Biological Age Clocks are Maturing, But Can They Actually Tell Us Who Will Stay Strong?

For a decade, “biological age” has promised to be the stethoscope of geroscience—a quick, quantifiable read on how fast we’re wearing down. In practice, most clocks were trained to predict birthdays or death, not day-to-day function. That’s a problem: what people care about first is how well they move, think, and live. A new analysis from the large INSPIRE-T cohort in France pushes the field toward that goal by asking a simple, clinically relevant question: which clocks line up best with real-world physical capacity across adulthood (20–104 years)? The short answer: GrimAge comes out on top, with iAge (an inflammatory clock) lighting up in late life—exactly where inflammation tends to dominate the aging phenotype.

The state of play: what do modern aging clocks actually measure?

Epigenetic clocks (DNA methylation). Early “first-generation” clocks (Horvath, Hannum) were trained to match chronological age. They’re excellent agometers, but not necessarily sensitive to healthspan. Second-generation clocks like PhenoAge and GrimAge pivoted toward risk: GrimAge was trained on DNAm surrogates for plasma proteins and smoking pack-years to predict morbidity and mortality, and repeatedly tops leaderboards for survival and disease prediction. A 2025 meta-analysis again found GrimAge and its update GrimAge2 best-in-class for mortality risk.

Inflammation clocks. The immune system sets the tenor of late-life decline. iAge—a deep-learning inflammatory clock built from the “immunome” of 1,001 people aged 8–96—tracks multimorbidity, immunosenescence, frailty, cardiovascular aging, and even exceptional longevity. It was not trained on birthdays, but on inflammatory patterning across the lifespan.

Beyond methylation and cytokines. Newer modalities (proteomics, multi-omics) are arriving. A 2024 proteomic aging clock in Nature Medicine predicted mortality and common diseases across diverse populations—evidence that multiplex protein signatures may rival or complement methylation in the clinic. Meanwhile, DunedinPACE measures the speed of aging (not accumulated age) and links tightly to strength, gait, balance, cognition, and future morbidity—useful for trials that need a responsive endpoint. Recent work even extends “pace” models to imaging-derived phenotypes.

Why function matters. Lower-extremity function tests like the Short Physical Performance Battery (SPPB) predict disability, hospitalization, and mortality across settings. If a clock can anticipate SPPB decline, VO₂max erosion, or chair-stand performance, it starts to look like a healthspan tool, not just a lifespan statistic.

What the INSPIRE-T study asked—and how it answered it

Design. Cross-sectional analysis of 1,014 community-dwelling adults (20–104 years; ~63% women) from the INSPIRE-T cohort. The team computed biological age acceleration (BAA) from four DNAm clocks (Horvath, Hannum, PhenoAge, GrimAge) and one inflammatory clock (iAge), then modeled associations with five functional endpoints:

• Five-time sit-to-stand (5-STS)
• SPPB (gait speed, balance, chair rise)
• 30-second chair-stand (30-s CST)
• VO₂max (cardiorespiratory fitness; n≈245)
• Isokinetic lower-limb strength (n≈239)

Models adjusted for age (including non-linear terms), sex, BMI, and comorbidity; the authors probed whether effects varied by age or sex.

Headline findings.

• GrimAge BAA consistently tracked worse function: slower 5-STS, lower SPPB scores, and notably lower VO₂max across adulthood. In early adulthood, higher GrimAge BAA also associated with poorer 30-s CST.
• iAge BAA “switches on” in late life: at older ages, higher iAge BAA associated with worse 5-STS and SPPB—squarely consistent with inflammaging theory.
• First-generation clocks were weaker: Horvath showed a modest link to worse 30-s CST; Hannum/PhenoAge had inconsistent, age-restricted effects. Some counterintuitive late-life associations likely reflect survival or ceiling effects.
• Sex interactions exist (e.g., GrimAge BAA predicted worse SPPB in men >74 and women 58–86), underscoring the need for sex-stratified analyses in trials.

Interpretation. Among clocks in wide use, GrimAge looks most aligned with multi-domain functional capacity, while iAge captures the late-life inflammatory burden where mobility and resilience falter. Together they provide a plausible “two-signal” readout across the adult lifespan: risk-weighted molecular aging (GrimAge) + immune burden (iAge).

Why these results make biological sense

• Risk-tuned clocks should mirror function. GrimAge was trained on DNAm surrogates of risk proteins and smoking exposure to predict disease and death; it’s unsurprising it correlates with VO₂max and lower-extremity performance, which themselves track survival.
• Inflammation’s late-life footprint. iAge’s strongest associations appearing in older adults fits decades of data on inflammaging—rising cytokine tone drives sarcopenia, frailty, and mobility loss. INSPIRE-T detects exactly that pattern.
• Pace vs. age. Clocks that read speed (e.g., DunedinPACE) often show robust ties to functional decline and could pair well with GrimAge/iAge in interventional trials where you need a sensitive, short-horizon endpoint.

How this fits with broader evidence

• Longitudinal function: In older Finnish women, GrimAge BAA predicted 3-year declines in Timed Up-and-Go, 10-m gait speed, knee extension strength, and 6-minute walk—while other clocks were largely null. INSPIRE-T extends this beyond older women to the full adult lifespan and adds VO₂max.
• Modality diversification: Proteomic clocks and multi-omic composites are catching up and sometimes outperform methylation for certain endpoints—this argues for pan-omic panels in future trials rather than one clock to rule them all.
• Intervention sensitivity: Exercise and higher physical activity levels are associated with slower epigenetic aging and, in some trials, small but measurable “age reductions” on methylation clocks—supporting the idea that at least some clocks are modifiable and function-linked.

What this means for clinics and trials—practical takeaways

1. If your endpoint is mobility or fitness, start with GrimAge (plus a pace-of-aging metric if available). It has the most consistent ties to performance in cross-section and over time, and it’s sensitive to risk biology that overlaps with functional decline.
2. Layer in iAge for older cohorts. When the biology shifts toward chronic, low-grade inflammation, iAge provides complementary signal that GrimAge may miss.
3. Always anchor clocks to hard functional tests. Keep collecting SPPB, gait speed, chair-stands, and ideally VO₂max. They’re prognostic on their own and validate whether your biomarker is “seeing” meaningful change.
4. Consider multi-omic composites. A practical stack for many studies: GrimAge + DunedinPACE + iAge, optionally adding a proteomic clock. That covers risk, rate, immune tone, and systems biology.
5. Expect heterogeneity by sex and age. Pre-specify stratified analyses and non-linear terms; the INSPIRE-T moderation results show signals can flip across the lifespan.

Limits to keep in mind

INSPIRE-T is cross-sectional—it nails correlation, not causation. Blood methylation/immune measures may not reflect tissue-specific aging (muscle, brain). And SPPB has ceiling effects in the young, which can obscure true associations. These are fixable with longitudinal follow-up, multi-tissue sampling, and age-appropriate performance tests (e.g., maximal power output in youth).

Where the field is going next

• Trial-readiness. Pace-of-aging measures (e.g., DunedinPACE) and risk-weighted clocks (GrimAge) are increasingly used as responsive endpoints in geroprotective trials—exactly the niche INSPIRE-T helps justify by tying clock readouts to functional capacity across ages.
• Multi-omic fusion. Proteomic clocks and inflammatory panels will likely be fused with methylation to improve sensitivity and mechanistic interpretability.
• Lifestyle and therapeutics. Exercise continues to show the clearest, repeatable effects on methylation age; observational and interventional data suggest training programs can nudge clocks in the “younger” direction alongside fitness gains. The convergence of molecular clocks with VO₂max (a gold-standard healthspan metric) is especially encouraging.

Bottom line

The INSPIRE-T study strengthens a pragmatic view: if you want a single methylation clock that best reflects whole-body functional capacity from young to old, pick GrimAge; if you’re working in late life or inflammatory conditions, add iAge. Then ground both in validated physical tests and, when possible, a pace-of-aging measure to see short-term change. That’s how biological age escapes the abstract and starts guiding the prevention of disability we can actually feel.

References

1. Argentieri, M. A., Xiao, S., Bennett, D. A., … van Duijn, C. M. (2024). Proteomic aging clock predicts mortality and risk of common age-related diseases in diverse populations. Nature Medicine, 30(9), 2450–2460. https://doi.org/10.1038/s41591-024-03164-7
2. Belsky, D. W., Caspi, A., Arseneault, L., … Moffitt, T. E. (2022). Quantification of the pace of biological aging in humans through a blood test, the DunedinPACE. eLife, 11, e70202. https://doi.org/10.7554/eLife.70202
3. Föhr, T., et al. (2022). The association between epigenetic clocks and physical functioning in older women: A three-year follow-up. The Journals of Gerontology: Series A, 77(8), 1569–1579. https://doi.org/10.1093/gerona/glab270
4. Guralnik, J. M., Simonsick, E. M., Ferrucci, L., Glynn, R. J., Berkman, L. F., Blazer, D. G., Scherr, P. A., & Wallace, R. B. (1994). A short physical performance battery assessing lower extremity function: Association with self-reported disability and prediction of mortality and nursing home admission. Journal of Gerontology, 49(2), M85–M94. https://doi.org/10.1093/geronj/49.2.M85
5. Lu, A. T., Quach, A., Wilson, J. G., Reiner, A. P., Aviv, A., Raj, K., Hou, L., Baccarelli, A. A., Li, Y., Stewart, J. D., Whitsel, E. A., Assimes, T. L., Ferrucci, L., & Horvath, S. (2019). DNA methylation GrimAge strongly predicts lifespan and healthspan. Aging (Albany NY), 11(2), 303–327. https://doi.org/10.18632/aging.101684
6. Lu, A. T., Binder, A. M., Zhang, J., Yan, Q., Reiner, A. P., Cox, S. R., Corley, J., Harris, S. E., Kuo, P.-L., Moore, A. Z., … Horvath, S. (2022). DNA methylation GrimAge version 2. Aging (Albany NY), 14(23), 9484–9549. https://doi.org/10.18632/aging.204434
7. Nagata, M., et al. (2024). Influence of physical activity on the epigenetic clock. Clinical Epigenetics, 16, 112. https://doi.org/10.1186/s13148-024-01756-1
8. Pavasini, R., Guralnik, J., Brown, J. C., di Bari, M., Cesari, M., Landi, F., Vaes, B., Legrand, D., Verghese, J., Wang, C., Stenholm, S., Ferrucci, L., Lai, J. C., Arnau Bartes, A., Espaulella, J., Ferrer, M., Lim, J.-Y., Ensrud, K. E., Cawthon, P., … Campo, G. (2016). Short Physical Performance Battery and all-cause mortality: Systematic review and meta-analysis. BMC Medicine, 14, 215. https://doi.org/10.1186/s12916-016-0763-7
9. Sayed, N., Huang, Y., Nguyen, K., … Wyss-Coray, T. (2021). An inflammatory aging clock (iAge) based on deep learning tracks multimorbidity, immunosenescence, frailty and cardiovascular aging. Nature Aging, 1, 598–615. https://doi.org/10.1038/s43587-021-00082-y
10. Sánchez-Sánchez, J. L., Vellas, B., Guyonnet, S., Bensadoun, P., Lemaitre, J.-M., Fuentealba Valenzuela, M., Pillard, F., Rolland, Y., Furman, D., & de Souto Barreto, P. (2025). Biological ageing acceleration and functional capacities across the lifespan in the INSPIRE-T cohort. Journal of Cachexia, Sarcopenia and Muscle, 16(4), e70046. https://doi.org/10.1002/jcsm.70046