Age and Ageing 1996:25:392-397
Does Education Level Determine the Course
of Cognitive Decline?
DIDIER LEIBOVICI, KAREN RITCHIE, BERNARD LEDESERT,
JACQUES TOUCHON
Summary
Many studies have implicated low education as a risk factor for cognitive impairment in elderly people. Findings
are, however, inconsistent and the mechanism by which education level may intervene in senescent cognitive
change is uncertain. The present study examines cognitive change over a 1-year period in 283 elderly persons
manifesting recent subclinical deterioration in at least one area of cognitive functioning. The results suggest that
the impact of both education level and young adult IQ on the degree of cognitive change over the year is greater
in the older age groups. Secondary memory and language functions were found to be more resistant to decline
in the high-education group, while attention, implicit memory and visuospatial skills are found to decline
irrespective of education level.
Keywords: Cognitive impairment, Dementia, Education.
Introduction
The impact of education level on cognitive ageing is of
interest to both epidemiologists seeking risk factors for
dementia and gerontologists attempting to define the
normal ageing process. Many studies have observed
low education to be consistently associated with a high
prevalence of cognitive impairment [1-4]. While this
observation derived initially from cross-sectional
studies in which cohort effects undoubtedly played an
important role, confirmation has more recently been
provided by longitudinal studies [5, 6].
Several epidemiological studies have also attempted
to determine whether low education may carry an
increased risk of dementia. Most report an inverse
relationship between dementia (both its presence and
severity) and education level [7-13]. In a smaller
number of studies no relationship was found between
age and education [14-16]. Teri et al. [17] have
reported an association between younger age and
higher education and a faster rate of decline in
Alzheimer's disease patients.
The education effect may be limited to specific types
of dementia. Fratiglioni et al. [15] have attributed the
higher prevalence of dementia in the poorly educated to
cases of alcoholic dementia. Mortel et al. [9] have also
reported differences in levels of education in patients
with Alzheimer's disease and patients with vascular
dementia. Ritchie et al. [18] observed that subjects with
'age-associated memory impairment' and a lower
young adult IQ had a much greater risk of evolving
towards dementia over a 2-year observation period than
subjects with a high premorbid IQ level. Some
evidence has also been presented to demonstrate that
elderly persons with a lower level of education have a
later age of onset of dementia [19], however, this may in
part be due to delays in contacting a health professional
by poorly educated people.
Three possible mechanisms may explain lower rates
of cognitive decline in elderly persons with higher
levels of education. First, persons with lower education
are at higher risk of central nervous system damage
(illness, injury, dietary deficiency, alcoholism); second,
persons with higher education may have greater
neuronal reserve capacity and/or reduced risk of
neuronal damage; third, persons with higher levels of
education may be better able to generate compensatory
strategies at both a behavioural and neuronal level.
There may also be an interaction between these factors.
It has not been possible to conclude in favour of any
of the above hypotheses because of shortcomings in the
studies conducted to date. First, with regard to the
measures of cognitive functioning themselves, these
have tended to be global measures (such as screening
tests or a single test of secondary memory functioning)
and little consideration has been given to the more
likely hypothesis that specific cognitive functions may
be differentially affected. The outcome of studies
examining the relationship between age and education
would thereby be determined by the type of tests
selected. Second, studies have been predominantly
cross-sectional, permitting comparisons of age cohorts
but unable to examine the rate of deterioration
EDUCATION LEVEL AND THE COURSE OF COGNITIVE DECLINE
manifested by different education groups over time.
The present study attempts to clarify findings relating
to education and cognitive change by examining
changes in a wide range of cognitive functions over
time.
Subjects
Subjects were recruited from a general-practitioner research
network in the South of France as part of the Eugeria
longitudinal study of cognitive ageing. A proxy screening
questionnaire on cognitive functioning over the past year was
sent to all persons over 60 years of age in each general practice.
This screening instrument, DECO (Deterioration Cognitive
Observee), has been shown in previous studies to be highly
sensitive to early changes in cognitive functioning due to
various causes [20-22]. DECO is a 19-item Likert scale in
which degree of change in behaviour over the past year is
estimated by a proxy, who has had at least monthly contact
with the subject over the past 3 years.
Three hundred and ninety-seven of the subjects included in
the cross-sectional study were found to have a score of less
than 38 on DECO; 38 being the maximum total score. These
persons were thus considered by an observer to have shown
some degree of deterioration in at least one area of cognitive
functioning over the past year. Of these, 283 have so far
completed the second year and are the subjects included in the
present study. Four levels of education were differentiated:
1 = no formal education; 2 = primary school education;
3 = secondary school education; 4 = tertiary education.
Methods
Each subject was visited at home annually by one of the
project interviewers and a computerized neuropsychometric
examination was administered. This examination, ECO
(Examen Cognitif par Ordinateur) was used to assess working
memory, verbal and visuospatial secondary memory, implicit
memory, language skills (words and syntax comprehension,
naming, verbal fluency, articulation), visuospatial performance (ideational, ideo-motor and constructional apraxia,
functional and semantic categorization of visual data, visual
reasoning and form perception), and focused and divided
attention (visual and auditory modalities). The development
of ECO and the theoretical basis for test selection is described
in detail in the report by Ritchie et al. [23].
From the 159 ECO variables, eight summary scores
representing six cognitive domains were used in the analysis.
The eight summary scores were derived from the mean of the
rescaled scores (between 0 and 100) representing six cognitive
domains:
Attention: measured by response time on a dual task
(simultaneous visual selection and counting of auditory
stimuli).
Primary memory (verbal and visual span): assessed by
immediate recall of first names which had the highest
frequency in the French language fifty years ago, and visual
trials of increasing length.
Secondary memory: measured by delayed recall of first
names and their associated faces, and prose recall.
Implicit memory: measured by reference to the number of
trials required to recognize previously presented items
compared with novel stimuli reconstructed progressively on
the computer screen.
Visuospatial ability: measured by reference to two scores;
number of correct responses on tasks of shape-matching,
semantic and functional categorization and reproduction of
393
three-dimensional figures (CR), and response time on shape,
functional and semantic matching tasks (RT).
Language: assessed by tests of word and syntax comprehension, naming and verbal fluency. Two scores are derived;
total number of correct responses (CR), and word and syntax
comprehension response time (RT).
All subjects completed the National Adult Reading Test
(NART). The scores obtained from this simple reading test
correlate highly with intelligence quotients obtained from
formal IQ batteries [24]. The present study examines changes
occurring in cognitive functioning between wave one and
wave two of the studv.
Results
The early cognitive impairment sample of 283 subjects
who fully completed both waves had a mean age of 74.7
years (SD = 7.0) and consisted of 87 men and 196
women. Of the sample, 3.5% had no education, 44.5%
had primary education, 35.7% secondary education and
16.3% tertiary education. Subjects not having yet
completed the second wave were not found to differ
from the present sample in distribution of either age or
educational status.
In order to assess the effect of education on change in
global cognitive functioning, a 'g' or overall change
factor was extracted by Principal Components Analysis. The g factor is the first principal component from
the difference scores (that is, the difference between
summary scores obtained on the first and second wave
of the study). With g as the dependent variable, the
independent variables age, NART, and education were
entered into a linear regression model. The regression
of g on age gave an R of9.3%(p < 0.01) suggesting that
age is a significant determinant of cognitive change over
the past year. Comparing subjects with low education
(education groups 1 + 2 ) with those with high education
(education groups 3 + 4), the age effect was found to be
significant for the low education group (R = 4.5%;
p < 0.01) but not for the high education group
(R = 1.4%; NS). Age thus appears to have a more
significant impact on rate of deterioration over a 1-year
period in persons with little education.
Regression on the NART score was not significant
for the cohort as a whole (R2 = 0.8%; NS). However, a
significant effect was found for subjects over 75 years
(R 2 = 3.5%; p < 0.03) but not for those under 75
(R = 1.0%; NS). Premorbid intelligence thus appears
to have little influence on the rate of cognitive decline in
younger cohorts but becomes significant at higher ages.
In order to answer the question of whether age or
education plays the most important role in cognitive
performance, correspondence analysis was used to
determine the factors which best explained separation
between the scores obtained on the second wave. The
second wave was chosen on the basis that improvements in scores due to learning by unimpaired subjects
would increase variability. The first axis, explaining
62.5% of the separation between groups of individuals
in terms of their cognitive profile, separates high and
low levels of education. The order of modalities on this
D. LEIBOVICI ET AL.
394
Table I. Mean rank scores and Kruskal-Wallis tests for each
cognitive function assessed, for two age ranges and two levels
of education
Education group
Wl
Attention
Language
RT
RC
Memory
primary
secondary
4
3
2
1
K-W( P )
OL
119
OL
126
OL
92
OL
106
OL
108
OH
137
OH
YL
154
YH
162
YH
OH
136
YL
140
YH
151
YL
138
YL
148
YL
144
OH
150
8.09
(0.04)
8.4
(0.03)
32.6
(0)
23.2
(0)
16.8
OL
97
OL
109
OH
130
OH
138
YL
YH
178
YH
OL
OL
OH
128
YL
101
139
YL
158
OH
148
OL
103
OL
102
OL
101
OH
134
OH
147
169
YH
171
YH
162
implicit
Visuospatial
RT
RC
149
YL
141 '
172
(0)
3
NS
34.1
(0)
25.1
(0)
W2
Attention
YH
169
YH
158
YH
162
YH
168
17.9
(0)
22.6
(0)
19.8
(0)
20.6
(0)
26
(0)
implicit
5.5
Visuospatial
RT
RC
134
YL
159
YL
143
YH
165
YH
174
NS
31.7
(0)
26.3
(0)
2
3
4
K-W(p)
112
Language
RT
RC
Memory
primary
secondary
Education group
OL
95
OL
105
1
143
OH
146
YL
134
OH
138
OH
YL
155
YL
148
OH
152
YH
160
W1-W2
Attention
Language
RT
RC
Memory
primary
secondary
OL
170
YL
136
YH
135
OH
129
OH
163
OL
147
YH
136
YL
124
implicit
Visuospatial
RT
RC
6.14
NS
9.2
(0.02)
1.31
NS
2.09
NS
9.1
(0.02)
6
YH
163
OL
135
YL
OH
134
129
NS
9.1
(0.02)
1.6
NS
YL = young and low (n = 73), YH = young and high
(n = 83), OL = old and low (n = 63), OH = old and high
(n = 64).
axis is low education/80-89 years; low education/70-79
years; low education/60-69 years; high education/8089 years; high education/70-79 years; high education/
60-69 years. From this general analysis, education
would appear to have a greater impact on cognitive
change than age.
In order to investigate this finding further, KruskalWallis tests were then conducted for each cognitive
function for the following age and education groups:
young (60-75 years)/low education (groups 1 + 2);
young/high education (groups 3 + 4); old (76+ years)/
low education; old/high education (see Table I).
Different age groups were chosen to obtain roughly
equal group size. This analysis demonstrated that the
effects of age and education vary according to the
cognitive domain examined. The education effect was
found to be greater for secondary memory and language
(RT) tasks only. It is interesting to note that the effect is
reversed with respect to visuospatial (RT) tasks, with
the young high-education group deteriorating more
rapidly than either the old group with high education or
the young group with low education.
Differences over the two waves for individual
cognitive tests are given in Table II. Generally speaking, it can be seen that education has its greatest effect at
higher ages on most tests. The Kruskal—Wallis test was
used to evaluate the significance of differences between
the two education groups with age. A decrease in
performance in the low-education group is seen with
age in wave one in attention, language (RC), primary
memory and visuospatial functioning. In the higheducation group, attention and primary memory also
decrease with age, there is no effect on the language
scores and only visuospatial reaction time (rather than
response accuracy) is seen to fall with age. A year later,
a more exaggerated version of the same pattern is
observed. At the second wave the high-education group
is now showing significant change on language (RT)
but not on language (RC).
Wilcoxon tests conducted between the two education
levels at each age show no significant differences in the
lowest age group apart from language (RT) where the
low-education group is already doing worse. Significant
differences are found on all tests in the age group 70—79
except attention and language (RT) in wave one, and
primary memory and implicit memory in wave two.
A repeated measure analysis of variance by age group
and global levels of education (Table I) indicated first a
time effect, showing an overall decline over the year in
attention (p < 0.001), but with some improvement in
primary (p < 0.001) and implicit memory (p < 0.001).
Second, an age-time interaction was observed with
significant decreases in attention (p < 0.05) and secondary memory (p < 0.001) being observed in the
oldest groups, and some improvement in secondary
memory in the youngest. Third, a significant education—time interaction could be observed with language
(RT) scores (p < 0.05) improving in the high education
group and decreasing in the low education group.
Finally, an education—time—age interaction was
EDUCATION LEVEL AND THE COURSE OF COGNITIVE DECLINE
395
Table II. Mean scaled scores (and standard deviations) by age group and education for each cognitive function
Wilcoxon-Z
60-69 years
Cognitive
function
70-79 years
^ 80 years
Kruskal-Wallis
High
(n = 44)
Low
(n = 63)
High
(n = 64)
Low
(n = 33)
(n = 40)
High
(n = 39)
85(13)
84(12)
82 (15)
84 (14)
74 (20)
77 (18)
8.1 (0.02)
3.1 (0.2)
64 (20)
1.4(0.5)
5 (0.07)
70(16)
13 (0.001)
1.8(0.4)
53 (16)
8.2 (0.02)
9.4 (0.01)
62 (20)
2.7 (0.3)
0.9 (0.6)
46(17)
0.6 (0.7)
0.08 (0.95)
Low
Low
High
Wl
Attention
NS
Language
RT
RC
69(17)
NS
72(17)
NS
72 (14)
59(18)
76(12)
62 (17)
NS
62(18)
45(11)
64(19)
RT
RC
71 (18)
44(13)
55 (19)
46 (15)
76 (14)
67 (14)
62 (17)
83(11)
69(21)
51 (16)
NS
66(17)
55 (20)
NS
46(13)
49(18)
NS
72 (14)
•
54 (20)
63 (20)
17 (0.0002)
8.9 (0.01)
75(17)
8.6 (0.01)
5.4 (0.07)
70(21)
11 (0.005)
11 (0.003)
65 (20)
2.5 (0.3)
6.9 (0.03)
70(18)
11 (0.005)
2.25 (0.3)
62 (18)
5.4 (0.07)
3.4 (0.2)
58 (18)
10 (0.007)
6.7 (0.03)
46 (17)
0.25 (0.9)
2.2 (0.3)
68 (15)
18 (0.0001)
5.5 (0.06)
72(15)
9.6 (0.008)
11 (0.004)
•
77 (14)
•
NS
60(17)
•••
NS
NS
76(21)
74(11)
•••
NS
Visuospatial
67 (17)
••
51 (16)
64(19)
NS
•••
NS
implicit
73(15)
NS
NS
Memory
primary
secondary
69(17)
NS
63 (24)
•
W2
Attention
82(19)
81 (16)
Language
RT
RC
66 (20)
75(13)
•
73 (19)
75(13)
NS
Memory
primary
secondary
63 (21)
69 (18)
51 (14)
68(16)
RT
RC
75(11)
46(16)
NS
57 (18)
59 (16)
51 (17)
73(13)
70(15)
71 (13)
75(15)
58 (23)
••
62(18)
53(19)
•
64(17)
51 (20)
NS
49 (14)
51 (20)
NS
75(11)
•
82(11)
58 (23)
NS
NS
NS
76(19)
73(12)
•
NS
Visuospatial
64(17)
• ••
68 (17)
64 (29)
NS
NS
NS
implicit
84(13)
NS
NS
65 (18)
76 (18)
•
NS
59(18)
••
80(12)
••
66(17)
NS
NS = not significant; • < 0.05; • • < 0.01; • • • < 0.001.
observed in relation to primary memory (p < 0.05),
with improvement being observed for the youngest and
oldest for the highly educated group.
Discussion
Examination of changes over one year in a cohort of
elderly persons with early signs of cognitive deterioration suggests that the impact of education is complex. It
would seem that while education does play a significant
role in the evolution of cognitive deficit, its impact
varies greatly according to the age of the subject at onset
of the impairment and the type of cognitive function.
Age would seem to have a far more significant effect
on rate of cognitive deterioration over a 1-year period in
persons with low education. With regard to premorbid
intelligence levels, young adult IQ seems to have little
effect on the rate of cognitive deterioration in the
younger elderly, but begins to exert an important
protective role over the age of 75. It might be suggested
therefore that IQ level becomes crucial at the age where
neurobiologists note a significant drop in neuronal
reserve capacity [25, 26]. The amount of variance
explained by the regression analyses conducted here are
admittedly very small, but the time over which change
is observed is very short and significance is none the less
obtained.
With regard to the question of whether age or
education level is the most significant determinant of
cognitive change, the study suggests that over the brief
396
D. LEIBOVICI ET AL.
time period examined here, education may have a more
important impact on changes in secondary memory and
language functioning, but that elsewhere age is the
more important factor.
Elderly persons with a high level of education appear
to show greatest resistance to change but only on tests
with a high learned component—that is, tests of
language and secondary memory. The results also
suggest that on cognitive functions such as attention,
implicit memory and visuospatial analysis, which might
be postulated to have a higher 'nature' rather than
'nurture' component, level of education seems to make
relatively little difference to the rate of change over
time. These latter functions have been attributed to
older nervous system structures. Corsini et al. [27] have
demonstrated, for example, that in ontogeny, visual
functions appear to precede verbal functions. If this is
so it would suggest that 'memory training' programmes
designed for elderly people might have a significant role
to play in helping to maintain skills with a high learned
component but may have little impact on cognitive
abilities, which are largely determined by genetic and
physiological factors and are not significantly modified
by learning.
These results tend to support a modification of our
third hypothesis, that is that elderly persons with
higher education have developed and are perhaps also
better able to continue developing explicit verbal
cognitive skills in the face of deterioration in other
areas. In the higher education groups, a higher average
score across cognitive tasks will therefore be masking a
greater underlying cross-task variability.
7.
8.
9.
10.
11.
12.
13.
14.
15.
A cknowledgemen ts
The authors wish to thank the Fondation de France, the
French Social Security (CNAM-TS), and the Direction
Generale de la Sante for their financial support of this project
and Joelle Faurobert and Catherine Gonzalez, the project
interviewers, for their important role in data collection.
16.
References
1. Holzer CE, Tischer GL, Leaf PJ, Myers J. An
epidemiologic assessment of cognitive impairment in a
community population. Research in community and mental
health. Greenwich: JAI Press, 1984.
2. Fillenbaum GG, Hughes DC, Heyman A. Relationship
of health and demographic characteristics to MiniMental-State-Examination score among community
residents. Psychol Med 1988;18:719-26.
3. Escobar JI, Burman A, Karno M, Forsythe A. Use of the
MMSE in a community population of mixed ethnicity:
cultural and linguistic artefacts. J Neurol Ment Disord
1982; 174:607-14.
4. Scherr PA, Albert MS, Funkenstein HH. Correlates of
cognitive function in an elderly community population.
Am J Epidemiol \988;l28:\084-10\.
5. Colsher PL, Wallace RB. Longitudinal application of
cognitive function measures in a defined population of
community-dwelling elders. Ann Epidemiol 1991 ;1:215—
30.
6. Farmer ME, Kittner SJ, Rae DS, Bartko JJ, Regier DA.
Education and change in cognitive function: the
18.
17.
19.
20.
21.
22.
23.
Epidemiologic Catchment Area Study. Ann Epidemiol
1995;5:l-7.
Ott A, Breteler MMB, Vanharskamp F, Claus JJ,
Vandercammen TJM, Grobbee DE. Prevalence of
Alzheimer's disease and vascular dementia: association
with education: the Rotterdam study. BMJ 1995;
310:970-3.
Stern Y, Gurland B, Tatemichi TK, Tang MX, Wilder
D, Mayeux R. Influence of education and occupation on
the incidence of Alzheimer's disease. JAMA 1994;
271:1004-10.
Mortel KF, Meyer JS, Herod B, Thornby J. Education
and occupation as risk factors for dementias of the
Alzheimer and vascular types. Dementia 1995;6:55-62.
McDowell I, et al. 1994 The Canadian Study of Health
and Aging: risk factors for Alzheimer's disease in Canada.
Neurology 1995;44:2073-80.
Mortimer JA, Graves AB. Education and other socioeconomic determinants of dementia and Alzheimer's
disease. Neurology 1993;43:S4, S39-44.
Dartigues JF. The Paquid research program on the
epidemiology of dementia: methods and initial results.
Rev Neurol 1991;147:225-30.
Zhang M, Katzman R, Salmon D, Jin H, Cai G, Wang Z.
The prevalence of dementia and Alzheimer's disease in
Shanghai, China: Impact of age, gender and education.
Ann Neurol 1990;27:428-37.
Kay DWK, Beamish P, Roth M. Old age mental
disorders in Newcastle upon Tyne: Part II. A study of
possible social and medical causes. Br J Psychiatry
1964;110:668-82.
Fratiglioni L, Grut M, Forsell Y, Viitanen M, Grafstrom
M, Holmen K, et al. Prevalence of Alzheimer disease and
other dementias in an elderly urban population: relationship with age, sex and education. Neurology 1991;
41:1886-92.
Beard CM, Kokmen E, Offord KP, Kurland LT. Lack of
association between Alzheimer disease and education,
occupation, marital status or living arrangement. Neurology 1992;42:2063-8.
Teri L, McCurry SM, Edland SD, Kukull WA, Larson
EB. Cognitive decline in Alzheimer's disease: a longitudinal investigation of risk factors for accelerated
decline. J Gerontol series A 1995;S0:M49-55.
Ritchie K, Leibovici D, Ledesert B, Touchon J. A
typology of sub-clinical senescent cognitive disorder.
Br J Psychiatry 1996;168:470-6.
Moritz DJ, Petitti DB. Association of education with
reported age of onset and severity of Alzheimer's disease
at presentation: implications for the use of clinical
sample. Am J Epidemiol 1993;137:456-62.
Ritchie K, Fuhrer R. A comparative study of the
performance of screening tests for senile dementia using
receiver operating characteristics analysis. J Clin Epidemiol 1992;45:627-37.
Ritchie K, Fuhrer R. La mise au point et la validation en
France d'un test de depistage de la demence senile. Rev
Geriatr 1994;19:233-42.
Ritchie K, Fuhrer R. The validation of an informant
screening test for irreversible cognitive decline in the
elderly: performance characteristics within a general
population sample. Int J Geriatr Psychiatry (submitted
1995).
Ritchie K, Allard M, Huppert FA, Nargeot C, Pinek B,
Ledesert B. Computerized cognitive examination of
the elderly: the development of a neuropsychological
EDUCATION LEVEL AND THE COURSE OF COGNITIVE DECLINE
examination for clinic and population use. Int J Geriatr
Psychiatry 1993;8:899-914.
24. Nelson HE. National Adult Reading Test, MART.
London: NFER, 2nd ed. 1991.
25. Terry RD, DeTeresa R, Hanse LA. Neocortical cell
counts in normal human adult aging. Ann Neurol
1987;21:530-7.
26. Mouritzen DA. The density of neurones in the human
hippocampus. Neuropathol Appl Neurobiol 1979;5:24953.
397
27. Corsini DA, Jacobus KA, Leonard SD. Recognition
memory of preschool children for pictures and words.
Psychonomic Sci 1969;16:192-3.
Authors' address
INSERM Equipe 'Vieillissement CognitiP, CRLC Val
d'Aurelle, F-34298 Montpellier Cedex 5, France
Received in revised form 3 April 1996