Physical Exercise, Aging, and Mild Cognitive Impairment: A Population-Based Study

Background  Physical exercise is associated with decreased risk of dementia and Alzheimer disease.

Objective  To investigate whether physical exercise is associated with decreased risk of mild cognitive impairment (MCI).

Design  Population-based case-control study.

Setting  The Mayo Clinic Study of Aging, an ongoing population-based cohort study in Olmsted County, Minnesota.

Participants  A total of 1324 subjects without dementia who completed a Physical Exercise Questionnaire.

Main Outcome Measures  An expert consensus panel classified each subject as having normal cognition or MCI based on published criteria.

Results  We compared the frequency of physical exercise among 198 subjects with MCI with that among 1126 subjects with normal cognition and adjusted the analyses for age, sex, years of education, medical comorbidity, and depression. The odds ratios for any frequency of moderate exercise were 0.61 (95% confidence interval, 0.43-0.88; P = .008) for midlife (age range, 50-65 years) and 0.68 (95% confidence interval, 0.49-0.93; P = .02) for late life. The findings were consistent among men and women. Light exercise and vigorous exercise were not significantly associated with decreased risk of MCI.

Conclusion  In this population-based case-control study, any frequency of moderate exercise performed in midlife or late life was associated with a reduced odds of having MCI.

Mild cognitive impairment (MCI) is an intermediate state between the cognitive changes of normal cognitive aging and dementia.^1-6 Subjects with MCI constitute a high-risk group because they develop dementia at a rate of 10% to 15% per year compared with 1% to 2% per year among the general population.⁷ Therefore, it is critical to identify potential protective factors against MCI.

Physical exercise is associated with reduced risk of heart disease, coronary artery disease, type 2 diabetes mellitus, some types of cancers, and overall mortality.^8,9 Several observational studies^10-17 showed that physical exercise may also be protective against dementia and Alzheimer disease, with few discrepant findings.¹⁸ Results of 2 studies^19,20 suggested a similar protective effect for MCI. We investigated whether physical exercise in midlife or proximate to the age at onset of MCI is associated with a reduced odds ratio (OR) for MCI in a case-control study derived from the Mayo Clinic Study of Aging.²¹

We conducted a population-based case-control study comparing subjects having MCI with subjects having normal cognition. This study was derived from the Mayo Clinic Study of Aging, which is described in detail elsewhere.²¹ Briefly, it is a population-based study designed to estimate the prevalence and incidence of MCI in Olmsted County, Minnesota. Subjects were recruited using stratified random sampling from the target population of almost 10 000 older individuals living in Olmsted County on October 1, 2004. The sampling involved equal allocation of men and women in 2 age strata (70-79 and 80-89 years). During the first follow-up period from April 2006 through July 2008, subjects were asked to complete a self-reported Physical Exercise Questionnaire; therefore, the sample of this study was restricted to 1324 subjects without dementia who completed the questionnaire.

Each subject in the Mayo Clinic Study of Aging underwent a baseline face-to-face evaluation that included the following 3 components: (1) a neurologic evaluation by a physician (Y.E.G., D.S.K., B.F.B., E.G.T., or R.C.P.), (2) a risk factor assessment by a nurse or study coordinator, and (3) neuropsychologic testing, which was interpreted by a neuropsychologist (R.J.I.). The interview by the nurse or study coordinator included administration of the Clinical Dementia Rating Scale (CDR)²² to the subject and to an informant. The neurologic evaluation was performed by a physician and included administration of the Short Test of Mental Status,²³ a medical history review, and a complete neurologic examination.

Neuropsychologic testing was performed using 9 cognitive tests to assess the following 4 cognitive domains: (1) memory (logical memory II and visual reproduction II [both delayed recall] from the Wechsler Memory Scale–Revised and delayed recall from the Auditory Verbal Learning Test),^24-27 (2) executive function (Trail Making Test B²⁸ and digit symbol substitution from the Wechsler Adult Intelligence Scale–Revised), (3) language (Boston Naming Test²⁹ and category fluency),³⁰ and (4) visuospatial skills (picture completion and block design from the Wechsler Adult Intelligence Scale–Revised). We transformed the raw scores on each test into age-adjusted scores using Mayo's Older American Normative Studies²⁷ data. These adjusted scores were also scaled to have a mean (SD) of 10 (3).^24-27

Cognitive domain scores were obtained for every subject by summing the age-adjusted scores within each domain. Because different numbers of tests were used to compute cognitive domain scores (ie, 2 tests for the executive function, language, and visuospatial skills domains vs 3 tests for memory), the domain scores were also scaled to allow comparisons across domains. In summary, the performance of a subject in a particular cognitive domain was measured by comparing his or her domain score with the score among persons with normal cognition, available from previous normative work among this same population.^24-27,31,32 However, the final decision about impairment in any cognitive domain was made by consensus agreement among the examining physician, nurse, and neuropsychologist, taking into account years of education, prior occupation, and other information.²¹

We considered as cases all subjects who met the following revised Mayo Clinic criteria for MCI^4,5: (1) cognitive concern expressed by a physician, informant, subject, or nurse; (2) cognitive impairment in 1 or more domains (memory, executive function, language, or visuospatial skills); (3) normal functional activities; and (4) without dementia. Subjects with MCI could have a CDR score of 0 or 0.5; however, the final diagnosis of MCI was not based exclusively on the CDR score but rather on all available data. We considered as controls all subjects who had normal cognition according to published normative data developed among this population.^24-27

We studied the frequency and intensity of exercise using a self-reported questionnaire with ordinal responses. We used questions from 2 previously validated instruments (the 1985 National Health Interview Survey³³ and the Minnesota Heart Survey intensity codes³⁴). Subjects were asked to provide information about physical exercise performed within 1 year of the date of cognitive assessment (late-life physical exercise) and performed at age 50 to 65 years (midlife physical exercise). The questionnaire inquired about light, moderate, and vigorous exercise.

Light exercise was defined as bowling, leisurely walking, stretching, slow dancing, and golfing using a golf cart. Moderate exercise was defined as brisk walking, hiking, aerobics, strength training, swimming, tennis doubles, yoga, martial arts, weight lifting, golfing without using a golf cart, and moderate use of exercise machines (eg, an exercise bike). Vigorous exercise was defined as jogging, backpacking, bicycling uphill, tennis singles, racquetball, skiing, and intense or extended use of exercise machines. For each category of intensity, further inquiry was made as to the frequency of exercise (times per month or per week).

We considered age, sex, years of education, medical comorbidity, and depression as covariates. We measured medical comorbidity using the weighted index by Charlson et al,³⁵ which considers the number and severity of diseases (range, 0-33). We measured depression using the Beck Depression Inventory.³⁶

We conducted a set of primary analyses considering only intensity of exercise to test whether any frequency of exercise was associated with decreased risk of MCI. The “once a month or less” category served as the reference. Prompted by the results of the primary analyses, we also conducted a set of secondary analyses considering the frequency and intensity of exercise. The strength of the association between physical exercise and MCI was measured using ORs and corresponding 95% confidence intervals (CIs) after adjusting for age (continuous variable), sex, years of education (continuous variable), medical comorbidity (weighted Charlson Comorbidity Index as a continuous variable), and depression (Beck Depression Inventory score of <13 vs ≥13).

Analyses were conducted separately for physical exercise performed at age 50 to 65 years and for physical exercise performed within 1 year of the date of cognitive assessment. We considered 3 levels of intensity of exercise (light, moderate, and vigorous) and 6 levels of frequency of exercise (≤1 time per month, 2-3 times per month, 1-2 times per week, 3-4 times per week, 5-6 times per week, and daily). Because the 3 categories of intensity of exercise were not mutually exclusive, it was impossible to collapse these categories.

We also conducted a set of sensitivity analyses using a composite score obtained by assigning a numeric score to the frequency of physical exercise and by adding the scores across the light, moderate, and vigorous strata (equal weighting was given to all strata). The scores were 0 for 1 time per month or less, 0.5 for 2 to 3 times per month, 1.5 for 1 to 2 times per week, 3.5 for 3 to 4 times per week, 5.5 for 5 to 6 times per week, and 7 for daily. The total composite score ranged from 0 to 21. We consider these as secondary (sensitivity) analyses because the results varied noticeably depending on the assumptions made in computing the scores (weights were assigned to different responses).

Statistical testing was performed at the conventional 2-tailed α = .05. All analyses were performed using commercially available statistical software (SAS, version 8; SAS Institute, Cary, North Carolina).

Table 1 summarizes the demographic characteristics of 198 subjects with MCI and 1126 subjects with normal cognition. Among subjects with normal cognition, there were equal numbers of men and women, whereas there were more men than women among subjects with MCI. The median ages were 83 years (interquartile range, 78-86 years) among subjects with MCI and 80 years (interquartile range, 76-84 years) among subjects with normal cognition.

We assessed the reliability of the Physical Exercise Questionnaire in 2 ways. First, we studied its internal consistency using Cronbach α and observed a value of 0.71 (in the moderate to good range). Second, among a subsample of 87 subjects who completed the questionnaire at 2 successive visits, we computed a test-retest Spearman rank correlation coefficient. The correlations in the overall group were 0.47 for light exercise, 0.50 for moderate exercise, and 0.33 for vigorous exercise. The test-retest correlations were similar among 73 subjects with normal cognition and among 14 subjects with MCI (of 87 subjects with 2 interviews [data not shown]).

Table 2 summarizes the results of the primary analyses dichotomizing physical exercise into any frequency of exercise vs none. Light exercise had ORs of 0.90 (95% CI, 0.55-1.47; P = .68) for midlife and 0.69 (95% CI, 0.47-1.00; P = .048) for late life. Moderate exercise had ORs of 0.61 (95% CI, 0.43-0.88; P = .008) for midlife and 0.68 (95% CI, 0.49-0.93; P = .02) for late life. The findings were consistent among men and women (Table 2 [footnotes b and c]). Vigorous exercise had ORs of 0.82 (95% CI, 0.59-1.15; P = .25) for midlife and 1.14 (95% CI, 0.72-1.81; P = .58) for late life. As expected, fewer subjects reported vigorous exercise in late life; therefore, the nonsignificant associations for this analysis may be due in part to the lack of statistical power. Results using a composite score of physical exercise were comparable (Table 2 [footnote a]).

Table 3 summarizes the results of our case-control analyses considering the frequency and intensity of exercise performed in midlife (age range, 50-65 years). The point estimates for almost all frequencies of light and vigorous exercise were between 0 and 1, suggesting a potential “protective” effect. However, none of these associations were statistically significant. In contrast, several frequencies of moderate exercise were significantly associated with decreased risk of MCI. Table 4 summarizes the results of our case-control analyses considering the frequency and intensity of exercise performed in late life (within 1 year of the date of cognitive assessment). Except for 1 variable, no significant association was noted between physical exercise and decreased risk of MCI in any of the analyses.

In this population-based case-control study, midlife moderate exercise was associated with a 39% reduced OR for MCI. Similarly, late-life moderate exercise was associated with a 32% reduced OR for MCI. The ORs for light and vigorous exercise were also consistently less than 1.00 in most primary analyses; however, most of these associations were not statistically significant. This may be due in part to the limited statistical power.

Observational studies^{10-14,16,17,37-42} have reported possible beneficial effects of physical exercise among older subjects with normal cognition and among subjects with dementia and Alzheimer disease. In contrast, investigators from the Chicago Health and Aging Project reported that physical activity conducted within 2 weeks of the date of cognitive assessment was associated with no decreased risk of cognitive decline among an older population.¹⁸ That negative finding may have been due in part to the timing of physical exercise proximate to the assessment of cognition.

One study¹⁹ reported a suggestive but nonsignificant association between physical activity and reduced risk of amnestic MCI. Several observational studies also have reported an association of physical exercise with decreased risk of cognitive decline. Although cognitive decline does not coincide with our definition of MCI, these studies are relevant to the interpretation of our findings. The Nurses' Health Study, involving 18 766 women aged 70 to 81 years, reported that long-term physical activity was associated with reduced risk of cognitive decline.⁴³ Similarly, the Monongahela Valley Independent Elders Survey project¹³ reported that higher exercise level (defined as aerobic exercise for ≥30 minutes performed ≥3 times per week) was associated with reduced risk of cognitive decline. The Monongahela Valley Independent Elders Survey project included a complete assessment of physical exercise (including the frequency, intensity, and duration); however, its outcome measure was limited to a Mini-Mental State Examination score.⁴⁴ The Canadian Health and Aging Study¹⁰ examined the association of physical exercise with cognitive impairment–no dementia (CIND) and dementia in a nested case-control study. Although CIND and MCI differ, they both describe the gray zone between normal cognitive aging and dementia. The Canadian Health and Aging Study investigators reported that physical activity was associated with a 42% reduced risk of CIND.

Recently, a team of Australian investigators conducted a clinical trial of 170 volunteers 50 years and older who reported memory problems but who did not meet the criteria for dementia.²⁰ Subjects were randomized to a program of education and usual care or to a 24-week home-based program of physical activity. Physical exercise improved cognitive function among older adults at risk for Alzheimer disease, including an unspecified number of subjects with MCI. These benefits were observed 6 months after initiation of the physical activity and were sustained 12 months after the intervention had been discontinued.

All of these observational studies used retrospective questionnaires and interviews to measure physical exercise; hence, some degree of recall bias is inherent. However, a University of California, San Francisco, study^11,40 that objectively measured physical fitness reported similar findings. The investigators prospectively followed up 349 community-dwelling older women for 6 to 8 years. At baseline, they objectively measured physical fitness using a treadmill duration test and a peak oxygen consumption test. They also used the oxygen uptake efficiency slope, which is a measure of cardiorespiratory fitness independent of motivation and effort. The investigators observed that subjects who were in the highest tertile of cardiorespiratory fitness experienced less cognitive decline over a 6-year follow-up period.

The findings of our study should be interpreted within the context of the following limitations. The first limitation pertains to study design. The exposure (physical exercise) and the outcome (MCI) were measured at a cross-sectional point in time. Therefore, it is difficult to study the direction of causality. The second limitation relates to the measurement of physical exercise. As in many other observational studies, we used a self-reported questionnaire to collect physical exercise data. Such measurement is prone to recall bias.^13,45 The third limitation is that few subjects engaged in vigorous exercise in late life; therefore, statistical power was limited for that analysis.

Our study did not address mechanisms of action. Based on the literature, we can speculate that physical exercise may be directly protective against MCI via increased production of neurotrophic factors,⁴⁶ greater cerebral blood flow, improved neurogenesis, enhanced neuronal survival, mobilization of gene expression affecting neuronal plasticity,^47,48 and decreased risk of cardiovascular and cerebrovascular diseases.⁴⁹ A second possibility is that physical exercise may be a marker for a healthy lifestyle. A subject who engages in regular physical exercise may also show the same type of discipline in dietary habits, accident prevention, adherence to preventive intervention, compliance with medical care, and similar health-promoting behaviors.

In summary, our findings contribute to the growing body of literature that indicates the potentially beneficial relationship between physical exercise and cognition. A future population-based cohort study is needed to confirm whether physical exercise is associated with decreased risk of incident MCI. The population-based setting will improve generalizability, and the prospective cohort design will strengthen causal inferences.

Correspondence: Yonas E. Geda, MD, MSc, Mayo Clinic, 200 First Street SW, Rochester, MN 55905 (geda.yonas@mayo.edu).

Accepted for Publication: August 20, 2009.

Author Contributions: Dr Geda had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Geda, Roberts, Knopman, Petersen, and Rocca. Acquisition of data: Geda, Knopman, Ivnik, Boeve, Tangalos, and Petersen. Analysis and interpretation of data: Geda, Christianson, Pankratz, and Rocca. Drafting of the manuscript: Geda. Critical revision of the manuscript for important intellectual content: Geda, Roberts, Knopman, Christianson, Pankratz, Ivnik, Boeve, Tangalos, Petersen, and Rocca. Statistical analysis: Christianson and Pankratz. Obtained funding: Geda and Petersen. Administrative, technical, and material support: Geda and Petersen. Study supervision: Roberts and Petersen.

Financial Disclosure: Dr Knopman serves on a data safety monitoring board for Eli Lilly and is an investigator for clinical trials sponsored by Baxter Pharmaceuticals, Elan Pharmaceuticals, and Forest Pharmaceuticals. He is an associate editor of Neurology, for which he receives compensation from the American Academy of Neurology. He served as a one-time consultant to GlaxoSmithKline in the past year.

Funding/Support: This study was supported by grant K01 MH068351, career transition award U01 AG006786 (Dr Geda), K01 AG028573, P50 AG016574, R01 AR030582, and R01 NS033978 from the National Institutes of Health; by the Robert H. and Clarice Smith and Abigail Van Buren Alzheimer's Disease Research Program; and by the Harold Amos (Robert Wood Johnson) Medical Faculty Development Program.

Additional Contributions: David A. Mrazek, MD, FRCPsych, provided administrative support in the design and conduct of grant K01 MH068351.