Diarmaid D. Houlihan and Ian A. Rowe are recipients of clinical research training fellowships from the Medical Research Council. Matthew J. Armstrong is the recipient of a clinical research training fellowship from the Wellcome Trust.
The association of nonalcoholic fatty liver disease (NAFLD) with type 2 diabetes mellitus, obesity, hyperlipidemia, and cardiovascular disease is well recognized.1 NAFLD is an independent risk factor for cardiovascular disease and cardiovascular disease–related death.2-4 Emerging data now suggest that NAFLD plays an important role in the genesis of chronic kidney injury (CKI) in patients with diabetes mellitus. In a large, prospective study, 1760 patients with type 2 diabetes mellitus were followed for an average duration of 6.5 years for the development of CKI. More than 30% of these patients developed CKI during the follow-up, and NAFLD was found to be an independent risk factor for this.5 Subsequently, a cross-sectional study by the same authors demonstrated similar findings in patients with type 1 diabetes mellitus.6 The mechanisms of NAFLD-induced renal dysfunction have yet to be elucidated.
Nonalcoholic steatohepatitis (NASH) cirrhosis has become a more frequent indication for liver transplantation (LT) over the last decade, yet the impact of LT on renal function in this at-risk group is not known. This has important clinical implications because the presence of CKI is a well-recognized determinant of long-term morbidity and mortality post-LT.7 Furthermore, it is now clear that early renal dysfunction after transplantation predicts long-term CKI,8, 9 and this has prompted some groups to introduce a variety of different immunosuppression regimens immediately after LT.10-12 Indeed, a recent large, multicenter randomized controlled trial demonstrated significantly less nephrotoxicity with a delayed calcineurin inhibitor (CNI) regimen versus the current standard of care.13
With the knowledge that the majority of patients with NASH cirrhosis also have metabolic disease at the time of LT, we undertook this study to examine the impact of LT on renal function in NASH patients, and here we compare their outcomes to those of a matched cohort of patients undergoing transplantation for other disease etiologies.
BMI, body mass index; CKI, chronic kidney injury; CNI, calcineurin inhibitor; eGFR, estimated glomerular filtration rate; LT, liver transplantation; MDRD, Modification of Diet in Renal Disease; MELD, Model for End-Stage Liver Disease; MMF, mycophenolate mofetil; NA, not applicable; NAFLD, nonalcoholic fatty liver disease; NASH, nonalcoholic steatohepatitis.
PATIENTS AND METHODS
A retrospective review of the case notes for adult patients who underwent deceased donor LT for NASH between January 2000 and December 2008 was undertaken. The following patients were included in the NASH group: (1) patients with a clinical and histological diagnosis of NASH14 established by liver biopsy before LT or by an explant examination (n = 37) and (2) patients with a histological diagnosis of cryptogenic cirrhosis established by an explant examination and a clinical phenotype compatible with underlying NASH, which was defined by the presence of 3 or more components of metabolic syndrome before LT (n = 11).15 In combination with the histopathological analysis, an extensive drug and alcohol history (excess alcohol consumption was defined as >30 g of alcohol/day for males and >20 g of alcohol/day for females) and a full serological screen (viral, genetic, metabolic, and autoimmune) were used to confidently exclude other etiologies of chronic liver disease. We also excluded patients with advanced CKI (stage IV or V) before LT and those listed for combined liver-kidney transplantation.
To benchmark the outcomes of our NASH cohort against a general LT population, we looked at patients who underwent transplantation over the same time period in the same tertiary center. The patients in the comparison group were selected from our transplant database and were matched to the NASH patients in a 1:1 ratio by age, sex, Model for End-Stage Liver Disease (MELD) score, and estimated glomerular filtration rate (eGFR) at the baseline. Only patients who underwent transplantation within 6 months of their comparators were selected for the study. Investigators were blinded to the post-LT clinical parameters, morbidity outcomes (including renal problems), and mortality outcomes when the comparison group was being selected. The patients in the non-NASH group had the following disease etiologies: primary sclerosing cholangitis (n = 10), primary biliary cirrhosis (n = 9), hepatitis B virus infection (n = 3), alcoholic liver disease (n = 11), hepatitis C virus infection (n = 6), alpha-1-antitrypsin deficiency (n = 2), hemochromatosis (n = 3), both hemochromatosis and alcoholic liver disease (n = 3), and autoimmune hepatitis (n = 1).
All subjects were started on a standard immunosuppression protocol that included a CNI (cyclosporine for 5 patients and tacrolimus for 91 patients), azathioprine, and prednisolone. Tacrolimus was introduced postoperatively at a dose of 0.1 to 0.15 mg/kg/day by mouth, which was subsequently adjusted to achieve target whole blood trough levels of 10 ng/mL for the first month post-LT. Prednisolone was discontinued in all patients at 3 months according to the protocol. In patients who developed progressive renal dysfunction on this regimen, mycophenolate mofetil (MMF) was introduced to facilitate a dose reduction of the CNI according to our unit's protocol. In those cases, MMF (1 g twice daily) was substituted for azathioprine, and this was followed by a tacrolimus dose reduction to 0.05 to 0.10 mg/kg/day, which was adjusted to achieve target trough levels <8 ng/mL. Such changes in immunosuppression for progressive renal dysfunction were recorded for this study.
The glomerular filtration rate was estimated with the 4-variable Modification of Diet in Renal Disease (MDRD) equation.16 Diabetes mellitus was confirmed with preexisting medical records or with 2 random venous plasma glucose measurements ≥11.1 mmol/L. Patients were considered to have hypertension if they had a preexisting diagnosis or blood pressure readings >135/85 mm Hg on more than 2 occasions in the clinic. Pre-LT cardiac dysfunction was defined as an ejection fraction <50% on echocardiography. Intraoperative variables, including the type of operation, the caval clamp time, the red blood cell transfusion requirements, the occurrence of hypotension (defined as a mean arterial pressure <60 mm Hg for >30 minutes or as a need for bolus vasopressors),17 and the occurrence of cardiac arrest, were recorded for patients who developed an acute kidney injury after LT. Acute kidney injuries were classified with the Acute Dialysis Qualitative Initiative criteria.18 The creatinine concentration on the day of LT (before the operation) was used as a baseline reference value. CKIs were staged according to the US Kidney Disease Outcomes Quality Initiative guidelines.19 Clinical and laboratory data were obtained from standard clinic visits post-LT and hospitalizations. These data included whole blood trough levels of tacrolimus and cyclosporine.
For comparisons of baseline demographics, continuous variables were compared with paired t tests or Wilcoxon tests (as applicable), and categorical variables were compared with Fisher's exact test. The differences in the eGFR values of the NASH and non-NASH patients were then analyzed with a 2-stage process. The first stage of the analysis considered the effects of NASH on eGFR values 3 months after LT. To take into account the pairing of the patients, we first calculated the difference between the eGFR values of each pair of patients at each time point. The value for this variable 3 months after LT was then entered as the dependent variable in a general linear model, and the difference in the eGFR values at the baseline was entered as a predictor. The intercept of the resulting model represented the difference in the eGFR values at 3 months after adjustments for any differences within the pairs at the baseline (Supporting Table 1).
In the second step, a repeated measures analysis (a generalized estimating equation) was used to adjust the differences in potential confounding factors between the groups and to test whether the differences varied over time. First, we subtracted the paired difference in the eGFR values at the baseline from the difference at each time point so that we could adjust for any differences in the baseline eGFR values within the pairs of patients. The result was used as the dependent variable in a model, and the time point at which the eGFR measurements were taken (3, 6, 12, or 24 months) was included as a factor. To take into account other potential confounding factors (not matched at the baseline), we expanded the analysis to adjust for differences in the body mass index (BMI), hepatocellular carcinoma status, hypertension status, diabetes mellitus status, and post-LT serum tacrolimus levels between the groups. We made the adjustments for the BMI and tacrolimus levels (continuous variables) by finding the differences in these variables within each pair at the baseline and then including these factors in the model. For hypertension, diabetes mellitus, and hepatocellular carcinoma (categorical variables), the pairs were divided into 4 mutually exclusive groups at the baseline: NASH patients positive for each variable, NASH patients negative for each variable, non-NASH patients positive for each variable, and non-NASH patients negative for each variable (Supporting Table 2). All analyses were performed with SPSS 19 (IBM SPSS, Inc.). P values less than 0.05 were considered significant.
The pre-LT characteristics of the NASH and non-NASH groups are presented in Table 1. The study population was almost exclusively white and Caucasian (>95%). The mean BMI immediately before LT was significantly greater in the NASH group versus the non-NASH group (29.66 versus 26.0 kg/m2, P < 0.001). Both diabetes mellitus (72.9% versus 29.2%) and hypertension (37.5% versus 4.2%) were significantly more common in the NASH group versus the non-NASH group (P < 0.001; Table 1). The follow-up was longer for the non-NASH group, although this was not statistically significant [the median follow-up was 52 months (range = 1-115 months) for the NASH group and 63 months (range = 1-128 months) for the non-NASH group; P = 0.08].
Table 1. Clinical Demographics Immediately Before LT
The data are expressed as means and ranges.
The patients were matched by the parameter at the baseline.
Seventy-three of the 96 study patients (76%) had an eGFR ≥ 60 mL/minute/1.73 m2 before LT. Sixteen of the 23 patients with an eGFR < 60 mL/minute/1.73 m2 had either diabetes mellitus or hypertension, which was deemed to be the cause of their renal impairment. Four of the remaining 7 patients (1 NASH patient and 6 non-NASH patients) had diuretic-induced renal dysfunction, 1 patient was diagnosed with renal stones and outflow obstruction, and 1 patient from each group was diagnosed with hepatorenal syndrome. The duration of significant renal dysfunction (creatinine level > 114 μmol/L) was previously shown to predict renal outcomes post-LT.20 We identified 13 patients who met these criteria. The median duration of significant renal dysfunction before LT was 35 days (range = 5-120 days) for the NASH patients (n = 7) and 53 days (range = 15-135 days) for the non-NASH patients (n = 6).
Acute Kidney Injury After LT (≤1 Month)
There was deterioration in renal function in both cohorts after LT. The median values at each time point after LT are highlighted in Fig. 1. In all, 38 of 96 patients (24 NASH patients and 14 non-NASH patients) developed significant renal dysfunction within 1 month of LT according to the criteria proposed by the Acute Dialysis Qualitative Initiative. Eleven of these 38 patients (7 NASH patients and 4 non-NASH patients) developed kidney function failure (baseline creatinine level × 3). The remaining 27 patients (17 NASH patients and 10 non-NASH patients) sustained a kidney injury (baseline creatinine level × 2). Eight patients in the NASH group and 4 patients in the non-NASH group required renal dialysis in the postoperative period. Cardiac dysfunction, the type of operation, red blood cell transfusion requirements, and intraoperative hypotension have previously been identified as risk factors for an acute kidney injury after LT.17 To check for an imbalance in these factors between the NASH and non-NASH patients, we retrospectively reviewed the notes of the 38 patients who developed significant post-LT renal dysfunction. We identified 9 patients who had a retrohepatic caval resection with a venovenous bypass; the average clamp times were 75 and 90 minutes for the NASH (n = 5) and non-NASH patients (n = 4), respectively. The remaining patients (n = 29) underwent piggyback LT. Only 1 patient in the NASH group had cardiac dysfunction before LT. One patient in each group experienced cardiac arrest perioperatively and received bolus vasopressors. The operative red blood cell requirements were similar in the 2 groups (4 U for NASH patients and 5 U for non-NASH patients). The incidence of early graft nonfunction was assessed through the recording of the peak aspartate aminotransferase levels in NASH and non-NASH patients within the first week of LT. There was no significant difference in the peak aspartate aminotransferase levels of the NASH patients [median = 922 U/L (interquartile range = 582-1460 U/L)] and the non-NASH patients [median = 882 U/L (interquartile range = 472-1853 U/L), P = 0.75].
Persistent Kidney Injury After LT (≥3 Months)
An analysis using general linear models showed that after adjustments for the baseline difference in the eGFR values, NASH patients had eGFR levels that were 8.85 mL/minute/1.73 m2 lower than those of patients in the comparison group 3 months after LT (P = 0.004, 95% confidence interval = 2.93-14.77; Supporting Table 1). The multivariate generalized estimating equation showed that after adjustments for the effects of the BMI, tacrolimus levels, diabetes mellitus, hypertension, and HCC, the difference in the eGFR values remained significant 3 months after LT (8.46 mL/minute/1.73 m2, P = 0.001). The data were then analyzed at numerous time points after LT (6, 12, and 24 months), and the time did not significantly affect the difference between the groups (P = 0.173; Supporting Table 2).
To appreciate the clinical long-term impact of LT on renal function in NASH patients, we summarize the proportions of patients with different stages of renal dysfunction at the baseline and 2 years after LT in Table 2. Notably, 31.2% of the patients in the NASH group progressed to a clinically relevant stage of CKI (stage IIIb) within 2 years, but only 8.3% of the patients in the control group did (P = 0.009). No patients in either group required long-term renal replacement therapy (≥3 months) during the 2-year follow-up period.
Table 2. CKI Stages of the NASH and Non-NASH Groups at the Baseline and 2 Years After LT
CKI Stage (eGFR)
Before LT (%)
2 Years After LT (%)
I (>90 mL/minute/1.73 m2)
II (60-89 mL/minute/1.73 m2)
IIIa (45-59 mL/minute/1.73 m2)
IIIb (30-44 mL/minute/1.73 m2)
IV (15-29 mL/minute/1.73 m2)
Patient Survival and Post-LT Morbidity
There were no significant differences in the actuarial patient survival rates of the NASH and non-NASH groups 1 year after LT (88% versus 86%, P > 0.99) or 5 years after LT (82% versus 82%, P > 0.99). In all, 11 patients in the NASH group and 14 patients in the non-NASH group died during the entire follow-up period. The causes of mortality in the 2 groups are summarized in Table 3. The prevalence of diabetes mellitus increased (albeit insignificantly) in both the NASH group (from 72.9% to 81%, P = 0.34) and the non-NASH group (from 29.2% to 43%, P = 0.14) within 2 years of LT. Three of 13 patients (23.1%) developed new-onset diabetes mellitus in the NASH group, whereas 5 of 34 patients (14.7%) did in the non-NASH group (P = 0.68). Glycemic control also deteriorated in patients with diabetes mellitus: the proportion of patients with hemoglobin A1c levels > 7% increased in both the NASH group (from 13% to 31.4%) and the non-NASH group (from 25% to 54.5%) within 2 years of LT.
Table 3. Causes of Death After LT
Cause of Death
NASH Group (n = 11)
Non-NASH Group (n = 14)
Cardiovascular event (n)
Disease recurrence (n)
More patients in the NASH group were given MMF because of renal dysfunction during follow-up (19 NASH patients and 10 non-NASH patients in the first 3 months after LT, 4 NASH patients and 2 non-NASH patients between 3 and 24 months after LT, and 9 NASH patients and 4 non-NASH patients more than 2 years after LT). Despite this observation, there were no significant differences in the mean serum levels of tacrolimus between the NASH and non-NASH groups at 3 months (8.4 versus 7.7 ng/mL, P = 0.14) and 2 years (8.0 versus 8.1 ng/mL, P = 0.95).
This single-center study highlights the fact that patients who underwent transplantation for NASH cirrhosis had worse renal outcomes after LT than a strictly matched group of patients with other chronic liver disease etiologies. This difference between the groups was pronounced at 3 months and did not change significantly for at least 2 years after LT. A multivariate analysis that allowed for the effects of confounding factors showed that NASH was an independent risk factor for post-LT renal dysfunction (P = 0.001). In agreement with this, more NASH patients (31.2%) than non-NASH patients (8.3%) progressed to a more advanced stage of renal dysfunction (glomerular filtration rate = 30-44 mL/minute/1.73 m2) within 2 years. Acute LT graft dysfunction (which was reflected by peak aspartate aminotransferase levels) did not contribute to the differences in postoperative renal dysfunction between the groups. Similarly, the frequencies of pre-LT cardiac dysfunction and perioperative hypotension, the caval clamping times, the red blood cell requirements, and the periods of significant baseline renal dysfunction were similar for the groups and were thus unlikely to be responsible for the differences in renal function. To the best of our knowledge, this is the first study that has identified NASH as an independent risk factor for early and persistent renal dysfunction after LT.
Although there are currently no other data focusing on the impact of NAFLD on renal function after LT, accumulating evidence suggests that NAFLD is an independent risk factor for CKI in patients with type 1 or 2 diabetes mellitus.5, 6 This association is independent of the multiple coexisting risk factors that these patients have for CKI (ie, the duration of diabetes, the extent of glycemic control, hypoglycemia, antihypertensive and antiplatelet medications, and components of metabolic syndrome). Furthermore, a recent study of 1361 patients with impaired glucose tolerance according to routine screening demonstrated that the presence of NAFLD on ultrasound was independently associated with the presence of microalbuminuria.21 In 4 large, prospective studies examining the factors that influence the development of renal disease over time, NAFLD was also identified as an independent risk factor for CKI.22 It is currently unclear whether the association between NAFLD and CKI results from shared cardiovascular and metabolic risk factors or specific pathophysiological mechanisms. There are a number of potential mechanisms by which NAFLD may exert an independent and deleterious effect on renal function: (1) persistent levels of low-grade inflammatory mediators from visceral adipose tissue and subsequent oxidative stress; (2) lower adiponectin levels, which lead to inflammatory and profibrotic cascades; (3) hypercoagulation/hypofibrinolysis; (4) atherogenic dyslipidemia; and (5) hepatic insulin resistance and dysglycemia.22 Further studies are required, however, to unravel the specific cascades that link NAFLD and renal dysfunction.
The development of CKI after LT is almost inevitable,23, 24 and its etiology is multifactorial: pretransplant kidney disease, immunosuppression therapies, and patient comorbidities (including hypertension and diabetes mellitus) all contribute to the pathogenesis.25 CKI is associated with a dramatic predisposition to cardiovascular events and death in the general population.26 This effect is magnified in the transplant population, and in some studies, an advanced stage of CKI (stage IV) has led to a 4-fold increase in overall mortality versus transplant patients without renal dysfunction.7, 27 Although the pathogenesis of renal dysfunction after LT is multifactorial, therapeutic strategies aimed at minimizing renal injury focus on reducing the exposure of patients to CNIs. Recognized immunosuppression strategies in this context include (1) reducing or withdrawing CNIs after the stable introduction of MMF, (2) switching from a CNI-based regimen to a non–CNI-based regimen (with a sirolimus or rapamycin backbone), and (3) using antibody induction therapy to delay a patient's exposure to CNIs.25 In our study, significantly more NASH patients (66.7%) than non-NASH patients (33.3%) were given MMF post-LT. Notably, more than 50% of the patients who were started on MMF in the NASH group began taking the drug during the first 3 months after LT. During this period, the differences in the eGFR values of the 2 groups were most apparent. In the absence of published data, we can only hypothesize that the early introduction of MMF after LT may have influenced the leveling off in the rate of deterioration of renal function in the NASH patients between 3 months and 2 years after LT (Fig. 1). The fact that a significant difference remained in the eGFR values of the 2 groups 2 years after LT suggests that further changes in immunosuppression regimens are required to optimize post-LT renal function in NASH patients. According to previous trials,13 the delayed introduction of CNIs or regimens entirely free of CNIs may prove to be the most effective strategies for preserving the renal function of NASH patients after LT.
This study has several strengths. This is the first study to examine the impact of NASH on the renal function of patients after LT. We strictly matched in a blinded fashion our NASH cohort to a carefully selected transplant cohort (by age, sex, MELD score, and eGFR at the baseline) to ensure the accuracy of our results. In the non-NASH group, we also included patients with a wide variety of disease etiologies as benchmarks for our comparison because they were more reflective of the general transplant population. The main limitation of this retrospective study is that we were restricted to calculating the eGFR, which depends on a number of patient variables (including muscle mass, body weight, age, sex, and ethnicity). A variety of formulas, including the Cockcroft-Gault and MDRD formulas, have been used in previous studies to monitor post-LT renal function.13, 24 A specific consideration of our study is that the NASH cohort had a significantly greater mean BMI than the non-NASH cohort. In one of the largest studies evaluating the predictive performance of the MDRD and Cockcroft-Gault formulas for the estimation of renal function, the MDRD formula was found to be significantly more accurate over a wide range of BMIs in comparison with the Cockcroft-Gault formula.28 Specifically, in comparison with the renal excretion of 51Cr-labeled ethylene diamine tetraacetic acid, the MDRD formula was found to underestimate the true glomerular filtration rate by approximately 1 to 2 mL/minute/1.73 m2 in patients with BMIs ranging from 18.5 to >30 kg/m2. The majority of the patients enrolled in our study (>95%) had BMIs between 18.5 and 34 kg/m2, and this made the MDRD equation an appropriate choice for calculating the eGFR in this study.
The fact that approximately 20% of the patients in our comparison group underwent LT for primary biliary cirrhosis may have exaggerated the observed differences in the post-LT renal morbidity rates of our NASH and non-NASH groups because patients with primary biliary cirrhosis have the best reported long-term prognosis after LT.29 Urinalysis for the albumin-to-creatinine ratio, which provides a useful clinical and research marker of glomerular function, was not routinely performed and thus was not available for analysis. Finally, it is noteworthy that our sample size was small and that only 3 patients in this study had stage IIIb CKD before LT (Table 2); the remainder had a baseline eGFR ≥ 45 mL/minute/1.73 m2. By excluding the more severe cases of CKI, our study may have underestimated the extent of renal deterioration and associated mortality post-LT. With longer follow-up periods, larger sample sizes, and the inclusion of patients with more advanced renal disease at the baseline, it is possible that differences in mortality and morbidity between NASH patients and patients with comparative etiologies might become more apparent post-LT.
In conclusion, this is the first study demonstrating that patients undergoing LT for NASH develop worse renal function than matched patients with other chronic liver diseases. Renal-sparing immunosuppression regimens should be considered at the time of LT to reduce the development of CKI in NASH patients. The optimization of such regimens requires a prospective study.