A word of caution on the regression to the mean phenomenon in (biological) age prediction

When predicting ‘biological age’, always consider whether regression to the mean could be an issue before drawing conclusions. Marije Sluiskes, Jelle Goeman, Marian Beekman, Eline Slagboom, Hein Putter and Mar Rodríguez-Girondo posted a PubPeer comment (a postpublication peer review platform) to a recent publication of Daunay et al. 2022 (Aging) which didn’t consider this. Daunay et al. evaluate and compare four previously proposed epigenetic clocks based on a small number of CpG sites in three sets of samples: semi-supercentenarians (mean age ± sd: 101.3 ± 1.4), offspring of nonagenarians and centenarians (61.2 ± 6.1), and individuals from the general population (56.0 ± 4.7). The authors conclude there is an indication of decelerated epigenetic and biological aging in the first two groups as compared to the last. In our comment, we raise a methodological issue that makes us skeptical about their conclusion.

Originally, we sent our comment as a letter to the editors of Aging, but our letter didn’t get considered due to “publication priorities”. Fortunately there’s PubPeer! See also the GitHub for a brief explainer on the regression to the mean phenomenon in a biological age prediction context.

https://pubpeer.com/publications/A9B0D98C173B74C41F55C340EF4945
https://github.com/marije-sluiskes/regression-to-the-mean-explainer/blob/main/Explainer.md