Serum creatinine in patients with advanced liver disease is of limited value for identification of moderate renal dysfunction : Are the equations for estimating renal function better ?

1Department of Medicine, Division of Gastroenterology (Hepatology Services); 2Office of Medical Outcomes Research and Biostatistics; 3Division of Nephrology; 4Department of Diagnostic Imaging, Division of Nuclear Medicine, Dalhousie University, Halifax, Nova Scotia Correspondence and reprints: Dr Kevork M Peltekian, Hepatology Services, 1278 Tower Road, Victoria General Site, Room 6–302, Halifax, Nova Scotia B3H 2Y9. Telephone 902-473-2898, fax 902-473-2783, e-mail Kevork.Peltekian@dal.ca Received for publication July 12, 2005. Accepted January 25, 2006 J MacAulay, K Thompson, BA Kiberd, DC Barnes, KM Peltekian. Serum creatinine in patients with advanced liver disease is of limited value for identification of moderate renal dysfunction: Are the equations for estimating renal function better? Can J Gastroenterol 2006;20(8):521-526.

Serum creatinine in patients with advanced liver disease is of limited value for identification of moderate renal dysfunction: Are the equations for estimating renal function better?B oth worldwide and in Canada, the burden of chronic liver disease and cirrhosis is significant (1,2).Renal dysfunction is a major complication that accompanies cirrhosis and is associated with poor prognosis (3), especially after liver transplantation.Serum creatinine (Cr) is increasingly being incorporated into prognostic models for patients with decompensated cirrhosis (4,5) and for allocation of organs for liver transplantation (6).However, renal dysfunction in patients with advanced liver disease (ALD) is often difficult to assess due to poor methods for estimating renal function in this patient population (7)(8)(9)(10).
In general, Cr and urea clearance are used to estimate glomerular filtration rate (GFR).Using Cr-based methods to estimate GFR in ALD patients is problematic for multiple reasons.Decline in hepatic functional capacity results in decreased creatine production and lower serum Cr levels.ALD patients are known to have less skeletal muscle mass, resulting in diminished creatine storage and less conversion of creatine to Cr.All of these factors lead to a decreased serum Cr level in ALD patients, making Cr an unreliable factor in estimating GFR (11).Analytical methods used to determine the serum Cr concentration can also result in variability in Cr levels.The kinetic Jaffe method is commonly used to measure serum Cr levels and can be interfered with by elevated serum bilirubin levels, resulting in false low Cr levels (12).
Many formulas have been published to estimate GFR using serum Cr levels.The simplest is the reciprocal of serum Cr calculated as 1/Cr, with Cr in mg/dL (13).The most widely used formula to estimate filtration is the Cockcroft-Gault formula (CGF): with age in years, weight in kg and Cr in mg/dL; this formula is modified for women (14): In 1999, the Modification of Diet in Renal Disease (MDRD) Study Group published a more accurate method to estimate GFR from serum Cr based on demographic and serum variables: with age in years, Cr, urea nitrogen and albumin in mg/dL (15).These equations have been validated in patients with end-stage renal disease (16,17) and in renal transplant recipients (18,19).The MDRD formula has been used to estimate renal function in liver disease and transplantation (20)(21)(22).Recently, Skluzacek et al ( 23) compared renal with plasma clearance of 125-iodine ( 125 I) iothalamate, as well as compared the CGF with the MDRD formula to estimate renal function in a small patient group with ALD.Gonwa et al (24) evaluated the performance of these formulas in a large database for both pre-and postliver transplantation.Neither study reported on bias, precision or accuracy of the various calculated GFR formulas.The proper validation of the MDRD formula in ALD patients is important because serum Cr, urea nitrogen and albumin levels are variables that change with ALD.More importantly, these formulas need to be assessed in the context of identifying patients with ALD who truly have significant renal dysfunction (stage 3 or worse as defined by the Kidney Disease Outcomes Quality Initiative [KDOQI] Chronic Kidney Disease Classification) (25).
The aim of the present study was to compare estimates of GFR using published formulas (excluding those requiring urine collections) with standard radionuclide measurements of GFR in patients with ALD.We used the reciprocal for Cr, the CGF and the MDRD formulas for GFR estimation.

METHODS
Fifty-seven consecutive patients with ALD (excluding those presenting with fulminant liver failure) referred for liver transplantation assessment underwent routine radionuclide GFR measurement using 99m-technetium-diethylenetriamine pentaacetic acid ( 99m Tc-DTPA) by the Russell two-point sample technique following single bolus intravenous injection (26).All patients undergoing radionuclide assessment had clinically stable hemodynamics and renal function.99m Tc-DTPA clearance was used as the reference GFR because it has the same volume distribution, transit time and plasma clearance as inulin (Pearson's correlation coefficient [r]=0.98)(27).
Height and weight were measured on the morning of the study.All patients ingested 500 mL to 750 mL of water over a 2 h period before the procedure.An intravenous bolus of 5 mCi of 99m Tc-DTPA (DRAXIMAGE Inc, Canada) was administered into one arm with the residual activity in the syringe checked after injection through imaging.The blood samples were drawn from the contralateral arm.All chemical analyses and gamma counting were performed by routine methods at the Queen Elizabeth II Health Sciences Centre (Halifax, Nova Scotia).Clearances were calculated using standard formulas and were normalized to a body surface area of 1.73 m 2 .
Demographic and clinical data were retrieved from the liver transplantation database.The reciprocal of Cr, the CGF and the MDRD formulas were used; for conversion, serum Cr of 1 mg/dL equalled 88.4 µmol/L.To render the calculated GFR comparable, the reciprocal of Cr in mg/dL multiplied by 100 (100/Cr) was used.
All data were entered into Microsoft Excel (Microsoft Co, USA).Simple statistics (mean difference, relative percentage difference) were used to compare the radionuclide GFR with each of the other three methods using SAS version 8.2 (SAS Institute Inc, USA).The association between GFR and formulas was evaluated using Pearson's r, as well as the Bland-Altman plot (28) to describe the agreement using S-Plus 6.1 (Insightful, USA).All variables are given as mean ± SD or with ±95% CI.P<0.05 was considered significant.
The mean absolute difference between GFR ( 99m Tc-DTPA clearance) and calculated GFR was used to calculate bias.The precision of the formulas were expressed as the root mean square error.The accuracy was measured as the proportion of GFR estimates within 10%, 30% and 50% deviation of the reference GFR.Finally, sensitivity and specificity for each of the calculated GFR were determined at the cut-off point of 60 mL/min/1.73m 2 (moderate renal dysfunction).

RESULTS
Etiology of ALD varied among the patients studied, with Laennecs cirrhosis and cryptogenic cirrhosis being most prevalent.Severity of ALD was measured using the Child-Pugh score with 14% Child-Pugh class A, 33% Child-Pugh class B and 53% Child-Pugh class C. All but one patient were Caucasian.Mean height and weight were 1.65 m and 81 kg, respectively; 170 × (Cr) -0.999 × (age) -0.176 × (0.762 if patient is female) × (1.18 if patient is black) × (urea nitrogen) -0.17 × (albumin)  1.
The simple statistics summarizing the comparison of each method with 99m Tc-DTPA are shown in Table 2.The mean MDRD did not differ from the mean 99m Tc-DTPA clearance (P=0.414).Both 100/Cr and the CGF overestimated the mean 99m Tc-DTPA clearance (P<0.001).All three methods demonstrated large standard errors in the difference, suggesting limited precision.The regression analysis showed good degrees of association with Pearson's r (100/Cr) =0.74, r (CGF) =0.80 and r (MDRD) =0.87 (Figure 1), all at P≤0.0001.Pearson's coefficient of determination or the r 2 value was best associated with the MDRD formula and 99m Tc-DTPA (Table 2).Figure 2 summarizes the accuracy of the three calculated GFR formulas within 10%, 30% and 50%; the MDRD formula was the best among the three formulas.
An alternative statistical method to compare methods of measurement is the Bland-Altman plot, which shows agreement between a test method and the reference.It uses 95% limits of agreement to define the range within which most differences between measurements lie (28).The Bland-Altman plot revealed that the 100/Cr formula resulted in a mean overestimation of 24 mL/min/1.73m 2 with limits of agreement between -81 mL/min/1.73m 2 and 22 mL/min/1.73m 2 (Figure 3A).The CGF overestimated the mean by 15 mL/min/1.73m 2 with limits of -72 mL/min/1.73m 2 and 27 mL/min/1.73m 2 (Figure 3B).The MDRD formula had the highest level of agreement with GFR measured by 99m Tc-DTPA clearance, with an overestimation of 3 mL/min/1.73m 2 , and limits of agreement between -78 mL/min/1.73m 2 and 52 mL/min/1.73m 2 (Figure 3C).These figures had a GFR of 60 mL/min/1.73m 2 or less; the difference was the least between the reference GFR and the MDRD formula.
Table 2 also provides the sensitivity and specificity of each calculated GFR formula for moderate renal dysfunction (cut-off point 60 mL/min/1.73m 2 ).The sensitivity was best for the MDRD formula.

DISCUSSION
The data shown here confirm previous studies stating that renal dysfunction is underestimated in patients with ALD (7)(8)(9).The use of 99m Tc-DTPA clearance as reference Although inulin clearance is the gold standard for the measurement of GFR (11), this method is cumbersome and is used mainly as a research tool.In clinical practice, GFR measurements by radioisotope filtration markers have gained prominence because of their simplicity.The renal clearance of 99m Tc-DTPA, 169-ytterbium-DTPA or 125 I-iothalamate, administered as a single intravenous or subcutaneous injection, can be used to accurately measure GFR compared with using inulin clearance (12).All three radioisotopes are routinely used; however, it has been shown that 169-ytterbium-DTPA and 125 I-iothalamate have higher clearance values compared with 99m Tc-DTPA or inulin (12).These radioisotope filtration markers are alternatives to inulin clearance as a standard reference for measurement of GFR (29), especially in routine clinical practice (30).

Comparison of GFR formulas in ALD patients
The analysis showed that 99m Tc-DTPA clearance correlated strongly with all three formulas.However, the scatter plot (Figure 1) illustrates the overestimation of 99m Tc-DTPA clearance.The correlation between two measurement tools looks at how well the measurements are linearly related, not the agreement between them (31).The correlation also ignores the bias in each estimate of GFR and measures the relative agreement rather than the absolute agreement.Graphical representation, such as the Bland-Altman plot, is one alternative method preferred over correlation analysis to measure agreement in method comparison studies (28,31).Using this method, we determined that the MDRD formula provides the narrowest 95% limits of agreement.
The MDRD formula was also more accurate in the prediction of GFR measured by 99m Tc-DTPA clearance as demonstrated by the lowest mean difference.This formula is based on age, sex and serum Cr levels (as is the more commonly used CGF), as well as serum albumin and serum urea nitrogen levels.ALD patients are often malnourished and have decreased albumin and urea production by the liver, that can lead to low serum albumin and urea levels (32).The formula seems to be structured to compensate for this.The other major difference between the MDRD formula and the CGF is that the MDRD formula does not factor in patient weight.Most ALD patients have moderate to significant ascites and edema, leading to increased weight.Patient weight is found in the numerator of the CGF, indicating that an overestimation in GFR would occur, which is exactly what is observed clinically.
Our results indicate that the MDRD formula may provide better estimates of 99m Tc-DTPA clearance on average compared with the more common CGF.The correlation analysis and Bland-Altman plots provided estimates of the measurement errors, not indications of acceptability.It is the result of the measurement errors that decides the appropriateness of the methods.The MDRD formula provided limits of agreement that were clinically acceptable in our patient population.

Clinical implications
Using the CGF or 100/Cr method to estimate GFR in patients with ALD can lead to an overestimation of GFR by 15 mL/min/1.73m 2 and 24 mL/min/1.73m 2 , respectively.In a clinical situation, this overestimation may result in overlooking renal dysfunction.Renal function has significant prognostic implications in the course of liver disease.It is a key factor in the determination of transplant status.As well, many medications that are used both pre-and postliver transplantation are cleared by the kidneys and, therefore, true renal function is an important factor in correct dosage and therapeutic monitoring.
If the clinician is unaware of the severity of renal dysfunction, further damage may be done to the kidneys through procedures and medications, possibly resulting in hepatorenal syndrome, which is associated with significant mortality (33,34).The present study indicates that the MDRD formula for GFR estimation is more accurate on average than the CGF and 100/Cr, providing clinically acceptable limits of agreement with small underestimated 99m Tc-DTPA clearance.
ACKNOWLEDGEMENTS: Many thanks to Marlene Hudgins of the Nuclear Medicine Department and Helena MacKinnon, liver transplant recipient coordinator at the time of data collection, for their indispensable assistance in completing this project.
Estimating GFR in advanced liver disease Can J Gastroenterol Vol 20 No 8 August 2006 523

Figure 3 )Figure 1 )Figure 2 )
Figure 3) Bland-Altman comparison plots, which measure the agreement between glomerular filtration rate (GFR) measured by 99m-technetium-diethylenetriamine pentaacetic acid clearance, and A 100/creatinine (Cr), B the Cockcroft-Gault formula (CGF) and the C Modification of Diet in Renal Disease (MDRD) formula.The mean difference is indicated by the centre line; 95% limits of agreement are indicated by the upper and lower lines.The MDRD formula shows the highest level of agreement with 99m-technetiumdiethylenetriamine pentaacetic acid clearance, overestimating renal function by 3 mL/min/1.73m 2 (P=0.414),while the CGF and 100/Cr overestimated renal function with significant difference (P<0.001)