BACKGROUND: Knee osteoarthritis (OA) is one of the most common joint diseases, and pharmacotherapy is necessary to control its symptoms. AIM: To evaluate efficacy and tolerability of duloxetine in patients with knee OA. METHODS: PubMed, Web of Science, Embase, Cochrane Library and ClinicalTrials.gov were searched to identify randomised controlled trials comparing duloxetine with placebo for knee OA. Data including pain, stiffness, physical function, and adverse events were extracted for meta-analysis. The protocol was prospectively registered in PROSPERO (CRD42018097110). RESULTS: Data from six randomised controlled trials including 2059 participants were pooled. Duloxetine achieved significant reductions in primary outcomes including Brief Pain Inventory 24-h average pain score (weighted mean difference (WMD) = -0.74, 95% confidence interval (CI) = -0.92 to -0.57), weekly mean of the 24-h average pain score (WMD = -0.76, 95% CI = -0.96 to -0.56), WOMAC stiffness score (WMD = -0.47, 95% CI = -0.60 to -0.34) and WOMAC physical function score (WMD = -4.44, 95% CI = -5.24 to -3.64). Furthermore, duloxetine demonstrated a higher number of treatment-emergent adverse events (risk ratio (RR) = 1.31, 95% CI = 1.20-1.44) and discontinuations (RR = 2.26, 95% CI = 1.63-3.12); however, no difference in serious adverse events (RR = 0.92, 95% CI = 0.40-2.11) was observed. CONCLUSION: Duloxetine is effective in the management of chronic pain and loss of physical function in knee OA with acceptable adverse events despite having no advantage in treating joint stiffness. Future trials should focus on determining the optimal treatment regimen.