Dysautonomia is prodromally driven by two distinct pathophysiological axes

Background. Dysautonomia spans a heterogeneous clinical spectrum, from orthostatic hypotension and syncope to peripheral neuropathy. Whether these entities share one prodromal biology or arise from distinct pathophysiological pathways has not been tested at population scale. Methods. In the UK Biobank (N = 501,946; median follow-up 13.6 years), we developed and internally validated prodromal risk models for five hospital-coded dysautonomia endpoints: a union composite (DYS; 25,714 events), orthostatic hypotension (OHT; 7,128), peripheral neuropathic disorders (NRP; 4,899), a cardiovascular autonomic composite (DYS_C; 22,155), and syncope (SYN; 17,017), with Parkinson's disease (PDS; 3,433) as a cross-prediction outcome. Models were $\ell_1$-regularized logistic regressions with nested cross-validation, built on baseline clinical, biochemical and hematologic measurements, prevalent conditions, and composite indices. We evaluated discrimination (Harrell's C-statistic), calibration, Cox proportional hazards, landmark analysis at 2 to 10 years, a cross-endpoint dissociation matrix with paired bootstrap intervals, a full-proteome model and a single-protein screen on the Olink subcohort (2,923 proteins; n = 52,997), and portable scoring rules. Findings. Nested C* ranged from 0.695 (SYN) to 0.801 (OHT), and every score remained associated with its endpoint after adjustment for age and sex (hazard ratio 1.77 to 3.19 per SD). Discrimination was stable over landmarks from 2 to 10 years for four endpoints and declined by 0.015 for NRP. The dissociation matrix resolved two axes: a cardiovascular autonomic cluster (OHT, DYS_C, SYN) in which no endpoint's own score beat the best other score (every within-cluster difference in C*{} at most 0.0056), and a peripheral neuropathic axis whose own score outperformed the best other score by $\pDelta$C*{} = +0.0121 [+0.0081, +0.0153] and each cardiovascular score by 0.021 to 0.025. The proteome separated the axes in the same way: 454 proteins were specific to NRP and 50 to the cardiovascular cluster, and the Jaccard overlap of protein sets was 0.788 to 0.895 within the cluster against 0.619 to 0.711 between NRP and the cluster. Full-proteome models reached nested concordance 0.671 to 0.797, with NEFL the leading protein in five of the six signatures. The cardiovascular scores predicted Parkinson's disease with age/sex-adjusted hazard ratios of 1.49 to 1.63 per SD, against 1.31 for the neuropathic score, although in discrimination they exceeded age alone by at most 0.009. Portable rules on 17 routine measurements retained 94.5 to 96.4% of full-model discrimination. Interpretation. Dysautonomia is not a single prodromal entity. It resolves into a cardiovascular autonomic axis and a peripheral neuropathic axis with different baseline signatures, different proteomic correlates, and different clinical implications. The cardiovascular axis carries a component of Parkinson's disease risk beyond age and sex, consistent with, but not proof of, a shared prodromal synucleinopathy. Both axes are detectable a decade before diagnosis from routine clinical data.