Baseline characteristics
193,198 participants were included in the analysis. The mean (SD) age of the participants at baseline was 64.1 (2.9) years, and 101,322 (52.4%) were female. They were followed up for 1,700,886 person-years (median [interquartile range] follow-up, 9.0 [8.3-9.7] years). One thousand seven hundred and eighty-eight cases of incident all-cause dementia were identified. Baseline characteristics of the participants are summarized in Table 1.
APOE genotype, polygenic risk and dementia risk
Dementia risk was significantly higher among APOE-ε4 allele carriers compared to non-ε4 allele carriers (Supplementary Fig. S2). Specifically, two APOE-ε4 allele carriers were at greatest risk of dementia followed by one-ε4 allele carriers compared with non-carriers, with an adjusted hazard ratio (HR) of 7.63 (95% confidence interval [CI] 6.46-9.01) and 2.47 (95% CI, 2.22-2.76), respectively (Table 2). The strength of the associations was slightly attenuated after additionally adjusting for smoking status with HR of 7.53 (95% CI, 6.36-8.91) for two-ε4 allele carriers and 2.45 (95% CI, 2.20-2.74) for one-ε4 allele carriers (Supplementary Fig. S3). Individuals with intermediate and high polygenic risk were at greater risk of developing dementia (Supplementary Fig. S2). The adjusted hazard ratio (HR) of dementia was 1.53 (95% CI, 1.29-1.81) for individuals with high polygenic risk and 1.28 (95% CI, 1.10-1.49) for those with intermediate polygenic risk (Table 2). Strength of the associations between polygenic risk and dementia risk were slightly stronger after additionally controlling for smoking status, with HR of 1.55 (95% CI, 1.31-1.83) for high polygenic risk and 1.29 (95% CI, 1.11-1.50) for intermediate polygenic risk individuals (Supplementary Fig. S3).
Smoking status with dementia risk
Dementia risk was significantly associated with smoking (Supplementary Fig. S2 and Table 2). Current and former smokers had a higher risk of dementia than those who never smoked (HR 1.36; 95% CI, 1.13-1.63) and HR 1.20; 95% CI, 1.07-1.34, respectively). The hazards ratio of dementia risk for current and former versus never smokers remained nearly unchanged after the additional adjustment of APOE genotype and polygenic risk (HR 1.37; 95% CI, 1.14-1.65 and HR 1.19; 95% CI, 1.06-1.32, respectively) (Supplementary Fig. S3).
Interaction between APOE genotype and smoking on dementia risk
Cumulative incidence of dementia according to APOE genotype and smoking status is shown in Figure 1a. Among participants with zero APOE-ε4 allele, 0.89% (95% CI, 0.73%-1.08%) current and 0.64% (95% CI, 0.58%-0.71%) former smokers developed dementia compared with 0.49% (95% CI, 0.44%-0.55%) of never smokers (HR 1.78; 95% CI, 1.39-2.29 and HR 1.23; 95% CI, 1.05-1.46, respectively) (Table 3). No significant difference was detected between smoking and dementia risk among two APOE-ε4 allele carriers (HR of 0.94 [95% CI, 0.49-1.79] for current smokers and HR of 1.32 [95% CI, 0.96-1.81] for former smokers, respectively) or among one APOE-ε4 allele carriers (HR of 1.06 [95% CI, 0.77-1.45] for current smokers and HR of 1.11 [95% CI, 0.94-1.32] for former smokers, respectively). There was significant negative multiplicative interaction between current smoking and one APOE-ε4 allele carriers (HR 0.54; 95% CI, 0.36-0.80). However, the additive interaction between current smoking and one APOE-ε4 allele carriers was not significant (RERI -0.85, 95% CI -1.84-0.13; AP -0.30, 95% CI -0.71-0.12; S 0.68, 95% CI 0.42-1.11). Neither multiplicative interaction (HR of 0.53 [95% CI, 0.27-1.05]) nor additive interaction (RERI -0.93, 95% CI -5.95-4.10; AP -0.12, 95% CI -0.82-0.58; S 0.88, 95% CI 0.43-1.81) between current smoking and two APOE-ε4 allele carriers was significant.
Interaction between polygenic risk and smoking status on dementia risk
Cumulative incidence of dementia according to polygenic risk category and smoking status is shown in Figure 1b. No significant association was observed between smoking and dementia risk among individuals within low polygenic risk (HR of 1.28 [95% CI, 0.79-2.07] for current smokers and HR of 1.04 [95% CI, 0.78-1.40] for former smokers, respectively) (Table 3). Among intermediate polygenic risk individuals, 1.01% (95% CI, 0.93%-1.11%) former smokers developed dementia compared with 0.80% (95% CI, 0.72%-0.87%) of never smokers (HR 1.23; 95% CI, 1.06-1.41), while no significant association between current smokers and dementia risk was detected (HR 1.25; 95% CI, 0.97-1.60). Among individuals with high polygenic risk, 1.77% (95% CI, 1.35%-2.27%) current smokers developed dementia compared with 1.05% (95% CI, 0.91%-1.21%) of never smokers (HR 1.63; 95% CI, 1.16-2.28), while no significant association between former smokers and dementia risk was detected (HR 1.21; 95% CI, 0.97-1.50) (Table 3). Neither multiplicative interaction (HR of 1.17 [95% CI, 0.66-2.08]) nor additive interaction (RERI 0.49, 95% CI -0.43-1.42; AP 0.22, 95% CI -0.15-0.58; S 1.62, 95% CI 0.59-4.44) between current smoking and high polygenic risk was significant. Moreover, there was no significant multiplicative interaction or additive interaction (RERI -0.05, 95% CI -0.77-0.67; AP -0.03, 95% CI -0.51-0.44; S 0.91, 95% CI 0.26-3.24) between current smoking and intermediate polygenic risk (Table 4).
Sensitivity analyses
The association between genetic predisposition and dementia risk was similar between men and women (Supplementary Fig. S4). However, current and former smoking were associated with greater dementia risk among women (HR 1.55; 95% CI, 1.16-2.07 and HR 1.29; 95% CI, 1.10-1.52, respectively) compared with never smokers but not in men (HR 1.25; 95% CI, 0.98-1.59 and HR 1.10; 95% CI, 0.95-1.28, respectively). There was significant multiplicative interaction between smoking status and APOE genotype among women (interaction P = 0.01) while no additive interaction was detected (Supplementary Table S1). Neither multiplicative interaction nor additive interaction between smoking status and polygenic risk was observed on dementia risk among women. For men, there was significant additive interaction (RERI 3.81, 95% CI 0.67-6.96; AP 0.40, 95% CI 0.14-0.66; S 1.81, 95% CI 1.09-3.01) between one APOE-ε4 allele carriers and former smoking on dementia risk while non-significant multiplicative interaction was identified (HR 1.49; 95% CI, 0.92-2.41) (Supplementary Table S2). The interaction between smoking and genetic predisposition was similar with the main analyses after additionally adjusting for self-reported depression (Supplementary Table S3) and after excluding participants followed up for less than three years (Supplementary Table S4).