Phenotypic illusion but genetic decoupling between osteoarthritis and Parkinson's disease

Observational studies have repeatedly linked osteoarthritis (OA) with Parkinson's disease (PD), but whether this reflects causality, shared genetic architecture, or artefactual mechanisms remains unresolved. We triangulated six layers of evidence: two population cohorts, genome-wide and local genetic architecture, bidirectional Mendelian randomization, shared-locus mapping, cell-type-specific analyses, and single-cell expression validation. In CHARLS (n = 7,587), knee OA was associated with higher PD risk (OR 1.88, 95% CI 1.24-2.84) with consistent depression-mediated attenuation (approximately 34%). In NHANES 2011-2018, the association was positive but imprecise (OR 1.94, 95% CI 0.44-8.52) and non-specific. Genetically, OA and PD were decoupled: genetic correlation was null-to-negative, no local correlation survived correction, and bidirectional MR was null across all pre-specified chains and sensitivity analyses. Conditional FDR identified 732 independent shared loci without directional bias; colocalization converged on a single robust shared region at MAPT 17q21.31 (posterior probability 0.975). Cell-type-specific and single-cell expression analyses did not support an expression-mediated pathway. The OA-PD association is therefore largely consistent with a phenotypic illusion (reproducible but proxy-sensitive and depression-mediated), with genome-wide genetic decoupling and only locus-restricted sharing at 17q21.31.