Genes Linked to Smoking May Shape Health Even in People Who Never Smoked

A new study of nearly 10,000 older Americans has uncovered a puzzle that is forcing scientists to rethink what genetic risk scores for smoking actually measure. Researchers analyzing decades of data from the Health and Retirement Study found that people who carry more genetic variants associated with smoking tend to have worse health in midlife and beyond, even after every measurable aspect of their smoking behavior is taken into account. Strikingly, the association persists among people who report having never smoked at all, suggesting that the genetics of smoking may be entangled with broader biological and behavioral differences that influence health in ways that have little to do with cigarettes themselves.

The study, published in SSM – Population Health by Yeongmi Jeong, Meghan M. Skira, and Robbee Wedow, focuses on two polygenic scores, which are numerical summaries of the many small genetic differences scattered across a person’s genome. The first, called the smoking initiation score, aggregates variants associated with whether a person ever becomes a regular smoker. The second, the cigarettes-per-day or CPD score, aggregates variants associated with how heavily someone smokes once they start. Both scores were built from genome-wide association studies, or GWAS, that scanned the genomes of more than a million people of European ancestry to find single-letter DNA differences, known as single nucleotide polymorphisms, that statistically track with smoking behavior. Each person’s score is calculated by adding up the number of copies of the risk-increasing variants they carry, with each variant weighted by the strength of its estimated association.

Neither score is a crystal ball. Out-of-sample predictions indicate that the initiation score explains only about 4 percent of the variation in whether people start smoking, and the CPD score explains about 4 percent of the variation in how much they smoke. Within the study sample itself, a one standard deviation increase in the initiation score raised the probability of ever smoking by roughly 9 percentage points, while a one standard deviation increase in the CPD score was associated with about two additional cigarettes smoked per day among smokers. These are modest effects, but they are large enough, in a sample of this size, to reveal patterns that would otherwise remain invisible.

To measure health, the researchers constructed a standardized index combining thirteen distinct outcomes: self-reported general health, hospitalization, body mass index, seven doctor-diagnosed chronic conditions including high blood pressure, diabetes, cancer, lung disease, heart problems, stroke, and arthritis, and limitations in basic and instrumental activities of daily living. The analysis covered 48,444 person-wave observations from 9,515 unique participants aged 50 to 65, drawn from survey waves between 1994 and 2018. The age range was chosen deliberately, to minimize survival bias, since people with the worst health may not survive to be surveyed at older ages, and to avoid conflating genetic associations with the health effects of retirement.

The central finding is stark. A one standard deviation increase in the CPD score was associated with a 0.043 standard deviation increase in the health index, where higher values indicate worse health. Adding detailed controls for smoking behavior, including current and lifetime smoking status, the exact number of cigarettes smoked per day, the maximum ever smoked, the age of starting and quitting, and teenage smoking, attenuated the association only modestly, to 0.035 standard deviations. For context, that residual association is roughly 38 percent of the average health gap between ever smokers and never smokers in the sample. In other words, even after accounting for everything the surveys record about smoking, the genetic score still carries substantial information about health.

The most provocative result emerged when the researchers restricted the analysis to never smokers, a group of more than 20,000 observations with no reported smoking behavior to explain away. Among these individuals, a one standard deviation increase in either the CPD score or the initiation score was associated with roughly a 0.03 standard deviation increase in the health index. The CPD score also tracked with higher body mass index and obesity among never smokers, with a one standard deviation increase associated with 0.645 units of additional BMI. Because these individuals never smoked, the association cannot be attributed to their own tobacco use, although it could still reflect secondhand smoke exposure, unmeasured smoking-related environments, or something else entirely.

One obvious candidate is the family environment. Because children inherit genes from their parents, a person’s smoking-related genetic score may be correlated with parental smoking, parental substance use, parental education, and childhood socioeconomic conditions, all of which shape later health. The researchers tested this by adding controls for whether parents smoked during the respondent’s childhood, whether parents drank or used drugs heavily enough to cause family problems, parental mortality timing, parental years of schooling, and self-reported childhood financial status. These factors accounted for a modest share of the association, but more than 75 percent of the CPD score’s relationship with health remained, and about 60 percent of the initiation score’s relationship among never smokers survived the most comprehensive specifications.

The researchers then turned to psychology. Prior work has shown that smokers tend to be more impatient, risk-tolerant, and impulsive than non-smokers, and the team asked whether people with higher smoking-related genetic scores differ along similar dimensions. They found little evidence of a link with measured risk preferences, but higher scores were significantly associated with shorter financial planning horizons, greater pessimism, and lower conscientiousness, traits that economists and psychologists have long connected to lower investment in health. When these behavioral and psychological characteristics were added to the models simultaneously, they explained at most 21 percent of the remaining association between the CPD score and health, and 18 percent of the initiation score’s association among never smokers. Even with the fullest set of controls, at least 69 percent of the CPD score’s association with health persisted.

The authors are careful to emphasize what the study does not show. Polygenic scores are not clean measures of direct genetic effects. The GWAS estimates underlying them can absorb indirect genetic effects, such as parental genotypes shaping the environments children experience, as well as confounding from assortative mating, population stratification, collider bias, and case-control ascertainment. There is also the problem of pleiotropy: the same genetic variants that predict smoking may influence other biological processes and traits that affect health directly. Genetic correlation analyses have indeed documented significant overlap between smoking behaviors and a wide range of health, behavioral, and socioeconomic traits. The associations reported here are descriptive, not causal, and the study’s restriction to individuals of European ancestry, a consequence of the limited predictive power of European-derived scores in other populations, means the findings may not generalize more broadly.

Even with those caveats, the implications are significant. Economists studying smoking have long suspected that people who smoke differ from those who do not in ways that extend beyond the behavior itself, and some modeling work has speculated that genetics or correlated risk-taking might explain why smokers face elevated chronic disease and mortality risks beyond what smoking alone explains. This study provides some of the most direct evidence yet that such unobserved heterogeneity is real and partially visible in the genome. Smoking-related polygenic scores, the results suggest, contain information about health that detailed behavioral histories do not capture, pointing toward shared genetic architecture linking tobacco use, personality, and long-run wellbeing. Untangling those threads, the authors argue, will be essential for interpreting genetic risk scores in health research, and for understanding why two people with identical smoking histories can end up with very different health outcomes decades later.

Subject of Research: Associations between smoking-related polygenic scores and health outcomes in midlife

Article Title: Exploring Links Between Health and the Genetic Risk for Smoking

Article References: Jeong, Y., Skira, M. M., & Wedow, R. (2026). Exploring Links Between Health and the Genetic Risk for Smoking. SSM – Population Health, Article 101982. https://doi.org/10.1016/j.ssmph.2026.101982

Image Credits: AI Generated

DOI: 10.1016/j.ssmph.2026.101982

Keywords: polygenic scores, smoking initiation, smoking intensity, Health and Retirement Study, GWAS, pleiotropy, chronic disease, health behavior, genetic risk, conscientiousness, never smokers, population health