Background: Many studies demonstrated that roxadustat (FG-4592) could increase hemoglobin (Hb) levels effectively in anemia patients with chronic kidney disease (CKD). However, its safety remains controversial. This study aims to explore the risk of infection for CKD patients treated with roxadustat, especially focused on sepsis.Methods: We thoroughly searched for the randomized controlled trials (RCTs) comparing treatment with roxadustat versus erythropoiesis stimulating agents (ESAs) or placebo in PubMed, Embase, Cochrane Library, ClinicalTrials.gov, European Union Clinical Trials Register. Both on and not on dialysis anemia patients with CKD were included. Primary outcomes contained the incidence rates of sepsis. Secondary outcomes included infection-related consequences (septic shock and other infection events), general safety outcomes [all-cause mortality, treatment-emergent adverse events (TEAEs) and treatment-emergent serious adverse events (TESAEs)] and iron parameters. Moreover, a trial sequential analysis (TSA) was conducted to assess if the results were supposed to be a robust conclusion.Results: Eighteen RCTs (n = 11,305) were included. Overall, the incidence of sepsis (RR: 2.42, 95% CI [1.50, 3.89], p = 0.0003) and cellulitis (RR: 2.07, 95% CI [1.24, 3.44], p = 0.005) were increased in the roxadustat group compared with placebo group. In non-dialysis-dependent (NDD) CKD patients, the incidence of cellulitis (RR 2.01, 95% CI [1.23, 3.28], p = 0.005) was significantly higher in roxadustat group than that in the ESAs or placebo group. Both groups showed similar results in the incidence of septic shock (RR 1.29, 95% CI [0.86, 1.94], p = 0.22). A significant increased risk of all-cause mortality [risk ratios (RR): 1.15, 95% confidence interval (CI) [1.05, 1.26], p = 0.002] was found in roxadustat treatment, and TSA confirmed the result. Compared with ESAs or placebo, both the incident rates of TEAEs (RR:1.03, 95% CI [1.01, 1.04], p = 0.008) and TESAEs (RR: 1.06, 95% CI [1.02, 1.11], p = 0.002) were significantly increased in roxadustat group. As for iron parameters, changes from baseline (Δ) of hepcidin (MD: -26.46, 95% CI [-39.83, -13.09], p = 0.0001), Δ ferritin and Δ TSAT were remarkably lower in the roxadustat group, while Δ Hb, Δ iron and Δ TIBC increased significantly versus those in ESAs or placebo group.Conclusion: We found evidence that incidence rates of sepsis and cellulitis are higher in roxadustat group compared with placebo. This may be the result of improved iron homeostasis. The risk of all-cause mortality, TEAEs and TESAEs in CKD patients also increased in patients treated with roxadustat. We need more clinical and mechanistic studies to confirm whether roxadustat really causes infection.