After eight weeks of combined aerobic and strength training, a blood-based epigenetic aging estimate fell by 1.4 years in the women who had the most accelerated readings at the start. The important denominator is eight.

The study enrolled 18 sedentary women aged 50 to 70. The eight classified in a higher baseline age-acceleration group moved from an average of 3.6 years above the age-adjusted prediction to 2.2 years above it. The ten women in the lower group did not change. The overall effect of time was not statistically significant, although the difference in change between the two groups was.

This is one small single-arm study, not settled consensus. With no non-exercising comparison group, it cannot establish that training caused the methylation change. This is reporting on exercise research, not medical or training advice.

The 1.4 years came from a blood-based algorithm

DNA methylation is a chemical tag that helps regulate how DNA is used without rewriting its sequence. Methylation at some sites changes predictably with age. An epigenetic clock combines selected measurements in an algorithm and returns an age-related score.

Researchers extracted DNA from whole blood and measured methylation with Illumina’s Infinium MethylationEPIC array, which covers roughly 850,000 sites. They used the DNA Methylation Age Calculator platform to produce an epigenetic age, then examined an age-acceleration residual.

A residual asks whether the clock’s estimate is higher or lower than expected for someone’s chronological age. It is not a second birth certificate. Moving from 3.6 to 2.2 does not mean that eight women literally became 1.4 years younger, gained 1.4 years of life, or reversed aging throughout the body. Their mean score remained on the accelerated side of the model after training.

The publicly available abstract names the calculator platform but not the precise clock output used for this residual. That detail matters because different clocks were trained for different jobs and do not always agree.

The intervention meant 24 planned one-hour sessions

The women took part in combined aerobic and strength training for 60 minutes, three times a week, for eight weeks. The schedule therefore offered 24 planned sessions. Blood was sampled before and after the program.

The design was a before-and-after intervention, not a randomized trial. Everyone trained. Nobody was assigned to continue without training, and the study included only women who had previously been sedentary. Its answer is therefore narrower than a comparison between exercise and no exercise.

The research team used a two-way analysis of variance followed by Fisher’s least significant difference tests. The work was supported by FAPESP and CAPES in Brazil, and the authors declared no potential conflict of interest. The paper appeared online in July 2023 and in the journal’s 2024 print volume.

The average result across all 18 women was not significant

The researchers separated participants by their baseline age-acceleration residual. Ten women entered the lower group. Their mean was minus 2.3 years before training and remained minus 2.3 afterward, with standard deviations of 3.2 and 3.6 years.

The other eight entered the higher group. Their mean fell from 3.6 years, with a standard deviation of 2.6, to 2.2 years, with a standard deviation of 2.7. That 1.4-year difference is the result in the headline.

The group-by-time interaction had a p value of .005, meaning the two groups changed differently under the model. Yet the main effect of time had a p value of .31, which did not meet the study’s .05 threshold. Put plainly, the analysis did not find a statistically significant before-and-after change shared across all 18 participants.

The subgroup result is not meaningless. It suggests that baseline state may matter. But it does not show that a typical sedentary woman should expect her methylation age to fall after eight weeks, and it certainly does not establish an average 1.4-year change across the full sample.

Regression to the mean is a serious alternative

There is a built-in problem when people are grouped by an extreme first measurement and then the same outcome is measured again. Extreme observations tend to move closer to the average on retesting because of ordinary biological fluctuation and measurement error. Statisticians call this regression to the mean.

That pattern resembles what happened here: the high group moved downward while the low group stayed level. Exercise may have contributed, but the design cannot tell how much. A non-exercising group selected with the same baseline rule could have shown whether the high readings also drifted down without training.

The sample adds uncertainty. A mean based on eight people can move sharply because of one or two individual changes. There was no random allocation, and no control for the attention, schedule, retesting or other behavior changes that can accompany joining an exercise study. A larger trial could also predefine whether the high-acceleration subgroup is the primary analysis rather than discovering its importance after the split.

Blood methylation is not whole-body rejuvenation

Whole blood is a practical tissue because it can be collected repeatedly. It is also a mixture of immune-cell populations, each carrying a different methylation pattern. Exercise can change which cells are circulating, so a blood signal can reflect shifts in cell composition as well as changes within cells. The public abstract does not say whether the clock result was adjusted for changing blood-cell proportions.

A 2026 review of 29 human exercise trials concluded that longer interventions often coincide with extensive methylation remodeling in peripheral blood, but it also emphasized heterogeneous methods and the limited ability of blood to represent tissue-specific changes. Methylation can respond to exercise without every organ becoming younger by the same number.

ScienceBlog previously examined a study in which aortic stiffness in first-time marathoners was translated into four years of “vascular age”. That estimate came from artery measurements, while the present one came from blood-cell methylation. Neither number edits chronological age, and the two should not be treated as interchangeable proof of rejuvenation.

This study did not measure disease, disability, cognition, organ function, survival or lifespan. It cannot tell us whether the 1.4-year biomarker change produced a health benefit.

A suggestive signal now needs a controlled test

The wider evidence is promising but not definitive. A 2026 systematic review of 44 studies found that greater physical activity was associated with lower acceleration on the Horvath and GrimAge clocks, but not significantly on the Hannum or PhenoAge clocks. Only seven cross-sectional studies entered its meta-analysis, so the authors called for longitudinal studies and clinical trials before drawing causal conclusions.

A convincing follow-up to the 18-woman study would randomize a much larger group to training or a credible comparison condition. It would prespecify the clock and subgroup rules, report adherence and exercise intensity, account for blood-cell composition, blind laboratory processing, repeat the assay, and ask whether any change persists after training stops. Fitness, metabolic health and clinical outcomes would show whether a clock shift travels with something people can feel or clinicians can measure.

I think the study is useful precisely because its limitations define the next experiment. Eight women with high baseline age acceleration had a lower model-derived blood score after a short combined program. That is an intriguing biomarker signal, not evidence that eight weeks of exercise reversed whole-body aging.

Exercise has health benefits established through evidence far broader than this experiment. Anyone with symptoms, a known condition, or concerns about starting a program can discuss the appropriate type and intensity with a qualified clinician.