Deciphering the Causal Relationship between Sodium-glucose Cotransporter 2 Inhibition and Cancer Risks: A Comprehensive Mendelian Randomization Study.

Controversy persists regarding the effects of sodium-glucose cotransporter 2 (SGLT2) inhibitors on cancer. The underlying causal relationship remains unclear. A two-sample Mendelian randomization (MR) strategy was employed to investigate the causal associations between SGLT2 inhibitors and 26 site-specific malignancies. Instrumental variants strongly associated with SLC5A2 gene expression and glycated hemoglobin A1c levels were identified as the genetic proxy for SGLT2 inhibition. Cancer-related outcome datasets sourced from the OpenGWAS project were separated into discovery and replication datasets. The meta-analysis was conducted to determine the final causality. Genetically proxied SGLT2 inhibition showed a significant association with bronchial and lung cancer (beta: -0.028 [-0.041, -0.015], P < 0.001), bladder cancer (beta: 0.018 [0.008, 0.027], P < 0.001), prostate cancer (beta: 1.168 [0.594, 1.742], P < 0.001), cervical cancer (beta: -0.019 [-0.031, -0.008], P = 0.001), corpus uterine cancer (beta: 0.015 [0.006, 0.025], P = 0.001) and non-melanoma skin cancer (beta: -0.080 [-0.116, -0.044], P < 0.001) in the discovery cohort. The suggestive causal effect of SGLT2 inhibition on the increased risk of cervical cancer (beta: 3.241 [0.855, 5.627], P = 0.008) and lymphoid leukemia (beta: 4.126 [0.383, 7.868], P = 0.031) was found in the replication cohort. The combined causality of the following types of cancer was observed to remain significant after meta-analysis: bronchial and lung cancer, bladder cancer, prostate cancer, corpus uterine cancer, and non-melanoma skin cancer (all P ≤ 0.001). For the first time we discovered that the SGLT2 inhibition may exert protection on bronchial and lung cancer and non-melanoma skin cancer from a genetic perspective. However, suggestive higher cancer risks of bladder, prostate, and corpus uteri were also noted, which warrants real-world data validation in the future.