Fitness and cognition in the elderly

Study population.

The Austrian Stroke Prevention Study is a community-based cohort study on the effects of vascular risk factors on brain structure and function in elderly participants without a history or signs of stroke and dementia on the inhabitants of Graz, Austria.^15,16 Study characteristics can be found in table 1. The participation rate was 32.4%. Telephone interviews with 200 random nonresponders yielded no significant differences to responders regarding demographics and frequency of vascular risk factors known to the participants.

The current study cohort comprised 877 participants from 2 panels (1991–1994 and 1999–2003). A subset of 762, 782, and 594 participants additionally underwent MRI measures for WML volume, BPF, and lacunes, respectively. APOE genotype was available for a subset of 780 participants.

Clinical history, blood tests, MRI, cognitive testing, and exercise ECG were done on the same day. Smoking status was coded as current, former, or never, and participants were considered current smokers if they smoked >10 cigarettes a day. Body mass index (BMI) was calculated in kilograms per square meter. Total cholesterol was determined using a commercially available kit (MA-Kit 100; Hoffmann-La Roche, Basel, Switzerland). Hypertension was considered present if repeated blood pressure was ≥160/95 mm Hg or if participants received medication. Diabetes was considered present if fasting blood glucose levels exceeded 140 mg/dL or participants received medication.^15,16 Assessment of metabolic syndrome and cardiovascular disease is described elsewhere.^17,18

Standard protocol approvals, registrations, and patient consents.

We received approval from the local standard ethics committee of the Medical University of Graz for experiments using human participants. Written informed consent was obtained from all study participants.

Exercise protocol.

All participants completed a graded exercise stress test on a treadmill ergometer (General Electric Case V6) supervised by a cardiologist. None of the participants who underwent exercise ECG had clinical or ECG evidence for myocardial infarction or atrial fibrillation.

The protocol consisted of progressive increments of 25 W at 2-minute intervals. According to standard guidelines, reasons for breakoff were occurrence of angina pectoris or dyspnea, ECG ST segment changes, malignant arrhythmias, hypertension >240 mm Hg systolic blood pressure, or achievement of maximum heart rate. The stress tests were performed in the morning hours. Heart rate and blood pressure were measured in sitting position at rest and during exercise in 2-minute intervals. ECG was recorded continuously. Mass-specific V̇o₂max was calculated by the following formula: (15 mL·kg⁻¹·min⁻¹) × maximum/resting heart rate.¹⁹

Cognitive testing.

A test battery assessing the cognitive domains of memory, motor skills, and executive function was administered as described previously^15,16 and references are listed in appendix e-1 on the Neurology® Web site at Neurology.org.

To evaluate memory, a test called Bäumler's Lern-und Gedächtnistest (LGT) was used. Motor skills were evaluated by the Purdue Pegboard Test. Executive function was assessed by the Wisconsin Card Sorting Test, Trail Making Test–Part B, and Digit Span Backward, which is part of the Wechsler Adult Intelligence Scale–Revised, the Alters Konzentrationstest, and a computerized complex reaction time task (Wiener Reaktionsgerät), which is part of the Schuhfried psychological test battery.

Summary measures of cognitive functions were calculated by converting test results to z scores and computing the average scores within each cognitive domain. The z scores were not adjusted for age, sex, and education.

We additionally calculated a measure of global cognitive ability from all cognitive domains.²⁰

The listwise n was 765 and the Pearson correlation factor among the 7 tests ranged from 0.130 to 0.531, with a mean of 0.331. Principal components analysis was applied to the 7 tests. The first unrotated principal component accounted for 42.9% of the total test variance and was used as the global cognition in our subsequent analyses. Loadings on the first unrotated principal component were as follows: Alters Konzentrationstest, 1 = −0.538; figural memory (LGT), 2 = 0.646; verbal memory (LGT), 3 = 0.730; complex reaction time task, 4 = −0.541; Digit Span Backward, 5 = 0.594; Purdue Pegboard Test, 6 = 0.723; and Trail Making Test–Part B, 7 = −0.771.

MRI.

All MRI scans were performed on 1.5T supraconducting magnets (Gyroscan S 15 and ACS; Philips, Eindhoven, the Netherlands) using proton density- and T2-weighted sequences (repetition time [TR]/echo time [TE] = 2,000–2,500 milliseconds [ms]/30–90 ms) in the transverse orientation and T1-weighted images (TR/TE = 600/30 ms) in the sagittal plane. The slice thickness was 5 mm and the matrix size was 128 × 256 pixels. An axial fluid-attenuated inversion recovery sequence (TR = 10,000 ms, TE = 69 ms, inversion time = 2,500 ms, number of slices = 40, slice thickness = 3 mm, in-plane resolution = 0.86 × 0.86 mm) was added to the initial magnetic resonance protocol in the second study phase between 1999 and 2003. Assessment of brain abnormalities was only done on the dual-echo T2-weighted but not on fluid-attenuated inversion recovery scans to rely on identical MRI sequences throughout the study.

For WML volume measurements, the scans were analyzed by 2 experienced investigators who were blinded to clinical data of study participants. WMLs were graded according to our scheme. All punctate early confluent and confluent WMLs in the deep and subcortical white matter and periventricular WMLs irregularly extending into the deep white matter were marked and outlined on a transparency that was overlaid on the proton density scan. Periventricular caps, pencil-thin lining, and periventricular halos were not included for WML volume measurement as these changes are considered to be of nonischemic origin.

Independent from visual analysis, lesion load measurements were done on proton density–weighted images on an UltraSPARC workstation (Sun Microsystems, Santa Clara, CA) by a trained operator using DISPImage. The operators used a hardcopy overlaid by the transparency, with each single lesion outlined by the experienced readers as reference. Every single lesion was segmented on the computer image, and its area was provided by the semiautomated thresholding algorithm implemented in DISPImage. Each lesion volume was calculated by multiplying the area by slice thickness. Total WML volume in cubic millimeters was the sum of volumes of single lesions in a given study participant.

Lacunes were defined as focal lesions that involved the basal ganglia, internal capsule, thalamus, or brainstem, not exceeding a diameter of 15 mm.

Brain volume was calculated from the T2-weighted spin-echo sequence using the automated structural image evaluation of atrophy (SIENAX, part of the FMRIB Software Library; http://www.fmrib.ox.ac.uk/fsl). BPF was estimated from the ratio of parenchymal volume to the total volume given by the outer surface of the brain.

To determine the measurement errors in our dataset, 20 participants were scanned and rescanned after shortly leaving the scanner. The median relative error for SIENAX was 1.2%.

Statistical analysis.

Only cases with no missing data were included in the analyses. Normal distributions were tested with the Kolmogorov-Smirnov test and evaluation of quantile-quantile plots. WML volume had a skewed distribution containing zero values and therefore the value 1 was added before natural log transformation.

We assessed the relationship between V̇o₂max and cognition in 2 linear regression models. Model 1 was adjusted for age, sex, years of education, and treatment with Ca²⁺ channel antagonists or β-blockers. Model 2 additionally incorporated vascular risk factors, namely, hypertension, diabetes, total cholesterol, smoking status, and BMI. Sources of potential bias include the indirect assessment of V̇o₂max and residual confounding through unaccounted lifestyle variables.

To assess the association between V̇o₂max and WMLs, lacunes, and BPF, linear regression models including the same set of covariates, except for years of education, were performed. Multicollinearity was assessed between the independent variable V̇o₂max and the confounders in the models using the variance inflation factor. The mean variance inflation factor was 1.250 with values ranging from 1.052 to 1.938, therefore no covariates had to be excluded from the models. For each regression coefficient, the 95% confidence interval and p value were determined. A 2-tailed p value <0.05 was considered to be statistically significant.

To test the hypothesis that V̇o₂max effects on cognition are mediated by MRI findings, we used simple bootstrapped mediation models for estimating indirect effect sizes using PROCESS.²¹ Mediation was assessed for each of the MRI measures individually.

In addition, stratified analysis was performed using the median for age, binary carrier status for APOE ε4, as well as 3 groups for BMI (≤25, 26–29, and ≥30 kg/m²). Interaction terms of age, APOE ε4, and BMI were calculated by multiplying the variables individually with V̇o₂max and added to model 2 of the regression to investigate interaction effects.

To determine possible dose-effect relationships of significant associations between V̇o₂max and cognitive outcomes, participants were categorized into V̇o₂max quartiles. Linear regression analyses with V̇o₂max quartiles and cognitive measures adjusted for confounders similar to model 2 were performed, in which the lowest quartile of V̇o₂max values served as the reference. A p value for the trend was calculated by including the quartiles as a continuous variable in the model.

Because of the explorative nature of the study, no corrections for multiple testing were made.²²