Details of the CKB study have been described previously [17,18]. Briefly, the CKB is a prospective cohort study of 512,715 adults aged 30 to 79 years recruited from 10 areas (five urban and five rural) of China. The baseline survey took place during 2004 to 2008. In each study assessment clinics, participants were interviewed by trained staff using laptop-based questionnaires, that covered demographic characteristics, socioeconomic status, lifestyle factors, general health, family history, mental health, and for women only, reproductive history. A range of physical measurements was undertaken by trained staff, including height, weight, and blood pressure. Blood samples were collected from each participant. All participants provided written informed consent. The Ethical Review Committee of the Chinese Center for Disease Control and Prevention (Beijing, China) and the Oxford Tropical Research Ethics Committee, University of Oxford (United Kingdom) approved the study.

For the present analysis, we excluded participants with previously diagnosed heart disease (n=15,472), stroke (n=8,884), or cancer (n=2,577), and those who had missing data for body mass index (n=2) or were lost to follow-up shortly after baseline (n=1). The final analysis included 487,198 participants.

The baseline questionnaire included a series of three questions regarding family history of stroke, which asked participant if their biological father, mother, or siblings ever had “stroke.” For siblings, the number with stroke was recorded. For each participant, the first-degree family members’ stroke history had been collected. In this study, a participant was classified as “having a family history of stroke” if they reported at least one first degree relatives had stroke. First-degree relatives included fathers, mothers, and siblings.

Our baseline questionnaire collected smoking information including frequency, amount, and types of tobacco smoked (factory cigarettes, hand-rolled cigarettes, pipe or water pipe, cigars), depth of inhalation, and the ages at which participants started smoking regularly and stopped smoking, with additional information on the main reason for cessation (ill or stopped by choice) among those who had stopped smoking. Among current and former (or ex) regular smokers, the amount of tobacco smoked (g/day) was calculated, assuming 1 g of tobacco per factory cigarette, and 2 g per cigar, with quantities smoked in pipes and hand-rolled cigarettes given as liang/month (1 liang equivalent to 50 g). Smoking duration in years was derived from the age at starting smoking regularly to age at baseline (current smokers) or duration since quitting at baseline (ex-smokers).

For the present study, participants were categorized into four groups according to the smoking status in line with previous CKB studies [19]. Never-smokers were defined as those who had not smoking at recruitment and who had not smoked more than 100 cigarettes during their lifetime. Ever-regular smokers were defined as those who had ever smoked at least one cigarette or equivalent per day for at least 6 months, or who had stopped smoking at least 6 months before recruitment because of ill. Regular smokers who had stopped smoking by choice for at least 6 months at baseline were classified as ex-smokers. Participants who did not meet the criteria for never-smoker, ever-regular smoker, or ex-smoker were classified as occasional smokers.

The baseline questionnaire also included other covariates, such as sociodemographic characteristics (age, sex, education, marital status, and household income), lifestyle behaviors (physical activity, alcohol consumption, intakes of red meat, fresh fruits, and vegetables), and menopausal status (for women). Trained staff measured weight and height using calibrated instruments. Prevalent diabetes was defined as measured fasting blood glucose ≥7.0 mmol/L, measured random blood glucose ≥11.1 mmol/L, or self-reported diagnosis of diabetes. Prevalent hypertension was defined as measured systolic blood pressure ≥140 mm Hg, measured diastolic blood pressure ≥90 mm Hg, self-reported diagnosis of hypertension, or self-reported use of antihypertensive agents at baseline.

All the participants were followed up from the time they enrolled in the study. We ascertained the vital status of each participant through the linkage with established chronic disease registries, and the disease surveillance points system in the study areas, as well as the national health insurance system. Almost all the participants (>98%) across the 10 study areas had been successfully linked to the health insurance databases. Annual active follow-up was performed by checking against local residential and health insurance records and was confirmed by street committees or village administrators. Fatal and non-fatal events were coded by the trained staff, blinded to baseline information, using the International Classification of Diseases, 10th Revision. All reported fatal or non-fatal major diseases events (including stroke) from different sources were reviewed and integrated centrally by trained clinical staff, who were blinded to baseline information. By December 2018, we have retrieved medical records for 31,069 newly reported IS cases during follow-up, among which 21,426 (69%) were confirmed as IS cases.

As the prevalence of smoking and the number of cigarettes smoked per day are much higher in men compared with women, all analyses were done separately in men and women. Percentages and mean±standard deviation of baseline characteristics were calculated across family history of stroke standardized by age and study area, when appropriate. We calculated person-years at risk from the baseline date to diagnosis of IS, death, loss to follow-up, or the censoring date (December 31, 2016), whichever occurred first. Cox regression was used to estimate the hazard ratios (HRs) of family history, with age as the underlying time scale and stratified jointly by 10 study areas and age at baseline in 5-year intervals. Multivariable models were adjusted for education (no formal school, primary school, middle school, high school, or college or university or higher), marital status (married, widowed, divorced/separated, or never married), household income (<10,000, 10,000 to 19,999, ≥20,000 Chinese renminbi/yr), alcohol consumption (less than weekly, ex-regular drinkers, weekly; or <15, 15 to 29, 30 to 59, or ≥60 g/day of pure alcohol), smoking status (never-smokers or occasional smokers, or ex-smokers, or current smokers of 1 to 14, 15 to 24, or ≥25 cigarettes or equivalents per day), physical activity as metabolic equivalent of task hours per day, intake of red meat, fresh fruits, and vegetables (daily, 4 to 6 days/week, 1 to 3 days/week, monthly, or rarely or never), menopause status (for women only, premenopausal, perimenopause, postmenopausal, or missing), body mass index (in kg/m²), and prevalent diabetes and hypertension at baseline (presence or absence).

Interactions between the family history of stroke and tobacco smoking on IS were tested by using likelihood ratio tests to compare models with and without cross-product terms (e.g., interaction term: family history of stroke × smoking categories). The HRs for IS were also calculated by strata of smoking characteristic subgroups, using individuals who reported no family history of stroke as the reference group. We further examined the risk of incident IS according to the joint classification of family history and smoking, in which smoking variables were included two categories (never-smokers and ever-regular smokers). We performed all statistical analyses using Stata version 14.2 (StataCorp., College Station, TX, USA).

Overall, the mean baseline age was 51.5±10.5 years, 67.8% (n=135,168) of men ever smoked regularly compared with 2.7% (n=7,775) of women. Of the 199,240 men and 287,958 women included, 17.8% of men and 17.5% of women had a family history of stroke. Both men and women who had a family history of stroke were, on average, more likely to reside in the urban area, have a higher level of education, and have hypertension at baseline (Table 1).

During 10 years of follow-up (4.7 million person years), we documented 16,923 incident IS cases in men and 20,656 in women. Multivariate-adjusted analyses showed that family history of stroke was associated with an increased IS risk (Supplementary Table 1). In the whole cohort, compared with participants who reported no family history of stroke, the adjusted HR for IS was 1.19 (95% confidence interval [CI], 1.16 to 1.22) for those who reported a family history of stroke. The respective HRs were 1.19 (95% CI, 1.15 to 1.23) among men and 1.19 (95% CI, 1.15 to 1.23) among women. There was no heterogeneity between men and women (P for interaction=0.92).

Among men, we observed a significant interaction between family history of stroke and IS risk across the stratum for tobacco smoking (P for interaction=0.03), with a stronger association among ever-regular smokers than among never-smokers (Table 2). Compared with men who reported no family history of stroke, the adjusted HRs for those reported a family history of stroke were 1.11 (95% CI, 1.01 to 1.23) among never-smokers and 1.21 (95% CI, 1.16 to 1.27) among ever-regular smokers. Among women, differences in the associations between family history and IS risk across the smoking status were observed, but the interaction was not significant (P for interaction=0.49). The respective HRs for those who reported a family history of stroke were 1.19 (95% CI, 1.15 to 1.23) among never-smokers, and 1.32 (95% CI, 1.14 to 1.53) among ever-regular smokers. Likewise, compared with never-smokers, the adjusted HRs for ever-regular smokers were greater in those who reported a family history of stroke than those who did not (Table 3).

We further examined the interaction between each type of family history and the smoking history (Supplementary Table 2). We only found a significant interaction between parental history and smoking status among men. Nevertheless, the associations between each type of family history and IS risk seemed to be less evident in never-smokers compared with ever-regular smokers.

Among ever-regular smokers, we further performed stratified analyses according to baseline smoking characteristics: types of tobacco smoked, age started smoking, smoking duration, smoking amount, and depth of inhalation. Among men, the HRs for family history on IS were markedly higher among all subgroups compared with never-smokers (Figure 1), but did not significantly differ across different smoking characteristics subgroups (all P for interaction >0.05). Similar but more evident associations were observed among women.

Among ex-smokers, no statistically significant heterogeneity was observed in the associations between family history of stroke and IS by years of smoking cessation (Table 4). Nevertheless, the associations between family history of stroke and IS seemed to be less evident in ex-smokers with longer years since quitting. Among men, the HRs were 1.32 (95% CI, 1.09 to 1.59) among those who had stopped for less than 10 years, and 1.01 (95% CI, 0.85 to 1.20) among those who had stopped for 10 years or more before baseline (P for interaction=0.05). Similar patterns were observed among women, with HRs of 1.01 (95% CI, 0.55 to 1.84) and 0.92 (95% CI, 0.51 to 1.65), respectively (P for interaction=0.86).

We further examined the joint association of family history and smoking with IS risk (Figure 2). The strongest association was observed in the combination of family history with ever-regular smoking. Compared with never-smokers who reported no family history of stroke, ever-regular smokers who reported a family history of stroke showed a 42% risk increasing in IS (HR, 1.42; 95% CI, 1.34 to 1.51) among men and a 43% risk increasing (HR, 1.43; 95% CI, 1.26 to 1.63) among women, respectively.