Dogs and Humans Share the Same Blood Metabolite Signatures of Death Risk

Understanding what molecular signals in the blood predict death risk has enormous implications for aging medicine, but human mortality cohort studies are prohibitively slow and expensive, often requiring thousands of participants tracked for decades. This study asks whether companion dogs — genetically diverse, living in human environments, and suffering similar age-related diseases — can serve as an accelerated translational model.

The research drew on the Dog Aging Project's Precision Cohort, 937 companion dogs across the United States with annual veterinary visits, owner surveys, and biospecimen collection including targeted LC-MS plasma metabolomics measuring 133 metabolites. Using time-dependent mixed-effects Cox proportional hazards models controlling for age, sex, weight, creatinine, and genetic relatedness, the team analyzed 104 deaths over a mean follow-up of 2.6 years (maximum 3.9 years). Roughly 17% of all metabolites were estimated to associate with mortality, with 23 reaching FDR < 5% significance.

The key finding is a striking concordance with human data. Across five human studies with sufficient metabolite overlap (37–64 shared metabolites each), the sign of the hazard ratio (HR) matched between dogs and humans in 64% of cases — far above chance (Fisher's exact p = 9.3×10⁻⁹). Pairwise Pearson correlations of HRs between dogs and each human study ranged from r = 0.46 to 0.74 (all p < 0.002; combined p = 4.0×10⁻¹⁷). When data from all nine human studies were aggregated into a geometric mean HR per metabolite across 93 shared metabolites, the correlation with dog HRs was r = 0.52 (p = 8.9×10⁻⁸). Notably, human studies themselves correlated strongly with each other (r = 0.37–0.85), suggesting a general mortality metabolome signature exists within the human blood metabolome — and dogs share it.

Among the specific metabolites most consistently elevated in dogs that died were pseudouridine, N2,N2-dimethylguanosine, and homocitrulline, while deoxycarnitine and homoarginine were depleted. These same metabolites are known markers of glomerular filtration rate in humans, pointing toward kidney function as a conserved physiological axis of aging and mortality across species. The fact that these signals emerged in under three years in dogs versus up to 22.5 years in human studies powerfully illustrates the translational efficiency of the canine model.

The study's implications are significant for geroscience. Dogs offer an ethically tractable, genetically variable, environmentally diverse, and clinically managed system for rapidly validating — and potentially discovering — human-relevant aging biomarkers. Future longitudinal waves of the Dog Aging Project may allow even more refined mechanistic insights, including whether interventions that alter these metabolites extend healthspan in dogs and, by inference, possibly humans.