Smoking beat lifespan genes as a mortality predictor

Polygenic scores now let researchers summarise thousands of common genetic variants into a single number estimating someone's inherited predisposition to a longer life. The obvious question is how much that inheritance actually matters once you account for the things people do — smoking, drinking, moving, body weight, education. If genetics dominated, lifestyle messaging would look weaker; if lifestyle dominated, the genetic score would be a curiosity. Quantifying both in the same population, with long enough follow-up for real mortality data to accumulate, is the only way to put the two on a comparable footing.

The researchers computed a genome-wide polygenic lifespan score for 5,575 participants in the older Finnish Twin Cohort, mean age 57.4 years, of whom 45.2% were men. Cox regression estimated the effect of the score on all-cause mortality risk, first alone and then after adding sex, physical activity, BMI, alcohol consumption, smoking behaviour and education level as covariates. Concordance indices were used to quantify how much each predictor added to the model's ability to discriminate who died from who did not. Follow-up averaged 17.5 years with a standard deviation of 8.3 years, during which 1,405 deaths occurred — 25.2% of the cohort.

Each one standard deviation increase in the polygenic lifespan score was associated with a lower risk of death (hazard ratio 0.838, 95% CI 0.792–0.887), and this barely shifted after adjusting for all covariates (HR 0.863, 95% CI 0.816–0.912) — meaning the genetic association was largely independent of lifestyle. The comparison, though, was stark: smoking 20 or more cigarettes per day carried the strongest association with increased mortality (HR 3.341, 95% CI 2.751–4.056), while female sex conferred the greatest risk reduction (HR 0.678, 95% CI 0.597–0.770). On model discrimination, smoking behaviour had the largest impact (ΔC-index = 0.027); every other covariate contributed less than 0.006.

This confirms that polygenic lifespan scores capture a real, statistically robust signal, while placing that signal firmly in perspective against behavioural risk. The independence of the genetic association from lifestyle is notable — it suggests the two operate through partly separate routes rather than the score simply proxying for healthier behaviour. The very small contributions to discriminatory performance also temper enthusiasm for such scores as predictive tools at the individual level. Open questions include whether the same balance holds in younger cohorts, in non-Finnish populations, and over longer follow-up, and whether polygenic scores add anything clinically useful beyond standard risk factors.

This is a reason to think that inherited longevity is real but modest, and that it does not compete with the effect size attached to heavy smoking. A genetic predisposition towards a longer life appears to act somewhat separately from how you live, meaning it neither guarantees anything nor cancels behavioural risk. The near-negligible improvement in predictive performance from the genetic score suggests these numbers are more informative about populations than about any one person's future. It is an observational study, so these are associations rather than demonstrated causes.