We aim to investigate the causal effect of blood lipids mediating sodium-glucose cotransporter 2 (SGLT2) inhibition in cardiovascular disease (CVD) using Mendelian randomization (MR). A two-sample two-step MR study was conducted to evaluate the association of SGLT2 inhibition with CVDs and the mediation effects of blood lipids linking SGLT2 inhibition with CVDs. Genetic instruments for SGLT2 inhibition were identified as genetic variants, which were associated with the expression of the SLC5A2 gene and glycated haemoglobin level (HbA1c). SGLT2 inhibition was associated with reduced risk of heart failure (HF) (OR 0.44 [95% CI 0.32-0.61]; P=6.0×10-7), atrial fibrillation (AF) (0.47 [0.37-0.61]; P=1.81×10-8), coronary artery disease (CAD) (0.47 [0.30-0.73]; P=7.46×10-4), myocardial infarction (MI) (0.30 [0.15-0.61]; P=7.44×10-4), any stroke (AS) (0.28 [0.18-0.42]; P=1.14×10-9), and ischaemic stroke (IS) (0.27 [0.17-0.44]; P=1.97×10-7). Our results indicated that the proportion mediated of the mediating effect of total cholesterol was 1.7% (OR 0.99 [95% CI 0.98, 0.99], P=0.004), 4.7% (0.96 [0.95, 0.98], P=0.002), and 2.7% (0.97 [0.95, 0.98], P=0.002) in the association between SGLT2 inhibition and the risk of HF, CAD, and MI, respectively. For low-density lipoprotein cholesterol, the proportion mediated of the mediating effect was 2.2% for HF (OR 0.98 [95% CI 0.98, 0.99], P=0.003), 8.6% for CAD (0.93 [0.91, 0.95], P=5.74×10-4), and 5.0% for MI (0.95 [0.94, 0.96], P=6.97×10-4). For non-high-density lipoprotein cholesterol, the proportion mediated of the mediating effect was 3.4% for HF (OR 0.98 [95% CI 0.97, 0.98], P=4.42×10-6), 11.8% for CAD (0.92 [0.90, 0.93], P=7.23×10-8), 5.7% for MI (0.94 [0.92, 0.95], P=8.17×10-7), 1.5% for AS (0.98 [0.98, 0.99], P=0.001), and 1.4% for IS (0.98 [0.98, 0.99], P=0.004). Our study showed the association of SGLT2 inhibition with the reduced risk of CVDs and blood lipids might mediate this association.