American Journal of Epidemiology
ª The Author 2011. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of
Public Health. All rights reserved. For permissions, please e-mail: [email protected].
Vol. 173, No. 9
DOI: 10.1093/aje/kwq476
Advance Access publication:
March 4, 2011
Original Contribution
Cognitive Lifestyle and Long-Term Risk of Dementia and Survival After Diagnosis
in a Multicenter Population-based Cohort
Michael Valenzuela*, Carol Brayne, Perminder Sachdev, Gordon Wilcock , and Fiona Matthews on
Behalf of the Medical Research Council Cognitive Function and Ageing Study
* Correspondence to Dr. Michael Valenzuela, Regenerative Neuroscience Group, Neuropsychiatric Institute, Prince of Wales
Hospital, Randwick, New South Wales 2031, Australia (e-mail: [email protected]).
Initially submitted July 19, 2010; accepted for publication December 6, 2010.
An active cognitive lifestyle has been linked to dementia incidence and survival, but the separate and combined
effects of its subcomponents are not clear. Data were derived from the Medical Research Council Cognitive
Function and Ageing Study, a population-based study of 13,004 individuals in England and Wales first interviewed
in 1991–1992 and followed over a 10-year period for dementia incidence and 12 years for mortality. A Cognitive
Lifestyle Score (CLS), defined as a composite of cognitive activity including education, occupational complexity,
and social engagement, was available for 12,600 individuals in 3 stages of life. A higher CLS was protective of
dementia (odds ratio ¼ 0.6, 95% confidence interval: 0.4, 0.9). Sensitivity analyses found this main effect to be
reliable and replicable even when considering just 2 components of the score, either education and occupation or
education and late-life social engagement. No single CLS factor was associated with dementia incidence on its
own. Survival differences did not reach statistical significance. Our data suggest that more years of education, as
well as further stimulatory experiences in either midlife or late life. are necessary for a protective link with dementia
incidence. There was little evidence of an effect of cognitive lifestyle on survival after dementia diagnosis.
cognitive reserve; dementia; incidence; survival
Abbreviations: AGECAT, Geriatric Mental State-Automated Geriatric Examination for Computer Assisted Taxonomy; CI,
confidence interval; CLS, Cognitive Lifestyle Score; LEQ, Lifetime of Experiences Questionnaire; OR, odds ratio.
Epidemiologic studies provide strong evidence linking an
enriched cognitive lifestyle with reduced dementia risk. For
example, a meta-analysis of 22 studies revealed a relative
risk reduction for incident dementia of 46% in those with
more complex cognitive life experiences (odds ratio (OR) ¼
0.54, 95% confidence interval (CI): 0.49, 0.59) (1). Interestingly, despite differences in the way these factors have been
measured, the relative effects of high versus low education
(OR ¼ 0.53), occupational complexity (OR ¼ 0.56), and
late-life leisure activities (OR ¼ 0.50), were highly consistent. Moreover, several more recent prospective studies
have revealed a dose-dependent association (2), whereby
dementia risk declines with each step of increasing cognitively loaded leisure activities (3–6). What is not clear is
the relative importance of the different cognitive lifestyle
components or their combinatorial interaction in relation to
dementia incidence or survival after diagnosis.
The Lifetime of Experiences Questionnaire (LEQ) was
developed to measure cognitive lifestyle differences between individuals by integrating information on stimulatory
educational, occupational, and cognitively loaded leisure
activities across the life span (7). Higher LEQ scores independently predict an attenuated rate of cognitive decline
over time, as well as a diminished rate of longitudinal hippocampal atrophy (8). However, it is unclear whether this integrated approach to cognitive lifestyle can predict incident
dementia. There are also no studies that we are aware of that
have systematically compared the impact of each of these
components in isolation after controlling for the other 2
factors or investigated whether a critical pairwise combination of factors is as powerful as the triple combination.
Furthermore, the effects of cognitive lifestyle on survival
after diagnosis may dissociate from the protective effects
against dementia incidence, because higher education in
1004
Am J Epidemiol. 2011;173(9):1004–1012
Cognitive Lifestyle, Dementia, and Survival
those already with dementia has been linked to increased
mortality (9). Whether a similar pattern also holds for the
other cognitive lifestyle factors of occupational complexity
and late-life cognitive activities is unknown.
The issue of combined versus stand-alone effects of cognitive lifestyle factors on both dementia incidence and then
survival after diagnosis has not been evaluated to date within
the same population-based longitudinal data set. The Medical Research Council Cognitive Function and Ageing Study
(CFAS) is a multicenter, community-based, cohort study
that has now been running for over 14 years and is well
designed to address questions of this nature (10). Using a
subset of questions derived from the LEQ covering the domains of education, occupational complexity, and late-life
cognitive activities that were available as part of baseline
and screening questions from the Cognitive Function and
Ageing Study, we were able to evaluate the impact of cognitive lifestyle on both incident dementia and survival time
after diagnosis. This study therefore aimed at investigating
the individual and combined contribution of 3 major cognitive lifestyle factors to incident dementia risk and survival
after diagnosis.
MATERIALS AND METHODS
Study design and population
Data are taken from the Cognitive Function and Ageing
Study (http://www.cfas.ac.uk), a large multicenter, population-based, prospective cohort study of individuals aged 65
years or older in community and residential settings. Full
details are published elsewhere and briefly described here
(11, 12). Individuals were randomly selected from the Family Health Service Authority lists in 5 areas of England and
Wales, including 2 rural (Cambridgeshire and Gwynedd) and
3 urban (Newcastle, Nottingham, and Oxford). Baseline interviews were undertaken in 1991–1992. A 2-phase screening procedure was used. At baseline screening, 13,004
individuals provided information on physical, behavioral,
and sociodemographic status in addition to aspects of
health, including self-reported chronic conditions, and cognition using the Mini-Mental State Examination (MMSE)
(13) and selected items from the Geriatric Mental StateAutomated Geriatric Examination for Computer Assisted
Taxonomy (AGECAT) (14). Following the baseline interview, a subsample of approximately 20% (n ¼ 2,640) were
selected on the basis of age, center, and cognitive ability and
weighted toward the cognitively frail, to participate in a
more detailed assessment interview that included the full
measures of mood and organicity of the AGECAT (15)
and so allowed dementia diagnosis. At 2 years, those who
had not undertaken an assessment interview were rescreened with a further 20% undergoing the diagnostic interview. Respondents who underwent further assessment at
baseline were asked to complete 1 or 2 yearly follow-up
interviews including a diagnostic interview (12). Further
interviews were undertaken at 3, 6, 8, and 10 years (different
subgroups targeted at each time with a maximum of 8 interviews in 10 years). At each diagnostic interview depression,
anxiety, and dementia status were derived from the full
Am J Epidemiol. 2011;173(9):1004–1012
1005
AGECAT diagnostic algorithm. Dementia was defined as
an AGECAT organicity rating case level of 3 or above and
is comparable to dementia as diagnosed by the Diagnostic
and Statistical Manual of Mental Disorders: DSM-IIIR (14,
15). Depression and anxiety were both defined as an AGECAT symptom level of 3 or above. Informant interviews
were undertaken on those with the detailed assessment interviews and from which the Blessed Dementia Rating Scale
can be measured. Incident dementia was measured for all
individuals who developed dementia at follow-up who did
not have dementia at the previous interview (16). A history
of vascular risk factors (diabetes, medicated high blood
pressure, heart attack, angina, stroke) was asked at each
interview, as well as smoking history at baseline. These
factors were combined to produce a vascular risk factor
score: Individuals were given 1 point for the presence of
each risk factor, except for smoking where individuals were
coded as 0 for nonsmokers, 1 for former smokers (more than
5 years ago), and 2 for current smokers. Individuals could
therefore score between 0 and 8, and the median score was
2. Data from all waves have been used in the analysis (Data
Version 8.2, December 2006), in addition to death notifications from the United Kingdom’s National Health Service
Central Register.
This study had ethical approval from the Eastern Anglia
Multicentre Research Ethics Committee and all local ethical
committees for the duration of the study (1990 to date). All
individuals gave written, informed consent.
Cognitive lifestyle score
The LEQ asks detailed information about an individual’s
range and intensity of educational, occupational, and cognitive lifestyle activities in all 3 phases of life (young adulthood, midlife, and late life) (7). The original Cognitive
Function and Ageing Study protocol did not cover these
domains in as much detail as the LEQ, so we constructed
a simplified proxy for the LEQ called the Cognitive Lifestyle Score (CLS) as follows. Matches were found between
the study’s baseline/screening questions and LEQ questions,
and the resultant abbreviated score was tested to determine
if it correlated sufficiently highly with the overall LEQ. This
test was completed by using data from the LEQ validation
study (7). The following combination of questions was
found to correlate significantly with the total LEQ score
(r ¼ 0.60, P < 0.0001, n ¼ 70; data not shown):
1. Educational level in young adulthood: assessed by the selfreport question, ‘‘how many years of full-time education?’’
2. Occupational complexity in midlife: assessed by recording the participant’s main occupation in terms of years
most worked and then recoding it using 2 systems, his/her
social class grouping (from I to VI) and socioeconomic
grouping (from 11 to 150). These 2 systems distinguish
different occupations by complexity and status (17). Social class groupings were exploded with the socioeconomic groupings and ranked to create a finer detail than
either grouping alone. These groups have then been reclassified into 14 groups, the 15th representing housewives who do not have a formal code within the United
1006 Valenzuela et al.
To enable each subscore to have an equal and unbiased
contribution to the combined score, the CLS was generated
by using a weighted function prior to any risk factor
analysis:
CLS ¼ 23ðeducationÞ þ 23ðoccupation complexityÞ
þ 33ðsocial engagementÞ:
Weighting factors for each component were empirically
derived at the whole-group level so that the resultant
weighted mean scores for each component were equivalent.
This method follows the structure of the LEQ, whereby
young adulthood, midlife, and late-life subscores are also
equally weighted. Housewives were given an occupation of
minimum complexity. Following common practice, sexspecific tertiles for the CLS were generated to investigate
contrasting high, medium, and low cognitive lifestyle
groups. The associations among the 3 CLS subscores were
compared by using Spearman’s rank correlation coefficient.
Statistical methods
Dementia incidence. A nested case-control analysis of
incident dementia versus nondementia has been undertaken
adjusting for study design, attrition, age at interview, and
wave of interview. The nested case-control approach
considers each wave of follow-up separately and defines
cases and controls for that time point. Cases are therefore
compared with the controls nested at the same follow-up
time. Multivariable logistic regression analyses have been
used.
Dementia survival. Median survival times by group and
Cox proportional hazards regression have been used to investigate differences in survival for incident dementia cases.
Proportional hazards assumptions have been checked by
using Schoenfield residuals. Deaths to the end of 2008 were
included in this analysis.
years for the 11,591 who survived to the follow-up
interview).
Cognitive lifestyle score
The continuous CLS measure was generated for 12,600
individuals (97%), with 157 missing education components,
320 missing occupation components, and 83 missing current
activity components. Of the 404 individuals with missing
data, 99 could be classified within the sex-specific tertiles
regardless of their missing information as their grouping did
not change after replacement with the maximum or minimum possible value, thereby leaving 12,699 individuals for
analysis of tertile groups. Figure 1 shows the distribution of
the CLS that approximated a normal distribution with slight
right-hand skewness. The CLS had a mean of 41.2 and a
standard deviation of 11.2. Table 1 also shows the general
characteristics of the sample.
Incident dementia
During the course of the longitudinal phase of the study,
there have been 438 incident dementia cases (16). Of these,
361 were diagnosed at interviews where a nested casecontrol study could be undertaken (120 cases at year 2,
138 at year 6, and 103 at year 10). The other cases were
diagnosed at annual follow-up points between interviews
where control status is not known for all individuals. These
361 individuals were compared with those individuals who
did not develop dementia.
As shown in Table 2, higher scores on the total CLS
appear to be protective against the incidence of dementia,
with a significant reduction in relative risk for the high LEQ
group compared with the low LEQ group. Controlling for
the covariates of age, sex, wave, and vascular risk factors did
not eliminate this association (OR ¼ 0.6, 95% CI: 0.4, 0.9).
We also analyzed each CLS component’s prediction of
incident dementia, before and after adjusting for the other
CLS components. No particular component reached statistical significance (for all levels, P > 0.2, except for enriched
0.06
0.05
0.04
Density
Kingdom system (n ¼ 2,171 women). These scores were
reversed so as to be in the same direction as education and
current social engagement (with low scores indicating
less cognitive lifestyle activity).
3. Social engagement (current levels in later life): calculated
on the basis of three 3-point Likert scale questions (i.e,,
minimum of 3 and maximum of 9): a) frequency of contact with children and other relatives; b) frequency of
face-to-face contact with neighbors; and c) frequency of
attending meetings, clubs, and other social events.
0.03
0.02
RESULTS
Study population
0.01
The study consists of 13,004 individuals aged 65 years or
above with baseline interview. There are 7,847 women and
5,157 men; 1,682 of 12,747 (13%) had a Mini-Mental State
Examination score of <22 at baseline interview. Individuals
were followed from baseline by using interviews at 0, 2, 6,
and 10 years (mean length of time on active follow-up: 4.1
0.00
20
40
60
80
Cognitive Lifestyle Score
100
Figure 1. Distribution of the Cognitive Lifestyle Score (n ¼ 12,699)
in the Cognitive Function and Ageing Study Sample, England and
Wales, 1991–2006.
Am J Epidemiol. 2011;173(9):1004–1012
Am J Epidemiol. 2011;173(9):1004–1012
Table 1. Characteristics of the Cognitive Function and Ageing Study Sample, England and Wales, 1991–2006
Low (n 5 4,178)
Medium (n 5 3,942)
High (n 5 4,579)
Total (n 5 12,699)
Missing (n 5 305)
Cognitive Lifestyle Score Group
No.
%
65–74
1,787
75–84
1,782
Median
IQR
No.
%
43
1,984
43
1,571
609
15
1,623
40
Median
IQR
No.
%
50
2,520
40
1,687
387
34
1,567
40
Median
IQR
No.
%
Median
IQR
No.
%
55
6,291
50
43
14
37
5,040
40
122
40
372
35
1,368
11
140
46
1,889
41
5,079
40
78
26
Median
IQR
9
9–10
10
6–17
Age, years
85
Men
Years of education
9
9–9
9
9–10
10
9–12
9
9–12
Occupational category
High
Medium
Low
Unknown
Housewife
69
36
372
35
2,873
63
3,314
26
1 <1
790
19
1,937
49
1,259
28
3,986
31
9
1,640
39
1,007
26
358
35
3,005
24
8
3
69
36
21 <1
30
1
120
1
283
93
1,610
39
4
1
605
15
3
Social class
I/II
III
IV/V
Unclassified
8 <1
5
2,759
60
2,983
23
1 <1
20
2,026
51
1,649
36
4,500
35
9
1,707
41
1,093
28
101
36
2,901
23
9
3
41 <1
282
92
28 <1
2 <1
11 <1
3
1,610
39
605
15
59
1
2,274
18
4
1
Visiting family frequentlya
1,730
42
2,189
56
2,301
50
6,220
49
54
31
Visiting neighbors frequentlya
2,575
62
3,105
79
3,592
79
9,272
73
212
94
763
18
1,908
49
2,733
60
5,404
43
17
35
Social groups
MMSE at baseline
25
22–28
27
24–28
28
26–29
27
24–28
Cognitive Lifestyle Score
Total score
31
28–33
39
37–41
51
47–57
40
33–48.5
Education subscore
9
9–9
9
9–10
10
9–12
9
9–11
9
9–10
Occupation subscore
2
1–4
6
3–7
11
8–12
6
2–10
3.5
3–6
Social engagement subscore
6
5–7
7
6–8
7
7–8
7
6–8
5
4–6
Abbreviations: IQR, interquartile range; MMSE, Mini-Mental State Examination; NA, not applicable.
Missing data excluded.
a
NA
Cognitive Lifestyle, Dementia, and Survival
Housewife
216
825
1007
1008 Valenzuela et al.
Table 2. Incident Dementia and Relation to Cognitive Lifestyle in the Cognitive Function and Ageing Study Sample, England and Wales, 1991–
2006a
Additional Adjustment for Other Cognitive
Lifestyle Score Components
Multiple Adjustment for Background Variables Only
Cognitive Lifestyle Score Group
Low
Cognitive lifestyle combination
Medium
High
Low
OR
95% CI
OR
95% CI
OR
95% CI
OR
1.0
Referent
1.0
0.7, 1.3
0.6*
0.4, 0.9
NA
1.0
Referent
1.1
0.7, 1.9
0.7
0.4, 1.3
1.0
95% CI
Medium
High
OR
95% CI
OR
NA
NA
NA
1.1
0.7, 1.9
0.8
95% CI
Single components
Education
Referent
0.4, 1.3
Occupation
1.0
Referent
0.9*
0.7, 0.9
0.6
0.4, 1.4
1.0
Referent
0.9
0.7, 1.2
0.7
0.5, 1.0
Social engagement
1.0
Referent
0.8
0.6, 1.2
0.7
0.5, 1.1
1.0
Referent
0.9
0.6, 1.2
0.7
0.5, 1.1
Pairwise combinations
Education þ occupation
1.0
Referent
0.7
0.5, 1.0
0.6*
0.4, 0.8
1.0
Referent
0.8
0.6, 1.1
0.6*
0.5, 0.9
Education þ social engagement
1.0
Referent
0.8
0.6, 1.0
0.6*
0.4, 0.8
1.0
Referent
0.7
0.6, 1.0
0.6*
0.4, 0.8
Occupation þ social engagement
1.0
Referent
1.1
0.8, 1.5
0.7
0.5, 1.0
1.0
Referent
1.1
0.8, 1.5
0.8
0.6, 1.1
Abbreviations: CI, confidence interval; CLS, Cognitive Lifestyle Score; NA, not applicable; OR, odds ratio.
* P < 0.05 (statistically significant).
a
All models were adjusted for background variables including age, sex, interview wave, and a combination of vascular risk factors. Additional
adjustment for each single CLS component consisted of controlling for the other 2 CLS subscores, while the pairwise combinations included
adjustment for the single remaining CLS subscore.
occupational complexity, P ¼ 0.07). In general, each of the
3 subcomponents of the CLS was only weakly correlated
with each of the others (Table 3).
We additionally tested the 3 different possible pairwise
combinations of CLS components after controlling for the
remaining single CLS factor (Table 2). High levels of education plus occupational complexity or of education plus
late life social engagement appeared to confer risk similar
to that of the complete combination (OR ¼ 0.6 for both).
The combination of enriched occupational complexity and
late life social engagement was not significantly associated
with a change in dementia incidence after taking educational level into account. Figure 2 shows how different
combinations of CLS components were associated with
dementia risk and combined to produce an overall risk.
Survival
Sensitivity analysis
Sensitivity analyses were carried out to determine
whether other potential factors moderated our key finding
of decreased dementia risk in the high CLS group. These
factors included time to diagnosis, removing individuals
Table 3. Spearman’s Correlations Among the 3 Cognitive Lifestyle
Score Subcomponents in the Cognitive Function and Ageing Study
Sample, England and Wales, 1991–2006
Education
r
r
0.27
0.25, 0.29
1.0
0.03
0.01, 0.05
0.01
1.0
Occupation
Social
engagement
95% CI
Social
Engagement,
r
0.03, 0.01
1.0
Occupation
95% CI
Education
who became demented in the first 2 years of follow-up,
splitting our summary vascular risk score into scores for
individual cardiovascular risk factors, history of head injury
or boxing, and comorbid depression or other emotional
problems. As can be seen in Table 4, none of these factors
significantly altered our key findings.
Because preclinical and undiagnosed dementia may have
affected our estimation of the risk of social engagement on
dementia incidence more than the other CLS components,
we also tested this relation before and after excluding individuals who became demented within the first 2 years of
follow-up. Again, the effects were minimal, with odds ratios
for high social engagement compared with low changing
from 0.71 to 0.69, respectively.
Abbreviation: CI, confidence interval.
Of the 438 incident dementia respondents with survival
data, there are 434 (99%) that have a CLS. Of these, 398
(92%) had died by December 31, 2008, with a median survival time of 4.5 years (interquartile range: 2.8–6.9). There
was no positive evidence that individuals with contrasting
CLS scores differed in terms of age at dementia onset or
survival time after diagnosis (Table 5). Some evidence of
survival differences was shown for 2 comparisons, although
larger numbers are needed to confirm these findings. The
medium and high CLS groups had an adjusted hazard ratio
of 1.3 (95% CI: 1.0, 1.7) for decreased survival time compared with the low CLS group; similarly, those with a high
occupational complexity had decreased survival time
(hazard ratio ¼ 1.4, 95% CI: 1.0, 1.9). No interactions were
found between the CLS and severity (for both the Blessed
Dementia Rating Scale and the Mini-Mental State Examination: P > 0.2).
Am J Epidemiol. 2011;173(9):1004–1012
Cognitive Lifestyle, Dementia, and Survival
B)
10
10
Odds Ratio
Odds Ratio
A)
1
1
0.1
0.1
Low
Medium
Low
High
Total CLS Score
C)
Medium
High
Education and Occupation
D)
10
10
Odds Ratio
Odds Ratio
1009
1
0.1
Low
Medium
High
Education and Social
1
0.1
Low
Medium
High
Occupation and Social
Figure 2. Adjusted odds ratios for incident dementia in individuals with low (reference), medium, and high cognitive lifestyle in the Cognitive
Function and Ageing Study Sample, England and Wales, 1991–2006. All graphs show point estimates as well as 95% confidence intervals after
controlling for background variables of age, sex, interview wave, and a combination of vascular risk factors. Risk estimates have been based on
complete Cognitive Lifestyle Score (CLS) (A), combined education and occupational factors (controlling in addition for social engagement) (B),
education and social engagement (controlling for occupational factors) (C), and occupational and social engagement factors (controlling for
education) (D).
DISCUSSION
Fourteen-year prospective data from a large, multicenter,
and community-based study in the United Kingdom were
used to evaluate the combined and independent associatons
of cognitive lifestyle on ongoing dementia risk and survival.
We found that the combination of education, occupational
complexity, and late-life social engagement, rather than any
individual component, was an independent predictor of dementia risk. Those with a higher overall CLS were at 40%
decreased risk for developing dementia. There was no clear
association between cognitive lifestyle and survival time
after diagnosis.
Our findings are consistent with previous epidemiologic
estimates of the protective effects of cognitive lifestyle. For
example, our meta-analysis of cognitive lifestyle found an
overall relative risk reduction of dementia incidence of 46%
(1), comparable to 40% in this study. Although reverse
causality cannot be completely accounted for, our results
are unlikely to be influenced by a diagnostic threshold
effect, because including time to diagnosis as a covariate
Am J Epidemiol. 2011;173(9):1004–1012
or excluding those individuals who became demented
within 2 years of baseline interview did not substantively
change our results. Furthermore, exclusion of individuals
who became demented within 2 years of follow-up did not
alter the estimated protective association of high late-life
social engagement, the CLS factor most vulnerable to
reserve causality bias if present. These analyses also found
no moderating influence of depression, head injury, or the
use of individual cardiovascular risk factors rather than a
vascular summary score in our models. On the basis of our
10-year population-based longitudinal study, an active cognitive lifestyle therefore appears to reliably predict lower
dementia risk.
Each of the CLS subcomponents measured relatively independent facets of cognitive lifestyle because they were
only weakly intercorrelated. It is therefore interesting
that no particular single cognitive lifestyle component
was significantly associated with dementia risk, either before or after adjusting for the other 2 components. Cohort
studies of occupational complexity that have simultaneously
controlled for education have produced conflicting
1010 Valenzuela et al.
Table 4. Logistic Regression Results of Dementia Incidence Based on Cognitive Lifestyle Score Tertiles After Accounting for Different
Combinations of Additional Predictor Variables Not Specified in the Main Model, the Cognitive Function and Ageing Study Sample, England and
Wales, 1991–2006
Cognitive Lifestyle Score Tertile
Sensitivity Analysis
Low
Adjusted just for time
Medium
High
OR
95% CI
OR
95% CI
OR
95% CI
1.0
Referent
0.86
0.6, 1.2
0.55
0.4, 0.8
Adjusted for age, sex, and time
1.0
Referent
0.94
0.7, 1.3
0.62
0.4, 0.9
Main analysis: adjusted for age, sex, time, and vascular factors (grouped)
1.0
Referent
0.96
0.7, 1.3
0.61
0.4, 0.9
Adjusted for age, sex, time, and single vascular factors
1.0
Referent
0.95
0.7, 1.3
0.62
0.4, 0.9
Adjusted for age, sex, time, vascular factors (grouped), and depression
1.0
Referent
0.81
0.5, 1.2
0.51
0.3, 0.8
Adjusted for age, sex, time, vascular factors (grouped), and any emotional problems
1.0
Referent
0.97
0.7, 1.3
0.61
0.4, 0.9
Adjusted for age, sex, time, vascular factors (grouped), and disability
1.0
Referent
0.97
0.7, 1.3
0.65
0.4, 0.9
Adjusted for age, sex, time, vascular factors (grouped), head injury, and boxing
1.0
Referent
1.01
0.7, 1.4
0.59
0.4, 0.8
Lagged for the first 2 years, adjusted for age, sex, time, and vascular factors (grouped)
1.0
Referent
1.05
0.7, 1.6
0.66
0.4, 1.0
Abbreviations: CI, confidence interval; OR, odds ratio.
results; some have noted a residual protective effect on incident dementia (18), while others have not (5, 19, 20). Only
1 previous study that controlled for education and occupation when examining late-life leisure activities found a signification protective effect (21). Perhaps more importantly,
our systematic review suggested that the overall cohort
event rate is a key consideration when testing for the differential effects of a cognitive lifestyle factor on dementia
incidence (1). In the current study, 434 incident dementia
cases in the subsample from the baseline cohort of 12,699
were followed over a 10-year period. After accounting for
study design, this averages to a new-case rate of 1.8% per
year (10). In our meta-analysis, dementia event rates ranged
from 0.2% to 4.9% per annum. Hence, the midrange event
rate observed in the Cognitive Function and Ageing Study
may explain why individual CLS factors did not reach significance. Even large cohort studies may be insufficiently
powered to fully analyze interactions among individual cognitive lifestyle components, a task that may require metaanalyses incorporating individual patient data, background
variables, and study design.
By contrast, paired combinations of cognitive lifestyle
components were also tested while controlling for the third
factor, yielding a different pattern. When a high level of
education was combined with either a complex occupation
or social engagement in later life, a significant 40% level of
risk reduction was found (Figure 2, B and C). Alternatively,
a complex job plus late-life social engagement was not sufficient to reduce incident dementia risk after adjusting for
education (Figure 2D). Particular combinations of CLS factors therefore appear to be more strongly predictive of reduced dementia risk than others or when considered in
isolation. Specifically, a higher level of education may be
most effective in providing some protection against dementia when in combination with subsequent cognitive complexity and stimulation in later life.
There was no compelling evidence that an enriched cognitive lifestyle is linked to decreased survival. Stern et al.
(22) found that those with increased cognitive reserve in the
form of higher education or occupational complexity experienced accelerated cognitive decline. Although this relation
has been replicated in some cohort studies (23–25), others
have found either no link (26, 27) or even the opposite
finding (28). The reasons for these discrepancies may include the selection of cognitive outcome measure, various
degrees of dementia severity, different follow-up time periods, or an interaction with comorbid depression (29).
Importantly, in this study we did not specifically examine
the rate of cognitive decline after diagnosis but rather survival time, and these variables may not necessarily be correlated. Although individuals with high educational levels
experienced faster cognitive decline after diagnosis on the
basis of repeated Mini-Mental State Examination tests,
Bruandet et al. (30) found there was no effect on time to
death. Generally, study results of differential survival after
diagnosis based on educational level have not been consistent, including positive (9, 31) and negative (18, 32, 33)
studies. As identified in our meta-analysis, predictions based
on educational levels exhibit a significant degree of heterogeneity (1), so the overall educational level of the cohort
may be important. In our study, individuals with a very low
level of education were not seen, reflecting its rare occurrence in the population of England and Wales. Most individuals, for example, had 9 years of full-time education,
with very few having less than 8 years. Because individuals
with higher levels of education may mask their dementia
and falsely achieve a higher score, our study benefited from
selecting individuals for an initial diagnostic interview on
the basis of 2 separate screening mechanisms. The potential
for diagnostic bias was thereby reduced as the result of a
more random sampling of the complete population.
Other important factors that may have contributed to our
null dementia survival findings include the level of medical
comorbidity in the sample, overall disease severity, and the
length of follow-up. Despite these issues, we noted nearsignificant findings in the direction of abbreviated survival
time in the higher cognitive lifestyle group. The effect may
therefore be subtle and require either a larger sample size for
Am J Epidemiol. 2011;173(9):1004–1012
Cognitive Lifestyle, Dementia, and Survival
Table 5. Median Age at Dementia Onset and Survival Time After
Diagnosis, as Well as Adjusted Hazard Ratios for Differential
Survival Time as a Function of Cognitive Lifestyle Score Category,
the Cognitive Function and Ageing Study Sample, England and
Wales, 1991–2006
Subgroup and
Cognitive Lifestyle
Score Category
Median Age
at Onset,
years
Survival
Years (IQR)
Hazard
Ratioa
95% CI
All
Low
83.9
4.6 (3.2–7.1)
1.0
Referent
Medium
84.0
4.1 (2.8–6.6)
1.3
1.0, 1.7
High
84.4
4.2 (2.6–6.9)
1.3
1.0, 1.7
Men
Low
82.4
4.2 (2.9–7.9)
1.0
Referent
Medium
84.0
3.9 (2.4–6.1)
1.3
0.7, 2.3
High
82.0
3.7 (2.2–6.2)
1.6
0.8, 3.3
Women
Low
84.4
4.6 (3.3–7.1)
1.0
Referent
Medium
84.0
4.4 (3.1–6.8)
1.3
0.9, 1.7
High
85.0
4.2 (2.7–7.3)
1.3
0.9, 1.7
Education
Low
86.7
3.8 (2.4–5.8)
1.0
Referent
Medium
83.7
4.6 (2.9–7.0)
1.2
0.9, 1.6
High
84.4
3.7 (2.6–6.6)
1.2
0.8, 1.8
Occupation
Low
84.0
4.5 (2.8–7.0)
1.0
Referent
Medium
83.6
4.6 (3.0–7.0)
1.2
0.8, 1.5
High
85.1
3.6 (2.4–6.7)
1.4
1.0, 1.9
1011
and cognitive lifestyle activities may change over time. A
more detailed assessment, using for example the Lifetime of
Experiences Questionnaire (7) in a longitudinal setting,
could therefore be more informative. Finally, many women
in this generation were not employed while married. It is
standard practice in the United Kingdom for married and
widowed women to be classified according to their husband’s occupation for social class, while here these individuals’ occupations were coded as housewives. Future studies
could therefore focus on whether married women who did
not participate in the workforce have differential dementia
risk based on their husbands’ cognitive lifestyles.
In conclusion, we compared dementia risk and survival in
individuals with a range of cognitive lifestyle patterns using
the Cognitive Function and Ageing Study longitudinal data
set. Our main finding was a 40% reduced risk for developing
dementia in those that maintain a high level of cognitive
complexity throughout their lives, independent of other
known risk factors. By comparison, individual components
of an active cognitive lifestyle, such as educational achievement, occupational complexity, or social engagement, were
not linked to a protective effect. The 2-factor combination
of higher education and either a more complex occupation
or late-life social engagement was as effective in reducing
dementia risk as the complete 3-factor combination. Some
level of cognitive enrichment beyond education in young
adulthood may therefore be required for more effective dementia prevention. No conclusive evidence was found for an
effect on survival time after diagnosis, yet this may be due to
a relatively subtle relation that requires more investigation.
Social engagement
Low
83.9
4.6 (3.1–7.6)
1.0
Referent
Medium
84.3
4.3 (2.8–6.5)
1.1
0.8, 1.4
High
84.2
4.3 (2.3–6.9)
1.2
0.8, 1.5
Abbreviations: CI, confidence interval; IQR, interquartile range.
Hazard ratios from a multivariable Cox proportional hazards regression model adjusting for sex (all only), age at dementia, living
alone, marital status, Blessed dementia severity, self-reported health,
Mini-Mental State Examination score, and vascular risk factors.
a
detection or the use of more complex longitudinal modeling
to dissociate the effects on time of presentation versus the
rate of progression and time to death. These are interesting
theoretical questions for future research.
There were also some limitations to this study. The information used to generate the CLS was limited to those
questions originally devised by the Cognitive Function
and Ageing Study team over 14 years ago and, for this
reason, the full range of cognitive lifestyle activities was
not assessed. Assessment of social engagement, for example, was particularly simplistic in comparison to current
methods. The CLS also uses information from the baseline
interview, and this may introduce bias, as engagement in
cognitive leisure activities will almost certainly be different
for the young-old than for the oldest old. Furthermore, our
component scores assume a static level of participation,
whereas individuals’ educational, occupational complexity,
Am J Epidemiol. 2011;173(9):1004–1012
ACKNOWLEDGMENTS
Author affiliations: Brain and Ageing Research Program,
University of New South Wales, Sydney, New South Wales,
Australia (Michael Valenzuela, Perminder Sachdev); School
of Psychiatry, University of New South Wales, Sydney, New
South Wales, Australia (Michael Valenzuela, Perminder
Sachdev); Regenerative Neuroscience Group, Faculty of
Medicine, University of New South Wales, Sydney, New
South Wales, Australia (Michael Valenzuela); Institute of
Public Health, University of Cambridge, Cambridge, United
Kingdom (Carol Brayne); Nuffield Department of Medicine, University of Oxford, Oxford, United Kingdom
(Gordon Wilcock); and Medical Research Council Biostatistics Unit, Institute of Public Health, University of Cambridge, Cambridge, United Kingdom (Fiona Matthews).
This work was supported by the Department of Health
and the Medical Research Council (grant MRC/G9901400).
G. W. is partly supported by the National Institute for Health
Research Oxford Comprehensive Biomedical Research
Centre. F. M. is currently funded by the Medical Research
Council (grant MRC/U.1052.00.013). The Medical Research Council Cognitive Function and Ageing Study is part
of the Cambridgeshire and Peterborough Collaboration for
Leadership in Applied Health Research and Care
(CLAHRC). M. V. is a University of New South Wales Vice
Chancellor’s Research Fellow.
1012 Valenzuela et al.
The Medical Research Council and the Department of
Health had no role in the study design, collection of data,
analysis, or decision to publish.
Conflict of interest: none declared.
16.
17.
18.
REFERENCES
1. Valenzuela MJ, Sachdev P. Brain reserve and dementia: a
systematic review. Psychol Med. 2006;36(4):441–454.
2. Valenzuela MJ, Sachdev P. Cognitive leisure activities, but not
watching TV, for future brain benefits. Neurology.
2006;67(4):729; author reply 729.
3. Fabrigoule C, Letenneur L, Dartigues JF, et al. Social and
leisure activities and risk of dementia: a prospective longitudinal study. J Am Geriatr Soc. 1995;43(5):485–490.
4. Fratiglioni L, Wang HX, Ericsson K, et al. Influence of
social network on occurrence of dementia: a community-based
longitudinal study. Lancet. 2000;355(9212):1315–1319.
5. Karp A, Andel R, Parker MG, et al. Mentally stimulating
activities at work during midlife and dementia risk after age
75: follow-up study from the Kungsholmen Project. Am J
Geriatr Psychiatry. 2009;17(3):227–236.
6. Verghese J, Lipton RB, Katz MJ, et al. Leisure activities and
the risk of dementia in the elderly. N Engl J Med. 2003;
348(25):2508–2516.
7. Valenzuela MJ, Sachdev P. Assessment of complex mental
activity across the lifespan: development of the Lifetime of
Experiences Questionnaire (LEQ). Psychol Med. 2007;
37(7):1015–1025.
8. Valenzuela M, Sachdev P, Wen W, et al. Lifespan mental
activity predicts diminished rate of hippocampal atrophy
[electronic article]. PLoS One. 2008;3(7):e2598.
9. Stern Y, Tang MX, Denaro J, et al. Increased risk of
mortality in Alzheimer’s disease patients with more advanced
educational and occupational attainment. Ann Neurol. 1995;
37(5):590–595.
10. Matthews F, Brayne C. The incidence of dementia in England
and Wales: findings from the five identical sites of the MRC
CFA Study [electronic article]. PLoS Med. 2005;2(8):e193.
11. Cognitive function and dementia in six areas of England and
Wales: the distribution of MMSE and prevalence of GMS
organicity level in the MRC CFA Study. The Medical Research Council Cognitive Function and Ageing Study (MRC
CFAS). Psychol Med. 1998;28(2):319–335.
12. Brayne C, McCracken C, Matthews FE. Cohort profile: the
Medical Research Council Cognitive Function and Ageing
Study (CFAS). Medical Research Council Coginitive Function
and Ageing Study (CFAS). Int J Epidemiol. 2006;35(5):
1140–1145.
13. Folstein MF, Folstein SE, McHugh PR. ‘‘Mini-Mental State.’’
A practical method for grading the cognitive state of patients
for the clinician. J Psychiatr Res. 1975;12(3):189–198.
14. Copeland JR, Dewey ME, Griffiths-Jones HM. A computerized psychiatric diagnostic system and case nomenclature for
elderly subjects: GMS and AGECAT. Psychol Med. 1986;
16(1):89–99.
15. Copeland JR, Kelleher MJ, Kellett JM, et al. A semi-structured
clinical interview for the assessment of diagnosis and mental
state in the elderly: the Geriatric Mental State Schedule. I.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
Development and reliability. Psychol Med. 1976;6(3):
439–449.
Xie J, Brayne C, Matthews FE. Survival times in people with
dementia: analysis from population based cohort study with 14
year follow-up. BMJ. 2008;336(7638):258–262.
Standard Occupational Classification. London, United
Kingdom: Her Majesty’s Stationery Office; 1990.
Schmand B, Smit JH, Geerlings MI, et al. The effects of
intelligence and education on the development of dementia.
A test of the brain reserve hypothesis. Psychol Med. 1997;
27(6):1337–1344.
Evans DA, Hebert LE, Beckett LA, et al. Education and
other measures of socioeconomic status and risk of incident
Alzheimer disease in a defined population of older persons.
Arch Neurol. 1997;54(11):1399–1405.
Helmer C, Letenneur L, Rouch I, et al. Occupation during life
and risk of dementia in French elderly community residents
[comment]. J Neurol Neurosurg Psychiatry. 2001;71(3):
303–309.
Scarmeas N, Levy G, Tang MX, et al. Influence of leisure
activity on the incidence of Alzheimer’s disease. Neurology.
2001;57(12):2236–2242.
Stern Y, Albert S, Tang MX, et al. Rate of memory decline in
AD is related to education and occupation: cognitive reserve?
Neurology. 1999;53(9):1942–1947.
Teri L, McCurry SM, Edland SD, et al. Cognitive decline
in Alzheimer’s disease: a longitudinal investigation of risk
factors for accelerated decline. J Gerontol A Biol Sci Med Sci.
1995;50A(1):M49–M55.
Rasmusson DX, Carson KA, Brookmeyer R, et al. Predicting
rate of cognitive decline in probable Alzheimer’s disease.
Brain Cogn. 1996;31(2):133–147.
Wilson RS, Li Y, Aggarwal NT, et al. Education and the course
of cognitive decline in Alzheimer disease. Neurology. 2004;
63(7):1198–1202.
Mortimer JA, Ebbitt B, Jun SP, et al. Predictors of cognitive
and functional progression in patients with probable
Alzheimer’s disease. Neurology. 1992;42(9):1689–1696.
Suh GH, Ju YS, Yeon BK, et al. A longitudinal study of
Alzheimer’s disease: rates of cognitive and functional decline.
Int J Geriatr Psychiatry. 2004;19(9):817–824.
Fritsch T, McClendon MJ, Smyth KA, et al. Effects of educational attainment on the clinical expression of Alzheimer’s
disease: results from a research registry. Am J Alzheimers Dis
Other Demen. 2001;16(6):369–376.
Geerlings MI, Deeg DJ, Penninx BW, et al. Cognitive
reserve and mortality in dementia: the role of cognition,
functional ability and depression. Psychol Med. 1999;
29(5):1219–1226.
Bruandet A, Richard F, Bombois S, et al. Cognitive decline
and survival in Alzheimer’s disease according to education
level. Dement Geriatr Cogn Disord. 2008;25(1):74–80.
Freels S, Nyenhuis DL, Gorelick PB. Predictors of survival in
African American patients with AD, VaD, or stroke without
dementia. Neurology. 2002;59(8):1146–1153.
Larson EB, Shadlen MF, Wang L, et al. Survival after initial
diagnosis of Alzheimer disease. Ann Intern Med. 2004;
140(7):501–509.
Qiu C, Bäckman L, Winblad B, et al. The influence of
education on clinically diagnosed dementia incidence and
mortality data from the Kungsholmen Project. Arch Neurol.
2001;58(12):2034–2039.
Am J Epidemiol. 2011;173(9):1004–1012