Characteristics of study population
Among the 723 eligible elders with normal left ventricular ejection fraction (mean age 69.2 ± 7.2 years, 49.93% female, average left ventricular ejection fraction 71.3 ± 6.7%), the average MMSE score was 27.9 ± 2.7 (28.3 ± 2.07 for men, 27.6 ± 3.08 for women, P = 0.0008), and 97 subjects had a MMSE < 26 (13.4%; 10.8% for men vs. 16.0% for women, P = 0.0395) (Table 1).
Table 1
Characteristics of the study population with lower and higher levels of education (n = 723)
Variable | Total (n = 723) | Lower education level (≤9 years) (n = 231) | Higher education level (> 9 years) (n = 492) | P value |
Age, years | 69.2 ± 7.2 | 72.3 ± 7.6 | 67.8 ± 6.5 | < 0.0001 |
Male gender, n (%) | 361 (49.9%) | 86 (36.9) | 275 (55.9) | < 0.0001 |
Formal education, years | 10.3 ± 4.2 | 5.2 ± 1.9 | 12.6 ± 2.7 | < 0.0001 |
Body mass index, kg/m2 | 24.7 ± 3.4 | 25.3 ± 3.3 | 24.4 ± 3.4 | 0.0013 |
Brachial systolic BP | 133.4 ± 17.7 | 136.3 ± 17.3 | 132.0 ± 17.8 | 0.0024 |
Brachial diastolic BP | 77.1 ± 10.3 | 77.2 ± 9.3 | 77.0 ± 10.7 | 0.8188 |
Triglycerides, mg/dL | 127.4 ± 77.5 | 134.1 ± 77.1 | 124.1 ± 77.4 | 0.1066 |
HDL-cholesterol, mg/dL | 54.3 ± 15.6 | 53.3 ± 14.1 | 54.8 ± 16.3 | 0.2293 |
LDL-cholesterol, mg/dL | 116.3 ± 34.6 | 116.7 ± 34.2 | 116.4 ± 35.0 | 0.9114 |
Total cholesterol, mg/dL | 195.8 ± 39.5 | 196.5 ± 38.5 | 195.6 ± 40.0 | 0.7766 |
Fasting glucose, mg/dL | 104.3 ± 26.3 | 107.6 ± 31.3 | 102.7 ± 23.4 | 0.0192 |
Cardiac index, L/min/m2 | 2.8 ± 0.7 | 2.92 ± 0.75 | 2.75 ± 0.65 | 0.0022 |
CFPWV, m/sec | 13.7 ± 4.6 | 14.6 ± 5.2 | 13.3 ± 4.3 | 0.0007 |
Ejection fraction, % | 71.3 ± 6.7 | 71.3 ± 6.8 | 71.3 ± 6.7 | 0.9677 |
MMSE | 27.9 ± 2.7 | 26.3 ± 4.3 | 28.6 ± 1.8 | < 0.0001 |
MMSE < 26, n (%) | 97 (13.42) | 68 (29.19) | 31 (6.30) | < 0.0001 |
Heart disease, n (%) | 151 (20.89) | 59 (25.32) | 92 (18.70) | 0.0403 |
Stroke history, n (%) | 24 (3.32) | 12 (5.15) | 12 (2.44) | 0.0567 |
BP = blood pressure; CFPWV = carotid-femoral pulse wave velocity; HDL = high-density lipoprotein; LDL = low-density lipoprotein; MMSE = Mini-Mental State Examination. |
Subjects with a lower education level (≤9 years of formal education, average 5.2 years) were significantly older, had a significantly lower MMSE score and higher prevalence of cognitive impairment (MMSE < 26), and had a significantly higher CI and CFPWV than those with a higher education level (> 9 years of formal education, average 12.6 years) (Table 1).
Subjects with a lower education level had a higher proportion of female sex, greater body mass index, higher brachial systolic blood pressure and fasting glucose, and a higher prevalence of heart disease, compared to those elders with a higher education level (Table 1).
Association Of MMSE With CI And CFPWV
In univariate analysis for the study population, age and CFPWV were negatively, and male sex and years of formal education were positively associated with MMSE (all P < 0.0001) (Table 2). CI was not significantly associated with MMSE in the univariable analysis. In subjects with a lower education level, age, years of formal education, and CFPWV were significantly associated with MMSE. In subjects with a higher education level, only age and CFPWV remained significantly associated with MMSE.
Table 2
Association of MMSE with CI and CFPWV, univariable and multivariable linear regression analyses stratified by levels of education
Variable | Total (n = 723) | Lower education level (≤9 years) (n = 231) | Higher education level (> 9 years) (n = 492) |
Univariable analysis | BETA (SE) | P value | BETA (SE) | P value | BETA (SE) | P value |
Age, years | -0.119 (0.013) | < .0001 | -0.104 (0.030) | 0.0006 | -0.076 (0.012) | < .0001 |
Gender, Male vs. Female | 0.661 (0.195) | 0.0008 | 0.717 (0.474) | 0.1319 | 0.165 (0.162) | 0.3115 |
Formal education, years | 0.248 (0.021) | < .0001 | 0.798 (0.108) | < .0001 | 0.036 (0.030) | 0.2275 |
CFPWV, m/sec | -0.131 (0.021) | < .0001 | -0.141 (0.044) | 0.0015 | -0.082 (0.019) | < .0001 |
CI, L/min/m2 | -0.013 (0.142) | 0.9278 | 0.423 (0.304) | 0.1654 | -0.022 (0.123) | 0.8602 |
Multivariable analysis | | | | | | |
Age, years | -0.071 (0.014) | < .0001 | -0.052 (0.030) | 0.089 | -0.068 (0.013) | < .0001 |
Gender, Male vs. Female | 0.490 (0.186) | 0.0086 | 0.737 (0.436) | 0.0928 | 0.346 (0.165) | 0.0362 |
Formal education, years | 0.197 (0.023) | < .0001 | 0.701 (0.110) | < .0001 | 0.013 (0.030) | 0.6507 |
CFPWV, m/sec | -0.070 (0.021) | 0.0008 | -0.090 (0.042) | 0.0341 | -0.045 (0.020) | 0.0250 |
CI, L/min/m2 | 0.344 (0.130) | 0.0082 | 0.649 (0.274) | 0.0187 | 0.093 (0.112) | 0.4467 |
BETA = standardized regression coefficient; CFPWV = carotid-femoral pulse wave velocity; CI = cardiac index; SE = standard error of BETA. |
In the multivariable analysis for the study population, CFPWV was significantly negatively, and CI was significantly positively associated with MMSE, when age, gender, and years of formal education were included in the model (Table 2). In subjects with a lower education level, CFPWV was significantly negatively, and CI was significantly positively associated with MMSE (Table 2). In contract, in subjects with a higher education level, CFPWV remained significantly negatively associated with MMSE, but CI was no longer associated with MMSE (Table 2).
Association between cognitive impairment and quartile analyses of CI and CFPWV
For the study population, subjects in the upper quartile of CI (≥75th percentile, sex-specific) were significantly associated with a lower risk, and subjects in the upper quartile (sex-specific) of CFPWV were associated with a higher risk of cognitive impairment when age, sex, and years of formal education were accounted for (Table 3, separate upper quartile analysis). In subjects with a lower education level, subjects in the upper quartile of CI were significantly associated with a lower risk, whereas subjects in the upper quartile of CFPWV were not significantly associated with a higher risk for cognitive impairment (P = 0.0592). In contrast, in subjects with a higher education level, subjects in the upper quartile of CFPWV were significantly associated with a higher risk, whereas subjects in the upper quartile of CI were not associated with a lower risk of cognitive impairment (Table 3, separate upper quartile analysis).
Table 3
Association between low MMSE and cardiac index, carotid-femoral pulse wave velocity and education years, multivariable logistic analyses stratified by levels of education
Variable | Total (n = 723) | Lower education level (< 9 years) (n = 231) | Higher education level (≥9 years) (n = 492) |
OR (95% CI) | P value | OR (95% CI) | P value | OR (95% CI) | P value |
Separate upper quartile analysis |
Cardiac index, ≥ vs. <75th percentile | 0.484 (0.268–0.872) | 0.0158 | 0.350 (0.155–0.790) | 0.0115 | 0.765 (0.306–1.913) | 0.5664 |
CFPWV, ≥ vs. <75th percentile | 2.218 (1.313–3.748) | 0.0029 | 1.986 (0.974–4.051) | 0.0592 | 2.565 (1.108–5.941) | 0.0279 |
Bivariate upper quartile analysis |
Cardiac index, ≥ vs. <75th percentile | 0.495 (0.274–0.896) | 0.0202 | 0.357 (0.158–0.808) | 0.0134 | 0.788 (0.314–1.977) | 0.6116 |
CFPWV, ≥ vs. <75th percentile | 2.187 (1.287–3.716) | 0.0038 | 1.947 (0.939–4.037) | 0.0732 | 2.553 (1.100-5.925) | 0.0292 |
Combined upper quartile analysis |
Higher CI (≥75th percentile) and lower CFPWV(< 75th percentile) (N = 133) | 0.246 (0.112–0.542) | 0.0005 | 0.211 (0.073–0.609) | 0.0040 | 0.297 (0.077–1.149) | 0.0786 |
Lower CI (< 75th percentile) and lower CFPWV(< 75th percentile) (N = 409) | 0.403 (0.221–0.737) | 0.0031 | 0.385 (0.167–0.889) | 0.0040 | 0.406 (0.156–1.057) | 0.0649 |
Higher CI (≥75th percentile) and higher CFPWV(≥75th percentile) (N = 49) | 0.364 (0.143–0.929) | 0.0345 | 0.171 (0.043–0.671) | 0.0254 | 0.845 (0.238–3.001) | 0.7940 |
Lower CI (< 75th percentile) and higher CFPWV (≥75th percentile) (N = 132) | 1.0 (referent) | | 1.0 (referent) | | 1.0 (referent) | |
All models were adjusted for age, sex, and years of formal education. CI = cardiac index; CFPWV = carotid-femoral pulse wave velocity. |
When upper quartiles of CI and CFPWV were included in the same logistic model, upper quartile CI was significantly associated with a lower risk, and upper quartile CFPWV was significantly associated with a higher risk for cognitive impairment for the study population when age, sex, and years of formal education were accounted for (Table 3, bivariate upper quartile analysis). In subjects with a lower education level, upper quartile CI was significantly associated with a lower risk, and upper quartile CFPWV was not significantly associated with a higher risk for cognitive impairment (P = 0.0732). In contrast, in subjects with a higher education level, upper quartile CFPWV was significantly associated with a higher risk, whereas upper quartile CI was not associated with a lower risk for cognitive impairment (Table 3, bivariate upper quartile analysis).
All subjects were then divided into 4 subgroups according to higher and lower CI and CFPWV using the sex-specific 75th percentile as the cut-points. For the study population, subjects in the subgroups of higher CI and lower CFPWV, higher CI and higher CFPWV, and lower CI and lower CFPWV had significantly lower risks of cognitive impairment as compared with the referent subgroup of lower CI and higher CFPWV when age, sex, and years of formal education were accounted for (Table 3, combined upper quartile analysis). Similar significant results were observed in subjects with a lower education level but not in those with a higher education level.
Modulating effects of CI and CFPWV on the causal relationship between age and MMSE
According to the Path analysis with years of formal education as a confounder, advancing age might directly decrease MMSE, and indirectly affect MMSE favorably and unfavorable by increasing CI and CFPWV for the total study population (Fig. 1, A), and in subjects with a lower education level (Fig. 1, B). In contrast, in subjects with a higher education level, advancing age might directly decrease MMSE and indirectly decrease MMSE through increased CFPWV. However, the favorable effect of increased CI on MMSE was no longer observed (Fig. 1, C).