Background and aim Trials of treatments for non-alcoholic steatohepatitis require endpoint assessment with liver biopsies. Previous large-scale trials have calculated their sample size expecting high retention but on average did not achieve this. We aimed to quantify the proportion of participants with a valid follow-up biopsy. Methods We conducted a systematic review of MEDLINE and Embase until May 2020 and included randomized clinical trials of any intervention in non-alcoholic steatohepatitis with at least 1-year follow-up. We were guided by Cochrane methods to run a meta-analysis with generalized linear mixed models with random effects. Results Forty-one trials (n = 6,695) were included. The proportion of participants with a valid follow-up biopsy was 82% (95%CI: 78%-86%, I² = 92%). There was no evidence of a difference by location, trial length, or by allocated treatment group. Reasons for missing follow-up biopsies were, in ranked order, related to participants (95 per 1,000 participants (95%CI: 69–129, I² = 92%), medical factors, protocol, trial conduct, and other/unclear. Biopsy-related serious adverse events occurred in 16 per 1,000 participants (95% CI: 8–33, I² = 54%). No biopsy-related deaths were reported. Conclusions The proportion of participants with a valid follow-up biopsy in therapeutic trials in non-alcoholic steatohepatitis is on average 82%, with around 1 in 10 participants declining a follow-up biopsy. These findings can inform adequately-powered trials.