Mid- and Late-Life Leisure-Time Physical Activity and Global Brain Amyloid Burden: The Atherosclerosis Risk in Communities (ARIC)-PET Study

INTRODUCTION

Reducing the burden of cognitive impairment in the United States aging population is a high priority that may be attainable by intervening on modifiable behaviors such as physical activity. Our prior work in the Atherosclerosis Risk in Communities (ARIC) Study suggests that compared to participants who were physically inactive in mid-life (aged 45–64 years), middle or high levels of leisure-time physical activity were associated with less global and domain-specific cognitive decline and a lower incidence of dementia over 14 years of follow-up [1]. The underlying mechanisms that link physical activity to brain-related outcomes are still unknown but hypothesized to occur through several pathways, including increased neurogenesis [2] and the reduction of vascular and metabolic risk factors, including blood pressure [3], blood glucose levels [4], and systemic inflammation [5].

Mouse models have related physical activity (i.e., wheel running) to lower amyloid levels in the brain [6]. However, the role of exercise on amyloid-β burden in the human brain has not been widely examined. The results to date appear inconsistent, and there have been no reports on physical activity in mid-life at a time when amyloid-β begins to accumulate in the brain [6]. The largest human investigation thus far to examine this association was conducted among 268 elderly French community-dwelling individuals with mild cognitive impairment. In this cross-sectional analysis, self-reported physical activity measured continuously was not associated with brain amyloid burden (Odds Ratio = 1.00, 95% Confidence Interval: 0.99–1.00), based on Florbetapir levels (defined as a standardized uptake volume ratio (SUVR)>1.10) [7]. In another cross-sectional analysis, there was no association between levels of self-reported physical activity and amyloid positron emission tomography (PET) among the total sample of 116 cognitively normal individuals. However, lower levels of amyloid-β were observed in the highest versus lowest tertile of self-reported physical activity, but only among apolipoprotein (APOE) ɛ4 allele carriers [8]. Other prior studies have observed a significant negative association between physical activity and amyloid burden measured with PET, including a report on 60 cognitively normal older adults showing that participants with elevated amyloid had significantly lower reported exercise [9]. An additional report of 317 middle-aged adults from the Wisconsin Registry for Alzheimer’s Prevention found that self-reported physical activity attenuated the adverse effects of age on amyloid burden [10].

Altogether, the current literature on the association between physical activity and brain amyloid burden is constrained by small sample sizes and cross-sectional assessments which are susceptible to a reverse causation interpretation. A further limitation is the lack of repeated measures of physical activity. Considering the variation in physical activity over the adult life span associated with changes in lifestyle and functional abilities, a one-time measurement of physical activity may not be a reliable or informative evaluation of an individual’s activity exposure. Considering the fact that amyloid deposition occurs decades prior to manifest clinical symptoms, quantifying physical activity levels across life epochs with repeated measures is imperative.

Therefore, we examined the association between leisure-time physical activity and PET-quantified brain amyloid burden in the ARIC cohort by incorporating both late-life and repeated mid-life assessments of physical activity with more than 25 years of follow-up. We also examined whether the associations differed by APOE ɛ4 carrier allele status, race-study center, and cognitive status. We hypothesized that higher levels of leisure-time physical activity in mid- or late-life were associated with lower brain amyloid burden in late-life.

MATERIALS AND METHODS

RESULTS

Sociodemographic and clinical characteristics of the study population (n = 326) are provided in Table 1 overall and by reported participation (yes/no) in LTPA in mid-life (visit 1, 1987-1989). Participants who did not participate in LTPA were more often Black and female and had lower educational attainment. Participants not reporting LTPA also had a worse cardiometabolic risk factor profile, including, a slightly higher prevalence of hypertension and smoking and a somewhat higher BMI. Our data suggest a medium/moderate correlation (0.33–0.45) of physical activity (in MET·min·wk^-1) across ARIC visits from mid- to late-life [22]. The frequency of elevated amyloid burden (global SUVR > 1.2) was also higher in participants who did not participate in LTPA. Figure 1 shows the negative linear relationship between total volume of LTPA (MET·min·wk^-1) in mid-life and global SUVR in late-life. At the median SUVR cutoff of 1.2, participants with non-elevated amyloid burden (SUVR ≤1.2) in late-life had on average higher total LTPA in mid-life (381.3 MET·min·wk^-1) compared to participants with elevated amyloid burden (SUVR > 1.2) in late-life (270.4 MET·min·wk^-1).

OPEN IN VIEWER

Fig. 1

Scatterplot of mid-life (visit 1) total leisure-time physical activity in MET·min·wk^-1 and global standardized uptake value ratio measured in late-life (visit 5). SUVR, standardized uptake value ratio.

The prevalence of SUVR > 1.2 in late-life in the analytic sample was 52% (n = 169; Table 1). Participation in LTPA in mid-life was not significantly associated with a lower prevalence of elevated amyloid burden (Odds ratio [OR] = 0.71 (95% confidence interval [CI]: 0.43, 1.18)) compared to non-participation in LTPA in mid-life (Fig. 2). The OR of elevated amyloid burden for those who met the 2018 U.S. Physical Activity Guidelines in mid-life was 0.83 (95% CI: 0.49, 1.42) compared to those who did not meet the guidelines in mid-life. A per standard deviation (SD) higher total volume of LTPA (970 MET·min·wk^-1) in mid-life was also not significantly associated with an elevated amyloid burden OR = 0.89 (95% CI: 0.69, 1.15) in late-life, although each point estimate favored an inverse association between physical activity and amyloid burden. No significant associations were observed for all measures of LTPA in mid- and late-life with global amyloid SUVR analyzed continuously (Table 2) or when using a more conservative global amyloid SUVR cut point >1.3. Specifically, the OR of elevated amyloid burden (based on an SUVR cut point >1.3) for participation in LTPA in mid-life compared to non-participation in LTPA in mid-life was 0.77 (95% CI: 0.44, 1.33).

OPEN IN VIEWER

Fig. 2

Adjusted odds ratio (OR) and 95% confidence interval of elevated brain amyloid burden (standardized uptake value ratio (SUVR) > 1.2) by measures of leisure-time physical activity (LTPA) in mid- and late-life. Models for mid-life and late-life leisure-time physical activity were run separately. *1 standard deviation (SD) = 970 MET·min·wk^-1.

An overall measure of total volume of LTPA in mid-life was estimated by averaging the MET·min·wk^-1 at visits 1 and 3. Results were similar to those observed with total volume of LTPA measured at visit 1 only (data not shown). In summary, regardless of how LTPA was operationalized, mid-life LTPA was not significantly associated with amyloid burden in late- life. For all models considering either mid- or late-life LTPA, we did not observe any significant interactions by race, APOE ɛ4, or cognitive status. Although results were slightly attenuated, the overall inferences were also not affected by additional adjustment for BMI, diabetes, and hypertension (data not shown).

DISCUSSION

In this largest known community-based sample of non-demented older adults with amyloid imaging and repeated measures of LTPA, neither mid- nor late-life LTPA was statistically significantly associated with brain amyloid burden. Our findings fail to support our a priori hypothesis that a mid-life measure of LTPA, free of any question of reverse causality, is significantly associated with brain amyloid burden 25 years later. This result is important because it shows the troublesome independence of amyloid accumulation from the healthy behavior of physical activity.

Our study adds new evidence to the literature relating physical activity to brain amyloid. Although previous findings from human studies have been inconsistent, rodent studies have consistently shown lower brain amyloid burden after intense exercise. Our lack of significant associations is consistent with those of the largest human study to date, based on 268 mild cognitively impaired participants with PET-quantified amyloid-β and self-reported physical activity [7]. This largest study to date raises potential concerns pertaining to its more cognitively impaired participant population, because the cognitive impairment might both affect habitual physical activity levels and reduce the accuracy of reported physical activity. However, the results of studies limited to cognitively normal participants have also been conflicting. In an investigation of 85 late-middle-aged adults from the Wisconsin Registry for Alzheimer’s Prevention, moderate intensity physical activity, ascertained with 1-week actigraphy, was inversely associated with cerebrospinal fluid measured amyloid-β [23]. In contrast, in a study of 139 cognitively normal participants, self-reported physical activity was not associated with cerebrospinal fluid measured amyloid-β in the Dominantly Inherited Alzheimer Network [24]. Using PET-quantified amyloid imaging, the data among cognitively normal participants has also been inconsistent, with some reports showing that higher levels of self-reported physical activity are associated with lower levels of amyloid burden [8–10 , 25]. Conversely, data from 182 clinically normal older adults enrolled in the Harvard Aging Brain Study did not show a statistically significant cross-sectional association between accelerometer-measured physical activity and amyloid burden after adjusting for age and sex [26]. Interpretation of these studies, however, is clouded by the fact that they relate amyloid burden with concurrent levels of physical activity, possibly indicating an effect rather than a cause of the amyloid burden.

There are limitations to the present study. The exposure assessment relied on self-report, which typically overestimates vigorous physical activity and underestimates and/or does not quantify light physical activity [27]. This supports the need for studies to also include device-based assessments of physical activity into data collection protocols. The criteria by which participants were selected for the ARIC-PET study may constrain generalizability. Specifically, participants enrolled in PET studies are those willing and able to tolerate the MRI and PET scans, and therefore may be potentially healthier. Lastly, though the results show a consistent, but not statistically significant, protective association of physical activity on amyloid burden, our study’s size potentially limits its power to detect small differences. A post-hoc power analysis conducted in G*Power 3.1 suggests that for an OR = 0.71 for elevated amyloid burden comparing participants who participate versus do not participate in physical activity, a larger sample size (n > 336) is required to detect a significant association at 80% or greater power and an alpha = 0.05. Another limitation of this analysis is that the outcome cut point was based on a median split of global amyloid PET SUVR which almost certainly includes persons with amyloid burdens that are closer to background levels than to levels that are more concerning. Thus, the elevated amyloid group will contain persons who are not destined to develop overt Alzheimer’s disease. This may have made it more difficult to detect an association with physical activity. However, the same inferences were supported with a more conservative SUVR cutpoint >1.3 conducted in sensitivity analyses. Several strengths should be mentioned. First, the well-characterized ARIC cohort provides over 25 years of collected data, allowing us to examine the role of physical activity across different life epochs on brain amyloid in older adulthood. Second, the cross-temporal analysis (mid-life exposure and late-life outcome) results provide much clearer results than those seen at late-life only due to the potential effects of reverse causation. Lastly, this is the largest known sample of non-demented older adults with amyloid imaging and repeat measures of physical activity.

Given our large sample size, we can reasonably conclude that while the point estimates favored an inverse association of LTPA and brain amyloid burden, the hypothesized association was not supported. Although we found that physical activity is not associated with Alzheimer’s disease pathology, it is possible that physical activity is related to cerebrovascular disease, which has important contributions for later life cognitive impairment. It has been documented that physical activity can reduce the risk of diabetes [4], hypertension [3], and obesity [28]. Reductions in these vascular risk factors may reduce the burden of cerebral atherosclerosis and arteriolosclerosis and subsequently reduce ischemic cerebrovascular pathology [29]. In this setting of less cerebrovascular pathology, individuals may be more tolerable to higher burdens of Alzheimer’s disease pathology. Therefore, it is important that future studies quantify the role of physical activity and its changes over time with measures of cerebrovascular disease.

The long-term impact of mid-life leisure-time physical activity as a modifiable element of lifestyle is of considerable interest, both in clinical settings and from a public health perspective. Pharmaceutical therapies aimed at targeting clinical symptoms of AD and underlying AD pathophysiological processes have not been shown to be effective thus far. Therefore, there is a greater need to understand how lifestyle interventions may not only affect AD symptoms but its pathophysiology as well. Randomized clinical trials of a physical activity lifestyle intervention have not shown clear benefits [30 –32], and implementing sustained modification of long-term physical activity patterns has proved to be challenging. This is particularly important in the context of amyloid-β accumulation, which is known to progress for decades prior to the appearance of clinical manifestations. Therefore, observational data with prolonged follow-up, and repeated measures of physical activity that are preferably objectively-measured, are still needed to help elucidate the potentially beneficial long-term role of higher physical activity levels in reducing or slowing the accumulation of brain amyloid in adulthood.