A generator-matrix model quantifies the limited contribution of measured biomarkers to human mortality acceleration
In the authors' words
Aging clocks and biomarkers are increasingly used as if they were the mechanism that drives mortality. But a mechanism, unlike a thermometer, must satisfy three conditions: it must account for the mortality signal, be causal, and be the layer that reverses when aging is reversed. Using public or provider-restricted de-identified data, we test all three. First (accounting), a Markov generator-matrix model with death as an absorbing state, fitted jointly to biomarkers and mortality in NHANES (n=23,512) and replicated in the Health and Retirement Study, assigns most of the Gompertz rise in mortality to a component latent to measured blood biomarkers (92.7%; 89.1% under a mean-field re-specification; 91.5% on replication). Second (causation), a positive-control-calibrated, two-platform cis-pQTL Mendelian-randomization design (UKB-PPP, deCODE) detects known causal proteins (LPA, IL6R) yet finds the measurable inflammatory, renal and growth-signalling markers null. Third (reversibility), at donor level reprogramming reverses a chronological clock (-9.7 and -22.4 yr), whereas a causality-enriched damage clock shows no detectable reversal while somatic identity is retained and moves only as pluripotency is approached. On these data the biomarkers meet none of the three conditions; aging measures appear to track mortality risk rather than its cause, a caution for their use as surrogate end-points.
Appeared: Wednesday, September 23. medRxiv. Preprint, not yet peer-reviewed.