This study examined the relationships between type 2 diabetes (T2D), metformin use, and Parkinson’s disease (PD) risk. The authors aimed to determine both overall and independent causal effects of T2D and metformin use on PD by combining genetic causal inference with pharmacovigilance data.
The genetic component used genome-wide association study (GWAS) summary datasets to perform Mendelian randomization (MR) analyses. Univariable MR assessed the overall causal association of genetically instrumented exposures (T2D and metformin use) with PD. Sensitivity analyses were applied to evaluate result robustness. Multivariable MR (MVMR) was performed to estimate the direct effects of each exposure while accounting for the other.
The abstract states that publicly released aggregated GWAS summary statistics were reanalyzed; ethical approval for the original GWAS datasets had already been granted. Specific details about GWAS cohort sizes, instrument selection criteria, genetic variants used, and full sensitivity-analysis methods are not reported in the abstract.
To complement genetic analyses, the study interrogated the FDA Adverse Event Reporting System (FAERS) covering 2004 Q1 through 2024 Q2. Reports listing metformin-related adverse events were reviewed for instances of Parkinsonism. A reporting odds ratio (ROR) was calculated to quantify disproportionate reporting of Parkinsonism among metformin-associated reports compared with other adverse event reports.
In univariable MR analyses the authors report that genetically predicted T2D was not associated with PD risk (OR 1.02, 95% CI 0.96–1.09, P = 0.47). By contrast, genetically proxied metformin use was associated with a modestly increased risk of PD (OR 1.08, 95% CI 1.02–1.14, P = 0.011). Sensitivity analyses were performed to assess robustness of the MR estimates, although the abstract does not present those detailed results or specific sensitivity-method outputs.
When both exposures were modeled jointly using MVMR, effects were in opposing directions. T2D appeared robustly associated with a decreased risk of PD (OR 0.63, 95% CI 0.50–0.78, P = 5.09E-5). In the same multivariable analysis, metformin use was linked to a higher PD risk with a larger effect size (OR 1.63, 95% CI 1.31–2.03, P = 1.29E-5). These MVMR estimates represent direct-effect estimates conditional on the other exposure in the model.
Analysis of FAERS identified 179 reports of Parkinsonism among 228,283 metformin-related adverse event reports. The reporting odds ratio (ROR) for metformin and Parkinsonism was reported as 5.25, indicating a statistically significant disproportionate reporting signal in spontaneous adverse event data.
The combined evidence from genetic MR and pharmacovigilance analyses led the authors to highlight contrasting signals: genetically predicted T2D showed no association in univariable MR and a protective association in multivariable MR, whereas genetically proxied metformin use and spontaneous-report data from FAERS were both associated with increased risk or reporting of PD/Parkinsonism.
These findings suggest that the relationship between diabetes, glucose-lowering therapy, and neurodegenerative risk is complex. The results raise the potential that metformin may contribute to increased PD risk or to increased reporting of parkinsonian symptoms, and that T2D per se may not increase PD risk in the same direction once medication effects are accounted for.
Clinicians and researchers should consider potential neurological adverse effects when evaluating long-term metformin use in patients with or at risk for Parkinson’s disease. However, the abstract does not provide clinical guidance on treatment changes; it emphasizes the need to weigh possible adverse effects in context.
The abstract notes that the study relied on publicly available GWAS summary statistics and on FAERS spontaneous reports. The authors state that ethical approvals for original GWAS were obtained and no separate approval was required for reanalysis. The abstract does not report detailed methods for genetic instrument selection, cohort characteristics, confounder control beyond MVMR, or full sensitivity-analysis outputs. FAERS is a voluntary reporting system subject to reporting bias and cannot establish incidence or causality on its own.
Using MR and FAERS data, the study reports opposing causal signals: a potential protective association of T2D with PD in multivariable analyses, and associations of metformin use with increased PD risk and with disproportionate reporting of Parkinsonism. The authors conclude that metformin’s potential adverse neurological effects warrant consideration in patients with or at risk for Parkinson’s disease. Further detailed methodological and cohort information is not provided in the abstract and would require review of the full text for comprehensive evaluation.