Performance of the Cockcroft-Gault and Modification of Diet in Renal Disease Equations in Estimating GFR in Ill Hospitalized Patients

Performance of the Cockcroft-Gault and Modification of Diet in Renal Disease Equations in Estimating GFR in Ill Hospitalized Patients Emilio D. Poggio, MD, Patrick C. Nef, MD, Xuelei Wang, MS, Tom Greene, PhD, Frederick Van Lente, PhD, Vincent W. Dennis, MD, and Phillip M. Hall, MD ● Background: Estimating glomerular filtration rate (GFR) in severely ill inpatients is clinically important for therapeutic interventions and prognosis, but notoriously difficult to do accurately. The Modification of Diet in Renal Disease (MDRD) equation and Cockcroft-Gault (CG) formula are widely used to estimate renal function in sick hospitalized patients; however, neither method has been validated in this setting. Methods: Iodine 125–iothalamate clearances (iGFR) performed in 107 sick inpatients with renal dysfunction were compared with estimated GFRs (eGFRs) from the 6- and 4-variable MDRD (MDRD eGFR) and CG (CG eGFR) equations. Results: Mean serum creatinine (SCr) level was 3.5 ⴞ 2.0 mg/dL (309 ⴞ 177 ␮mol/L), and mean iGFR was 17.1 ⴞ 17.9 mL/min/1.73 m2 (0.29 ⴞ 0.30 mL/s/1.73 m2). Six-variable MDRD eGFR was 22.5 ⴞ 17.4 mL/min/1.73 m2 (0.38 ⴞ 0.29 mL/s/1.73 m2), 4-variable MDRD eGFR was 23.9 ⴞ 16.3 mL/min/1.73 m2 (0.40 ⴞ 0.27 mL/s/1.73 m2), and CG eGFR was 26.0 ⴞ 17.1 mL/min/1.73 m2 (0.43 ⴞ 0.29 mL/s/1.73 m2). Blood urea nitrogen (BUN)/SCr ratios greater than 20 were seen in 58% of patients. Overall, the CG and MDRD equations overestimated iGFR, with poor agreement. Overestimation of at least 25% of measured iGFR was seen in 63%, 67%, and 70% of all inpatients when using the 6-variable MDRD, 4-variable MDRD, and CG equations, respectively. Accuracy of eGFR within 50% of measured iGFR was 55% for the 6-variable MDRD equation, 49% for the 4-variable MDRD equation, and 40% for the CG formula. The performance of both methods deteriorated further in patients with a BUN/SCr ratio greater than 20. Conclusion: Estimation equations are performed poorly compared with iGFR and are not reliable measures of actual level of function in sick hospitalized patients, especially those with a high BUN/SCr ratio. Although use of the 6-variable MDRD equation provides a better estimation of GFR, it still is unsuitable for clinical application in this population. Am J Kidney Dis 46:242-252. © 2005 by the National Kidney Foundation, Inc. INDEX WORDS: Glomerular filtration rate (GFR); iodine 125 (125I)–iothalamate GFR; hospitalized patients; CockcroftGault formula; Modification of Diet in Renal Disease (MDRD) equation; serum creatinine.

E

STIMATION OF glomerular filtration rate (GFR) is important in the clinical management of hospitalized patients. GFR is used to assess the presence and degree of underlying renal impairment, follow up the course of kidney disease, and allow appropriate dosing of renally excreted drugs.1,2 Serum creatinine (SCr) level and creatinine clearance are used commonly to

From the Departments of Nephrology and Hypertension, Internal Medicine, Biostatistics, and Clinical Pathology, Cleveland Clinic Foundation, Cleveland, OH. Received December 13, 2004; accepted in revised form April 21, 2005. Originally published online as doi:10.1053/j.ajkd.2005.04.023 on June 20, 2005. E.D.P. was a recipient of a fellowship grant from the National Kidney Foundation (2003 to 2004). Presented in part at the American Society of Nephrology meeting, San Diego, CA, November 12-17, 2003. Address reprint requests to Emilio D. Poggio, MD, Department of Nephrology and Hypertension (Desk A51), Cleveland Clinic Foundation, 9500 Euclid Ave, Cleveland, OH 44195. E-mail: [email protected] © 2005 by the National Kidney Foundation, Inc. 0272-6386/05/4602-0007$30.00/0 doi:10.1053/j.ajkd.2005.04.023 242

estimate GFR; however, they are inaccurate compared with inulin clearance, particularly in patients with chronic kidney disease.3-8 Moreover, accurate collection of timed urine specimens is cumbersome, especially in the context of the current pace of hospital stays. Alternative methods of renal function evaluation involve the use of such methods as the Cockroft-Gault (CG) and Modification of Diet in Renal Disease (MDRD) formulas.9,10 Each is based mainly on the inverse of SCr concentration. These equations are used in both inpatient and outpatient settings to estimate creatinine clearance or GFR, respectively. The CG formula is based on 24-hour creatinine clearance measurements performed in an inpatient cohort with mild renal dysfunction.9 The MDRD equation is based on iodine 125 (125I)–iothalamate clearances in a cohort of outpatients with moderate to severe renal impairment.10 An abbreviated 4-variable MDRD equation that uses only age, sex, race, and SCr level has been the most accepted and used.11 A 6-variable MDRD equation also adds albumin and blood urea nitrogen (BUN) concentrations into the model. None of these equations

American Journal of Kidney Diseases, Vol 46, No 2 (August), 2005: pp 242-252

ESTIMATING GFR IN ILL HOSPITALIZED PATIENTS

has been thoroughly validated in a large number of ill hospitalized patients, for whom the anthropometric and laboratory variables used in these formulas may be distorted.12,13 Elevated BUN/SCr ratios often are used by clinicians to determine the cause of certain renal conditions, such as prerenal states or gastrointestinal bleeding, among others. In these situations, an elevated BUN/SCr ratio is caused mainly by an increase in BUN generation, rather than a decrease in SCr level.1,14-17 Moreover, in sick hospitalized patients, high BUN/SCr ratios could be caused by a decrease in SCr level (because of muscle mass loss, malnourishment, and decreased production, as in liver failure) despite the presence of renal failure; thus, elevated BUN/ SCr ratios also may represent a marker for severity of illness in this situation. The 6-variable MDRD equation may be a more applicable method in circumstances in which these variables are affected by overall illness. Several investigators have studied and validated radionuclide clearance, such as 125Iiothalamate (iGFR), as an alternative method of measuring GFR.18,19 In this regard, the Department of Nephrology and Hypertension at the Cleveland Clinic Foundation (CCF; Cleveland, OH) has been performing iGFR measurements in hospitalized patients and has served as a resource laboratory for this procedure. The present study compares the performance of estimated GFR (eGFR) using the 4- and 6-variable MDRD equations and CG formula with measured iGFR in ill hospitalized patients. METHODS Patients hospitalized at CCF with varying degrees of kidney dysfunction who had iGFR measurements performed between November 1996 and February 2003 were identified. Inpatient selection was based on the individual nephrologist’s perception of laboratory values not reflecting actual GFR. Demographic information (age, sex, and race), anthropometric data (body size), and laboratory data (SCr, BUN, and albumin levels) available at the time of GFR measurement were collected for 107 patients. Patients were excluded from the study if: (1) they had incomplete data, (2) they were undergoing some form of dialysis therapy at or around the time of GFR measurement, (3) SCr level was less than 0.3 mg/dL (⬍27 ␮mol/L), and (4) they had unstable renal function. Stability of renal function was defined as a 15% or less change in SCr value both the day before and the day after iGFR measurement. The performance of estimation equations also was compared between individuals with a BUN/SCr ratio greater or less than 20. This study was

243

approved by the Institutional Review Board at the CCF. Informed consents were waived by the board because of the nature of the study. Blood samples obtained simultaneously with the iGFR were used to measure SCr by means of the modified kinetic Jaffé reaction, using a Hitachi 747-200 Chemistry Analyzer (1996 to 2001) or Hitachi D 2400 Modular Chemistry Analyzer thereafter (Roche Diagnostics, Indianapolis, IN). CCF SCr assays were calibrated by using College of American Pathology (CAP) samples to ensure that a large calibration bias was absent. Paired SCr measurements by means of the MDRD and CCF laboratories were compared for 89 CAP samples obtained at 14 times between 1996 and 2002. Values within ⫾0.3 mg/dL (⫾27 ␮mol/L) or 15% of mean SCr level are considered by the CAP to be within an acceptable range.20 Overall, mean SCr for the 89 paired CAP specimens was similar between the MDRD and CCF laboratories (MDRD – CCF mean difference ⫽ 0.04 ⫾ 0.02 [SE] mg/dL [4 ⫾ 2 ␮mol/L]; P ⫽ 0.12), suggesting that a large calibration bias was unlikely to adversely affect interpretation of eGFR results in this study. However, in 28 CAP specimens in which average SCr level for the 2 laboratories was less than 2 mg/dL (⬍177 ␮mol/L), mean SCr level was significantly greater for the MDRD laboratory (0.09 ⫾ 0.03 mg/dL [8 ⫾ 3 ␮mol/L]; P ⫽ 0.006), suggesting the possibility of a limited calibration bias at lower SCr levels. Nevertheless, effects of limited calibration bias would not be expected to be significant because of the relatively high SCr levels of patients in this study. No calibration data were available for measurements of serum albumin and serum urea nitrogen between the MDRD and CCF laboratories. GFR was measured in study patients by using plasma disappearance of 125I-sodium iothalamate using the 2-compartment model.21 Each patient was administered 5 drops of supersaturated potassium iodine diluted in 15 mL of water to inhibit thyroid uptake. A background blood sample was collected and saved. Twenty-five microcuries of 125I-sodium iothalamate (Glofil; Questor Pharmaceuticals Inc, Union City, CA) was injected intravenously at time 0. Blood samples were drawn at 5, 10, 15, 300, 330, and 360 minutes after injection. To accurately obtain GFR measurement in patients with an expected GFR less than 30 mL/min/1.73 m2 (⬍0.50 mL/s/1.73 m2), age older than 65 years, or SCr level greater than 2.5 mg/dL (⬎221 ␮mol/L), a sample at 24 hours was also collected to calculate the final phase of the disappearance curve.22,23 Isotope activity was determined by gamma counting of 0.5 mL of plasma on a Packard Minaxi 5000 series counter (Perkin Elmer Life Sciences, Downers Grove, IL). For this study, iGFR is defined as plasma clearance of 125I-iothalamate using the 2-compartment version of the Sapirstein equation24: GFR ⫽ D(In 2) ⁄ [A(t)a ⫹ B(t)b] where D is the injected dose; A and B are intercepts of 2 compartments on a semilog plot of plasma activity versus time; and ta and tb are serum count-per-minute half-lives for the 2 serum sample series fitted to a 2-compartment pharmacokinetic model. GFR results were corrected to standard body surface area (1.73 m2).

244

POGGIO ET AL Table 1. Demographic Characteristics of Study Population

Sample size Age (y) Female sex Non–African-American race Weight (kg) Height (cm) Body surface area (m2) Patients in intensive care unit

All Inpatients

BUN/SCr Ratio ⱕ20

BUN/SCr Ratio ⬎20

P*

107 65 ⫾ 15 (39-82) 48 (45) 92 (86) 80.1 ⫾ 21.3 (57.1-108.5) 169 ⫾ 10 (153-183) 1.89 ⫾ 0.26 (1.59-2.32) 46 (43)

44 61 ⫾ 16 (37-79) 14 (32) 34 (77) 79.7 ⫾ 21.0 (58.3-108.5) 171 ⫾ 10 (157-183) 1.90 ⫾ 0.25 (1.60-2.32) 12 (28)

63 68 ⫾ 14 (53-83) 34 (54) 58 (92) 80.4 ⫾ 21.7 (57.1-103.9) 167 ⫾ 10 (152-180) 1.89 ⫾ 0.26 (1.56-2.24) 34 (54)

0.02 0.02 0.03 0.79 0.13 0.58 ⬍0.01

NOTE. Values expressed as mean ⫾ SD (10th to 90th percentiles) or number (percent). *P corresponds to a BUN/SCr ratio of 20 or less versus greater than 20.

GFR was calculated using the CG and 4- and 6-variable MDRD equations. CG eGFR9: ●

For men: CG eGFR ⫽ [(140 – age) ⫻ weight (kg)]/SCr ⫻ 72

●

For women: CG eGFR ⫽ ([(140 – age) ⫻ weight (kg)]/SCr ⫻ 72) ⫻ 0.85

Both equations then were adjusted for body surface area of 1.73 m2. Four-variable MDRD eGFR11: ●

4-Variable MDRD eGFR ⫽ 186 ⫻ [SCr]–1.154 ⫻ [age]–0.203 ⫻ [0.742 if patient is female] ⫻ [1.212 if patient is African American]

Six-variable MDRD eGFR10: ●

6-Variable MDRD eGFR ⫽ 170 ⫻ [SCr]–0.999 ⫻ [age]–0.176 ⫻ [0.762 if patient is female] ⫻ [1.180 if patient is African American] ⫻ [BUN]–0.170 ⫻ [albumin]⫹0.318

Statistical Analysis Subgroups were compared by using 2-sample t-test or Wilcoxon rank sum test, when appropriate. MDRD eGFR and CG eGFR were plotted against measured iGFR. The 45° line indicates the line of identity. Concordance correlation coefficient (Rc) was presented as an index of agreement, and Pearson correlation coefficient as a simple linear correlation. Both correlation coefficients were performed after log transformation of the data. Concordance correlation reflects both bias and linear association; thus, it provides a more comprehensive measure of agreement than the Pearson correlation, which evaluates linear association only. Both correlation coefficients are reported. The high Pearson correlation seen in the results is caused in part by the large range of GFRs. Bias, a measure of systematic error, was assessed by both mean difference and percentage of median difference between eGFRs using the MDRD or CG equation and measured iGFR. Overall agreement was evaluated by the median absolute difference and median percentage of absolute difference, as well as percentage of eGFR values within 30% and 50% of measured iGFR (accuracy within 30% and 50%). Mean bias for each estimation equation was compared

between subgroups with BUN/SCr ratios of 20 or less and greater than 20 after controlling for age, weight, race, and sex. P for comparisons of agreement of eGFR and iGFR between the 6-variable MDRD eGFR and either 4-variable MDRD eGFR or CG eGFR were obtained by using the McNemar test for percentage of eGFR values within 30% and 50% of iGFR and the bootstrap method (using 400 independent replications) for other indices.

RESULTS

Table 1 lists characteristics of the study group. Mean age was 65 years, 45% were women, and 86% were of nonblack race. Mean SCr level was 3.5 ⫾ 2.0 mg/dL (309 ⫾ 177 ␮mol/L), whereas mean iGFR was 17.1 ⫾ 17.9 mL/min/1.73 m2 (0.29 ⫾ 0.30 mL/s/1.73 m2). Table 2 lists other indices of renal function for the entire cohort of study patients and subgroups. Figure 1 shows the association between eGFR and iGFR in 107 hospitalized patients. Figures 2 and 3 show the relation between eGFR and iGFR in patients with BUN/SCr ratios of 20 or less (n ⫽ 44) and greater than 20 (n ⫽ 63), respectively. As listed in Table 3, the 6-variable MDRD equation shows a better concordance correlation coefficient compared with the 4-variable MDRD equation and CG formula (Rc of 0.66 versus 0.57 and 0.46, respectively; P ⬍ 0.01). Further analysis showed that concordance correlation indices were lower for patients with BUN/SCr ratios greater than 20 (Rc of 0.55 for 6-variable MDRD versus 0.43 and 0.31 for 4-variable MDRD and CG, respectively; P ⬍ 0.01) and higher for patients with a BUN/SCr ratio less than 20 (Rc of 0.74 for 6-variable MDRD equation versus 0.69;

NOTE. Values expressed as mean ⫾ SD (10th, median, 90th percentiles). To convert SCr in mg/dL to ␮mol/L, multiply by 88.4; GFR in mL/min/1.73 m2 to mL/s/1.73 m2, multiply by 0.01667. *BUN/SCr ratio of 20 or greater versus less than 20 by means of nonparametric test.

⬍0.01 0.04 ⬍0.01 0.02 ⬍0.01 ⬍0.01 2.8 ⫾ 1.0 (1.5, 2.7, 4.1) 17.1 ⫾ 12.4 (7.0, 15.0, 30.0) 26.5 ⫾ 15.6 (13.0, 21.9, 39.6) 23.5 ⫾ 15.0 (12.6, 18.4, 42.5) 28.1 ⫾ 14.3 (15.2, 24.9, 46.1) 37 ⫾ 14 (23, 35, 56) 4.5 ⫾ 2.5 (2.1, 4.1, 7.7) 17.1 ⫾ 24.0 (7.0, 10.0, 35.0) 20.1 ⫾ 16.7 (8.2, 15.0, 34.2) 21.0 ⫾ 20.4 (7.7, 15.5, 37.6) 22.8 ⫾ 20.1 (8.7, 18.9 38.2) 14 ⫾ 5 (7, 14, 19) 3.5 ⫾ 2.0 (1.7, 3.1, 6.0) 17.1 ⫾ 17.9 (7.0, 12.0, 30.0) 23.9 ⫾ 16.3 (9.8, 20.1, 37.1) 22.5 ⫾ 17.4 (10.1, 17.4, 38.8) 26.0 ⫾ 17.1 (13.0, 21.3, 41.8) 28 ⫾ 16 (10, 24, 46) SCr (mg/dL) iGFR (mL/min/1.73 m2) 4-Variable MDRD eGFR (mL/min/1.73 m2) 6-Variable MDRD eGFR (mL/min/1.73 m2) CG eGFR (mL/min/1.73 m2) BUN/SCr ratio

BUN/SCr Ratio ⱕ20 All Inpatients

Table 2. Renal Function Characteristics of Study Population

BUN/SCr Ratio ⬎20

P*

ESTIMATING GFR IN ILL HOSPITALIZED PATIENTS

245

P ⫽ not significant for 4-variable MDRD and 0.59 for CG; P ⬍ 0.01). Overall median bias for the 6-variable MDRD equation was 4.9 mL/min/1.73 m2 (0.08 mL/s/ 1.73 m2) versus 6.2 and 8.2 mL/min/1.73 m2 (0.10 and 0.14 mL/s/1.73 m2) for the 4-variable MDRD and CG formulas, respectively (P ⬍ 0.01; Table 3). Overestimation of iGFR by at least 25% is seen in 64% of patients by using the 6-variable MDRD equation, 67% when using the 4-variable MDRD equation, and 71% of patients when the CG formula is applied. This bias is larger for patients with a BUN/SCr ratio greater than 20 compared with those with a preserved BUN/SCr ratio after adjusting for age, sex, race, and weight. The smallest bias is seen in patients with a BUN/SCr ratio less than 20 and using the 6-variable MDRD equation to estimate GFR (median difference, 3.4 mL/min/1.73 m2 [0.06 mL/s/1.73 m2]), whereas the largest bias is seen when applying the CG formula to patients with a BUN/SCr ratio greater than 20 (mean difference, 8.9 mL/min/1.73 m2 [0.15 mL/s/1.73 m2]). Nevertheless, all equations performed poorly with respect to iGFR in any given subgroup (see median percentage of differences for each equation in each subgroup). Accuracy within 30% and 50% (percentage of estimated values within 30% and 50% of measured value) were used as indices of agreement between eGFR and iGFR. Overall, less than 60% of estimated values by either the MDRD or CG equation were within 50% of median iGFR (Table 3). Moreover, this poor agreement was even lower for the subgroup of patients with a BUN/ SCr ratio greater than 20, for which only approximately 50% of estimated values were within 50% of measured GFR. Whereas the 6-variable MDRD equation performs moderately better than the 4-variable MDRD and CG equations, overall performance is still poor compared with iGFR. DISCUSSION

This study evaluates the performance of SCrbased equations (MDRD and CG equations) compared with GFR by means of iothalamate clearances in severely ill hospitalized patients with stable renal dysfunction. The daily need by clinicians to estimate GFR in inpatients requires a method that is precise, accurate, and easy to perform. The current Kidney Disease Outcomes

246

POGGIO ET AL

Fig 1. Association of eGFR with iGFR in 107 hospitalized patients. Note that iGFR is plotted on the Y axis. To convert GFR in mL/min/1.73 m2 to mL/s/1.73 m2, multiply by 0.01667.

Quality Initiative guidelines propose the use of the abbreviated MDRD equation for estimation of GFR in outpatients with chronic kidney disease.25 This equation includes only age, sex, race, and SCr level.11 However, the 6-variable MDRD equation incorporates albumin and BUN levels, and it may be a more appropriate equation in sick inpatients, for whom these variables also may reflect severity of illness.10 These methods have been validated to some extent in different populations26-31; however, their applicability in ill hospitalized patients is still unclear. In the current study, we sought to address this issue, and not surprisingly, we report on the weak performance of estimation equations in this popu-

lation. Performance was especially poor in individuals with an elevated BUN/SCr ratio; nevertheless, in this setting, use of the 6-variable MDRD equation moderately improved overall precision and accuracy of the method in estimating GFR, albeit not enough to warrant clinical applicability. The poor performance of the CG and MDRD equations in ill hospitalized patients may be explained, in part, by the methods and populations used to develop these equations. The CG formula was derived from creatinine clearances based on 24-hour urine collections in 249 subjects in a medical ward in a Veterans Administration facility,9 for which 96% of patients were

ESTIMATING GFR IN ILL HOSPITALIZED PATIENTS

247

Fig 2. Association of eGFR with iGFR in 47 hospitalized patients with a BUN/SCr ratio of 20 or less. Solid line, line of identity; dotted line, 30% agreement; dashed line, 50% agreement. eGFR using the (A) 6-variable MDRD equation, (B) 4-variable MDRD equation, and (C) CG formula. Note that iGFR is plotted on the Y axis. To convert GFR in mL/min/1.73 m2 to mL/s/1.73 m2, multiply by 0.01667.

men. Because of the relative difference in amounts of fat and muscle in men compared with women, this formula had to be adjusted for women. Moreover, in this study, less than 25% of study patients had a mean creatinine clearance less than 50 mL/min/1.73 m2 (⬍0.83 mL/s/1.73 m2). The MDRD equation was based on 125Iiothalamate urinary clearances in an outpatient cohort of 1,628 subjects10 with stable chronic kidney disease. The MDRD equation was derived in an outpatient cohort and thus may not be applicable to the inpatient setting. Because of decreased urine output in patients with a low

GFR, GFR measurements in hospitalized patients are performed routinely by the CCF laboratory by using plasma clearances of iothalamate instead of urine clearances. Previous reports comparing urinary versus plasma disappearance methods of iothalamate GFR have shown excellent correlation.23,32 It is important to note that in patients with a very low GFR, late (even up to 24-hour) plasma samples should be obtained to build an accurate late-phase disappearance curve. This maneuver will allow more accurate measurement of plasma clearance of isotopes in patients with severe renal dysfunction.

248

POGGIO ET AL

Fig 3. Association of eGFR with iGFR in 64 hospitalized patients with a BUN/SCr ratio greater than 20. Solid line, line of identity; dotted line, 30% agreement; dashed line, 50% agreement. eGFR using the (A) 6-variable MDRD equation, (B) 4-variable MDRD equation, and (C) CG formula. Note that iGFR is plotted on the Y axis. To convert GFR in mL/min/1.73 m2 to mL/s/1.73 m2, multiply by 0.01667.

Because these equations are creatinine based, other important factors affecting their performance are creatinine metabolism and SCr measurement. SCr is used commonly by clinicians to estimate GFR; however, its use in sick inpatients may be limited because SCr concentration could be affected by factors other than creatinine filtration.1 Some of these factors include sex, body mass, age, diet, and tubular secretion of creatinine.8,33,34 These variables further complicate the applicability of creatinine-based formulas in accurately estimating GFR in hospitalized patients. Another important issue to consider is SCr

assay calibration. A small systematic bias in SCr measurement could greatly affect eGFR results in situations in which SCr values are within the normal range,20,30 whereas large systematic errors could affect results of equations in individuals with more advanced kidney dysfunction. In our study, large calibration errors are unlikely to have affected our results because there was good agreement between the CCF and MDRD laboratories on CAP samples (mean ⫾ SE ⫽ 0.04 ⫾ 0.02 mg/dL). Conversely, effects of potential limited SCr calibration bias would not be expected to greatly affect results because of the

Pearson R* (log scale)

All inpatients (n ⫽ 107) 6-Variable MDRD eGFR 4-Variable MDRD eGFR† CG eGFR† BUN/SCr ratio ⱕ20 (n ⫽ 44) 6-Variable MDRD eGFR 4-Variable MDRD eGFR CG eGFR† BUN/SCr ratio ⬎20 (n ⫽ 63) 6-Variable MDRD eGFR 4-Variable MDRD eGFR† CG eGFR†

Concordance R (log scale)

Mean Difference (mL/min/1.73 m2)

Median Difference (mL/min/1.73 m2)

Median Difference (%)

Median Absolute Difference (mL/min/1.73 m2)

Median Absolute Difference (%)

Agreement Within (%) 30%

50%

0.77 0.71 0.66

0.66 0.57 0.46

5.4 6.8 8.9

4.9 6.2 8.2

46 53 71

6.1 7.2 8.3

47 53 71

36 31‡ 26

55 49§ 40

0.83 0.77 0.74

0.74 0.69‡ 0.59

3.9 3.0‡ 5.8

3.4 3.4‡ 6.5

45 45‡ 71

4.9 4.7‡ 6.9

47 47‡ 71

32 32‡ 25‡

55 57‡ 48‡

0.69 0.63 0.53

0.55 0.43 0.31

6.1 8.5 8.9

48 60 69

6.5 8.6 9.2

48 60 69

38 30‡ 27

56 43 35

6.4 9.4 11.0

ESTIMATING GFR IN ILL HOSPITALIZED PATIENTS

Table 3. Analysis of Correlation, Bias, and Agreement Between eGFR by the MDRD and CG Equations and iGFR

NOTE. To convert GFR in mL/min/1.73 m2 to mL/s/1.73 m2, multiply by 0.01667. *High Pearson R caused in part by the high range of GFRs. †Unless indicated, P for each result ⬍0.01 for 6-variable MDRD eGFR versus 4-variable MDRD eGFR and CG eGFR. ‡P ⫽ not significant versus 6-variable MDRD eGFR (P ⬎ 0.05). §P ⫽ 0.05 versus 6-variable MDRD eGFR.

249

250

relatively low GFR of patients in this study. Finally, calibration bias assessment for albumin and BUN levels was not investigated in this study; therefore, it is unknown whether any potential calibration bias of these variables could affect interpretation of results. The reason for subdividing our study population into 2 subgroups (ie, BUN/SCr ratio ⱕ20 and ⬎20) was based on the fact that BUN/SCr ratios greater than 20 may represent individuals with distorted values for the anthropometric and laboratory variables used in these equations. Patients with an elevated BUN/SCr ratio were more likely to be older non–African-American women and have more preserved renal function (Tables 1 and 2). High BUN/SCr ratios are seen commonly in hospitalized patients, and estimation of GFR in this population is challenging for clinicians. In this situation, our results suggest that use of estimation equations in general is poor and not sufficient for clinical applicability. The 6-variable MDRD equation showed moderately improved performance compared with the 4-variable MDRD equation in the subgroup with higher BUN/Scr ratios, but showed no improvement in subjects with a normal BUN/SCr ratio. This finding also would support the notion that the 4-variable MDRD equation in general performs similar to the 6-variable MDRD equation in healthier subjects with a preserved BUN/SCr ratio. Overall, when BUN/SCr ratio is preserved, bias and agreement of the equations improve considerably, albeit not enough to warrant their use in sick hospitalized subjects. Because of the importance of assessing GFR in daily clinical practice, the performance of different estimation equations has been a matter of extensive research in recent years. Because of the nature of the MDRD equation, this method performs with good precision and accuracy when applied to outpatients with stable chronic kidney disease and expected GFR less than 60 mL/min/ 1.73 m2 (⬍1.00 mL/s/1.73 m2).30,31 However, application of this equation in individuals with higher levels of renal function has been more problematic, and it appears to underestimate GFR in healthy individuals.30,31 This underestimation could be explained in part by the regression to the mean effect. The MDRD equation was developed from subjects with a median iGFR of approximately 40 mL/min/1.73 m2 (⬃0.67 mL/s/

POGGIO ET AL

1.73 m2); therefore, the error of this equation in subjects with normal GFR tends to be directed toward a lower GFR of that particular population. Similarly, the overestimation seen in sick hospitalized patients could be explained in part by this same mechanism, which, in this case, is opposite to what happens in healthy subjects. These equations also have been studied in renal transplant recipients,28,35 liver transplant recipients,36 and Asians subjects with chronic kidney disease37 with different degrees of performance. However, the main limitation of these studies has been lack of calibration of SCr measurement, which could significantly affect the interpretation of their results. In our study, compared with iGFR, the CG and MDRD formulas performed poorly with respect to their ability to predict actual GFR, shown by their poor agreement and high degree of bias (Table 3). These equations reflect the population characteristics from which they were derived; therefore, it is expected that a new equation/ modification derived directly from pooled data obtained from inpatients at different institutions will be needed to develop a more applicable and consistent method. The 6-variable MDRD equation could be improved further by providing more strength to BUN and albumin variables when it is intended to be applied in sick hospitalized patients. To solve this problem, there currently is an ongoing National Institutes of Health– sponsored study designed to address this important issue. In the meantime, routine reporting of GFR for sick inpatients by hospital laboratories using the current formula methods should be reconsidered. In conclusion, in sick hospitalized patients with advanced renal failure, the CG and MDRD equations perform poorly when estimating GFR and are not reliable measures of actual level of renal function. Performance of the formulas deteriorates further in hospitalized patients with a BUN/SCr ratio greater than 20. The moderately better performance of the 6-variable MDRD equation over the 4-variable equation suggests that improved estimation of GFR may be possible in sick hospitalized patients by incorporating albumin and BUN levels when developing new equations.

ESTIMATING GFR IN ILL HOSPITALIZED PATIENTS

ACKNOWLEDGMENT The authors thank Henry Rolin, Diane Pexa, and Gerald Roberson for implementation of iGFR and data collection.

REFERENCES 1. Levey AS: Measurement of renal function in chronic renal disease. Kidney Int 38:167-184, 1990 2. Barratt TM, Chantler C, Lederman S, Rigden SP: Assessment of renal function. Am J Kidney Dis 7:347-349, 1986 3. Kim KE, Onesti G, Ramirez O, Brest AN, Swartz C: Creatinine clearance in renal disease. A reappraisal. Br Med J 4:11-14, 1969 4. Walser M: Assessing renal function from creatinine measurements in adults with chronic renal failure. Am J Kidney Dis 32:23-31, 1998 5. Rapoport A, Husdan H: Endogenous creatinine clearance and serum creatinine in the clinical assessment of kidney function. Can Med Assoc J 99:149-156, 1968 6. Perrone RD, Madias NE, Levey AS: Serum creatinine as an index of renal function: New insights into old concepts. Clin Chem 38:1933-1953, 1992 7. Shemesh O, Golbetz H, Kriss JP, Myers BD: Limitations of creatinine as a filtration marker in glomerulopathic patients. Kidney Int 28:830-838, 1985 8. Rolin HA III, Hall PM, Wei R: Inaccuracy of estimated creatinine clearance for prediction of iothalamate glomerular filtration rate. Am J Kidney Dis 4:48-54, 1984 9. Cockcroft DW, Gault MH: Prediction of creatinine clearance from serum creatinine. Nephron 16:31-41, 1976 10. 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 130:461470, 1999 11. Levey AS, Greene T, Kusek JW, Beck GJ: A simplified equation to predict glomerular filtration rate from serum creatinine. J Am Soc Nephrol 11:155A, 2000 (abstr) 12. Wells M, Lipman J: Measurements of glomerular filtration in the intensive care unit are only a rough guide to renal function. S Afr J Surg 35:20-23, 1997 13. Robert S, Zarowitz BJ: Is there a reliable index of glomerular filtration rate in critically ill patients? DICP 25:169-178, 1991 14. Chalasani N, Clark WS, Wilcox CM: Blood urea nitrogen to creatinine concentration in gastrointestinal bleeding: A reappraisal. Am J Gastroenterol 92:1796-1799, 1997 15. Dossetor JB: Creatininemia versus uremia. The relative significance of blood urea nitrogen and serum creatinine concentrations in azotemia. Ann Intern Med 65:1287-1299, 1966 16. Papadakis MA, Arieff AI: Unpredictability of clinical evaluation of renal function in cirrhosis. Prospective study. Am J Med 82:945-952, 1987 17. Sherman DS, Fish DN, Teitelbaum I: Assessing renal function in cirrhotic patients: Problems and pitfalls. Am J Kidney Dis 41:269-278, 2003 18. Bajaj G, Alexander SR, Browne R, Sakarcan A, Seikaly MG: 125Iodine-iothalamate clearance in children. A

251

simple method to measure glomerular filtration. Pediatr Nephrol 10:25-28, 1996 19. Perrone RD, Steinman TI, Beck GJ, et al: Utility of radioisotopic filtration markers in chronic renal insufficiency: Simultaneous comparison of 125I-iothalamate, 169YbDTPA, 99mTc-DTPA, and inulin. The Modification of Diet in Renal Disease Study. Am J Kidney Dis 16:224-235, 1990 20. 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 39:920-929, 2002 21. Groth T, Tengstrom B: The usefulness of 125I-sodium iothalamate as a GFR-indicator in single intravenous injection tests. Ups J Med Sci 83:53-63, 1978 22. Frennby B, Sterner G, Almen T, et al: The use of iohexol clearance to determine GFR in patients with severe chronic renal failure—A comparison between different clearance techniques. Clin Nephrol 43:35-46, 1995 23. Jagenburg R, Attman PO, Aurell M, Bucht H: Determination of glomerular filtration rate in advanced renal insufficiency. Scand J Urol Nephrol 12:133-137, 1978 24. Sapirstein LA, Vidt DG, Mandel MJ, Hanusek G: Volumes of distribution and clearances of intravenously injected creatinine in the dog. Am J Physiol 181:330-336, 1955 25. National Kidney Foundation: K/DOQI Clinical Practice Guidelines for Chronic Kidney Disease: Evaluation, classification, and stratification. Am J Kidney Dis 39:S1266, 2002 (suppl 1) 26. 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 13:2140-2144, 2002 27. Hallan S, Asberg 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 44:84-93, 2004 28. Mariat C, Alamartine E, Barthelemy JC, et al: Assessing renal graft function in clinical trials: Can tests predicting glomerular filtration rate substitute for a reference method? Kidney Int 65:289-297, 2004 29. Vervoort G, Willems HL, Wetzels JF: Assessment of glomerular filtration rate in healthy subjects and normoalbuminuric diabetic patients: Validity of a new (MDRD) prediction equation. Nephrol Dial Transplant 17:1909-1913, 2002 30. 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 16:459-466, 2005 31. 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 141:929-937, 2004 32. Gaspari F, Mosconi L, Vigano G, et al: Measurement of GFR with a single intravenous injection of nonradioactive iothalamate. Kidney Int 41:1081-1084, 1992 33. Levey AS, Berg RL, Gassman JJ, Hall PM, Walker WG: Creatinine filtration, secretion and excretion during progressive renal disease. Modification of Diet in Renal

252

Disease (MDRD) Study Group. Kidney Int Suppl 27:S73-S80, 1989 34. Bleiler RE, Schedl HP: Creatinine excretion: Variability and relationships to diet and body size. J Lab Clin Med 59:945-955, 1962 35. Gaspari F, Ferrari S, Stucchi N, et al: Performance of different prediction equations for estimating renal function in kidney transplantation. Am J Transplant 4:1826-1835, 2004

POGGIO ET AL

36. Gonwa TA, Jennings L, Mai ML, et al: Estimation of glomerular filtration rates before and after orthotopic liver transplantation: Evaluation of current equations. Liver Transpl 10:301-309, 2004 37. Zuo L, Ma YC, Zhou YH, et al: Application of GFR-estimating equations in Chinese patients with chronic kidney disease. Am J Kidney Dis 45:463-472, 2005