A randomized controlled trial of metformin versus vitamin E or prescriptive diet in nonalcoholic fatty liver disease

Metformin proved useful in the treatment of nonalcoholic fatty liver disease (NAFLD), but its superiority over nutritional treatment and antioxidants has never been demonstrated. We aimed to compare the usefulness of metformin versus prescriptive diet or vitamin E.

In an open label, randomized trial, nondiabetic NAFLD patients were given metformin (2 g/day; n = 55) for 12 months. The control cases were given either vitamin E (800 IU/day; n = 28) or were treated by a prescriptive, weight-reducing diet (n = 27). Outcome measures were liver enzymes, insulin resistance (homeostasis model assessment), parameters of the metabolic syndrome, and histology.

Aminotransferase levels improved in all groups, in association with weight loss. The effects in the metformin arm were larger (p < 0.0001), and alanine aminotransferase normalized in 56% of cases (odds ratio (OR) versus. controls, 3.11; 95% confidence interval (CI), 1.56-6.20; p= 0.0013). In multivariate analysis, metformin treatment was associated with higher rates of aminotransferase normalization, after correction for age, gender, basal aminotransferases, and change in body mass index (OR, 5.98; 95% CI, 2.05-17.45). Differences were maintained when the two control groups were separately analyzed. The distribution of positive criteria for the metabolic syndrome was reduced only in the metformin arm (p= 0.001, signed rank test). A control biopsy in 17 metformin-treated cases (14 nonresponders) showed a significant decrease in liver fat (p= 0.0004), necroinflammation, and fibrosis (p= 0.012 for both). No side effects were observed during metformin treatment.

Metformin treatment is better than a prescriptive diet or vitamin E in the therapy of NAFLD patients receiving nutritional counseling. Limited histological data support an association between improved aminotransferases and biopsy findings, which require confirmation in a double-blind trial with appropriate statistical power based on liver histology.