Minimum physical activities protective against Alzheimer’s disease in late life: a systematic review

Literature search

The systematic review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [34] and utilized MEDLINE via the PubMed and Cumulative Index to Nursing and Allied Health Literature (CINAHL) databases. The search strategy involved medical subject headings (MeSH) and keywords such as “exercise” OR “physical activity” AND “Alzheimer’s disease” OR “cognitive impairment” OR “dementia.” We utilized a combination of these terms and filters to access the full search results. See Supplementary Material for a detailed search strategy. The search was conducted in August 2023.

Inclusion and exclusion criteria

The inclusion criteria for the study were as follows: prospective cohort or interventional studies written in English that measured PA in a cohort of older adults without dementia, AD, or cognitive decline at the beginning of the study and that measured the incidence of AD as an outcome. Reviews, case reports, cross-sectional studies, retrospective studies, and other studies that did not meet the inclusion criteria were excluded.

Study selection

Covidence software was utilized to screen the articles [35]. Two reviewers (AS and AL) independently screened the articles by title and abstract, while a third reviewer (PL) was consulted to address any discrepancies. Following this initial screening, the selected studies underwent full-text reviews in accordance with the specified inclusion and exclusion criteria.

Risk of bias

The risk of bias was assessed utilizing the Robins-E tool to assess the risk of bias in nonrandomized follow-up studies of exposure effects [36]. Studies were assessed according to the Robins-E criteria by independent reviewers in seven domains: confounding, measurement of exposure, selection of participants, postexposure intervention, missing data, measurement of the outcome, selection of reported results, and overall bias judgment Figures were created utilizing the Risk-of-bias Visualization (Robvis) tool [37].

Study characteristics

Study characteristics were separated into those studying PA during midlife, late life, and unspecified times of life (Tables 1–3). Overall, the study populations were from the United States, Canada, and various European countries, including Sweden, Finland, France, Germany, Denmark, Italy, and the United Kingdom. Sixteen of the studies analyzed the general population, whereas one study focused on patients. The sample sizes varied from 466 to 90,320, with a total of 206,463 participants. The average age was 67.7 years, and the follow-up time varied from 3.9 to 44 years. Our method identified 92 duplications compared to 3,580 in a previous review [8], which seems to be due to the efficiency of machine learning–assisted literature screening. Of the 17 studies, 13 reported a statistically significant inverse association between PA and the incidence of AD.

Midlife

Four studies measured PA during midlife (defined as participants before the age of 65 years or if the average age was 50 years or younger), as shown in Table 1.

The study populations were the general populations of Sweden and Finland, with an average age of 49.1 years and an average follow-up time of 29.2 years. Three studies measured PA utilizing questionnaires [38], 40], 41], and one study conducted structured interviews [39]. Two of the four studies had significant findings that regular PA at midlife had a protective effect against AD [38], 40]. Only one of the two studies conducted in Sweden found a significant association between PA and AD incidence, while only one of the two studies conducted in Finland found a significant association. These findings suggest that the country of study may not be a significant factor influencing outcomes. Other aspects of the studies, such as the differences in methods of assessing PA, and the outcomes and covariates included in the multivariate models, may be driving these differences across studies. Two of the three studies measuring leisure-time PA were significant [38], 40], and one measuring occupational PA was not significant [41]. Najar et al. [39] and Rovio et al. [40] included a measurement of intensity and duration in their thresholds of PA. Rovio et al. [40] reported a significant decrease in the incidence of AD with PA, with associated breathlessness and sweating for at least 20 min at a time and occurring at least twice a week, compared to less than twice a week. Najar et al. [39] did not report a significant relationship with a threshold of light PA of a minimum of 4 h, regular physical training for at least 2–3 h, or intense physical training several times per week.

Late life

Eleven studies measured PA during late life (defined as an average age of 70 years or older and/or inclusion criteria of participants aged 65 years or older), as shown in Table 2.

The study participants comprised the general populations of the United States, Canada, France, Italy, and Denmark, and a patient population in Germany. The study population had an average age of 75.9 years and an average follow-up time of 5.9 years. Based on the country of study, all six studies conducted in the United States found a significant association between PA and AD incidence. Only one study was conducted in France, Germany, and Canada. One study from each of these countries found a significant association. The study in Denmark and the study in Italy did not find a significant association. One study assessed PA utilizing accelerometer data [39], one study conducted structured interviews [45], and nine studies utilized questionnaires [43], 44], [46], [47], [48], [49], [50], [51], [52]. Eight out of 11 (75 %) studies reported significant findings [42], [43], [44], [45], [46, 48], 50], 52]. Six studies measured leisure PA [44], [45], [46], [47], [48], [50], and four reported a significant relationship. [44], [45], [46], [48], 50]. One study measured household/transportation PA and was significant [43]. Four measured total PA [42], 49], 51], 52], and three reported a significantly protective benefit [42], 51], 52].

Several studies have included the duration or intensity in PA in their assessments. Both Larson et al. [44] and Middleton et al. [46] found that PA at least three times a week conferred a significant protective benefit, but the former required at least 15 min at a time, while the latter required an intensity equal to walking. Luck et al. [45] found similar results with a threshold of PA every day or several times a week. Scarmeas et al. [50] found that the combinations of light, moderate, or vigorous exercise several times a week decreased incidence of AD, compared to those who did low activity or 0 h of PA. Podewils et al. [48] observed that at least four physical activities in a 2-week period confer a significant benefit, compared to one or fewer physical activities.

Studies also calculated subscores and percentiles to compare PA levels. Dupré et al. [43] calculated a household or transportation subscore and found that the intermediate score conferred a protective benefit, compared to no activity at all. Two studies reported 24-h PA calculated as the average hours per day spent in different basal states. Taaffe et al. [51] found that a high PA subscore was protective from developing AD, compared to a low PA score. Similarly, Tan et al. [52] found that the second quintile of PA was protective, compared to the lowest quintile. Buchman et al. [42] utilized accelerometer data for total daily PA and found that the 90th percentile of PA had a protective benefit, compared to the 10th percentile of accelerometer data.

We examined the age-stratified effects of these studies. Although all studies adjusted for age in their multivariate analyses, not all studies reported an association between age and AD incidence. Three studies specifically discussed this association utilizing their multivariate models. Buchman et al. [42] found that total daily PA and incident AD did not change with age, when age was treated as a continuous variable. This finding was based on a multivariate proportional hazards model that included age, sex, race, and interaction terms between these variables and the total PA. In contrast, Dupré et al. [43] stratified participants into two age groups: those below 80 years old and those aged 80 years and older. They discovered that a household/transportation subscore of 1.6–2.0 (with a score of 1.6 or lower as the reference) was significantly associated with AD incidence in individuals aged 80 and older, whereas no significant association was found for those under 80. Lastly, both Luck et al. [45] and Neergaard et al. [47] found that in their multivariate proportional hazards models, older age was associated with a higher incidence of AD. However, the covariates included in these models differed between studies.

Unspecified

Two studies analyzed a population in the United Kingdom that did not specify a targeted age range. The study populations had an average age of 60.1 years (ranging from 37 to 73 years old) and an average follow-up time of 6.6 years (Table 3). Both studies utilized accelerometer data to assess the total daily PA. Petermann-Rocha et al. [53] and Zhong et al. [54] utilized 900 metabolic equivalents (METs) per minute per week and 27 milli-gravity units per week, respectively, for the active group threshold. Both studies conferred protective benefits at these thresholds.

Risk of bias assessment

The risk of bias assessment in the overall quality of evidence was “low risk” in 13 studies and “some concerns” in four studies (Figure 2). The studies were placed in the “some concerns” category due to potential bias in confounding, exposure measurement, and missing data. None of the studies were deemed to have a high or very high risk of bias.

Figure 2:

The risk of bias of various studies of physical activity (PA).

Midlife

Two of the four included studies that assessed midlife PA resulted in a significant protective benefit against developing AD. Three studies – Andel et al. [38], Najar et al. [39], and Rovio et al. [40] – measured leisure-time PA at midlife, with at least two decades of follow-up . Of these, only Najar et al. did not find significant protection from AD, although the study reported a reduced risk of mixed dementia and dementia with cerebral vascular disease [39].

In contrast, both Andel et al. and Rovio et al. presented a strong significant reduction in AD risk [38], 40]. Andel et al. specifically assessed regular exercise involving sports from age 25 to50 years old [32], which highlights the significant benefit of cumulative PA leading up to midlife. This concurs with previous meta-analytic data that concluded that PA provides a significant risk reduction, especially when comparing moderate levels of PA to extreme sedentary behavior [5], 7], 49], 55]. However, Najar et al. did not find a significant difference between light PA (such as walking) and inactivity [39], suggesting that this level of activity may not be sufficient to offer significant protective benefits. In contrast, Rovio et al. assessed leisure-time PA utilizing more specific criteria: at least 20 min of activity that caused breathlessness and sweating, performed at least twice a week [40]. The inclusion of a measure of intensity in the definition of PA may provide valuable insights into the minimum intensity required for a protective effect, with breathlessness and sweating potentially serving as indicators of more protective activity levels.

One study measured midlife occupational PA and determined that it did not provide a significant protective effect against AD [41]. This suggests that leisure-time and occupational activities may have different effects on AD development. There are many hypotheses about why this is the case, with some evidence finding that the population who worked occupations requiring manual labor are associated with lower socioeconomic status and more adverse health outcomes after retirement, including dementia and cardiovascular disease [30], 56], 57]. However, Rovio et al. found that occupational PA did not demonstrate a statistically significant result, even when controlled for socioeconomic variables, including income and education levels [41]. Leisure-time PA was also controlled for, suggesting that occupational PA likely does not have an independently significant protective effect on the incidence of AD. Further studies are needed to address these nuances.

Late life

The current analysis showed that PA, especially during later life, had a significant protective effect against the development of AD. This supports previous studies indicating that PA has a protective effect on the development of neurodegenerative diseases [58], 59], specifically AD [4], 6], 8], 30], [60], [61], [62], and among older adults [62], 63]. Our review found that various types of PA are protective, including household or transportation, leisure, sports, and even total daily activity measured by accelerometers.

Thus far, there has been a lack of a clear consensus on specific practical recommendations for PA [8], and our review can help contribute to clearer guidelines for late-life intervention. Although PA methods and thresholds vary, we conclude that engaging in PA at least three times per week for at least 15 min per session can be suggested for the minimum threshold for protection against AD in late life. Studies by both Larson et al. [44] and Luck et al. [45] indicate that engaging in PA at least three times a week, compared to less frequently, offers a significant protective effect. In particular, Larson et al. [44] specified that engaging in activities for at least 15 min at a time, at least three times a week, were associated with lower AD incidence. These findings are in agreement with those of Podewils et al. [48] who reported significant protection for individuals engaging in at least four physical activities in a 2-week period [48]. Although Neergard et al. [47] found that PA at least once a week did not significantly prevent AD, the study did find that PA at least three times a week conferred a significant benefit in preventing all-cause dementia and vascular dementia. Taken together, these studies suggest that a frequency of at least three times a week of PA is likely necessary for meaningful protection against AD development. We acknowledge that a practical threshold of at least 15 min/session at least three times per week is not an absolute threshold and warrants further investigation.

The most recent Physical Activity Guidelines for Americans by the US Department of Health and Human Services recommend a combination of frequency and intensity of PA per week to achieve health benefits. The recommendation for adults suggests engaging in a minimum 150–300 min of moderate-intensity activities each week (such as brisk walking), or 75–150 min a week of vigorous-intensity aerobic exercises each week (such as jogging or running), or an equivalent combination of moderate or vigorous activity, as well as muscle-strengthening exercises at least twice a week [64]. The guidelines are rooted in decades of substantial research aimed at providing broad protection from a large range of health risks, including cardiovascular diseases, several types of cancers, dementia, anxiety and depression, and all-cause mortality [64], 65].

Unfortunately, only about one-quarter of Americans reach the suggested levels of PA [65]. However, there is evidence that even a frequency of three times a week, with even mild to moderate intensity, can provide substantial benefits. A recent study identified that a minimum of 40 min per week of vigorous PA can reduce AD-related mortality, with an optimal duration of 140 min per week [66]. Remarkably, the researchers found that even 40 min a week can prevent 12,238 deaths per year from AD [66]. These findings align with the broader consensus that while 140–150 min of higher-intensity PA is optimal, many older adults can have significant benefits from a lesser amount.

Furthermore, if public health authorities or clinicians can recommend adequate PA as a form of prevention that can be beneficial even over the age of 65, it can encourage those in late life to continue to exercise and adopt a “never too late” mentality on preventing disease [62]. Three of the included studies had an average age of over 80 years and still had a significant benefit [42], 43], 45], with a variety of activities, including household chores [43]. This suggests that routine activities that are inclusive of the capabilities of the older population can have significant benefits. Additionally, given that there is less clarity regarding the relationship between PA and improved cognitive outcomes after developing cognitive decline or dementia [67], 68], the emphasis should shift to utilizing PA as a key preventative therapy. This recommendation can benefit patients at any time, because PA is relatively accessible, safe, and low-cost. Therefore, physicians and health providers may counsel their patients about the many flexible ways that can allow them to accomplish this minimum threshold. With studies suggesting the associations of multiple forms of PA with decreasing AD incidence, patients can engage in activities of their liking, whether it is walking outside, cleaning the house, gardening, swimming, etc., and are still likely reduce their risk of AD. Hopefully, this can encourage patients to engage in PA that is enjoyable to them and empower them to feel that they can do meaningful PA, even if it is not very rigorous.

The use of PA as a means of prevention for AD strongly aligns with the principles of osteopathic medicine. For example, the first Tenet of Osteopathic Medicine states that the body is a unit, and the person is a unit of the body, mind, and spirit. With our systematic review, we can see that physical exercise of the body can influence AD incidence, which is related to cognitive functioning and the mind. Moreover, our findings highlight the significance of preventive care within osteopathic medicine. An increase in PA is a lifestyle change that can greatly reduce the risk of AD, leading to an improved quality of life.

Strengths and limitations

Strengths of the current analysis include the inclusion of studies with large cohorts with sufficient years of follow-up and the completion of bias assessment that ensured the quality of the included studies. Our review was the first to provide practical guidelines for optimal PA levels that can be protective of developing AD in late life.

This present study is not without limitations. Although a meta-analysis may have been performed, there are several reasons that we opted for a systematic review only, mostly due to heterogeneity in analysis, study design, and PA measurement differences. First, each study employed unique parameters for statistical analysis and different reference values, making head-to-head comparisons and collective analysis difficult. Second, while calculated subscores, percentiles, and quintiles are useful for independent analysis, it can be difficult to create guidelines regarding minimum requirements from this type of data. Third, assessment of PA levels also varied, with some studies utilizing accelerometer data, and some utilizing self-report questionnaires or interviews. With the lack of standardized effect sizes or reported metrics (hazard ratios [HRs] vs. odds ratios [ORs], percentiles vs. cutoffs), it would be difficult to perform data harmonization. Additionally, the variables for PA differed across studies, with some utilizing quantitative data (METs measured by accelerometer data), while others utilized categorical data with a different number of subgroups.

Additionally, research could perpetuate biases that might influence outcome measurements. For example, self-report data are susceptible to self-report bias and social desirability bias. While free from self-report bias, the accelerometer could be susceptible to the Hawthorne effect while the patients were monitored. However, overall, the biases of these studies were assessed, and none of them were high risk. Another limitation includes the lack of a longitudinal measure of PA. Most of the studies included in this present review assessed PA at a cross-sectional time point, making it difficult to assess whether subjects were experiencing benefits from PA over a short time period or a cumulative effect over time. It has been stated that the benefit of PA was only associated with the protection of a 4-year follow-up period [69]. Therefore, studies can benefit from more years of follow-up, with levels of PA assessed at more frequent time periods, to provide further insight into the protective benefits.