The phrase "metabolic health" can mean many things. In this 2026 Swedish twin study, researchers focused on one specific research tool: the MetaboHealth score, a blood-metabolomics score built from 14 metabolites and originally trained to predict mortality risk.
Using data from the Swedish TwinGene cohort, the researchers asked how much variation in that score could be statistically attributed to genetics, shared family environment and individual-specific environment. The answer was not genetic determinism — and it was not proof that lifestyle can freely rewrite the score either.
Quick answer
Was the MetaboHealth score mostly genetic?
No. In ACE modelling of 4,751 complete Swedish twin pairs, 40% of score variation was attributed to additive genetic factors, 57.5% to unique environmental factors and 2.5% to shared environment. These are population-level variance estimates and do not mean that an individual score is "60% modifiable".
Key takeaways
- The main cohort included 12,256 Swedish TwinGene participants with metabolomics data.
- Heritability analysis used 4,751 complete twin pairs: 1,370 monozygotic and 3,381 dizygotic pairs.
- The estimated additive genetic component was 40% (95% CI 37–43%).
- The estimated unique-environment component was 57.5%; the shared-environment estimate was small and imprecise.
- The score was associated with mortality risk and with factors including smoking, BMI and frailty, but the study did not prove that changing those factors will necessarily change the score or clinical outcomes.
What is the MetaboHealth score?
MetaboHealth is a research biomarker calculated from 14 blood metabolites measured with nuclear magnetic resonance metabolomics. It was originally trained against mortality outcomes and has been associated with several ageing-related phenotypes.
That makes it useful for research, but it is important not to over-translate it. The score is not a medical diagnosis, not a general measure of every aspect of metabolism and not a direct readout of a person’s total biological age.
Why twins are useful for this question
The study included 12,256 participants from the Swedish TwinGene cohort and followed mortality through February 2024. For the heritability model, the researchers had 4,751 complete twin pairs with the necessary data.
Classical ACE modelling separates statistical variance into additive genetic factors, shared environmental factors and unique environmental factors. Because monozygotic twins are genetically much more similar than dizygotic twins, differences in pair similarity can help estimate those components at population level.
The 40% result — and the 57.5% result
The MetaboHealth score’s additive genetic component was estimated at 40%, with a 95% confidence interval of 37–43%. Unique environmental factors accounted for an estimated 57.5% of the variance, while shared environment contributed an estimated 2.5% with a confidence interval that included zero.
The tempting headline would be that the score is therefore "mostly modifiable". That goes too far. Unique-environment variance includes many measured and unmeasured influences, plus measurement error. A variance component does not tell us which intervention would change the score, by how much, or whether a changed score would improve health.
Smoking, BMI and frailty were associated with the score — but causality remains open
The researchers found associations between the MetaboHealth score and smoking, BMI and frailty-related measures. They also examined identical twin pairs who differed in selected exposures or health states to explore whether the score tracked some non-genetic differences.
Those discordant-pair analyses were informative but limited by small subgroup sizes, and some were underpowered. The paper therefore treats intervention responsiveness as a research possibility rather than a demonstrated clinical effect.
TwinPare perspective: heritable does not mean fixed
This study is a strong example of why heritability and modifiability should never be treated as opposites. A trait can show a clear genetic component and still vary substantially between people who share family background or even nearly identical DNA.
For TwinPare Health & Fitness, the relevant principle is individual response over time. Consumer health tracking can help people follow habits and patterns, but TwinPare does not measure MetaboHealth, estimate molecular heritability or convert research biomarkers into personal medical risk.
Source notes
The source has been verified and editorially reviewed for this article. The limitations below show which level of conclusion the sources support.
- [smulders-2026-metabohealth] The MetaboHealth score is 40% heritable and is influenced by frailty status, BMI, and smoking in Swedish twins Larissa Smulders; Jonathan K. L. Mak; Sara Hägg; Joris Deelen. GeroScience, 2026. Evidence type: Swedish TwinGene metabolomics study of 12,256 participants; ACE heritability modelling in 4,751 complete twin pairs Limitation: ACE modelling estimated 40% additive genetic, 57.5% unique-environment and 2.5% shared-environment variance in the score. The score is a mortality-trained metabolomic research biomarker; observational associations do not prove that interventions will change the score or improve outcomes. PubMed DOI
Editorial source review
This section shows how the article's key factual claims are linked to the source.
Phrasings that require caution
- Do not say the MetaboHealth score is 60% modifiable for an individual.
- Do not equate MetaboHealth with a clinical diagnosis or a universal biological-age score.
- Do not claim smoking cessation, weight change or frailty improvement will necessarily change the score by a predictable amount.
- Keep the 40%, 57.5% and 2.5% figures explicitly tied to population-level ACE modelling in the studied cohort.
| ID | Claim | Source support | Caution |
|---|