Abstract
Summary
Background and objectives
A specific method is required for estimating glomerular filtration rate GFR in hospitalized patients. Our objective was to validate the Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation and four cystatin C (CysC)–based equations in this setting.
Design, setting, participants, & measurements
This was an epidemiologic, cross-sectional study in a random sample of hospitalized patients (n = 3114). We studied the accuracy of the CKD-EPI and four CysC-based equations—based on (1) CysC alone or (2) adjusted by gender; (3) age, gender, and race; and (4) age, gender, race, and creatinine, respectively—compared with GFR measured by iohexol clearance (mGFR). Clinical, biochemical, and nutritional data were also collected.
Results
The CysC equation 3 significantly overestimated the GFR (bias of 7.4 ml/min per 1.73 m2). Most of the error in creatinine-based equations was attributable to calculated muscle mass, which depended on patient's nutritional status. In patients without malnutrition or reduced body surface area, the CKD-EPI equation adequately estimated GFR. Equations based on CysC gave more precise mGFR estimates when malnutrition, extensive reduction of body surface area, or loss of muscle mass were present (biases of 1 and 1.3 ml/min per 1.73 m2 for equations 2 and 4, respectively, versus 5.9 ml/min per 1.73 m2 for CKD-EPI).
Conclusions
These results suggest that the use of equations based on CysC and gender, or CysC, age, gender, and race, is more appropriate in hospitalized patients to estimate GFR, since these equations are much less dependent on patient's nutritional status or muscle mass than the CKD-EPI equation.
Introduction
Hospitalized patients undergo intensive medical care that often includes examinations using radiologic contrast agents and/or potentially nephrotoxic drugs. A precise and reliable method for measuring the renal function is essential in this setting, characterized by an extremely heterogeneous population. The methods usually employed in hospitalized patients are serum creatinine and endogenous creatinine clearance. However, they both have important limitations (1–3). The production rate and tubular secretion of creatinine may be substantially altered in hospitalized patients due to associated comorbidities and/or drug therapy. The endogenous creatinine clearance tends to overestimate the glomerular filtration rate GFR as renal insufficiency progresses and is subject to error depending on samples collection method (1–3).
Estimation of GFR using equations based on serum creatinine has been suggested as an alternative. The most popular examples are Cockroft–Gault (CG) (4) and modification of diet in renal disease (MDRD) (5) equations. Use of isotope dilution mass spectrometry (IDMS) traceable creatinine in these equations results in a more accurate estimated GFR (eGFR) (6). However, these were developed in patients with stable renal dysfunction, and the Kidney Disease Outcomes Quality Initiative (KDOQI) guidelines recommend not using them in clinical circumstances particularly prevalent in hospitalized patients, such as extremes of age and body size, severe malnutrition, or obesity (7). A recent study (8) in hospitalized patients with renal dysfunction concluded that both CG and MDRD significantly overestimate GFR in this setting.
Recently, the Chronic Kidney Disease Epidemiology Collaboration has developed a new equation (CKD-EPI) based on serum creatinine, age, gender, and race (9). In comparison to the IDMS-MDRD equation, CKD-EPI has greater precision and reliability, especially for glomerular filtration rates >60 ml/min per 1.73 m2. However, to date, this equation has not been evaluated in hospitalized patients.
Cystatin C (CysC) is a 13-kD protein freely filtered by the glomerulus and completely reabsorbed by renal tubular epithelial cells (10,11). For these reasons, several CysC-based equations have been proposed (12,13). Although there are some factors that influence CysC concentrations (14–16), CysC-based equations are less dependent on muscle mass (17), nutritional status (18), and age (16) than creatinine, so they may be more precise in hospitalized patients.
The purpose of the study was to evaluate the performance (accuracy, bias, and precision) of the CKD-EPI equation and four CysC-based equations compared with measured GFR (mGFR) in hospitalized patients with stable renal function.
Materials and Methods
Study Subjects and Sampling
The study included patients who were admitted to the Vall d'Hebron General Hospital between November 2008 and October 2009. Exclusion criteria were patients from critical units, transplant recipients, history of allergy to iodized contrast agents, hemodialysis, patients whose anthropometric parameters could not be measured, and unstable renal function (≥25% increase or decrease in creatinine since admission).
Patients were selected by simple random sampling. All patients were required to grant their informed consent in writing, and procedures were performed according to the principles of the Declaration of Helsinki.
GFR Measurement and Estimation
mGFR by iohexol clearance (19–21) was determined during the first three days of admission using fasting plasma sampling and HPLC (20,21). The same samples were used to measure creatinine levels by the Roche Lab “compensated” IDMS-traceable method (Hitachi Modular P-800 Roche Diagnostics, Germany) and CysC by particle enhanced immunonephelometry on a BN II system (Dade Behring Marburg GMBH, USA). eGFR was estimated from creatinine values by the CKD-EPI equation (9):
eGFR = 141 × min (SCr/k, 1)a × max (Scr/k, 1)−1.209 × 0.993age × 1.018 (if female) × 1.159 (if black)
where Scr is serum creatinine, k is 0.7 for females and 0.9 for males, a is −0.329 for females and −0.411 for males, min indicates the minimum of Scr/k or 1, and max indicates the maximum of Scr/k or 1.
The three equations described by Stevens et al. (12) and Grubb's equation (13) were used to estimate CysC-based GFR:
eGFR = 76.7 × CysC−1.19
eGFR = 127.7 × CysC−1.17 × age−0.13 × (0.91 if female) × (1.06 if black)
eGFR = 177.6 × SCr−0.65 × CysC−0.57 × age−0.20 × (0.82 if female) × (1.11 if black)
eGFR = 87.62 × CysC−1.693 × (0.94 if female)
Variables and Definitions
We examined all medical records and interviewed patients to record age, gender, ethnic group, limb amputations, diagnoses of muscle diseases with amyotrophy, chronic liver disease or cirrhosis, thyroid conditions, diabetes, or known chronic kidney disease (CKD). We also performed biochemical determinations and an anthropometric evaluation (weight, height, arm and leg girth, tricipital fold, thigh-fold and girth, leg-fold and girth, and body mass index[BMI]). Obesity was defined as BMI >27 kg/m2 (22). Elmore's equation (23) was used to diagnose malnutrition. Total-body muscle mass was calculated by Lee's equation (24). The Child–Pugh classification was used to determine liver disease (25). Patients were considered to suffer from systemic inflammatory response syndrome (SIRS) if they presented at least one of the following: temperature >38°C or <36°C; heart rate >90 beats/min; respiratory rate >20 breaths/min; or leukocyte count >12,000 or <4000 cells/mm3 (26). The diagnosis of inflammatory abdominal disease included acute cholecystitis or acute pancreatitis of any etiology.
The criteria mentioned in the KDOQI guidelines (7) and in the consensus of the Spanish Society of Nephrology (27) were used to define subgroups who did not meet the conditions for using the MDRD equation (see Figure 1).
Figure 1.
Study population and analyzed subgroups.
Statistical Analysis
To determine whether the study sample represented all hospitalized patients, we compared the main characteristics of the cohort with those from the patients who were admitted during the same period but not included in the study. Categorical variables were compared by the Pearson chi-squared or Fisher exact test as required. Continuous variables were compared by t test or the Mann–Whitney U-test.
Univariate and stepwise multivariate linear regression models were developed to determine the independent predictors of calculated muscle mass, creatinine, and CysC levels (adjusting by mGFR in the latter two cases).
To validate eGFR equations, bias, precision, accuracy, and Pearson correlation coefficients with respect to mGFR by iohexol (considered as the standard reference and expressed in standardized values per 1.73m2) were calculated. Bias was defined as the mean of individual differences between eGFR and mGFR. Precision was defined as the SD of bias. Accuracy was evaluated by the percentages of patients with eGFR within 30% and 50% of mGFR. T tests for paired samples were used to assess differences between eGFR and mGFR values. Validation of eGFR equations was performed in the overall sample and in each subgroup of patients who did not meet the criteria for using creatinine-based equations.
Bland–Altman plots were made to analyze whether differences between eGFR and mGFR were related to the magnitude of GFR. In addition, to analyze the variables statistically associated to the differences from mGFR, we performed four multivariate regression models in which the standardized residuals for each method were taken as dependent variables, and the following were taken as potential predictor variables: age, gender, malnutrition, chronic liver disease, inflammation, and thyroid disease.
A P < 0.05 was considered statistically significant. MedCalc (MedCalc Software, Broekstraat 52, 9030 Mariakerke, Belgium) and SPSS 15.0 (SPSS, Inc., Headquarters, Chicago, IL) software were used for the analyses.
Results
Study Population
Figure 1 shows the selection process and the analyzed populations. A total of 3114 patients with varied diagnoses, admitted in 20 different services, fulfilled all selection criteria. Their mean age (SD) was 62.7 (19.1) years, and 55% were men. Median (Q1, Q3) hospital stay was 8 days (5, 13), and mortality rate was 3.85% (120 patients died). Main comorbidities were hypertension (31.1%), diabetes (14.9%), chronic obstructive pulmonary disease (10.9%), myocardiopathy (10.0%), and chronic kidney disease (6.7%). No significant differences were found in demographic and clinical characteristics between included (n = 3114) and nonincluded patients (n = 22,840; data not shown).
Table 1 summarizes anthropometric and biochemical variables in both genders and in the total sample. The prevalence of obesity was 45% (14% women 46% men) and the prevalence of BMI >35 kg/m2 was 1.8%. Prevalence of malnutrition and SIRS were 49.9% and 28.0%, respectively. A poor correlation was observed between nutritional status according to BMI and Elmore's equation (data not shown).
Table 1.
Description of the main biochemical and anthropometric characteristics in the overall study population and in gender subgroups
| Males (n = 1713) Mean ± SD or n (%) | Females (n = 1401) Mean ± SD or n (%) | Total (n = 3114) Mean ± SD or n (%) | |
|---|---|---|---|
| Hemoglobin, g/dl | 10.8 ± 1.9 | 10.7 ± 2.0 | 10.9 ± 2.1 |
| Creatinine, mg/dl | 1.01 ± 0.38 | 0.91 ± 0.42 | 0.96 ± 0.51 |
| Cystatin C, mg/dl | 0.91 ± 0.46 | 0.89 ± 0.38 | 0.94 ± 0.36 |
| Albumin, g/dl | 3.45 ± 0.56 | 3.54 ± 0.71 | 3.38 ± 0.6 |
| mGFR by iohexol, ml/min per 1.73 m2 | 90.3 ± 36.3 | 84.2 ± 27.7a | 87.5 ± 32.5 |
| <30 | 26 (1.5) | 16 (1.1) | 42 (1.4) |
| 31–60 | 161 (9.4) | 98 (6.9) | 259 (8.3) |
| 61–90 | 709 (41.4) | 657 (46.8) | 1366 (43.9) |
| 91–120 | 674 (39.3) | 517 (36.8) | 1191 (38.2) |
| >120 | 142 (8.3) | 114 (8.1) | 256 (8.2) |
| Transferrin, mg/dl | 192.4 ± 87.5 | 193.8 ± 101.3 | 193.0 ± 92.0 |
| Lymphocytes, /mm3 | 1496 ± 1180 | 1478 ± 993 | 1483 ± 976 |
| Weight, kg | 71.5 ± 18.2 | 65.8 ± 14.3b | 68.9 ± 15.8 |
| Height, cm | 173.3 ± 9.7 | 162.5 ± 8.3b | 167.1 ± 8.6 |
| BMI, kg/m2 | 26.1 ± 6.2 | 27.1 ± 7.3a | 26.8 ± 7.6 |
| Calculated muscle mass, kg/1.73 m2 | 25.4 ± 8.1 | 21.8 ± 7.9b | 23.7 ± 6.9 |
BMI, body mass index; mGFR, measured GFR.
P < 0.05 versus males.
P < 0.01 versus males.
Independent Predictors of Muscle Mass
The best model for predicting calculated muscle mass (R2 = 0.42, P < 0.001) included age (β = −0.08, SE = 0.02, P = 0.002), gender (β = 2.18 for male versus female, SE = 0.38, P = 0.001), BMI (β = 0.02, SE = 0.06, P = 0.004), and malnutrition (β = −0.33, SE = 0.25, P = 0.02). Alone, neither serum albumin (R2 = 0.18) nor BMI (R2 = 0.24) were good predictors of calculated muscle mass variability.
Independent Predictors of Serum Creatinine and CysC Levels
Table 2 summarizes the independent determinants of creatinine and CysC levels. The main determinants of creatinine were mGFR, total muscle mass, malnutrition, and age. In simple linear regression analysis, calculated muscle mass predicted 40.8% of the variability in creatinine (R2 = 0.48, P = 0.0002).
Table 2.
Independent predictors of serum levels of creatinine and cystatin C
| (A) Creatinine | |||
|---|---|---|---|
| Variable | Change (mg/dl) | 95% Confidence Interval | P-value |
| mGFR (per 5 ml/min per 1.73 m2) | −11.90 | −10.63 to −13.17 | <0.001 |
| Age (per 10 years) | −0.08 | −0.04 to −0.12 | 0.001 |
| Calculated muscle mass (per 5 kg) | 0.90 | 0.89 to 0.91 | 0.001 |
| Constant | 15.33 | 2.90 to 27.75 | |
| Analysis of variance: F, 21.97; P < 0.0001; R2, 0.63. | |||
|---|---|---|---|
| (B) Cystatin C | |||
| Variable | Change (mg/dl) | 95% Confidence Interval | P-value |
| mGFR (per 5 ml/min per 1.73 m2) | −0.20 | −1.96 to 1.56 | <0.001 |
| Age (per 10 years) | −0.05 | −0.40 to 0.30 | 0.002 |
| BMI (per 1 kg/m2) | −0.09 | −0.08 to −0.10 | 0.04 |
| Gender (male vs female) | 0.01 | 0.002 to 0.018 | 0.04 |
| SIRS (presence vs absence) | 0.05 | −0.20 to 0.30 | 0.02 |
| Diabetes (presence vs absence) | 0.02 | −0.17 to 0.21 | 0.04 |
| Constant | 12.50 | 3.78 to 21.22 | |
Analysis of variance: F, 23.25; P < 0.0001; R 2, 0.52. BMI, body mass index; mGFR, measured GFR; SIRS, systemic inflammatory response syndrome.
In the case of CysC, most of the variability could be attributed to mGFR. Other variables such as BMI, gender, and inflammation were statistically significant but added little further value to the model.
Measured GFR
Table 1 shows the distribution of GFR categories, as measured by iohexol clearance. Table 3 shows the prevalence of mGFR <60 ml/min per 1.73 m2 by age and gender quartiles. The prevalence of mGFR <60 ml/min per 1.73 m2 increased in proportion to age (P = 0.0003) and was significantly higher in men for each age quartile (Table 3).
Table 3.
Prevalence of mGFR by iohexol <60 ml/min per 1.73 m2 by age and gender quartiles in the study population
| Age Quartiles |
Total | |||||
|---|---|---|---|---|---|---|
| <50 | 51–65 | 66–80 | >80 | |||
| Femalesa | mGFR <60 ml/min per 1.73 m2, n (%)* | 14 (4.0) | 24 (6.9) | 36 (10.2) | 40 (11.4) | 114 (8.1) |
| Malesb | mGFR <60 ml/min per 1.73 m2, n (%)* | 28 (6.5) | 39 (9.15) | 55 (12.9) | 65 (15.2) | 187 (10.9) |
| Totalc | mGFR <60 ml/min per 1.73 m2, n (%)* | 42 (5.4) | 63 (8.1) | 91 (11.7) | 105 (13.5) | 301 (9.7) |
mGFR, measured glomerular filtration rate.
Number and percentage of people in each particular age-gender subgroup who have mGFR values <60 ml/min per 1.73 m2.
Differences between age categories in females, χ2 = 19.05, P < 0.0001.
Differences between age categories in males, χ2 = 17.09, P < 0.0001.
Differences between males and females, χ2 = 6.56, P = 0.01.
Estimated GFR
Table 4 shows performance of the five equations for estimating GFR in each of the analyzed populations. In the overall sample, the CysC 3 equation overestimated mGFR. The bias in CKD-EPI and equations 1, 2, and 4 was significantly lower and gave more precise results.
Table 4.
Glomerular filtration rate estimations according to the five analyzed equations and validation parameters using mGFR by iohexol as reference
| Equation | Mean ± SD ml/min per 1.73 m2 | Bias | Precision | P-valuea | Accuracy within 30% (%)b | Accuracy within 50% (%)b | R |
|---|---|---|---|---|---|---|---|
| All (n = 3114) | |||||||
| mGFR | 87.5 ± 32.5 | ||||||
| CKD-EPI | 89.8 ± 38.9 | 1.8 | 23.1 | 0.124 | 72 | 88 | 0.84 |
| CysC 1 | 90 ± 39.1 | 2.5 | 14.1 | 0.167 | 80 | 87 | 0.90 |
| CysC 2 | 89.4 ± 40.2 | 1.9 | 11.3 | 0.063 | 82 | 89 | 0.91 |
| CysC 3 (+Cr) | 94.9 ± 37.4 | 7.4 | 15.7 | 0.006 | 66 | 78 | 0.79 |
| CysC 4 | 89.2 ± 41.2 | 1.7 | 18.4 | 0.413 | 84 | 89 | 0.93 |
| Malnourished patients (n = 1555) | |||||||
| mGFR | 76.3 ± 26.1 | ||||||
| CKD-EPI | 81.2 ± 33.4 | 5.9 | 12.6 | 0.035 | 70 | 83 | 0.77 |
| CysC 1 | 76.9 ± 28.7 | 0.6 | 13.2 | 0.154 | 78 | 86 | 0.88 |
| CysC 2 | 77.3 ± 22.5 | 1.0 | 11.7 | 0.432 | 85 | 92 | 0.91 |
| CysC 3 (+Cr) | 84.1 ± 29.8 | 7.8 | 13.1 | 0.001 | 58 | 64 | 0.73 |
| CysC 4 | 77.6 ± 31.2 | 1.3 | 14.3 | 0.165 | 86 | 91 | 0.92 |
| All excluding patients with malnutrition or amputations >25% body surface area (n = 1673) | |||||||
| mGFR | 86.6 ± 31.3 | ||||||
| CKD-EPI | 85.7 ± 28.2 | −0.6 | 11.4 | 0.368 | 84 | 98 | 0.84 |
| CysC1 | 87.9 ± 27.4 | 1.3 | 13.6 | 0.198 | 82 | 98 | 0.85 |
| CysC 2 | 86.5 ± 28.6 | −0.1 | 14.8 | 0.547 | 84 | 98 | 0.84 |
| CysC 3 (+Cr) | 85.3 ± 30.1 | −1.3 | 10.7 | 0.483 | 77 | 97 | 0.86 |
| CysC 4 | 87.8 ± 28.6 | 1.2 | 8.6 | 0.471 | 85 | 98 | 0.89 |
| mGFR >60 ml/min per 1.73 m2 (n = 2813) | |||||||
| mGFR | 98 ± 28.1 | ||||||
| CKD-EPI | 96.5 ± 30.3 | −1.5 | 23.9 | 0.224 | 82 | 91 | 0.88 |
| CysC 1 | 97.8 ± 36.3 | −0.2 | 12.8 | 0.541 | 81 | 91 | 0.87 |
| CysC 2 | 96.9 ± 28.7 | −1.1 | 14.1 | 0.341 | 84 | 92 | 0.84 |
| CysC 3 (+Cr) | 103.2 ± 31.9 | 5.2 | 17.6 | 0.003 | 69 | 81 | 0.71 |
| CysC 4 | 98.1 ± 29.6 | 0.1 | 11.8 | 0.598 | 89 | 97 | 0.89 |
| GFR >60 ml/min per 1.73 m2 excluding patients with moderate or severe malnutrition or amputation >25% body surface area (n = 2248) | |||||||
| mGFR | 104 ± 36.1 | ||||||
| CKD-EPI | 104 ± 39.2 | 0.4 | 22.3 | 0.312 | 86 | 95 | 0.91 |
| CysC 1 | 101 ± 36.3 | −3.0 | 11.8 | 0.131 | 82 | 92 | 0.91 |
| CysC 2 | 106 ± 28.7 | 2.0 | 12.3 | 0.410 | 83 | 94 | 0.93 |
| CysC 3 (+Cr) | 103.2 ± 31.9 | −0.8 | 11.8 | 0.517 | 81 | 90 | 0.92 |
| CysC 4 | 107 ± 29.6 | 3.0 | 7.8 | 0.298 | 90 | 97 | 0.92 |
| Age >70 years (n = 1307) | |||||||
| mGFR | 69.6 ± 16.4 | ||||||
| CKD-EPI | 70.3 ± 19.8 | 2.7 | 19.2 | 0.356 | 78 | 89 | 0.88 |
| CysC 1 | 65.6 ± 29.2 | −4.0 | 25.2 | 0.004 | 69 | 74 | 0.82 |
| CysC 2 | 68.7 ± 30 | −0.9 | 11.1 | 0.132 | 79 | 93 | 0.87 |
| CysC 3 (+Cr) | 74.6 ± 33.5 | 5.1 | 12.3 | 0.046 | 61 | 68 | 0.84 |
| CysC 4 | 70.2 ± 24.6 | 0.6 | 8.7 | 0.554 | 76 | 91 | 0.97 |
| Age >70 years excluding patients with moderate or severe malnutrition or amputation >25% body surface area (n = 850) | |||||||
| mGFR | 68.1 ± 21.5 | ||||||
| CKD-EPI | 68.6 ± 27.2 | −0.5 | 24.1 | 0.221 | 84 | 95 | 0.82 |
| CysC 1 | 61.3 ± 19.2 | −6.8 | 34.2 | 0.0001 | 82 | 94 | 0.83 |
| CysC 2 | 70.2 ± 24.3 | 2.1 | 21.3 | 0.214 | 79 | 95 | 0.85 |
| CysC 3 (+Cr) | 70.6 ± 26.8 | 2.5 | 18.3 | 0.056 | 73 | 92 | 0.86 |
| CysC 4 | 68.8 ± 19.4 | 0.7 | 13.2 | 0.221 | 74 | 91 | 0.93 |
| Child Class C Liver Disease or cirrhosis of the liver (n = 63) | |||||||
| mGFR | 89.1 ± 41.3 | ||||||
| CKD-EPI | 95.3 ± 29.5 | 4.2 | 28.6 | 0.037 | 77 | 81 | 0.78 |
| CysC 1 | 89.3 ± 21.2 | 0.1 | 27.2 | 0.342 | 80 | 91 | 0.82 |
| CysC 2 | 88.4 ± 25.6 | −0.7 | 31.3 | 0.214 | 79 | 92 | 0.83 |
| CysC 3 (+Cr) | 97.8 ± 27.3 | 8.7 | 36.5 | 0.012 | 76 | 84 | 0.74 |
| CysC 4 | 88.8 ± 29.5 | −0.3 | 23.6 | 0.363 | 79 | 91 | 0.88 |
| BMI >35 kg/m2 (n = 56) | |||||||
| mGFR | 108.9 ± 46.8 | ||||||
| CKD-EPI | 86.5 ± 24.3 | −22.4 | 31.8 | 0.000 | 60 | 71 | 0.66 |
| CysC 1 | 88.9 ± 31.6 | −20.0 | 38.1 | 0.000 | 58 | 66 | 0.63 |
| CysC 2 | 87.9 ± 32.4 | −21.0 | 29.7 | 0.000 | 54 | 68 | 0.65 |
| CysC 3 (+Cr) | 85.8 ± 28.9 | 23.1 | 33.2 | 0.000 | 56 | 70 | 0.68 |
| CysC 4 | 87.3 ± 36.2 | −21.6 | 40.2 | 0.000 | 55 | 67 | 0.62 |
Results are presented for the general hospitalized patients and for each subgroup of patients in which, according to Kidney Disease Outcomes Quality Initiative guidelines, creatinine-based equations cannot be used to estimate GFR. BMI, body mass index; CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration equation; CysC, cystatin C; mGFR, measured GFR; R, Pearson correlation coefficient; eGFR, estimated GFR.
P-value for t tests between eGFR and mGFR values.
% patients with eGFR within 30% or 50% of mGFR.
Equations tested:
CKD-EPI: eGFR = 141 × min (SCr/k, 1)a × max (Scr/k, 1)−1.209 × 0.993age × 1.018 [if female] × 1.159 [if black]
[Scr is serum creatinine, k is 0.7 for females and 0.9 for males, a is −0.329 for females and −0.411 for males, min indicates the minimum of Scr/k or 1, and max indicates the maximum of Scr/k or 1.]
CysC 1: eGFR = 76.7 × CysC−1.19
CysC 2: eGFR = 127.7 × CysC−1.17 × age−0.13 × (0.91 if female) × (1.06 if black)
CysC 3: eGFR = 177.6 × SCr−0.65 × CysC−0.57 × age−0.20 × (0.82 if female) × (1.11 if black)
CysC 4: eGFR = 87.62 × CysC−1.693 × (0.94 if female)
Figure 2 shows the error in the eGFR depending on the mean of eGFR and mGFR. The accuracy of the CKD-EPI equation in the overall sample decreased with increasing GFR. CysC equations 2 and 4 gave more precise results in GFR range of 15 to 100 ml/min per 1.73 m2, but overestimated mGFR for values >100 ml/min per 1.73 m2.
Figure 2.
Distribution of differences for each analyzed equation with respect to mean of measured GFR (mGFR) and estimated GFR (eGFR) values. The Y-axes display differences between “estimated” minus “measured” GFR values, to obtain positive values in case of overestimation.
The main predictors of error are depicted in Table 5. Neither the presence of SIRS at admission nor diabetes mellitus or thyroid disease showed significance.
Table 5.
Independent predictors of difference with respect to measured GFR for each equation estimating GFR in the study population
| (A) CKD-EPI Equation | |||
|---|---|---|---|
| Variable | Difference (ml/min per 1.73 m2) | 95% Confidence Interval | P-value |
| BMI >35 kg/m2 (versus ≤35 kg/m2) | 0.26 | 0.24 to 0.28 | 0.002 |
| Calculated muscle mass (per 5 kg) | −0.30 | −0.40 to −0.20 | <0.0001 |
| Measured GFR >90 ml/min per 1.73 m2 (versus ≤90 ml/min per 1.73 m2) | −0.43 | −0.53 to −0.33 | 0.004 |
| Constant | 29.2 | 18.75 to 39.65 | |
| Analysis of variance: F, 38.1; P < 0.0001; R2, 0.34. | |||
|---|---|---|---|
| (B) Cystatin C Equation 1 | |||
| Variable | Difference (ml/min per 1.73 m2) | 95% Confidence Interval | P-value |
| Age (per 10 years) | −3.5 | −4.87 to −2.13 | 0.001 |
| BMI (per 1 kg/m2) | −0.02 | −0.08 to 0.04 | 0.001 |
| Measured GFR >100 ml/min per 1.73 m2 (versus ≤100 ml/min per 1.73 m2) | −0.51 | −1.35 to 0.33 | 0.001 |
| Constant | 34.69 | 21.54 to 47.84 | |
| Analysis of variance: F, 53.94; P < 0.001; R2, 0.19. | |||
|---|---|---|---|
| (C) Cystatin C Equation 2 | |||
| Variable | Difference (ml/min per 1.73 m2) | 95% Confidence Interval | P-value |
| BMI (per 1 kg/m2) | −0.06 | −0.24 to 0.12 | 0.001 |
| Measured GFR >100 ml/min per 1.73 m2 (versus ≤100 ml/min per 1.73 m2) | −0.43 | −0.86 to 0.001 | 0.001 |
| Constant | 16.15 | 0.55 to 31.75 | |
| Analysis of the variance: F, 16.77; P < 0.001; R2, 0.16. | |||
|---|---|---|---|
| (D) Cystatin C Equation 3 | |||
| Variable | Difference (ml/min per 1.73 m2) | 95% Confidence Interval | P-value |
| Calculated muscle mass (per 5 kg) | −0.95 | −1.06 to −0.84 | 0.001 |
| BMI (per 1 kg/m2) | −0.07 | −0.10 to −0.03 | 0.01 |
| Measured GFR >100 ml/min per 1.73 m2 (versus ≤100 ml/min per 1.73 m2) | −0.18 | −0.24 to −0.12 | 0.001 |
| Constant | 17.33 | −0.21 to 34.87 | |
| Analysis of variance: F, 13.27; P < 0.0001; R2, 0.27 | |||
|---|---|---|---|
| (E) Cystatin C Equation 4 | |||
| Variable | Difference (ml/min per 1.73 m2) | 95% Confidence Interval | P-value |
| BMI (per 1 kg/m2) | −0.09 | −0.13 to −0.05 | 0.001 |
| Measured GFR >100 ml/min per 1.73 m2 (versus ≤100 ml/min per 1.73 m2) | −0.26 | −0.51 to −0.005 | 0.001 |
| Constant | 7.33 | −2.37 to 17.03 | |
Analysis of variance: F, 11.32; P < 0.0001; R2, 0.14. BMI, body mass index.
Subgroup Analyses
In patients with malnutrition criteria and reduced muscle surface or mass, CysC equations provided more accurate and precise eGFR estimates than the CKD-EPI equation, although CysC equation 3 was significantly less precise than the others (Table 4).
The accuracy of the CKD-EPI equation after excluding patients with malnutrition and/or amputation of >25% of body surface area (BSA) was significantly increased, remaining approximately constant in the range from 15 to 100 ml/min per 1.73 m2 (data not shown).
In the subpopulation with mGFR >60 ml/min per 1.73 m2, CysC equation 3 overestimated mGFR, whereas CKD-EPI and CysC equations 1, 2, and 4 had lower bias (Table 4).
CysC equation 1 underestimated mGFR in patients aged >70 yr. The accuracy of the CKD-EPI equation remained constant in the subgroup of patients aged >70 years (Table 4).
In patients with cirrhosis of the liver, both the CKD-EPI equation and CysC equation 3 overestimated mGFR. The bias in equations 1, 2, and 4 was significantly lower. Total muscle mass was the main source of eGFR errors in cirrhotic patients (R2 0.38, P = 0.0015), after adjustment for age and gender.
In patients with a BMI >35 kg/m2, CysC equation 3 significantly overestimated, whereas CKD-EPI and CysC equations 1, 2, and 4 underestimated mGFR (Table 4).
Discussion
The present study analyzes the accuracy of the CKD-EPI equation and of four equations based on CysC, either as a single variable or after adjustment for demographic characteristics and creatinine level, for estimating GFR in a large and representative sample of hospitalized patients with stable creatinine. The study population showed high prevalences of malnutrition, obesity, and SIRS, similar to those described previously (28–34), confirming the need for validating eGFR equations in this setting.
When we analyzed the main sources of error for CKD-EPI equation, we observed that total muscle mass and serum creatinine were the most powerful predictors. Total muscle mass was significantly conditioned by patients' nutritional status, after adjusting for age, gender, and BMI. These data suggest that the overestimation was a consequence of the high prevalence of malnutrition among hospitalized patients. Proof of this is the increased accuracy of the CKD-EPI equation when patients with malnutrition or amputations >25% of BSA were excluded from the analysis. In these conditions, the CKD-EPI equation gave a precise estimate in the 15 to 100 ml/min per 1.73 m2 range (for values above 100 ml/min per 1.73 m2, it significantly underestimated mGFR).
Regarding applicability of CKD-EPI in patients over 70 years of age (42% in our study), our study identified reduced muscle mass and BSA as the main sources of error, and showed little effect of age. The accuracy in patients aged >70 years with no malnutrition or reduced BSA was good, which suggests that age alone is not a factor limiting its use.
With regard to CysC-based equations, our data show that CysC equations 1, 2, and 4 give more precise results than the CysC equation 3 in general hospitalized patients, and that they do not systematically overestimate mGFR for values between 15 to 100 ml/min per 1.73 m2. As in other studies (15–16), we found that CysC levels were significantly associated with age, inflammation, and BMI, but these variables had little impact on eGFR differences. Equation 1 performed differently in elderly patients, leading to an underestimation of mGFR.
Cys C equations 2 and 4 gave comparable GFR values for mGFR values of 100 ml/min per 1.73 m2 or lower. The accuracy and precision of CysC equation 3 in our study differed significantly from those reported in nonhospitalized patients with CKD (12) This is probably due to the fact that it takes into account serum creatinine, and muscle mass was the main source of error in our sample.
In patients with malnutrition criteria and reduced muscle mass or BSA, all of the CysC equations provided more accurate and precise eGFR than the CKD-EPI equation. The impact of BMI on differences for the CysC equations was opposite than that found for the CKD-EPI equation. This difference may be explained by the correlation described between CysC levels and body fat mass (16).
The activation of inflammatory response may alter muscular creatinine production and increase CysC levels (16,35), which could, theoretically, influence both CKD-EPI and CysC equations. However, in our study, the presence of SIRS at admission did not predict error in eGFR. This could be due to the exclusion of patients admitted to critical units and patients with unstable renal function, who tend to have more severe inflammatory responses (34,36).
In patients with liver cirrhosis or Child class C liver disease, low creatinine production secondary to reduced muscle mass implies systematic overestimation of GFR with both creatinine-based equations and creatinine clearance (37–40). Some studies have highlighted that CysC in these patients could provide a more precise estimate (41–44). However, results are inconsistent. In our study, the CKD-EPI equation systematically overestimated mGFR in patients with liver cirrhosis, and the error was associated with muscle mass. This suggests that cirrhotic patients are comparable to patients with malnutrition due to other factors.
On the whole, our data suggest that CysC equations 2 and 4 might be more appropriate than the CKD-EPI for assessing renal function in hospitalized patients. However, cystatin-based methods are much more expensive than creatinine-based methods. Therefore, the clinical context in which they should be used needs to be defined in detail. Our data show that, in patients with stable kidney function, mGFR ≤100 ml/min per 1.73 m2, and no malnutrition or reduced BSA, the CKD-EPI equation adequately estimates GFR in all age groups. Detection of malnutrition and moderate loss of muscle mass is not easy without trained personnel and a huge investment in time. Therefore, systematic determination in all hospitalized patients is not possible. According to our data and to previous studies (8,45), the use of creatinine-based equations requires correction by muscle mass variations, particularly in hospitalized patients whose muscle mass depends heavily on their nutritional status. Rule et al. (45) showed that equations that include age and gender show a higher prevalence of renal disease than equations that only include muscle mass. It has been suggested that the six-variable MDRD equation (8) could be adapted to the hospitalized population, giving more weight to albumin. In our patients, however, although albumin was significantly related to the presence of malnutrition, its relation to muscle mass was quantitatively weak. These findings also apply to BMI.
The present study has some limitations. The exclusion of patients whose clinical status prevented the measurement of anthropometric variables could have led to an underestimation of malnutrition. No interviews on diet were conducted to determine daily protein intake, and this variable could not be included in the multivariate analysis for predicting creatinine. Finally, only one method was used to determine CysC, and important variations between methods have been reported (46).
In view of our findings, we conclude that, since the performance of the CKD-EPI equation is highly dependable on the presence of malnutrition or muscle mass loss, and measuring these variables is difficult in the clinical practice, the use of an equation based on CysC and gender, or on CysC, age, gender, and race, could be more appropriate to estimate GFR in hospitalized patients with reduced body surface area or in whom malnutrition is suspected.
Disclosures
Writing assistance was supported by Amgen S.A., but this company did not participate in the study design, data collection, or analyses. The authors declare no other conflict of interest.
Acknowledgments
Our acknowledgments to Montserrat Hervás, Helena Angulo, Maria Teresa Arraiz, and Mar Aguilar, the nurses who conducted the anthropometric studies and interviewed the patients.
We thank Dr. Ximena Alvira from Health Co SL (Madrid, Spain) and Dr. Neus Valveny from Trial Form Support SL (Barcelona, Spain) for assistance in the preparation of the manuscript. This study was supported by a grant from Amgen, S.A.
Footnotes
Published online ahead of print. Publication date available at www.cjasn.org.
Access to UpToDate online is available for additional clinical information at www.cjasn.org.
References
- 1. Walser M: Assessing renal function from creatinine measurements in adults with chronic renal failure. Am J Kidney Dis 32: 23–31, 1998 [DOI] [PubMed] [Google Scholar]
- 2. 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 [DOI] [PubMed] [Google Scholar]
- 3. 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 [PubMed] [Google Scholar]
- 4. Cockcroft DW, Gault MH: Prediction of creatinine clearance from serum creatinine. Nephron 16: 31–41, 1976 [DOI] [PubMed] [Google Scholar]
- 5. Levey AS, Bosch JP, Lewis JB, Greene T, Rogers N, Roth D: 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: 461–470, 1999 [DOI] [PubMed] [Google Scholar]
- 6. Levey AS, Coresh J, Greene J: [F-FC142]. Expressing the MDRD study equation for estimating GFR with IDMS traceable (goldstandard). http://nkdep.nih.gov/labprofessionals/F_Fc142.pdf
- 7. National Kidney Foundation K/DOQI clinical practice guidelines for chronic kidney disease: Evaluation, classification, and stratification. Am J Kidney Dis 39: S1–S266, 2002 [PubMed] [Google Scholar]
- 8. Poggio ED, Nef PC, Wang X, Greene T, Van Lente F, Dennis VW, Hall PM: Performance of the Cockcroft-Gault and modification of diet in renal disease equations in estimating GFR in ill hospitalized patients. Am J Kid Dis 46: 242–252, 2005 [DOI] [PubMed] [Google Scholar]
- 9. Levey AS, Stevens LA, Schmid CH, Zhang YL, Castro AF, 3rd, Feldman HI, Kusek JW, Eggers P, Van Lente F, Greene T, Coresh J, CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration): A new equation to estimate glomerular filtration rate. Ann Intern Med 150: 604–612, 2009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Grubb A, Simonsen O, Sturfelt G, Truedsson L, Thysell H: Serum concentration of cystatin C, factor D and beta 2-microglobulin as a measure of glomerular filtration rate. Acta Med Scand 218: 499–503, 1985 [DOI] [PubMed] [Google Scholar]
- 11. Simonsen O, Grubb A, Thysell H: The blood serum concentration of cystatin C (gamma-trace) as a measure of the glomerular filtration rate. Scand J Clin Lab Invest 45: 97–101, 1985 [DOI] [PubMed] [Google Scholar]
- 12. Stevens LA, Coresh J, Schmid CH, Feldman HI, Froissart M, Kusek J, Rossert J, Van Lente F, Bruce RD, Zhang YL, Greene T, Levey AS: Estimating GFR using serum cystatin C alone and in combination with serum creatinine: A pooled analysis of 3418 individuals with CKD. Am J Kidney Dis 51: 395–406, 2008 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Grubb A, Nyman U, Björk J, Lindström V, Rippe B, Sterner G, Christensson A: Simple Cystatin C-based prediction equations for glomerular filtration rate compared with the modification of diet in renal disease prediction equation for adults and the Schwartz and the Counahan-Barratt prediction equations for children. Clin Chem 51: 1420–1431, 2005 [DOI] [PubMed] [Google Scholar]
- 14. Fricker M, Wiesli P, Brändle M, Schwegler B, Schmid C: Impact of thyroid dysfunction on serum cystatin C. Kidney Int 63: 1944–1947, 2003 [DOI] [PubMed] [Google Scholar]
- 15. Rule AD, Bergstralh EJ, Slezak JM, Bergert J, Larson TS: Glomerular filtration rate estimated by cystatin C among different clinical presentations. Kidney Int 69: 399–405, 2006 [DOI] [PubMed] [Google Scholar]
- 16. Stevens LA, Schmid CH, Greene T, Li L, Beck GJ, Joffe MM, Froissart M, Kusek JW, Zhang YL, Coresh J, Levey AS: Factors other than glomerular filtration rate affect serum cystatin C levels. Kidney Int 75: 652–660, 2009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Thomassen SA, Johannesen IL, Erlandsen EJ, Abrahamsen J, Randers E: Serum cystatin C as a marker of the renal function in patients with spinal cord injury. Spinal Cord 40: 524–528, 2002 [DOI] [PubMed] [Google Scholar]
- 18. Delanaye P, Cavalier E, Radermecker RP, Paquot N, Depas G, Chapelle JP, Scheen AJ, Krzesinski JM: Cystatin C or creatinine for detection of stage 3 chronic kidney disease in anorexia nervosa. Nephron Clin Pract 110: c158–c163, 2008 [DOI] [PubMed] [Google Scholar]
- 19. Effersöe H, Rosenkilde P, Groth S, Jensen LI, Golman K: Measurement of renal function with iohexol. A comparison of iohexol, 99mTc-DTPA, and 51Cr-EDTA clearance. Invest Radiol 25: 778–782, 1990 [DOI] [PubMed] [Google Scholar]
- 20. Sterner G, Frennby B, Hultberg B, Almen T: Iohexol clearance for GFR-determination in renal failure—single or multiple plasma sampling? Nephrol Dial Transplant 11: 521–525, 1996 [PubMed] [Google Scholar]
- 21. Soman RS, Zahir H, Akhlaghi F: Development and validation of an HPLC UV method for determination of iohexol in human plasma. J Chromatogr B 816: 339–343, 2005 [DOI] [PubMed] [Google Scholar]
- 22. Sociedad Española para el Estudio de la Obesidad (SEEDO): Consenso español 1995 para la evaluación de la obesidad y para la realización de estudios epidemiológicos. Med Clin (Barc) 107: 782–787, 1996 [PubMed] [Google Scholar]
- 23. Elmore MF, Wagner DR, Knoll DM, Eizember L, Oswalt MA, Glowinski EA, Rapp PA: Developing an effective adult nutrition screening tool for a community hospital. J Am Diet Assoc 94: 1113–1121, 1994 [DOI] [PubMed] [Google Scholar]
- 24. Lee RC, Wang Z, Heo M, Ross R, Janssen I, Heymsfield SB: Total-body skeletal muscle mass: Development and cross-validation of anthropometric prediction models. Am J Clin Nutr 72: 796–803, 2000 [DOI] [PubMed] [Google Scholar]
- 25. Pugh RN, Murray-Lyon IM, Dawson JL, Pietroni MC, Williams R: Transection of the oesophagus for bleeding oesophageal varices. Br J Surg 60: 646–649, 1973 [DOI] [PubMed] [Google Scholar]
- 26. Bone R, Balk R, Cerra F, Dellinger R, Fein A, Knaus W, Schein R, Sibbald W: Definitions for sepsis and organ failure and guidelines for the use of innovative therapies in sepsis. The ACCP/SCCM Consensus Conference Committee. American College of Chest Physicians/Society of Critical Care Medicine. Chest 101: 1644–1655, 1992 [DOI] [PubMed] [Google Scholar]
- 27. Gracia S, Montañés R, Bover J, Cases A, Deulofeu R, Martín de Francisco AL, Orte LM: Sociedad Española de Nefrología. Documento de consenso: Recomendaciones sobre la utilización de ecuaciones para la estimación del filtrado glomerular en adultos. Nefrologia 26: 658–665, 2006 [PubMed] [Google Scholar]
- 28. Planas M, Audivert S, Pérez-Portabella C, Burgos R, Puiggrós C, Casanelles JM, Rosselló J: Nutritional status among adult patients admitted to an university-affiliated hospital in Spain at the time of genoma. Clin Nutr 23: 1016–1024, 2004 [DOI] [PubMed] [Google Scholar]
- 29. de Ulibarri Pérez JI, Picón César MJ, García Benavent E, Mancha Alvarez-Estrada A: Early detection and control of hospital malnutrition. Nutr Hosp 17: 139–146, 2002 [PubMed] [Google Scholar]
- 30. Weinsier RL, Hunker EM, Krumdieck CL, Butterworth CE, Jr: Hospital malnutrition: A prospective evaluation of general medical patients during the course of hospitalization. Am J Clin Nutr 32: 418–426, 1979 [DOI] [PubMed] [Google Scholar]
- 31. Singh H, Watt K, Veitch R, Cantor M, Duerksen DR: Malnutrition is prevalent in hospitalized medical patients: are housestaff identifying the malnourished patient? Nutrition 22: 350–354, 2006 [DOI] [PubMed] [Google Scholar]
- 32. Pérez de la Cruz A, Lobo Támer, Orduña Espinosa R, Mellado Pastor C, Aguayo de Hoyos E, Ruiz López M: Desnutrición en pacientes hospitalizados: prevalencia e impacto económico. Med Clin (Barc) 123: 201–206, 2004 [DOI] [PubMed] [Google Scholar]
- 33. Kyle UG, Unger P, Mensi N, Genton L, Pichard C: Nutrition status in patients younger and older than 60 y at hospital admission: a controlled population study in 995 subjects. Nutrition 18: 463–469, 2002 [DOI] [PubMed] [Google Scholar]
- 34. Brun-Buisson C: The epidemiology of the systemic inflammatory response. Intensive Care Med 26[Suppl 1]: S64–S74, 2000 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Doi K, Yuen PS, Eisner C, Hu X, Leelahavanichkul A, Schnermann J, Star RA: Reduced production of creatinine limits its use as marker of kidney injury in sepsis. J Am Soc Nephrol 20: 1217–1221, 2009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. 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 [PubMed] [Google Scholar]
- 37. Cocchetto DM, Tschanz C, Bjornsson TD: Decreased rate of creatinine production in patients with hepatic disease: Implications for estimation of creatinine clearance. Ther Drug Monit 5: 161–168, 1983 [DOI] [PubMed] [Google Scholar]
- 38. Sherman DS, Fish DN, Teitelbaum I: Assessing renal function in cirrhotic patients: Problems and pitfalls. Am J Kidney Dis 41: 269–278, 2003 [DOI] [PubMed] [Google Scholar]
- 39. Caregaro L, Menon F, Angeli P, Amodio P, Merkel C, Bortoluzzi A, Alberino F, Gatta A: Limitations of serum creatinine level and creatinine clearance as filtration markers in cirrhosis. Arch Intern Med 154: 201–205, 1994 [PubMed] [Google Scholar]
- 40. Papadakis MA, Arieff AI: Unpredictability of clinical evaluation of renal function in cirrhosis. Prospective study. Am J Med 82: 945–952, 1987 [DOI] [PubMed] [Google Scholar]
- 41. Woitas RP, Stoffel-Wagner B, Flommersfeld S, Poege U, Schiedermaier P, Klehr HU, Spengler U, Bidlingmaier F, Sauerbruch T: Correlation of serum concentrations of cystatin C and creatinine to inulin clearance in liver cirrhosis. Clin Chem 46: 712–715, 2000 [PubMed] [Google Scholar]
- 42. Orlando R, Mussap M, Plebani M, Piccoli P, De Martin S, Floreani M, Padrini R, Palatini P: Diagnostic value of plasma cystatin C as a glomerular filtration marker in decompensated liver cirrhosis. Clin Chem 48: 850–858, 2002 [PubMed] [Google Scholar]
- 43. Gerbes AL, Gulberg V, Bilzer M, Vogeser M: Evaluation of serum cystatin C concentration as a marker of renal function in patients with cirrhosis of the liver. Gut 50: 106–110, 2002 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44. Pöge U, Gerhardt T, Stoffel-Wagner B, Klehr HU, Sauerbruch T, Woitas RP: Calculation of glomerular filtration rate based on cystatin C in cirrhotic patients. Nephrol Dial Transplant 21: 660–664, 2006 [DOI] [PubMed] [Google Scholar]
- 45. Rule AD, Bailey KR, Schwartz GL, Khosla S, Lieske JC, Melton LJ, 3rd: For estimating creatinine clearance measuring muscle mass gives better results than those based on demographics. Kidney Int 75: 1071–1078, 2009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 46. Tidman M, Sjöström P, Jones I: A Comparison of GFR estimating formulae based upon s-cystatin C and s-creatinine and a combination of the two. Nephrol Dial Transplant 23: 154–160, 2008 [DOI] [PubMed] [Google Scholar]
