The chronic kidney disease epidemiology collaboration equation combining creatinine and cystatin C accurately assesses renal function in patients with cirrhosis

Background Accurate measurement of renal function in cirrhotic patients is still challenging. To find the best test for the determination of the true glomerular filtration rate (GFR) in cirrhotic patients this study prospectively compared measured (m)GFR, the gold standard, with estimated (e)GFR using equations based on serum levels of creatinine and cystatin C. Methods GFR was measured by sinistrin clearance using the bolus method in 50 patients with cirrhosis (Child Turcotte Pugh score A, B and C) and 24 age-matched healthy subjects as controls. Measured (m)GFR was compared to eGFR using bias, accuracy 10 % and 30 %, as well as correlation coefficients. Results Creatinine-based equations generally overestimated GFR in patients with cirrhosis and showed a bias (average difference between mGFR and eGFR) of −40 (CG), −12 (MDRD) and −9 (CKD-EPI-Cr) ml/min/1.73 m2. Cystatin C-based equations underestimated GFR, especially in patients with Child Turcotte Pugh score C (bias 17 ml/min/1.73 m2for CKD-EPI-CysC). Of these equations, the CKD-EPI equation that combines creatinine and cystatin C (CKD-EPI-Cr-CysC) showed a bias of 0.12 ml/min/1.73 m2 as compared to measured GFR. Conclusions The CKD-EPI equation that combines serum creatinine and cystatin C measurements shows the best performance for accurate estimation of GFR in cirrhosis, especially at advanced stages. Electronic supplementary material The online version of this article (doi:10.1186/s12882-015-0188-0) contains supplementary material, which is available to authorized users.


Background
Renal function has a pivotal prognostic value in patients with cirrhosis [1]. Its critical prognostic impact is indicated by the inclusion of serum creatinine (Cr) levels in the widely used MELD (Model for end-stage liver disease) score, whose value is an accurate predictor of 3-month mortality in cirrhosis. As a result, the MELD score is widely used for determining priority for liver transplantation [2].
An accurate evaluation of renal function is of utmost importance in patients with cirrhosis and patients with acuteon-chronic liver failure, especially those being evaluated for transplantation. Although the serum level of Cr is an easily measurable and widely available marker of excretory renal function, it has limitations in assessing glomerular filtration rate (GFR) in patients with cirrhosis [3][4][5][6]. Due to malnutrition, low protein intake, decreased Cr synthesis and increased tubular secretion, GFR in patients with cirrhosis is overestimated when estimated by serum Cr levels [3]. The equations that are most frequently used to estimate GFR include Cockcroft Gault (CG) [7], Modification of Diet in Renal Disease (MDRD) [8] and the Chronic Kidney Disease Epidemiology Collaboration formula (CKD-EPI); these equations, however, require corrections for age, gender, ethnicity and body weight. Estimated (e)GFR based on serum levels of cystatin C (CysC) has been claimed to be more accurate than Cr-based equations [9,10]. CysC is a nonglycosylated low molecular weight protein of the cystatin superfamily of cysteine protease inhibitors [11]. In renal impairment, its levels increase faster than Cr levels, and have been considered as valuable for the early detection of renal dysfunction [12,13]. Unlike Cr, CysC is independent of muscle mass, age and gender, and not influenced by serum bilirubin or malignancy [3,4,11,14]. However, measurement of CysC, has recently been reported to be influenced by factors such as low serum albumin levels, elevated white blood cell count, and elevated CRP levels [15]. These abnormalities are frequently present in cirrhotic patients and consequently are likely to impair the reliability of CysC-based equations. Therefore, measurement rather than estimation of GFR seems to be mandatory for these patients.
Although technically demanding, time-consuming and costly, inulin clearance has been considered the gold standard for determination of GFR. Inulin is freely filtered by the glomerulus and neither secreted, reabsorbed, synthesized nor metabolized by the kidney [16]. In Europe, sinistrin, an inulin-like polyfructosan, exhibiting the same properties as inulin, is preferentially used for the determination of GFR [17,18].
The aim of this study was to identify the best equation for the determination of eGFR in cirrhotic patients without concomitant kidney disease or conditions that are prone to structural kidney damage (e.g. diabetes). For that purpose, GFR was determined in 50 cirrhotic patients using sinistrin clearance measured after single injection technique (mGFR, measured GFR) and compared to that estimated by Cr-and CysC-based equations (eGFR). Results obtained from cirrhotics were compared to those of healthy controls.

Study population and collection of data
This prospective study was performed in accordance with the Declaration of Helsinki and the applicable laws of the Republic of Austria in 74 patients referred to our center between 2012 and 2014. The study protocol was approved by the ethics committee of the Medical University of Graz (23-497 ex 10/11). No sample size calculation was done since this study was performed as a pilot study. To compare different methods of GFR estimation with GFR measurement, 50 patients with compensated or decompensated cirrhosis of different etiologies as well as 24 age-matched healthy living kidney donors representing the healthy control group were included. Patients with known underlying kidney disease, diabetes, insufficiently treated arterial hypertension, collagen vascular diseases, known malignancy of the urogenital tract, recurrent urinary tract infections, ingesting nephrotoxic drugs such as aminoglycosides, or having a regular intake of non-steroidal anti-inflammatory drugs were excluded; patients who were pregnant, and/or breastfeeding and patients being unable to give informed consent due to cognitive impairment were also excluded. Patients were recruited during their visit to the liver outpatient clinic or while being on the inpatient ward. A written informed consent was obtained from each patient.
Measurements All investigations for the study took place at the Department of Nephrology at the Medical University of Graz. After placing a catheter in the antecubital vein, basement blood samples for serum analyses were taken. Presence of ascites was determined by abdominal ultrasound and patients were clinically checked for hepatic encephalopathy. Determination of mGFR by sinistrin clearance was performed using the bolus method, which has been considered to be advantageous over the continuous infusion since neither urine samples nor steady state conditions are required [17,19]. We employed a singleinjection technique with sufficiently long sinistrin serum concentration contours adapted to a two-compartment kinetic model with variable parameters for transfer-and elimination rates, which was extensively described earlier [18,[20][21][22]. The computer model used in our study allows a self-validation of the data that in turn results in GFR measurements of very high precision [17,[21][22][23]. Application of an exogenous marker performed as single injection experiment offers the opportunity of exact clearance determination without potential errors of incorrect steady state conditions or incomplete urinary collection. Each study participant received an injection of 2500 mg of sinistrin (Inutest ® , Fresenius KABI, Graz, Austria) intravenously. Serum concentrations of sinistrin were determined after 10, 15, 20, 30, 45, 60, 90, 120, 180, 210, 240, and 270 min after injection. An enzymatic method was used to measure the serum concentration of sinistrin as described previously [24]. Cr was measured using a rateblanked and compensated modified Jaffé method on a Cobas analyzer (Roche Diagnostics, Mannheim, Germany). The Cr assay was standardized by isotope-dilution using mass spectrometry (ID-MS). Liver tests were measured enzymatically and C-reactive protein (CRP) by immunoturbidometry (Roche Diagnostics, Mannheim, Germany). CysC was determined by particle-enhanced immunonephelometry using N Protein Standard UY from Siemens Healthcare Diagnostics, Marburg, Germany.
Statistical Analysis Measured GFR (mGFR) was compared to eGFR determined by different Cr-and/or CysCbased equations (CG, MDRD4, Hoek, Larsson, CKD-EPI equations using Cr, CysC and both) (Additional file 1). For calculating the CG equation, the measured body weight of the patients was used. Agreement between mGFR and eGFR was assessed using the mean bias (average difference between mGFR and eGFR), standard error of the mean bias, the Pearson correlation coefficient and the concordance correlation coefficient as described previously [25][26][27]. Accuracy 10 % and accuracy 30 % of each equation were calculated [28]. Bland-Altman plots were prepared showing correlation and mean bias between mGFR and eGFR [29]. Patients' characteristics were compared by Student's T test for continuous variables and Fisher exact test for categorical ones. Accuracies (P10 and P30) and correlation coefficients were compared using McNemar's test. Statistical analyses were performed using the commercial software SPSS (IBM SPSS Statistics 21) and STATA (Stata Statistical Software: Release 13. Stata-Corp, 2009, College Station, TX, USA).
In controls, the performance of the Hoek, Larsson and CKD-EPI-CysC equation was not superior to Crbased equations. Bias, P10 and P30 of the Hoek and Larson formula even showed worse performance compared to Cr-based equations and overestimated mGFR. Bland-Altman-plots comparing the most commonly used CysC-based equation (CKD-EPI-CysC) to mGFR are shown in Fig. 1B.

Superiority of the CKD-EPI equation combining Cr and CysC in patients with cirrhosis
In patients with cirrhosis, the CKD-EPI formula combining Cr and CysC showed the best performance amongst all equations. Mean bias was low (0.1 ± 16.3 mL/min/  Values are expressed as means ± standard deviations or n (%); AIH, autoimmune hepatitis; BMI, body mass index; Cr, Creatinine; CRP, C-reactive protein; CTP, Child-Turcotte-Pugh Score; CysC, Cystatin C; MELD, model for end-stage liver disease; n.a., not applicable; PSC, primary sclerosing cholangitis. * p <0.05; statistical significant difference between cirrhotics (all) and controls, ** p <0.05; statistical significant difference between CTP A and CTP B, *** p <0.05; statistical significant difference between CTP A and CTP C, **** p <0.05; statistical significant difference between CTP B and CTP C, ***** p <0.05; statistical significant difference between CTP A and control, ****** p <0.05; statistical significant difference between CTP B and control, ******* p <0.05; statistical significant difference between CTP C and control    (Table 2). Due to the relatively low number of study patients, statistical significant differences for P10 and P30 between the CKD-EPI-Cr-CysC equation and the other equations were only found for CG (p < 0.0001 for P10 and P30 determined by McNemar's test, Table 2). Also for correlation coefficients, statistical significant differences between CKD-EPI-Cr-CysC and all other eGFR equations were only found for Crbased equations in all cirrhotics, CG in CTP A and CysCbased equations in healthy controls. Bland-Altman-plots comparing mGFR to eGFR determined by the CKD-EPI-Cr-CysC equation are shown in Fig. 2. Of 50 cirrhotics, 8 had an mGFR < 60 ml/min/1.73 m 2 (Additional file 2). The combined CKD-EPI-Cr-CysC equation correctly identified 7 out of those 8 patients as patients with impaired renal function (eGFR < 60 ml/min/1.73 m 2 ). In contrast, less than half of those patients were correctly identified using Cr-based equations and even a greater proportion of patients were found to have a GFR < 60 ml/min/ 1.73 m 2 by using CysC-based equations (Additional file 3).

Discussion
There is an urgent need for early and precise detection of impaired GFR in cirrhotic patients, especially in those suffering from acute on chronic liver failure (ACLF) and being evaluated for liver transplantation. We, therefore, prospectively evaluated GFR in 50 patients with cirrhosis and 24 healthy controls by measuring the renal function using sinistrin clearance (mGFR) and comparing its value with estimated (e)GFR using creatinine (Cr)-and Cystatin (Cys)Cbased equations. We found that Cr-and CysC-based equations were inaccurate for the assessment of renal function in cirrhotic patients. Patients with known renal disease, diabetes or nephrotoxic drugs were excluded in order to exclusively study patients with impaired renal function most likely due to their liver disease. Only by the use of the combined CKD-EPI equation (CKD-EPI-Cr-CysC) we could obtain results comparable to those measured by sinistrin clearance. This represents the first prospective study with direct comparison of several equations for estimating GFR with the determination of GFR by measurement of sinistrin clearance using the bolus method. Due to various limitations (e.g. malnutrition, muscle atrophy) the commonly used Cr-based equations for estimating renal function in cirrhosis are unreliable. Although having been considered a more sensitive indicator of renal  [25,[30][31][32][33][34], CysC is influenced by factors independent of GFR which are frequently present in patients with cirrhosis such as elevated CRP or low serum albumin levels [15]. Since direct measurement of GFR by inulin clearance is technically demanding, time-consuming and costly, easier techniques using synthetic inulin-like polyfructosans (e.g. sinistrin) have been developed [17,18]. In contrast to previous studies we determined mGFR by bolus intravenous injection of sinistrin, a simpler method without urine collection, which is considered to be more precise compared to the constant infusion standard method in healthy subjects and in patients with ascites.
Apart from severity of liver disease, which was less pronounced in our study cohort, patients' characteristics were well comparable to previous studies in terms of age, gender and etiology of cirrhosis. Accordingly, we showed that Crbased equations tend to overestimate mGFR and therefore could not serve as reliable parameters for assessing renal function in cirrhosis ( Table 2). The accordance between eGFR and mGFR was much better in healthy controls, although the number of patients in this group was low. CysC-based equations rather underestimated mGFR in cirrhosis, however, the performance of these equations seemed to be less influenced by CTP score and was better A B Fig. 2 Bland-Altman-plots of eGFR determined by the combined CKD-EPI-Cr-CysC equation. Bland-Altman-plots of the CKD-EPI-Cr-CysC equation compared to mGFR determined by sinistrin clearance show an excellent agreement between eGFR and mGFR for the combined equation. The agreement between mGFR and eGFR using the combined formula in cirrhotic patients (a) is as good as that in healthy controls (b). a-b Horizontal long-dashed lines depict the bias (average difference between mGFR and eGFR), short-dashed lines show the limits of agreement between mGFR and eGFR according to the Bland-Altman method, and filled circles denote the measurement points. The oblique straight lines show the regression line between difference and average, and the shaded areas show the hyperbolic 95 % confidence limits of this regression line. R-values represent the Pearson's linear correlation coefficients between difference and average. CKD-EPI, Chronic kidney disease epidemiology collaboration; Cr, Creatinine; CysC, Cystatin C compared to Cr-based ones, which has as well been confirmed by other studies [25,32]. Although showing a better performance in accurately assessing mGFR than Cr-based equations, underestimation of GFR in end-stage cirrhotic patients by the use of CysC-based equations bears the potential risk of unnecessary simultaneous liver and kidney transplantation. Amongst all formulas, the CKD-EPI formula combining both, Cr and CysC, showed by far the best performance in cirrhosis (Table 2). Its diagnostic performance in cirrhotic patients was even as good as in healthy subjects and was independent of CTP score ( Table 2). We confirmed the results of a recently published study that compared the performance of the CKD-EPI-Cr-CysC equation to mGFR determined by non-radiolabeled iothalamate plasma clearance in cirrhosis and foundsimilar to our study -that this equation was superior to other frequently used Cr-and CysC-based equations, although its performance was still worse than reported by others for non-cirrhotics [34]. However, measurement of GFR by renal clearance of iothalamate has been suggested to be not as accurate as compared to the gold standard inulin or sinistrin [35]. The performance of the CKD-EPI-Cr-CysC equation was superior in our study with only 16 % of the estimated GFR values differing from the mGFR by more than 30 % in cirrhotic patients (24 % in the study by Mindikoglu et al.) [34]. Our findings of decreasing accuracy of Cr-based formulas with the increase of CTP score are in line with results from a study by de Souza et al., who evaluated a dataset of 202 consecutive liver transplantation candidates. They described a better performance of CysC-based equations with CKD-EPI-CysC being considered as the most accurate equation whatever the magnitude of ascites and even in the presence of significant renal dysfunction. However, diabetics were also included and GFR was measured by the continuous infusion method, which has been reported to be inferior to the bolus method we used [17,19,36].
The main limitation of this study is the number of study patients that is relatively small. Due to that small number which was due to careful selection of included patients, only a few patients appeared to have renal dysfunction and only slight differences in the performance of CKD-EPI-Cr-CysC as compared to the other CKD-EPI equations were observed in patients with more advanced cirrhosis.