The Physiological Aging Index Only Slightly Improves on the PhenoAge Clock
Any sufficiently complex set of biological measures can be used to produce an aging clock: researchers establish a database of the measures in people of different ages and apply machine learning techniques to produce an algorithm that maps an individual's measured data to a predicted age. That doesn't mean it is a good clock, however. One then has to validate the algorithm against data from other populations and see how well it does in predicting disease, mortality, and other outcomes of interest. Much of the development of clocks is focused on epigenetic data, but distinctly from that line of research, the scientific community is also exploring clocks built on clinical measures, such as blood chemistry, physical performance, and so forth.
Of the clinical measure clocks, PhenoAge is probably the most widely used, both inside and outside the scientific community. Its popularity may derive from ease of use, as it employs only 9 parameters that can be obtained from a complete blood count and a couple of other blood chemistry measures. If proposing a new, more complicated clinical measure clock, one would have to demonstrate that it improves on PhenoAge to a meaningful degree. In today's open access paper, researchers fail to achieve this goal. Their Physiological Aging Index uses 17 parameters and only marginally improves on PhenoAge. It is perhaps interesting to consider why it is that clocks with fewer parameters can still perform well, even in diverse populations.
It has been known that for individuals of the same chronological age (CA), those with obesity, long-term nicotine exposure, or lower socioeconomic status are more likely to experience adverse health outcomes and increased mortality risk. Thus it is important to measure one's biological age (BA) to identify individuals with accelerated aging and to develop precision prevention and intervention strategies for major chronic diseases in an aging population. To date, researchers have developed a variety of predictors of BA using biomarkers such as telomere length, DNA methylation, gene expression, metabolites, or clinical biomarkers. While BA indices based on genomic data, such as DNA methylation, are accurate in predicting CA, clinical biomarkers are generally more affordable, interpretable, and modifiable. However, existing clinical biomarker predictors were primarily based on supervised models with CA as the training label and thus may have limited value to predict disease risks independent of CA.
In this study, we propose a physiological aging index (PAI) based on 36 clinical biomarkers from the Dongfeng-Tongji (DFTJ) cohort of elderly Chinese. In the DFTJ training set (n = 12,769), we identified 25 biomarkers with significant nonlinear associations with mortality, of which 11 showed insignificant linear associations. By incorporating nonlinear effects, we selected CA and 17 clinical biomarkers to calculate PAI. PAI aims to measure an individual's BA based on routine clinical biomarkers in the blood. We use restricted cubic spline (RCS) Cox models to capture potential U-shaped relationships between clinical biomarkers and mortality, and determine the optimal value of each biomarker for subsequent piece-wise linear transformation. We define PAI as a linear combination of CA and the transformed biomarkers, as well as ΔPAI as the residual of PAI after regressing on CA. Thus, ΔPAI measures physiological aging acceleration independent of CA.
In the DFTJ testing set (n = 15,904), PAI predicts mortality with a concordance index (C-index) of 0.816, better than CA (C-index = 0.771) and PhenoAge (0.799). ΔPAI was predictive of incident cardiovascular disease and its subtypes, independent of traditional risk factors. In the external validation set of the UK Biobank (n = 296,931), PAI achieved a C-index of 0.749 to predict mortality, remaining better than CA (0.706) and PhenoAge (0.743). In both DFTJ and UK Biobank, PAI calibrated better than PhenoAge when comparing the predicted and observed survival probabilities. Furthermore, ΔPAI outperformed any single biomarker to predict incident risks of eight age-related chronic diseases.