BACKGROUND: In non-alcoholic fatty liver disease, histological lesions display a significant sampling variability that is ignored when interpreting histological progression during natural history or therapeutic interventions. AIM: To provide a method taking into account sampling variability when interpreting crude histological data, and to investigate how this alters the conclusions of available studies. METHODS: Natural history studies detailing histological progression and therapeutic trials were compared with the results of a previously published sampling variability study. RESULTS: Natural history studies showed an improvement in steatosis, which was significantly higher than expected from sampling variability (47% vs. 8%, P