Dr J. B. Schwimmer, Department of Pediatrics, Division of Pediatric Gastroenterology, Hepatology and Nutrition, University of California, San Diego, 200 West Arbor Drive, San Diego, CA 92103-8450, USA. E-mail: firstname.lastname@example.org
Background : Children with non-alcoholic steatohepatitis are insulin-resistant and metformin has been proposed as a potential therapy. However, paediatric safety and efficacy data are absent.
Aim : To test the hypothesis that metformin therapy will safely improve markers of liver disease in paediatric non-alcoholic steatohepatitis.
Methods : Single-arm open-label pilot study of metformin 500 mg twice daily for 24 weeks in non-diabetic children with biopsy-proven non-alcoholic steatohepatitis.
Results : Ten obese children (mean body mass index 30.4) enrolled and completed the trial. Mean alanine aminotransferase and aspartate aminotransferase (AST) improved significantly (P < 0.01) from baseline (184, 114 U/L) to end of treatment (98, 68 U/L). Alanine aminotransferase normalized in 40% and AST normalized in 50% of subjects. Children demonstrated significant improvements in liver fat measured by magnetic resonance spectroscopy (30–23%, P < 0.01); insulin sensitivity measured by quantitative insulin sensitivity check index (0.294–0.310, P < 0.05); and quality of life measured by pediatric quality of life inventory 4.0 (69–81, P < 0.01).
Conclusion : Open-label treatment with metformin for 24 weeks was notable for improvement in liver chemistry, liver fat, insulin sensitivity and quality of life. A large randomized-controlled trial is needed to definitively determine the efficacy of metformin for paediatric non-alcoholic steatohepatitis.
Childhood obesity has been described as a pandemic of the new millennium.1 One of the most common consequences of obesity in children, non-alcoholic fatty liver disease (NAFLD), is estimated to occur in 1–10% of children and adolescents in the industrialized world.2 The diagnosis of NAFLD represents a spectrum of liver pathology characterized by the accumulation of large droplet fat within hepatocytes in the absence of alcohol exposure. Histologically the features may range from fat alone to the more worrisome findings of non-alcoholic steatohepatitis (NASH), advanced fibrosis and cirrhosis. A diagnosis of NASH requires a liver biopsy and is characterized as steatosis in conjunction with inflammation and liver cell injury with or without fibrosis.3 The natural history of paediatric NASH is unknown but cirrhosis does develop in some children,4 and in adults is recognized as an important cause of liver transplantation and hepatocellular carcinoma.5 Although lifestyle changes and some pharmacological interventions have shown promise in pilot studies no treatment has been established for NASH.
Hepatic steatosis can be associated with multiple causes including medications, metabolic disorders and toxin exposures. However, the most common condition associated with fatty liver is insulin resistance.5 Furthermore, insulin resistance is present in the majority of children and adolescents with biopsy-proven NAFLD.6 Thus, insulin resistance is a rational therapeutic target and pilot studies of therapy for NASH in adults have tested medications used for the treatment of type 2 diabetes mellitus.7–9 Metformin reduces hepatic glucose production and increases insulin sensitivity (SI) in patients with type 2 diabetes mellitus. Safety and efficacy have been established for metformin in the treatment of type 2 diabetes for paediatric subjects aged 10–16 years.10 An open-label pilot study in adults with NASH demonstrated that short-term treatment with metformin improved SI and serum alanine aminotransferase (ALT).7
Paediatric NASH often demonstrates histopathological differences from adult NASH in the degree and location of fat, inflammation and fibrosis.11 Furthermore, children often demonstrate different pharmacokinetics than adults. Thus, one cannot assume that adult data regarding treatment will extrapolate to children. We therefore conducted an open-label phase 2 clinical trial in non-diabetic children and adolescents with biopsy-proven NASH to assess safety and test the potential for efficacy. Our primary hypothesis was that treatment with metformin would decrease markers of disease activity. Methods novel for treatment trials in paediatric hepatology including magnetic resonance (MR) spectroscopy and quality of life (QoL) assessment were utilized.
Selection of subjects
We sought children with biopsy-proven NASH representative of typical patients who had persistent ALT elevation despite dietary counselling. Inclusion criteria were age <18 years, a biopsy-proven diagnosis of NASH and an ALT that remained above the upper limit of normal (>40 U/L) for >3 months after diagnosis. NASH was defined as steatosis along with hepatocyte necrosis, inflammation (lobular and/or portal) and fibrosis (perisinusoidal and/or portal). Exclusion criteria were the use of any pharmacological therapy intended as a treatment for NASH, hepatic synthetic dysfunction as measured by prothrombin time and serum albumin, abnormal renal function as measured by blood urea nitrogen and serum creatinine, a history of diabetes or a fasting blood glucose >6.9 mm. Before participating in the study, each subject and parent gave written voluntary assent and consent. The protocol was approved by the institutional review board of the University of California, San Diego.
As there was no safety data available in children with liver disease we selected the lowest dose of metformin previously demonstrated to have beneficial effects on SI in non-diabetic adolescents.12 The design was a single-arm open-label study of oral metformin 500 mg twice daily for 24 weeks. In order to minimize gastrointestinal side-effects, subjects were initially started at metformin 500 mg once daily. After subjects demonstrated tolerance of metformin for a minimum of 1 week the dosing was increased to 500 mg twice daily. Subjects and their parents received a one-time counselling session with a registered dietician in order to simulate routine clinical care. Details of the American Heart Association diet13 were reviewed and the subjects were encouraged to increase their physical activity. Adherence to lifestyle recommendations were neither measured nor reinforced.
Monitoring and safety
Although rare, hepatoxicity and lactic acidosis have been reported with metformin use.14, 15 Therefore, subjects made monitoring visits to the General Clinical Research Center at the University of California, San Diego at 8-week intervals to assure safety and adherence. At these visits we assessed anthropometrics, liver chemistry, serum lactate and pill counts.
Height was measured to the nearest 1 mm using a wall-mounted stadiometer and weight was measured to the nearest 0.1 kg using a balance scale. Body mass index (BMI) was calculated as the weight (kg) divided by the height (m) squared. Obesity was defined as a BMI ≥ 95th percentile for age and sex.16, 17 BMI varies with age and gender; therefore in order to compare subjects independently of age and gender BMI Z-scores were calculated. The BMI Z-score represents the number of standard deviations from the national reference mean a subject is for a given age and gender. Waist circumference was measured with a non-stretchable tape midway between the inferior costal margin and the superior border of the iliac crest. Fat mass and lean body mass were measured by bioelectrical impedance analysis (Quantum II, RJL Systems, Clinton, MI, USA).18
Acanthosis nigricans (AN) is a lesion characterized by velvety thickening and hyperpigmentation affecting localized areas of the skin on neck, axillae and other flexural areas in subjects with hyperinsulinaemia.19 Scoring was performed at the neck as this has the greatest reproducibility.20 The degree of acanthosis present at the neck was recorded using a previously reported scale20 as follows: absent (0) – not detectable on close inspection; present (1) – clearly present on close visual inspection, not visible to the casual observer, extent not measurable; mild (2) – limited to the base of the skull, does not extend to the lateral margins of the neck (<7.5 cm in breadth); moderate (3) – extending to the lateral margins of the neck (posterior border of the sternocleidomastoid, 7.5–15 cm in breadth), should not be visible from the subject's front; severe (4) – extending anteriorly (>15 cm) visible when the participant is viewed from the front.
Liver chemistry. The primary end-point was change in serum ALT and AST from baseline to the end of treatment at 24 weeks. A complete response was defined as ALT and AST returning to values within the normal range (≤40 U/L). A partial response was defined as a decrease from baseline in ALT and AST ≥20% but with a level that remained >40 U/L. Baseline and treatment ALT and AST were measured in the UCSD Medical Center clinical chemistry laboratory using an enzymatic assay on a Beckman LX20 PRO analyzer (Beckman Coulter, Fullerton, CA, USA).
Liver fat. All subjects underwent MR spectroscopy before and 6 months after metformin treatment on a Siemens 1.5 Tesla Symphony scanner (Syngo 2002B software; Siemens, New York, NY, USA). Proton MR spectroscopy to quantify liver fat is a well known technique,21, 22 and was performed in this study (TR = 1500 ms; TE = 30, 32 or 120 ms) by placing a single 30 × 30 × 30 mm voxel over the right lobe of the liver where there were no large vessels or other water-like structures present. Fat-water ratios were derived from the MR spectroscopic data. They were corrected for differences in the T2 of fat and water, and for weight-to-signal differences between the various fat peaks.
Insulin sensitivity. Impaired SI, termed insulin resistance, has been defined as being the SI of the lowest 10% of a non-obese, non-diabetic population.23 Sera were analysed for glucose by the specific glucose oxidase method (YSI 2300 Stat Plus Analyzer, YSI Inc., Yellow Springs, OH, USA) and insulin concentrations using human insulin-specific radioimmunoassay (Linco Research Inc., St Charles, MO, USA). We used a mathematical model for SI, the quantitative insulin sensitivity check index (QUICKI)24 that was calculated as: QUICKI =1/[log(fasting insulin, μU/mL) + log(fasting glucose, mg/dL)]. Although other models exist we chose QUICKI because we have previously demonstrated that QUICKI is significantly related to the histological severity of steatosis in children with NAFLD.6 For children impaired SI is defined as QUICKI < 0.33925 and is reliable in both the euglycaemic and hyperglycaemic range.
Quality of life. Health-related QoL was assessed at baseline and at the completion of treatment prior to revealing any clinical results to the subjects. The 23-item pediatric quality of life inventory (PedsQL) 4.0 encompasses: (i) Physical Functioning (8-items), (ii) Emotional Functioning (5-items), (iii) Social Functioning (5-items) and (iv) School Functioning (5-items).26, 27 A total scale score is calculated from the means of all 23-items. The PedsQL 4.0 Generic Core Scales are composed of parallel child self-report and parent proxy-report formats and have been validated for use in obese children.28 A 5-point response scale is utilized (0 =never a problem; 4 = almost always a problem). Items are reverse-scored and linearly transformed to a 0–100 scale (0 = 100, 1 = 75, 2 = 50, 3 = 25, 4 = 0), so that higher scores indicate better HRQoL. Parents and subjects completed the instrument separately in either English or Spanish based upon the subject's stated preference. Responses were pooled across languages and ages as previously validated.26, 27
All study variables were treated as continuous variables. Summary outcome measures were reported as mean ± s.d. or range. Changes from baseline to end of treatment (24 weeks) with metformin were tested for significance using Student's t-tests or Wilcoxon rank sum tests, depending on whether the normality and homogeneity of variance assumptions were met. Correlation was tested using Pearson correlation test. All hypothesis tests were two-tailed and assumed a 5% level of significance.
Ten subjects (eight boys, two girls) with a mean age of 11.2 ± 2.7 years (range: 8–17) were enrolled. The parent-reported ethnic/racial distribution was Hispanic seven of 10; white, non-Hispanic two of 10 and native American Indian one of 10. All subjects were obese with a mean BMI of 30.4 ± 3.8 kg/m2 and BMI Z-score of 2.3 ± 0.2. Baseline and end of treatment data are shown in Table 1. All subjects completed the study and are included in the analysis. There were two subjects who experienced large weight loss (7 and 11 kg) and one subject who had a large weight gain (7 kg). The remaining seven subjects stayed within 1 kg of baseline weight. Therefore, there was no significant change in body weight or percentage body fat with treatment. However, height increased significantly (P < 0.001) and accounted for a significant (P < 0.05) decrease in BMI. AN was present in all subjects and significantly (P < 0.01) decreased with treatment.
Table 1. Clinical characteristics in paediatric NASH (n = 10)
End of treatment
Data shown as mean (±s.d.).
Significant changes are marked as *P < 0.05, **P < 0.01 and ***P < 0.001.
NASH, non-alcoholic steatohepatitis; BMI, body mass index; QUICKI, quantitative insulin sensitivity check index.
Waist circumference (cm)
Body fat (%)
All subjects had a biopsy diagnosis of NASH prior to enrolment. Steatosis was mild (<33%) in one, moderate (33–66%) in three and severe (>66%) in six subjects. Inflammation was lobular alone in one, lobular and portal in four and portal alone in five subjects. Fibrosis was perisinusoidal alone in one, perisinusoidal and portal in two and portal alone in seven subjects.
Safety and adherence
Gastrointestinal side-effects, diarrhoea and/or abdominal pain, were reported by three subjects with initiation of metformin. All subjects were able to tolerate twice daily metformin without symptoms by 4 weeks. Plasma lactate levels remained normal throughout the course of the trial, and there were no episodes of acidosis. Pill counts revealed excellent adherence to therapy with a mean consumption of 93% of expected tablets (range 83–98).
Biochemical response. Baseline mean aminotransferase values were ALT 184 ± 182 U/L and AST 114 ±109 U/L. Other liver chemistry (albumin, total protein, bilirubin, alkaline phosphatase) were normal. Following 24 weeks of treatment, ALT and AST were reduced in all 10 subjects from pre-treatment values (Figure 1). The changes were significant (P < 0.01) with mean changes of −86 for ALT and −46 for AST. ALT decreased significantly even in the absence of weight loss (Table 2). After treatment ALT was normal in 40% and AST was normal in 50% of subjects.
Table 2. Efficacy end-points in the absence of weight loss (n = 8)
End of treatment
Data shown as mean (±s.d.).
Significant changes are marked as *P < 0.05, **P < 0.01 and ***P < 0.001.
ALT, alanine aminotransferase; BMI, body mass index; QUICKI, quantitative insulin sensitivity check index.
Liver fat (%)**
Quality of life**
Quantitative liver fat. Liver fat as measured by MR spectroscopy at baseline was a mean of 30 ± 11%. After 24 weeks of treatment, liver fat decreased in 90% of subjects. The change was significant (P < 0.01) with a post-treatment mean of 23 ± 9%. A significant (P < 0.01) decrease in liver fat occurred even in the absence of weight loss (Table 2). The percentage change in liver fat was inversely correlated (r = −0.42) with the amount of metformin taken on an mg/kg/day basis as determined by baseline weight and pill counts (Figure 2).
Insulin sensitivity. Fasting glucose at baseline was normal in all subjects (range: 73–93 mg/dL). Baseline SI was impaired in all subjects with a mean QUICKI value of 0.294 ± 0.014. SI remained abnormal in nine of 10 subjects but did improve significantly (P < 0.05) to 0.310 ± 0.019.
Quality of life. Baseline QoL was impaired as assessed both by self-report (69 ± 11) and parent-proxy report (61 ± 18). No subject experienced a decline in QoL with treatment. Overall, subjects reported a significant (P < 0.01) improvement in QoL following treatment with a mean total score of 81 ± 8 (Figure 3). The largest gains were made in physical health and emotional functioning.
All of the 10 children with biopsy-established diagnosis enrolled and completed a pilot study of the safety and efficacy of metformin as a treatment for paediatric NASH. No adverse outcomes were experienced. The primary outcome was achieved as all subjects demonstrated improvement in ALT and AST. Metformin therapy was associated with improved SI and AN score although weight loss was not achieved. Furthermore, liver fat decreased and QoL improved.
Metformin has previously been tested in a few small studies of obese non-diabetic adolescents as a treatment for polycystic ovary syndrome,29 insulin resistance12 and weight loss.30 Glueck et al. studied 11 white, non-Hispanic adolescent girls with polycystic ovary syndrome with a mean age of 16.2 years and weight of 91 kg.29 They were treated in open-label fashion with a high protein, low carbohydrate, reduced-calorie diet and metformin at an initial dose of 1500 mg/day with an increase to 2250 mg/day in nine subjects. Weight loss and restoration of regular normal menses were reported in nine of 11 subjects. Freemark and Bursey treated 14 obese adolescents with fasting hyperinsulinaemia with metformin 500 mg twice daily and 15 adolescents with placebo for 6 months.12 The treated group had a mean age of 14.4 years, BMI of 41.5 and was 79% female. No subjects were Hispanic. AN was present only in 29% and was described as mild. Metformin treatment was associated with a small decrease in BMI Z-score and insulin resistance. Kay et al. treated 12 morbidly obese adolescents with metformin 850 mg twice daily and 12 with placebo for a total of 8 weeks.30 All received a low-calorie meal plan. The subjects had a mean age of 15.6 years and BMI of 41.2. Those in the metformin group had greater weight loss and improved SI. None of these prior studies presented data regarding liver chemistry, imaging or disease. The current study is the first clinical trial of metformin to include non-diabetic prepubertal children. The subjects in our study population also differ from those in other trials by being predominately male and Hispanic. This is important because male gender and Hispanic ethnicity are major risk factors for the development of NAFLD in children.31
Several pilot studies of varied design using metformin for adults with NAFLD or NASH have been reported. In the first of these, Marchesini et al. prescribed 500 mg of metformin thrice daily for 4 months to 20 consecutive patients with NAFLD (median age 40, mean BMI 28.8).7 Among the 14 subjects who completed treatment, liver volume as measured by ultrasound was reduced by 18% and serum ALT had normalized in 50% at the end of the study period.
Uygun et al. reported results from 34 subjects of an initial 36 adults with NASH (mean age 41, mean BMI 29.3) who completed a 6-month randomized-controlled study of a calorie-restricted diet with or without metformin 850 mg twice daily.32 ALT improved significantly with metformin, to a greater degree than with diet alone, and at 6 months, ALT was normal in 60% of subjects treated with metformin. In addition, liver fat as measured by ultrasound decreased by 23% more with metformin than with diet alone. In the longest study to date, Nair et al. reported on 15 completers (10 with follow-up liver biopsy) of 28 adults with NAFLD (mean age 51, mean BMI 32) who were treated with metformin 20 mg/kg/day for 48 weeks in open-label fashion.33 There was significant improvement in BMI, waist–hip ratio, and insulin resistance after 1 year of treatment. The completer population demonstrated a decrease in ALT at 3 months that was not sustained thereafter. However, ALT was improved at 48 weeks in four of the five subjects who had some degree of histological improvement. Taken as a whole the extant literature on the use of metformin for adults with NAFLD or NASH suggests a beneficial role for some patients but many questions remain unanswered.
In one previous study, treatment with metformin was not associated with a change in liver fat content. However, the study population was very different from ours. Tiikkainen et al. treated 11 Finnish adults who had type 2 diabetes with metformin 1000 mg twice daily for 16 weeks.34 None of these subjects was noted to have NASH. The study population was notable for normal fasting insulin levels at baseline with hyperglycaemia and a mean liver fat content of 13%. In marked contrast, the children in our study had biopsy-proven NASH, hyperinsulinaemia, normal fasting glucose and a mean liver fat content of 30% at baseline. The large difference in baseline liver fat content may be one explanation for the difference in outcome. In a study of weight of weight loss in women, Tiikkainen et al. demonstrated that women with high baseline liver fat have a greater change in liver fat in response to equivalent weight loss than women with low liver fat at baseline.35 Whether the same relationship would be true in response to treatment with metformin is unknown. Another potential explanation for the difference in liver fat outcomes between the two studies may be differences in the development of steatosis between insulin-resistant patients with and without diabetes. Insulin and glucose independently regulate the synthesis of fatty acids in the liver.36 The activation of lipogenesis by insulin is mediated by the transcription factor, sterol regulatory element-binding protein-1c (SREBP-1c)37 which leads to activation of lipogenic genes and a decrease in fatty oxidation. The activation of lipogenesis by glucose is mediated by a different transcription factor, carbohydrate response element-binding protein (ChREBP),38 which acts via liver-type pyruvate kinase to stimulate both glycolysis and lipogenesis. Thus, the activation of SREPB-1c by hyperinsulinaemia and of ChREBP by hyperglycaemia act synergistically to promote hepatic triglyceride production and storage.39 Non-diabetic children with NASH have not experienced pancreatic β-cell failure and hyperglycaemia. Therefore, the balance of factors contributing to steatosis should differ from adults with type 2 diabetes mellitus that may in turn explain the difference in response to treatment with metformin.
In the current study, the absence of a control group was appropriate for a pilot study but does limit the ability to determine true impact of pharmacotherapy. For example, any potential contribution to study outcomes by diet and/or exercise cannot be addressed. Their role will be important to address in future studies as cross-sectional data from a study of healthy white men demonstrated that cardiorespiratory fitness correlated with liver fat independently of BMI.40 However, several lines of reasoning do support the belief that the effects seen in this study were due to the actions of metformin. First of all, subjects were required to have persistent ALT elevation for a minimum of 3 months despite lifestyle counselling. Thus, subject selection reduced the likelihood that the children would lose weight in response to nutritional counselling or physical activity. Moreover, 80% of the subjects did not lose weight, but they too experienced significant improvements in ALT and liver fat. The possibility remains that physical activity may have beneficial effects on liver fat in the absence of anthropometric changes. In addition, the percentage change in liver fat correlated with the amount of metformin actually delivered on a per kilogram basis. This is the first study in humans with NASH to demonstrate that treatment with metformin leads to a decrease in liver fat content as measured by MR spectroscopy. This result is consistent with findings in leptin-deficient mice41 and changes in liver volume noted in adults with NAFLD.42 Whether reduction in steatosis is clinically meaningful is not known. However, fat is believed to be the substrate for reactive oxygen species generated in an environment of increased oxidative stress. Therefore, a reduction in liver fat may present less potential for necroinflammatory response with progressive fibrosis. Future studies that include both MR and biopsy before and after therapy will be able to determine if changes in liver fat detectable non-invasively correlate with changes in inflammation and fibrosis.
At baseline the subjects with NASH reported impaired QoL similar to that we have previously reported for severely obese children and adolescents.28 The improvement reported in QoL is notable and is the first clinical trial data showing improvement in paediatric QoL in the context of liver disease. The study was not designed to distinguish the reason for improvement. Possibilities include beneficial effects of metformin, physical activity or the psychological support of participating in a clinical trial.
This is the first paediatric treatment trial to only include subjects with biopsy-proven steatohepatitis. Subjects were carefully selected; excluding those with any evidence of hepatic synthetic dysfunction or renal dysfunction. All subjects were retained and we demonstrated tolerance to metformin in children as young as 8 years of age. We chose not to perform liver biopsy following treatment given the short duration of the study, and the prior lack of preliminary efficacy data. The degree to which changes in serum ALT or the amount of liver fat correlate with changes in histology remains to be established. Furthermore, the mean serum aminotransferase levels were still decreasing at the final time point. Longer study duration will be necessary to determine the optimal duration of therapy.
Targeting insulin resistance is but one of several potential therapeutic strategies for NASH. As the majority of children with NASH are obese, weight reduction via dietary change and physical activity is widely promoted. A case series of nine obese children with persistent elevation of ALT demonstrated that weight loss >10% of excess body weight achieved with a hypocaloric diet and physical exercise was associated with normalization of aminotransferase values.43 An open-label treatment trial of α-tocopherol in 11 obese children with elevated serum ALT and ultrasound evidence of fatty liver demonstrated normalization of serum ALT in all subjects in the absence of weight loss.44 Because these studies were based on surrogate measures of NAFLD it is not known which subjects truly had NAFLD nor is it known which ones had NASH. Furthermore, whether interventions based on weight loss, antioxidants or improving SI would be additive if used in combination merits consideration.
In the context of a pilot open-label clinical trial of short duration, metformin appears to be safe in children and adolescents with NASH. Furthermore, treatment with metformin at a dose of 1000 mg/day was associated with improvements in serum aminotransferase activity, total liver fat, SI and QoL. The true role of metformin including the safety and efficacy as well as optimal dosing and duration in paediatric NASH must be established in randomized-controlled trials as will be conducted by the National Institutes of Health NASH Clinical Research Network.45
Supported in part by grant M01 RR00827 from the NCRR of the NIH for the General Clinical Research Center at UCSD. Presented in abstract form at the Annual Meeting of the North American Society for Pediatric Gastroenterology, Hepatology and Nutrition, Montreal, Canada, 2 October 2003. Authors thank the nurses, dieticians and laboratory technicians from the GCRC for their contributions to subject care and data collection.