Review article: renal function assessment in cirrhosis – difficulties and alternative measurements
Professor A. K. Burroughs, Liver Transplantation and Hepatobililary Medicine, Royal Free Hospital, Pond Street, Hampstead, London NW3 2QG, UK.
Background Renal function in patients with cirrhosis is important prognostically, both before and following liver transplantation. Its prognostic impact is reflected by the inclusion of serum creatinine in the model for end-stage liver disease score, which is now used for recipient prioritization on liver transplantation waiting lists in the USA.
Aim To review the accuracy of the surrogate markers for the assessment of renal function, i.e. glomerular filtration rate, particularly in patients with cirrhosis.
Method We reviewed the available literature in PubMed regarding the markers for GFR evaluation and the factors which affect their accuracy in cirrhosis.
Results Although creatinine is widely available, it is an unreliable marker of glomerular filtration rate, particularly in patients with cirrhosis. Clearance of exogenous markers is considered the ‘gold standard’, but this methodology has many drawbacks, particularly poor applicability. Several mathematical formulae for estimated glomerular filtration rate are used to overcome some of these limitations: Cockcroft-Gault and Modification of Diet in Renal Disease formulae are the most frequently applied, but they are based on serum creatinine.
Conclusions Due to the inaccuracy of serum creatinine and its derived formulae in estimating glomerular filtration rate, alternative serum markers, such as cystatin C, and new formulae are desirable. These need formal evaluation in patients with cirrhosis so as to have a reliable surrogate of glomerular filtration rate, and to obviate many problems that are associated with using creatinine and estimated glomerular filtration rate.
Understanding predictive factors for survival in cirrhotic patients is very useful in therapeutic decision-making, including allocation for liver transplantation (LT). Renal dysfunction is a well-established adverse predictor associated with increased mortality in both acute liver failure and cirrhosis, particularly following acute complications, such as sepsis,1–3 and also following LT.4, 5
Cirrhotics who develop hepatorenal syndrome (HRS) have very high mortality (particularly for HRS type I, as median survival without specific therapy is only 2 weeks),6 and even with terlipressin and albumin, only approximately 40% respond and survive for 1 month after treatment.7–9 However, cirrhotics are also exposed to therapies and complications that increase the risk of renal failure.9–12 In patients who initially develop HRS, septic, ischaemic and/or hypotensive episodes, it can lead to acute tubular necrosis (ATN) and to a degree of interstitial fibrosis and tubular atrophy, which remains irreversible. Patients who become haemofiltration (or dialysis)-dependent, because of HRS, often remain so, until LT or death. Recovery from HRS after successful transplant sometimes takes several weeks.
The prognostic impact of renal function is reflected by the inclusion of serum creatinine (Cr) in the model for end-stage liver disease (MELD) score, which predicts the likelihood of death within 3 months and is used for prioritization of recipients (‘sickest first’) in the US.13, 14 MELD score is considered better than Child-Pugh score in part because of the inclusion of Cr as a prognostic factor. Although Cr is a routine laboratory test and widely accepted as a measure of renal function, it is only an indirect marker of renal function, i.e. of glomerular filtration rate (GFR). A problem, not often recognized by hepatologists, is that measurement of Cr suffers from a variety of interferences having significant problems of standardization,15 and no international standard for measurement.16
However, as renal dysfunction is so important prognostically, both before and after LT, and because of its inclusion in MELD, the accurate assessment of renal function in cirrhosis has become even more important and necessary than hitherto.
Measurement of glomerular filtration rate
Glomerular filtration rate is a direct measure of renal function, which is impaired before the onset of symptoms, correlating with the severity of renal impairment (Table 1). GFR estimation is considered the best index of renal function;17 if <60 mL/min, it defines chronic kidney disease (CKD) and represents a 50% or more reduction of normal adult renal function.14
Table 1. Different methods for estimation of GFR
|Clearance of exogenous marker||Inulin|
Non-radioactive contrast media (iohexol or iothalamate)
Radioisotopic methods (51Cr- EDTA, [99Tcm]-DPTA and 125I-iothalamate)
|Serum marker||Creatinine, urea, cystatin C, β2-microglobin|
|Creatinine clearance||24-h urine collection|
|Mathematical formulae*||Equations based on serum Cr or cystatin C|
The clearance of exogenous markers, such as inulin (considered the ‘gold standard’) and synthetic inulin-like polyfructosans, or non-radioactive contrast media (iohexol or iothalamate)18–20 are the most accurate methods for GFR estimation. Radioisotopic methods, such as 51Cr-EDTA, [99Tcm]-DPTA and 125I-iothalamate have comparable accuracy.21–24 Measurement of Cr clearance, performed by 24 h urine collection, has several pitfalls (see next section). To overcome some of these limitations, GFR is estimated by mathematical formulae based on serum Cr, such as the Cockcroft-Gault (C-G)25 and Modification of Diet in Renal Disease (MDRD) formulae.24, 26 Recently, in the UK, estimated GFR (based on MDRD formulae) has become a routine calculated measurement,27 but Cr alone is still used clinically as a surrogate index of GFR.
The limitations of the above methods of assessing GFR are not always understood amongst non-nephrologists including gastroenterologists and hepatologists, because of an incomplete understanding of what serum Cr represents, its pathophysiological pathway and the factors that affect its metabolism and measurement.17, 24
Different methods of serum cr measurement
Creatinine measurements suffer from interlaboratory differences, only some accounted for, by calibration differences between assays. A major problem, which also affects the results of GFR formulae, is that Cr assays are subject to interference by chromogens, bilirubin being the major one. This makes Cr an even less precise surrogate of GFR in jaundiced patients. Creatinine is routinely measured by either colorimetric or enzymatic methods. Most methods employ the Jaffe reaction: Cr reacts with alkaline picrate, to form an orange-red complex (Janovski complex) and is measured spectrophotometrically at 520 nm. The widespread use is attributable to its simplicity, ease of automation and relatively low cost.
Chromogens such as bilirubin, glucose, uric acid, ketoacids, pyruvate and certain antibiotics interfere with Cr measurements to varying degrees, so that various modifications of the original Jaffe and enzymatic methods were introduced to separate these chromogens such as acid blanking and absorption techniques with Fuller’s earth or Lloyd’s reagent, but these are laborious and unsuitable for routine use. The kinetic alkaline picrate method is sometimes used giving a differential rate of colour development between Cr and non-CR chromogens. Several enzymatic methods using creatininases and creatininase hydrolases are available. The end product is sarcosine oxidized to glycine, formaldehyde and hydrogen peroxide. The latter forms a chromogen, the colour intensity being proportional to the starting Cr. Recently, to improve the correlation between the Jaffe method and HPLC, Roche Diagnostics (Burgess Hill, UK) has introduced new calibrators, which adjust Cr values by 26 μm to compensate for the systematic error. However, this fixed correction does not reflect differences across the range of bilirubin concentrations found in patients thus introducing another systematic measurement error.
Dry chemistry systems (e.g. Ortho Diagnostics, Rochester, NY, USA) are less prone to chromogen interference. However, this methodology is not suitable for large-scale use as it is costly, and the equipment is less available. The negative interference of bilirubin, both conjugated and unconjugated, in serum or plasma Cr measurements by both Jaffe and enzymatic methods is an additional unresolved problem in patients with liver disease. To try to overcome this, deproteinization of samples prior to analysis, increasing the picrate concentration, use of bilirubin oxidase or addition of potassium ferricyanide were used in the Jaffe reaction. We recently showed15 that, interference occurred with bilirubin concentrations >62 μm (3.7 mg/dL), resulting in significant differences in Cr values between different methods. The Jaffe unmodified method showed the greatest interference with bilirubin and the modified Jaffe method (addition of potassium ferricyanide) the least interference. The variation of MELD score according to the method of Cr measurement, was considerable.15
A further problem associated with accuracy and precision of Cr measurements is, a lack of reference standard. Calibration of individual methods to a readily available reference standard would reduce discrepancies, but this currently does not exist. This also affects the calculation of MDRD, for which Cr measurement method was the Beckman CX3 Jaffe assay.28 Recently, reporting of estimated GFR, based on MDRD, has been recommended internationally,28 so that attempts have been made to realign the methods using correction factors traceable to a reference method for measurement of Cr using isotope dilution mass spectrometry (ID-MS) method. This should reduce the interlaboratory variations in estimating GFR to some extent. The ID-MS traceable MDRD equation, improves the estimation of GFR when using assays that account for chromogen interference.28
Factors affecting accurate assessment of glomerular filtration rate
Limitations of exogenous markers
Although clearance of exogenous markers for estimating GFR is the best method, there are several technical difficulties, and are expensive and impractical for routine use (Table 2). In addition, clearance rates of biomarkers from the circulation are inherently unreliable in patients with oedema, ascites and pleural effusions because of abnormal volumes of distribution and may be affected by circulation time.
Table 2. Characteristics of the established methods for renal function assessment
| Cr||Widely available||High specificity but low sensitivity for detection of CKD|
Is influenced by several factors unrelated to renal function
Not early detection of acute renal failure
Unreliable in particular group, such as cirrhotics
Absence of standardization of the laboratory method
Particularly in cirrhotics: is influenced by bilirubin
|Clearance of exogenous marker||‘Gold standard’||Technical difficulties|
Expensive impractical for routine use
Not available in all countries
Unreliable in patients with oedema, ascites and pleural effusions
|Clearance of endogenous marker|
| Creatinine (24 h urine collection)||More accurate compared to Cr||Underestimates GFR in children|
Inconvenient, unpleasant Overestimates GFR in proteinuria and renal dysfunction
Influenced by muscle metabolism and diet, inflammatory disease and malnutrition
25% unexplained variation due to incomplete urine collection and errors in urine volume measurement
In cirrhosis: overestimation of GFR
|Mathematical formulae based on Cr||Easier method compared to 24-h urine collection||The accuracy may be lower in specific clinical settings (none of these have been developed in cirrhotics)|
Unreliable in acute renal failure, muscle wasting disorders
Does not overcome the limitations in serum Cr
|C-G formula||Requires only gender, age, body weight||Not take into account the ethnicity|
|MDRD formula||Body weight is not needed|
Ethnicity, gender and age are taken into account
|Has not been extensively validated in patients without CKD|
6-variables formula: needs albumin, urea
Limitations of endogenous markers
Factors affecting serum creatinine
Creatinine is an end product of metabolism in humans. Creatine released from the liver, is taken up by muscle and non-enzymatically dehydrated to Cr.17 Serum Cr concentration is influenced by several factors unrelated to renal function, such as dietary intake of Cr (meat ingestion), turnover rate of creatine to Cr, renal tubular secretion of Cr, urinary flow rate, state of hydration, as well as the total pool of body creatine (total muscle mass).17 The latter is an important extra-renal factor leading to discrepancies in Cr concentration between individuals with the same renal function (same GFR), but of different age, sex, race, body mass index or having concomitant chronic disease, such as cirrhosis, malignancies, inflammation and endocrinopathies.17 Thus, women have lower Cr compared to men with the same renal function29 and the elderly may have overestimation of their renal residual function, compared to younger individuals. Larger muscle mass in Afro-Caribbean patients leads to increased GFR for the same Cr, compared to other ethnic groups, whereas South Asians have lower GFR (lower muscle mass).24, 30
Creatinine has high specificity but low sensitivity for detection of CKD,24, 30, 31 but there is a stage, when with a normal Cr, there is an abnormal GFR, whatever the aetiology of renal dysfunction. Interestingly, although normally, there is a decline in renal function with ageing, there are no differences in Cr values in the reported normal range between healthy elderly and younger people,32, 33 again suggesting a poor correlation of Cr with GFR. GFR and muscle mass fall with age but Cr may remain unchanged. Acute anuria may be associated initially with only a slight increase of serum Cr, even if GFR is almost zero30– Cr usually increases by 50–100 μm/day when GFR ceases. Very few correlations have been made outside of patients with chronic renal failure between GFR and Cr.
In cirrhotic patients the additional reasons why Cr is inadequate for renal function assessment are as follows: (i) decreased hepatic production of creatine (reduction in creatine pool), (ii) the oedematous state that complicates end-stage liver disease, leading to large distribution of Cr in the body and lower serum Cr concentration30 and (iii) use of drugs such as cephalosporins (used as they are non-nephrotoxic) and calcitriol, which affect the tubular secretion of Cr.30 As a result, in a recent meta-analysis, GFR estimation by inulin clearance was proposed as the only way for accurate assessment of renal function,34 but this remains impractical for routine clinical use.
In addition, patients with cirrhosis can develop renal dysfunction because of many causes, often in combination. Under these circumstances, Cr is an even more unreliable indicator of reduced GFR. Acute renal failure in cirrhotics usually develops on the basis of complications such as variceal bleeding, spontaneous bacterial peritonitis or sepsis, conditions that are related to increased Cr tubular excretion.17 The latter, combined with the lower muscle mass and dietary Cr intake because of malnutrition, can lead to a great discrepancy between the Cr and severity of renal dysfunction.
Factors affecting renal clearance of creatinine
Creatinine clearance underestimates GFR in children, and when the serum levels are high, where re-absorption from the tubules is greater.31 In association with proteinuria and reduced renal function, when tubular secretion is increased, Cr clearance overestimates GFR.35 Other pitfalls are because of extra-renal elimination of Cr by micro-organisms. Creatinine clearance is influenced by muscle metabolism and diet (a heavy meat meal can increase 24-h creatinine clearance by 37%), and generation of creatinine is increased in inflammatory disease and malnutrition.36 If muscle mass remains the same, a new steady-state of creatinine excretion is achieved in 4 months after any change in the body creatinine pool. In addition, several studies have revealed that there is up to 25% unexplained variation in GFR estimation based on Cr clearance,31 because of incomplete urine collection, errors in urine volume measurements, variations in tubular excretion or re-absorption of Cr and other unpredictable factors.17, 24, 37 In cirrhosis, standard measurement of creatinine clearance from urine collections has shown that it overestimates the true GFR34 (Table 3).
Table 3. Studies comparing different methods for assessment of renal function with a ‘gold standard’ method in cirrhotic patients
|Proulx et al.34 (meta-analysis)||193||Inulin||Creatinine clearance (24-h urine collection) overestimates GFR (particularly in low GFR)|
|Gonwa et al.38||1447||125I-iothalamate (90.7 ± 40.5)||MDRD, C-G: overestimate and underestimate renal function in GFR < 40 min/mL and GFR > 40 mL/min, respectively.|
MDRD better than C-G
|Woitas et al.54||44||Inulin||Cys-C more sensitive than Cr to detect reduced GFR|
|Orlando et al.55*||36 cirrhotics, 56 controls||Inulin||Cys-C more sensitive than Cr and Cr clearance (in both groups)|
|Demitras et al.56||26 (hepatorenal syndrome)||[99Tcm]-DTPA||Cys-C had better correlation with GFR, compared to Cr and Cr clearance|
|Samyn et al.60||62 children||51Cr-EDTA||Cys-C better correlation with GFR, compared to Cr |
|Poge et al.72||44||Inulin (28.3 mL/min/1.7 m2)||Cr and Cys-C overestimate GFR|
|Skluzacek et al.73|| ||125I-iothalamate (58.2 ± 5.1)||MDRD and C-G overestimate GFR|
MDRD superior to C-G
|MacAulay et al.74||57||[99Tcm]-DTPA (83 mL/min/1.7 m2)||MDRD better correlation and agreement with GFR, compared to G-C and 100/Cr|
All methods overestimate GFR
Factors affecting mathematical formulae based on serum creatinine
Different formulae which incorporate Cr for GFR estimation, have tried to overcome some of these limitations of creatinine clearance-based GFR estimation and are being increasingly used, particularly the C-G and MDRD formulae. Others, such as Nankivell, Gates, Jelliffe, Walser, Levey and Wright are recommended for use in particular patients groups. When the C-G formula25 that requires weight, gender and age is used, the estimated GFR is within 20% of the reference procedure which was a duplicate creatinine clearance measured in the original study for GFR estimation. However, only 4% of the reference cohort were women; creatinine was measured on Technicon autoanalyser (Saskatoon, SK, Canada) using Jaffe reagent. This formula does not take into account the ethnic variation in muscle mass. For the MDRD formula,26 a body weight variable (which is difficult to assess as lean body mass in ascitic and malnourished patients) is not needed, and it uses ethnicity, gender and age and is calculated for 1.7 m2 body surface area; estimated GFR is within 30% of the reference procedure which was 125I-iothalamate in 90% of patients (the creatinine measurement used was kinetic Jaffe on Beckman Astra CX3; Boehringer Mannheim, Mannheim, Germany). The accuracy of MDRD has been only estimated on a large scale, in patients with CKD. In cirrhosis, although there is discrepancy when compared to 125I-iothalamate,38 it is considered the best formula, compared to C-G (Table 3), possibly because it incorporates urea and albumin, which are abnormal in cirrhotics.21 However, usually and most often, the formula, excluding urea and albumin (four variables), is used to calculate GFR, as it is as accurate as the original six-variable formula.39 However, its usefulness has not been proven in healthy individuals,21, 40 and its accuracy may be lower in specific clinical settings. For example, MDRD is not applicable in children. So, Coulahan-Barratt and Schwartz formulae are considered more accurate21 and may be superior to [99Tcm]-DPTA.42 In kidney transplant patients, the Walser and Jelliffe formulae have been shown to be superior to MDRD.22, 42 This would suggests that a specific formula should be derived for patients with cirrhosis.
However, the use of such formulae does not overcome the limitations in serum Cr measurement. Even if their performance may be enhanced after cimetidine administration (which blocks tubular secretion of creatinine, and which in practice may only make a difference if 24-h urine collection for creatinine clearance is undertaken), their accuracy for estimating GFR17, 42 is still limited. They cannot substitute for the inulin clearance,22 particularly in children, the elderly,33 in cirrhosis34 and those with renal insufficiency (with or without abnormal serum Cr)24, 37, 43, 44 (Table 3).
In addition, these formulae have several intrinsic drawbacks due to limitations inherent in the original cohorts from which they were derived (for example, the C-G formula was developed in cohorts of healthy men, and might not be suitable in women or other groups of patients).24, 30 Importantly, neither C-G nor MDRD were developed in patients with cirrhosis.22, 40 MDRD correlates less well with GFR in cirrhosis.38 Estimates for GFR are also unreliable in acute renal failure and muscle wasting disorders because of kinetics of creatinine accumulation.33
Alternative serum markers for assessment of renal function
The clinical implications of the problems of Cr measurement in cirrhotics are very important. Calculation of MELD scores using different Cr assays and therefore different creatinine values yields different scores.15 For example, using four different methods for Cr measurement, in patients with bilirubin between 200 and 400 μm, the range of MELD scores was found to be 0–5 points and between 100 and 200 μm, it was 0–4 points.15 Moreover, failing to correct for the Cr method used as a surrogate marker for renal function, with respect to gender and ethnicity, may also have important consequences when therapeutic decisions are taken.15 We have shown that if the same GFR is taken as a basis for creatinine values in eGFR formulae and in EDTA Cr45 measurement, then women would have higher MELD scores than if the measured creatinine is used.46 This could be a reason, why women on LT waiting lists in the US have higher mortality and thus are less likely to undergo LT than men, using the MELD-based allocation system,47 although the preliminary studies had shown similar MELD scores at transplantation and mortality on the waiting list.48, 49 Moreover, a recent paper5 using UNOS data, has shown significantly more males (P < 0.0001) being transplanted following the introduction of MELD, compared to the period before its introduction. We believe that the unimodal assessment of Cr may be one reason for this inequality, such that women have a worse prognosis for an equivalent MELD score, because in reality their renal dysfunction is worse,46 as pre-LT renal function has important prognostic significance.2, 4
Based on the inaccuracy of Cr and its derived formulae and 24-h urine collection for creatinine clearance, a simple and more accurate serum marker would be preferable for reliable estimation of renal function. Use of several low-molecular weight proteins, such as β2-microglobin and β-trace protein are proposed, but cystatin C has received the most interest.24 Cystatin C is a non-glycosylated basic protein, produced by all nucleated cells; its renal clearance is approximately 94% of 51Cr-EDTA.24 Cystatin C is more sensitive for early detection of renal dysfunction, compared to Cr44, 50, 51 and reflects GFR independently of diet, sex, age beyond 1 year, muscle mass, serum bilirubin and body composition. In addition, concentrations are not influenced by inflammation or malignancy.45, 52 In a recent meta-analysis, cystatin C was clearly superior to Cr and had better correlation with the GFR, compared to Cr (r = 0.82 vs. 0.74, respectively).53 These qualities would have particular value in cirrhotic patients (Table 3). Indeed, cystatin C may give similar or more information, compared to Cr.54, 55 In cirrhotics with HRS, cystatin C is at least as accurate as Cr.56 In a recent study, Gerbes et al. evaluated 97 patients and concluded that cystatin C is a useful tool for renal function assessment, especially in patients with end-stage liver disease, in women and for earlier diagnosis of renal impairment, compared to enzymatic Cr and C-G formula.57 Renal dysfunction after renal or LT, mainly because of long-term exposure to calcineurin inhibitors, may also be best monitored by cystatin C.58, 59 Cystatin C is a reliable index of GFR after renal or LT, superior to Cr and C-G or Schwartz formulae.60–63 In addition, several equations for GFR calculation from serum cystatin C have been developed.64 These formulae are simpler and they have given similar or more accurate estimation of GFR, compared to those based on serum Cr.65, 66
However, cystatin C is not without its drawbacks. Although cystatin C is easily measured by commercially available immunoassays, the cost is significantly higher compared to that of Cr,24 and the assays may need further standardization.67 In addition, with cystatin C, larger intraindividual variation has been reported and its concentrations may be influenced by the presence of infection, drugs, such as corticosteroids, cotrimoxazole, angiotensin-converting enzyme inhibitors or calcineurin inhibitor therapy.63, 68 However, these interactions have not been confirmed by other studies.63 More studies are required to elucidate the usefulness of cystatin C as a marker of GFR in cirrhotics, particularly pre-LT and post-LT and to confirm its potential significant advantages over Cr.59 These would obviate the increased cost of its measurement, and may reduce the need for isotope measurement of GFR. It could also replace Cr in prognostic formulae.
Accurate assessment and early detection of renal dysfunction in cirrhotics, both before and after LT, is very important for therapeutic decision-making and for evaluating prognosis. This is particularly so in candidates on the waiting list for LT, for whom the MELD allocation protocol represents a justice system (‘the sickest goes first’). It is important that there are no systematic biases in the MELD calculation, which could influence prioritization of some patients vs. others. One such variation is the use of different laboratory methods for Cr measurement between laboratories.12 The second is not accounting for gender46 and age differences in creatinine measurements. The third is not accounting for differences in nutrition and muscle mass. Problems with different assays have also been found for INR measurements.69 To eliminate these biases, standardization or conversion factors are needed for the Cr measurement used in the MELD score to maintain it as a justice system, as it is likely that Cr will remain as a variable for the foreseeable future, as it is cheap to assay and is widely available, despite being an inaccurate surrogate marker for renal function.
Unfortunately, all the available methods for renal function assessment have several limitations (Table 2) and also do not have a good correlation with GFR (Table 3). The C-G formula does not offer any advantage, compared to Cr alone.70 MDRD formulae should be evaluated as a substitute for creatinine as it seems the best calculated formula for GFR in cirrhosis.38 Inulin is still available in Europe (but not in the US). Using an accurate pump and a loading dose, an inulin infusion can be used to achieve a steady-state. Then, the amount infused (which is accurately measured by the pump) can be assumed to be the amount excreted in the urine. There would be no need to collect urine. The measurement of inulin can be automated.
Finally, in cirrhotic patients being considered for kidney transplantation alone or after LT, the use of transvenous renal biopsy should be helpful, particularly in cases where the nature of renal dysfunction is uncertain and in patients to be transplanted. The percentage of the biopsy core replaced by scar tissue/fibrosis, as an index of chronic damage, has been shown to predict renal outcome in non-liver patients. This could be looked at cirrhotics as well.71 The use of newer serum markers, such as cystatin C, are promising (Table 3) and may be particularly suitable for accurate assessment of renal function in patients with liver disease before or after LT, and could substitute for Cr in prognostic scores. Cystatin-based GFR estimates have been shown to perform equally well or better than Cr-based ones, and therefore could be useful in patients with high bilirubin for calculation of GFR.64
Declaration of personal and funding interests: None.