Abstract:

Objective: To explore the correlation between metformin treatment and the onset of tuberculosis (TB) by Mendelian randomization (MR). Methods: We searched the MR Base database (http://app.mrbase.org) using metformin treatment as the exposure variable and TB incidence as the outcome variable. A set of publicly available genome-wide association studies (GWAS) were used to summarize the data. Mendelian randomization causal association analysis was performed using inverse variance weighting (IVW), weighted median, and MR-Egger regression analysis on two samples (462933 cases of metformin treatment, including 11552 cases and 451381 controls; 212453 cases of TB, including 549 cases and 211904 controls). Cochran’s Q statistic and I² statistic were used to evaluate heterogeneity among single-nucleotide polymorphisms (SNPs). The “leave-one-out-method” sensitivity analysis and funnel plot were used to verify the reliability of the results. Results: Forty independent SNPs significantly associated with metformin treatment at the genome-wide level were selected as instrumental variables from GWAS, and 168 independent SNPs significantly associated with TB were selected. Inverse variance-weighted analysis showed a significant positive correlation between metformin treatment and TB incidence (P=0.022). MR-Egger regression analysis found that directional pleiotropy was unlikely to affect the results (intercept=-0.043, P=0.394), there was no evidence of a direct correlation between metformin treatment and TB incidence (P=0.150). However, the weighted median method analysis supported a positive correlation between metformin treatment and TB incidence (P=0.010). No heterogeneity was found in the inverse variance-weighted and MR-Egger analyses (Q=46.890, I²=0.168, P=0.175; Q=45.990, I²=0.174, P=0.181), increasing the reliability of the Mendelian randomization analysis estimates. The “leave-one-out-method” sensitivity analysis found that the overall Mendelian randomization estimates were stable, with no significant bias, and the results were reliable. The funnel plot also did not show asymmetry, indicating no directional pleiotropy and supporting the robustness and reliability of the results. Conclusion: After heterogeneity and sensitivity testing, the Mendelian randomization results obtained through inverse variance weighting, weighted median, and MR-Egger regression analysis support a positive correlation between metformin treatment and TB onset.

Key words: Metformin, Tuberculosis, Statistics as topic, Mendelian randomization