Abstract. Non-alcoholic fatty liver disease (NAFLD) related to insulin resistance (IR) is a growing global health concern. Recent studies have indicated that metformin could improve IR and may be beneficial in the treatment of NAFLD. This study aimed to assess the beneficial or harmful effects of metformin in NAFLD. We searched Medline and four other databases during April 2012. Selection criteria were randomized clinical trials comparing metformin with placebo or other interventions for treating NAFLD patients. The primary outcome was histological response. The secondary outcomes included alanine aminotransferase (ALT), aspartate aminotransferase (AST), homeostasis model assessment of IR (HOMA-IR), body mass index (BMI) and adverse events. Dichotomous data were reported as odds ratio (OR), while continuous data were calculated as the mean difference (MD), both with 95% confidence intervals (CI). Random and fixed effects meta-analyses were performed. Nine studies were included, involving 417 participants, and conducted for a time period ranging from 4 to 12 months. In the treated participants, improvements were observed in ALT (MD, -8.12 U/l; P=0.03), AST (MD, -4.52 U/l; P=0.04), HOMA-IR (MD, -0.61; P=0.005) and BMI (MD, -0.82 kg/m 2 ; P=0.04), but not in histological response: steatosis (P=0.66), inflammation (P=0.91), hepatocellular ballooning (P= 0.25) and fibrosis (P= 0.90). Sub-analysis of non-alcoholic fatty steatohepatitis showed that metformin failed to improve any pooled outcome. Adverse events were poorly reported. Current information indicates that metformin improves liver function, HOMA-IR and BMI to some extent, but not histological response in NAFLD patients. This finding could serve as a stimulus for future studies investigating issues such as dose-responsiveness, safety and patient tolerance to metformin therapy.