Introduction

Multiple metabolic disorders, such as diabetes [1], insulin resistance and dyslipidaemia [2], are associated with non-alcoholic fatty liver disease (NAFLD), ranging from simple fatty liver to non-alcoholic steatohepatitis (NASH). Steatosis of the liver is closely associated with insulin resistance. However, the toxic lipids are not intrahepatic triacylglycerols but, rather, it is non-esterified cholesterol [3, 4] and some NEFA [5] that contribute to inflammation and insulin resistance in hepatocytes.

The level of cholesterol is tightly regulated by endogenous synthesis in the liver and dietary absorption/biliary reabsorption in the small intestine. Niemann–Pick C1-like 1 (NPC1L1) plays a pivotal role in cholesterol incorporation in enterocytes [6]. Ezetimibe, a potent inhibitor of cholesterol absorption, inhibits NPC1L1-dependent cholesterol transport at the brush border of the intestine and the liver [6]. This suggests that ezetimibe ameliorates toxic-lipid-induced inflammation and insulin resistance by inhibiting cholesterol absorption. Indeed, ezetimibe improves liver steatosis and insulin resistance in mice [7] and Zucker obese fatty rats [8], although the beneficial effects of ezetimibe are observed only when the animals are fed a high-fat diet. Ezetimibe can also ameliorate liver pathology in patients with NAFLD [9, 10]; however, these studies lack a control group, which precludes meaningful conclusions as liver pathology can improve over the natural course of the disease or with tight glycaemic control in some NAFLD patients [1]. In the present study, we investigated the efficacy of ezetimibe treatment in patients with NAFLD for 6 months in an open-label randomised control study by examining liver pathology, as well as hepatic enzymes, glucose metabolism, hepatic fatty acid composition and hepatic gene expression profiles.

Methods

Patient selection

Study staff recruited participants from outpatients at Kanazawa University Hospital, Ishikawa, Japan. Patients were recruited from April 2008 to August 2010, with follow-up visits during the 6 months thereafter. The study lasted from April 2008 to February 2011.

The inclusion criterion was a biopsy consistent with the diagnosis of NAFLD. Exclusion criteria included hepatic virus infections (hepatitis C virus [HCV] RNA–PCR-positive, hepatitis B and C, cytomegalovirus and Epstein–Barr virus), autoimmune hepatitis, primary biliary cirrhosis, sclerosing cholangitis, haemochromatosis, α1-antitrypsin deficiency, Wilson’s disease, history of parenteral nutrition and use of drugs known to induce steatosis (e.g. valproate, amiodarone and prednisone) or hepatic injury caused by substance abuse and/or the current or past consumption of more than 20 g of alcohol daily. None of the patients had any clinical evidence of hepatic decompensation, such as hepatic encephalopathy, ascites, variceal bleeding or an elevated serum bilirubin level more than twofold the upper normal limit.

A random allocation sequence was computer-generated elsewhere and assigned participants in a 1:1 ratio to treatment with ezetimibe or to the control group. All patients and responsible guardians underwent an hour of nutritional counselling by an experienced dietitian before starting the 6 month treatment period. The experienced dietitians were unaware of the study assignments. In addition, all patients were given a standard energy diet (125.5 kJ/kg per day; carbohydrate 50–60%, fat 20–30%, protein 15–20%) and exercise (5–6 metabolic equivalent estimations for 30 min daily) counselling before the study. Patients remained on stable doses of medications for the duration of the study. The patients in the ezetimibe group received generic ezetimibe (10 mg/day; Zetia, [Merck, Whitehouse Station, NJ, USA]) for 6 months.

The study was conducted with the approval of the Ethics Committee of Kanazawa University Hospital, Ishikawa, Japan, in accordance with the Declaration of Helsinki. Written informed consent was obtained from all individuals before enrolment. This trial is registered with the University Hospital Medical Information Network (UMIN) (Clinical Trials Registry, no. UMIN000005250).

Primary and secondary outcomes

The primary endpoint was change in serum alanine aminotransferase (ALT) level at month 6 from baseline. Secondary outcomes included changes in the histological findings for NAFLD, hepatic gene expression profiling, fatty acid compositions of plasma and liver biopsy samples, lipid profiles, insulin resistance and anthropometric measures, as well as assessment of ezetimibe safety. We had planned to enrol 80 patients in the trial, as we had estimated that with this sample size, the study would have 90% power at an α (two-tailed) value of 0.05 showing a 50% decrease of serum ALT values with 6 months of pioglitazone therapy on the basis of a previous study [11]. At the time of adverse event analyses, 32 of the targeted 80 patients had been randomly assigned and were included in the safety analyses.

Data collection

Clinical information, including age, sex and body measurements, was obtained for each patient. Venous blood samples were obtained after the patients had fasted overnight (12 h) and were used to evaluate blood chemistry. Insulin resistance was estimated by HOMA-IR, calculated as [fasting insulin (pmol/l) × fasting glucose (mmol/l)]/22.5 [12] and insulin sensitivity was estimated as the quantitative insulin sensitivity check index (QUICKI)[13]. The adipose tissue insulin resistance index (adipose IR) was calculated as fasting NEFA (mmol/l) × fasting insulin (pmol/l) [1416]. The indocyanine green retention rate at 15 min after venous administration (ICG15) was assessed using standard laboratory techniques before and after treatment. Serum fatty acids were measured with a gas chromatograph (Shimizu GC 17A, Kypto, Japan) at SRL (Tokyo, Japan).

Evaluation of insulin sensitivity derived from an OGTT

After an overnight fast (10–12 h), a 75 g OGTT was performed at 08:30 hours. The OGTT-derived index of beta cell function, the insulinogenic index, computed as the suprabasal serum insulin increment divided by the corresponding plasma glucose increment in the first 30 min (ΔI30/ΔG30) [15, 17, 18] was calculated. From the OGTT data, the Matsuda index [19] was calculated. The hepatic insulin resistance index (H-IR) was calculated as the product of the total AUCs for glucose and insulin during the first 30 min of the OGTT (glucose 0–30 [AUC] [mmol/l] × insulin 0–30 [AUC] [pmol/l]). Skeletal muscle insulin sensitivity can be calculated as the rate of decline in plasma glucose concentration divided by plasma insulin concentration, as follows. Muscle insulin sensitivity index = dG/dt/mean plasma insulin concentration, where dG/dt is the rate of decline in plasma glucose concentration and is calculated as the slope of the least square fit to the decline in plasma glucose concentration from peak to nadir [20]. See the electronic supplementary material (ESM) for further details.

Evaluation of insulin sensitivity derived from the euglycaemic insulin clamp

Insulin sensitivity in 23 of the 31 patients (10 control and 13 ezetimibe patients) was also evaluated in a hyperinsulinaemic–euglycaemic clamp study [21]. Patients did not receive any medication on the morning of the examination. At ∼09:00 hours, after an overnight fast of at least 10 h, an intravenous catheter was placed in an antecubital vein in each individual for infusion, while a second catheter was placed in the contralateral hand for blood sampling. The euglycaemic–hyperinsulinaemic clamp technique was performed using an artificial pancreas (model STG-22; Nikkiso, Tokyo, Japan), as described previously [22]. See ESM for further details. The mean glucose metabolic clearance rate (MCR) in healthy individuals (n = 9; age, 26.6 0± 2.9 years; body mass index, 22.3 ± 2.1 kg/m2) was 13.5 ± 3.4 mg kg−1 min−1 [2].

Liver biopsy pathology

A single pathologist, who was blinded to the clinical information and the order in which the biopsies were obtained, analysed all biopsies twice and at separate times. The sections were cut from a paraffin block and stained with haematoxylin and eosin, Azan–Mallory and silver reticulin impregnation. The biopsied tissues were scored for steatosis (from 0 to 3), stage (from 1 to 4) and grade (from 1 to 3) as described [2], according to the standard criteria for grading and staging of NASH proposed by Brunt et al [23]. The NAFLD activity score (NAS) was calculated as the unweighted sum of the scores for steatosis (0–3), lobular inflammation (0–3) and ballooning (0–2), as reported by Kleiner et al [24].

Gene expression analysis of liver biopsied samples

Gene expression profiling was performed in samples from nine patients in the ezetimibe group and six in the control group. Liver tissue RNA was isolated using the RNeasy Mini kit (QIAGEN, Tokyo, Japan) according to the manufacturer’s instructions. See ESM for further details. Data files (CEL) were obtained using the GeneChip Operating Software 1.4 (Affymetrix). Genechip data analysis was performed using BRB-Array Tools (http://linus.nci.nih.gov/BRB-ArrayTools.html). The data were log-transformed (log10), normalised and centred. To identify genetic variants, paired t tests were performed to define p values <0.05 and fold change > 1.5. Pathway analysis was performed using MetaCore (GeneGo, St Joseph, MI, USA). Functional ontology enrichment analysis was performed to compare the gene ontology (GO) process distribution of differentially expressed genes (p < 0.01).

Fatty acid composition of liver

Aliquots (0.2 mg) of liver samples snap-frozen by liquid nitrogen were homogenised in 1 ml normal NaCl solution (NaCl 154 mmol/l). Briefly, fatty acids were extracted by using pentadecanoic acid, and saponified with alkaline reagent (0.5 mmol/l KOH/ CH3OH). The fatty acid methyl esters were analysed in a gas chromatograph (Shimadzu GC-2014 AF/SPL; Shimadzu Corporation, Kyoto, Japan) equipped with a flame ionisation detector and an auto injector. See ESM for further details. Mass spectra were analysed using GC solution (v. 2.3) software (Shimadzu Corporation, Kyoto, Japan, www.shimadzu.com). The changes in hepatic fatty acid composition are expressed as 10−4 mg/mg liver.

Statistical analysis

Data are expressed as mean ± one standard error, unless indicated otherwise. The Statistical Package for the Social Sciences (SPSS; version 11.0; Chicago, IL, USA) was used for the statistical analyses. For univariate comparisons between the patient groups, Student’s t test or Mann–Whitney’s U test was used, as appropriate, followed by the Bonferroni multiple-comparison test. A value of p < 0.05 was considered to indicate statistical significance.

Results

Enrolment and discontinuation

The data and safety monitoring board recommended that the study intervention and enrolment be discontinued because of the higher proportion of adverse events (significant elevation in HbA1c) in the ezetimibe group than in the control group. At the time of adverse event analyses, 32 of the targeted 80 patients had been randomly assigned and were included in the safety analyses. In our open-label trial, 32 patients with NAFLD were enrolled. They were randomised to treatment with ezetimibe (n = 17) or a control (n = 15) with no significant clinical differences in variables between the groups. Of the 32 randomly assigned patients, 31 had completed the 6 month intervention period; one patient dropped out of the study. One case in the control group withdrew consent after randomisation and before intervention (ESM Fig. 1). The patient who withdrew was excluded from analysis because he did not start his course of treatment. Two analyses were conducted in the remaining patients. In the intention-to-treat analysis (ESM Tables 1 and 2), measures that were missing for participants who discontinued the study were replaced with baseline measures. In the second analysis, the only data included were from participants who completed the study to the end of the 6 month follow-up period. We performed a completed case analysis because there were few dropouts unrelated to baseline values or to their response.

Patient characteristics

The 31 study patients (mean age 52.7 ± 2.1 years; mean BMI 29.2 ± 1.0) included 14 randomised to the control group and 17 to the ezetimibe group (ESM Table 3).

At baseline, the characteristics of patients in the ezetimibe and control groups were comparable except for the waist circumference (p = 0.085) and the Matsuda index (p = 0.060). The histological features of the liver are summarised in Table 1. At baseline, neither the severity of the individual histological features nor the proportion of patients distributed in the three NAS categories was significantly different between the two groups. All 31 participants agreed to complete the follow-up venous blood samples including OGTT. The ICG15 was conducted in 24 patients (ten control and 14 ezetimibe patients).

Table 1 Histological characteristics of the livers of patients who completed the study at baseline and 6 months

Changes in laboratory variables

The primary study outcome, serum alanine aminotransferase levels, did not change after ezetimibe treatment (Table 2).

Table 2 Laboratory values, insulin sensitivity and insulin resistance derived from the euglycaemic insulin clamps and OGTTs of patients who completed the study at baseline and 6 months

After 6 months of ezetimibe treatment, systolic blood pressure, HbA1c, glycated albumin, and lathosterol were significantly increased, while total cholesterol levels, campesterol, sitosterol and ferritin were significantly decreased. In contrast, body weight, BMI, fasting plasma glucose, plasma γ-glutamyltransferase, triacylglycerols, HDL-cholesterol, small dense LDL (sdLDL), remnant-like particle cholesterol (RLP-C), type IV collagen 7 s levels, NEFA, total bile acid, high-sensitivity C-reactive protein (hsCRP), adiponectin, TNF-α, plasminogen activator inhibitor-1 (PAI-1), 8-isoprostanes and ICG15 did not change after ezetimibe treatment (Table 2). Adipose IR tended to increase in the ezetimibe group (from 88.1 ± 25.5 to 107.5 ± 25.5, p = 0.070), but not in the control group.

When changes in the groups were compared, the ezetimibe group, but not the control group, had a significant decrease in total cholesterol (ezetimibe, −0.49 ± 0.19 vs control, 0.06 ± 0.14 mmol/l; p = 0.037), whereas the ezetimibe group, but not control group, showed a significant elevation in HbA1c (ezetimibe, 0.46 ± 0.12% [4.95 ± 1.28 mmol/mol] vs control, 0.08 ± 0.13% [0.78 ± 1.46 mmol/mol]; p = 0.041). Also, there were significant differences between the groups in cholesterol and HbA1c levels at 6 months. The multiple-comparison Bonferroni test revealed highly significant differences in the changes in total cholesterol (p = 0.037) and HbA1c (p = 0.040) between the ezetimibe and control groups.

Increased concentrations of the cholesterol synthesis markers lathosterol (ezetimibe, 1.49 ± 0.32 nmol/l vs control, 0.58 ± 0.14 nmol/l; p = 0.018) and decreased concentrations of the cholesterol absorption markers campesterol (ezetimibe, −1.28 ± 0.41 nmol/l vs control, 1.88 ± 0.54 nmol/l, p = 0.000) and sitosterol (ezetimibe, −0.91 ± 0.27 nmol/l vs control, 0.85 ± 0.45 nmol/l; p = 0.002) were observed on treatment. The ezetimibe group had an increase, whereas the control group had a decrease, in the level of 8-isoprostanes (ezetimibe, 11.6 ± 6.4 pmol/mmol creatinine vs control, −19.9 ± 12.9 pmol/mmol creatinine; p = 0.031).

When changes between groups were compared, the ezetimibe group had a greater decrease in the Matsuda index (ezetimibe = −0.78 ± 0.57 vs control = −1.35 ± 0.55, p = 0.013), QUICKI (ezetimibe = −0.02 ± 0.01 vs control = 0.03 ± 0.0, p = 0.019), and muscle insulin sensitivity (ezetimibe = −0.002 ± 0.004 vs control = 0.019 ± 0.014, p = 0.067) than the control group.

Changes in liver histology

Twenty-eight of 31 participants, 16 in the ezetimibe group and 12 in the control group, agreed to complete the follow-up and undergo a liver biopsy at 6 months, allowing for complete case analysis of the data (Table 1). After 6 months, the changes in staging score (from 1.75 ± 0.28 to 1.53 ± 0.26) and ballooning score (from 0.69 ± 0.20 to 0.41 ± 0.15) were significantly improved in the ezetimibe group compared with the control group, whereas the scores of steatosis, lobular inflammation and NAS were not significantly changed in either group. The degree of all of these histological features was not significantly different between the two groups (Table 1).

Serial changes in liver gene with ezetimibe treatment

Gene expression profiling was conducted in samples from nine patients in the ezetimibe group and six in the control group (ESM Table 4). In the ezetimibe group, 434 genes were upregulated and 410 genes downregulated, while in the control group, 643 genes were upregulated and 367 genes downregulated. Pathway analysis of the process network of differentially expressed genes showed coordinate downregulation of genes involved in skeletal muscle development and cell adhesion molecules in the ezetimibe group, suggesting a suppression of stellate cell development into myofibroblasts (Table 3). In addition, ezetimibe activated the immune response pathway. In contrast, genes involved in skeletal muscle development were upregulated and those in the immune response downregulated in the control group (Table 4). Pathway analysis of the metabolic network also revealed decreased l-carnitine pathway and increased steroid metabolism with ezetimibe treatment, but decreased CoA biosynthesis and increased glycerol 3-phosphate pathway in the control group (ESM Fig. 2).

Table 3 Signalling pathway gene expression changes in the ezetimibe group
Table 4 Signalling pathway gene expression changes in the control group

Changes in plasma fatty acid composition and fatty acid composition extracted from liver tissue

The changes in plasma fatty acid composition are shown in Table 5. Compared with baseline levels, only eicosatrienoic acid was significantly increased in the ezetimibe group.

Table 5 Changes in plasma fatty acid composition

Fatty acid composition in extracted liver tissue was available for 16 NAFLD patients treated with ezetimibe and 12 controls (Table 6). Ezetimibe treatment for 6 months significantly and markedly increased hepatic lauric, myristic, palmitic, palmitoleic, margaric and stearic acids compared with the control group. The changes in hepatic fatty acid composition did not correlate with the changes in serum fatty acid composition before and after ezetimibe treatment (ESM Table 5).

Table 6 Changes in hepatic fatty acid composition

Discussion

This is the first report of the efficacy of ezetimibe treatment on liver pathology in patients with NAFLD in an open-label randomised controlled trial. Treatment with 10 mg/day ezetimibe for 6 months did not alter the primary study outcome, serum aminotransferase levels. Ezetimibe significantly decreased serum cholesterol levels and cholesterol absorption markers as expected, whereas, in contrast to previous reports, ezetimibe treatment did not decrease serum levels of triacylglycerol. Our initial hypothesis was that ezetimibe treatment ameliorates liver pathology by inhibiting the absorption of toxic lipids such as oxidised cholesterol and palmitate. In our animal model, cholesterol feeding to mice increased not only cholesterol but also triacylglycerols in the liver, and upregulated the gene for sterol regulatory element binding protein (SREBP)-1c that governs fatty acid synthesis [3], probably via activation of liver-X-receptor (LXR) in the liver [25]. Therefore, in experimental models of high-cholesterol-diet-induced steatohepatitis, ezetimibe ameliorated liver steatosis by reducing cholesterol-induced activation of LXR and SREBP-1c [26, 27]. In the present study, however, treatment with ezetimibe unexpectedly ameliorated liver fibrosis staging and ballooning scores without significantly changing hepatic steatosis and insulin resistance.

One possible explanation for the improvement of hepatic fibrosis by ezetimibe treatment may be related to the direct effect of cholesterol on hepatic fibrogenesis. The cholesterol molecule affects membrane organisation and structure, which are critical determinants of membrane bilayer permeability and fluidity [28]. Altered cholesterol metabolism has several toxic effects on hepatocytes, resident macrophages, Kupffer cells and hepatic stellate cells, which promote NASH through diverse mechanisms. Hepatic stellate cells, in particular, are responsible for liver fibrosis in NASH. It has recently been reported that intracellular cholesterol accumulation directly activates hepatic stellated cells through a toll-like receptor-4-dependent pathway and triggers hepatic fibrosis [29]. These effects might be more evident in humans because, unlike rodents, where NPC1L1 is primarily expressed in the intestine, in humans NPC1L1 mRNA is highly expressed both in the small intestine and liver. Therefore, ezetimibe is estimated to inhibit not only dietary and biliary cholesterol absorption through the small intestine, but also reabsorption of biliary cholesterol in the liver [30, 31]. Thus, ezetimibe may inhibit liver fibrosis by ameliorating cholesterol-induced activation of hepatic stellate cells in patients with NAFLD. This hypothesis was well supported by the hepatic gene expression profile induced by ezetimibe administration. Ezetimibe treatment coordinately downregulated genes involved in skeletal muscle development and cell adhesion molecules, suggesting that ezetimibe suppressed stellate cell development into myofibroblasts and thereby inhibited fibrogenesis.

Another important finding of the present study was that treatment with ezetimibe significantly deteriorated glycaemic control. Ezetimibe therapy also altered the hepatic profile of fatty acid components by significantly increasing hepatic levels of lauric, myristic, palmitic, palmitoleic, margaric, stearic, oleic and linoleic acids. Experimentally, palmitate induces interleukin-8 [32], endoplasmic reticulum stress, and c-Jun amino-terminal kinase activation and promotes apoptosis in the liver [5, 33, 34]. Lipid-induced oxidative stress and inflammation are closely related to insulin resistance [3, 5], which could be relevant to the ezetimibe-induced deterioration of glucose homeostasis. Indeed, urinary excretion of 8-isoprostanes was significantly increased in the ezetimibe group compared with the control, and showed significant negative correlation with insulin sensitivity indices such as the Matsuda index and QUICKI in the present study (ESM Table 6). Moreover, hepatic gene expression in the ezetimibe group showed coordinated upregulation of genes involved in the immune response compared with those in the control group, suggestive of oxidative stress caused by ezetimibe treatment.

Pathway analysis of the metabolic network showed unique metabolic changes in the ezetimibe group compared with the control group. In the control group, genes involved in the CoA-biosynthesis pathway were coordinately downregulated, and those in the glycerol-3 phosphate pathway coordinately upregulated, suggesting activated triacylglycerols biosynthesis. In the ezetimibe group, genes involved in the l-carnitine pathway, including CPT1A, were coordinately downregulated. A decreased l-carnitine pathway could be associated with reduced β-oxidation of palmitic acids in mitochondria, resulting in an increase in long-chain fatty acids (lauric, myristic, palmitic, palmitoleic, margaric, stearic, oleic and linoleic acids). Unbalanced fatty acid composition could induce oxidative stress and lead to insulin resistance in the ezetimibe group. In addition, genes involved in the cholesterol and NEFA biosynthesis, including SREBF2, were coordinately upregulated in the ezetimibe group (Table 3), probably as a result of deceased absorption of exogenous cholesterol. Upregulation of SREBF2 potentially represses the expression of hepatocyte nuclear factor 4, which is required for CPT1 transcription [35]. Moreover, recent reports have demonstrated that microRNA (miR)-33, encoded by an intron of Srebp2 [36], inhibits translation of transcripts involved in fatty acid β-oxidation, including CPT1 [37]. miR-33 is also implicated in decreased insulin signalling by reducing insulin receptor substrate-2 [38, 39]. Hepatic gene expression profiles may, to some extent, explain hepatic fatty acid composition and impaired glycaemic control in the ezetimibe group. These novel SREBP-2-mediated pathways in the gene expression network may be relevant to a recent report that a polymorphism in the SREBF2 predicts incidence and the severity NAFLD and the associated glucose and lipid dysmetabolism [15]. These unique hypotheses should be confirmed in future in vitro and in vivo studies.

Our study has some limitations. First, the number of patients is relatively small because the data and safety monitoring board recommended that the study intervention and enrolment be discontinued in light of the higher proportion of adverse events in the ezetimibe group than in the control group. Second, our trial was a 6 month open-label study that resulted in subtle changes in liver pathology compared with previous reports [40]. Indeed, a 6 month duration may be too short a period to expect improvement of fibrosis, which is a slowly progressive process [40]. Third, the average serum aminotransferase levels were lower than those in previous studies [9, 10], and most of the patients had mild steatosis, fibrosis and lower NAS at baseline before ezetimibe treatment. Serum ALT levels did not decrease with ezetimibe treatment in the present study, in contrast to the significant improvement reported previously [9, 10]. And finally, secondary outcomes are always at risk of false-positive associations. Therefore, we not only presented the changes in HbA1c (p = 0.001 for ezetimibe treatment and p = 0.041 for the intergroup difference at the end of the study), but also showed the signature of hepatic fatty acid composition and hepatic gene expression profiles that support the hypothesis that ezetimibe increases HbA1c and hepatic fatty acids contents possibly through the SREBP-2–miR33 pathway. No previous studies have raised this issue, which is worth investigating. The same mechanism may underlie a statin-induced deterioration of glucose tolerance, which remains a serious concern. Furthermore, the SREBP-2–miR33 pathway may raise a concern for a safety issue of combination therapy with ezetimibe and statins because these agents may additively upregulate SREBF2 expression [41]. Future large-scale, long-duration studies involving more severely affected patients are required to determine the definite efficacy and risks of ezetimibe in the treatment of NAFLD.

In conclusion, the present study represents the first randomised controlled clinical trial of the efficacy of ezetimibe on liver pathology, energy homeostasis, hepatic fatty acid composition and hepatic gene expression profiles in patients with NAFLD. The lipid profile and liver histology of cell ballooning and fibrosis were significantly improved by ezetimibe treatment. However, our findings suggest an increase in oxidative stress, insulin resistance and HbA1c on treatment with ezetimibe, which should be taken into consideration in NAFLD patients.