ABSTRACT Background Renal replacement therapy (RRT) is essential in the presence of life-threatening complications associated with acute kidney injury (AKI). In the absence of urgent indications, the optimal timing for RRT initiation is still under debate. This meta-analysis aims to compare the benefits between early and late RRT initiation strategies in critically ill patients with AKI. Methods Studies were obtained from three databases [Medical Literature Analysis and Retrieval System Online (MEDLINE), Cochrane Central Register of Controlled Trials (CENTRAL) and Scopus], searched from inception to May 2021. The selected primary outcome was 28-day mortality. Secondary outcomes included overall mortality, recovery of renal function (RRF) and RRT-associated adverse events. A random-effects model was used for summary measures. Heterogeneity was assessed through Cochrane I2 test statistics. Potential sources of heterogeneity for the primary outcome were sought using sensitivity analyses. Further subgroup analyses were conducted based on RRT modality and study population. Results A total of 13 randomized controlled trials including 5193 participants were analysed. No significant differences were found between early and late RRT initiation regarding 28-day mortality [risk ratio (RR) 1.00; 95% confidence interval (CI) 0.89–1.12, I² = 30%], overall mortality (RR 1.00; 95% CI 0.90–1.12, I² = 42%) and RRF (RR 1.02; 95% CI 0.92–1.13, I² = 53%). However, early RRT initiation was associated with a significantly higher incidence of hypotensive (RR 1.34; 95% CI 1.17–1.53, I² = 6%) and infectious events (RR 1.83; 95% CI 1.11–3.02, I² = 0%). Conclusions Early RRT initiation does not improve the 28-day and overall mortality, nor the likelihood of RRF, and increases the risk for RRT-associated adverse events, namely hypotension and infection.