image: Association between dementia protein risk score and incident dementia among individuals with prediabetes ordiabetes in the testing and external validation sets. A, B: the dose–response association between the protein risk score fordementia and incident dementia, with adjustments for age, sex, education, SBP, BMI, total cholesterol, physical activity andAPOE. C, D: Receiver operating characteristic curves showing the performance of the dementia protein risk score for predictingdementia at 3, 5 and 10 years. APOE, apolipoprotein E; AUC, area under the curve; BMI, body mass index; CI, confidence interval;HR, hazard ratio; SBP, systolic blood pressure; SD, standard deviation.
Credit: Yuanyuan Zhang, Yu Huang, Junyu Xue, Yiwei Zhang, Sisi Yang, Yanjun Zhang, Ziliang Ye, Xiaoqin Gan, Yiting Wu, Fanfan Hou, Xianhui Qin.
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.
Journal
General Psychiatry
Method of Research
Observational study
Article Title
Dementia risk prediction in individuals with prediabetes or diabetes: A novel multi-protein score with biological pathway analysis
Article Publication Date
15-Mar-2026