Gompertz Law-Based Biological Age (GOLD BioAge): A Simple and Practical Measurement of Biological Ageing to Capture Morbidity and Mortality Risks

Abstract

Biological age reflects actual ageing and overall health, but current ageing clocks are often complex and difficult to interpret, which limits their clinical application. This study introduces a Gompertz law-based biological age (GOLD BioAge) model designed to simplify the assessment of ageing. We calculated GOLD BioAge using clinical biomarkers and found significant associations between the difference from chronological age (BioAgeDiff) and the risks of morbidity and mortality in the NHANES and UK Biobank. Using proteomics and metabolomics data, we developed GOLD ProtAge and MetAge, which outperformed the clinical biomarker models in predicting mortality and chronic disease risk in UK Biobank. Benchmark analyses demonstrated that the models outperformed common ageing clocks in predicting mortality across diverse age groups in both the NHANES and UK Biobank cohorts. Additionally, a simplified version called Light BioAge is created, which uses three biomarkers to assess ageing. The Light model reliably captured the mortality risk across three validation cohorts (CHARLS, RuLAS, and CLHLS). It significantly predicted the onset of frailty, stratified frail individuals, and collectively identified individuals at high risk of mortality. In summary, the GOLD BioAge algorithm provides a valuable framework for the assessment of ageing in public health and clinical practice.

Keywords: aging clocks; biological age; frailty index; metabolomics; proteomics.

Figure 1

GOLD BioAge and its Association with Health‐Related Factors. Panel A illustrates the exponential relationship between mortality hazard and biological age (BA) and chronological age (CA). The “Diff” referred to the difference between GOLD BioAge and CA, termed GOLD BioAgeDiff. The scatter plot B) shows the strong correlation between GOLD BioAge (estimated biological age) and CA. The estimated coefficients for CA and biomarkers C), used to calculate GOLD BioAge, were displayed, with the mean biomarker values of young adults serving as the reference. Abbreviation: MCV: mean cell volume; RDW: red cell distribution width, ALP: alkaline phosphatase, LYM%: lymphocyte percent, WBC: white blood cell count, GGT: gamma glutamyl transferase (GGT). The distribution of GOLD BioAgeDiff in NHANES D). The correlations of GOLD BioAgeDiff with counts of age‐related chronic diseases E), self‐rated health F), and unhealthy lifestyles (G).

Figure 2

The associations of BioAgeDiff and risks of mortality. Survival plots for individuals categorized by BioAgeDiff, PCA age, and MDS in the NHANES cohort are presented. The high and low risk groups represent the top and bottom 25% of the age‐stratified population (ages 45–54, 55–64, 65–74, and 75–85 years). PCA: principal component analysis; MDS: Mahalanobis distance statistics.

Figure 3

The associations of GOLD ProtAge, MetAge, and BioAge with mortality in UK Biobank. A) Correlations between the three aging clocks and chronological age. B) C‐index values from survival analysis and the AUC for 10‐year mortality prediction, comparing the three aging clocks and chronological age, with results for all‐cause (age‐stratified) and cause‐specific mortality. C) Survival curves for individuals classified by ProtAgeDiff, MetAgeDiff, and BioAgeDiff in the general population (top panel) and young adults (bottom panel, <45 years old). High and low risk groups are defined as the top and bottom 25% of the population. ProtAgeDiff, MetAgeDiff, and BioAgeDiff represent the differences between ProtAge, MetAge, and BioAge and chronological age, respectively. The C‐index for ProtAgeDiff and its subpanels and proteins are shown. ProtAgeDiff consisted of CardioDiff, InfamDiff, NeuroDiff, and OncoDiff, which were linear combinations of cardiometabolic, inflammatory, neurological, and oncological proteins. E) Density plots and G) a correlation heatmap (filled with Pearson correlation coefficients) of these subpanels are presented. F) Survival plots based on ProtAgeDiff subpanels and H) the risk score, which was the count of high‐risk factors derived from ProtAgeDiff subpanels.

Figure 4

Associations between GOLD ProtAge, MetAge, and BioAge and the incidence of age‐related chronic diseases. The forest plots A) illustrates the hazard ratios and C‐index for ProtAge, MetAge, and BioAge across chronic diseases in the UKB. These associations were adjusted for age and sex. MI: myocardial ischemia; COPD: chronic obstructive pulmonary disease; CI: confidence interval. Survival plots are displayed based on the differences between ProtAge, MetAge, and BioAge relative to chronological age, referred to as ProtAgeDiff, MetAgeDiff, and BioAgeDiff. The high and low risk groups correspond to the top and bottom 25% of the population, respectively.

Figure 5

Comparison of GOLD BioAge and other common aging clocks in predicting mortality in NHANES and UKB. The C‐index in survival analysis A) and AUC value of 10‐year mortality prediction B) of these aging clocks are shown. Both all‐cause (age‐stratified) and cause‐specific mortality are considered. The highest value is marked with bold. The BioAge, Light, Levine and KDM referred to the GOLD BioAge, its light version, Levine's phenotypic age, KDM algorithm derived age, respectively.

Figure 6

The Light BioAge and its associations with age‐related factors and outcomes. The correlation of Light BioAge with age A). And the difference of Light BioAge with age also correlated with age B) and its distribution C). The correlations of Light BioAgeDiff with counts of age‐related chronic diseases D), self‐rated health E), and unhealthy lifestyles F). The survival plot G) based on Light BioAgeDiff levels, with the top and bottom 25% of the population representing the high and low risk groups. The forest plots H) show the hazard ratios and C‐index of Light BioAge in relation to chronic diseases, adjusted for age and sex. MI: myocardial ischemia; COPD: chronic obstructive pulmonary disease; HR: hazard ratio; CI: confidence interval.

Figure 7

Validations of the Light BioAge in three independent cohorts. The correlations A) of Light BioAge with age in CHALS, RuLAS and CLHLS. The ROC curves B) of Light BioAge (solid lines) and age (dotted lines) for predicting mortality across all samples, and within age‐stratified groups (<80, ≥80 years old). Survival plots C) depict mortality trajectories of individuals categorized based on Light BioAgeDiff levels, with the top and bottom 25% represented as high and low risk groups in CHARLS, RuLAS, and CLHLS.

Figure 8

The Light BioAge, its dynamics and mortality in CHARLS. Illustration A) detailing the study designs across five waves in CHARLS. Correlation B) between Light BioAge values in wave 1 (2011) and wave 3 (2015). Scatter plot C) displays Light BioAgeDiff in wave 1 (2011) and wave 3 (2015), with dotted lines indicating Light BioAgeDiff values of 0 and 5. Individuals were divided into 7 groups based on the changes in Light BioAgeDiff, with survival plots D) and forest plots E) provided. Model 1 represented the crude model, while Model 2 adjusted for age and sex.

Figure 9

The Light BioAgeDiff, frailty and mortality in CHARLS. The boxplots A) of Light BioAgeDiff across robust, prefrail, and frail individuals in CHARLS waves 1 (2011) and 3 (2015), with statistical significance determined using Wilcoxon tests. Forest plots B) illustrates the associations between Light BioAgeDiff and incidence of frailty. The odd ratios were calculated through continuous and category Light BioAgeDiff (Q1‐4: quartiles), adjusted by age and sex. The mediation models C) of Light BioAge (wave 1, 2011), frailty index (wave 3, 2015) and mortality (wave 3–5, 2015–2020). The change in frailty index was calculated based on assessments from waves 1 and 3. SRH: self‐rated health; ADL: activities of daily living; ADE: average direct effect; ACME: average causal mediated effect. The survival plots of individuals according to frailty status D), and both frailty status and levels of Light BioAgeDiff E).