Researchers led by Piergiorgio Grillo report that 225 people with newly diagnosed idiopathic Parkinson's disease fall into three distinct biological groups when sorted by 22 biomarkers rather than by their symptoms. The analysis, posted to the preprint server medRxiv on 28 July 2026 and not yet peer reviewed, used data from the Parkinson's Precision Medicine Initiative (PPMI) cohort and excluded people with known genetic forms of the disease. The markers, measured in cerebrospinal fluid, whole blood, serum, plasma or urine, covered neurodegeneration, proteinopathy, neuroinflammation, mitochondrial impairment and lipid metabolism with autophagy-lysosomal dysfunction. Hierarchical clustering produced three groups the authors call Mitochondrial-Predominant, Mixed Pathology and Cryptic.

The separations were chemical, not clinical. The Mitochondrial-Predominant and Mixed Pathology groups both showed higher cerebrospinal fluid mitochondrial DNA deletions and ND1 than the Cryptic group. The Mixed Pathology group also showed higher serum neurofilament light chain, cerebrospinal fluid phosphorylated tau, GFAP and YKL-40, with lower plasma total ceramide, glucosylceramide and sphingomyelin than the other two. In the Cryptic group no dominant biological profile emerged. Members of the Mixed Pathology group were older and mostly male, but otherwise the clusters did not differ in baseline clinical or demographic measures, and the authors report no difference in clinical progression between the groups over five years of follow-up.

That absence is the point of interest and also the caution. Existing subtyping schemes divide patients by motor phenotype, tremor dominant against postural instability and gait disorder, or by a composite of motor, sleep, cognitive and autonomic features; earlier work, such as a 2022 study in npj Parkinson's Disease, largely attached biomarkers to those clinically defined categories. This analysis reverses the order and finds biological groupings that clinical description did not capture. The authors frame the result against the failure of antibody trials aimed at alpha-synuclein alone, among them the 52-week phase 2 SPARK trial of cinpanemab in 357 people with early Parkinson's, published in the New England Journal of Medicine in August 2022, which did not meet its primary endpoint.

The counterpoint is that a grouping which predicts nothing about a patient's course has not yet earned a place in trial design. The clusters were derived once, in one cohort, with the biomarkers that PPMI happens to have collected, and were not tested in an independent sample. The authors write that studies with a broader set of biomarkers are needed to confirm and expand the findings. Until a peer-reviewed version and a replication exist, the three subtypes describe patterns in stored samples, not categories a clinician can assign at the bedside.