Sample characteristics are displayed in . We analyzed data from 187 individuals with a diagnosis of AD (87.2% with AD only, 8.6% with AD and VaD, and 4.3% with AD and another form of dementia). The sample was predominantly Caucasian (99%) and female (64.7%). Males had completed more years of education (M = 13.9, SD = 3.3) than females, M = 12.8, SD = 2.6; t(111) = −2.3, p = 0.02. Mean age of AD onset was 82.4 (SD = 6.0), and severity of dementia at diagnosis was generally mild with mean CDR of 1.4 (SD = 0.9). Mean duration of dementia at the initial DPS visit was 4.0 (SD = 2.0) years. Most participants (84.4%) were rated as either in good or in better physical health on the GMHR.
Sample Characteristics at the Initial Dementia Progression Study Visit
Participants in these analyses were examined up to five times, including the initial DPS visit. Of the original 187, 66 (35.3%) completed the fifth visit. Reflecting the ongoing status of the work, an additional 25 participants are pending their fifth assessment. Reasons for losses to follow-up included death (n = 89), moving (n = 1), and refusal (n = 6). Individuals who were pending or who dropped out prior to the final visit were older, M = 87.5, SD = 5.4 for drop-out/pending vs. M = 84.5, SD = 5.9 for completers, t(185) = 3.5, p = .001; in worse physical health (19.2% of pending/drop-out were in fair to poor health vs. 9.2% of completers, χ2(2, N = 185) = 7.7, p = .02; were less physically active, M = 8.3, SD = 16.0 for drop-out/pending versus M = 16.2, SD = 26.0 for completers, t(90) = −2.2, p = .03; had completed fewer years of formal education, M = 12.8, SD = 2.8 for drop-out/pending versus M = 13.8, SD = 3.0 for completers, t(185) = −2.2, p = .03; and engaged in fewer cognitively stimulating activities, M = 3.6, SD = 2.8 for drop-out/pending versus M = 5.2, SD = 3.0 for completers, t(185) = −3.5, p = .001.
At the initial DPS visit, most individuals (87.2%) reported weekly or daily participation in at least one cognitively stimulating activity. A majority reported two or more activities, whereas 29.4% reported engaging in over five. Number of such activities declined with follow-up, from an initial mean of 4.0 (SD = 3.0) to 2.4 (SD = 2.0, t(65) = 8.4, p < .0001) at the final visit. As expected, moderate correlations were found between number of cognitively stimulating activities and dementia duration at each visit, such that increasing duration was associated with fewer activities. For instance, at Visit 1, the association between dementia duration and cognitive activity at the initial DPS visit was −0.34 (p < .0001). At Visit 4, the correlation was −.47 (p < .0001). Initially, mean physical activity totaled 11.2 (SD = 20.5) hrs per month, but this figure also declined over time to 5.8, SD = 11.1, t(60) = 3.9, p < .0001, hrs per month at the last visit, an average of 2.7 years later. Unlike cognitive activity, however, hours of physical activity and dementia duration were not significantly correlated, r(173) = −0.11, p = .16. Participants were relatively highly educated with a mean (SD) of 13.2 (2.9) years.
Prior to modeling the association between engagement in cognitively stimulating activities, other indicators of cognitive reserve, and the other covariates, we examined their correlations with each other as a check for collinearity. Although education was statistically correlated with premorbid IQ, r(187) = .36, p < .0001, and with occupational attainment, r(187) = .40, p < .001, and number of cognitive activities was statistically correlated with occupational attainment, r(187) = .16, p = .037, the magnitude of each association was low, suggesting that each variable would add unique information when added to the linear mixed models.
Lower cognitive activity was associated with increasing duration of dementia, r(180) =−.33, p < .001, and varied by sex, with men engaging in significantly more cognitively stimulating activities than females, t(185) =−2.83, p = .005. Number of cognitive activities was negatively associated with age r(187) = −.18, p = .01, and positively associated with overall health, r(185) = .19, p = .01, and physical activity, r(173) = .29, p < .001. APOE status was not significantly associated with cognitive activity, M = 4.6, SD = 3.2 for APOE E4+ vs. M = 3.8, SD = 2.7 for APOE E4-, t(185) =−1.84, p = .07.
Association of Cognitive Activities, Other Indicators of Cognitive Reserve, and Rate of Cognitive Decline
As expected, in linear mixed models significant negative estimates for time indicated deterioration in MMSE scores. A significant three-way interaction between number of activities, dementia duration, and time observed in initial models remained significant with the addition of covariates. In the final model, a significant three-way interaction (F = 5.5, p = .02) indicated a differential effect for stimulating activity on decline in MMSE performance depending upon duration of dementia at the initial DPS visit. Specifically, with shorter illness duration, increased engagement in cognitively stimulating activities was associated with a slower rate of decline. For example, the model predicted that for those with less than one-year dementia duration, engagement in zero activities would be associated with a decline of 3.9 points per year on the MMSE (), whereas those engaged in five stimulating activities would experience a decline of 2.2 points per year, a difference of 1.7 points. By contrast, among those with a three-year dementia duration, the model predicted a decline of 2.2 points per year for those engaged in zero activities and a rate of decline of 2.1 points per year for engagement in five activities ().
Figure 1. (a) and (b) The graphs display the predicted decline in Mini-Mental State Examination (MMSE) scores for varying numbers of cognitively stimulating activities by dementia duration at Visit 1. Note that the graph is based on the regression equation for (more ...)
Estimate of premorbid IQ on average was associated with higher MMSE score (Estimate = 0.11, p = .02) but not with rate of decline (Estimate =−0.01, p = .62). Occupational attainment (Estimate = 0.15, p = .55) and years of education (Estimate = 0.09, p = .61) were not significantly associated with MMSE. However, these were retained in the final model because of their theoretical importance as indicators of cognitive reserve. Engagement in physical activity was not associated with cognitive performance (Estimate = −0.01, p = .54). The results of the final model are displayed in .
Effect of Engagement in Cognitively Stimulating Activities and Other Indicators of Cognitive Reserve on Mini-Mental State Examination Scores: Results of the Final Model
Association of Cognitive Activities, Other Indicators of Cognitive Reserve, and Rate of Functional Decline
Linear mixed models of the CDR-SB yielded a significant positive estimate for time, indicating that functional impairment increased over time. Models examining the effects of covariates and their interactions with time also revealed a significant main effect for GMHR rating, indicating better functional ability with better physical health. The final model, controlling for physical health, revealed a significant interaction between dementia duration at the initial visit and number of cognitive activities (F = 7.5, p = .006). Specifically, a higher number of cognitive activities was associated, on average, with better functional performance, particularly with longer dementia duration. For every 5-unit increase in dementia duration and cognitive activity, there was a corresponding reduction of 0.5 points on the CDR-SB. depicts the relationship between cognitive activities and dementia duration on functional ability. Engagement in physical activity (Estimate = 0.01, p = .24), estimate of premorbid IQ (Estimate =−0.05, p = .15), education (Estimate = 0.06, p = .61), and occupational attainment (Estimate = −0.09, p = .64) were not associated with overall functional ability or rate of decline. However, in view of their theoretical importance as indicators of cognitive reserve, the latter terms were retained in the final model (see ).
Figure 2. (a) and (b) The graphs 2 display the predicted change in Clinical Dementia Rating sum of boxes (CDR-SB) scores for varying numbers of cognitively stimulating activities by dementia duration at Visit 1. Note that the graph is based on the regression equation (more ...)
Effect of Engagement in Cognitively Stimulating Activities and Other Indicators of Cognitive Reserve on Clinical Dementia Rating Sum of Boxes Scores: Results of the Final Model
In this population-based study of incident AD, engagement in cognitively stimulating activities early in the course of AD was associated with slower cognitive decline. This effect was striking, particularly at two and a half years of follow-up where, for example, statistical modeling predicted that those engaged in four or more activities scored approximately four points higher on the MMSE than those engaged in no stimulating activities. Although differences in methodology prohibit direct comparisons, the results differ from those of Helzner and colleagues (2007)
and Wilson and colleagues (2000)
who reported, respectively, no effect or a faster deterioration in cognitive functioning with greater engagement in premorbid
mental activity. The latter study restricted the assessment to reading activity five years prior to dementia onset and did not reassess the activity through the course of dementia. Unlike these studies, we required at least weekly participation in cognitive pursuits, which could suggest that higher levels of engagement are needed to observe a relationship with slower decline. Alternatively, higher levels of engagement in stimulating activities may be a reflection of less severe dementia. Indeed, over time, participants declined in their level of engagement in stimulating activities. Furthermore, the association between stimulating activity and cognition was evident only early in the course of AD.
Engagement in cognitively stimulating activities was also associated with better functional ability, but here, the associations were greater for those with longer dementia duration at assessment. Our lack of observation for similar association in early dementia may reflect the nature of the CDR, a measure that emphasizes functional abilities that are more routinized and exhibit greater declines later in the course of dementia (Hughes et al., 1982
; Morris, 1993
). There was no association between engagement in cognitive activity and rate
of functional decline.
Physical activity was not associated with change in rate of cognitive or functional decline in our participants. Physical activity has known positive benefits in healthy individuals, and engagement throughout mid- and late-life has been shown to be protective against late-life cognitive decline and AD in some studies (Colcombe et al., 2004
; Dik, Deeg, Visser, & Jonker, 2003
; Laurin, Verreault, Lindsay, MacPherson, & Rockwood, 2001
; Weuve et al., 2004
; Yaffe, Barnes, Nevitt, Lui, & Covinsky, 2001
) but not others (Carlson et al., 2008
; Scarmeas, Levy, Tang, Manly, & Stern, 2001
; Verghese et al., 2003
, 2006). Thus, the effects of physical activity in AD are not well understood. Although our results are negative, it is possible that our crude measurement of total physical activity was insufficient in capturing the effects of potentially neuroprotective aerobic activities. Nonetheless, our results, while observational, suggest limited benefits of physical activity to cognition following the onset of dementia.
Among the strengths of the study were its large population-based sample, systematic assessment of cognitive activity, and examination of both cognitive and functional outcomes. In addition, the six-month follow-up intervals allowed for increased sensitivity in tracking decline. The comprehensive protocol for determining a dementia diagnosis and high participation rates among participants and their caregivers were additional strengths.
Several limitations of this study merit discussion. We did not examine differential effects of specific activities. Previous research has suggested a positive effect for social engagement and the maintenance of cognitive functioning (e.g., Barnes, Mendes de Leon, Wilson, Bienias, & Evans, 2004
). Additionally, social disengagement may lead to a reduction in cognitive stimulation (Wang et al., 2002
). Given that many leisure activities are both cognitive and social in nature, the relative effect of each represents an area for future research. There was also significant participant attrition (mostly due to death) over the course of the study. This may have reduced statistical power in examining the longitudinal effects of engagement in stimulating activities. Our results may also reflect the effects of differential attrition, for example, if greater attrition had occurred among those with severe dementia or engaged in fewer activities. Finally, our sample, while population based, was 99% Caucasian and relatively well educated. Our results may not generalize to other populations with more diverse ethnic representation or levels of educational attainment.
A worthwhile endeavor may be to explicitly examine whether engagement in cognitively stimulating activities after dementia onset may produce beneficial effects. A randomized controlled intervention trial of stimulating activities in early dementia may clarify this issue. To date, there have been no such trials after the onset of dementia. Studies among the non-demented elderly suggest some benefit of cognitive intervention (Willis et al., 2006
). From a cognitive perspective, late-life volunteer activity in cognitively at-risk older adults has been associated with protection (Carlson et al., 2008
, 2009). These findings suggest the potential for use-dependent brain plasticity among those with dementia, but their limits remain to be determined.
National Institute of Health grants R01AG21136 and R01AG11380.