Display omitted •Magnetic resonance-based images are frequently used in phase I-II studies for NASH.•It is assumed that changes in liver fat will predict histological changes.•We analyzed data from 121 patients with paired biopsies and magnetic resonance imaging.•Changes in liver fat did not predict changes in inflammation, ballooning or fibrosis.•Quantification of liver fat after treatment may be misleading as a surrogate marker of treatment response. Proof-of-concept studies frequently assess changes in intrahepatic triglyceride (IHTG) content by magnetic resonance-based techniques as a surrogate marker of histology. The aim of this study was to establish how reliable this strategy is to predict changes in liver histology in patients with non-alcoholic steatohepatitis (NASH). Patients with NASH who had participated in our prior randomized controlled trials of pioglitazone with complete paired data for IHTG content by magnetic resonance spectroscopy and liver histology were included in the study. A total of 121 patients were included. Changes in IHTG were assessed in several ways: as a continuous variable (correlations), as categorical groups (IHTG change ≥0%; or IHTG reduction of 1–30%; 31–50%; 51–70%; or >70%), and in a binomial way as steatosis resolution or not (defined as achieving IHTG <5.56%). Changes in IHTG correlated with steatosis on histology (r = 0.54; p <0.01). However, the magnitude of IHTG reduction was not associated with the rate of response of the primary histological outcome (2-point improvement in the NAFLD activity score from 2 different parameters, without worsening of fibrosis) or resolution of NASH without worsening of fibrosis, neither in patients receiving pioglitazone nor placebo. Changes in lobular inflammation, hepatocyte ballooning, or liver fibrosis were also independent of changes in IHTG, irrespective of treatment arm. Steatosis resolution was not associated with better histological outcomes either. Changes in IHTG predict changes in steatosis but not of other liver histological parameters. This implies that IHTG response to treatment should be interpreted with caution, as it may not be as reliable as previously believed to predict a treatment’s overall clinical efficacy in patients with NASH. Quantification of liver fat by magnetic resonance imaging (MRI) is currently used to assess treatment responses in patients with fatty liver, with the assumption that improvements in liver fat translate into less inflammation, necrosis, and fibrosis in the liver. However, in this article, we showed that changes in liver fat do not necessarily translate into changes in these parameters. This means that MRI may not be as useful to assess treatment response in patients with fatty liver as previously believed.