GOLD BioAge Uses Gompertz Math to Turn Routine Blood Tests Into a Biological Age Score

Chronological age is an imperfect proxy for biological aging. Two people of the same calendar age can differ dramatically in physiological function, disease burden, and remaining lifespan. Existing biological age clocks — from DNA methylation arrays to complex machine-learning models — often require expensive or technically demanding assays that limit routine clinical use. This paper introduces GOLD BioAge, a Gompertz law-based algorithm designed to be both mathematically interpretable and practically deployable.

The Gompertz distribution describes the exponential rise in mortality hazard with age, and the authors exploit its hazard function to anchor biological age to actual mortality risk. Starting with NHANES data (39,348 participants, mean age 49.5 ± 18.0 years), they applied LASSO-Cox regression to select nine biomarkers from an initial panel of 26: red blood cell distribution width (RDW), albumin (ALB), creatinine, alkaline phosphatase (ALP), gamma-glutamyl transferase (GGT), white blood cell count (WBC), lymphocyte percent, mean cell volume (MCV), and one additional marker. GOLD BioAge is the linear sum of chronological age and weighted biomarker values, achieving a correlation of R = 0.969 with chronological age. The key clinical metric is BioAgeDiff — biological age minus chronological age — which directly quantifies accelerated or decelerated aging at the individual level.

In NHANES and UK Biobank analyses, higher BioAgeDiff correlated significantly with counts of age-related chronic diseases, worse self-rated health, and unhealthy lifestyle behaviors. Benchmark comparisons showed GOLD BioAge outperformed several established aging clocks (including phenotypic age and PCAge variants) in predicting all-cause mortality across diverse age groups in both cohorts. The algorithm was then extended to UK Biobank omics data, producing GOLD ProtAge (proteomics-based) and GOLD MetAge (metabolomics-based), which exceeded the clinical biomarker model in predicting mortality and chronic disease risk — suggesting the framework scales naturally to richer molecular data.

Recognizing that even nine biomarkers may be impractical in resource-limited settings, the team developed Light BioAge using only three biomarkers. This simplified model was validated in three independent Chinese longitudinal cohorts: CHARLS, RuLAS, and CLHLS. Light BioAge reliably predicted mortality risk across all three datasets, predicted onset of frailty, stratified already-frail individuals by risk, and identified high-risk individuals when combined with frailty assessment — demonstrating cross-population generalizability.

The study's key contribution is methodological elegance: by grounding biological age directly in the Gompertz hazard function, every coefficient has a biologically interpretable meaning tied to mortality risk. The linear formula is calculable without specialized software, making it suitable for point-of-care or population-level health screening.