Chronos Biosciences
ESContact
← Articles

State of the field

The measurement problem

November 5, 2026 · 3 minute read · 4 references

Interventions with a reproducible effect on the lifespan of laboratory mice do exist. What does not exist, to date, is an accepted functional endpoint that would let anyone state whether one of them rejuvenated a human tissue.

What can be measured

In the mouse the endpoint is direct: you count how long an animal lives and how long the control lives. On that basis the field has accumulated reproducible results. Inhibiting IL-11 signalling, for instance, extends median lifespan in male mice by 22.5 %, and by up to roughly 25 % in females Widjaja et al., Nature, 20241. The figure is solid and the methodology behind it is not in dispute.

The problem appears when the species changes. A clinical trial measuring survival in healthy humans would have to run for decades and recruit populations no sponsor can sustain. An intermediate endpoint is therefore needed: something measurable in months that works as a credible stand-in for the effect sought.

What is used instead

The most widespread stand-in is the DNA methylation age estimator. The most cited one was developed from 8,000 samples across 82 datasets, covering 51 healthy tissues and cell types Horvath, Genome Biology, 20132. As a predictor of chronological age its performance is remarkable, and that is what it was built and validated for.

That is also its limitation. An estimator trained to predict the age of a sample is not validated to measure the effect of an intervention on that sample. These are two different statistical problems. In practice it means the CpG positions the model weights can shift without any tissue function changing, and the shift would read the same on the estimator.

A correlative marker accompanies the phenomenon. A causal marker responds to the intervention in the same direction and magnitude as function does. Evaluating a treatment requires the second, and the first does not become the second through long use.

The practical consequence

With no accepted endpoint, every result in the field sits in an awkward position. A positive result is arguable, because one can always object that the marker does not measure what is claimed. A negative result is dismissible, for the same reason in reverse. In that situation evidence accumulates without resolving.

The first senolytic studies in people illustrate the problem from another angle. The two published in 2019 — in idiopathic pulmonary fibrosis Justice et al., EBioMedicine, 20193 and in diabetic kidney disease Hickson et al., EBioMedicine, 20194 — were open-label and without a control group. They entered through a specific pathology, which is the only route available: no agency recognises ageing as an indication, so there is no direct approval path for a treatment that targets it.

What a usable measure would have to satisfy

Three conditions, all at once:

  1. Functional. That it measures a capacity of the tissue — responding to damage, recovering proliferation, depositing organised matrix — and not a molecular signature correlated with age.
  2. Measurable within a useful window. Weeks or months, not decades, and with available instrumentation.
  3. Causal with respect to the intervention. That its variation mediates the observed effect, established by mediation analysis rather than by correlation after the fact.

None of the markers in use satisfies all three. Building one that does is a problem of method before it is a problem of biology, and it is the one with the widest consequences: it would unlock the evaluation of any candidate, including those of other groups.

How we handle it

In the designs we draw up, the primary endpoint is functional and methylation estimators appear as secondary, explicitly declared correlative. When a hypothesis proposes a mechanism, the design incorporates measurement of that mechanism at the moment the intervention is applied, so that the effect can be attributed or ruled out by mediation rather than by association.

We publish no outcome figures of our own because we have not yet run any experiment. When there are any, they will state the model they come from.

References

  1. Widjaja AA, et al. (2024) Inhibition of IL-11 signalling extends mammalian healthspan and lifespan Nature. 10.1038/s41586-024-07701-9 Checked against the primary source
  2. Horvath S. (2013) DNA methylation age of human tissues and cell types Genome Biology 14(10):R115. 10.1186/gb-2013-14-10-r115 Checked against the primary source
  3. Justice JN, et al. (2019) Senolytics in idiopathic pulmonary fibrosis: first-in-human, open-label, pilot study EBioMedicine 40:554-563. 10.1016/j.ebiom.2018.12.052 Checked against the primary source
  4. Hickson LJ, et al. (2019) Senolytics decrease senescent cells in humans: dasatinib plus quercetin en enfermedad renal diabética EBioMedicine 47:446-456. 10.1016/j.ebiom.2019.08.069 Corrigendum: EBioMedicine 2020;52:102595 (doi 10.1016/j.ebiom.2019.12.004). Checked against the primary source