Reconsidering a “chopped liver”: The need for improving glomular filtration rate estimation for hepatic transplantation

Authors

  • Chirag R. Parikh M.D., Ph.D,

    Corresponding author
    1. Program of Applied Translational Research, Yale University School of Medicine, New Haven, CT
    2. Section of Nephrology, Yale University School of Medicine, New Haven, CT
    3. Clinical Epidemiology Research Center, West Haven, CT
    • Address reprint requests to: Chirag Parikh, M.D., Ph.D., Section of Nephrology, Yale University and VAMC, Temple Medical Center, 60 Temple Street, Suite 6C, New Haven, CT 06510. E-mail: chirag.parikh@yale.edu; fax: 203-764-8373.

    Search for more papers by this author
  • Justin M. Belcher M.D.

    1. Program of Applied Translational Research, Yale University School of Medicine, New Haven, CT
    2. Section of Nephrology, Yale University School of Medicine, New Haven, CT
    3. Clinical Epidemiology Research Center, West Haven, CT
    Search for more papers by this author

  • See Articles on Pages 1514, 1522, and 1532

  • Potential conflict of interest: Nothing to report.

Abbreviations
CKD-EPI

Chronic Kidney Disease Epidemiology Collaboration

eGFR

estimated GFR

GFR

glomular filtration rate

MDRD

Modification of Diet in Renal Disease Study

MELD

Model of End-Stage Liver Disease

mGFR

measured glomerular filtration rate

SLKT

simultaneous liver-kidney transplants.

Kidney dysfunction is very common in patients with cirrhosis and independently associated with decreased survival. Such dysfunction may be related to structural injury or may be functional and result from a decrease in renal blood flow that can be ameliorated with liver transplant. In light of the independent prognostic effect of kidney dysfunction in patients with cirrhosis, serum creatinine was incorporated into the Model of End-Stage Liver Disease (MELD) score, which was subsequently adopted as the primary means of allocating liver transplants. Because creatinine is heavily weighted in the calculation of MELD, patients with apparent impairment in renal function are necessarily prioritized for transplant. As a result, since the adaptation of MELD, the incidence of simultaneous liver-kidney transplants (SLKT) has increased dramatically, with 425 such transplants performed in the United States in 2013, up 200% from the pre-MELD era.[1] With the demand for transplant organs far outstripping supply, accurate assessment of renal function is therefore critical both for equitable allocation of solitary livers and for complex decisions regarding combined liver-kidney transplants.

The gold standard for assessment of glomerular filtration rate (GFR), direct measurement by insulin clearance, or radiolabeled tracers, is expensive, technically complex and time intensive and therefore not widely utilized. Similarly, renal biopsies are rarely performed in patients with advanced cirrhosis because of concerns over excessive bleeding. In lieu of direct measurement, GFR is typically estimated using one of over five estimation equations. Unfortunately, the most common marker used for estimation of renal function, serum creatinine, is often an unreliable surrogate for GFR because of the effect on its levels of nonrenal determinants, and these shortcomings are significantly magnified in patients with cirrhosis.[2] In recognition of this, patients are considered for SLKT when measured GFR (mGFR) is <30 mL/min, but when estimated GFR (eGFR) is <40.3 The accuracy of creatinine in reflecting GFR declines with worsening stages of cirrhosis. Because of this inexactness, creatinine-based equations for estimating GFR, including the most frequently used (Chronic Kidney Disease Epidemiology Collaboration [CKD-EPI] and Modification of Diet in Renal Disease Study [MDRD]), perform significantly worse in patients with cirrhosis than in other settings.[4] In addition, infusion of albumin and elevation of blood urea nitrogen resulting from gastrointestinal bleeds, tube feeds, and steroids frequently observed in patients with cirrhosis add to the imprecision of MDRD-6. The consequences of inaccurate estimation in this setting can be profound, with potential both for wasting of critically scarce organs and denial of critical transplants.

Acknowledging these limitations, researchers have proposed novel creatinine-based equations intended to increase the accuracy of estimation. In addition, investigators have sought alternatives to creatinine as a marker of glomerular filtration. Cystatin C is a 13.36-kDa cysteine proteinase inhibitor synthesized at a constant rate by all nucleated cells. Cystatin C levels are less influenced by nonrenal factors than creatinine and, especially, in patients with cirrhosis, ought to be a superior marker of glomerular filtration. However, cystatin levels are affected by inflammation, steroid use, and infection, which are prevalent in the setting of cirrhosis. Figure 1 outlines factors potentially compromising of each of these markers for accurate GFR assessment.

Figure 1.

(A) Nonrenal influences of serum creatinine levels. Those factors that are particularly problematic in patients with cirrhosis are highlighted in red. The figure emphasizes that creatinine levels are affected by non-renal determinant which impact its production, measurement and excretion. (B) Nonrenal influences of serum cystatin C levels. Those factors that are particularly problematic in patients with cirrhosis are highlighted in red. Unlike creatinine, non-renal determinants only impact cystatin C levels via alterations in production and volume of distribution.

In this issue of Hepatology, three articles evaluate novel GFR-estimating equations utilizing both creatinine and cystatin C, as compared to traditional equations against the gold standard of mGFR. When assessing their performance, three distinct, yet complimentary, metrics must be taken into consideration: bias; precision; and accuracy. Bias refers to the average of the differences between the eGFR and mGFR across the patients studied and thus can be thought of as the extent to which, on average, an eGFR will diverge from the true mGFR for a given patient. Precision refers to the standard deviation of these differences across a patient population and thus estimates the extent to which the difference between eGFR and mGFR will differ between patients. Finally, accuracy, which can be calculated in various fashions, typically looks at a given threshold for acceptable estimation performance, for instance, within 20% of the mGFR, and calculates what percentage of patients who receive an eGFR fall within this threshold.

Francoz et al.[5] and De Souza et al.[6] studied patients on the wait list for liver transplant comparing the MDRD-4 and -6 and CKD-EPI equations to mGFR. In the study by Francoz et al., MDRD-4 and CKD-EPI overestimated GFR, whereas MDRD-6 underestimated it. MDRD-4 and CKD-EPI overestimation increased in patients with mGFR ≤60 mL/min. Correlation of all three equations with mGFR was poor (R2, 0.37-0.40), but declined markedly in patients with mGFR <40. De Souza et al. also compared cystatin C equations (CKD-EPI cystatin, CKD-EPI creatinine-cystatin, and Hoek) against mGFR. The cystatin-containing equations were, in general, superior to those based on creatinine, demonstrating lower bias and greater precision, and the CKD-EPI cystatin equation performed best, irrespective of cirrhosis severity. Mindikoglu et al.[7] evaluated the performance of the newly proposed CKD-EPI creatinine-cystatin equation in 72 outpatients with cirrhosis against an exhaustive list of existing equations as well as 24-hour urine creatinine clearance. Whereas several equations performed well on individual metrics, CKD-EPI creatinine-cystatin was uniformly strong and, unlike in De Souza et al., outperformed CKD-EPI cystatin. Critically, the precision and accuracy of the CKD-EPI creatinine-cystatin equation was higher when mGFR was <60 mL/min than when it was >60. Despite its impressive performance, CKD-EPI creatinine-cystatin nonetheless was significantly less accurate in patients with cirrhosis than in other settings. Whereas only 22.8% of patients who had eGFR differ from mGFR by >20% and 8.5% by >30% in a noncirrhotic cohort, such inaccuracy occurred in 44% and 24% of patients in this study, respectively. Though comprehensive, the study had 18 exclusion criteria, including patients with hepatocellular carcinoma (who, in fact, made up 41% of the population in the study by Francoz et al.), potentially limiting its generalizability. The results of the three studies are summarized in Table 1 with data aggregated when possible.

Table 1. Performance and Potential Misclassification of Estimation Equationsa
 MDRD-4MDRD-6CKD-EPICKD-EPI CystatinCKD-EPI Creatinine-Cystatin
  1. Difference score (DS) refers to mGFR-eGFR for each subject, bias refers to mean DS, precision refers to standard deviation of DS, and accuracy refers to percentage of eGFRs that were within 30% (P30) or 20% (P20) of mGFR.

  2. Abbreviation: KDOQI, kidney disease outcomes quality initiative.

  3. a

    Weighted averages of composite data, where possible, from Francoz et al., De Souza et al., and Mindikoglu et al.

  4. b

    Absolute value; bias was positive in Francoz et al. and negative in Mindikoglu et al.

  5. c

    P20 provided by Mindikoglu et al., calculated for De Souza et al. as average of P30 and P10.

  6. d

    Percentage of patients where estimation equation incorrectly classified the KDOQI CKD stage, as compared to gold standard of mGFR. In patients with GFR under 60 mL/min, underestimation thus implies primarily that patients with stage 4 CKD (mGFR, 15-30) were estimated as having stage 3 CKD (eGFR, 30-60).

Bias     
GFR ≥60, mL/min−1211.5b−64.75
GFR <60, mL/min−237.6−22−5.11
      
Precision     
GFR ≥60, mL/min27.126.3
GFR <60, mL/min13.712
      
Accuracy, %c     
GFR ≥60     
P3050.766.27086.788.5
P2033.342.748.36265.2
GFR <60     
P3019.246.517.373.148
P2013.534.212.55237
      
KDOQI CKD stage misclassification, %d     
GFR ≥60     
Underestimation37.428.6441024.7
Overestimation3.34.7228.77.3
GFR <60     
Underestimation69.253.873.123.148.1
Overestimation0005.81.9

The consistency of findings and comparatively strong performances of MDRD-6, CKD-EPI cystatin, and CKD-EPI creatinine-cystatin are encouraging. However, not only has the performance of GFR estimation equations typically declined with declining renal function, but it is also in the very setting of significant CKD that these differences become clinically meaningful. Across all three studies, equations based on creatinine consistently overestimated GFR, whereas those based on cystatin underestimated it. It is important to be mindful that all misestimations of GFR are not equally nefarious. Most significant harm occurs when estimation equations erroneously misclassify patients across the 40-mL/min GFR barrier that is the threshold for eGFR-mediated decisions regarding combined liver-kidney transplants. Such damaging misclassification is bidirectional and could occur either with creatinine overestimation or cystatin underestimation. Francoz et al. attempted to estimate the prevalence of such misclassification by identifying what they termed “discordant” patients, where the mGFR was >30 mL/min and eGFR ≤40. Twenty-three of two hundred and ninety (8%) patients fit this definition by MDRD-6 and, were they transplanted based on eGFR, may have inappropriately received an SLKT. The potential wastefulness this would engender is evidenced by the excellent serum creatinine at 2 years of those surviving discordant patients who received only a liver transplant. Alternatively, 2 of 10 (20%) patients with mGFR <30 had eGFR >40, meaning that they could have mistakenly been denied an SLKT. Similarly, the equation that performed best in the study by De Souza et al., CKD-EPI cystatin, misclassified the CKD stage of 23% of patients with mGFR <60 mL/min.

Given the heightened importance of accurate estimation in the setting of depressed GFR, the lack of patients with significant renal dysfunction across these studies is a significant limitation in deciding their clinical utility. In the studies mentioned above, only a minority (<10%) of the patients had GFR <30. Along with being compromised by advanced CKD, the performance of traditional estimation equations declines with progressively more advanced, decompensated cirrhosis.[2] It is, of course, in just such patients where critical decisions regarding transplant must be made. It may well be that the preservation of performance in late-stage cirrhosis is one of the advantages to cystatin C–based equations, as suggested by the results of De Souza et al. However, the strikingly low MELD scores in the majority of included patients prevents this from being definitively concluded.

Despite these limitations, the data regarding estimation equations presented in this issue are important and of immediate use to clinicians. Where cystatin C is not available (sadly, the majority of institutions), MDRD-6 should be used in patients with cirrhosis in an out-patient setting. When cystatin can be measured, CKD-EPI cystatin and CKD-EPI creatinine-cystatin appear to be the best available options. However, the inferior performance of all equations in patients with cirrhosis, relative to other settings, remains frustrating. A large component of this suboptimal showing likely stems from the fact that none of these equations was derived in a population of patients with cirrhosis. Given the particular importance of accurate, precise GFR estimation in this setting, further research to develop a cirrhosis-specific equation is urgently needed. In addition, unaddressed by these new equations is the primacy of importance that the cause of kidney dysfunction has on renal recovery post–liver transplant.[8] It is critical that future equations incorporate elements to distinguish structural from functional disease as well as improving estimation of GFR.[9] Until this time, clinicians faced with critical clinical decisions hinging upon accurate estimation of GFR would do well to rely upon these three equations while remaining cognoscente of their residual limitations.

  • Chirag R. Parikh, M.D., Ph.D.1,2,3 Justin M. Belcher, M.D.1,2,3

  • 1Program of Applied Translational Research Yale University School of Medicine New Haven, CT

  • 2Section of Nephrology Yale University School of Medicine New Haven, CT

  • 3Clinical Epidemiology Research Center VAMC West Haven, CT

Ancillary

Advertisement