Measurement of Glomerular Filtration Rate
A measured GFR remains the single best index of kidney function. As renal mass declines in the presence of age-related loss of nephrons or disease states such as hypertension or diabetes, there is a progressive decline in GFR. The rate of decline in GFR can be used to predict the time to onset of stage 5 CKD, as well as the risk of complications of CKD. Accurate measurement of GFR in clinical practice is a critical variable for the individualization of the dosage regimens of renally excreted medications so that one can maximize their therapeutic efficacy and avoid potential toxicity.
The GFR is expressed as the volume of plasma filtered across the glomerulus per unit time, based on total renal blood flow and capillary hemodynamics. The normal values for GFR are 127 ± 20 mL/min/1.73 m2 (1.22 ± 0.19 mL/s/m2) and 118 ± 20 mL/min/ 1.73 m2 (1.14 ± 0.19 mL/s/m2) in healthy men and women, respectively. These measured values closely approximate what one would predict if the normal renal blood flow were approximately 1.0 L/min/1.73 m2 (0.01 mL/s/m2), plasma volume was 60% of blood volume, and filtration fraction across the glomerulus was 20%: then the normal GFR would be expected to be approximately 120 mL/min/1.73 m2 (1.16 mL/s/m2).
Optimal measurement of GFR involves determining the renal clearance of a substance that is freely filtered without additional clearance because of tubular secretion or reduction as the result of reabsorption. Additionally, the substance should not be susceptible to metabolism within renal tissues and should not alter renal function. Given these conditions, the measured GFR is equivalent to the renal clearance of the solute marker:
GFR = renal CL = (Ae)/AUC0–t
where renal CL is renal clearance of the marker, Ae is the amount of marker excreted in the urine in a specified period of time, t, and AUC0-t is the area under the plasma-concentration-versus-time curve of the marker.
Under steady-state conditions, for example during a continuous infusion of the marker, the expression simplifies to
GFR = renal CL = (Ae)/[Css) × t]
where Css is the steady-state plasma concentration of the marker achieved during continuous infusion. The continuous infusion method can also be employed without urine collection, where plasma clearance is calculated as CL = infusion rate/Css. This method is dependent on the attainment of steady-state plasma concentrations and accurate measurement of infusate concentrations. Plasma clearance can also be determined following a single-dose IV injection with the collection of multiple blood samples to estimate area under the curve (AUC0-∞). Here, clearance is calculated as CL = dose/AUC. These plasma clearance methods commonly yield clearance values 10% to 15% higher than GFR measured by urine collection methods.75,76
Several markers have been used for the measurement of GFR and include both exogenous and endogenous compounds. Those administered as exogenous agents, such as inulin, sinistrin, iothalamate, iohexol, and radioisotopes, require specialized administration techniques and detection methods for the quantification of concentrations in serum and urine, but generally provide an accurate measure of GFR. Methods that employ endogenous compounds, such as creatinine or cystatin C, require less technical expertise, but produce results with greater variability. The GFR marker of choice depends on the purpose and cost of the test, ranging from $2,000 per vial for radioactive for 125I-iothalamate (Glofil-125, QOL Medical) to $6 per vial for nonradiolabeled iothalamate (Conray-60, Mallincrodt) or iohexol (Omnipaque-300, GE Medical) (eTable 18-4).
eTable 18-4 Sensitivity and Clinical Utility of Renal Function Tests |Favorite Table|Download (.pdf)
eTable 18-4 Sensitivity and Clinical Utility of Renal Function Tests
|Nonisotopic contrast agents||+++||++||$$$|
Inulin and Sinistrin Clearance
Inulin is a large fructose polysaccharide (5,200 Da), obtained from the Jerusalem artichoke, dahlia, and chicory plants. It is not bound to plasma proteins, is freely filtered at the glomerulus, is not secreted or reabsorbed, and is not metabolized by the kidney. The volume of distribution of inulin approximates extracellular volume, or 20% of ideal body weight. Because it is eliminated by glomerular filtration, its elimination half-life is dependent on renal function and is approximately 1.3 hours in subjects with normal renal function. Measurement of plasma and urine inulin concentrations can be performed using high-performance liquid chromatography.77 Sinistrin, another polyfructosan, has similar characteristics to inulin; it is filtered at the glomerulus and not secreted or reabsorbed to any significant extent. It is a naturally occurring substance derived from the root of the North African vegetable red squill, Urginea maritime, which has a much higher degree of water solubility than inulin. Assay methods for sinistrin have been described using enzymatic procedures, as well as high-performance liquid chromatography with electrochemical detection.78 Alternatives have been sought for inulin as a marker for GFR because of the problems of availability, high cost, sample preparation, and assay variability.
Iothalamate is an iodine-containing radiocontrast agent that is available in both radiolabeled (125I) and nonradiolabeled forms. This agent is handled in a manner similar to that of inulin; it is freely filtered at the glomerulus and does not undergo substantial tubular secretion or reabsorption. The nonradiolabeled form is most widely used to measure GFR in ambulatory and research settings, and can safely be administered by IV bolus, continuous infusion, or subcutaneous injection.76 Plasma and urine iothalamate concentrations can be measured using high-performance liquid chromatography.79,80 Plasma iothalamate clearance methods that do not require urine collections have been shown to be highly correlated with iothalamate renal clearance, making them particularly well-suited for longitudinal evaluations of renal function.76,81 These plasma clearance methods require two-compartment modeling approaches because accuracy is dependent on duration of sampling. For example, Agarwal et al.81 demonstrated that short sampling intervals can overestimate GFR, particularly in patients with severely reduced GFR. In individuals with GFR >30 mL/min/1.73 m2 (>0.29 mL/s/m2), a 2-hour sampling strategy yielded GFR values that were 54% higher compared with 10-hour sampling, whereas the 5-hour sampling was 17% higher. In individuals with GFR <30 mL/min/1.73 m2 (<0.29 mL/s/m2), the 5-hour GFR was 36% higher and 2-hour GFR was 126% higher than the 10-hour measurement. The authors proposed a 5 to 7 hour sampling time period with eight plasma samples to be the most appropriate and feasible approach for most GFR evaluations.
Iohexol, a nonionic, low osmolar, iodinated contrast agent, has also been used for the determination of GFR. It is eliminated almost entirely by glomerular filtration, and plasma and renal clearance values are similar to observations with other marker agents: Strong correlations of 0.90 or greater and significant relationships such with iothalamate have been reported.82–84 These data support iohexol as a suitable alternative marker for the measurement of GFR. A reported advantage of this agent is that a limited number of plasma samples (as few as two collected at 120 and 300 min after injection) can be used to quantify iohexol plasma clearance.85 For patients with a reduced GFR more time must be allotted—more than 24 hours if the estimated GFR is less than 20 mL/min (0.33 mL/s).
The GFR has also been quantified using radiolabeled markers, such as 125I-iothalamate (614 Da, radioactive half-life of 60 days), 99mTc-diethylenetriamine pentaacetic acid (99mTc-DPTA; 393 Da, radioactive half-life of 6.03 hours), and 51Cr-ethylenediaminetetraacetic acid (51Cr-EDTA; 292 Da, radioactive half-life of 27 days).86 These relatively small molecules are minimally bound to plasma proteins and do not undergo tubular secretion or reabsorption to any significant degree. 125I-iothalamate and 99mTc-DPTA are used in the United States, whereas 51Cr-EDTA is used extensively in Europe. The use of radiolabeled markers allows one to determine the individual contribution of each kidney to total renal function.87 Various protocols exist for the administration of these markers and subsequent measurement of GFR using either plasma or renal clearance calculation methods. The nonrenal clearance of these agents appears to be low (3 to 8 mL/min [0.05 to 0.13 mL/s]), suggesting that plasma clearance is an acceptable technique except in patients with severe renal insufficiency (GFR <30 mL/min [<0.50 mL/s]). Indeed, highly significant correlations between renal clearance among radiolabeled markers has been demonstrated.88 Although total radioactive exposure to patients is usually minimal, use of one of these agents does require compliance with radiation safety committees and appropriate biohazard waste disposal.
Measured Creatinine Clearance
Although the measured (24-hour) CLcr has been used as an approximation of GFR for decades, it has limited clinical utility for a multiplicity of reasons. Short-duration witnessed measured CLcr correlates well with iothalamate clearance performed using the single-injection technique. In a multicenter study89 of 136 patients with type 1 diabetic nephropathy, the correlations of simultaneous measured CLcr, and 24-hour CLcr (compared to CLiothalamate) were 0.81 and 0.49, respectively, indicating increased variability with the 24-hour clearance determination. In a selected group of 110 patients, measurement of a 4-hour CLcr during water diuresis provided the best estimate of the GFR as determined by the CLiothalamate. Furthermore, the ratio of CLcr to CLiothalamate did not appear to increase as the GFR decreased. These data suggest that a short collection period with a water diuresis may be the best CLcr method for estimation of GFR.
A limitation of using creatinine as a filtration marker is that it undergoes tubular secretion. Tubular secretion augments the filtered creatinine by approximately 10% in subjects with normal kidney function. If the nonspecific Jaffe reaction is used, which overestimates the serum creatinine concentration by approximately 10% because of the noncreatinine chromogens, then the measurement of CLcr is a very good measure of GFR in patients with normal kidney function. Tubular secretion, however, increases to as much as 100% in patients with renal insufficiency.30 The result is that measured CLcr markedly overestimates GFR. For example, Bauer et al.30 reported that the CLcr-to-CLinulin ratio in subjects with mild impairment was 1.20; for those with moderate impairment, it was 1.87; and in those with severe impairment, it was 2.32. Thus, a measured CLcr is a poor indicator of GFR in patients with moderate to severe renal insufficiency, that is, stages 3 to 5 CKD.
Because cimetidine blocks the tubular secretion of creatinine the potential role of several oral cimetidine regimens to improve the accuracy and precision of measured CLcr as an indicator of GFR has been evaluated. The CLcr-to-CLDPTA ratio declined from 1.33 with placebo to 1.07 when 400 mg of cimetidine was administered four times a day for 2 days prior to and during the clearance determination.90 Similar results were observed when a single 800-mg dose of cimetidine was given 1 hour prior to the simultaneous determination of CLcr and CLiothalamate; the ratio of CLcr to CLiothalamate was reduced from a mean of 1.53 to 1.12.91 Thus a single oral dose of 800 mg of cimetidine should provide adequate blockade of creatinine secretion to improve the accuracy of a CLcr measurement as an estimate GFR in patients with stages 3 to 5 CKD.
To minimize the impact of diurnal variations in serum creatinine concentration on CLcr, the test is usually performed over a 24-hour period with the plasma creatinine obtained in the morning, as long as the patient has stable kidney function. Collection of urine remains a limiting factor in the 24-hour CLcr because of incomplete collections, and interconversion between creatinine and creatine that can occur if the urine is not maintained at a pH <6.
Because of the invasive nature and technical difficulties of directly measuring GFR in clinical settings, many equations for estimating GFR have been proposed over the past 10 years (eTable 18-5). A series of related GFR estimating equations (eGFR) have been developed for the primary purpose of identifying and classifying CKD in many patient populations.92–99 The initial equation was derived from multiple regression analysis of data obtained from the 1,628 patients enrolled in the Modification of Diet in Renal Disease Study (MDRD) where GFR was measured using the renal clearance of 125I-iothalamate methodology. The first regression model yielded a six-variable Modification of Diet in Renal Disease Study (MDRD6) equation.92 This equation (r2 = 0.902) provided a more precise estimate of the measured GFR in the MDRD population than measured CLcr (r2 = 0.866) or CLcr estimated by the Cockcroft-Gault equation (r2 = 0.842). A newer four-variable version of the original MDRD equation (MDRD4), based on plasma creatinine, age, sex, and race, was shown to provide a similar estimate of GFR results compared to the six-variable equation predecessor93 and has been subsequently updated according to nationally recognized serum creatinine concentration (IDMS) assay results.94 The MDRD4–IDMS equation is recommended by the NKF and the National Kidney Disease Education Program (NKDEP) for calculating the estimated GFR (eGFR) in patients with a history of CKD risk factors and a GFR <60 mL/min/1.73 m2 (<0.58 mL/s/m2). The performance of the MDRD4 equations has been assessed in a variety of patient populations with GFR < 60 mL/min/1.73 m2 (<0.58 mL/s/m2), including those with diabetic nephropathy and renal transplant patients. However, the MDRD4 equations have been shown to be significantly biased in healthy subjects, diabetic patients with normal GFR (88 to 182 mL/min/1.73 m2 [0.85 to 1.75 mL/s/m2]), and healthy potential kidney donors.95 In patients with GFR values >60 mL/min/1.73 m2 (>0.58 mL/s/m2), the values provided by the MDRD4 equation have been shown to be highly variable and less accurate than other traditional estimation methods, perhaps in part due to the fact that the original study population consisted of only patients with low GFR values (<60 mL/min/1.73 m2 [<0.58 mL/s/m2]). The study also included few Asians, elderly patients, and those with diabetes, or ill, hospitalized patients.96,97 A recent study conducted by the FDA compared the eGFR estimated by the MDRD4 equation to the CLcr estimated by the Cockcroft-Gault equation in 973 subjects enrolled in pharmacokinetic studies conducted for new chemical entities submitted to the FDA from 1998 to 2010.98 The MDRD4 equation eGFR consistently overestimated the CLcr calculated by the Cockcroft-Gault method. The FDA investigators concluded that “For patients with advanced age, low weight, and modestly elevated serum creatinine concentration values, further work is needed before the MDRD equations can replace the Cockcroft-Gault equation for dose adjustment in approved product information labeling.”
eTable 18-5 Equations for the Estimation of GFR in Adults with Stable Renal Function |Favorite Table|Download (.pdf)
eTable 18-5 Equations for the Estimation of GFR in Adults with Stable Renal Function
|Levey et al.92 (MDRD6)||GFR = 170 × (Scr)–0.999 × [age]–0.176 × [0.762 if patient is female] × [1.180 if patient is black] × [BUN]–0.170 × [Alb]0.318|
|Levey et al.93 (MDRD4)||GFR = 186 × (Scr)–1.154 × (age)–0.203 × (0.742 if patient is female) × (1.210 if patient is black)|
|Levey et al.94 (MDRD4-IDMS)||GFR = 175 × (Scr)–1.154 × (age)–0.203 × (0.742 if patient is female) × (1.210 if patient is black)|
|Levey et al.100 (CKD-EPI)||GFR = 141 × min(Scr/κ, 1)α × max(Scr/κ, 1)-1.209 × 0.993age × 1.018 [if female] × 1.159 [if black]|
|Rule et al.102 (MCQ)|
|Larsson et al.109||GFR = 77.24 × (CysC in mg/L)–1.2623|
|Macdonald et al.108|
|CKD-EPI Equation 8112|
eGFR = 127.7 × (CysC in mg/L)−1.17 × (age in years)−0.13 × 0.91 (if female) × 1.06 (if black)
*eGFR = 127.7 × (−0.105 + 1.13 × standardized SCysC)−1.17 × age−0.13 × (0.91 if female) × (1.06 if black)
|CKD-EPI Equation 9112|
eGFR (mL/min/1.73 m2) = 76.7 × (CysC in mg/L)−1.19
*eGFR (mL/min/1.73 m2) = 76.7 × (−0.105 + 1.13 × CysC in mg/L)−1.19
|CKD-EPI Equation 10112|
eGFR (mL/min/1.73 m2) = 177.6 × (Scr in mg/dL)−0.65 × (CysC in mg/L)−0.57 × (age in years)−0.20 × 0.82 [if female] × 1.11 [if black]
*eGFR (mL/min/1.73 m2) = 177.6 × (Scr in mg/dL)−0.65 × (−0.105 + 1.13 × CysC in mg/L)−0.57 × (age in years)−0.20 × 0.82 [if female] ×1.11 [if black]
A single GFR equation may not be best suited for all populations, and choice of equation has been shown to impact CKD prevalence estimates.99 This has led to a revitalized interest in the development of new equations to estimate GFR. The newest equations to be proposed for the estimation of GFR have been derived from wider CKD populations than the MDRD study, and include the Chronic Kidney Disease Epidemiology Study (CKD-EPI)100 and the Mayo Clinic Quadratic Equation102 or MCQ. The CKD-EPI equation was developed from pooled study data involving 5,500 patients (including the original MDRD population), with mean GFR values of 68 ± 40 mL/min/1.73 m2 (0.65 ± 0.39 mL/s/m2) (range 2 to 190 mL/min/1.73 m2 [0.02 to 1.83 mL/s/m2]). It has been reported that the CKD-EPI equation is less biased (2.5 vs. 5.5 mL/min/ 1.73 m2 [0.024 vs. 0.053 mL/s/m2]) but similarly imprecise compared to MDRD4.101
Sidebar: Clinical Controversy…
Some practitioners are advocating the use of the four-variable Modification of Diet in Renal Disease Study equation (MDRD4) in patients without chronic kidney disease (CKD), although it appears to have a weaker correlation with glomerular filtration rate (GFR) than the Cockcroft-Gault equation. Recent evidence suggests that the MDRD4 equation should be reserved for patients with a GFR <60 mL/min/1.73 m2 (<1.0 mL/s/m2). The use of newer equations, such as the CKD-EPI, has improved accuracy in patients with GFR >60 mL/min/1.73 m2 (<1.0 mL/s/m2); however, further studies incorporating additional biomarkers such as cystatin C are underway.
Chronic Kidney Disease–EPI Equation
The CKD-EPI study equation was recently compared to the MDRD equation using pooled data from patients enrolled in research or clinical outcomes studies, where GFR was measured by any exogenous tracer.100 The findings of the study indicated that the bias of CKD-EPI equation was 61% to 75% lower than the MDRD equation for patients with eGFR of 60 to 119 mL/min/1.73 m2 (0.58 to 1.15 mL/s/m2). Based on these findings, the CKD-EPI equation is most appropriate for estimating GFR in individuals with eGFR values >60 mL/min/1.73 m2 (>0.58 mL/s/m2). Currently the Australasian Creatinine Consensus Working Group is one of the first to recommend that clinical laboratories switch from the MDRD4 to CKD-EPI for routine automated reporting.103 If one’s clinical lab does not automatically calculate eGFR using the CKD-EPI, it becomes a bit of a challenge since the equation requires a more complex algorithm than the MDRD equation.
Most labs now use IDMS–calibrated creatinine assays and with the recommendation that a revised “CKD-EPI–IDMS” equation be used based on the estimated 5% lower creatinine values that result from the using the IDMS–calibrated assay are near fully implemented.4,104
Limitations of the pooled analysis approach used to develop the MDRD and CKD-EPI equations include the use of different GFR markers between studies (iothalamate, 51Cr-EDTA, 99mTc-DTPA), different methods of administration of the GFR markers (subcutaneous and IV), and different clearance calculations (renal clearance vs. plasma disappearance). These limitations may partly explain the reduced accuracy observed with the MDRD4 equation at GFR values >60 mL/min/1.73 m2 (>0.58 mL/s/m2). Additionally, a recent inspection of the MDRD GFR study data showed that large intrasubject variability in GFR measures was a likely contributor to the inaccuracy of the gold standard method ([125I] iothalamate urinary clearance) that was used to create the MDRD equation.105
Numerous studies have consistently reported that the MDRD4 eGFR equation overestimates CLcr estimated by the Cockcroft-Gault equation in more than 40,000 patients. For example, Wargo et al.106 evaluated 409 patients with CKD admitted to a tertiary care clinic. The CKD-EPI equation significantly overestimated the Cockcroft-Gault equation (39.9 mL/min vs. 34.8 mL/min [0.67 mL/s vs. 0.58 mL/s], respectively; p < 0.001), with 95% of cases ranging from –5.1 mL/min to +15.3 mL/min (–0.09 mL/s to +0.26 mL/s). In the largest retrospective study comparing the Cockcroft-Gault and MDRDIND conducted to date, Melloni et al.107 reported that use of MDRDIND eGFR resulted in a failure to make manufacturer-recommended dose reductions for enoxaparin and the glycoprotein IIb/IIIa inhibitor (GPI) eptifibatide in up to 50% of their cohort of more than 49,000 patients. The excessive MDRDIND-derived doses were also correlated with major bleeding episodes (odds ratio 1.57 [95% CI 1.35–1.84]). Thus, the MDRDIND equation should not be used in lieu of Cockcroft-Gault for renal dose adjustments for these drugs. Thus, it is important to understand that eGFR equations such as MDRD4 and CKD-EPI were developed for the purpose of identifying and stratifying CKD based on large multicenter epidemiologic studies. Extension of their use for individualized drug dosing has not been fully evaluated and automatic substitution of MDRD4 or CKD-EPI in place of estimated or measured CLcr for drug dose calculations should be avoided (eFig. 18-3).
Algorithm for estimating kidney function using eGFR and/or eCLcr approaches. Creatinine clearance is converted from mL/min to mL/s by multiplying by 0.0167. (BMI, body mass index; C-G, Cockcroft-Gault; CLcr, creatinine clearance; eGFR, estmated glomerular filtration rate.)
Mayo Clinic Quadratic Equation
The MCQ equation was developed by Rule et al.102 to estimate GFR from a population of CKD (n = 320) and healthy individuals (n = 580), using 1/Scr, 1/Scr2, age, and sex as the covariates. Using the NHANES database, Snyder et al.99 reported that MCQ GFR values resulted in 28% fewer patients being categorized as CKD stage 3 or 4 when compared to MDRD4.
Combined Serum Creatinine and Cystatin C Equations
Addition of serum cystatin as a covariate in equations to estimate GFR has been employed as a means to improve creatinine-based estimations that historically were limited to lean body mass (LM), age, sex, race, and serum creatinine concentration (Scr).108–113
Serum Cystatin C Equations
A significant limitation of serum cystatin C as a renal biomarker is the influence of body mass on serum concentrations. Studies by MacDonald et al.108 and Vupputuri et al.111 reported that fat-free mass is a significant covariate in GFR determination using cystatin C. Using GFR measured as plasma inulin clearance, lean body mass accounted for at least 16.3% of the variance in GFR values obtained using cystatin C (p < 0.001). When using a cystatin C-based estimate of GFR, which incorporates the CysC, age, race, and sex, a higher prevalence of CKD was reported in obese patients when compared to the MDRD4 equation.111 In a recent retrospective analysis of over 1,000 elderly individuals (mean age 85 y) enrolled in the Cardiovascular Health Study, GFR was estimated using the CKD-EPI and CKD-EPI-CysC equations.113 In this population, all-cause mortality rates were significantly different between equations. The CKD-EPI equation yielded a U-shaped association, whereas the CysC equation yielded a linear relationship at eGFR values <60 mL/min/1.73 m2 (<1.0 mL/s/m2), suggesting that CysC does not accurately predict mortality risk in patients with low serum creatinine concentration, reduced muscle mass and malnutrition. This further highlights the important relationship between muscle mass and serum CysC concentrations, and cautions against the use of CysC-based equations in the elderly.
Use of CysC-based equations to estimate GFR may also be problematic in patients with autoimmune disorders. It has been reported that cystatin C levels are associated with disease activity in patients with rheumatoid arthritis, which may be due to the presence of C-reactive protein or analytical interference from drugs such as glucocorticoids and infliximab. A single dose of infliximab (3 mg/kg) resulted in a 14% (p < 0.05) increase in CysC, whereas no change was observed in serum creatinine concentration or GFR values.114
In 167 patients with rheumatoid arthritis and 91 controls, CysC levels were significantly higher in RA than controls (15%; p < 0.001]. Furthermore, CysC levels correlated positively with erythrocyte sedimentation rate (p < 0.001), and C-reactive protein (p = 0.01), indicating that CysC levels are associated with severity of disease.115 Thus, CysC-based equations should be used with caution in the obese, elderly and those with inflammatory conditions such as rheumatoid arthritis.
Additional research on the utility of these newer eGFR equations in diverse ethnic groups has resulted in modifications or “correction factors,” such as those for Japanese116 and Chinese117 populations.
A variety of online resources provide GFR calculators such as the NKDEP website,118 which provides aGFR calculator based on the MDRD4–IDMS equation. The NKDEP also recommends that laboratories report eGFR values greater than or equal to 60 as “>60 mL/min/1.73 m2 (>1.0 mL/s/m2), not as an exact number,” due to inaccuracies of the MDRD4 equation at higher levels of GFR. It should be noted that one must verify that a given equation is appropriate for the institutional creatinine reporting method.
Sidebar: Clinical Controversy…
Drug-dose individualization is often required in patients with chronic kidney disease (CKD). Approved drug labeling typically includes dose-adjustment information based on the patient’s estimated creatinine clearance (CLcr) using the Cockcroft-Gault method. Automated estimated glomerular filtration rate (eGFR) reporting by many hospital laboratories has now raised the question if the four-variable Modification of Diet in Renal Disease Study equation (MDRD4)/ CKD-EPI or any other GFR estimation equations should be used as a guide for drug-dose adjustments.
Estimation of Creatinine Clearance
Many equations describing the mathematical relationships between various patient factors and measured CLcr, the most widely recognized surrogate for GFR in clinical settings. Most equations incorporate factors such as age, gender, weight, and serum creatinine concentration, without the need for urine collection. The most widely used of these estimators is the Cockcroft-Gault equation,119 which identified age and body mass as factors, which significantly contribute to the estimate of CLcr. This relationship was based on observations from 249 male patients with stable kidney function in whom the creatinine production rates were estimated. Estimated creatinine clearance (eCLcr), using the Cockcroft-Gault equation, is one of the methods endorsed by the FDA for stratifying patients in drug development pharmacokinetic studies, and has been reported most often in FDA-approved package inserts for new drug entities since the 1990s.7,96
One of the key considerations with the use of this equation is whether or not a modified weight index should replace actual body weight. Several modified weight indices have been proposed and this remains a controversial issue. For obese individuals, defined as those with a body mass index (BMI) greater than or equal to 30 kg/m2 but less than 40 kg/m2, it is generally recommended that total or actual body weight be used. This is based on a recent analysis by the FDA indicating that the Cockcroft-Gault equation had <10% bias in nearly 600 normal weight, overweight, and obese individuals enrolled in drug pharmacokinetic studies.120 In morbidly obese individuals (BMI ≥ 40 kg/m2, obesity class III) an alternate measure of body weight such as lean body weight was shown to significantly reduce bias in the Cockcroft-Gault equation, where lean body weight (LBW) is calculated as:
LBW (kg, males) (9270 × weight) / (6680 + 216 × BMI)
LBW (kg, females) (9270 × weight) / (8780 + 244 × BMI)
And BMI is calculated as:
BMI (kg/m2) = weight (kg) / height (m)2
Regardless of the approach used to estimate renal function in obese patients, it is imperative that drug therapy outcomes be monitored closely in this population.
Luke et al.121 evaluated the ability of the Cockcroft-Gault method and four other methods to determine eCLcr, with inulin clearance being considered the standard measure of GFR. The simultaneously determined inulin and measured creatinine clearances correlated best, r2 = 0.85, and the measured CLcr overestimated CLinulin by approximately 15% due to tubular secretion of creatinine. Of the five estimated creatinine clearances, the ones calculated by the Cockcroft-Gault and Mawer et al. methods122 correlated the best with GFR. Other methods, such as Jelliffe123 and Hull et al.,124 consistently underestimated the measured CLcr (eTable 18-6).
eTable 18-6 Equations for the Estimation of Creatinine Clearance in Adults with Stable Renal
Function |Favorite Table|Download (.pdf)
eTable 18-6 Equations for the Estimation of Creatinine Clearance in Adults with Stable Renal
|Cockroft and Gault119|
- Men: CLcr = (140 − age)ABW/(Scr × 72)
- Women: CLcr × 0.85
- Men: CLcr = (100/Scr) − 12
- Women: CLcr = (80/Scr) − 7
- Men: CLcr = 98 − [0.8 (age − 20)]/Scr
- Women: CLcr ×0.9
|Mawer et al.122|
- Men: IBW [29.3 − (0.203 × age)] [1 − (0.03 × Scr)]/(14.4 × Scr)
- Women: IBW [25.3 − (0.175 × age)] [1 − (0.03 × Scr)]/(14.4 × Scr)
|Hull et al.124|
- Men: CLcr = [(145 − age)/Scr] − 3
- Women: Clcr × 0.85
Patients undergoing screening for participation in the African American Study of Kidney Disease (AASK) were evaluated for kidney function based on eCLcr, simultaneous 125I-iothalamate and measured 24-hour CLcr.125 The simultaneous measured CLcr provided the best estimate of GFR. The Cockcroft-Gault method was the preferred method for eGFR, based on performance and ease of use. This method was noted to underestimate the GFR by 9%, perhaps because of the increased excretion rate of creatinine by black patients.126
Administration of cimetidine has also resulted in improved performance of the Cockcroft-Gault equation as means to calculate eGFR. Ixkes et al.127 gave patients three 800-mg doses of cimetidine in 24 hours, and measured creatinine plasma levels from 3 to 7 hours following the final dose. During this 4-hour period, the CLiothalamate was determined as the measure of GFR. The Cockcroft-Gault calculations were performed with the plasma creatinine measurement 3 hours after the last dose of cimetidine. The ratio of the Cockcroft-Gault eCLcr-to-CLiothalamate decreased from 1.28 ± 0.21 to 0.98 ± 0.11 in the presence of cimetidine. This cimetidine dosing schedule also improved the accuracy of Cockcroft-Gault eCLcr relative to GFR in renal transplant patients with GFR values ranging from 20 to 80 mL/min/1.73 m2 (0.19 to 0.77 mL/s/m2).128,129
Sidebar: Clinical Controversy…
The use of ideal or lean body weight for estimating creatinine clearance (CLcr) in obese patients is controversial. Some clinicians recommend using a weight adjustment in patients that are greater than 30% above ideal body weight (IBW), such as using an adjusted body weight. However, recent studies by the FDA indicate that use of IBW should be avoided. Further research evaluating weight-based adjustments and drug pharmacokinetic outcomes is needed.
The estimation of CLcr or GFR is particularly problematic in patients with preexisting liver disease and renal impairment. Lower-than-expected serum creatinine concentration values may result from reduced muscle mass, protein-poor diet, diminished hepatic synthesis of creatine (a precursor of creatinine), and fluid overload can lead to significant overestimation of CLcr. Orlando et al.130 evaluated 10 healthy subjects, 10 patients with mild liver disease, and 10 with severe liver disease, and observed a measured CLcr-to-CLinulin ratio of 1.05, 1.03, and 1.04 for each group, respectively. When the CLcr of patients with severe liver disease was estimated using the Cockcroft-Gault equation, the resultant ratio (eCLcr-to-CLinulin) was 1.23. Lam et al.131 likewise noted an overestimation by Cockcroft-Gault of the measured CLcr in patients with severe disease, by 40% to 100%.
Studies of renal function in patients with severe hepatic disease confirm the earlier observations of Hull et al.124 and Caregaro et al.132 who reported that measured CLcr overestimated GFR by up to 50% in hepatic patients with a GFR of 56 ± 19 mL/min/1.73 m2 (0.54 ± 0.18 mL/s/m2) because of increased tubular secretion of creatinine. The effect of cimetidine administration on measured CLcr was evaluated in a small study by Sansoe et al.133 In 12 patients with compensated cirrhosis, serum creatinine concentration values increased from 0.68 ± 0.11 to 0.94 ± 0.14 mg/dL (69 ± 10 μmol/L to 83 ± 12 μmol/L) during coadministration of cimetidine (1,000 mg given as 400 mg × 1 then 200 mg every 3 hours) during a 9-hour clearance period. The measured CLcr declined from 138 ± 20 prior to cimetidine administration to 89 ± 13 mL/min (2.30 ± 0.33 to 1.49 ± 0.22 mL/s), with no change in measured GFR.
In cirrhotic patients being evaluated for liver transplant (GFR 58 ± 5.1 mL/min/1.73 m2 [0.56 ± 0.049 mL/s/m2]) the eGFR by the MDRD4 and MDRD6, and the eCLcr by Cockcroft-Gault significantly overestimated measured GFR by 30% to 50%, and all three were considered unacceptable methods for renal function assessment in liver transplant patients.134,135
Recently, Gerhart et al.136 evaluated the performance of the CKD-EPI and MDRD4–IDMS equations in patients with liver disease following transplantation (group 1, n = 59) and those with cirrhosis (group 2; n = 44). When compared to measured GFR, both equations yielded positively biased estimates of GFR (4 to 9 mL/min/1.73 m2 [ 0.04 to 0.09 mL/s/m2]) in transplanted patients. However, in patients with hepatic cirrhosis, both equations were more significantly positively biased (40 to 42 mL/min/1.73 m2 [0.39 to 0.40 mL/s/m2]), with low precision (21 to 26 mL/min/1.73 m2 [0.20 to 0.25 mL/s/m2]) and low accuracy with only 7% of patients having eGFR values within 30% of the measured GFR.
Thus, renal function assessment in patients with hepatic disease should be performed by measuring glomerular filtration, and estimation equations for GFR or CLcr should be avoided.
Other Special Populations
Davis and Chandler137 confirmed the accuracy of the Cockcroft-Gault eCLcr method in trauma patients with stable kidney function, and Thakur et al.40 demonstrated its acceptable performance in 42 paraplegic patients. Renal transplant recipients are frequently monitored for renal function, as numerous complications may occur during the life of the allograft. Ruiz-Esteban et al.138 evaluated the bias and precision of the MDRD4 and CKD-EPI relative to Cockcroft-Gault in 153 postrenal transplant patients. Here, the mean bias for MDRD4 was –10.6 ± 12.7 compared to –9.8 ± 11.3 mL/min/1.73 m2 (–0.10 ± 0.12 compared to –0.09 ± 0.11 mL/s/m2) for CKD-EPI (p = 0.006), with the CKD-EPI having a higher percentage of patients within 30% of the Cockcroft-Gault value than the MDRD equation (86.9% vs. 81.7%), p <.001). Huang et al.139 reported the inability of several CLcr equations to predict renal function in hospitalized patients with advanced human immunodeficiency virus (HIV) disease. All of the prediction methods overestimated the measured 24-hour CLcr. The reasons for the poor predictability of these methods are unclear, although 24-hour collection methods result in increased variability, often because of inadequate collection of urine.
Renal function assessment during pregnancy is usually performed using a 24-hour CLcr determination, and estimation equations have been shown to perform poorly particularly in the preeclampsia population. For example, Alper et al.140 recently evaluated the Cockcroft-Gault, MDRD4 and CKD-EPI equations in 543 women, aged 16 to 49 years, with preeclampsia after the 20th week of gestation. When compared to 24-hour measured CLcr (mean 133 ± 43 mL/min [2.22 ± 0.72 mL/s]), the Cockcroft-Gault equation was positively biased (36 ± 2 mL/min [0.60 ± 0.03 mL/s]) whereas both the MDRD4 and CKD-EPI were negatively biased (–20 ± 1.5 mL/min [–0.33 ± 0.025 mL/s]). Thus, kidney function estimating equations should not be used during pregnancy.
Patients with unstable kidney function or AKI present a unique situation because serum creatinine concentration values are changing, and steady state cannot be assumed, which is one of the assumptions of all the above-mentioned eCLcr methods. It is now widely accepted that a change in the serum creatinine concentration of more than 50% over a period of 7 days, or an increase in serum creatinine concentration by at least 0.3 mg/dL (≥27 μmol/L) over a 24- to 48-hour period indicates the presence of AKI.141 Methods to measure GFR in this population, such as 125I-iothalamate clearance, are cumbersome and costly especially in the acute care setting. Although several equations have been proposed the calculate eCLcr in AKI patients or those with rapidly progressive renal disease,142–144 a rigorous evaluation of the accuracy and precision of each of these proposed methods is lacking and none of them is currently recommended for clinical use. Use of semiquantitative approaches is preferred for the purpose of estimating severity of disease using RIFLE and AKIN criteria (see Chap. 28 for more details).
Use of abbreviated CLcr measurements (<24 h) may be valuable for detecting early evidence of AKI in critically ill patients. Pickering et al.145 recently reported that 4-hour CLcr measurements were significantly better at predicting AKI events than serum creatinine concentration alone in 484 patients. However, the lack of true measured GFR in these patients prevented assessment of the accuracy and precision of the CLcr values. Hoste et al.146 used 1-hour CLcr measurements to evaluate the Cockcroft-Gault and MDRD equations in critically ill patients within 1 week of ICU admission. Both equations were poorly correlated with CLcr (R < 0.5), and were similarly imprecise with Bland-Altman 95% confidence intervals ranging from –77 to 64 mL/min (–1.29 to 1.07 mL/s) for Cockcroft-Gault and –77 to 58 mL/min/1.73 m2 (–0.74 to 0.56 mL/s/m2) for MDRD. Changes in serum creatinine concentration have also been shown to be more sensitive than cystatin C in early AKI.
It is thus, ultimately, most important to recognize that renal function in patients with AKI is generally markedly lower than one would estimate using steady-state methods, and dose adjustments should be made if necessary to avoid drug toxicity (see Chaps. 28 and 33).
Kidney Function in Children
Kidney function in the neonate is difficult to assess because of difficulty in urine and blood collection, the frequent presence of a non–steady-state serum creatinine concentration, and apparent disparity between development of glomerular and tubular function. Preterm infants demonstrate significantly reduced GFR prior to 34 weeks, which rapidly increases and becomes similar to term infants within the first week of life.147 Evaluation of GFR in preterm infants on day 3 of life, using an inulin infusion, failed to identify a relationship between patient weight and GFR. Gestational age, which ranged from 23.4 to 36.9 weeks (mean: 30.2 weeks), however, correlated with both GFR and reciprocal of serum creatinine concentration (Scr). The inulin clearance increased from 0.67 to 0.85 mL/min (0.011 to 0.014 mL/s) in those with gestational age <28 weeks versus those of 32 to 37 weeks of age, while Scr decreased from 1.05 to 0.73 mg/dL (93 to 65 μmol/L), respectively. Creatinine was measured using a specific enzymatic method to avoid interference from bilirubin or drugs.148 Creatinine clearance has also been evaluated in infants younger than 1 week of age, and values of 17.8 mL/min/ 1.73 m2 (0.171 mL/s/m2) on day 1 increased to 36.4 mL/min/1.73 m2 (0.351 mL/s/m2) by day 6.149 In light of these rapid changes in GFR, estimation of GFR is not recommended for infants younger than 1 week of age. Kidney function expressed as GFR standardized to body surface area increases with age and stabilizes at approximately 1 year. In older children, GFR is best assessed using standard measurement techniques for GFR. Subcutaneous administration of 125I-iothalamate has been effectively used to measure GFR in children ranging in age from 1 to 20 years.150 The original equation to estimate GFR as described by Schwartz et al.151 is dependent on the child’s age and length:
GFR = [length (cm) × k] / (Scr in mg/dL)
where k is defined by age group: infant (1 to 52 weeks) = 0.45; child (1 to 13 years) = 0.55; adolescent male = 0.7; and adolescent female = 0.55. Serum creatinine in mmol/L can be converted to mg/dL by multiplication using 0.0113 as the conversion factor.
A newer version of the Schwartz equation152 was developed from a population of 349 children (age 1 to 19 y) with mild-to-moderate CKD enrolled in the Chronic Kidney Disease in Children (CKiD) study. This simple equation is commonly referred to as the Schwartz “Bedside” formula:
GFR = 0.41 * [length (cm)/Scr in mg/dL]
Lee et al.153 recently reported that this new Bedside Schwartz equation performed better than the original Schwartz equation for patients with mild-moderate CKD, but was less accurate in patients with mild CKD. Thus, the appropriate use of the Bedside Schwartz equation and accuracy in subpopulations of CKD is yet to be fully determined.
Equations derived from adult populations have also been evaluated in pediatric patients. Peirrat et al.154 compared the MDRD, Schwartz, and Cockcroft-Gault equations in children 3 to 19 years of age. In children <12 years, the Schwartz and MDRD equations were significantly more biased than Cockcroft-Gault, and Cockcroft-Gault provided the best prediction of GFR in children >12 years. The results of these investigations suggest that further studies will be needed to clarify the value of any of these predictive methods in children. Most recently, an equation for GFR based on beta-trace protein was shown to yield similar values of GFR compared to the Schwartz equation in 387 pediatric patients (age 10.7 ± 7.1 y) who underwent a 99mTC-DTPA GFR scan.70 The most recent GFR equation evaluated in pediatrics includes use of cystatin C, BUN, serum creatinine concentration (in mg/dL) and demographic data derived from over 600 pediatric patients enrolled the CKiD study155:
eGFR (mL/min/1.73 m2) = 39.8 × [ht(m)/Scr]0.456 × (1.8/cystatin C)0.418
× (30/BUN)0.079 × 1.076male × [ht(m)/1.4]0.179
This equation had the lowest root-mean square error (0.147), highest R2 (0.863) and highest frequency of values within 30% of iohexol-measured GFR (91.3%) when compared to seven other GFR equations.
Kidney Function in the Elderly
Cross-sectional studies have historically shown that GFR declines as a function of age.156,157 The largest prospective study conducted in healthy elderly individuals is the Baltimore Longitudinal Study on Aging (BLSA).157 In an initial analysis of 254 BLSA participants without kidney disease, it was reported that measured CLcr decreases at the rate of approximately 0.75 mL/min/1.73 m2/y (0.0072 mL/s/m2/y) beginning at the fourth decade of life. These subjects were then evaluated prospectively for up to 23 years. Interestingly, approximately one-third of the subjects showed no change in renal function from their baseline value, and a small number showed an increased clearance. These changes may be a result of normal physiologic changes or of subclinical insults to the kidneys initiating the events leading to chronic progressive loss of renal function. Fliser et al.158 studied renal functional reserve in healthy young (23 to 32 years) and elderly (61 to 82 years) volunteers using an amino acid infusion technique. Measured GFR increased by 16% in young and 17% in elderly subjects following the infusion. Renal functional reserve thus appears to be maintained in healthy elderly individuals.
Interpretation of the serum creatinine concentration alone is difficult in the elderly patient primarily because of the decreased muscle mass and resultant lower production rate of creatinine. Thus, the serum creatinine concentration often remains within the normal range despite a reduction in the number of functional nephrons. As renal function declines, the kidneys excrete a larger fraction of creatinine. This perpetuates the “normal” serum creatinine concentration. Recent recommendations such as the adoption of standardized creatinine assays by clinical laboratories and reporting of serum creatinine concentration values to two decimal places will likely improve the accuracy of renal function estimation in the elderly population.34
The Cockcroft-Gault formula119 continues to provide a valid estimate of the CLcr of elderly patients. Smythe et al.159 estimated CLcr in 23 patients >60 years of age using seven different methods, and compared the results to a measured 24-hour CLcr determination. Estimations were performed with the actual serum creatinine concentration and also with the serum creatinine concentration rounded up to 1.0 mg/dL (88 μmol/L) if the actual value was <1.0 mg/dL (<88 μmol/L). Rounding the serum creatinine concentration to 1.0 mg/dL (88 μmol/L) resulted in a significantly lower eCLcr (–28.8 mL/min [–0.481 mL/s]) compared to the unadjusted serum creatinine concentration (+2.3 mL/min [+0.038 mL/s]). In patients >60 years of age with serum creatinine concentration <1.0 mg/dL (<88 μmol/L), rounding the serum creatinine concentration value up to 1.0 mg/dL (85 μmol/L) resulted in dose estimates for gentamicin that were significantly lower (–90 ± 67 mg/day) than doses calculated based on the actual serum creatinine concentration value.160 Until further data are available, fixing or rounding serum creatinine concentration to an arbitrary value in elderly patients is not recommended.
An alternative to the estimation of GFR or a 24-hour measured CLcr is a 4-hour measured clearance performed during water diuresis. This approach correlated with the inulin clearance as well as with an observed inpatient 24-hour measured CLcr.87 However, one must be aware of the potential risk of hyponatremia in the geriatric patient who is unable to tolerate an oral water load, as well as the need for complete bladder emptying to ensure accurate results. O’Connell et al.161 assessed the accuracy of 2- and 8-hour urine collections compared with 24-hour CLcr determinations in 45 hospitalized patients >65 years old with indwelling urethral catheters. The 8-hour timed urine collection for CLcr showed minimal bias (2.2 mL/min [0.037 mL/s]) as compared with the 24-hour value, whereas the 2-hour determination was both positively biased (11 mL/min [0.18 mL/s]) and less precise (25 mL/min [0.42 mL/s]).
Sidebar: Clinical Controversy…
Estimation of creatinine clearance in elderly with low serum creatinine values is controversial. Some clinicians advocate correction of serum creatinine to 1.0 mg/dL (88 μmol/L) to account for reduced muscle mass. This practice should be avoided, and the impact of this correction factor on glomerular filtration rate (GFR) estimates using the four-variable Modification of Diet in Renal Disease Study equation (MDRD4) or other equations has yet to be evaluated in this population.
Impact on Drug Dosing Recommendations
The automated reporting of eGFR in the clinical setting has led some practitioners to consider substituting eGFR in place of eCLcr for renal dose adjustments. Concerns with this approach, particularly in the elderly, include the uncertainty of applying eGFR values to eCLcr-based dosage adjustment algorithms in package inserts, which may result in dosing errors and toxicity especially for drugs with narrow therapeutic indices.104,162–169 Roberts et al.162 reported that the MDRD eGFR values overestimated gentamicin clearance by 29% (p < 0.001), whereas the Cockcroft-Gault yielded only 10% overestimation (p < 0.01), and MDRD overestimated renal function as age increased. Retrospective studies in more than 1,200 patients with renal disease have shown that overestimation of renal function using the MDRD4 with or without IDMS creatinine equation results in up to 30% to 60% higher doses for digoxin, amantadine and various antimicrobials compared to doses calculated using eCLcr.104,164–169 A recent study by Stevens et al.170 reported that use of a back-corrected value of eGFR, based on a calculated body surface area (BSA), yields dose calculations that are similar to those calculated using a measured GFR. This back-correction approach has not been validated and calculating BSA in clinical settings is inconvenient and unlikely to occur. A more detailed discussion of application of renal function estimates and renal dosing approaches is provided in Chapter 33.