Physical fitness is associated with higher global cognition in individuals with dementia, independent of past levels of exercise and illness duration.

Dementia is one of the leading causes of disability in later life, ahead of cancer, cardiovascular disease, and stroke (1) and is characterized by a decline in many aspects of cognitive ability. Despite promising new research (2), there is currently no pharmacological cure for dementia (3). It is, therefore, important to help identify behavioral and environmental modifiable risk factors that could reduce the progression of the disease. One such risk factor is physical activity (PA), which is defined as any bodily movement produced by skeletal muscles that result in energy expenditure. PA differs from physical fitness (PF), which represents a set of attributes that people have or achieve that relates to the ability to perform PA.

Current recommendations from the World Health Organization (4) suggest that throughout the week, older adults (age 65+) should perform either at least 150 min of moderate-intensity aerobic PA or 75 min of vigorous-intensity aerobic PA. The National Institute for Health and Care Excellence recommends that adults increase their PA levels to reduce the risk of dementia (5), but does not recommend guidelines for the amount of exercise recommended for people with dementia.

Previous research has suggested increased levels of PA as an intervention for maintaining cognitive performance in older adults (6, 7); for reviews, see Ref. (8, 9). Results of a meta-analysis of 29 randomized control trials (RCTs) found that sustained PA was associated with improved scores in memory, attention, and processing speed in both healthy adults and those with mild cognitive impairment (MCI) (10). Similarly, a recent cross-sectional study of 7,000 older adults found that higher levels of objectively measured PA were associated with better performance in memory and executive function and lower prevalence of cognitive impairment (11). However, not all research has supported this association. For example, Young et al. (12) reviewed 12 RCTs of 750 participants and found that aerobic PA to have no cognitive benefit in cognitively healthy older adults.

The effect of PA has also been investigated in people with dementia, with a substantial body of research reporting a positive association between increased levels of PA and improved cognition [for reviews, see Ref. (13–16)]. A RCT of 185 older adults with dementia found a significant effect of PF on dual-task performance (e.g., 10-m walking while counting backwards from 50 or naming the months backwards) and other tests of cognition [e.g., Mini-Mental State Examination (MMSE), Symbol Digit Modalities Test, the Stroop Colour and Word Test, and Lexical Verbal Fluency Test] (17). However, not all research has supported this positive effect. Forbes et al. (18) found no significant impact of exercise programs on improving cognitive function in a review of 17 RCTs studying 1,067 older adults with dementia, but did notice improvements in activities of daily living. Beyond improving outcomes for people with dementia, a systematic review of 26 studies reported a 14% reduction in dementia risk for those with high PA levels (RR 0.86, 95% CI 0.76–0.97) compared with those inactive (19), see also Ref. (20–22).

The overall findings suggest an advantage of PA on cognitive ability even in individuals with dementia, but some of the present inconsistencies indicate a highly complex interaction between the level of PF, the type of PA, and the neuropsychological domains measured in various studies. In particular, studies found selective improvements in executive function rather than general cognitive benefits due to PA (23–25). Therefore, it is recommended to use a battery of neuropsychological tests rather than a single measure of cognition (14, 26). For example, in a study that used multiple cognitive tests in oldest-old women (self-reported), PA was associated with better scores in the modified MMSE, category fluency, and Trails B, but not for digit span backwards, phonemic fluency, and delayed recall at 5 years follow-up (27). In another study, Matthews and Williams (28) found that Tai Chi was effective at improving performance in the Trails B and clock drawing tests, but not in the digit-symbol substitution test. Interactions between age and PA were also found for measures of immediate memory and visuospatial ability, and a trend for verbal ability (29).

Although most of the aforementioned work has focused on PA, there is less research addressing the role of PF in cognitive performance in dementia. A review of 16 intervention studies revealed improvements in physical functioning irrespective of the stage of dementia (30). For example, Santana-Sosa et al. (31) demonstrated improved PF after a 12-week exercise intervention period, but they did not assess cognitive function. However, Huh and colleagues (32) observed a cross-sectional association between executive functioning and PF, regardless of PA level. As of yet, there are no studies looking at both PA and PF on cognitive performance in dementia. Given the importance of ascertaining the potential differential contributions of PA and PF, more research is necessary to evaluate these specific associations with cognitive functioning, particularly among people with dementia.

Table 1 presents descriptive statistics for the 70 participants grouped according to their health status (healthy and dementia) and PF level. Healthy controls had significantly higher levels of current PF than the dementia group. Individuals with dementia were 5.4 years older than the healthy group. There were more women with low PF compared to men. Interestingly, high levels of past exercise were associated with high PF but were not related to health status.

Table 2 presents the means and SDs of the test scores separated by PF level and health status.

A MANCOVA was conducted to assess the effects of PF and health status on the dependent measures, including as covariates age and previous levels of exercise, as these were the only demographic variables that showed at least one significant univariate test result (see Table S1 in Supplementary Material). The pattern of results remained the same when these covariates were excluded from the analysis.

There were significant multivariate effects of health status [Pillai’s Trace¹ = 0.68, F(10, 55) = 11.58, p < 0.001, observed power = 1.00] and level of PF [Pillai’s Trace = 0.31, F(10, 55) = 2.47, p < 0.05, observed power = 0.91], and an interaction between these factors [Pillai’s Trace = 0.28, F(10, 55) = 2.13, p < 0.05, observed power = 0.85]. To investigate these effects, discriminant analysis (method = enter) was used in which the combined set of 10 measures predicted the health status by the level of PF.

The overall test was significant (Wilks λ = 0.15, χ² = 119.42, df = 30, canonical correlation = 0.87, R² = 0.75, p < 0.001). Reclassification of cases based on the canonical variables was successful: 71.4% of the cases were correctly reclassified into their original categories, but 54.3% of the cross-validated cases were correctly classified. The misclassifications were predominantly due to dementia cases with high PF being misclassified as healthy (45% compared to 16% for low PF dementia cases), whereas the reverse misclassification was less likely (7% for low PF healthy and 4% high PF healthy).

The analysis extracted two functions (eigenvalues: 3.045 and 0.455), accounting for 95.4% of the variance in cognitive performance, with the first function accounting for 83%. This function was labeled “Global cognition,” as for all dependent measures, better scores increased the value from the discriminant function. Figure 1 shows the resulting values from function 1 against the reference of the minimal value (=−4.14) in the sample. The interaction observed from the MANCOVA reflects better global cognition for dementia cases who are physically fit compared to those who are not, combined with a lack of such an effect for the healthy controls. Using fitted values for each individual, a two-way ANOVA crossing health status and PF revealed main effects of status [F(1, 66) = 145.42, p < 0.001, partial η² =0.69, observed power = 1.00] and PF [F(1, 66) = 11.69, p = 0.001, partial η² = 0.15, observed power = 0.92] and an interaction [F(1, 66) = 8.58, p < 0.01, partial η² = 0.12, observed power = 0.82], due to a significant simple effect of PF for dementia cases [t(28) = 3.18, p < 0.01], but not for healthy controls (p > 0.90). These results remained significant when including age, past exercise, and time since diagnosis (for the dementia group) as covariates.

Whereas the above two-way ANOVA takes into account all dependent measures in a linear-weighted manner, univariate analyses on each measure (corrected for multiple testing) lack the necessary univariate power. The pattern of results differed across dependent measures (see Table 2). All dependent measures revealed better performance for healthy individuals compared to dementia patients (all p < 0.001). Only MMSE and clock drawing, exhibited better performance for high compared to low PF. There were interaction effects just for clock drawing, with larger effects of PF for those with dementia in contrast to the control group when taking into account age, past exercise, and years since diagnosis as covariates.

This study explored the influence of PF on cognitive functioning in participants with dementia and healthy controls. As expected, those with dementia showed lower levels of cognitive functioning than healthy group. Of interest are the findings that current PF levels were found to have a significant overall effect on cognition specifically on MMSE and clock drawing.

The association between PF and MMSE is an interesting finding, given that previous research into the effects of PA on MMSE in people with dementia and MCI has provided inconsistent results. For example, whereas some RCTs (42–45) show positive effects of PA on MMSE, Venturelli et al. (46) did not find a strong advantage for their walking intervention. However, the decline in MMSE scores was lower in the intervention group compared to the control group, suggesting a relative attenuation of decline with exercise, although the interaction did not reach significance after correcting for multiple testing.

Performance on the clock drawing task was significantly affected by PF levels in dementia participants but not in control participants. The abilities required for clock drawing depends on executive functions, such as visuoconstructive abilities, spatial function, planning, abstract thinking, and attention (47). This research supports the hypothesis proposed by Kramer et al. (24) that executive control processes showed the most considerable benefit of improved fitness [see also Ref. (23, 25, 48)].

Only clock drawing revealed a reliable interaction between health status and PF, highlighting different sensitivities of the tests to the combination of health status and PF level. The reason for this may be related to the requirement of visual processing which has been shown to be particularly sensitive to dementia-related cognitive decline [e.g., Ref. (49)].

With regard to the demographic information, although high levels of past exercise were associated with high PF, it was not related to health status. This suggests that high levels of PA may not in and of itself reduce the incidence of dementia, but when dementia does occur a history of high levels of physical exercise may limit the amount (or perhaps the rate) of cognitive deterioration. With a cross-sectional design, this study is unable to address this pattern of decline in more detail but provides a direction for longitudinal analyses.

Physical fitness is associated with higher global cognition in individuals with dementia, independent of their levels of past exercise and illness duration. This research finds significant positive effects of PF on some, but not all cognitive tests in individuals with dementia. Results from this study support the WHO recommendation for increased PF for older adults in general and those with dementia in particular. Future research should focus on how this can be made accessible to older adults who may have mobility issues, physical disabilities, and cognitive impairment.

This study was carried out in accordance with the recommendations of the UCL Research Ethics Committee with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki. The protocol was approved by the UCL Research Ethics Committee (Project ID Number: 6439/003).

DC and AH designed the study and collected the data. DC, AH, and ED analyzed the data and wrote the paper.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The reviewer LV and handling editor declared their shared affiliation.