Sturman MT, Morris MC, Mendes de Leon CF, Bienias JL, Wilson RS, Evans DA. Physical Activity, Cognitive Activity, and Cognitive Decline in a Biracial Community Population. Arch Neurol. 2005;62(11):1750-1754. doi:10.1001/archneur.62.11.1750
Findings from studies investigating whether physical activity reduces the risk of cognitive decline in old age have been inconsistent.
To examine whether participation in physical activity by older adults reduces the rate of cognitive decline after accounting for participation in cognitively stimulating activities.
A prospective population study conducted from August 1993 to January 2003, with an average follow-up of 6.4 years.
A biracial community population on the south side of Chicago.
Participants were 4055 community-dwelling adults 65 years and older who were able to walk across a small room and had participated in at least 2 of the 3 follow-up assessments.
Main Outcome Measure
Annual rate of cognitive change as measured by a global cognitive score, which consisted of averaged standardized scores from 4 cognitive tests.
In a mixed model adjusted for age, sex, race, and education, each additional physical activity hour per week was associated with a slower rate of cognitive decline by 0.0007 U/y (P = .04). However, with further adjustments (1) for participation in cognitive activities (β̂ = .0006, P = .10),(2) for depression and vascular diseases (β̂ = .0005, P = .19), and (3) by excluding participants whose global cognitive score at baseline was at or below the 10th percentile (β̂ = .0002, P = .45), the coefficients were smaller and no longer statistically significant.
These data do not support the hypothesis that physical activity alone protects against cognitive decline among older adults.
Participation in physical activity has been shown to protect against the development of stroke1 and coronary heart disease,2 2 conditions implicated in the development of dementia.3- 5 A neuroprotective benefit of exercise seems plausible given its demonstrated effects on enhanced cerebral angiogenesis in animals.6,7 In humans, physical activity has been shown to increase cerebral blood flow8,9 and to prevent or control hypertension,10,11 diabetes mellitus,12 and hypercholesterolemia,13 3 vascular-related conditions that also are being investigated as risk factors for dementia.14,15 Several prospective studies16- 23 have examined the association of physical activity, cognitive function, and cognitive decline in old age. These previous investigations were limited to relatively homogeneous populations, single measures of cognitive function, or short follow-up times. Moreover, most previous studies22,24,25 have investigated physical activities without accounting for the reported beneficial effects of participation in cognitively stimulating activities on preventing cognitive decline. Although it is difficult to separate the cognitive and physical components of many activities, any distinct effects they may have on cognitive change are important to elucidate.
In the present study, we examined the association of physical activity in older adults with cognitive decline, as measured by multiple tests at 3 assessment periods over 6 years in a large biracial community population. We also address the question of whether participation in predominantly physical activity has a protective effect on cognitive decline independent of participation in cognitively stimulating activities.
The Chicago Health and Aging Project (CHAP) is an ongoing longitudinal study of risk factors for chronic disease among black and white older adults.26,27 CHAP is conducted in 3 adjacent neighborhoods on the south side of Chicago: Morgan Park, Beverly, and Washington Heights. From August 1993 to November 1996, a door-to-door census identified 8501 residents 65 years and older, of whom 439 died, 249 moved from the area, 1655 declined participation, and 6158 participated in the study. The effective response was 78.8% of those living and remaining in the area. In-home interviews were conducted at baseline between October 1993 and May 1997 and twice more during follow-up at approximately 3-year intervals. The interviews included brief cognitive measures of episodic memory, global cognition, and perceptual speed.
Of 6158 potential participants, 202 had no information on physical activity or cognitive function at baseline. Of the remaining 5956 persons, we restricted all analyses to those who were thought to be able to participate in physical activity by excluding 592 individuals who reported that they were unable to walk across a small room. This left 5364 participants, 803 of whom were unavailable because of death before the follow-up interview and 506 of whom were unavailable because of nonparticipation (ie, refusal, moving, or lost to follow-up). The remaining 4055 participants were included in the present analysis. Each participant signed a written consent form, and the study was approved by the Institutional Review Board of Rush University Medical Center.
Four cognitive tests were administered at each population interview, and included the East Boston Tests of Immediate Memory and Delayed Recall,28,29 the Mini-Mental State Examination,30 and the Symbol Digit Modalities Test.31 As described previously, we used a single global measure of cognitive function by combining the 4 tests.32 For each period of cognitive testing, we computed standardized scores (z scores) for each of the individual tests using the baseline population means (SDs), and then averaged the standardized scores into a single global measure of cognitive function.
Level of participation in physical activity was assessed with questions from the US Health Interview Survey that were adapted for use with older individuals.33 The activities queried included the following: (1) walking for exercise, (2) jogging, (3) yard work, (4) dancing, (5) calisthenics or general exercise, (6) bicycle riding, (7) swimming or water exercises, (8) bowling, and (9) golfing. Participants were asked whether they engaged in each of these activities during the past 2 weeks, and if they did, the number of times and average number of minutes each time. A composite index of physical activity, expressed as hours per week, was computed by summing the products of the number of minutes in each activity and the number of activity occasions for the 9 activities. This number was divided by 60 minutes, and then by 2, to produce the number of hours of physical activity per week.
Medical conditions were identified at baseline and follow-up interviews. Hypertension was considered present if the participant reported a history of high blood pressure, if measured systolic blood pressure was greater than 160 mm Hg, or if measured diastolic blood pressure was greater than 95 mm Hg. Heart disease was present if the participant reported a history of myocardial infarction, was taking digitalis, or had evidence of angina pectoris based on answers to a standardized questionnaire. Diabetes mellitus was defined as a self-reported history of diabetes mellitus or use of diabetic medications. A history of stroke was determined by participant report. A modified version of the Center for Epidemiological Studies Depression Scale was used at baseline to assess depressive symptoms.34,35
Participation in cognitive activities was assessed at the baseline interview by asking persons to rate the frequency of their participation in 7 cognitively stimulating activities: viewing television, reading newspapers, reading magazines, reading books, going to a museum, listening to a radio, and playing games, such as cards, crossword or other puzzles, or checkers. Participants rated their frequency of participation on a 5-point scale: (1) once a year, (2) several times a year, (3) several times a month, (4) several times a week, and (5) every day or about every day. As previously described, a composite index of cognitive activity ranging from 1 to 5 was then determined by averaging the individual ratings. The higher scores represented increasing levels of participation.24,25
To test whether the composite measure of physical activity was associated with the rate of cognitive decline, we used mixed models to explore change in cognitive function during follow-up.36 Such models assume that an individual’s initial level of cognitive performance and rate of change (slope) over time follow those of the population, with the exception of those random effects that may contribute variability in baseline cognitive function and rate of cognitive change over time. This approach is advantageous because it accounts for the correlation between cognitive scores at repeated assessments while also allowing a more precise estimate of the error variability.
The basic model consisted of terms for physical activity (hours per week), age (years, centered at 75 years), sex, education (years, centered at 12 years), race (black and nonblack), the time undergoing study (years since baseline), and the interaction of time with each covariate. The term for time refers to the annual rate of change in the global cognitive score in the reference group, and the interaction of each covariate with time reflects the additional effect of the covariate on the annual rate of cognitive change. We repeated the basic model with terms to control for participation in cognitive activities.
To more fully adjust for the contribution of confounding variables to the initial level and the rate of change in cognitive score, before adding model terms for physical activity, we found the best-fitting model of change in cognitive score. We investigated curvilinear associations of the covariates (using polynomial terms) and all possible interactions among the covariates, retaining only statistically significant terms (P≤.05). In this fully adjusted model, we used terms to control for participation in cognitive activities and for other factors that may be related to change in cognitive function, including depression and vascular-related medical illnesses (hypertension, diabetes mellitus, heart disease, and stroke).
We repeated the fully adjusted model, eliminating those participants with the lowest 10% cognitive scores. This sensitivity analysis was done to examine the potential influence of inaccurate recall of physical activity and to account for the possibility that these participants represented individuals with preclinical dementia, thereby making it difficult to determine whether level of physical activity was the consequence of rather than the reason for cognitive decline.
In secondary analysis, we examined the degree to which the association of physical activity with level of, and change in, cognitive function varied as a function of age, sex, race, and education by including 3-way interaction terms for these variables with physical activity and time in the model.
Of 4055 participants, 60.7% were women and 61.1% were black. The mean age was 73.5 years, and the mean education level was 12th grade. The average total follow-up was 6.4 years. The study population participated in an average of 3.7 h/wk (SD, 5.7 h/wk) of physical activity. The most common activities in which participants engaged were walking for exercise (51.2%), gardening or yard work (35.8%), and calisthenics or general exercise (24.1%). Other activities reported less frequently included bicycle riding, including stationary bikes (12.8%), dancing (7.9%), golf (3.8%), bowling (3.3%), swimming (2.9%), and jogging (1.3%). Persons with the highest participation in physical activities tended to be younger, men, white, and more highly educated and were less likely to have a major chronic condition or depressive symptoms vs persons with little to no physical activity, as tested by a 1-way analysis of variance (Table 1).
Physical activity had no significant association with level of baseline cognitive function in the basic or adjusted models. In the basic model, however, more hours of participation in physical activity at baseline were associated with a slower rate of decline in cognitive function during follow-up. More specifically, with each additional activity hour per week, the rate of cognitive decline was slower by 0.0007 U/y. By adjusting the basic model for participation in cognitive activities, the effect of physical activity on rate of change was slightly attenuated and not significant (Table 2).
In the fully adjusted model, further adjustment for vascular illnesses and depression also resulted in a smaller estimated effect, which was not significant. By using this last model, we performed a sensitivity analysis by excluding those participants whose global cognitive score at baseline was at or below the 10th percentile. This was done to examine the increased likelihood for inaccurate recall by these participants and to account for the possibility that those in the lowest cognitive performers’ group may have early cognitive impairment. The effect of physical activity on rate of change was substantially modified and not statistically significant (Table 2).
Based on the fully adjusted model after dropping the lowest cognitive performers, an average 75-year-old nonblack woman with 12 years of education and 0 h/wk of physical activity had an estimated rate of change in the global cognitive score of −0.022 U/y (standard). In contrast, a similar woman engaging in 3 h/wk of physical activity had an estimated rate of change in global cognitive score of −0.021 U/y (standard), a difference of 2.8% per year. Finally, a similar woman participating in 6 h/wk of physical activity would have an estimated rate of decline of −0.020 U/y (standard), an improvement of 5.6% per year over the woman with no activity.
We did not find a curvilinear association between physical activity and baseline cognition (β̂ = .0000, P = .98) or cognitive decline (β̂ = .0000, P = .18). An analysis of physical activity modeled in quintiles of hours per week produced similar results.
In secondary analysis, we examined whether the association was modified by age, sex, race, or education. There was a suggestion of effect modification by age, with a greater effect at older ages (β̂ for interaction = .0001, P = .05). However, this relation did not remain in the fully adjusted model or after dropping those participants whose cognitive scores were in the lowest 10%.
In further secondary analysis, we investigated the relation of the 3 most commonly performed activities on the rate of cognitive decline. We found no significant protective association on cognitive decline with walking (β̂ = .0004, P = .47), gardening (β̂ = .0010, P = .08), or calisthenics (β̂ = .0014, P = .29) when considered individually.
In this large biracial community of older adults, more hours of physical activity was associated with a small beneficial effect on the rate of cognitive decline over 6 years. The association was reduced and no longer statistically significant when we adjusted for participation in cognitively stimulating activities and in analyses that eliminated persons with the lowest cognitive performance at baseline. These results, therefore, suggest that predominantly physical activity in old age does not have a substantial inverse relation to cognitive decline.
Prospective epidemiological studies have examined this topic and found physical activity to be inversely related to cognitive decline16- 20,23 and dementia,16,37 although others21,22,24,38- 40 failed to observe an association. One plausible explanation for the inconsistencies among studies is confounding by participation in cognitively stimulating activities, which has been reported to protect against cognitive decline22,25,38 and Alzheimer disease.22,24,40 Because participants who engage in cognitively stimulating activities might be more likely also to participate in physical activities, such activity may confound any protective association observed with physical activity. Of course, some activities (eg, gardening) possess cognitive and physical benefits. Although it is difficult to separate the contribution of physical and cognitive activities to cognition, our study investigated predominantly physical activities and then adjusted for those activities with a principally cognitive component. None of the previous studies adjusted for participation in cognitively stimulating activities as a potential confounding factor.
It is possible that the weak association between physical activity and cognitive decline observed in our study is explained by the moderately low levels of physical activity reported by this older urban population compared with reports in some previous studies. For example, approximately 60% of our population was in the lowest quartile of walking (<1.9 metabolic equivalents) reported in the Nurses’ Health Study, which reported less 2-year decline in cognitive score among nurses in the upper 2 quartiles of physical activity.20 Of our population, 61.8% walked less than 0.4 km/d compared with 27% of the elderly men participating in the Honolulu Heart Program. That study37 found greater risk of incident dementia among men who walked fewer than 3.2 km/d. On the other hand, 40% of CHAP participants participated in 3 h/wk or more of activity, a level that was inversely associated with 5-year incidence of cognitive impairment among older participants of the Canadian Health and Aging Study.16
It also may be that lifetime participation in physical activity, rather than current levels, is what is important for preserving cognitive function in later life. A limitation of the CHAP study is that the measure of current frequency of participation in physical activity (in the past 2 weeks) may not have adequately captured differences among subjects’ long-term participation in physical activity.
The CHAP results are based on a prospective study of older residents of a racially and socioeconomically diverse community that enjoyed high participation and follow-up. Cognitive change was measured by multiple cognitive tests at several points during a long follow-up, which minimizes random and nonrandom error in the assessment of cognitive change. The mixed model used for analysis is superior to other analytic methods in its ability to control for initial level of cognitive function when measuring risk factor effects for cognitive change. In addition to the lack of statistical control for cognitively stimulating activities in their analyses, another limitation of previous studies is the reliance on 2 cognitive assessment periods or a single cognitive test to measure the relation of physical activity and cognitive change.
In summary, although physical activity has been shown to have many favorable effects on health, the results of this study do not provide strong support that it is inversely related to cognitive decline associated with aging. In the absence of randomized trials of long-term effects of physical activity on cognitive change, which may not be feasible, future observational studies will need to address more carefully the role of cognitively stimulating activities vs physical activities in observed associations.
Correspondence: Maureen T. Sturman, MD, MPH, Rush Institute for Healthy Aging, Rush University Medical Center, 1645 W Jackson, Suite 675, Chicago, IL 60612 (email@example.com).
Accepted for Publication: March 28, 2005.
Author Contributions:Study concept and design: Morris, Mendes de Leon, Wilson, and Evans. Acquisition of data: Morris and Evans. Analysis and interpretation of data: Sturman, Morris, Mendes de Leon, Bienias, and Evans. Drafting of the manuscript: Sturman, Morris, and Mendes de Leon. Critical revision of the manuscript for important intellectual content: Sturman, Morris, Bienias, Wilson, and Evans. Statistical analysis: Morris and Bienias. Obtained funding: Wilson and Evans. Administrative, technical, and material support: Morris, Wilson, and Evans. Study supervision: Morris, Mendes de Leon, and Evans.
Funding/Support: This study was supported by grant AG11101 from the National Institute on Aging, Bethesda, Md; and by grant ES10902 from the National Institute of Environmental Health Sciences, Research Triangle Park, NC.
Acknowledgment: We thank the residents of the Morgan Park, Washington Heights, and Beverly communities who participated in the study; Ann Marie Lane for community development and oversight of project coordination; Michelle Bos, Holly Hadden, Flavio LaMorticella, and Jennifer Tarpey for coordination of the study; and Hye-Jin Nicole Kim, MS, for statistical programming.