Longevity genes helped. Not smoking helped far more.

The interesting thing about longevity genetics is not whether it exists. It is how much of it there is compared with the things you can change. A study in GeroScience measured both in the same people, over the same 17 years, and the answer is not close.

The researchers examined whether a genome-wide polygenic lifespan score is associated with all-cause mortality and how this association compares with associations between long-term lifestyle factors and mortality.

Why a twin cohort

The design is what makes the comparison credible. Twin studies exist precisely to separate what is inherited from what is lived, because identical twins share a genome and fraternal twins share half of one.

The score was computed for the older Finnish Twin Cohort, mean age 57.4 years, 45.2% men, with 5,575 people. Finland keeps unusually complete national records, so nobody was lost to follow-up because they moved house.

The follow-up was long enough for the outcome to be unambiguous. Over a mean follow-up of 17.5 years, 1,405 deaths occurred, a quarter of the cohort. Death is the one endpoint nobody misclassifies.

What the genes were worth

The genetic signal was real, and it was small.

A one standard deviation increase in the polygenic lifespan score was associated with a lower all-cause mortality risk - 0.838 on the study’s risk measure, roughly a sixth lower across a substantial genetic gap.

Then the researchers added everything else: sex, physical activity, body mass index, alcohol, smoking and education. The association remained relatively unchanged after adding all of them, at 0.863.

That stability is informative in its own right. It means the genetic effect is not lifestyle wearing a disguise. People with favorable lifespan genetics were not simply people who smoked less. The two run on separate tracks, and the genetic track is the shorter one.

What behavior was worth

The contrast arrives in the same paragraph of the paper.

Smoking 20 or more cigarettes a day showed the strongest association with increased mortality risk, at 3.341. More than tripled, against a sixth lower for the genetics.

Female sex was associated with the greatest risk reduction, at 0.678 - itself a larger effect than the entire polygenic score, and equally outside anyone’s control.

Which variable actually did the sorting

There is a second measure hiding behind those numbers: how much each factor improved the model’s ability to separate who died from who did not.

Smoking behavior had the largest impact on model performance, and everything else contributed less. The gap was substantial - smoking’s contribution was several times that of everything else combined.

So on both measures, the effect size and the sorting power, the same variable wins. It is not the genome.

What this cannot tell you

It is observational, and a polygenic score is a population instrument. It predicts reliably across thousands of people and says very little about any one of them. Two people with identical scores can die thirty years apart.

The cohort is also Finnish, middle-aged at baseline and genetically fairly homogeneous, which is what makes twin analysis tractable and also what limits how far the numbers travel. Polygenic scores built on European samples routinely perform worse elsewhere.

The authors’ conclusion states the ranking plainly: genetic predisposition to a longer lifespan was associated with a modest reduction in all-cause mortality risk, independent of lifestyle and other factors, while smoking and female sex were stronger predictors of mortality than the polygenic lifespan score.

The commercial version of longevity genetics sells the first half of that sentence. The study’s actual finding is the second half.