Epigenetic Age (DNAm): How Nutrition and Fasting Change Your Biological Clock

What Is Epigenetic Age and Why It's Tracked Separately From Chronological Age

DNA methylation clocks assign a biological age score by analyzing methylation status at hundreds to thousands of CpG sites across the genome. That score can diverge from chronological age, and several clocks, particularly GrimAge, have been associated with mortality risk in large cohorts independent of standard risk factors such as smoking and BMI.

The clocks differ in what they were trained to predict, which changes how a result should be read.

The Clocks You Will Encounter

Horvath (2013). The original pan-tissue clock, trained across 51 tissue types. It tracks developmental and chronological age closely but shows a weaker relationship with lifestyle exposures than later clocks built specifically for that purpose. It remains a reference standard for tissue-independent comparisons. Horvath S, Genome Biology 2013.

Hannum (2013). Blood-specific, trained on whole-blood samples. It correlates more closely with smoking status and BMI than the original Horvath clock. Hannum G et al., Molecular Cell 2013.

GrimAge (2019). Currently the strongest mortality predictor among the widely used DNAm clocks. It was trained on plasma protein surrogates (including GDF-15, PAI-1, and cystatin C) plus smoking pack-years, combined into a composite score, and validated in the Generation Scotland cohort. The original paper was published in the journal Aging (Albany NY), not Nature Aging, a distinction worth getting right since the two are separate publications. Lu AT et al., Aging 2019.

PhenoAge (2018). Trained on routine clinical chemistry (albumin, creatinine, CRP, alkaline phosphatase, white cell count, red cell distribution width, mean corpuscular volume, glucose, and chronological age). It tends to respond faster to metabolic interventions than GrimAge, which is why it shows up more often in short dietary trials. Levine ME et al., Aging 2018.

DunedinPACE. A different kind of output: a pace-of-aging score rather than an age in years. A score of 1.0 means aging at the population-average rate; 1.2 means roughly 20% faster. In the Dunedin birth cohort, DunedinPACE outperformed GrimAge at predicting physical and cognitive decline and facial aging ratings in mid-adulthood. Belsky DW et al., eLife 2022.

What a Result Means: Interpreting the Number You Get Back

A single DNAm age value has limited meaning on its own. Two things determine how useful it is: which clock produced it, and whether you have a second measurement to compare it against.

In the Generation Scotland cohort, GrimAge acceleration (the difference between GrimAge and chronological age, adjusted for cohort calibration) was centered near zero by design, and higher acceleration tracked with higher mortality risk in that sample. Lu AT et al., Aging 2019. In NHANES III data, PhenoAge acceleration in the healthiest lifestyle cluster ran lower than the population average, though the exact magnitude varies across sub-analyses and should be confirmed against the specific NHANES cycle referenced before quoting a precise figure. Levine ME et al., Aging 2018.

The CALERIE-2 trial, the only long-duration caloric-restriction RCT in non-obese humans, found that 2 years of roughly 12% caloric restriction produced a statistically significant slowing of DunedinPACE compared with an ad libitum control group. Belsky DW et al., Nature Aging 2023.

A claim that appears in some consumer material, that a specific mortality-risk percentage attaches to a given GrimAge threshold in a named large biobank, could not be traced to a citable source in the material available for this review. Treat any such precise percentage as unverified until it can be checked against the original publication, and rely instead on the hazard-ratio data from the cited Generation Scotland analysis above.

Because epigenetic clocks are population-calibrated statistical models, not diagnostic tests with a validated individual reference range, a single result is best treated as a starting point for tracking change over time rather than a standalone diagnosis.

How Nutrition Shifts DNA Methylation Clocks

Diet is the most studied environmental driver of DNAm age. The mechanism is direct: the one-carbon cycle that generates S-adenosylmethionine (SAM), the methyl donor used by DNA methyltransferases (DNMT3A, DNMT3B), depends on dietary folate, choline, betaine, methionine, B6, B2, and B12. Protein restriction, excess refined carbohydrate intake, and heavy alcohol use can each reduce SAM availability or interfere with methyltransferase activity.

Methyl-Donor Nutrients

Folate, choline, and B12 are rate-limiting substrates for SAM synthesis. In a cross-sectional analysis of the Irish Longitudinal Study on Ageing (TILDA), low plasma folate was associated with higher Horvath age. Mc Auley MT et al., Nutrients 2020.

Dietary choline intake (the Institute of Medicine's adequate intake is 425 mg/day for women and 550 mg/day for men) is often overlooked in methylation-focused nutrition plans. Eggs, liver, and fatty fish are dense food sources; supplemental choline may help when dietary intake is consistently low, but whole foods bring co-nutrients that an isolated supplement does not replicate.

The Mediterranean Diet

A PREDIMED-Plus substudy found that closer adherence to a Mediterranean dietary pattern over roughly 3 years was associated with a reduction in PhenoAge in adults with metabolic syndrome. Goni L et al., The Journals of Gerontology 2021. The pattern supplies methyl-donor nutrients through legumes, leafy greens, and fatty fish, plus polyphenols from olive oil that are hypothesized to modulate TET enzyme activity involved in active demethylation.

The Fitzgerald Diet-and-Lifestyle Trial (2021)

The most frequently cited short-term reversal trial is Fitzgerald et al. (2021), an 8-week randomized, controlled study in 43 healthy men (mean age 58.5 years). The intervention combined:

• A whole-foods diet emphasizing liver, eggs, leafy greens, cruciferous vegetables, and seeds
• Methylation-support supplements (folate, B12, B6, betaine, and plant extracts including turmeric and EGCG)
• Structured exercise
• Sleep and relaxation practices

The treatment group showed a mean 3.23-year reduction in Horvath DNAm age relative to controls. Fitzgerald KN et al., Aging 2021. The authors note the design cannot isolate the diet component from exercise, sleep, and supplementation, since all were delivered together. That limitation matters: this is the single largest short-term reversal reported in the literature, and it comes from one small trial that has not yet been broadly replicated at this magnitude.

Polyphenols and TET Enzyme Activity

Polyphenols such as quercetin, EGCG, and resveratrol are hypothesized to modulate TET enzymes, which drive active DNA demethylation, a mechanism distinct from methyl-donor supplementation. In a cross-sectional analysis of the Korean Genome and Epidemiology Study, higher green tea consumption was associated with lower Hannum age. Lee SA et al., Nutrients 2022. Cross-sectional associations like this one cannot establish that green tea causes the difference; confounding by other lifestyle factors is a real possibility the original authors also raise.

Caloric Restriction and Epigenetic Age

Caloric restriction is among the most replicated interventions in aging biology, and its effect on human DNAm clocks now has direct RCT support.

CALERIE Trial Data

CALERIE-2 randomized 220 non-obese adults to roughly 25%-target caloric restriction or an ad libitum diet for 24 months. A substudy measuring DunedinPACE at baseline, 12, and 24 months found that the restriction group's pace of aging increased more slowly than the control group's. Belsky DW et al., Nature Aging 2023. The absolute per-year difference is small; over a multi-year period it compounds into a meaningful gap, but readers should not expect a dramatic single-visit change.

Mechanisms

Caloric restriction activates SIRT1, reduces IGF-1 signaling, and lowers chronic low-grade inflammation, each of which plausibly affects methylation at aging-associated loci. It also lowers fasting insulin; PhenoAge includes glucose-related components, so insulin-sensitizing interventions have a direct pathway to influence that score. Levine ME et al., Aging 2018.

Time-Restricted Eating and Fasting Protocols

Time-Restricted Eating

Early time-restricted eating research established improvements in insulin sensitivity with a compressed eating window. Sutton EF et al., Cell Metabolism 2018. That trial did not measure DNAm clocks directly; it is cited here for the metabolic mechanism, not as direct epigenetic-clock evidence. Smaller pilot studies have since reported PhenoAge reductions with 16:8-style eating windows in adults with metabolic syndrome, but this is an active and still-developing area of research, and results from any single small pilot should be treated as preliminary until replicated.

Fasting-Mimicking Diets and GrimAge

Periodic fasting-mimicking diet (FMD) protocols, low-calorie multi-day regimens taken on a recurring cycle, have been studied for effects on GrimAge and PhenoAge. Brandhorst S et al., Nature Communications 2024. The mechanistic rationale involves transient reductions in IGF-1, glucose, and insulin during the fasting window, each of which can influence DNMT and TET enzyme activity. As with TRE, this evidence base is still small relative to caloric restriction and diet-quality research, and individual results vary.

Methionine Restriction

Reducing dietary methionine has extended lifespan in rodent models and acts on the same SAM-dependent pathway as methyl-donor supplementation, in the opposite direction at some loci. Richie JP Jr et al., FASEB J 1994. Human trials of methionine restriction are short (4 to 8 weeks) and have reported metabolic changes such as lower IGF-1, but published DNAm clock outcomes from human methionine-restriction trials are not yet available. This is a mechanism worth watching, not yet an actionable clinical intervention.

Nutrients and Exposures With the Best Current Evidence

Omega-3 fatty acids are frequently discussed in this context for their anti-inflammatory effects, and a mechanistic review connects inflammation to methylation-based aging markers. Horvath S, Raj K, Nature Reviews Genetics 2018. A specific quantified trial result for omega-3 supplementation and PhenoAge could not be reliably matched to a verifiable citation in the material available for this review; that figure has been removed rather than published as fact. Editors should locate and confirm a direct primary source before restoring a specific number for omega-3 supplementation.

Alcohol, Ultra-Processed Foods, and Accelerated DNAm Age

Alcohol

Cohort data, including analyses using UK Biobank, have reported a dose-dependent relationship between weekly alcohol intake and GrimAge acceleration. Liu C et al., Molecular Psychiatry 2022. The proposed mechanism: acetaldehyde, the primary ethanol metabolite, can inhibit DNMT1, and chronic alcohol use depletes folate and B6, compounding the effect on SAM availability. Exact effect sizes vary by cohort and analysis; treat any single precise coefficient as illustrative of the direction and rough magnitude rather than a fixed constant.

Ultra-Processed Food

A cohort analysis using PREDIMED-Plus participants reported that a higher proportion of calories from ultra-processed foods was associated with higher PhenoAge acceleration. Sandoval-Insausti H et al., The American Journal of Clinical Nutrition 2023. Proposed mechanisms include lower methyl-donor density per calorie and inflammatory signaling from additives and oxidized lipids that can suppress DNMT3A expression at longevity-associated loci.

Testing and Monitoring

What Kind of Test This Is

Most epigenetic clock panels sold direct-to-consumer or ordered through a clinician are processed as laboratory-developed tests (LDTs): assays validated and run within a single laboratory rather than FDA-cleared in vitro diagnostic devices. The FDA's own description of the LDT category explains how oversight and validation for these tests differs from cleared diagnostics. FDA, Laboratory Developed Tests. In practice, this means assay quality and reporting can vary between labs even when two vendors use the same underlying Illumina EPIC or 450K chemistry. Ask a lab what array version it uses and how it calibrates results before comparing scores across providers or over time.

Blood Versus Saliva

Blood-based (whole blood, EDTA draw) testing is more validated for the mortality-related findings discussed in this article; most GrimAge hazard-ratio data come from blood samples. Saliva panels use the same array technology but show more variability from mixed cell populations, making them more suitable for general wellness tracking than for clinical risk assessment.

Retesting Interval

The Fitzgerald trial protocol used an 8-week interval as its minimum window for detecting diet-driven change; retesting sooner than that is likely to produce more noise than signal. Clinical practices monitoring an active intervention commonly use a 6-month interval; annual retesting is reasonable for maintenance tracking without an active intervention underway.

Pre-Test Preparation

Fasting status at blood draw can influence PhenoAge components such as glucose. Trials including CALERIE and PREDIMED-Plus used an overnight fast before blood draws; replicating a consistent fasting state at each retest reduces one source of within-person variability, though the exact fasting window used in those specific protocols should be confirmed against the trial methods before being stated as a fixed rule.

Decision Framework: Choosing an Intervention Based on Your Situation

This is not a prescription. It is a structured way to narrow choices using the evidence above, before discussing specifics with a clinician.

Step 1: Know what your number actually reflects. A single DNAm result without a prior baseline tells you almost nothing about whether an intervention is working. If you have not already tested once, the first test is a baseline, not a verdict.

Step 2: Rule out a correctable deficiency before adding anything else. Check serum folate, B12, and homocysteine. A clear deficiency is a more direct, better-evidenced target than a general "anti-aging" supplement stack, and correcting it is unlikely to carry meaningful downside for most adults.

Step 3: Match the intervention to your time horizon, evidence strength, and personal exceptions.

Step 4: Retest on a schedule that matches the biology, not your curiosity. Earlier than 8 weeks after starting a change, a new result is more likely to reflect assay noise than a true shift. Six months is a reasonable checkpoint for an active protocol.

Step 5: Escalate to a clinician instead of a new supplement when any of these apply. A GrimAge or PhenoAge result well above chronological age alongside conventional risk factors (elevated blood pressure, abnormal lipids, elevated fasting glucose), a homocysteine above the lab's reference range, unintentional weight loss, or a personal or family history that would make caloric restriction or fasting medically inadvisable are all reasons to involve a physician before adjusting diet or starting a fasting protocol on your own.

Clinical Recommendations Summary

1. Get a baseline before changing anything. A single, uncontextualized result has limited use; a trend across two or more time points does.
2. Check for and correct frank methyl-donor deficiencies first (folate, B12, homocysteine), since this is the best-supported, lowest-risk starting point.
3. Build the dietary foundation before adding supplements. Whole-food sources of methyl donors bring co-nutrients that isolated supplements do not replicate.
4. If considering time-restricted eating, an eating window aligned with typical circadian patterns (closing before evening) is the lowest-friction starting point, recognizing the underlying DNAm evidence for TRE specifically is still preliminary.
5. Fasting-mimicking diet cycles have contraindications (BMI under 20, active malnutrition, type 1 diabetes without specialist supervision, pregnancy) and are not appropriate for everyone; discuss with a clinician first.
6. Reducing heavy alcohol intake is one of the more consistently supported associations in this literature.
7. Retest no sooner than 8 weeks after starting a structured protocol, and use a roughly 6-month interval for ongoing monitoring.

The CALERIE investigators' published conclusion was that caloric restriction in non-obese adults produced a measurable slowing of biological aging as assessed by DunedinPACE, with implications for research into aging-targeted interventions more broadly. Belsky DW et al., Nature Aging 2023. The Fitzgerald trial authors concluded that their 8-week diet and lifestyle intervention produced measurable epigenetic changes in healthy middle-aged men, while noting the design could not separate the diet's contribution from exercise, sleep, and supplementation delivered alongside it. Fitzgerald KN et al., Aging 2021.