Can Calorie Restriction Move Biological Aging Metrics?

Calorie restriction moved some aging measures. It did not validate a universal aging test.

Two years of sustained calorie restriction changed several biological-aging measures in healthy adults, including blood-chemistry composites and one DNA-methylation pace measure. That is useful evidence. It does not mean a blood test can now certify that someone is aging more slowly, much less predict extra years of life.

The distinction matters because biological-age testing has become a consumer product category before it has become a settled clinical science. The strongest human evidence here comes from CALERIE, a randomized trial in which calorie restriction altered some aging-related signals while leaving others unchanged. The results offer a practical standard: a metric earns more confidence when it responds to a randomized intervention over time, but response alone does not make it a validated surrogate for longevity.

The most informative test: CALERIE

CALERIE Phase 2 randomized 220 healthy, nonobese adults in a 2:1 ratio to a 25% calorie-restriction program or an ad libitum control diet for two years. Participants did not reach the prescribed restriction: average achieved restriction was about 11.7% in a biobank analysis and about 11.9% in later reporting. So this trial tested moderate, sustained restriction in younger-to-middle-aged healthy adults, not a full 25% reduction and not an intervention for people with obesity, frailty, chronic illness, or advanced age.

In a secondary analysis of the randomized trial, Belsky et al. (2017) applied blood chemistry and organ-function data to two composite measures: Klemera-Doubal biological age and homeostatic dysregulation. Intent-to-treat mixed-effects models found that calorie restriction slowed the increase in both measures relative to the control group over 24 months. The authors also reported that weight loss did not explain the effect.

That finding answers a narrow but worthwhile question: can a long-term human intervention shift certain aging-related biomarkers in the expected direction? In this trial, yes. It does not answer whether the participants aged more slowly in the sense that matters most—delayed disease, preserved function over decades, or longer life. CALERIE did not measure morbidity, mortality, or lifespan as its endpoint.

The composites also create a built-in interpretive problem. Klemera-Doubal biological age and homeostatic dysregulation draw on physiological biomarkers. If calorie restriction improves those component biomarkers, the composite can improve as a mathematical consequence. That may still be meaningful, but it cannot by itself prove a deeper change in the biology that causes aging.

DNA methylation gave a more selective answer

A later CALERIE DNA-methylation analysis by Waziry et al. (2023) examined blood samples from 197 participants who had baseline data and at least one follow-up measure. Calorie restriction slowed DunedinPACE by about 2% to 3% over two years. The simplified main reporting found no significant change in two DNA-methylation biological-age clocks.

This split is more informative than a blanket claim that "epigenetic clocks work." DunedinPACE aims to estimate the pace of aging, whereas biological-age clocks estimate age-related methylation patterns. Those are related ideas, not interchangeable endpoints. Under the same intervention, in the same broad trial population and time frame, one type of methylation measure responded and two others did not.

The methylation results also come from fewer people than the 220 originally randomized participants. Missing methylation measurements do not erase the trial's randomized design, but they reduce precision and make subgroup conclusions less secure. The study supports intervention responsiveness for DunedinPACE under this specific two-year calorie-restriction program; it does not establish that every methylation clock should move under the same conditions.

A separate CALERIE epigenome-wide analysis by Ramaker et al. (2023) found that long-term calorie restriction changed methylation at specific CpG sites. This shows that the intervention reached the blood methylome. It does not show that a summary clock score measures slowed aging, because a collection of locus-level methylation changes can reflect many biological responses.

Telomeres complicate the story further

Telomere length did not supply a clean, uniformly favorable genomic-aging signal. In the 2024 CALERIE telomere analysis by Hastings et al., the authors described the wider genomic-aging evidence as mixed. That does not disprove telomeres as a biomarker in every setting. It does mean that telomere length did not behave as a dependable confirmation signal for this two-year intervention.

Put the CALERIE findings together and the pattern is plain: calorie restriction changed blood-chemistry composites, slowed one methylation pace measure, altered specific methylation sites, and produced mixed telomere evidence. Biology rarely signs up for a single dashboard.

Why existing clocks may disagree

The disagreement is not merely an inconvenience for marketers. It may reflect how clocks are built. In a 2024 Nature Aging paper, Ying et al. used epigenome-wide Mendelian randomization across 420,509 CpG sites with meQTL data from 36 cohorts totaling 27,750 European participants. They tested links with eight aging-related traits and concluded that existing epigenetic clocks were not enriched for putatively causal CpGs.

The authors then developed DamAge and AdaptAge, intended to distinguish methylation changes associated with damage from those associated with adaptation. The premise is sensible: an intervention can cause methylation changes that reflect a healthy adjustment rather than a direct reduction in damage. A conventional clock that blends both signals could move weakly, move in an ambiguous direction, or miss a beneficial response.

That paper improves interpretation, but it does not demonstrate that its new measures track long-duration human anti-aging interventions better than current clocks. Its intervention-sensitivity work used small, short-term datasets involving plasma fraction treatment, cigarette-smoke-condensate exposure, and omega-3 supplementation. Those experiments cannot substitute for a multi-year randomized human trial.

Genetics can prioritize targets, not prove treatments

Rosoff et al. (2023) approached aging from another direction: shared genetics. Their multivariate analysis created a genetic aging factor with reported fit statistics of CFI 0.97 and SRMR 0.069, used an effective sample size of 1,958,774, and identified 52 lead SNPs in 38 loci, including 20 novel loci.

Those findings can help researchers rank pathways and possible drug targets. They do not show that changing any target slows aging in people. Mendelian-randomization-based target prioritization generates hypotheses; randomized intervention trials test them.

What to do with an aging test now

If you use biological-aging measurements, treat them as research-style tracking tools rather than personal longevity certificates. The best-supported hierarchy is straightforward:

• Give the most weight to metrics that have changed under a randomized human intervention, such as the CALERIE blood-chemistry composites and DunedinPACE.
• Interpret composite biomarkers cautiously, because direct changes in their component measurements can drive the score.
• Do not assume that a nonresponsive clock is useless everywhere, but do not call it intervention-responsive for CALERIE.
• Use causal-clock redesign and genetic studies to guide future experiments, not to declare that an intervention works.

CALERIE shows that moderate calorie restriction can move selected aging-related biomarkers over two years in healthy, nonobese adults. It does not show that any current metric is a validated proxy for slower aging, lower disease risk, or longer life.

That is a more limited conclusion than the usual anti-aging sales pitch. It is also the conclusion the human trial can carry.