REF-1188 · Rottensteiner et al., 2015 · Medicine & Science in Sports & Exercise · Paper · PubMed

Does activity itself make the difference if genes are equal?

In 10 identical male twin pairs aged 32-36, the twin more active for 3 years was leaner and more insulin-sensitive. Cross-sectional, abstract only: not proof of cause.

Blimited evidenceMeasured on non-athletes and recreational people

Rottensteiner and colleagues (University of Jyväskylä and University of Helsinki, the FITFATTWIN study) asked whether physical activity is linked to body composition, glucose control and brain structure when genes are held constant.

From a population-based Finnish twin cohort they systematically selected 10 young adult male monozygotic twin pairs aged 32-36 who differed in leisure-time physical activity over the past 3 years, quantified from interviews as a leisure and commuting activity index in MET-hours a day.

They measured body fat by DXA, glucose homeostasis by HOMA and by the Matsuda index from an oral glucose tolerance test, and regional brain volumes by whole-brain MRI.

In pairwise analysis the active twins had lower body-fat percentage (P = 0.029), lower HOMA (P = 0.031) and higher Matsuda index (P = 0.021) than their inactive co-twins, and larger grey-matter volumes in the striatum and prefrontal cortex of the non-dominant hemisphere at a statistical threshold of P < 0.001.

The authors conclude that more activity is associated with better glucose homeostasis and brain grey-matter differences "independent of genetic background".

Our reading: the co-twin design does remove genes and the shared upbringing, which strengthens the association, but with 10 pairs, a 3-year discordance and no intervention it cannot show that the activity caused the differences rather than something else that made one twin more active.

Protocol

Participants (from the abstract): 10 male monozygotic twin pairs aged 32-36, systematically selected from a population-based twin cohort for discordance in leisure-time physical activity during the past 3 years.
Exposure: a mean sum index of leisure-time and commuting activity over the past 3 years from interviews (the "3-yr LTMET index", MET-hours per day). The size of the within-pair difference is not in the abstract.
Outcomes: body composition including DXA fat percentage; glucose homeostasis by homeostatic model assessment (HOMA) and the Matsuda insulin-sensitivity index from an oral glucose tolerance test; whole-brain MRI with regional volumetric analysis. The title names fitness; no fitness results appear in the abstract.
Analysis: pairwise (within-pair) comparisons; brain volumes at a threshold of P < 0.001 (the abstract does not say whether this was corrected for multiple comparisons across brain regions).
Funding and conflicts: not stated in the abstract; the PubMed / Europe PMC record lists NIH (NIAAA) grants AA-09203, AA-12502 and AA-00145. Conflicts of interest not checked (full text not read).

Key Results

Outcome (active vs inactive co-twin, from the abstract)DirectionP
Body-fat percentage (DXA)lower in the active twin0.029
HOMA indexlower in the active twin0.031
Matsuda insulin-sensitivity indexhigher in the active twin0.021
Grey-matter volume, striatum and prefrontal cortex (subgyral, inferior frontal gyrus), non-dominant hemispherelarger in the active twinthreshold P < 0.001
n: 10 pairs (20 men). Means, differences and confidence intervals are not in the abstract.

Mechanisms

Measured: within genetically identical pairs, the more active twin had less body fat and better insulin sensitivity (lower HOMA, higher Matsuda index).
Authors' interpretation: the associations are "independent of genetic background" and may contribute to a lower later risk of type 2 diabetes and of mobility limitations. The future-risk part is not measured here.
Inference: the lower fat percentage could account for part of the better insulin sensitivity; whether activity improves glucose control beyond its effect on fatness cannot be told from the abstract.
Inference: the co-twin design controls genes and shared upbringing, not the non-shared reasons one twin became more active (job, injury, illness, diet, motivation); reverse causation (the leaner, healthier twin finds activity easier) is also open in a cross-sectional comparison.

Practical Application

Does activity itself make the difference if genes are equal? In 10 identical male twin pairs aged 32-36, the twin more active for 3 years was leaner and more insulin-sensitive. Cross-sectional, abstract only: not proof of cause.
For a health question from an athlete or a parent: being more active goes with lower fat and better glucose control even between identical twins (B, from the abstract); genes do not account for the whole difference. Training trials, not twin snapshots, show what training changes.
No training dose can be taken from this: the activity measure was an interview-based index of leisure and commuting MET-hours over 3 years, and the abstract does not give the size of the gap between twins.
Brain findings: do not quote as "exercise grows the brain" — larger regional grey-matter volumes in 10 pairs at P < 0.001, without a stated correction across regions in the abstract, is a lead, not an effect size.

Limitations

Read from the abstract only; the publisher page and the Finnish repository copy could not be opened, so means, effect sizes, fitness results and the size of the activity gap are not checked.
Ten pairs; men only, aged 32-36; healthy.
Cross-sectional: discordance in activity over the past 3 years from interviews, not lifelong and not measured objectively in the abstract's description; no intervention.
Co-twin control removes genes and shared upbringing but not non-shared confounders or reverse causation.
The brain threshold (P < 0.001) is given without a stated correction for multiple comparisons in the abstract.
Funding and conflicts of interest not stated in the abstract.
How this entry was checked and changed
Why this level

cross-sectional comparison of 10 pairs (n = 20), from the abstract (the publisher and repository copies could not be opened). The co-twin design removes genetic and shared-family confounding, which a usual cross-sectional study cannot, but 10 pairs, a 3-year interview-based activity history and no intervention keep it at B; the abstract gives P values without effect sizes