Therefore, the aim of this cross-sectional study was to examine associations between self-reported physical activity carried out in midlife and late life and self- and informant-reported neuropsychiatric symptoms in a sample of community-dwelling older adults who were either cognitively unimpaired or had mild cognitive impairment (MCI). We hypothesized that physical activity would be associated with a lower likelihood of having neuropsychiatric symptoms and lower scores on continuous measures of neuropsychiatric symptoms.
Results
The study included 3,222 individuals (N=1,655 men) ≥70 years of age (mean±SD=79.2±5.6); 2,723 of these individuals were cognitively unimpaired, and 499 individuals had MCI. The mean midlife physical activity score was 8.93±4.63, and the mean late-life score was 6.54±3.90. The presence of neuropsychiatric symptoms ranged from 12% for depression to 3% for agitation; 7% of participants had clinical depression (i.e., BDI-II score ≥13), and 7% had clinical anxiety (i.e., BAI score ≥10). Women had significantly fewer years of education; lower body mass index; lower Charlson index scores; lower frequency of cognitive impairment; lower frequency of agitation, apathy, appetite or eating changes, nighttime disturbances, and irritability; and higher BAI total scores. The demographic and other characteristics of the study sample are shown in
Table 1.
Logistic regression analyses revealed statistically significant inverse associations between engagement in late-life physical activity and the presence of neuropsychiatric symptoms: each one-unit increase in late-life physical activity was associated with lower odds of having apathy (OR=0.89, 95% CI=0.84–0.93, p<0.001), appetite or eating changes (OR=0.92, 95%=0.87–0.98, p=0.006), sleep or nighttime disturbances (OR=0.95, 95% CI=0.91–0.99, p=0.024), depression (OR=0.94, 95% CI=0.90–0.97, p<0.001), irritability (OR=0.93, 95% CI=0.89–0.97, p=0.001), clinical depression (OR=0.92, 95% CI=0.88–0.97, p=0.001), and clinical anxiety (OR=0.90, 95% CI=0.86–0.94, p<0.001). We did not observe any significant associations between midlife physical activity and neuropsychiatric symptoms. The logistic regression models of the associations between midlife and late-life physical activity and neuropsychiatric symptoms are shown in
Table 2.
Linear regression analyses showed that higher levels of late-life physical activity were associated with less depressive and anxiety symptom severity as indicated by lower BDI-II (β estimate=−0.042, 95% CI=−0.051 to −0.033, p<0.001) and BAI (β estimate=−0.030, 95% CI=−0.040 to −0.021, p<0.001) total scores. Higher levels of midlife physical activity were associated with higher BDI-II total scores (β estimate=0.011, 95% CI=0.004 to 0.019, p=0.004). Because the BDI-II and BAI scores were log-transformed, we can approximately interpret the results in terms of percentages: each one-unit increase in late-life physical activity was associated with an approximately 4.2% decrease in BDI-II score, on average. The linear regression models of the associations between midlife and late-life physical activity and BDI-II and BAI scores are shown in
Table 3.
To examine whether sex modifies the association between midlife and late-life physical activity and neuropsychiatric symptoms, we ran models that included an interaction term with sex. We observed two statistically significant interactions; these interactions included physical activity in midlife predicting the presence of anxiety and physical activity in late life predicting clinical anxiety. Thus, we stratified these analyses by sex and found that higher levels of midlife physical activity were statistically significantly associated with lower odds of the presence of anxiety in women (OR=0.92, 95% CI=0.86–0.98, p=0.013) but not in men (OR=1.03, 95% CI=0.98–1.08, p=0.299). Similarly, higher levels of late-life physical activity were associated with lower odds of clinical anxiety in women (OR=0.84, 95% CI=0.78–0.90, p<0.001) but not in men (OR=0.97, 95% CI=0.90–1.03, p=0.284).
Discussion
Engagement in physical activity in late life was associated with a lower likelihood of having neuropsychiatric symptoms in our sample of community-dwelling older adults free of dementia. In line with our hypothesis, we observed significant inverse associations between late-life physical activity and apathy, appetite or eating changes, sleep or nighttime disturbances, depression, and irritability (all assessed by the NPI-Q), as well as associations with clinical depression and anxiety (assessed by the BDI-II and BAI, respectively). Because it is not possible to infer cause and effect based on our cross-sectional study design, there are two potential explanations for our findings: engaging in physical activity in late life may decrease the risk of having neuropsychiatric symptoms, and participants who have neuropsychiatric symptoms may be less likely to engage in late-life physical activity.
For women, there were significant associations between higher levels of midlife physical activity and a lower likelihood of the presence of anxiety and higher levels of late-life physical activity and a lower likelihood of having clinical anxiety; these associations did not hold for men. These findings may indicate that the associations between physical activity and anxiety are modified by sex; more research is needed to examine the potential impact of sex on the associations between physical activity and neuropsychiatric symptoms.
Furthermore, there was a significant association between higher levels of midlife physical activity and higher BDI-II scores. We conducted additional analyses to further explore this association (data not shown) and found that particularly vigorous midlife physical exercise in women was associated with higher depressive symptom severity. In this study, as noted above, midlife was defined as 50 to 65 years of age, which coincides with menopause and associated mood changes. We and others have reported that menopause is associated with depression (
34). Some women might be using vigorous physical exercise as a nonpharmacological way of dealing with depressive symptoms associated with physiological changes in midlife. Future studies are warranted to explore the potentially different associations between midlife and late life physical activity with neuropsychiatric symptoms by accounting for potential sex effects. In addition, future research should investigate the potential impact of social determinants of health factors on the association between physical activity and neuropsychiatric symptoms.
Overall, the findings from our study are in line with previous research. For example, investigators from the Netherlands have reported that adults >60 years of age with depression were less physically active than peers without depression based on cross-sectional data (
35). In addition, a recent review concluded that there is an inverse relationship between physical activity and depression in older adults, although it was acknowledged that the dose-response relationship between physical activity and depression, as well as the potentially different effects of various types of physical activity, remain unknown (
36). Another review of clinical trials and cohort studies also showed that aerobic and resistance training are effective in decreasing symptoms of depression in older adults (
37). On the basis of these findings, it has been postulated that physical activity may affect biological and psychosocial processes that are implicated in the pathophysiology of depression. For example, physical activity may have effects on neuroplasticity (e.g., increased hippocampal neurogenesis), inflammation, oxidative stress, the endocrine system (e.g., cortisol), self-esteem, social support, and self-efficacy (
38,
39).
With regard to anxiety symptoms, a recent review of 24 prospective studies revealed a potentially protective effect of physical activity on anxiety symptoms and disorders (
40). Similar to the aforementioned review on depression, the authors concluded that the data were not sufficient to make conclusions about the dose-response relationship between physical activity and lower anxiety symptoms; furthermore, the use of different methods to assess physical activity may limit the comparability of studies. Evidence of physical activity reducing neuropsychiatric symptoms can also be drawn from interventional research. For example, a home-based, 6-month physical activity program was effective in reducing depression and anxiety, particularly in persons with elevated anxiety and depressive symptoms at baseline (
41). In addition, there is a rather large body of research on the associations between physical activity and sleep. For example, a randomized controlled trial showed that a 4-week, low-impact walking intervention was associated with better sleep quality in healthy adults (
42). Furthermore, a longitudinal study derived from the Wisconsin Sleep Cohort showed that an intermediate level of physical activity, defined as 500–1,500 metabolic-equivalent-of-task minutes per week, was associated with a lower risk of short sleep times (
43). Furthermore, investigators from Spain reported that higher levels of objectively measured physical activity were associated with better subjective sleep quality and quantity (
44).
In our study, we did not examine the association between changes in physical activity levels between midlife and late life and neuropsychiatric symptoms. However, this potential association may be interesting with regard to the linear model predicting BDI-II scores, as midlife physical activity was not statistically significant for any other model, and the result that higher levels of midlife physical activity were associated with higher BDI-II scores seems counterintuitive. Based on the model we report in
Table 3, we can say that participants with low midlife and high late-life physical activity will have the lowest BDI-II scores, and those with high midlife and low late-life physical activity will have the highest BDI-II scores, on average. In an additional analysis (data not shown), we created a four-level physical activity variable by using median splits of the midlife and late-life activity data: low in midlife and low in late life, high in midlife and low in late life, low in midlife and high in late life, and high in midlife and high in late life. Using the participants with low activity in both midlife and late-life periods as the reference, those who went from high to low levels of activity were not statistically different from those who always had low levels of activity; those who went from low to high levels of activity had the lowest BDI-II scores (β estimate=−0.258, 95% CI=−0.351 to −0.166, p≤0.001); and those who maintained high levels of activity also had lower BDI-II scores (β estimate=−0.220, 95% CI=−0.289 to −0.151, p≤0.001). Although these results were not completely consistent across models, it seems likely that a change from high levels of midlife physical activity to low levels of late-life physical activity would be associated with a greater likelihood or severity of neuropsychiatric symptoms.
The strengths of our study are the large sample size and the rigorous assessment of self-reported and informant-observed neuropsychiatric symptoms. In addition, we focused on both midlife and late-life physical activity as presumed predictor variables. To our knowledge, our study is one of the few that has examined the association between engaging in physical activity in different periods of life and the presence of neuropsychiatric symptoms. Furthermore, we examined potential sex interactions, given that previous research indicated that the association between physical activity and neuropsychiatric symptoms may be moderated by sex (
42). Finally, to adjust physical activity in midlife for physical activity in late life and vice versa, we included physical activity composite scores from both time periods in all models. We made this adjustment because it is likely that physical activity habits in late life are influenced by those in midlife. The main limitations of our research pertain to the cross-sectional study design, which does not allow for drawing conclusions about cause and effect in the associations between physical activity and neuropsychiatric symptoms. Furthermore, physical activity was assessed with a self-report questionnaire, which may be prone to recall bias, and reliability was only low to moderate. Thus, use of body-worn sensors should be considered in future research to ensure objective assessments of physical activity and overcome the risk of bias from self-reported data. In addition, we did not differentiate between physical activity and exercise in the composite score, and we did not weigh the scores on the basis of degree of intensity. However, as noted above, we conducted additional analyses by including the three exercise intensities (i.e., light, moderate, and vigorous physical exercise in midlife and in late life) separately in the models to further clarify rather unexpected findings (e.g., when we observed a significant association between higher levels of midlife physical activity and higher BDI-II scores). Furthermore, we did not adjust our analyses for multiple comparisons, which may have increased the possibility of a type I error. However, when considering a Bonferroni correction for our analyses, the alpha significance level would be 0.006 (i.e., 0.05/9) because we had a total of nine neuropsychiatric symptoms per predictor (i.e., levels of midlife and late-life physical activity) in
Table 2. In this situation, six out of the seven significant p values for the association between late-life physical activity and neuropsychiatric symptoms would have remained significant, and none of our major conclusions would have been affected by the correction. As expected in our population-based sample, the numbers of participants with depression and anxiety as measured by the BDI-II and BAI were low, and this may have influenced the conclusions. Finally, our sample is relatively highly educated, and 99% of the study participants were White. However, it has been shown that data from Olmsted County are generalizable to the U.S. population of Minnesota and the upper Midwest (
45). Nevertheless, the data may not be generalizable to minority populations such as non-White race or ethnicities. This lack of cultural and linguistic sensitivity may have also had an impact on the study findings. Thus, we acknowledge that future studies in populations with more racial and ethnic diversity are needed to confirm our findings.