Introduction
Neurodegenerative diseases result from deterioration of neurons, which over time lead to neurodegeneration and disabilities. Despite neurodegenerative diseases presenting a major health problem in ageing communities, modifiable risk factors such as diet and exercise have gained relatively little attention. Physical activity is known to prevent a number of chronic diseases including cardiovascular disease, hypertension, type II diabetes, and certain cancers (O'Donovan et al. Reference O'Donovan, Biddle, Blazevich, Boreham, Cooper, Crank, Ekelund, Fox, Gately, Gill, Hamer, McDermott, Murphy, Mutrie, Reilly, Riddoch, Saxton and Stamatakisin press). The association between physical activity and neurodegenerative diseases is, however, less well established. Randomized controlled trials have demonstrated that fitness training has a robust effect on improving certain cognitive processes (Colcombe & Kramer, Reference Colcombe and Kramer2003), which may be important for future risk of dementia (Linn et al. Reference Linn, Wolf, Bachman, Knoefel, Cobb, Belanger, Kaplan and D'Agostino1995; Small et al. Reference Small, Fratiglioni, Viitanen, Winblad and Backman2000). Research that attempts to examine the risk of neurodegenerative disease is, however, often prone to biases attributable to pre-existing mental or cognitive dysfunction that may influence physical activity behaviours. Various cardiovascular risk factors might also contribute to the development of neurodegenerative disease (Kivipelto et al. Reference Kivipelto, Ngandu, Fratiglioni, Viitanen, Kareholt, Winblad, Helkala, Tuomilehto, Soininen and Nissinen2005; Rosendorff et al. Reference Rosendorff, Beeri and Silverman2007), thus confounding possible associations with physical activity. Using a prospective epidemiological design it is possible to examine the cumulative effects of physical activity over several years. We therefore sought to review evidence from prospective epidemiological studies of physical activity and neurodegenerative disease risk.
Method
Study selection and data extraction
We adhered to the guidelines for reporting meta-analysis of observational studies in epidemiology (Stroup et al. Reference Stroup, Berlin, Morton, Olkin, Williamson, Rennie, Moher, Becker, Sipe and Thacker2000). We searched Medline, the Cochrane Database of Systematic Reviews and Web of Science databases from 1990 to 2007 using the key words physical activity or exercise in combination with dementia, Alzheimer's disease, Parkinson's disease, and cognition and also examined the reference lists of the retrieved articles. We limited the current systematic review to prospective cohort epidemiological studies in an attempt to eliminate selection and recall bias. Criteria for inclusion were as follows: (1) English-language full-length publication in a peer-reviewed journal; (2) prospective cohort studies in non-demented men and women at baseline, with a diagnosis of dementia or cognitive impairment at follow-up; and (3) measures of physical activity (frequency/time/type) at baseline. Information on participants and study characteristics, measurement of exposure and outcome, adjustment for potential confounders, and hazard ratio (HR) or relative risks (RRs) with 95% confidence intervals (CIs) was abstracted by one author (M.H.) and verified by another (Y.C.). Assessment of quality was made based on the validity and accuracy of the physical activity exposure measure and adjustment for potential confounding. Quality and validity were assessed independently by both reviewers and any discrepancies were resolved by discussion.
Statistical analysis
We used random effects modelling to meta-analyse the data, which allows for heterogeneity between studies (Dersimonian & Laird, Reference Dersimonian and Laird1986). HR or RR was used as a measure of the association between physical activity and risk of neurodegenerative disease. For the present analyses we assumed RRs to be a valid approximation of HRs, thereby enabling the use of one consistent measure throughout. In each case we extracted a RR for the highest versus the lowest reported physical activity group (referent group). RRs or HRs were transformed by taking their natural logarithms and standard errors were calculated from ln RR or ln HR and corresponding 95% CIs. In all analyses we used RRs from multivariate models with the most complete adjustment for potential confounders. If HR or RR was not presented we calculated RR and 95% CI from the study data with the use of the number of participants (N) and the number of cases (A) in both exposed (1, physically active) and unexposed groups (0, sedentary): RR=(A1/N1 divided by A0/N0) and 95% CI=exp[ln RR ±1.96 √(1/A1+1/A0)]. We used the Q-test for homogeneity between studies, which tests whether the between-study variability in effect sizes exceeds that expected from corresponding within-study variability. To detect publication biases we explored heterogeneity in funnel plots and the degree of asymmetry by using Begg's asymmetry method (Begg & Berlin, Reference Begg and Berlin1989). Data for dementia, Alzheimer's and Parkinson's outcomes were analysed separately. We also performed sensitivity analyses based on gender and quality of studies. All analyses were performed on a Macintosh G4 using a meta-analysis program (Masui, Reference Masui2003).
Results
We identified 1866 studies in the literature search, of which 16 studies met inclusion criteria (see Table 1). We did not include cohort studies that examined physical activity and cognitive decline without diagnosis of a neurodegenerative disease, although these studies are displayed in Table 2. The included studies incorporated 163797 participants with 3219 cases (2731 dementia/Alzheimer's, 488 Parkinson's) at follow-up. Two studies were performed in men only, whereas the remainder were mixed cohorts. For dementia and Alzheimer's disease outcomes we identified five high-quality studies (Laurin et al. Reference Laurin, Verreault, Lindsay, MacPherson and Rockwood2001; Wilson et al. Reference Wilson, Bennett, Bienias, Aggarwal, Mendes De Leon, Morris, Schneider and Evans2002; Abbott et al. Reference Abbott, White, Ross, Masaki, Curb and Petrovitch2004; Podewils et al. Reference Podewils, Guallar, Kuller, Fried, Lopez, Carlson and Lyketsos2005; Rovio et al. Reference Rovio, Kareholt, Helkala, Viitanen, Winblad, Tuomilehto, Soininen, Nissinen and Kivipelto2005) that had used an adequate and valid assessment of physical activity and adjusted for relevant confounders, including age, education, vascular risk factors, existing medical conditions, and the apolipoprotein allele 4 (ApoE4) genotype. The two studies that examined risk of Parkinson's disease were also both considered to be high quality (Chen et al. Reference Chen, Zhang, Schwarzschild, Hernan and Ascherio2005; Logroscino et al. Reference Logroscino, Sesso, Paffenbarger and Lee2006).
Table 1. Cohort studies of physical activity and neurodegenerative disease risk
EE, Energy expenditure; MET, metabolic equivalent tasks; AD, Alzheimer's disease; CI, cognitive impairment; D, dementia; PD, Parkinson's disease; MMSE, Mini Mental State Examination; 3MS, Modified Mini Mental State Examination; CDR, Clinical Dementia Rating Scale; CASI, Cognitive Ability Screening Instrument; CAPE, Clifton Assessment Procedures for the Elderly; DSM-IV, Diagnostic and Statistical Manual of Mental Disorders; ApoE4, apolipoprotein allele 4; BMI, body mass index; BP, blood pressure; CVD, cardiovascular disease.
Table 2. Cohort studies of physical activity and cognitive decline (excluded from analyses)
MMSE, Mini Mental State Examination.
Figures 1–3 show the associations between physical activity and Alzheimer's disease, dementia and Parkinson's disease respectively. The pooled RR of overall dementia in the highest physical activity category compared with the lowest was 0.72 (95% CI 0.60–0.86, p<0.001), for Alzheimer's 0.55 (95% CI 0.36–0.84, p=0.006), and for Parkinson's 0.82 (95% CI 0.57–1.18, p=0.28). There was significant heterogeneity for the association between physical activity and dementia [χ2 (13)=46.66, p<0.001] and Alzheimer's [χ2 (6)=29.12, p<0.001], but not for Parkinson's [χ2 (2)=3.09, p=0.21]. We found no evidence of publication biases in any analyses using Begg's asymmetry method (p's>0.1). To further explore the heterogeneity issue we performed sensitivity analyses based on gender and study quality. In studies that presented separate gender effects, the RR was more robust for men [0.72, 95% CI 0.57–0.90, p=0.004; χ2 (3)=0.99, p=0.80] than for women [0.46, 95% CI 0.18–1.14, p=0.09; χ2 (2)=9.95, p=0.007] in relation to risk of dementia. There was also a more robust effect in high-quality studies [0.61, 95% CI 0.46–0.82, p=0.001; χ2 (10)=43.96, p<0.001] compared with lower quality studies [0.75, 95% CI 0.60–0.93, p=0.01; χ2 (9)=33.99, p<0.001] in relation to dementia and Alzheimer's risk, although heterogeneity was displayed in both. In two studies that examined gene interactions with the ApoE4 genotype, one observed more pronounced associations of physical activity and dementia among ApoE ε4 carriers (Rovio et al. Reference Rovio, Kareholt, Helkala, Viitanen, Winblad, Tuomilehto, Soininen, Nissinen and Kivipelto2005) whereas in a different cohort the effects were absent in the carriers (Podewils et al. Reference Podewils, Guallar, Kuller, Fried, Lopez, Carlson and Lyketsos2005).
Fig. 1. The association between physical activity and Alzheimer's disease in prospective cohorts.
Fig. 2. The association between physical activity and dementia in prospective cohorts.
Fig. 3. The association between physical activity and Parkinson's disease in prospective cohorts.
Discussion
The present meta-analysis of prospective cohort studies suggests that physical activity reduces the risk of dementia and Alzheimer's disease by 28% and 45% respectively. Physical activity was not associated with a significant reduction in risk of Parkinson's disease, although this finding should be viewed in light of limited evidence in this area. Further evidence presented in Table 2 also strongly supports the role of physical activity in preventing cognitive decline.
There was a large amount of heterogeneity in our findings that was possibly due to different gender effects although there were limited studies to specifically examine this issue. Possible gender effects might reflect differences in biological responses to exercise in men and women (Day, Reference Day2008) that have relevance for cognitive health, although future studies should be conducted to confirm this finding. The higher quality studies demonstrated more robust effects and this may have been partly due to more precise categorization of physical activity exposure. A substantial proportion of the included studies did not accurately define physical activity levels in terms of frequency, duration and intensity, which may be important in relation to risk of neurodegenerative disease. Nevertheless, in studies that examined relative risk of dementia across well-defined physical activity quintiles, a linear dose–response relationship was not consistently demonstrated. For example, Abbott et al. (Reference Abbott, White, Ross, Masaki, Curb and Petrovitch2004) showed a linear reduction in risk of dementia across quintiles of increasing daily walking distance, although in other reports that assessed physical activity from combined frequency, duration and intensity data, there was weaker evidence for a dose–response relationship (Laurin et al. Reference Laurin, Verreault, Lindsay, MacPherson and Rockwood2001; Podewils et al. Reference Podewils, Guallar, Kuller, Fried, Lopez, Carlson and Lyketsos2005). Data from the Nurse's Health Study (Weuve et al. Reference Weuve, Kang, Manson, Breteler, Ware and Grodstein2004) have demonstrated that walking for a minimum of 1.5 h per week is associated with better cognitive performance and in a cohort of Dutch men, decreasing activity duration by more than 60 min/day over 10 years resulted in significant cognitive decline (van Gelder et al. Reference van Gelder, Tijhuis, Kalmijn, Giampaoli, Nissinen and Kromhout2004). The optimal dose of physical activity for risk reduction therefore remains to be accurately defined.
One important factor that is likely to bias the association between physical activity and neurodegenerative disease risk is the presence of subtle cognitive changes at baseline that may influence physical activity and other lifestyle behaviours. Thus studies with short follow-ups and a lack of adjustment for cognitive function at baseline may be particularly vulnerable to such biases. In addition, the timing of physical activity assessment in relation to neurodegeneration might be crucial. For example, it is not clear at what stage in an individual's life physical activity is most important for future protection.
Physical activity may be protective against cognitive decline and neurodegenerative diseases through a number of possible mechanisms. Regular exercise is likely to promote vascular health by lowering blood pressure, lipids, obesity and inflammatory markers, and improving endothelial function, which are risk factors for dementia and Alzheimer's disease (Kivipelto et al. Reference Kivipelto, Ngandu, Fratiglioni, Viitanen, Kareholt, Winblad, Helkala, Tuomilehto, Soininen and Nissinen2005; Rosendorff et al. Reference Rosendorff, Beeri and Silverman2007). In particular, cerebral circulation appears to be important for cognitive performance and exercise adaptations may involve improved blood flow and oxygen supply to these areas. Beneficial effects of exercise on reducing amyloid β plaques in the frontel cortex have been demonstrated recently in a transgenic mouse model (Adlard et al. Reference Adlard, Perreau, Pop and Cotman2005; Lazarov et al. Reference Lazarov, Robinson, Tang, Hairston, Korade-Mirnics, Lee, Hersh, Sapolsky, Mirnics and Sisodia2005), which might be mediated through insulin control (Farris et al. Reference Farris, Mansourian, Chang, Lindsley, Eckman, Frosch, Eckman, Tanzi, Selkoe and Guenette2003). Studies of physical activity and cerebrovascular structural abnormalities have reported no association between physical activity levels and white matter lesions (Carmelli et al. Reference Carmelli, Swan, Reed, Miller and DeCarli1999; Podewils et al. Reference Podewils, Guallar, Beauchamp, Lyketsos, Kuller and Scheltens2007), although improvements in aerobic fitness were shown be related to greater white matter integrity in select brain regions (Marks et al. Reference Marks, Madden, Bucur, Provenzale, White, Cabeza and Huettel2007). Other mechanisms may involve effects on brain plasticity and cognitive reserve, angiogenesis, neurogenesis, synaptogenesis, and increased levels of neurotrophic factors (Cotman & Berchtold, Reference Cotman and Berchtold2002; Kramer et al. Reference Kramer, Colcombe, McAuley, Scalf and Erickson2005; Cotman et al. Reference Cotman, Berchtold and Christie2007). An emerging theory is that exercise enhances several growth factors, such as brain-derived neurotrophic factor and insulin-like growth factor, which mediate the protective and therapeutic effects of exercise on brain function (Cotman et al. Reference Cotman, Berchtold and Christie2007). There is evidence to suggest the pro-inflammatory cytokines impair some of the growth factor signalling pathways in the brain, thus anti-inflammatory actions of exercise may be important.
In conclusion, our results suggest that physical activity is protective against future risk of dementia and Alzheimer's disease. However, the optimal dose of physical activity for risk reduction remains to be accurately defined and this should be a focus of future research.
Acknowledgements
Dr Hamer is supported by a British Heart Foundation (UK) grant.
Declaration of Interest
None.
