Researchers led by Raquel Real and Huw Morris at University College London measured 3,998 protein-detecting aptamers in baseline serum samples using the SomaScan assay and report that 32 protein targets were significantly associated with later progression to dementia in people with Parkinson's disease. The work, posted to medRxiv on 30 June 2026, is a preprint and has not been peer reviewed. The discovery cohort was Tracking Parkinson's, a UK multicentre observational study of 834 participants recruited within 3.5 years of symptom onset and reassessed every 18 months; 335 protein targets were nominally significant there. Those candidates were then tested in 371 participants from the Oxford Parkinson's Disease Centre Discovery cohort, and results from the two were combined by meta-analysis. Every target significant in the meta-analysis pointed the same way in both cohorts. The median time between the baseline blood sample and the point at which dementia criteria were met was 3.11 years in the Tracking Parkinson's cohort and 3.79 years in the Oxford cohort.

The proteins named do not track the familiar Alzheimer's story. The authors group them into synaptic plasticity, protein degradation and lysosomal function, and extracellular matrix organisation, and their pathway analysis after adjustment for APOE genotype pointed to extracellular matrix organisation and regulation of synapse organisation. Two of the hits, TBCA and S100A13, turned out to be strongly correlated with the APOE e4-tagging variant rs429358 in both cohorts, and their association with dementia was attenuated once APOE genotype was added as a covariate. Fourteen of the original targets remained significant with that adjustment and 13 further targets appeared. A separate Mendelian randomisation analysis, which uses genetic variants as proxies, suggested that changes in RTN4R, the Nogo receptor, may be causally associated with Lewy body dementia risk.

The context is a field with no clinically validated blood test for predicting Parkinson's disease dementia. The authors note that small targeted studies have suggested Alzheimer's-related plasma markers and neurofilament light might predict cognitive decline in Parkinson's, but nothing has been validated for clinical use. Markers that identify who will decline would matter most for trial design, since studies enrolling unselected Parkinson's populations dilute any cognitive signal they are trying to detect.

The counterpoints are substantial and several are the authors' own. The findings rest on a single proteomic platform and have not been checked against a different assay, so a signal could reflect the aptamer rather than the protein. There is no amyloid co-pathology data, which the authors say makes it impossible to separate changes specific to Lewy body pathology. Both cohorts are UK-based and the authors call for testing in more diverse populations. The survival models described adjust for age at baseline, sex and latent technical variables, with APOE genotype added in sensitivity analysis; baseline cognitive score is not listed among the covariates, and the preprint reports associations rather than a validated risk score benchmarked against routine cognitive testing.