Midlife is good for more than a crisis: Exercise for dementia prevention

Abstract

Alzheimer's disease (AD) will have touched the lives of many readers, as nearly 7 million older Americans currently live with the disease.¹ While there has been recent progress toward disease-modifying treatment, these medications have limited efficacy, carry known risks (e.g., brain bleeding and brain swelling), and are expensive.^{2, 3} Given the current Alzheimer's landscape, identifying effective ways to prevent dementia is of paramount importance. Engaging in physical activity, or exercise, is a widely accepted strategy for dementia prevention.⁴ However, questions about the causal nature of the relationship remain. While epidemiological and cohort studies have consistently shown benefits of exercise for reducing dementia risk,^5-7 evidence from randomized controlled trials (RCTs), the gold standard for assessing causality, is a mixed bag.^{8, 9} Indeed, trials provide inconsistent results on the benefits of exercise, possibly due to difference in exercise duration, modality, and intensity, in addition to trial design and participant characteristics.^8-11

A new study by Wei and colleagues hones in on one key feature: trial duration. They take the view that the benefits of exercise require several years to be realized and are thereby missed by RCTs with relatively short follow-up periods. Target trial emulation, a statistical analysis technique used to estimate the effectiveness of a hypothetical intervention using observational data, can be used to evaluate changes over a much longer time period, 12 years in fact in the present study. Through adjustment for confounding variables and accounting for loss to follow-up and censoring, this type of study can limit the effects of confounding and selection bias, thereby increasing the reliability of the estimated causal effect.

The study sample consisted of 1505 participants aged 45–65 years from the nationally representative Health and Retirement Study (HRS) (average age 57.6 ± 4.8 years, 67% women, 76.5% White). The objective of the study was to measure longer term effects of moderate to vigorous physical activity on three main outcomes: cognitive status, conversion to cognitive impairment and dementia, and mortality over 12 years, a duration that would be impossible in a RCT design. At baseline, all participants were cognitively unimpaired and self-reported as physically inactive for the previous 2 years. Cognitive status was assessed using Langa–Weir classification, which categorized participants as “normal cognition,” “cognitively impaired but no dementia,” or “dementia.” For outcome measures, both “cognitively impaired but no dementia,” and “dementia” were considered cognitive impairment. Participants were categorized as initiating physical activity if they self-reported engaging in moderate or vigorous activity at least twice per week at the next wave of data collection (occurring 2 years after baseline). The authors then conducted intention-to-treat (ITT) and per-protocol analyses. ITT analysis included all individuals who initiated physical activity after being inactive at baseline, even if they self-reported as physically inactive at a later wave. Per-protocol analysis included only individuals who sustained moderate or vigorous activity at least twice per week across all study waves.

Of the 1505 study participants, there were 72 incidents of dementia and 409 incidents of cognitive impairment (which included those categorized as “cognitively impaired but no dementia,” or “dementia”) over the 12-year follow-up period. For the ITT analysis, there was a 30% relative risk reduction of dementia but no statistically significant reduction in cognitive impairment over the 12-year follow-up. For the per-protocol analysis, the relative risk reduction of dementia was 49% while the relative risk reduction of cognitive impairment was 23%. Of note, when stratified by sex, the benefits of physical activity for reduced dementia risk persisted in women but not in men. In the ITT analysis, dementia risk was reduced in women after 4, 6, and 10 years but was not observed at 12 years of follow-up. In the per-protocol analysis, dementia risk was reduced in women at each follow-up period. In stratified analyses with cognitive impairment as the outcome of interest, sex differences were reversed. In men, physical activity reduced the risk of cognitive impairment after 2, 4, 6, and 12 years of follow-up in the ITT analysis, and after 10 and 12 years of follow-up in the per-protocol analysis. In additional analyses using composite outcomes of incident dementia and death, and incident cognitive impairment and death, physical activity initiated in midlife resulted in significant protection in the per-protocol analysis, and the effect was more pronounced in women. When the groups were restricted to later-life participants over 65, there were no significant reductions in risk except for the composite outcome. Given these findings, the authors conclude that physical activity initiated during midlife appears to significantly reduce the risk of dementia and cognitive impairment.

The major strength of this study is the ability to test causal hypotheses that would be impossible within a real-world RCT due to cost and feasibility. Measuring the causal effects of moderate or vigorous intensity physical activity with a sample of more than 1000 individuals for 12 years provided the necessary follow-up time to observe benefits to cognition, dementia risk, and mortality. Although the technique of target trial emulation provides a strong causal framework, some limitations remain. Despite the use of inverse probability of treatment weight to balance participant characteristics across initiators and noninitiators, nonrandomization remains an inherent limitation. For example, it is possible that there were confounding variables (e.g., social support, genetics) that differed between initiators and noninitiators but were not accounted for. This challenge is also true for RCTs when imbalanced confounders exist between arms. A second limitation is that physical activity was characterized using a self-report questionnaire in which participants were asked about their engagement in vigorous (e.g., running, swimming, cycling) and moderate (e.g., gardening, walking, dancing) physical activity, with choices consisting of “every day,” “more than once a week,” “once a week,” “one to three times a month,” or “hardly ever or never.” Granular details such as the type, intensity (i.e., moderate versus vigorous), duration, or regularity of physical activity are unknown.

The findings by Wei and colleagues contribute to a literature burdened with mixed reports on the causal relationship between exercise and dementia risk. Cohort and case–control studies have consistently provided compelling evidence for an association between physical activity and brain health, with systematic reviews and meta-analyses concluding that physical activity is associated with lower risk of all-cause dementia and AD⁷ and cognitive decline and dementia.⁶ Data from RCTs have also reported favorable outcomes, including a pivotal study conducted in participants with subjective memory problems aged 50 and above, which reported an improvement in cognition (as measured by the Alzheimer Disease Assessment Scale- Cognitive Subscale) after a 6-month home-based moderate aerobic physical activity intervention.¹⁰ However, results from other RCTS have been less favorable. A recent 5-year RCT investigating the effect of moderate- or high-intensity interval training two times per week on cognition (as measured by the Montreal Cognitive Assessment) in older adults (mean age at baseline = 78.2 ± 2.02) also found no significant differences between the control and exercise groups.¹² Subgroup analyses revealed men in the exercise group performed better on a test of global cognition and had 32% lower risk of MCI, but no significant effects were observed in women. This type of equivocal evidence mirrors conclusions of a systematic review of RCTs in cognitively unimpaired participants aged 55 and above,⁹ which issued caution regarding the causal link between exercise and cognition. The light at the end of the tunnel is that the present study offers a possible explanation for the discrepancies observed in the literature, that is, RCTs may initiate exercise too late in life, and/or do not have long enough follow-up periods to observe the protective effects of the intervention.

How do these findings affect research, public policy, and clinical care? In research, replication is key: Target trial emulation design should be used to validate these findings in other large, representative cohorts with long follow-up time periods. Ideally, other datasets may offer more objective or detailed characterizations of physical activity engagement. These findings underscore the importance of public policy aimed at increasing engagement in physical activity during midlife. This could range from increasing access to physical activity (e.g., more green spaces¹³) to the implementation of walking groups in community and/or work settings.^{13, 14} Within clinical care, querying about physical activity engagement should be standard practice with actionable, realistic, and personalized recommendations. This patient-specific prescription needs to be followed and supported for accountability.¹⁵ Overall, a focus on precision prevention would greatly facilitate the development of precise, evidence-based recommendations to the public. Amidst growing evidence for protective effects of physical activity on brain health, the natural next questions about factors such as what type of physical activity (e.g., aerobic exercise), when to exercise (e.g., midlife), and how does it confer benefits (e.g., mechanisms such as BDNF, synaptic plasticity)?^{16, 17} continue to persist (Figure 1). While pursuing these questions, however, do not let “perfect be the enemy of good.” Striving for optimal recommendations should be balanced with efforts to meet people where they are.

LF and JP contributed equally to the conception and writing of this editorial.

The authors declare no conflicts of interest.

The authors have no sponsor role to report.