Highlights
– In a 20‑year real‑world cohort (TriNetX), SGLT2 inhibitor use in T2DM was associated with a 28% lower adjusted hazard of Parkinson’s disease compared with metformin (aHR 0.72; 95% CI 0.62–0.84).
– A concordant reduction in dementia risk (aHR 0.73) and lower all‑cause mortality (aHR 0.85) were observed, with negative controls supporting specificity.
– Findings are hypothesis‑generating; residual confounding, indication bias, and lack of randomized exposure assignment limit causal inference and highlight the need for mechanistic and prospective studies.
Background: clinical context and unmet need
Parkinson’s disease (PD) is a progressive neurodegenerative disorder characterized by loss of dopaminergic neurons in the substantia nigra and heterogeneous motor and nonmotor features. Age is the strongest risk factor, but epidemiological and mechanistic evidence links metabolic dysfunction—especially type 2 diabetes mellitus (T2DM)—to increased PD risk and faster progression. Proposed mediators include peripheral and central insulin resistance, chronic low‑grade inflammation, oxidative stress, and mitochondrial dysfunction. Identifying antidiabetic therapies that modify neurodegenerative risk is attractive because these agents are widely used, have established safety profiles, and could be repurposed to prevent or slow neurodegeneration.
Study design and methods
The referenced study by Sun et al. (2025) presents the first large, head‑to‑head real‑world comparison of sodium–glucose cotransporter‑2 inhibitors (SGLT2is) versus metformin for incident PD risk among adults with T2DM. Using the TriNetX electronic health record network spanning 142 healthcare organizations from 2005 to 2025, investigators identified 913,428 eligible patients: 96,018 who received SGLT2is and 817,410 metformin users. Patients with prior PD, other neurodegenerative diseases, or exposure to known neuroprotective or neurotoxic antidiabetic drugs were excluded.
To reduce confounding, 1:1 propensity score matching balanced cohorts on demographic, clinical, and pharmacologic variables. Time‑to‑event analyses used Cox proportional hazards models to estimate adjusted hazard ratios (aHRs) for incident PD, with additional analyses for dementia and all‑cause mortality as positive and global outcomes. Positive and negative control outcomes were used to assess residual bias and specificity.
Key findings
Main result: After propensity matching, SGLT2i use was associated with a 28% lower risk of incident PD versus metformin (aHR 0.72; 95% CI 0.62–0.84; p < 0.0001). The magnitude of effect was clinically meaningful and statistically robust in the primary adjusted model.
Secondary and validation outcomes: Dementia—used as a positive control given prior suggestions of SGLT2i cognitive benefit—was also reduced in SGLT2i users (aHR 0.73; 95% CI 0.68–0.78; p < 0.0001), lending internal consistency to the neuroprotective signal. Negative controls (outcomes not plausibly affected by the drug exposure) showed no association, supporting specificity and mitigating concerns about systematic bias in outcome ascertainment.
Mortality: All‑cause mortality was lower among SGLT2i users (aHR 0.85; 95% CI 0.83–0.89; p < 0.0001), consistent with established cardiovascular and renal survival benefits of SGLT2is in T2DM populations and potentially relevant as a competing risk in neurodegenerative outcome analyses.
Sensitivity analyses: The authors report robustness across multiple analytic approaches, though details on duration of exposure, dose, adherence, and timing relative to diabetes duration were not fully detailed in the summary and remain important caveats.
Biological plausibility and potential mechanisms
Mechanistic plausibility strengthens the argument that SGLT2is could influence neurodegenerative processes. Candidate mechanisms include improved systemic and central metabolic profiles (reduced insulin resistance), lowered inflammation, enhanced mitochondrial function and autophagy, reduced oxidative stress, modulation of cerebral blood flow, and direct central nervous system effects via SGLT expression in brain tissue. SGLT2is also alter energy substrate utilization (ketogenesis), which may provide neuroprotective metabolic effects. Conversely, metformin has shown neuroprotective signals in preclinical models and some epidemiological contexts, but its effects may differ by dose, duration, and population.
Expert commentary: strengths, limitations, and interpretation
Strengths of this analysis include the very large, multi‑institutional real‑world dataset with long temporal coverage, careful exclusion of prevalent neurodegenerative disease, use of propensity score matching across many covariates, and the application of positive and negative control outcomes to assess bias. These features increase confidence that the observed associations merit attention.
However, key limitations temper causal interpretation:
- Residual confounding: Despite matching, unmeasured confounders—such as diabetes duration, cumulative glycemic burden (A1c trajectory), socioeconomic status, physical activity, environmental toxin exposure, family history, prodromal nonmotor PD symptoms (e.g., REM sleep behavior disorder), and smoking—can influence both drug choice and PD risk.
- Channeling and indication bias: SGLT2is are more likely prescribed in patients with specific cardiovascular or renal comorbidities or in later treatment lines; metformin is typically first‑line. Differences in follow‑up windows and competing morbidities could bias associations.
- Exposure measurement: EHR prescription records do not guarantee adherence, persistence, or accurate measurement of cumulative exposure. Timing of initiation relative to diabetes onset and the latency for neurodegenerative processes are critical but often unavailable in administrative datasets.
- Outcome ascertainment: PD diagnosis in EHRs may be delayed or misclassified, and reduced healthcare contact in some subgroups could affect detection rates. Neurology specialist confirmation and standardized diagnostic criteria were not prospectively applied.
- Competing risks: Lower mortality among SGLT2i users could paradoxically increase observed PD incidence if survivors live longer to manifest PD, or conversely lower incidence if competing causes censor patients before PD develops. The study reports lower mortality, but the net impact on PD ascertainment requires careful competing‑risk analysis.
In summary, the association is compelling and biologically plausible but not yet definitive evidence of causality.
Clinical implications and next steps
For practicing clinicians, these findings should not yet change guideline‑directed prescribing. SGLT2is remain indicated for T2DM patients with appropriate cardiovascular, renal, or heart failure indications. The potential neuroprotective effect adds another rationale to consider when patients are eligible for SGLT2i therapy, particularly in older patients at higher neurodegenerative risk, but shared decision‑making should weigh established benefits and known risks (e.g., genital mycotic infections, euglycemic diabetic ketoacidosis in select patients, volume depletion).
Research priorities:
- Prospective randomized trials or pragmatic trials with neurodegenerative endpoints or validated surrogate biomarkers (e.g., imaging, CSF markers) are required to establish causality and quantify absolute risk reduction.
- Mechanistic studies in animal models and human biomarker studies should delineate pathways (insulin signaling, mitochondrial function, autophagy, inflammation) and identify subgroups most likely to benefit.
- Detailed observational studies addressing diabetes duration, glycemic control, dosing, adherence, and competing risks—ideally with specialist‑confirmed PD diagnoses—can further refine effect estimates.
- Exploratory pharmacoepidemiology comparing different SGLT2i agents and evaluating interactions with other antidiabetic agents (including metformin) could clarify additive or synergistic effects.
Conclusion
The large TriNetX study by Sun et al. provides the first head‑to‑head real‑world evidence that SGLT2 inhibitors may be associated with a substantially lower risk of Parkinson’s disease compared with metformin in patients with type 2 diabetes. The finding is supported by concordant reductions in dementia and lower mortality, and it is biologically plausible. Nevertheless, the observational design and potential for residual confounding preclude definitive causal claims. These results should be viewed as hypothesis‑generating and motivate prospective clinical trials, mechanistic investigations, and careful assessment of patient‑level benefit–risk when choosing glucose‑lowering therapies.
Funding and clinicaltrials.gov
The source article lists authors and affiliations; funding and trial registration details should be checked in the original publication: Sun M, Wang X, Lu Z, et al. SGLT2 inhibitors vs. metformin for Parkinson’s disease risk reduction in type 2 diabetes. J Parkinsons Dis. 2025. DOI: 10.1177/1877718X251359391. PMID: 40671477.
References
1. Sun M, Wang X, Lu Z, Yang Y, Lv S, Miao M, Chen WM, Wu SY, Zhang J. SGLT2 inhibitors vs. metformin for Parkinson’s disease risk reduction in type 2 diabetes. J Parkinsons Dis. 2025 Jul 17:1877718X251359391. doi: 10.1177/1877718X251359391. Epub ahead of print. PMID: 40671477.
(Readers are encouraged to consult the primary article for full methodological details and to follow emerging randomized and mechanistic studies that will test these provocative findings.)