3.1 Baseline characteristics
A total of 34,856 NHANES participants were included in this study, which is sufficient to represent 150,333,463 U.S. citizens. Participants were 46.33 ± 16.34 years of age, including 16,891 males and 17,965 female. Table 1 categorizes participants into non-stroke (n = 33,687) and stroke groups (n = 1,169) based on stroke status. In addition to KDMAge acceleration (p = 0.055), all factors including age, sex, race, education level, PIR, marital status, smoking status, alcohol consumption, BMI, diabetes, hypertension, triglycerides, total cholesterol, KDMAge, PhenoAge, PhenoAge acceleration, KDMAge accelerated aging, and PhenoAge accelerated aging were significantly different between the two groups (p ≤ 0.001).
Table 1
Weighted comparison in basic characteristics.
Characteristic | Total (n = 34,856) | Non-Stroke (n = 33,687) | Stroke (n = 1,169) | P | |
Age, years | 46.33 ± 16.34 | 45.92 ± 16.18 | 62.79 ± 14.17 | < 0.001 | |
Sex, n (%) | | | | < 0.001 | |
Male | 16,891 (48.6) | 16,316 (48.8) | 575 (41.6) | | |
Female | 17,965 (51.4) | 17,371 (51.2) | 594 (58.4) | | |
Race, n (%) | | | | < 0.001 | |
Non-Hispanic white | 15,822 (69.6) | 15,236 (69.5) | 586 (71.1) | | |
Non-Hispanic black | 6,741 (10.3) | 6,435 (10.1) | 306 (14.7) | | |
Others | 12,293 (20.2) | 12,016 (20.3) | 277 (14.3) | | |
Educational level, n (%) | | | | < 0.001 | |
Less than high school | 9,505 (16.9) | 9,071 (16.6) | 434 (27.2) | | |
High school or equivalent | 8,206 (24.8) | 7,896 (24.6) | 310 (31.3) | | |
College and above | 17,145 (58.3) | 16,720 (58.7) | 425 (41.1) | | |
PIR, n (%) | | | | < 0.001 | |
<1.0 | 6,797 (13.2) | 6,504 (13.0) | 293 (19.7) | | |
1-2.9 | 14,863 (36.1) | 14,268 (35.8) | 595 (47.7) | | |
≥3.0 | 13,196 (50.8) | 12,915 (51.2) | 281 (32.5) | | |
Marital status, n (%) | | | | < 0.001 | |
Married | 18,799 (56.8) | 18,219 (56.9) | 580 (52.1) | | |
Never married | 8,507 (24.6) | 8,355 (24.8) | 152 (14.3) | | |
Others | 7,550 (18.6) | 7,113 (18.3) | 437 (33.6) | | |
Smoking, n (%) | | | | < 0.001 | |
Never smoker | 18,699 (53.0) | 18,225 (53.3) | 474 (41.7) | | |
Ever smoker | 8,682 (25.0) | 8,270 (24.8) | 412 (31.7) | | |
Current smoker | 7,475 (22.0) | 7,192 (21.9) | 283 (26.6) | | |
Drinking, n (%) | | | | < 0.001 | |
Never drinker | 5,006 (11.3) | 4,774 (11.1) | 232 (20.0) | | |
Occasional drinker | 5,224 (13.4) | 5,013 (13.2) | 211 (18.3) | | |
Frequent drinker | 24,626 (75.3) | 23,900 (75.7) | 726 (61.6) | | |
BMI, n (%) | | | | < 0.001 | |
<25.0 | 10,061 (31.3) | 9811 (31.5) | 250 (21.7) | | |
25-29.9 | 11,983 (33.5) | 11588 (33.5) | 395 (33.2) | | |
≥30.0 | 12,812 (35.2) | 12288 (35.0) | 524 (45.1) | | |
Diabetes, n (%) | 5,334 (11.2) | 4,896 (10.6) | 438 (33.7) | < 0.001 | |
Hypertension, n (%) | 13,126 (33.0) | 12,216 (31.9) | 910 (74.6) | < 0.001 | |
Triglycerides, mmol/L | 1.67 ± 1.28 | 1.66 ± 1.28 | 1.93 ± 1.25 | < 0.001 | |
Total cholesterol, mmol/L | 5.10 ± 1.05 | 5.10 ± 1.05 | 4.97 ± 1.21 | 0.001 | |
KDMAge, years | 39.55 (28.37, 52.90) | 39.13 (28.17, 52.23) | 58.62 (45.54, 74.92) | < 0.001 | |
PhenoAge, years | 42.26 (29.79, 55.67) | 41.82 (29.49, 54.96) | 64.94 (51.96, 75.23) | < 0.001 | |
KDMAge acceleration, years | -5.00 (-13.62, 3.95) | -5.02 (-13.59, 3.87) | -3.42 (-15.53, 9.18) | 0.055 | |
PhenoAge acceleration, years | -2.98 (-5.76, -0.04) | -3.02 (-5.80, -0.10) | -0.72 (-4.23, 3.00) | < 0.001 | |
KDMAge accelerated aging, n (%) | 12,597 (35.3) | 12,092 (35.1) | 505 (42.1) | 0.001 | |
PhenoAge accelerated aging, n (%) | 9,609 (24.8) | 9,039 (24.3) | 570 (44.3) | < 0.001 | |
Continuous variables were presented as mean ± SD or median (IQR). Categorical variables were presented as n (%). BMI, body mass index; PIR, poverty income ratio; PhenoAge, Phenotypic age; KDMAge, biological age as calculated by the Klemera-Doubal method; SD, standard deviation; IQR, interquartile range; n, numbers of participants; %, weighted percentage; P, weighted P. |
3.2 Multivariable regression analysis
Table 2 provides a summary of the potential effects of biological aging on the incidence of stroke. The data were divided into quartiles based on the level of KDMAge acceleration and PhenoAge acceleration (quartiles 1–4). In the crude model, we found that each 10-year increase in KDMAge significantly increased the risk of stroke by 1.57 times (95%CI: 1.44–1.51, p < 0.001). Similarly, the risk of stroke also increased significantly with each 10-year increase in PhenoAge (OR: 1.92, 95%CI: 1.85-2.00, p < 0.001). Further analysis revealed a significant positive association between the fourth quartile of KDMAge acceleration and stroke (p for trend = 0.007) and a positive association between all quartiles of PhenoAge acceleration and stroke (p for trend < 0.001). The incidence of stroke in the KDMAge accelerated aging group was significantly higher than that in the non-KDMAge accelerated aging group (OR: 1.36, 95%CI: 1.21–1.53, p < 0.001). Similarly, the incidence of stroke was significantly higher in the PhenoAge accelerated aging group than in the non-PhenoAge accelerated aging group (OR: 2.59, 95%CI: 2.31–2.92, p < 0.001). These associations remained significant after controlling for multiple potential confounders. For example, the potential impact of PhenoAge on stroke incidence every 10-year remains significantly positive after adjusting for sex, age, and race/ethnicity (OR: 2.53, 95%CI: 2.26–2.83, p < 0.001). After further adjustment for education level, PIR, marital status, smoking status, alcohol consumption and BMI, the risk of stroke increased to 2.16 times (95%CI: 1.92–2.43, p < 0.001). After further adjustment for diabetes mellitus and hypertension, the risk of stroke remained significantly increased (OR: 1.72, 95%CI: 1.52–1.95, p < 0.001). In model 4, which further adjusted for triglycerides and total cholesterol, the risk of stroke remained significantly increased with each 10-year increase in PhenoAge (OR: 1.66, 95%CI: 1.46–1.88, p < 0.001).
Table 2
Weighted logistic regressions of associations between biological aging and stroke.
| Crude model | Model 1 | Model 2 | Model 3 | Model 4 |
| OR (95% CI) | P | OR (95% CI) | P | OR (95% CI) | P | OR (95% CI) | P | OR (95% CI) | P |
KDMAge, 10 years | 1.47 (1.44–1.51) | < 0.001 | 1.22 (1.18–1.26) | < 0.001 | 1.19 (1.15–1.22) | < 0.001 | 1.11 (1.07–1.14) | < 0.001 | 1.13 (1.10–1.17) | < 0.001 |
KDMAge acceleration, 10 years | | | | | | | | | | |
Continuous | 1.15 (1.10–1.19) | < 0.001 | 1.22 (1.18–1.26) | < 0.001 | 1.19 (1.15–1.22) | < 0.001 | 1.11 (1.07–1.14) | < 0.001 | 1.13 (1.10–1.17) | < 0.001 |
Quartile 1 | Reference | | Reference | | Reference | | Reference | | Reference | |
Quartile 2 | 0.71 (0.60–0.85) | | 1.16 (0.98–1.38) | | 1.12 (0.95–1.33) | | 1.01 (0.85–1.20) | | 1.05 (0.89–1.25) | |
Quartile 3 | 0.60 (0.50–0.71) | | 1.16 (0.97–1.39) | | 1.10 (0.92–1.31) | | 0.92 (0.77–1.10) | | 0.99 (0.83–1.18) | |
Quartile 4 | 1.22 (1.05–1.42) | | 2.20 (1.89–2.57) | | 1.94 (1.66–2.26) | | 1.42 (1.21–1.66) | | 1.58 (1.35–1.86) | |
| | 0.007a | | < 0.001a | | < 0.001a | | < 0.001a | | < 0.001a |
Non-KDMAge accelerated aging | Reference | | Reference | | Reference | | Reference | | Reference | |
KDMAge accelerated aging | 1.36 (1.21–1.53) | < 0.001 | 1.82 (1.62–2.05) | < 0.001 | 1.65 (1.47–1.86) | < 0.001 | 1.33 (1.18–1.51) | < 0.001 | 1.42 (1.26–1.61) | < 0.001 |
PhenoAge, 10 years | 1.92 (1.85-2.00) | < 0.001 | 2.53 (2.26–2.83) | < 0.001 | 2.16 (1.92–2.43) | < 0.001 | 1.72 (1.52–1.95) | < 0.001 | 1.66 (1.46–1.88) | < 0.001 |
PhenoAge acceleration, 10 years | | | | | | | | | | |
Continuous | 3.02 (2.70–3.39) | < 0.001 | 2.53 (2.26–2.83) | < 0.001 | 2.16 (1.92–2.43) | < 0.001 | 1.72 (1.52–1.95) | < 0.001 | 1.66 (1.46–1.88) | < 0.001 |
Quartile 1 | Reference | | Reference | | Reference | | Reference | | Reference | |
Quartile 2 | 1.25 (1.02–1.54) | | 1.26 (1.03–1.54) | | 1.17 (0.95–1.42) | | 1.07 (0.88–1.31) | | 1.05 (0.86–1.28) | |
Quartile 3 | 1.62 (1.34–1.98) | | 1.59 (1.31–1.94) | | 1.36 (1.12–1.65) | | 1.15 (0.94–1.39) | | 1.11 (0.91–1.35) | |
Quartile 4 | 3.40 (2.85–4.05) | | 2.95 (2.47–3.52) | | 2.32 (1.93–2.79) | | 1.72 (1.42–2.07) | | 1.62 (1.34–1.96) | |
| | < 0.001a | | < 0.001a | | < 0.001a | | < 0.001a | | < 0.001a |
Non-PhenoAge accelerated aging | Reference | | Reference | | Reference | | Reference | | Reference | |
PhenoAge accelerated aging | 2.59 (2.31–2.92) | < 0.001 | 2.26 (2.01–2.54) | < 0.001 | 1.93 (1.71–2.17) | < 0.001 | 1.58 (1.39–1.79) | < 0.001 | 1.52 (1.34–1.72) | < 0.001 |
Model 1: adjusted for sex, age, and race. Model 2: Model 1 + adjusted for education level, poverty–income ratio, marital status, smoking, drinking, and body mass index. Model 3: Model 2 + adjusted for diabetes and hypertension. Model 4: Model 3 + adjusted for triglycerides and total cholesterol. PhenoAge, Phenotypic age; KDMAge, biological age as calculated by the Klemera-Doubal method; OR, odds ratio; 95% CI, 95% confidence interval. a P for trend. |
3.3 Dose–response relationship between biological aging and stroke
According to the RCS analysis, there was a nonlinear relationship between KDMAge and Phenoage and the prevalence of stroke after adjusting for age, sex, race/ethnicity, education level, PIR, marital status, smoking status, alcohol consumption, BMI, diabetes, and hypertension. Figure 2 shows that there was a significant “S-shaped” relationship between KDMAge and the prevalence of stroke (p for nonlinear = 0.001). In addition, we also found a “J-type” relationship between PhenoAge and the prevalence of stroke (p for nonlinear < 0.001). It is noteworthy that although the straight line showed an upward trend with increasing PhenoAge or PhenoAge acceleration, no significant nonlinear relationship was found with the prevalence of stroke.
3.4 Subgroup analysis
Subgroup analyses further examined the association between biological aging and stroke risk in various subgroups defined by age, sex, race, education level, PIR, marital status, smoking, drinking, BMI, hypertension, and diabetes. The results showed that the incidence of stroke was positively associated with KDMAge acceleration among participants age ≥ 60 years, women, non-Hispanic black, less than high school, other marital status, smoking, diabetes, and hypertension (p < 0.05). In contrast, the incidence of stroke was negatively associated with KDMAge acceleration in the group of PIR ≥ 3.0 and frequent drinker (p < 0.05). It is noteworthy that the stratified analysis of drinking showed an interaction (p for interaction = 0.002). In addition, age ≥ 60 years, sex, non-Hispanic white, non-Hispanic black, less than high school, high school or equivalent, married, other marital status, smoking, 30 > BMI ≥ 25, BMI ≥ 30, hypertensive, non-hypertensive, diabetes, and non- diabetes was positively associated with PhenoAge acceleration (p < 0.05). In contrast, the incidence of stroke in the PIR ≥ 3.0 group was negatively associated with PhenoAge acceleration (p = 0.048). We did not find a statistically significant interaction in our stratified analysis of the association between PhenoAge acceleration and stroke. The results of subgroup analysis are shown in Figure.3.
3.5 Causal effects of biological aging on stroke
A variety of MR methods were employed to assess the impact of genetically predicted biological aging (including four accelerated aging phenotypes, TL, FI, and FA) on stroke and stroke subtypes (including IS, CES, LAS, and SVS). The results of the IVW method are presented in Fig. 4. The analysis of stroke GWAS data from the GIGASTROKE consortium revealed a positive association between genetically predicted FI and stroke (OR: 1.61, 95% CI: 1.35–1.92, p < 0.001). Further research revealed that FI was associated with IS (OR: 1.58, 95% CI: 1.29–1.93, p < 0.001) and LAS (OR: 2.69, 95% CI: 1.45–4.99, p = 0.002). The analysis of stroke GWAS data from the MEGASTROKE consortium revealed a positive association between genetically predicted FI and stroke (OR: 1.49, 95% CI: 1.12–1.98, p < 0.001), while TL was negatively associated with stroke (OR: 0.91, 95% CI: 0.84–0.99, p = 0.033). After meta-merging the IVW results of stroke GWAS data from the two consortiums, the positive correlation between FI and stroke was still observed (OR: 1.57, 95% CI: 1.36–1.83, p < 0.001). Further analysis of stroke subtypes revealed that GrimAge acceleration was negatively associated with LAS (OR: 0.86, 95% CI: 0.76–0.96, p = 0.010), while intrinsic epigenetic age acceleration was positively associated with CES (OR: 1.04, 95% CI: 1.01–1.07, p = 0.004). FI is positively correlated with IS (OR: 1.52, 95% CI: 1.29–1.79, p < 0.001), LAS (OR: 2.78, 95% CI: 1.74–4.42, p < 0.001), and SVS (OR: 1.79, 95% CI: 1.15–2.79, p = 0.010). A negative correlation was observed between FA and SVS (OR: 0.50, 95% CI: 0.30–0.83, p = 0.007).
3.6 Causal effects of stroke on biological aging
A reverse MR analysis was conducted to investigate the relationship between biological aging and stroke. The results of the IVW method are presented in Fig. 5. The analysis of stroke GWAS data from the GIGASTROKE consortium revealed that stroke was associated with PhenoAge acceleration (OR: 1.64, 95% CI: 1.14–2.36, p = 0.008), FI (OR: 1.08, 95% CI: 1.04–1.12, p < 0.001), and FA (OR: 1.02, 95% CI: 1.00-1.03, p = 0.010). In the MEGASTROKE consortium, stroke was positively associated with FI (OR: 1.14, 95% CI: 1.08–1.21, p < 0.001) and FA (OR: 1.02, 95% CI: 1.00-1.04, p = 0.020). The results of meta-analysis showed that stroke was associated with PhenoAge acceleration (OR: 1.54, 95% CI: 1.12–2.12, p = 0.008), FI (OR: 1.11, 95% CI: 1.05–1.17, p < 0.001), and FA (OR: 1.02, 95% CI: 1.01–1.03, p = 0.001). Further analysis revealed a positive correlation between IS and FI (OR: 1.09, 95% CI: 1.04–1.13, p < 0.001), as well as a positive correlation between LAS and FI (OR: 1.04, 95% CI: 1.00-1.08, p = 0.041).
3.7 Sensitivity analysis
Following the removal of samples with missing covariates from the NHANES data, the same analysis was performed, resulting in the confirmation of the robustness of the analysis. Furthermore, the other four MR methods highlight the overall robustness of the MR results (Supplementary Table 2–3). Our MR analysis demonstrated no heterogeneity or pleiotropy. The MR-Egger regression method was employed to assess the potential for horizontal pleiotropy between the SNP and the results. This analysis did not identify any evidence of such pleiotropy (Supplement Table 4). As illustrated in Supplementary Table 5, according to Cochran's Q tests, the results of the sensitivity analysis did not indicate any evidence of heterogeneity. Furthermore, the leave-one-out sensitivity analysis demonstrated that no single SNP could influence the causal relationship, thereby reinforcing the robustness of our conclusions.