Epigenetic Clocks Track Cardiometabolic Health in Obesity but Don't Budge With Dieting
A 12-month RCT finds DunedinPACE mirrors cardiometabolic risk at baseline in obese adults, but epigenetic clocks fail to shift meaningfully with weight-loss diets.
Summary
The MACRO trial tested whether three epigenetic aging clocks — DunedinPACE, PCPhenoAge acceleration, and PCGrimAge acceleration — responded to 12 months of low-carbohydrate or low-fat dieting in 144 obese adults. At baseline, DunedinPACE was meaningfully linked to insulin, HOMA-IR, cholesterol, HDL, CRP, adiponectin, and ghrelin. After the intervention, most of those associations disappeared, with only CRP and adiponectin remaining significant. Changes in epigenetic clocks did not track changes in biomarkers over time, and mediation analyses found no evidence that epigenetic aging measures explained the cardiometabolic benefits of weight loss. The findings suggest DunedinPACE is a useful snapshot of cardiometabolic burden but may not be a reliable short-term intervention target.
Detailed Summary
Epigenetic clocks have gained enormous traction as potential biomarkers of biological aging, with some researchers treating them as causal targets in anti-aging interventions. The MACRO trial offered a rare opportunity to test this assumption rigorously: 148 obese adults (BMI 30–45 kg/m²) aged 22–75 were randomized to either a low-carbohydrate diet (<40 g digestible carbs/day) or a low-fat diet (<30% energy from fat, <7% from saturated fat) for 12 months, with no specific calorie targets imposed. DNA methylation was profiled on 392 samples using Illumina EPIC 850K arrays at baseline, 3 months, and 12 months, and three clocks — PCPhenoAge acceleration, PCGrimAge acceleration, and DunedinPACE — were computed. A comprehensive panel of 12 cardiometabolic biomarkers was measured at each timepoint.
At pre-intervention baseline, DunedinPACE showed robust associations with multiple cardiometabolic markers after false discovery rate (FDR) correction (FDR < 0.05): insulin, HOMA-IR, total cholesterol, HDL-C, CRP, adiponectin, and ghrelin were all significantly linked. PCGrimAge acceleration showed a significant baseline association only with adiponectin. PCPhenoAge acceleration showed no significant baseline associations after FDR correction. This hierarchy suggests that DunedinPACE, which models the rate of change in 19 biomarkers over two decades, is the most sensitive of the three clocks to concurrent cardiometabolic burden in this obese population.
Following 12 months of dietary intervention — during which both groups lost weight and improved cardiometabolic profiles — the picture changed substantially. Cross-sectional associations at 12-month follow-up were markedly attenuated: only CRP (positively associated with DunedinPACE) and adiponectin (inversely associated with DunedinPACE and PCGrimAge acceleration) retained significance. This attenuation was notable because the biomarkers themselves improved meaningfully; the epigenetic clocks simply did not track those improvements in the same way at post-intervention timepoints.
Longitudinal change-to-change analyses — examining whether individuals whose clocks decreased the most also showed the greatest biomarker improvements — found no significant associations for any of the three clocks with any of the 12 biomarkers at either 3 or 12 months (all FDR > 0.05). Formal mediation analyses tested whether changes in epigenetic clocks explained the effect of weight loss on cardiometabolic outcomes; none of the three clocks mediated any of the biomarker changes, with all mediation proportion estimates statistically indistinguishable from zero.
The authors interpret these findings cautiously but provocatively. DunedinPACE appears to be a valid cross-sectional marker of cardiometabolic aging burden in obese individuals, but its failure to respond dynamically to intervention-induced improvements suggests it may not be causally upstream of those changes. One interpretation is that 12 months is too short a window for DNA methylation patterns to reorganize in response to lifestyle change. Another is that epigenetic clocks are statistical proxies for accumulated physiological burden rather than causal drivers — meaning improving biomarkers does not necessarily rewrite the epigenetic signature of prior metabolic stress. These findings are a meaningful caution for clinical trials that use epigenetic clocks as primary outcomes and assume changes in them signal genuine biological rejuvenation.
Key Findings
- DunedinPACE was significantly associated with 7 cardiometabolic biomarkers at baseline (FDR < 0.05): insulin, HOMA-IR, total cholesterol, HDL-C, CRP, adiponectin, and ghrelin
- After 12 months of dietary intervention, DunedinPACE associations were largely attenuated — only CRP and adiponectin remained significant
- PCGrimAge acceleration showed only a single significant baseline association (adiponectin), and PCPhenoAge acceleration showed none after FDR correction
- Change-to-change analyses found no significant associations between changes in any epigenetic clock and changes in any of the 12 cardiometabolic biomarkers at 3 or 12 months
- Mediation analyses found no evidence that any epigenetic clock mediated the cardiometabolic benefits of weight-loss dietary intervention (all mediation proportions non-significant)
- 144 of 148 randomized participants had usable DNA methylation data at baseline; 112 completed 12-month longitudinal follow-up
- Both diet groups (low-carbohydrate and low-fat) were analyzed together in primary analyses, with dietary arm included as a covariate
Methodology
The MACRO trial (NCT00609271) was a 12-month parallel-arm RCT at Tulane University (2008–2011) randomizing 148 obese adults to low-carbohydrate (<40 g/day) or low-fat (<30% energy) diets. DNA methylation was measured on Illumina EPIC 850K arrays at baseline, 3 months, and 12 months (392 samples), with samples randomized across plates to minimize batch effects; cell-type composition was estimated via the Houseman method and regressed out. Three epigenetic clocks were computed using the dnaMethyAge R package, and associations with 12 cardiometabolic biomarkers were assessed via linear regression (cross-sectional), linear mixed-effects models (longitudinal), and formal causal mediation analysis, all with FDR correction for multiple comparisons.
Study Limitations
The study enrolled only 144–148 participants, limiting statistical power for mediation analyses and subgroup comparisons between dietary arms. The 12-month intervention window may be insufficient to detect meaningful epigenetic reprogramming, and longer-term studies are needed to rule out delayed clock responses. The trial excluded individuals with type 2 diabetes or established cardiovascular disease, limiting generalizability to higher-risk obese populations; no conflicts of interest were declared by the authors.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.
Enter your email to subscribe: