Assessment of glomerular filtration rate in healthy subjects and normoalbuminuric diabetic patients: validity of a new (MDRD) prediction equation

Transcription

1 Nephrol Dial Transplant (2002) 17: Original Article Assessment of glomerular filtration rate in healthy subjects and normoalbuminuric diabetic patients: validity of a new () prediction equation Gerald Vervoort 1, Hans L. Willems 2 and Jack F. M. Wetzels 1 1 Department of Nephrology and 2 Department of Clinical Chemistry, University Medical Centre Nijmegen, The Netherlands Abstract Background. Based on the data derived from the Modification of Diet in Renal Disease () study, a new equation was developed for the estimation of glomerular filtration rate (GFR). This equation, which takes into account body weight, age, sex, serum creatinine, race, serum urea, and serum albumin, provided a more accurate estimation of GFR in patients with renal insufficiency. However, this prediction equation has not been validated in subjects with normal or supra-normal GFR. Methods. In a cross-sectional study, we measured GFR by inulin clearance in 46 healthy controls and 46 non-complicated type 1 diabetic patients. In this study population, GFR was predicted by measured creatinine clearance, the Cockcroft Gault formula, and the equation. Results. In the healthy subjects, mean GFR ("SD) was 107"11 as compared to 122"18 mlumin per 1.73 m 2 in the diabetic patients. This difference in GFR was reflected by a lower serum creatinine (76"8 vs 71"8 mmolul) in the diabetic patients. In the healthy controls, median absolute differences (and the 50th 75th 90th percentile of percentage absolute differences) between predicted and measured GFR were 5.2 mlumin per 1.73 m 2 ( %) for creatinine clearance, 9.0 mlumin per 1.73 m 2 ( %) for the Cockcroft Gault formula, and 10.7 mlumin per 1.73 m 2 ( %) for the equation. In the diabetic patients, these differences were 8.3 mlumin per 1.73 m 2 ( %) for creatinine clearance; 11.8 mlumin per 1.73 m 2 ( %) for the Cockcroft Gault formula, and 18.8 mlumin per 1.73 m 2 ( %) for the equation. Correspondence and offprint requests to: Gerald Vervoort, MD, Department of Medicine, Division of Nephrology 545, University Medical Centre Nijmegen, PO Box 9101, 6500 HB Nijmegen, The Netherlands. g.vervoort@aig.azn.nl Conclusions. In subjects with a normal or increased GFR, the new -prediction equation of GFR is less accurate than creatinine clearance or the Cockcroft Gault formula, and offers no advantage. Keywords: Cockcroft Gault formula; diabetes mellitus; GFR; inulin clearance; prediction equation Introduction Measurement of the true glomerular filtration rate (GFR) is time consuming and difficult to perform. Therefore, various formulas have been developed that allow a prediction of creatinine clearance or GFR from serum creatinine and demographic characteristics. The Cockcroft Gault formula is probably used most frequently [1]. Recently, a new prediction equation for estimation of the GFR was developed based on the data derived from the Modification of Diet in Renal Disease () study in patients with renal dysfunction [2]. This equation, which takes into account body weight, age, sex, serum creatinine, race, serum urea, and serum albumin, provided a more accurate estimation of GFR than measured creatinine clearance or the Cockcroft Gault formula. However, this prediction equation has not been validated in subjects with normal or supra-normal GFR. Therefore, we measured GFR by a standard technique (inulin clearance) in healthy subjects and non-complicated type 1 diabetic patients. We tested whether in these circumstances the new prediction equation is a better predictor of GFR than creatinine clearance or the Cockcroft Gault formula. # 2002 European Renal Association European Dialysis and Transplant Association

2 1910 G. Vervoort et al. Subjects and methods Study population After approval of the study protocol by the hospital ethics committee, 46 healthy persons and 46 non-complicated type 1 diabetic patients gave their written informed consent for participation in the study. The diabetic patients were recruited from the outpatient diabetic clinic of the University Medical Centre Nijmegen. The onset of diabetes was before the age of 40 years, insulin treatment was started within 1 year of diagnosis, and the duration of diabetes was 5 12 years. All patients were normotensive (blood pressure -140u90 mmhg, mercury sphygmomanometer), normoalbuminuric (urinary albumin excretion -20 mgumin), and without retinopathy (except for simple background retinopathy). The healthy persons were recruited from the local population. Before inclusion in the study, these subjects were screened for the absence of hypertension, cardiovascular disease, renal dysfunction, and microalbuminuria. Medication was not allowed except for oral contraceptives in the female participants. Study design All subjects were on an ad libitum diet the week before the study and were instructed to abstain from alcohol and caffeine consumption for at least 24 h and to refrain from smoking 12 h before the experiments. All participants were studied in the morning after a light breakfast. At a.m. they consumed a water load of 20 mlukg and two intravenous lines were inserted, one for constant infusion of 250 mluh glucose 2.5%uNaCl 0.45% and inulin (Inutest 1, Laevosan- GmbH, Linz, Austria), the other for blood sampling. In the diabetic patients, blood glucose was measured every 20 min using a glucocard (Menarini Diagnostics, Utrecht, The Netherlands). In view of the effects of hyperglycaemia on GFR, glucose concentrations were kept constant between 5 and 8 mmolul during the whole study period [3]. If necessary, additional glucose or insulin was given to maintain these target glucose levels. In addition to the infusion, oral tap water was given to match urine output. The subjects were supine during the clearance study, but were allowed to stand to urinate. After 30 min a constant infusion of inulin was started, preceded by a priming dose. After an equilibration period of at least 70 min, in which urine output was at least 10 mlumin, three urine samples were collected at 20-min intervals. Creatinine levels in serum and urine were determined according to the Jaffé method. Albumin and urea were determined according to the bromcresol green and enzymatic urease methods, respectively. All these assays were performed on a Hitachi 747 (Roche, Almere, The Netherlands). Inulin concentrations were determined in duplicate by an enzymatic assay [4]. Calculations and statistics The calculated urinary clearance of inulin was used as marker of GFR. We compared creatinine clearance, the Cockcroft Gault formula and the equation as predictors of GFR. The following equations were used: Creatinine clearance: (U creat (mmolul)3volume (mlumin))u S creat (mmolul)3(1:73 (m 2 )ubsa (m 2 )) Fig. 1. Ratio of non-corrected C-creat and GFR in relation to GFR in all study subjects; the regression line is represented by Ys 0.004Xq1.54. Table 1. Baseline characteristics of study population (means"sd) Cockcroft Gault: ((140 age (years))3body weight (kg))u (0:813S creat (mmolul))3(1:73 (m 2 )u BSA (m 2 ))(30:85 if female) formula: (1703(S creat (mgudl)) 0:999 )3(age 0:176 ) Healthy subjects Diabetic patients Number Femalesumales 23u23 25u21 Age (years) 28.2" "6.7 Weight (kg) 69.9" "9.7 Body surface (m 2 ) 1.85" "0.15 Serum albumin (gul) 43.6" "2.7 a Serum urea nitrogen (mmolul) 4.4" "1.1 Serum creatinine (mmolul) 76"8 71"8 a Urinary creatinine excretion (mmolu24 h) 13.8" "2.4 a P-0.01 vs healthy subjects. To convert creatinine values in mmolul to mgudl, multiply by ; to convert urea nitrogen values in mmolul to mgudl, multiply by ((S urea nitrogen (mgudl)) 0:170 )3((albumin (gudl)) q0:318 ) (30:762 if female; 31:180 if black) where U creat surinary creatinine concentration; S creat sserum creatinine concentration; S urea nitrogen sserum urea nitrogen concentration; BSAsbody surface area calculated by: BSAs( p height (cm) 3weight (kg)u3600. (To convert creatinine values in mmolul to mgudl, multiply by ; to convert urea nitrogen values in mmolul tomgudl, multiply by 2.8.) Creatinine clearance usually exceeds GFR by 10 15% because of urinary creatinine that is derived from tubular secretion [5]. To correct for this bias, a correction factor can be used. In the original study, this factor was derived from the patients sample. We used a formula correction factor derived from the linear regression line of the ratio of creatinine clearanceugfr and GFR (Figure 1). This line is represented by Ys 0.004Xq1.54. This means that in the GFR range as measured the (percentage) quantity of tubular secreted creatinine will fall with an increase in

3 Assessment of normal and supra-normal GFR Table 2. Comparison of creatinine clearance, Cockcroft Gault formula and the equation for assessment of GFR 1911 GFR C-creatinine C-creatinine C-creatinine Cockcroft Gault Cockcroft Gault Cockcroft Gault (non-corrected) (corrected)* (corrected)** (non-corrected) (corrected)* (corrected)** Healthy subjects All 107"11 118"16 106"17 107"15 112"17 100"15 100"15 103"13 Male 104"9 122"18 109"20 110"16 116"10 103"10 104"9 111"10 Female 109"13 115"13 103"15 103"12 107"21 97"19 96"19 95"12 Diabetic patients All 122"18 125"17 120"23 113"16 119"16 114"20 107"14 108"18 Male 115"15 123"22 115"26 111"19 123"12 114"15 111"11 120"16 Female 128"18 127"13 125"19 114"12 116"18 114"24 104"16 98"14 GFR, glomerular filtration rate (mluminu1.73 m 2 ); C-creatinine, creatinine clearance; Cockcroft Gault, estimation of creatinine clearance by Cockcroft Gault formula;, estimation of GFR by equation; *corrected by Ys 0.004Xq1.54; **corrected by multiplication factor of 0.9. GFR. As such, the overestimation of GFR by creatinine clearance will decrease by an increase in GFR. Since the Cockcroft Gault formula is an estimation of creatinine clearance, the same formula correction factor was used for prediction of GFR. Given that the use of a correction factor derived from the patient s data introduces a bias in favour of the corrected value, we also used a multiplication factor of 0.9. This reflects the average contribution of tubular secretion to creatinine clearance and is based on present and previous studies from our department (data not shown). Since in clinical practice such adjustments are not usually made, we also compared the three methods without using the multiplication factors. The accuracy of each method was determined by calculating the median of the absolute difference and the 50th 75th and 90th percentile of the percentage absolute differences between predicted and measured GFR [2]. Repeated measures of ANOVA (SPSS 9.0.1, SPSS Inc, Chicago, USA) were used to analyse differences between the three equations. Post hoc analysis (Tukey s test) was performed for comparisons between groups only when ANOVA provided a P-value less than Furthermore, Bland Altman plots were made to compare predicted and measured GFR [6]. Differences between healthy subjects and diabetic patients were analysed by unpaired Student s t-test. A P-value less than 0.05 was considered statistically significant. Results The characteristics of the diabetic patients and the healthy controls are summarized in Table 1. There were no differences in demographic or anthropomorphic parameters between the groups. All but two subjects were Caucasian. Serum albumin concentration was significantly lower in the diabetic patients (43.6"3.6 vs 40.3"2.7 gul, P-0.01), and the difference in urea nitrogen values was not significant (4.4"1.1 in healthy controls vs 4.0"1.1 mmolul in diabetic patients). The mean plasma glucose level in diabetic patients was 13.1"0.6 mmolul on arrival at the hospital and 7.1"0.3 mmolul throughout the experiment. Urinary creatinine excretion was comparable in both groups (Table 1). GFR ranged from 80 to 132 mlumin per 1.73 m 2 in the healthy subjects and from 88 to 182 mlumin per Fig. 2. Comparison of equations to predict GFR. C-creatsformulacorrected creatinine clearance (Ys 0.004Xq1.54; Cockcrofts formula-corrected Cockcroft Gault formula; s equation. Black circles, healthy subjects; grey circles, diabetic patients m 2 in the diabetic patients. Mean GFR ("SD) was significantly higher in diabetic patients (107"11 vs 122"18 mlumin per 1.73 m 2, P-0.01). These differences in GFR were reflected by significant differences in serum creatinine (Table 1). The results of the different prediction equations are summarized in Table 2 and Figure 2. The median absolute differences and the 50th 75th 90th percentile of percentage absolute differences between predicted and measured GFR in healthy subjects and diabetic patients are summarized in Table 3. Post hoc analysis showed that formula corrected creatinine clearance (by Ys 0.004Xq1.54) was superior to equation with respect to prediction of measured GFR in healthy subjects. Corrected (by Ys 0.004Xq1.54 as well as by multiplication factor 0.9) and non-corrected creatinine clearances as well as formula corrected (by Ys 0.004Xq1.54) and non-corrected Cockcroft Gault formulae were superior to equation in diabetic patients. Bland Altman plots are shown in Figures 3 5.

4 1912 G. Vervoort et al. Table 3. Differences between predicted and measured GFR (mlumin per 1.73 m 2 ). Median absolute differences (percentage absolute differences, 50th 75th 90th percentile) between predicted and measured GFR Subjects and parameters Differences between predicted and measured GFR P-value (ANOVA) Healthy subjects C-creatinine (corrected)* 5.2 ( %) Cockcroft Gault (corrected)* 9.0 ( %) 10.7 ( %) a C-creatinine (corrected)** 6.5 ( %) NS Cockcroft Gault (corrected)** 8.5 ( %) 10.7 ( %) C-creatinine 13.0 ( %) NS Cockcroft Gault 13.1 ( %) 10.7 ( %) Diabetic patients C-creatinine (corrected)* 8.3 ( %) Cockcroft Gault (corrected)* 11.8 ( %) 18.8 ( %) b,c C-creatinine (corrected)** 10.1 ( %) Cockcroft Gault (corrected)** 16.8 ( %) 18.8 ( %) c C-creatinine 8.8 ( %) Cockcroft Gault 15.1 ( %) 18.8 ( %) b,c GFR, glomerular filtration rate; C-creatinine, creatinine clearance; Cockcroft Gault, estimation of creatinine clearance by Cockcroft Gault formula;, estimation of GFR by equation; *corrected by Ys 0.004Xq1.54; **corrected by multiplication factor 0.9; Post-hoc analysis: a P-0.05 compared with C-creatinine; b P-0.05 compared with Cockcroft Gault; c P-0.01 compared with C-creatinine. Fig. 3. Bland Altman plot of measured GRF and formula-corrected (Ys 0.004Xq1.54) C-creatinine. Black triangles, healthy males; black circles, healthy females; grey triangles, diabetic males; grey triangles, diabetic females. Fig. 4. Bland Altman plot of measured GFR and formula-corrected (Ys 0.004Xq1.54) Cockcroft Gault equation. Black triangles, healthy males; black circles, healthy females; grey triangles, diabetic males; grey circles, diabetic females. To find out whether the difference in the usefulness of GFR evaluation by the equation between diabetic patients and non-diabetic controls was due to the diabetes itself or the higher GFR values in diabetic patients, a stepwise multiple regression analysis was performed with the absolute difference between measured GFR and prediction as a dependent variable, GFR as independent continuous variable, and healthy or diabetes and gender as independent nominal variables. GFR proved to be the most powerful and significant predictor for GFR difference in this model (P-0.01). Discussion We assessed the validity of the new equation for the prediction of GFR in healthy subjects and noncomplicated diabetic patients with increased GFR. From our study, we conclude that the new prediction equation, at least in Caucasians, is less accurate than corrected creatinine clearance or the Cockcroft Gault formula in persons with normal or supra-normal glomerular filtration. The new prediction equation was developed by computer modelling using a set of variables determined

5 Assessment of normal and supra-normal GFR Fig. 5. Bland Altman plot of measured GFR and equation. Black triangles, healthy males; black circles, healthy females; grey triangles, diabetic males; grey circles, diabetic females. by stepwise multiple regression analyses in a large sample of patients with renal failure participating in the study [2]. Persons without renal disease and persons with type 1 diabetes were excluded from that study. Our study shows that the new prediction equation has a tendency to underestimate GFR in subjects with a GFR in the supra-normal range. Most probably there is no linear correlation between the parameters used in the prediction equation (urea nitrogen, serum albumin, and race) and this GFR range. Creatinine clearance usually exceeds GFR because of tubular secretion of creatinine [5]. Therefore, when calculating GFR from creatinine data, a correction factor may be used to account for tubular secretion. In the study, a correction factor derived from the sample data was used [2]. Even when using such a correction factor the creatinine clearance and Cockcroft Gault formula proved inferior to the -equation in patients with renal failure. We initially also used a correction factor (Ys 0.004Xq1.54) based on the sample data. When using this correction factor, creatinine clearance was superior to the formula in reflecting true GFR as measured by inulin clearance. However, the use of such a correction factor introduces a bias, particularly if the factor was directly derived from the sample population. Furthermore, the correction factor is dependent on the level of GFR, since tubular secretion of creatinine increases with decreasing GFR (Figure 1). In routine clinical practice, GFR is not known, and the correction factor cannot be calculated from measured serum creatinine or creatinine clearance. We used a multiplication factor of 0.9, which reflects the average ratio of creatinine clearanceugfr, observed in previous experiments. It is generally agreed that tubular excretion of creatinine is responsible for ;10 15% of creatinine excretion [5]. Therefore, in general a correction factor of seems appropriate for routine use. Admittedly, the use of such average estimate of this factor introduces some error. However, our data clearly indicate that this will have no major impact on the conclusions. In fact, even when using non-corrected creatinine clearance and Cockcroft Gault formula, the equation was not of any additional value. Thus, the use of the formula is of no value for the estimation of GFR in subjects with normal or above normal GFR. We feel, however, that the formula may have advantages in patients with severe renal failure. Prospective long-term studies are needed to see if the use of this formula is of real benefit in clinical practice, particularly with respect to a better assessment of the rate of decline of GFR. Notably, subgroup analysis showed that the equation and the Cockcroft Gault formula underestimated GFR, particularly in the female diabetic patients. Most probably this is due to changes in body composition. Although there were no significant differences in body weight (66"10 kg in female controls vs 68"8 kg in diabetic female patients), urinary creatinine excretion was significantly higher in diabetic female patients than in non-diabetic female subjects (urinary creatinine excretion 11.8"1.4 vs 13.0" 1.7 mmolu24 h, Ps0.01). There were no significant differences in body weight (74"8 vs 75"10 kg in diabetic male patients) or in creatinine excretion between diabetic and non-diabetic male patients (15.7"2.6 vs 14.8"2.7 mmolu24 h in diabetic male patients). We conclude, therefore, that the new prediction equation is less precise than corrected creatinine clearance or the Cockcroft Gault formula, and offers no advantage in persons with normal or increased GFR. References Cockcroft DW, Gault MH. Prediction of creatinine clearance from serum creatinine. Nephron 1976; 16: 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; 130: Wiseman MJ, Mangili R, Alberetto M et al. Glomerular response mechanisms to glycemic changes in insulin-dependent diabetics. Kidney Int 1987; 31: Degenaar CP, Frenken LA, v Hooff JP. Enzymatic method for determination of inulin. Clin Chem 1987; 33: Shemesh O, Golbetz H, Kriss JP et al. Limitations of creatinine as a filtration marker in glomerulopathic patients. Kidney Int 1985; 28: Bland JM, Altman DG. Comparing methods of measurement: why plotting against difference against standard method is misleading. Lancet 1995; 346: Received for publication: Accepted in revised form: