Tier 2 — strong

DunedinPACE, a DNA methylation biomarker of the pace of aging

Belsky DW, Caspi A, Corcoran DL, Sugden K, Poulton R, Arseneault L, Baccarelli A, Chamarti K, Gao X, Hannon E, Harrington HL, Houts R, Kothari M, Kwon D, Mill J, Schwartz J, Vokonas P, Wang C, Williams BS, Moffitt TE
eLife 2022 Volume 11, article e73420

Bibliography

DOI
10.7554/eLife.73420
Funding
Verbatim from the article: "This research was supported by US-National Institute on Aging grants AG032282, AG061378, AG066887, and UK Medical Research Council grant MR/P005918/1."
Competing interests
The article carries a competing-interest declaration in its back matter (see page 18 of the published PDF). Readers evaluating commercial biological-age testing should consult it directly, as several authors are associated with the measure.

Study snapshot

DesignBiomarker development and validation study using elastic-net regression on genome-wide DNA methylation data, restricted to probes with high test-retest reliability, then validated in five additional datasets.
ModelDunedin Study 1972–1973 birth cohort, plus validation in five further human datasets including the Normative Aging Study and E-Risk.
SampleDevelopment cohort: the Dunedin Study birth cohort of 1,037 individuals, with 19 organ-system integrity indicators measured at four time points (ages 26, 32, 38 and 45). Validated in five additional datasets.
InterventionNone. This is an observational measurement study, not a supplementation trial.
DurationTwo decades of within-individual follow-up distilled into a single-timepoint blood test.
EndpointsTest-retest reliability of the DunedinPACE measure; Association with morbidity, disability and mortality; Association with childhood adversity in young adults; Incremental prediction beyond GrimAge

What the study showed, in plain terms

DunedinPACE is the blood test behind most claims that a supplement lowered someone's biological age. Understanding what it actually measures is the fastest way to read those claims correctly.

Most biological-age clocks estimate how old you are. DunedinPACE estimates how fast you are ageing right now — a rate, not an age. It was built from the Dunedin Study, which followed every one of the 1,037 babies born in one New Zealand city in 1972–73 and measured the integrity of nineteen organ systems at ages 26, 32, 38 and 45. That gave the researchers something almost nobody else has: two decades of directly observed decline in the same people. They used machine learning to compress that into a pattern readable from DNA methylation in a single blood sample.

The version matters. An earlier measure, DunedinPoAm, used only three assessment waves and had only moderate test-retest reliability — which made it poorly suited to detecting change within one person before and after a treatment. DunedinPACE added a fourth wave and kept only the methylation probes that give consistent readings between technical replicates. Reliability improved substantially. That is precisely why it turns up in supplement studies: it is one of the few clocks stable enough to be worth measuring twice.

None of this makes a DunedinPACE change a health outcome. It predicted morbidity, disability and mortality in this research, and it added prediction beyond GrimAge. A supplement study that reports a DunedinPACE shift in a handful of uncontrolled users has produced a movement in a surrogate measure, not a demonstration that anyone lived longer or better.

Key findings

  • DunedinPACE quantifies the pace of biological ageing from a single blood sample, expressed as a rate rather than an age.
  • It was distilled from two decades of within-individual decline across 19 organ-system integrity indicators measured at four time points in the Dunedin 1972–73 birth cohort.
  • Restricting the methylation dataset to high test-retest-reliability probes produced high reliability — the specific property that makes a clock usable for before-and-after measurement in a trial.
  • DunedinPACE was associated with morbidity, disability and mortality, with effect sizes similar to GrimAge, and added incremental prediction beyond GrimAge for incident morbidity, disability and mortality.
  • Faster ageing was indicated in young adults who had experienced childhood adversity.
  • The measure was validated in five datasets beyond the development cohort.

What this study can and cannot tell us

This is a biomarker paper, not an intervention trial. It establishes that the measure is reliable and predicts hard outcomes in observational data. It does not establish that changing the measure changes those outcomes — the surrogate-endpoint problem in its purest form.

For reading supplement marketing, the critical distinction is between DunedinPACE and DunedinPoAm, and between either of those and age-estimating clocks such as GrimAge, PhenoAge, Horvath or TruAge. A reported reduction in biological age may come from any of these, and they are not interchangeable. A claim that omits which clock was used, or reports a change without a control group, cannot be evaluated.

The development cohort is a single-year birth cohort from one New Zealand city, which is a strength for internal consistency and a constraint on generalisability. Validation datasets extend but do not eliminate that constraint.

Several authors are associated with the measure and the article carries a competing-interest declaration. That is normal for methods papers and is disclosed, but it matters when the same measure is sold commercially.

Editorial review

Reviewed by the Biohack Blueprint research team

Last verified

Dr. Khaled Mahmoud, MBBChCritical Care Medicine, Longevity Science & Editorial Direction Dr. Abdulla Mostafa, PhDCellular Senescence, Neurological, Mitochondrial & Brain-Ageing Biology Dr. Amr Yasser, PhDCardiometabolic, Musculoskeletal, Gut & Bioavailability