← All articles
An older adult seen from behind on a bright riverside promenade in the morning, a teal ribbon of light at one wrist, with pale sky and water on the left of the frame.

The artery-age score was age and pressure in disguise

ResearchBiomarkers

A person past sixty who is handed a second age can see which kind of number still sorts deaths, and which kind only repeats the page it was built from. Nothing here is a reason to chase a lower score.

In the US National Health and Nutrition Examination Survey, Hehe Wang, Jialei Chen, Yingying Wu, Junge Zhang and Ji Li took 3,972 adults aged 60 or older. In that group, 1,034 people later died, and 330 of those deaths were cardiovascular. The same comparison was run in two Chinese cohorts named in the paper, CHARLS (7,892 people) and CLHLS (560 people). The report is in GeroScience, dated 19 November 2026.

This article is information, not a substitute for a physician. The study records who died. It does not assign a drug, a supplement, or a change in any age score.

A vascular-age figure rebuilt from two numbers already on the page

Estimated pulse wave velocity is often offered as a stand-in for vascular age. In the equation these authors examined, it is computed from chronological age and mean blood pressure. They then removed age and looked at what remained. They call that remainder ePWV acceleration, and they asked whether it is a new quantity.

Inside each risk branch of the equation, the remainder could be reproduced from its own inputs. The fit was essentially exact: R² about 1, and a residual standard deviation below 1 × 10⁻¹¹ metres per second. The remainder correlated 0.990 with mean blood pressure. Once age, entered as a spline, and mean blood pressure were already in a survival model, ePWV acceleration added none (−0.0001) and could not be estimated jointly with mean blood pressure. The abstract states that figure in the sentence after the Harrell C changes for the blood scores.

Two blood scores still moved the ranking of deaths

Against that empty remainder the authors set two blood-based biological ages: PhenoAge, which the 2026 abstract does not assign to the clinical composite or to the methylation measure, and the estimator Klemera and Doubal published in 2006. The 2026 abstract does not list which assays went into either score in this analysis.

PhenoAge acceleration was tied to mortality at a hazard ratio of 1.82 per standard deviation (95% confidence interval 1.68 to 1.97). Klemera–Doubal acceleration was tied to mortality at 1.64 per standard deviation (1.53 to 1.76). The abstract attaches those ratios to mortality. It does not say that the ratios themselves were computed after age and blood pressure were already in the model. What it states separately is that both scores were well calibrated, and that after a spline for age and mean blood pressure were in the model they still improved discrimination. Harrell's C rose by 0.036 for PhenoAge acceleration and by 0.032 for Klemera–Doubal acceleration. The blood-based accelerations were uncorrelated with mean blood pressure. The same direction held in both Chinese cohorts and under machine-learning survival models. The abstract does not name those models.

A rise of about 0.03 on a C-statistic is a small shift in how a model ranks people. It is not a personal forecast, and it is not a count of deaths averted.

What an association of this kind does not license

The cohorts are observational. Nobody was assigned a lower PhenoAge. The paper does not report that changing either blood score lengthened life, delayed disease, or improved function. The US sample is adults of 60 and older. The abstract does not give ages for the Chinese cohorts, and it does not show the same split in younger adults. Follow-up length, the assay list, and separate hazard ratios for the 330 cardiovascular deaths are not in the abstract. The full text was closed when this was written, so those figures are not stated here.

The 2018 abstract names a composite clinical phenotypic age that captures differences in lifespan and healthspan and, in a two-step process, the DNA-methylation measure DNAm PhenoAge (PMID 29676998). The 2006 paper is a method for estimating biological age from biomarkers, and it treats chronological age itself as one input (PMID 16318865). Neither is a trial of a treatment. A second age printed beside a birthday is not a target to chase.

How these scores are built, and what a gap between them and a birthday does not mean, is in How biological age is calculated and Biological age versus chronological age.

Sources

  • Wang H, Chen J, Wu Y, Zhang J, Li J. Estimated pulse wave velocity acceleration is an algebraic function of its own inputs: incremental prognostic value of blood-based biological age beyond chronological age and blood pressure in older US and Chinese adults. GeroScience, 19 November 2026. DOI 10.1007/s11357-026-02668-z. PMID 43055908.

  • Levine ME and colleagues. An epigenetic biomarker of aging for lifespan and healthspan. Aging (Albany NY), 2018. PMID 29676998.

  • Klemera P, Doubal S. A new approach to the concept and computation of biological age. Mechanisms of Ageing and Development, 2006. PMID 16318865.

What this piece relies on

Cards

Articles in this section are educational and are not medical advice, a diagnosis, or a prescription. Consult a qualified professional before acting on anything you read here.

Read next