Advances in Therapy

, Volume 31, Issue 1, pp 30–43

Effect of Metformin on Ballooning Degeneration in Nonalcoholic Steatohepatitis (NASH): When to Use Metformin in Nonalcoholic Fatty Liver Disease (NAFLD)

Authors

  • Iliana Doycheva
    • Division of Gastroenterology, Department of MedicineUniversity of California, San Diego
    • Division of Gastroenterology, Department of MedicineUniversity of California, San Diego
    • Division of Epidemiology, Department of Family and Preventive MedicineUniversity of California, San Diego
Review

DOI: 10.1007/s12325-013-0084-6

Cite this article as:
Doycheva, I. & Loomba, R. Adv Ther (2014) 31: 30. doi:10.1007/s12325-013-0084-6

Abstract

The key histologic feature of nonalcoholic steatohepatitis (NASH) is hepatocellular ballooning (HB). It plays an important role in NASH progression and is an independent predictor of liver mortality. In this review, we identified all studies using metformin in the treatment of nonalcoholic fatty liver disease (NAFLD) that included pre- and post-treatment liver biopsies. We specifically reviewed the effects of metformin on HB. Improved HB was noted in pediatric populations and in those adult patients who were able to lose weight and improve or normalize transaminases during therapy. Previous studies have supported the beneficial effects of metformin in reduction of body weight, improvement of insulin resistance, prevention of complications related to diabetes and chemo-preventive benefits in reducing hepatocellular carcinoma. All these effects make it an attractive treatment consideration for patients with diabetes, and prediabetes who have co-existing NAFLD. Future studies are warranted in order to confirm this effect of metformin on HB and its association with improving long-term outcomes in patients with NAFLD.

Keywords

Body mass indexHepatocellular ballooningLiver histologyMetforminNonalcoholic fatty liver diseaseNonalcoholic steatohepatitisType 2 diabetes mellitus

Introduction

The increasing prevalence of nonalcoholic fatty liver disease (NAFLD) parallels the continuously growing epidemic of obesity worldwide. Depending on the assessment tools, the prevalence of NAFLD in adults ranges between 20% and 30% [1], reaching up to 46% in some studies [2]. Furthermore, NAFLD has become the most common cause of chronic liver disease in children and adolescents, with a prevalence of 3–10%, and even up to 70% in obese children [1, 3]. NAFLD encompasses two major entities: nonalcoholic fatty liver (NAFL) or simple steatosis and nonalcoholic steatohepatitis (NASH). The diagnosis of NASH rests upon histologic evidence of hepatocyte injury in the presence of steatosis and lobular inflammation with or without varying degrees of fibrosis [4]. Patients with NASH, but not those with simple steatosis, can progress to cirrhosis with its related complications, including hepatocellular carcinoma (HCC) [5]. Thus, accurate differentiation of these two forms is imperative in order to risk stratify patients and start timely surveillance and management.

At present, liver biopsy remains the gold standard for diagnosis of NASH, and hepatocellular ballooning (HB) is a key histologic finding. Ballooned hepatocytes are enlarged with pale rarefied cytoplasm and frequently contain Mallory–Denk bodies (MDB) (Fig. 1a) [6]. The most important mechanisms implicated in the pathogenesis of HB are intermediate cytoskeleton alterations, accumulation of small droplet fat, and endoplasmic reticulum dilatation [6]. Marked reduction or absence of immunohistochemically detectable cytoplasmic keratin 8/18 facilitates identification of these cells [7, 8]. The first classification system of NAFLD by Matteoni et al. [9] used hepatocyte necrosis to differentiate between advanced and mild disease. Patients who had hepatocellular injury, with or without MDB or fibrosis, progressed more frequently to cirrhosis and experienced a higher rate of liver-related death [9]. Other authors have confirmed the role of HB and MDB as independent risk factors for fibrosis and disease progression as well [10]. The principal role of HB in the progression of NASH poses the question for therapeutic approaches that can potentially ameliorate or treat it.
https://static-content.springer.com/image/art%3A10.1007%2Fs12325-013-0084-6/MediaObjects/12325_2013_84_Fig1_HTML.jpg
Fig. 1

a Pre-treatment liver biopsy illustrating ballooning injury (large cell in the center of the field) with moderate steatosis (H & E, ×600). b Pre-treatment liver biopsy illustrating prominent perisinusoidal fibrosis (Masson, ×400). c Post-treatment liver biopsy showing zone 3 steatosis without ballooning (H & E, ×600). d Post-treatment liver biopsy showing residual perisinusoidal fibrosis (Masson, ×400)

Although currently there is no established treatment for NAFLD, most of the therapeutic strategies target insulin resistance (IR) as the most important underlying pathophysiological mechanism of NAFLD and NASH [11]. In the setting of IR and hyperinsulinemia, adipocytes release high level of free fatty acids (FFA) into the liver, muscles and pancreatic beta cells due to increased adipose tissue lipolysis [12]. Thus, intrahepatic lipid accumulation exceeds secretion capacity and promotes hepatic steatosis. Mitochondrial dysfunction plays a central role in the pathogenesis of NAFLD along with altered expression of proinflammatory cytokines and accelerated production of reactive oxygen species (ROS) [13, 14]. Oxidation of FFA leads to accumulation of lipid intermediates such as ceramides and diacylglycerols, which are potent inhibitors of insulin signaling in both liver and muscle and contribute to the development of IR in these tissues [15]. Hepatic IR is a strong predisposing factor for metabolic abnormalities, independent of body mass index (BMI) and visceral fat mass [16]. However, the question whether NAFLD is a cause or a consequence of metabolic dysfunction remains unanswered.

Different factors have been implicated to play a role in NAFLD progression. It has been suggested that increased adrenergic activity may be involved in the pathogenesis and progression of NAFLD [17]. Patients with diabetes mellitus are at increased risk of NASH and advanced fibrosis as well as liver-related complications [18]. In addition, family history of diabetes, especially in non-diabetics, has been associated with NASH and fibrosis [19].

Metformin, a biguanide derivative, is an insulin-sensitizing agent. It is the first-line oral therapy for treatment of type 2 diabetes mellitus (T2DM) recommended by both the American Diabetes Association and the European Association of the Study of Diabetes [20, 21]. Several animal and human studies have demonstrated the beneficial effect of metformin on biochemical or histological features of NAFLD [2227]. This effect is primarily related to the reduction of hepatic glucose production, mainly attributed to inhibition of gluconeogenesis and alteration of the main steps of fatty acid metabolism by inhibiting adipose tissue lipolysis, increasing hepatic fatty acid oxidation, and inhibiting lipogenesis in the liver [28, 29].

The aim of this review was to summarize all adult and pediatric studies that included follow-up liver biopsies after treatment of NAFLD with metformin, and to provide guidelines for the best target subpopulation of patients with NAFLD in which metformin has the greatest benefit.

Methods

A PUBMED search was performed to identify relevant publications between the year 2000 and August, 2013 using the words “metformin”, “nonalcoholic fatty liver disease”, “nonalcoholic steatohepatitis”, “treatment”, or “histology” in different combinations. All reviews, randomized control trials (RCTs), open-label, single-arm or randomized studies were reviewed. Studies were included only if they met the following selection criteria: (1) included treatment of NAFLD with metformin, (2) included pre- and post-treatment liver biopsies, (3) there was a separate group treated with metformin only, (4) both pediatric and adult studies were included and (5) published as a full manuscript in English. Titles and abstracts of identified studies were evaluated against eligibility criteria. Four studies were not included as they did not have pre- or post-treatment liver biopsies or metformin was not studied in a separate group. Additional review was performed on the results of those that reported HB as a separate histologic feature.

All included studies were reviewed for design (duration of study, treatment groups, medication doses, type of histological score used, number of patients with follow-up liver biopsies), clinical (change in weight and BMI), laboratory [change in aminotransferases and IR measured by homeostasis model assessment of insulin resistance (HOMA-IR)], and histologic (overall and HB) outcomes. Studies included adult and pediatric patients with or without diabetes. We summarized our finding in the results section. This review article is based on previously conducted studies, and does not involve any new studies of human or animal subjects performed by any of the authors.

Results

Eleven studies were found [24, 25, 27, 3037] that satisfied the search criteria. All of them involved patients with biopsy-proven NAFLD and had histological evaluation after treatment with metformin. Two of the studies involved pediatric patients only [32, 33]. Most of the studies were small, open-label, single-arm studies, or randomized pilot trials that used 1–2 g/day of Metformin and only one study used a higher dose of 2.5–3 g/day (Table 1). HB, as a separate feature of NASH, was assessed in seven studies [27, 3033, 36, 37]. Two of them reported improvement in hepatocellular injury compared to baseline [27, 32]. In the largest pediatric study, HB improved compared to controls [33]. The study design of all 11 papers reviewed is summarized in Table 1.
Table 1

Summary of studies’ design

Author

Year

Country

Design

Duration (months)

Patients with T2DM (%)

# Patients in the treatment group

# Patients in the control group

Treatment group medication and dose

Control/placebo group

Hist. scoring system used

# Patients with follow-up LB in treatment group

# Patients with follow-up LB in control group

Nair et al. [35]

2004

USA

Open-label single-arm

12

6.7

15

NA

MF 20 mg/kg/day

NA

Brunt

10

NA

Uygun et al. [24]

2004

Turkey

Open-label randomized

6

None

17

17

MF 850 mg bid + diet

Diet

Brunt

13

10

Bugianesi et al. [25]

2005

Italy

Open-label randomized

12

11 diet

55

28 Vitamin E

MF 2 g/day

Vitamin E 800 IU/day/diet

Brunt, NAI

17

None

11 Vitamin E

27 diet

18 MF

de Oliveira et al. [36]

2008

Brazil, USA

Open-label single-arm

12

80

20

NA

MF 850–1,000 mg/day + NAC 1.2 g/day

NA

NAS

20

NA

Idilman et al. [30]

2008

Turkey

Open-label randomized

12

None

24 MF

25

MF 850 mg bid/rosiglitazone 8 mg/day

Diet

Brunt, NAS

10 MF

8

25 Rosigl.

11 Rosigl.

Nobili et al. [32]a

2008

Italy

Open-label observational

24

None

30

30

MF 22 mg/kg/day

Placebo

NAS

12

15

Shields et al. [31]

2009

USA

Open-label randomized

12

None

9

10

MF 500–1,000 mg/day

Placebo

Brunt, NAS

9

7

Haukeland et al. [37]

2009

Norway

RCT

6

33 placebo

24

24

MF 2.5–3 g/day

Placebo

NAS

20

24

20 MF

Loomba et al. [27]

2009

USA

Open-label single-arm

12

27

28

NA

MF 2 g/day

NA

NAI

26

NA

Omer et al. [34]

2009

Turkey

Open-label randomized

12

All DM or IGT

22 MF

NA

MF 1.7 g/day/rosiglitazone 4 mg/day/MF 1.7 g/day + rosiglitazone 4 mg/day

NA

NAS

10 MF

NA

20 Rosigl.

13 Rosigl.

22 MF + Rosigl.

12 MF + Rosigl.

Lavine et al. [33]a

2011

USA

RCT

24

None

57 MF

58

MF 500 mg bid + placebo/58 Vitamin E 400 IU/day + placebo

Placebo + placebo

NAS

50 MF

47

58 Vitamin E

50 Vitamin E

The table includes author and year of publication, design and duration of the study, histologic scoring system used and number of follow-up biopsies for all of the 11 studies reviewed

IGT impaired glucose tolerance, LB liver biopsy, MF metformin, NA not available, NACN-acetylcysteine, NAI nonalcoholic steatohepatitis (NASH) activity index, NAS nonalcoholic fatty liver disease (NAFLD) activity score, RCT randomized controlled trial, Rosigl. rosiglitazone, T2DM type 2 diabetes mellitus

aPediatric study

Effect of Metformin in Non-Diabetic Patients

Five of the eleven studies were conducted in non-diabetic patients [24, 3033]. Uygun et al. [24] randomized 36 patients with NASH and treated with diet alone or diet plus metformin for 6 months. The diet plus metformin group showed a significant improvement in BMI (p = 0.01), aspartate aminotransferase (AST) (p = 0.0001), alanine aminotransferase (ALT) (p = 0.003), and HOMA-IR index (p = 0.001) in comparison to control group. There was a slight decrease in necroinflammatory activity, but the result did not reach statistical significance. Idilman et al. [30] divided 74 patients with NASH into three groups: diet alone, diet plus metformin or diet plus rosiglitazone, a thiazolidinedione (TZD) insulin sensitizer, for 48 weeks. The diet plus metformin group showed statistically significant weight loss, decrease in ALT, and HOMA-IR compared to baseline. However, there was no histological improvement, and direct comparison between groups was not performed. Shields et al. [31] conducted a small study of 19 patients with NASH, randomized to diet alone or diet plus low-dose metformin (0.5–1 g daily) for 12 months. No significant difference in weight loss, IR, ALT, or histology was found between the groups.

Effect of Metformin in Pediatric Population

Two pediatric trials have been conducted [32, 33]. In an open-label, observational trial, Nobili et al. [32] assessed the effect of diet plus metformin in children ages 9–18 years with NAFLD (either NASH or NAFL on biopsy) for 2 years. At the end of the study, there was no difference in biochemical parameters and weight loss between the groups. In all patients, weight loss was associated with improved ALT and HOMA-IR. Compared to baseline, a significant reduction was noted in steatosis, lobular inflammation, HB, and NAS, but all parameters decreased equally in both groups. The second study was the largest pediatric study performed to date. It was a randomized, multicenter, double-blinded, placebo-controlled trial evaluating 173 patients between the ages of 8 and 17 years with NAFLD (either NASH or NAFL on biopsy) who received vitamin E, metformin, or placebo for 96 weeks (TONIC) [33]. None of the groups showed any sustained reduction in ALT level, though notably, metformin improved HB in 41% of patients vs. 21% in the placebo group (p = 0.02). This change was not associated with weight loss or improved insulin sensitivity. Among children on vitamin E, NASH resolution was greater than placebo (p = 0.006), which was mainly attributed to improvement in HB. None of the above-described studies reported any significant adverse events with the use of metformin in non-diabetic patients.

Effect of Metformin in Diabetic and Mixed Population

Only one of the eleven studies was conducted solely on patients with T2DM or impaired glucose tolerance [34]. Omer et al. [34] performed a study of 64 patients with NAFLD activity score (NAS) of 5 or more on liver biopsy (correlates with presence of NASH and provides a tool to assess treatment response in the setting of a clinical trial) randomized to metformin, rosiglitazone, or combination metformin plus rosiglitazone for 12 months. The only beneficial effect in the metformin group was a substantial reduction of BMI. There was no significant change in transaminases, insulin sensitivity, or NAS. There was, however, a significant effect on all these parameters in the rosiglitazone and combination groups. The findings of Omer et al. [34] are in concordance with the results from another study [38] which aimed to compare the effects of rosiglitazone and metformin on liver fat content (assessed by proton spectroscopy), insulin clearance, and insulin action in diabetic patients. Metformin increased hepatic insulin sensitivity without changing liver fat content and ALT level, while rosiglitazone increased insulin clearance and serum adiponectin level and reduced liver fat content and ALT.

The remaining five studies were conducted on a mixed patient cohort, including both diabetic and non-diabetic patients [25, 27, 3537]. Three of these studies were open-label, single-arm studies [27, 35, 36]. Nair et al. [35] conducted a small study of 15 patients with NAFLD (both NASH and NAFL on biopsy) who were treated with metformin for 1 year. Only ten participants underwent post-treatment biopsy. An improvement in HOMA-IR was noted at 3 months of therapy, which corresponded to a decline in AST and ALT levels. However, over the course of the study, insulin sensitivity plateaued and transaminases gradually increased again. Histological changes were very mild and HB was not assessed separately. De Oliveira et al. [36] studied 20 patients with NASH given a combination of metformin and N-acetylcysteine (NAC) for 12 months. There was a significant reduction in ALT (p = 0.034) and HOMA-IR (p = 0.0006), but not in BMI. Liver steatosis and fibrosis improved, but there was no improvement in HB and lobular inflammation. The effects observed in this study could probably be attributed to metformin only, as NAC did not show significant beneficial effect on liver enzymes in human [39] or animal [40] studies. Loomba et al. [27] studied the effect of metformin for 48 weeks in 28 patients with NASH. Overall, the greatest changes in histology were for hepatocellular injury (p < 0.001) and parenchymal inflammation (p = 0.03) (see Fig. 1). Eight out of twenty-six patients (31%) achieved a three-point decrease in NASH activity index (NAI) [a histologic score consisting of parenchymal inflammation (0–4) ballooning (0–4), and steatosis (0–4) ranging from 0 to 12], with a decrease in at least two of the components and no worsening of fibrosis or increase in MDB. All responders had a BMI <40 kg/m2 and were more likely to have a lower BMI (although not statistically significant) than nonresponders. Histologic responders were more likely to lose weight during therapy (p = 0.02) and have reduction or normalization of ALT, although the latter was not significant. Decrease of NAI showed a strong correlation with percent weight loss during treatment (r = 0.79, p < 0.0001) and with degree of improvement in ALT (r = 0.6, p < 0.01), but there was no correlation with improved insulin sensitivity.

The last two studies were randomized trials [25, 37]. Bugianesi et al. [25] studied 110 patients with NAFLD (either NASH or NAFL on biopsy) randomized to metformin or control (either vitamin E or diet) for 12 months. There was no difference in weight loss between the groups. However, patients on metformin showed a significant decrease in ALT levels (p < 0.0001), and metformin use was associated with a higher rate of aminotransferase normalization in multivariate analysis [odds ratio (OR) 5.98, p = 0.0011]. Follow-up biopsies were performed in only 17 (out of 55) of the participants on metformin, all of whom had NASH at baseline. This showed improved steatosis (p = 0.0004), necroinflammation (p = 0.012), fibrosis (p = 0.012), and NASH index (p < 0.0001). The beneficial effect was present even in those with no biochemical improvement and it did not correlate with weight loss.

Haukeland et al. [37] examined 48 patients with NAFLD (NASH or NAFL on biopsy) randomized to metformin or placebo for 6 months. There was a significant decrease in BMI (p < 0.001) in the metformin group, but there was no difference in ALT, HOMA-IR, and histology between groups. The metformin group showed a trend towards decrease in HB (p = 0.058), but this reduction was not statistically significant.

Eight of the eleven studies [24, 27, 30, 31, 3335, 37] enrolled obese patients with a mean BMI ranging between 30 and 34 kg/m2. Only two of the adult studies [25, 36] and one of the pediatric studies [32] included patients with mean baseline BMI in the overweight range. Interestingly, the studies of Loomba et al. [27] and Lavine et al. [33] where improvement in ballooning degeneration on metformin was noted had patients with higher mean BMI (~34 kg/m2) than the rest of the studies, where BMI ranged between 30 and 32 kg/m2. However, in the pediatric study of Nobili et al. [32], where mostly overweight patients were enrolled, also observed HB improvement in the metformin group although similar effect was noted in the control group as well.

A summary of the eleven studies of metformin in NAFLD is provided in Table 2. Most of the studies associated metformin with weight loss, improved insulin sensitivity or ALT reduction to baseline or to controls [24, 25, 30, 31, 37]. Metformin treatment showed histologic benefit in five studies [25, 27, 32, 33, 36]. Three studies observed improvement in HB [27, 32, 33] while two did not find any difference [30, 31] (Table 3). It is important to note that the studies of Loomba et al. [27] and Lavine et al. [33], where HB improvement was present, had the largest number of follow-up liver biopsies in the treatment groups.
Table 2

Effect of metformin on ALT, histology, BMI and insulin sensitivity

Study

Change in ALT compared to baseline

Change in ALT compared to controls

Change in histology

HB assessed

Change in HB compared to baseline

Change in HB compared to controls

Change in BMI compared to baseline

Change in BMI compared to controls

Change in HOMA compared to baseline

Change in HOMA compared to controls

Nair et al. [35]

NA

Weak effect

No

NA

NA

NA

NA

Uygun et al. [24]

No

NA

NA

Bugianesi et al. [25]

Improved steatosis, necroinflammatory, fibrosis, NASa

No

NA

NA

de Oliveira et al. [36]

NA

Steatosis and fibrosis improveda; HB and inflammatory—no

Yes

NA

NA

NA

Idilman et al. [30]

Yes

NA

NA

NA

Nobili et al. [32]b

Improved lobular inflammatory, steatosis, HB and NASa

Yes

Shields et al. [31]

Yes

Haukeland et al. [37]

Yes

Loomba et al. [27]

NA

Improved HB, inflammatory, NAIa

Yes

NA

NA

NA

Omer et al. [34]

NA

No

NA

NA

NA

NA

Lavine et al. [33]b

Improved HB

Yes

The results in this table are compared to baseline and to controls where possible

ALT alanine aminotransferase, BMI body mass index, HB hepatocellular ballooning, HOMA homeostatic model assessment, NA not available, NAI nonalcoholic steatohepatitis (NASH) activity index, NAS nonalcoholic fatty liver disease (NAFLD) activity score

→, no significant change; ↓, significant improvement

aCompared to baseline

bPediatric study

Table 3

Effect of metformin on hepatocellular ballooning

Study

# Patients in MF group

Mean HB score at baseline

Mean HB score at the end of treatment

p value (end of treatment vs. baseline)

p value (treatment vs. control group)

Idilman et al. [30]

10

2

2

>0.05

NA

Nobili et al. [32]a

12

1.27

0.55

0.008

Similar decrease in the control group, but no p value

Shields et al. [31]

9

1.69

1.47

NA

0.967

Loomba et al. [27]

26

2.1

1.0

<0.001

NA

Lavine et al. [33]a

50

0.8

0.5

NA

0.02

HB hepatocellular ballooning, MF metformin

aPediatric study

Discussion

Based on these results, the literature suggests that metformin is likely to be beneficial in pediatric populations, as a reduction of HB was recorded in both pediatric studies [32, 33]. The studies reviewed here also suggest that metformin can be beneficial in adult patients with BMI <40 kg/m2 who during therapy tend to lose more weight (>4–5 kg) and experience a decrease or normalization of liver enzymes [27]. Similar conclusions regarding weight loss as an important mechanism through which metformin exerts its effect were reached in an analysis of the Diabetes Prevention Program (DPP) study, a large randomized clinical trial comparing the effects of exercise and diet with either metformin or placebo in patients with prediabetes [41]. Throughout the mean follow-up period of 3.2 years, the metformin group had a slightly lower ALT level, but this effect disappeared when adjusted for changes in weight, fasting insulin, and glucose. Weight loss was the only significant predictor of a 4-year cumulative incidence for development of abnormal ALT in both placebo and metformin groups [41]. The group assigned to lifestyle intervention was not included in this analysis due to lack of follow-up transaminases.

It is important to note that other therapeutic strategies including vitamin E, pioglitazone, and lifestyle interventions also improve HB. In the study by Promrat et al. [42], ≥7% weight loss via lifestyle modifications improved all histologic parameters, including HB. Thiazolidinediones, such as pioglitazone and vitamin E are other currently available options for treatment of NASH. TZDs, similar to metformin, improve insulin sensitivity in liver, adipose tissue and muscles. Two RCTs comparing the effect of pioglitazone vs. placebo in diabetics or patients with IGT [43] and in non-diabetics [44] showed improvement in metabolic and all histologic parameters. Another small single-arm study on the effect of pioglitazone in non-diabetic patients with NASH also showed improvement in all histologic features of NASH, including fibrosis [45]. The largest trial to date on the effect of TZDs in patients with biopsy-proven NASH is Pioglitazone, Vitamin E, or placebo for Nonalcoholic Steatohepatitis (PIVENS) trial [46]. The primary outcome was improvement in histology and decrease in HB score was part of the requirements. Although pioglitazone did not meet the pre-specified significance level for the primary outcome and the percent of subjects with HB improvement was not statistically significant (p = 0.08), greater proportion of patients on pioglitazone (47%) vs. placebo (21%) had complete resolution of steatohepatitis on end-of-treatment biopsy (p = 0.001). The failure of pioglitazone group to meet the primary outcome was explained by a disproportionate misclassification of the presence of HB on enrollment in this group (28%) vs. 17% in the placebo group and 18% in vitamin E group. Vitamin E therapy, however, was associated with significantly higher rate of improvement in HB and all NASH components (43% vs. 19%, p = 0.001). In contrast, a study using rosiglitazone in patients with NASH for 1 year showed improvement in steatosis only [47]. The major side effect of TZD treatment in all studies was weight gain, which ranged between 3 and 5 kg and persisted after cessation of therapy.

The findings of this review of 11 studies [24, 25, 27, 3037] must be interpreted cautiously due to several methodological limitations of the studies that were included, which prevent us from drawing definite conclusions. First, this is a literature review, and a meta-analysis was not performed. Second, most of the studies were small, pilot trials with even smaller numbers of patients who underwent a follow-up liver biopsy assessment. Of the 11 studies, three did not have a control group, this meant that a direct comparison was not possible [27, 35, 36]. The clinical implications of biochemical and histological changes in placebo-treated patients with NASH have been previously quantified in a pooled analysis of five randomized clinical trials [48]. The result showed that placebo-treated patients also demonstrate a decline in serum ALT levels. However, while changes in steatosis, inflammation, fibrosis and overall NAS may be seen in 22–33% of placebo-treated patients, improvement in HB is only noted in 15% of the patients. These data suggest that HB improvement is less likely to be seen with chance alone. Similarly, in the TONIC trial, HB improved less often in the placebo group than other histological parameters [33]. These findings support the idea that significant improvements in HB are more likely a true treatment or weight loss effect. Third, inclusion criteria varied between studies with some of them involving only patients with biopsy-proven NASH [24, 27, 31, 34, 36] while others consisted of patients with either simple steatosis or NASH at baseline histology [32, 35, 37]. The proportion of diabetic patients (who may have more advanced liver disease) also differed between studies (Table 1). Fourth, the duration of treatment varied between 6 and 24 months, and the metformin dose ranged between 1 and 3 g. Although there is no established dose for the treatment of NAFLD, higher doses are probably more likely to induce weight loss, reduce IR, and thus, more likely to improve liver histology. Two of the eleven studies that used a higher dose of metformin did note a significant weight reduction [27, 37]. The average weight loss was 4 kg with a metformin dose of 2.5–3 g/day for 6 months [37], and 6 kg weight loss was seen with a metformin dose of 2 g/day for 1 year [27]. It remains uncertain whether the beneficial effects of metformin persist after discontinuation of therapy. Although two of the studies reported stable weight and biochemical parameters after discontinuation of treatment [24, 30], others have reported weight gain and increase in aminotransferases soon after stopping therapy [27]. Based on these data, studies with large sample size using higher doses of metformin and for longer duration of treatment are needed to truly assess and quantify the beneficial effects of metformin in NASH.

Another possible limiting factor that should be considered when assessing HB in general is the sampling error and intra- and interobserver variability with liver biopsy. In a study with 51 patients who underwent liver biopsy with two samples collected, the discordance rate for the presence of HB was as high as 18%. Ballooning would have been missed in 24% of cases if only one biopsy had been performed [49].

All eleven studies [24, 25, 27, 3037] reported only mild gastrointestinal side effects and just two of them [34, 37] had one patient who dropped out due to intolerability of these side effects. There were no episodes of lactic acidosis detected and treatment was safe in both adults and children.

In summary, there is preliminary evidence that metformin may be beneficial in patients with NAFLD especially those with early diabetes, and further studies are needed to confirm whether these beneficial effects are related to improvement in HB. Metformin remains attractive due to its safety profile and long-term chemo-preventive anti-cancer effects in reducing (HCC) risk. Whether higher doses of metformin for longer duration of therapy would increase response rates in patients with NASH remains to be tested. In particular, metformin might be useful in children with type 1 NASH (adult type), as those with type 2 or borderline zone 1 NASH lack this characteristic feature [50]. Children with metabolic syndrome would also benefit from treatment with metformin as large pediatric studies with biopsy-proven NASH found metabolic syndrome (OR 2.01, p = 0.03), central obesity (OR 2.15, p = 0.01), and IR (OR 0.9, p = 0.01) to be predictive for HB [51]. Similarly, based on histological data from the NASH Clinical Research Network in adults, the diagnosis of definite NASH and HB was associated with female gender, T2DM, metabolic syndrome, elevated AST, ALT, gamma-glutamyl transpeptidase, HOMA-IR and hemoglobin A1C [52]. These risk factors for HB in conjunction with the results from the studies that observed improvement in HB in adults [27] suggest that metformin would be a valuable option in patients with NAFLD and concomitant metabolic syndrome or T2DM, especially if they demonstrate a decrease in body weight and transaminases during therapy. In addition, if metformin is well tolerated, we recommend a dose of at least 2 g daily to be used for the treatment of NAFLD.

Interestingly, a recent large case–control study revealed that use of metformin is associated with a decreased risk of HCC in diabetics in a dose-dependent manner [53]. The authors found that each incremental year of use resulted in a 7% reduction in HCC risk. These results suggest that metformin may have a desirable beneficial effect in patients with NAFLD, especially when it is associated with T2DM, as these patients are at increased risk for more advanced liver disease that can eventually lead to cirrhosis and HCC [5].

Conclusion

In conclusion, current literature suggests that metformin therapy in children and adults with NASH is likely to be associated with improvement in HB, a key histologic feature of NASH. This effect of metformin in improvement of HB is associated with weight loss and correlates with improvement in serum ALT. Patients who lose ≥5% of body weight and achieve normalization of serum ALT are likely to benefit and show improvement in liver histology. In order to achieve this effect, a dose of at least 2 g a day and sufficient duration of treatment may be needed. The additional beneficial effects of metformin in improving IR, reducing the risk of incident diabetes and HCC prevention make it an attractive treatment consideration for patients with prediabetes or diabetes with NASH who are at high risk of developing liver-related complications. Future larger randomized, controlled trials evaluating the effects of metformin on long-term clinical outcomes in patients with NASH are warranted especially assessing its role in HCC chemoprevention. Histological and biochemical markers of hepatocyte apoptosis such as immunohistochemical staining for cytokeratin 8/18 and detection of its fragments in plasma should be further studied in order to confirm the effect of metformin on HB, and to establish a reliable noninvasive marker to follow treatment response. However, with metformin being out of patent, it is less likely that any company or private institution will invest in a large RCT. Considering this limitation, available evidence is suggestive, and further investigator-initiated studies are needed especially in assessing chemo-preventive benefits of metformin in patients with NAFLD for reducing long-term HCC risk.

Acknowledgments

Authors are thankful to Dr. David Kleiner (NCI/NIH) who provided the liver histology images shown in this paper and Dr. Mamie Dong for her assistance in manuscript editing. No funding or sponsorship was received for this study or publication of this article.

Conflict of interest

Dr. Iliana Doycheva and Dr. Rohit Loomba declare they have no conflict of interest.

Compliance with ethics guidelines

This review article is based on previously conducted studies and does not involve any new studies of human or animal subjects performed by any of the authors.

Supplementary material

12325_2013_84_MOESM1_ESM.pdf (169 kb)
Supplementary material 1 (PDF 168 kb)

Copyright information

© Springer Healthcare 2013