glomerular filtration rate. Ann Intern Med 2006; 145: 247 254. 45. Levey AS, Stevens LA, Schmid CH et al. A new equation to estimate glomerular filtration rate. Ann Intern Med 2009; 150: 604 612. 46. Cockcroft DW, Gault MH. Prediction of creatinine clearance from serum creatinine. Nephron 1976; 16: 31 41. 47. Efron B. The Jackknife, the Bootstrap and Other Resampling Plans, 1st edn. Society for Industrial and Applied Mathematics: Philadelphia, 1982. Kidney International (2010) 78, 1305 1311 1311