Objective: To examine the causal relationship between inflammatory cytokines and premature ovarian insufficiency (POI) using bidirectional Mendelian randomization (MR) approach, so as to provide the basis for the prevention and treatment of POI. Methods: The data for 91 inflammatory cytokines were sourced from the IEU OpenGWAS database, comprising 14 824 participants. GWAS data for POI were sourced from the FinnGen database, including 118 484 individuals (among which 254 were POI cases). MR analysis was performed using the inverse variance weighted (IVW) method with inflammatory cytokines as exposure and POI as the outcome for forward MR analysis and POI as the exposure and inflammatory cytokines as outcome for reverse MR analysis. Sensitivity analysis were conducted using Cochran's Q test, MR-Egger regression, and the MR-PRESSO test. Results: Forward MR analysis demonstrated statistically significant associations between POI and interleukin-10 (OR=0.410, 95%CI: 0.233-0.721), interleukin-33 (OR=2.826, 95%CI: 1.228-6.504), C-C motif chemokine ligand 19 (OR=0.583, 95%CI: 0.364-0.932), monocyte chemoattractant protein-3 (OR=0.559, 95%CI: 0.335-0.936), interleukin-18 receptor 1 (OR=1.370, 95%CI: 1.030-1.821), and interleukin-13 (OR=1.990, 95%CI: 1.034-3.832). Reverse MR analysis revealed significant negative associations between POI and 15 inflammatory cytokines, including C-C motif chemokine ligand 23 (OR=0.981, 95%CI: 0.968-0.994) and axin-1 (OR=0.978, 95%CI: 0.963-0.994). Sensitivity analysis showed no evidence of heterogeneity or horizontal pleiotropy (all P>0.05). Conclusion Elevated levels of interleukin-33, interleukin-18 receptor 1 and interleukin-13 were associated with an increased risk of POI, while POI may be associated with decreased levels of 15 inflammatory cytokines including C-C motif chemokine ligand 23 and axin-1.