Peripheral GDF15 as an early biomarker for brain disorders: A large prospective cohort study.
Level 3 - non-randomized controlled study
Prospective observational cohort study with Mendelian randomization
PubMed 42571855 · doi:10.1016/j.pnpbp.2026.111886
What was done
Plasma growth differentiation factor 15 (GDF15) was measured at baseline in UK Biobank participants followed prospectively for a median of 14 years. Cox proportional hazards models evaluated associations between log2-transformed GDF15 and incident brain disorders, including all-cause dementia (ACD), Alzheimer's disease (AD), Parkinson's disease (PD), anxiety, depression, sleep disorders, stroke, and epilepsy. Mediation analyses examined biochemical and hematological pathways, and one-sample Mendelian randomization (MR) was performed to test causality.
What was found
Higher baseline GDF15 was significantly associated with increased risk for overall brain disorders and all subtypes (all P < 0.001). Per 1-unit increase in log2-transformed GDF15, hazard ratios (95% CIs) were: - Overall brain disorders: 1.54 (1.48-1.60) - All-cause dementia: 1.98 (1.80-2.17) - Stroke: 1.92 (1.81-2.05) - Alzheimer's disease: 1.84 (1.60-2.10) - Epilepsy: 1.71 (1.46-2.01) - Sleep disorders: 1.40 (1.26-1.55) - Depression: 1.38 (1.29-1.49) - Parkinson's disease: 1.37 (1.18-1.59) - Anxiety: 1.26 (1.16-1.37) HDL cholesterol accounted for an estimated 7.51% of the association with depression and 11.47% with sleep disorders, and neutrophil count partially mediated multiple outcomes. MR analyses showed no direct causal effect.
Why it matters
Plasma GDF15 serves as a non-specific long-term risk marker across neurodegenerative, vascular, and psychiatric brain disorders. The lack of MR support indicates GDF15 is likely a downstream indicator of systemic stress or inflammation rather than a direct causal driver.
Limits
The abstract does not state the total sample size or case counts. GDF15 was measured at a single baseline timepoint without longitudinal reassessment. Observational associations remain susceptible to residual confounding, and UK Biobank participants may not represent broader global populations.