Dementia is a growing global health challenge, particularly among individuals with prediabetes or diabetes, who face a significantly higher risk of developing cognitive decline. However, existing tools for predicting dementia risk in these populations remain limited in accuracy and biological specificity.
A new large-scale study published in General Psychiatry introduces a novel protein-based risk score that can predict dementia risk in individuals with dysglycaemia more accurately. With proteomic data from over 10,000 participants in the UK Biobank, researchers identified a panel of 23 proteins associated with dementia risk, with six key proteins accounting for most of the predictive power.
The study demonstrated that this protein-based model significantly outperforms traditional clinical risk tools, such as those based on age, sex, and cardiovascular risk factors. According to the analysis results, protein score achieved a C-index of up to 0.85, compared with 0.78–0.80 for conventional models, indicating substantially improved predictive accuracy.
Biological pathway analysis revealed that these proteins are linked to mechanisms such as neuroinflammation, extracellular matrix remodelling, and cholesterol metabolism. As illustrated, dementia risk increases progressively with higher protein risk scores, highlighting a strong dose–response relationship.
Importantly, several identified proteins, such as APOE, GDF15, and HPGDS, may also serve as potential therapeutic targets. Some are already linked to existing or investigational drugs, suggesting opportunities for drug repurposing and precision medicine approaches.
Overall, the findings highlight the potential of proteomics to transform dementia prevention in high-risk populations. By enabling earlier and more precise risk stratification, this approach could support personalised interventions and improve long-term cognitive health outcomes for individuals with prediabetes or diabetes.
General Psychiatry
Observational study
Dementia risk prediction in individuals with prediabetes or diabetes: A novel multi-protein score with biological pathway analysis
15-Mar-2026