Moving Beyond Mortality Clocks: Responsive Rate of Aging Biomarkers with Assoc. Professor Dan Belsky
I. Executive Summary
In this expert transcript synthesis, Associate Professor Daniel W. Belsky (Columbia University Mailman School of Public Health) outlines a paradigm shift in biogerontology: moving from static cumulative “biological age” metrics (“odometers”) to dynamic, intervention-responsive “pace of aging” biomarkers (“speedometers”). First- and second-generation epigenetic clocks (e.g., Horvath, PhenoAge, GrimAge) quantify accumulated biological damage over a lifetime. Consequently, in a standard 1-to-2-year geroprotective clinical trial—which alters only ~2% of a human lifespan—these odometer clocks lack the statistical sensitivity to detect meaningful therapeutic signal. To resolve this, Belsky and colleagues developed DunedinPACE, an epigenetic algorithm trained on 20-year longitudinal physiological decline within the Dunedin Birth Cohort (1,037 individuals born in New Zealand in 1972–1973).
By tracking 19 multiorgan biomarkers across four timepoints from age 26 to 45 with a 95% cohort retention rate, the Dunedin Study eliminated two fundamental epidemiological confounds: survival bias (where unhealthy aging individuals die or drop out of older cohorts) and cohort effects (where cross-sectional age comparisons confound aging with historical differences in childhood nutrition, healthcare, and environmental exposures). The Dunedin data demonstrated that multiorgan physiological decline begins in early adulthood (age 26) and varies widely among chronologically identical peers (ranging from <0 to >3 years of biological aging per chronological year), directly predicting midlife motor balance, grip strength, cognitive decline, and facial aging.
This speedometer framework was validated in human clinical trials through the 2-year CALERIE trial (~12% caloric restriction in non-obese adults). While traditional odometer clocks failed to show significant divergence between intervention and control groups, DunedinPACE demonstrated a statistically significant reduction in the pace of aging (Belsky et al., 2023). This sensitivity was replicated using Framingham PACE (developed in an older cohort), proving that responsiveness is a feature of the longitudinal rate-of-change methodology rather than sample age. Finally, Belsky details the ARPA-H-funded FAST (Functionally Activated Surrogate Targets) initiative, a multi-million-dollar program designed to identify multi-omic biomarkers (proteomics, metabolomics, epigenetics) that respond across multiple FDA-approved geroprotective candidate classes (metformin, rapamycin, GLP-1 agonists, SGLT2 inhibitors) to establish validated surrogate endpoints for preventive longevity medicine.
II. Insight Bullets
- The “Odometer vs. Speedometer” Distinction: Cumulative biological age algorithms (Horvath, PhenoAge, GrimAge) function like odometers by measuring total lifetime biological damage, whereas DunedinPACE functions as a speedometer by quantifying the real-time rate of biological change per unit of chronological time (Belsky et al., 2022).
- Clinical Trial Sensitivity Gap: A 2-year geroprotective trial edits only ~2% of an adult’s lifespan; cumulative “odometer” clocks are mathematically insensitive to such short-term rate modifications, whereas “speedometer” metrics immediately reflect therapeutic rate-of-change deceleration.
- Architecture of the Dunedin Birth Cohort: The Dunedin Study tracked 1,037 individuals born in New Zealand in 1972–1973, collecting comprehensive physiological assessments at ages 26, 32, 38, 45, and 50 (Belsky et al., 2015).
- Eradication of Survival Bias: Studying aging in a birth cohort up to age 45 with a 95% retention rate eliminates survival bias, preventing the selective dropout or mortality of rapid agers that distorts gerontological studies in older populations (e.g., ages 70+).
- Mitigation of Demographic Cohort Confounders: Comparing a 20-year-old to a 70-year-old in cross-sectional studies conflates biological aging with cross-generational differences in birth-year exposures (e.g., historical nutrition, healthcare availability, environmental toxins); longitudinal birth cohort tracking isolates true intra-individual aging.
- Early Adulthood Onset of Aging (Age 26): Comprehensive physiological tracking reveals that multiorgan decline (cardiovascular, pulmonary, renal, hepatic, metabolic, immune) begins in early adulthood and follows the exact same physiological vectors seen in advanced age.
- Inter-Individual Divergence in Young Adults: Chronologically identical 38-year-olds display paces of aging ranging from under 0 years of physiological change per chronological year to nearly 3 years of biological aging per calendar year.
- Functional Manifestations of Accelerated Midlife Aging: Individuals with an accelerated pace of aging by age 38/45 exhibit objective functional deficits, including weaker grip strength, impaired motor balance (unipedal stance test), diminished neurocognitive performance, and older subjective self-rated health.
- Validation of Perceived Facial Aging: Facial photograph ratings by independent observers (and machine-learning computer vision models) significantly correlate with internal multiorgan pace of aging, validating facial aesthetics as a biomarker of systemic biological decline.
- Modern Framework for Aging Biomarker Validation: Modern geroscience rejects simple correlation with chronological age or mouse-model fidelity as sufficient validation, adopting the Biomarkers of Aging Consortium consensus framework (Moqri et al., 2023).
- Three-Tier Biomarker Criteria: Valid biological aging biomarkers must demonstrate: (1) prediction of future morbidity, functional decline, and mortality; (2) retroactive sensitivity to early-life adversity/exposures; and (3) responsiveness to geroprotective interventions.
- CALERIE Trial Human Caloric Restriction: The Phase 2 CALERIE 2 trial demonstrated that 2 years of ~12% caloric restriction in non-obese adults reduced cardiometabolic risk, downregulated systemic inflammation, and maintained organ integrity without causing organ wasting or severe adverse events (Ravussin et al., 2015).
- Clock Discordance in CALERIE: In the CALERIE trial, traditional chronological and mortality-prediction clocks (PhenoAge, GrimAge) ticked at identical rates (+2 years) in both control and intervention groups, whereas DunedinPACE demonstrated a statistically significant reduction in pace of aging (Belsky et al., 2023).
- Methodological Validation via Framingham PACE: Developing a pace-of-aging algorithm in the older Framingham Heart Study offspring cohort (Framingham PACE) yielded the same responsiveness in CALERIE as DunedinPACE, confirming that clock sensitivity is driven by longitudinal rate-of-change modeling, not cohort age.
- The ARPA-H FAST Initiative Strategy: The ARPA-H-funded FAST (Functionally Activated Surrogate Targets) initiative flips biomarker discovery by analyzing bio-specimens from completed trials of proven geroprotective candidate drugs to identify molecules that respond dynamically to therapy.
- Geroprotective Candidate Selection Criteria for FAST: Candidate drug classes included in the FAST pipeline must demonstrate: (1) extension of healthspan and hallmark modulation in animal models, and (2) protection against all-cause mortality in human epidemiological or clinical studies.
- Target Drug Classes in the FAST Pipeline: The FAST initiative evaluates bio-specimens from completed human RCTs of five repurposed drug classes: Metformin, Rapamycin (mTOR inhibitors), GLP-1 receptor agonists, SGLT2 inhibitors, and Bisphosphonates.
- Multi-Omic Integration for Surrogate Endpoints: FAST generates multi-omic profiles (proteomics, metabolomics, DNA methylation epigenetics) across trial samples to identify response molecules that alter in parallel across distinct therapeutic mechanisms.
- Predicting FDA-Approved Clinical Endpoints: Response molecules discovered in FAST must predict hard clinical outcomes (disease incidence, functional decline, all-cause mortality) and mediate the drug’s therapeutic benefit to serve as surrogate endpoints.
- Point-of-Care Diagnostic Testing Development: In collaboration with Stanford University (Mike Snyder’s laboratory), FAST data is being compiled to engineer low-cost, point-of-care multi-omic test kits for clinical practice and decentralized trial readouts.
- Predictive vs. Responsive Biomarker Divergence: Biomarkers optimized for patient risk stratification (predictive) are functionally distinct from those optimized for evaluating short-term therapeutic efficacy (responsive); clinical trial design requires prioritizing responsiveness.
- Overcoming Regulatory Barriers to Preventative Geroscience: Regulators currently permit medical interventions only after clinical disease diagnosis; establishing validated surrogate biomarkers for biological aging rate is essential to enable early, disease-preventative clinical trials.
- Future EHR Integration: Within 10 years, validated biomarkers of biological age and pace of aging are projected to be embedded directly into Electronic Health Records (EHR) to guide preventative clinical decision-making.
- Pharmaceutical Industry Adoption: Major pharmaceutical companies are increasingly profiling trial bio-specimens with high-dimensional multi-omics to validate longevity surrogate endpoints and repurpose existing assets for healthspan extension.
- Rate-Metric Expansion Across Life Course: Advanced longitudinal rate metrics are being applied to pediatric and adolescent cohorts to determine how early-life social determinants and biological stress accelerate lifetime aging trajectories.