Smoking Patterns in the Six Years Before Diagnosis Predict Vascular Disease Risk

A nationwide study of more than 1.1 million adults in South Korea has found that the way people’s smoking habits evolve in the six years before a diagnosis—rather than their smoking status at any single moment—is closely linked to their odds of developing cardiovascular and cerebrovascular disease. Using repeated health screening records, researchers identified five distinct smoking trajectories and showed that the odds of vascular disease climbed steadily with the intensity and persistence of smoking, reaching nearly double those of never smokers among people who smoked heavily and consistently. The strongest associations appeared for ischemic outcomes, particularly acute myocardial infarction and cerebral infarction.

The study, published in the International Journal of Public Health, drew on the National Health Insurance Service Customized Database, a population-based registry covering the entire Korean population. The researchers assembled a matched case-control study of 1,106,228 adults aged 30 to 69 who completed national general health screening in 2015 or 2016. From a baseline population of more than 17.7 million individuals after exclusions, they identified 618,109 people who were first hospitalized with a principal diagnosis of cardiovascular or cerebrovascular disease between January 2017 and December 2022. After applying strict preindex eligibility criteria—including a health screening record within two years before diagnosis and at least three nonmissing smoking measurements, and excluding anyone with a prior history of cancer or the vascular diseases under study—244,679 eligible cases remained. These were matched to controls on sex, age, and insurance premium quantile, an income-related indicator, yielding a final matched set of 241,163 cases and 865,065 controls.

What sets this study apart is its treatment of smoking as a moving picture rather than a snapshot. Smoking behavior was recorded at each health screening as never smoker, former smoker, or current smoker consuming 1–9, 10–19, or 20 or more cigarettes per day. Because electronic cigarette questions were introduced only in 2018 and were not consistently available, the analysis covered combustible cigarettes only. The researchers then applied group-based trajectory modeling, a semiparametric technique that identifies latent subgroups sharing similar longitudinal patterns, treating time as the longitudinal scale and smoking behavior as the repeated outcome. A five-group solution offered the best fit according to the Akaike and Bayesian information criteria and satisfied minimum group size requirements, producing never (50.6 percent), former (23.6 percent), high-decreasing (5.2 percent), moderate-stable (12.4 percent), and high-stable (8.2 percent) trajectory groups.

The composition of these groups revealed sharp demographic contrasts. The never group was the largest and contained higher proportions of women, people in their 60s, lower-income individuals, and rural residents. The former group was overwhelmingly male and included the highest proportion of participants with a family history of heart disease or stroke. The high-decreasing and moderate-stable groups were predominantly male and relatively young, concentrated in their 40s and 50s. The high-stable group stood out for the unhealthiest overall profile: the highest prevalence of rural residence, disability, frequent alcohol consumption, physical inactivity, and metabolic conditions such as hypertension, diabetes, and dyslipidemia.

To quantify the association between trajectory group and disease, the researchers used conditional logistic regression within matched strata, building models of increasing adjustment. The primary fully adjusted model controlled for region, family history of heart disease or stroke, disability, alcohol consumption, physical activity, body mass index, and the Charlson Comorbidity Index, which captures overall comorbidity burden. Compared with never smokers, former smokers had 11 percent higher odds of cardiovascular and cerebrovascular disease (odds ratio 1.11; 95 percent confidence interval 1.09–1.13). The high-decreasing group showed an odds ratio of 1.36 (1.33–1.40), the moderate-stable group 1.59 (1.56–1.62), and the high-stable group 1.97 (1.93–2.01)—essentially double the odds of never smokers. Crucially, the high-decreasing group fared consistently better than both the moderate-stable and high-stable groups, suggesting that cutting down, while not equivalent to quitting, was associated with lower odds than sustained heavy smoking.

The subtype-specific results carried the most striking signal. Among high-stable smokers, the strongest association was with acute myocardial infarction, at an odds ratio of 3.51 (95 percent confidence interval 3.34–3.69)—more than three and a half times the odds seen in never smokers. Cerebral infarction followed at 2.46 (2.35–2.56), then ischemic heart disease at 2.04 (1.99–2.09), overall cerebrovascular disease at 1.90 (1.84–1.96), and intracranial hemorrhage at 1.49 (1.37–1.61). All associations were statistically significant, and the pattern was clear: ischemic events, in which blood supply to the heart or brain is blocked, showed larger associations than hemorrhagic events, in which a vessel bleeds. The authors note this is biologically plausible, given smoking’s known effects on endothelial dysfunction, thrombosis, oxidative stress, and atherosclerotic progression—processes more closely tied to ischemic than hemorrhagic events.

Sensitivity analyses bolstered the findings. When participants were stratified by whether their most recent screening occurred within one year or between one and two years before the index date, the pattern of associations remained similar. More tellingly, the trajectory-based smoking measures showed better model fit than single time-point smoking measures assessed either one or six years before diagnosis, as indicated by lower AIC and BIC values. This supports the central argument of the paper: repeated assessments capture aspects of cumulative exposure that a single measurement misses. The matched design also preserved temporal ordering, since trajectories were defined entirely from preindex screening records and disease status was determined afterward, reducing the likelihood of reverse causation.

The findings align with, and extend, an international body of trajectory research. Previous studies have reported higher cardiovascular hazards among persistent smokers compared with never smokers using cohort-based trajectory designs, with reported hazard ratios ranging from roughly 1.46 to 2.77 depending on the population and methodology. What distinguishes the new work is the focus on the years immediately before diagnosis. Earlier trajectory studies often defined smoking patterns from cohort enrollment, leaving a long gap between exposure assessment and disease onset. By anchoring trajectories in the six preindex years—a window chosen because Korean screening occurs annually or biennially, guaranteeing at least three measurement opportunities—the study captures both accumulated exposure and recent changes in intensity at the moment when they may matter most for clinical disease onset.

The authors are careful to spell out the limitations. The population was restricted to health screening participants, raising the possibility of healthy screenee bias and limiting generalizability to nonparticipants. Smoking status was self-reported and subject to measurement or reporting error, and age at smoking initiation was not collected. Residual confounding by unmeasured factors, including education, employment status, diet, and sleep, cannot be excluded. Reverse causality remains possible, since some participants may have reduced or quit smoking in response to preclinical illness before their index date, although the consistency of associations across screening-timing subgroups offers some reassurance. Group assignment in trajectory modeling is probabilistic and depends on model-selection decisions, and the six-year window may not fully capture the cumulative vascular effects of decades-long smoking exposure, since smoking-related cardiovascular risks can persist over long periods.

For prevention, the practical implication is that screening systems that repeatedly collect behavioral data may be sitting on an underused resource. In South Korea, the national health examination program gathers smoking information as part of general screening for cardiovascular risk prevention, and the study suggests such longitudinal records can move risk assessment beyond current smoking status toward a characterization of prediagnostic exposure patterns. Identifying sustained high-intensity smoking profiles could, the authors argue, support earlier cessation intervention. The graded pattern across trajectories—former smokers lowest, high-stable smokers highest, with reducers in between—also reinforces a hierarchy of benefit: complete cessation remains the most effective preventive strategy, but reducing smoking intensity appears preferable to continued heavy smoking. As health systems worldwide grapple with how to target cardiovascular prevention, the message from this nationwide dataset is that the shape of a smoking history, not just its endpoint, carries information about vascular risk.

Subject of Research: Associations between prediagnostic smoking trajectories and subtype-specific cardiovascular and cerebrovascular disease risk in a nationwide South Korean health screening population

Article Title: Prediagnostic smoking intensity trajectories and subtype-specific vascular disease risk in a nationwide health screening population

Article References: Ra, G., Lee, H., Jeong, H.-S., Lee, K.-S., Won, Y.-J., & Yoo, K.-B. (2026). Prediagnostic smoking intensity trajectories and subtype-specific vascular disease risk in a nationwide health screening population. International Journal of Public Health, 71, Article 1610024. https://doi.org/10.3389/ijph.2026.1610024

Image Credits: AI Generated

DOI: 10.3389/ijph.2026.1610024

Keywords: smoking trajectories, cardiovascular disease, cerebrovascular disease, case-control study, group-based trajectory modeling, health screening, acute myocardial infarction, cerebral infarction, South Korea, National Health Insurance Service, smoking cessation, ischemic heart disease