Atherosclerosis is a risk factor for dementia. However, little is known about the association between cognitive performance and a widely used indicator of coronary heart disease, at rest electrocardiography. We identified 839 older residents (mean age 81 years, 58% black) from a geographically defined biracial community in Chicago, Illinois, who had undergone extensive cognitive performance testing and met the electrocardiographic eligibility criteria, including a QRS duration of <120 ms. We then examined multivariate regression coefficients that described the associations between global cognitive performance and 4 novel descriptors of ventricular repolarization waveforms. All analyses were adjusted for age, gender, education, and race. The T wave nondipolar voltage had a significant association with global cognitive performance (p = 0.01), and this association largely remained after adjustment for cardiovascular disease risk factors (p = 0.03). In contrast, global cognitive performance was not significantly associated with the rate-adjusted QT interval, the voltage change from the beginning to end of the ST segment in lead V 5 , or the spatial angle between the mean QRS and T wave vectors. In conclusion, the strengths of the associations varied between the novel electrocardiographic descriptors of ventricular repolarization and global cognitive performance. Nevertheless, the significant association observed with T wave nondipolar voltage suggests that the cardiac effects of heart disease are associated with cognitive declines.
Previous investigations of electrocardiograms and cognitive decline have focused on ventricular excitation (the QRS complex). However, recent evidence has suggested that ventricular repolarization (ST segment and T wave) might be more sensitive to coronary heart disease (CHD). In addition, newer computerized processing techniques for standard 12-lead electrocardiograms have allowed quantitative descriptions of repolarization waveforms. In general, quantitative disease markers will be more powerful than qualitative ones. We hypothesized, therefore, that quantitative descriptors of ventricular repolarization would be associated with lower cognitive performance.
Methods
The Chicago Health and Aging Project (CHAP) is a multi-wave, community-based longitudinal study of aging that was approved by our institutional review board. In 1993, we conducted a door-to-door census (including nursing homes) in a geographically defined area of Chicago, Illinois, and identified 8,501 residents aged ≥65 years. Of these residents, 79% signed a written consent form to undergo in-home interviews that included brief tests of cognitive function. We combined these results with the participants’ age, race, and gender to draw a stratified random sample of participants for additional detailed clinical evaluation that included medical history, physical examination, phlebotomy, and extensive cognitive performance testing (see below). We have conducted this cycle of in-home interviews followed by clinical evaluations of a stratified random sample every 3 years since 1993. For each cycle since 2001, we have also identified additional age-eligible residents who had either moved into the community or became 65 years old after the census. Because at rest electrocardiograms were not obtained in the first 2 cycles, the present analyses were based on the data from the last 2 completed cycles (2001 to 2006; Figure 1 ).
Of the 1,170 unique participants from these 2 cycles, we excluded 30 participants with incomplete cognitive testing and 151 with missing or poor quality electrocardiograms. To limit our analyses to primary repolarization abnormalities, which have a normal ventricular conduction sequence, we excluded 150 participants with a QRS duration of ≥120 ms—after which, no electrocardiograms were consistent with the Wolff-Parkinson-White pattern or the presence of a ventricular electronic pacemaker. Of these 839 unique participants, 123 had repeat observations, because they had been randomly sampled once per eligible Chicago Health and Aging Project cycle. Because we accounted for correlations between repeat observations (see below), we pooled together the single and repeat observations to yield 962 total electrocardiograms. Atrial fibrillation or flutter was present in 5.4% of all electrocardiograms (52 of 962). We did not exclude the patients with these electrocardiograms, because neither rhythm disturbance necessarily affects ventricular repolarization.
Masked to participants’ sampling category, the field investigators tested 17 measures of cognitive performance during the clinical evaluations. To construct a global cognitive performance score, we first standardized each test score to a z-score (using the mean and SD of the entire group) and then averaged the z-scores from the 17 component tests to yield a composite score. Higher scores indicated greater global cognitive performance.
The field investigators used strictly standardized procedures to record the standard 12-lead resting electrocardiograms (MAC 1200, version 1.1, software version 5.1, GE Marquette, Milwaukee, Wisconsin). Then, masked to the participants’ cognitive performance data, central laboratory technicians (Epidemiological Cardiology Research Center, Wake Forest University, Winston-Salem, North Carolina) visually inspected the electrocardiograms for quality and processed them using the Marquette 12-SL Program (GE Marquette).
We a priori chose 4 novel electrocardiographic (ECG) descriptors of ventricular repolarization that have been shown to predict incident CHD events. Technicians from the central laboratory derived these variables, which are neither in routine clinical use nor measurable by sole visual inspection, by special processing of the participants’ standard 12-lead ECG signals. First, QTrr was calculated from a linear transformation of the QT interval based on the RR interval; this linear transformation is more accurate than those based on power functions. To reduce measurement error due to small waveform amplitudes in individual leads, the QT intervals were measured from composite waveforms (also known as median complexes) derived from multiple ECG leads. Second, quasi-orthogonal X, Y, and Z leads were constructed through a matrix transformation of the standard 12-lead ECG signals to generate mean QRS and T wave vectors in 3-dimensional space. The QRS/T angle was then calculated as the spatial angle between these mean vectors. Third, mathematical models (modified Chebyshev polynomials) were applied to the composite waveforms from the standard lead V 5 to calculate the linear change, or gradient, during the ST interval (STV 5 gradient). Fourth, a principal component analysis was applied to T wave signals from the 8 independent leads (I, II, and V 1 to V 6 ) to derive 8 orthogonal components, or eigenvectors. The 3 largest of these 8 orthogonal components formed an idealized and customized 3-dimensional coordinate system that represents the T wave energy that can be ascribed to a single electrical dipole (often diagrammatically represented by a vector between an anode and a cathode). The remaining T wave energy, which is represented by the 5 smaller orthogonal components, cannot be ascribed to a single electrical dipole and, therefore, represents nondipolar voltage. When repolarization is normal and uniform, the T wave energy is largely dipolar; however, as the repolarization becomes fragmented, such as in cardiomyopathy, the amount of nondipolar T wave energy increases. We calculated the T wave nondipolar voltage (TNDPV) as the square root of the sum of orthogonal components 4 through 8.
Hypertension was considered present if the participants reported a history of high blood pressure, if they were currently taking antihypertensive medications, or if their measured systolic blood pressure was >140 mm Hg or diastolic blood pressure was >80 mm Hg. Diabetes was considered present if the participants reported a history of diabetes mellitus or the use of medication to treat it. Smoking status was classified as never versus current or past. The total serum cholesterol level was measured using routine enzymatic and colorimetric methods.
To determine the association between the 4 ECG descriptors and cognitive performance, we performed multiple linear regression analysis with global cognitive performance as the dependent variable. We analyzed no additional candidate ECG descriptors. We selected other variables known to be associated with either cognitive performance or cardiovascular risk. These were age (centered at 75 years), gender, education (centered at 12 years), race (black/nonblack), hypertension (yes/no), diabetes (yes/no), cigarette smoking (yes/no), cholesterol (mg/dl), and body mass index (BMI; centered at 25 kg/m 2 ). Because the relation between BMI and cognitive performance was curvilinear in our population, we also examined BMI × BMI in our model.
Age, gender, education, and race were included in our core model. Each ECG descriptor was then added singly to this core model. We then explored whether cardiovascular disease risk factors modified the relation between the ECG descriptors and cognitive performance by adding the following terms: hypertension, diabetes, cigarette smoking, cholesterol, and BMI. Finally, because our population was composed primarily of 2 races, we added interaction terms of the ECG descriptors by race.
To calculate the p values for the parameters in these models, we used sampling weights and a jackknife replication method of variance estimation. This variance calculation, together with our stratified Bernoulli sampling method, accounted for the inherent correlations between repeat measurements.
Results
For the 839 unique participants, there were 962 clinical evaluations ( Table 1 ). Of the 4 ECG descriptors, only TNDPV was significantly associated with global cognitive performance (p = 0.01, Table 2 ). The significance of this association was only slightly attenuated after adjustment for the cardiovascular disease risk factors (p = 0.03). The associations between the other 3 repolarization abnormalities (QTrr, QRS/T angle, and STV 5 gradient) and global cognitive performance were small and nonsignificant, although each association was in the direction of more abnormal repolarization being associated with lower global cognitive performance scores. As the QTrr and QRS/T angle widened and the STV 5 gradient became flatter and, therefore, more negative, the global cognitive performance scores worsened slightly and nonsignificantly. All interaction terms of ECG descriptors by race were not significant and were removed from both the core and the fully-adjusted models.
| Characteristic | Available Data (n) | Value |
|---|---|---|
| Age (years) | 962 | 81 (76–85) |
| Women | 962 | 603 (63%) † |
| Black | 962 | 564 (59%) † |
| Education (years) | 962 | 12 (11–14) † |
| Current or former smoker | 962 | 471 (49%) |
| Hypertension | 941 | 746 (78%) |
| Diabetes mellitus | 933 | 182 (19%) |
| Body mass index (kg/m 2 ) | 912 | 27 (24–31) |
| Total serum cholesterol (mg/dl) | 836 | 205 (179–234) |
| TNDPV (μV) | 962 | 9.5 (8.2–11.1) |
| QTrr (ms) | 962 | 415 (406–427) |
| STV 5 gradient (μV) | 962 | 30 (5–54) |
| QRS/T angle ( ° ) | 962 | 75 (51–106) |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
