Clarifying the biological and statistical assumptions of cross-sectional biological age predictors

There is variability in the rate of aging among people of the same chronological age. The concept of biological age is postulated to capture this variability, and hence to better represent an individual’s true global physiological state than chronological age. Predictors of biological aging are often generated based on cross-sectional data, using biochemical or molecular data (also called markers) as predictor variables. It is assumed that the difference between predicted and chronological age of an individual is informative of that person’s rate of aging Δ, having either a younger or older marker profile. We show that the most popular cross-sectional biological age predictors—based on multiple linear regression, the Klemera-Doubal method or principal component analysis—all rely on the same strong underlying assumption, namely that a candidate marker of aging’s association with chronological age is directly informative of its association with the aging rate Δ. We call this the identical-association assumption and prove that it is untestable in a cross-sectional setting. Using synthetic data, we illustrate the consequences if the assumption does not hold: in such scenarios, there is no guarantee that the weights that a cross-sectional method assigns to candidate markers are informative of the underlying truth, to the extent that one might as well assign weights to candidate markers at random. This critical limitation holds for any statistical method based on cross-sectional data. We use a real data example to illustrate that the extent to which the identical-association assumption holds is of direct practical relevance for any researcher interested in developing or interpreting cross-sectional biological age predictors. ### Competing Interest Statement The authors have declared no competing interest.