Researchers modelled predicted-age distributions rather than point estimates in over 450,000 UK Biobank participants, separating the mean deviation (biological age gap) from the width of the distribution (biological age uncertainty). Uncertainty was weakly correlated with the gap and rose with disagreement between ageing signals across molecules and organs. Per standard deviation it carried mortality hazard ratios of 1.24 for NMR-derived and 1.40 for proteomic-derived measures, independent of the gap, with replication in three external cohorts using routine blood tests.
Why it is interesting: Treats the width of a predicted-age distribution as a predictor of mortality and healthspan loss separate from its mean.