Late-Life Depression as a
Possible Predictor of Dementia
Cross-Sectional and Short-Term Follow-Up Results
Robert van Reekum, M.D., FRCPC, Martine Simard, Ph.D.
Diana Clarke, B.Sc., Malcolm A. Binns, M.Sc.
David Conn, M.B., FRCPC

The authors explored cognitive functioning of a group of elderly subjects with depression. The group as a whole, and, in particular, the late-onset group (LOD), demonstrated cognitive impairment on the Mattis Dementia Rating Scale (MDRS). Subgroup
differences were significant at P‫ .400.0ס‬This between-group difference was not seen
when age and level of education were controlled. In the LOD group, 47.5% (vs. 31.5%
of the early-onset group [P‫ ,)]520.0ס‬scored below the cutoff for dementia. Age-at-onset
status in a logistic regression model predicted MDRS category, and treatment of the
depression had little effect on cognition. Results support the hypothesis that late-life
depression, particularly LOD, is associated with cognitive impairment that may represent early AD. (Am J Geriatr Psychiatry 1999; 7:151–159)

The era of new treatments for Alzheimer’s disease
(AD) and related disorders is upon us.1–4 Even greater
hope, however, is presented by the possibility that this
disorder may actually be prevented or delayed. This
strategy will require the ability to predict an increased
risk of future occurrence in individuals who do not currently meet the criteria for AD.5,6 One possible predictor of future AD may be late-life depression.
Our understanding of the relationship between preAD depression and AD has evolved over the last two
decades. The early studies of Wells7 and Caine8 suggested that the “pseudodementia” of late-life depression
was reversible with treatment for the depression, and
hence this group was not expected to develop demen-

tia. Kral’s9,10 naturalistic data suggested instead that
late-life depression in the presence of cognitive impairment may be an early or “prodromal” sign of
impending AD. Alexopoulos et al.11 followed a similar
clinical group for approximately 34 months and found
that “irreversible dementia” (i.e., DSM-III-R12 dementia
in the absence of depression) occurred in 10 of 23 subjects with initially reversible dementia (43%) and in 4
of 34 subjects (12%; PϽ0.01) who were initially only
depressed. The risk ratio was 4.69. Alexopoulos’ data
were more systematic than the earlier research and included a control group. However, the outcome assessment was not blinded to initial status, and the two
groups differed in terms of the severity of depression at

Received February 24, 1998; revised July 5, 1998; accepted August 31, 1998. From Baycrest Centre for Geriatric Care, North York, Ontario, Canada.
Address correspondence to Dr. van Reekum, Department of Psychiatry, Baycrest Centre for Geriatric Care, 3560 Bathurst St., North York, Ontario,
M6A 2E1 Canada. e-mail: robert.vanreekum
utoronto.ca
Copyright ᭧ 1999 American Association for Geriatric Psychiatry

Am J Geriatr Psychiatry 7:2, Spring 1999

151

Depression and Dementia
baseline. Other research13,14 began to suggest that latelife depression confers a high risk for dementia even in
the absence of objectively verifiable significant cognitive impairment.
Jorm et al.’s15 initial analysis of pooled data from
available case-control studies found a relative risk (RR)
of 1.82 for developing AD in those with a pre-AD history
of depression, and this finding was supported by Speck
et al.16 These data appear to suggest that a lifetime history of depression is associated with increased risk of
AD. However, when Jorm et al.15 analyzed data by subgroup, they found that the RR in the late-onset depression (LOD) subgroup (cutoff age not given) was 2.44,
whereas that for the early-onset (EOD) group revealed
no increased risk of AD (RR‫ .)10.1ס‬Other data17–19 also
suggest that it is the LOD subgroup, in particular, who
are at increased risk of developing AD.
One question that remains to be addressed is
whether identified predictors of AD are simply markers
for the disease, or whether they are signs of the disease
process already at work, but not yet at the level of causing the classical signs and symptoms of AD. There is
growing evidence20–22 that changes in the brain are occurring before the full expression of AD, suggesting that
a “preclinical” or “prodromal” condition exists. Hence,
it is at least possible that late-onset depression is an early
manifestation of AD, rather than simply a marker of future disease onset.
In summary, there is considerable preliminary evidence for an association between depression and subsequent onset of dementia, especially when the depression 1) occurs in elderly patients; 2) is associated with
reversible cognitive impairment; and 3) has its first onset later in life. However, significant limitations to the
data exist, given that most of the evidence is derived
from naturalistic follow-up studies or case-control studies. We do not yet know the incidence of AD in these
populations, and, therefore, larger populations of depressed elderly patients will ultimately need to be followed over long periods in an appropriate prospective
cohort study. In the interim, cross-sectional and shortterm follow-up data may be valuable if, in fact, late-life
depression is an early manifestation of AD, because we
might expect there to be early evidence of cognitive
impairment even in this prodromal stage.
The hypothesis behind the current research is that
late-life depression, especially of late onset and/or with
reversible cognitive impairment, is associated with dementia, and indeed is likely to be a prodromal feature

152

of AD for a large subgroup of the LOD population. Corollaries of this hypothesis that may be addressed with
cross-sectional data might include the following: 1) Cognitive impairment is likely to be frequent in elderly patients with depression and may be more frequent or
more severe in those with LOD; 2) The cognitive impairment seen is likely to resemble that seen in early
AD; and 3) Cognitive impairment in the LOD group, or
in the group with cognitive impairment equivalent to
those with dementia, may be less likely to improve with
treatment for the depression than is the impairment in
the EOD group. The objective for this research is to add
to the available knowledge regarding the relationship
between pre-AD depression and subsequent onset of
AD, in the hope that these data will ultimately assist in
developing clinical markers for predicting AD, which,
in turn, will facilitate the testing of AD prevention strategies. This knowledge may also contribute to our understanding of the pathophysiology of AD and to our
knowledge regarding the outcome of late-life depression.

SUBJECTS AND METHODS
Subjects
The Department of Psychiatry at Baycrest Centre
for Geriatric Care, in Toronto, Canada, offers a day hospital for treating elderly persons with depression and
related disorders. Multidisciplinary treatment is provided, including medical and psychiatric care for mood
disorders, and a host of therapies, in an intensive program lasting an average of 119 days. Clinical data at admission to and discharge from the program have been
systematically collected for more than 10 years and
coded onto a computerized database. This is the same
database used by Steingart and Herrmann;19 however
the sample is now much larger, and the current study
did not enforce the requirement that subjects needed
to have been previously hospitalized for depression.
There were 667 possible subjects in the database,
and, of these, 452 had a diagnosis of “major depressive
episode or disorder” (as assessed by the attending psychiatrist per the DSM manuals appropriate to the time
period). However, some of these records represented
repeat admissions; after taking data from only the first
admissions, there remained 331 subjects, and 264 subjects remained after we excluded those with a history

Am J Geriatr Psychiatry 7:2, Spring 1999

van Reekum et al.
of central nervous system disease, including dementia
of any type (as diagnosed by the admitting psychiatrist).
This was narrowed to 245 subjects for whom data on
age at onset of first episode of depression were available. This represents the study sample; however, the
sample size for some of the analyses is smaller because
Mini-Mental State Exam (MMSE23) and Mattis Dementia
Rating Scale (MDRS24) data were not available (largely
because of the clinical nature of the database) for all
subjects (n‫ 922ס‬for admission MMSE; n‫ 521ס‬for discharge MMSE; and n‫ 191ס‬for admission MDRS).
Methods
The measures of cognitive function used by the day
hospital include the MMSE and the MDRS. The Hamilton
Rating Scale for Depression (Ham-D)25 is used as a measure of the severity of depressive symptoms. All of these
data, as well as demographic and other health-related
data, are collected at admission into the program by the
day hospital staff. The MMSE is collected at discharge
as well as at admission. The MDRS is administered by a
psychologist, and the other scales by healthcare professionals (nurse, social worker, or occupational therapist).
The analyses used are described in the title of each
table. All comparisons were evaluated for statistical significance, adjusting for multiple comparisons with the
Bonferroni correction (0.05/number of comparisons in
the table). For example, there are eight comparisons in
Table 1; only those differences with PϽ0.006 would be
accepted as significant.

RESULTS
Table 1 presents the demographic and clinical characteristics of the groups; it shows that the subjects are,
on average, in their mid-70s, and 73.9% are women. Less
than 20% of the group had more than high school training. A significant difference in age and a trend-level difference in level of education were found, with the LOD
group being both older and less well educated. The
clinical characteristics at admission that are relevant to
depression in the two groups were similar, with no significant differences in length of stay, Ham-D score at
admission, rate of major depression as the primary Axis
I diagnosis (note that those subjects who had another
Axis I disorder listed as the primary diagnosis received
a diagnosis of depression as a secondary diagnosis), and

Am J Geriatr Psychiatry 7:2, Spring 1999

frequency of Axis II diagnoses. However, there was a
trend-level difference seen in Ham-D scores at discharge, with the LOD group having more symptoms at
discharge (mean Ham-D‫ )3.11ס‬than the EOD group
(mean Ham-D‫ ;4.9ס‬P‫ .)50.0ס‬Both groups averaged
well below the cutoff score for depression (17) on the
Ham-D at discharge.
Table 2 presents the cognitive functioning of the
study sample, and the EOD and LOD subgroups, as measured by the MMSE at admission and discharge and by
the MDRS at admission. The LOD group shows lower
functioning on the MMSE at baseline, and there is a
trend-level difference in MMSE score at discharge; however the clinical significance of these small differences
is questionable. The MDRS, on the other hand, shows
a clinically and statistically significant difference of
nearly seven points, with the LOD group averaging below the 123 cutoff score for dementia.24 As described
in the note to the table, the change in MMSE from admission to discharge was statistically significant, at
PϽ0.05 for the entire group and for the EOD group, but
not for the LOD group. The magnitude of the differences was small for all groups, with the largest magnitude of average difference seen in the LOD group at just
over one point. Because of the difference in age and
level of education between the two groups, the EODvs.-LOD comparisons were repeated (Table 3; ANCOVAs) while controlling for these covariates; no significant differences remained, and the LOD group’s mean
adjusted MDRS score no longer remained in the dementia range.
Table 4 shows that the groups differed significantly
on the orientation subscale of the MMSE and on the
conceptualization subscale of the MDRS. There were
strong trend differences between the MMSE Language,
and the MDRS Memory subscales and a weak trend difference on the construction subscale of the MDRS. For
all of these comparisons, the LOD group was more impaired than the EOD group. These differences were not
seen when age and level of education were covaried
(Table 5).
Table 6 shows that nearly 40% of the entire group
scored below 123 on the MDRS. The two subgroups
differed (P‫ )520.0ס‬in the frequency of scoring below
this threshold, with the LOD group having a larger proportion of subjects (n‫ ,74ס‬or 47.5%) scoring below
threshold than the EOD group (n‫ ,92ס‬or 31.5%). The
odds ratio (OR) for this comparison was 1.96.
Table 7 shows that all the MMSE scores for the

153

Depression and Dementia
groups with MDRS scores in the dementia range are
about 25. There is very little change in these scores over
the course of admission to the Day Hospital.
TABLE 1.

Logistic regression analyses, with MDRS status
(score greater than or less than 123) as dependent variable were also completed. The first analysis included all

Comparison of demographics between the early-onset-of-depression (EOD) and the late-onset-of-depression (LOD)
samples (Student’s t-test and chi-square)
Onset of Depression
Total Group (TG) EOD and LOD
(N؄245)

EOD
(n؄114)

LOD
(n؄131)

75.28‫31.7ע‬
118.69‫24.36ע‬
19.89‫48.6ע‬
10.30‫95.6ע‬

72.19‫49.6ע‬
125.75‫25.06ע‬
20.03‫90.7ע‬
9.35‫94.6ע‬

77.96‫71.6ע‬
112.53‫44.56ע‬
19.76‫36.6ע‬
11.28‫85.6ע‬

64 (26.1%)
181 (73.9%)

28 (24.6%)
86 (75.4%)

36 (27.5%)
95 (72.5%)

7 (2.9%)
90 (36.7%)
45 (18.4%)
68 (27.8%)
10 (4.1%)
20 (8.2%)
5 (2.0%)

1 (0.9%)
36 (31.6%)
24 (21.1%)
34 (29.8%)
6 (5.3%)
10 (8.8%)
3 (2.6%)

6 (4.6%)
54 (41.2%)
21 (16%.0)
34 (26.0%)
4 (3.1%)
10 (7.6%)
2 (1.5%)

240 (98.0%)
5 (2.0%)

110 (96.5%)
4 (3.5%)

130 (99.2%)
1 (0.8%)

217 (88.6%)
28 (11.4%)

104 (91.2%)
10 (8.8%)

113 (86.3%)
18 (13.7%)

Age, years, mean‫ע‬SD
Length of stay, days, mean‫ע‬SD
Ham-D, baseline, mean‫ע‬SD
Ham-D, discharge, mean‫ע‬SD
Sex, n (%)
Male
Female
Level of education, n (%)
No formal education
Elementary, partial or complete
High school or technical school, partial
High school, complete
College/University, partial
College/University, complete
Graduate studies, partial or complete
DSM-IV diagnoses, n (%)
Primary Axis I
Major depression
Othera
Axis II
No diagnosis
Personality disorderb

P
0.001
0.100
0.759
0.048
0.604

0.062

0.187

0.224

Note: EOD: onset Ͻage 60; LOD: onset Նage 60. SD‫ס‬standard deviation.
a
includes dysthymic disorder (TG: 0.4%; EOD: 0.9%), adjustment disorder (TG: 0.4%; EOD: 0.9%), delusional disorder (TG: 0.4%; EOD: 0.9%),
bereavement (TG: 0.4%; EOD: 0.9%), and generalized anxiety disorder (TG: 0.4%; LOD: 0.8%).
b
includes borderline personality disorder (TG: 0.8%; EOD: 0.9%; LOD: 0.8%), histrionic personality disorder (TG: 3.7%; EOD: 4.4%; LOD:
3.1%), narcissistic personality disorder (TG: 2.1%; EOD: 0.9%; LOD: 3.1%), dependent personality disorder (TG: 2.1%; EOD: 0.9%; LOD: 3.1%),
avoidant personality disorder (TG: 0.4%; LOD: 0.8%), and ‘‘diagnosis deferred’’ (TG: 1.2%; LOD: 2.3%).

TABLE 2.

Comparison of cognitive functioning between
early-onset-of-depression (EOD) and late-onset-ofdepression (LOD) groups (ANOVA)
Total
Group

Onset of Depression
EOD

LOD

P

Mini-Mental State Exam, baseline
N/n/n
229
110
119
Mean‫ע‬SD 26.30‫62.3ע‬
27.01‫84.2ע‬
25.81‫500.0 67.3ע‬
Mini-Mental State Exam, discharge
N/n/n
125
71
54
Mean‫ע‬SD 27.58‫54.2ע‬
27.96‫13.2ע‬
27.07‫040.0 75.2ע‬
Mattis Dementia Rating Scale, baseline
N/n/n
191
92
99
Mean‫ע‬SD 124.07‫400.0 90.81ע19.021 94.21ע74.721 49.51ע‬
Note: EOD: onset Ͻage 60; LOD: onset Նage 60. SD‫ס‬standard
deviation.
The difference between MMSE scores at baseline and discharge
was found to be statistically significant for the total group (t ‫ס‬
‫ ;72.2מ‬PϽ0.05) and the EOD group (t ‫ ;89.2מ ס‬PϽ0.05) but not
for the LOD group (t ‫ ;91.0מ ס‬PϾ0.05).

154

TABLE 3.

Comparison of cognitive functioning between
early-onset-of-depression (EOD) and late-onset-ofdepression (LOD) groups with age and level of
education as covariates (ANCOVA)
Onset of Depression
EOD

LOD

Mini-Mental State Exam (MMSE), baseline
n
110
119
Adjusted mean
26.79
27.37
Mini-Mental State Exam (MMSE), discharge
n
71
54
Adjusted mean
27.51
27.52
Mattis Dementia Rating Scale (MDRS), baseline
n
92
99
Adjusted mean
127.72
127.81

P

0.271

0.980

0.969

Note: EOD: onset Ͻ age 60; LOD: onset Ն age 60. Multivariate
analysis of covariance revealed no significant difference between
the two groups (EOD vs. LOD) with respect to performance on the
MDRS and the MMSE (Wilks F[1, 98]‫ ;89.0ס‬PϾ0.05).

Am J Geriatr Psychiatry 7:2, Spring 1999

van Reekum et al.
possible subjects in the database, including those with
CNS disease (including dementia), those with and without an admitting diagnosis of depression, and those returning to the program for repeat treatment (maximum
of four repeats); 667 possible subjects were available,
but only 353 had complete data. Level of education
(R‫ ;822.0ס‬PϽ0.0001; OR: 0.60), age decile (R‫;671.0ס‬
PϽ0.0001; OR: 2.16), presence of CNS disease
(R‫ ;441.0ס‬P‫ ;5000.0ס‬OR: 2.78), and age at onset of
depression ([EOD or LOD] R‫ ;060.0ס‬P‫ ;50.0ס‬OR:
1.63) all entered the model. Severity of depression
(Ham-D score greater than or less than 17) and gender
did not enter the model. When this analysis was repeated in the absence of age and education as covariates, there were 356 possible subjects, and the model
included only age at onset of depression (R‫;961.0ס‬
P‫ ;1000.0ס‬OR: 2.44) and presence of CNS disease
(R‫ ;620.0ס‬P‫ ;620.0ס‬OR: 1.79). A third logistic regression analysis controlled for age effects by creating a new
dependent variable based on data from a community
sample of elderly subjects,26 which suggested that the
effects of age on MDRS score could be predicted by the
equation X‫ס‬MDRS score ‫( 80.0 מ‬age minus mean age
of sample). In this analysis, 355 possible subjects were
included (same criteria as that for the previous two regression analyses), and age-at-onset category (R‫;702.0ס‬
PϽ0.0001) and education level (R‫ ;302.0ס‬PϽ0.0001)
were the only two variables to enter the model.
TABLE 4.

Comparison of cognitive functioning (per
subscales of the MMSE and MDRS at baseline)
between early-onset-of-depression (EOD) and lateonset-of-depression (LOD) groups, meanؓSD
(ANOVA)
Total
Group

TABLE 5.

Comparison of cognitive functioning (per the
subscales of the MMSE and MDRS at baseline)
between early-onset-of-depression (EOD) and lateonset-of-depression (LOD) groups, adjusted
means, with age and level of education as
covariates (ANCOVA)
Onset of
Depression
EOD

LOD

P

Mini-Mental State Exam (MMSE), baseline
Orientation
9.65
Registration
2.97
Calculation
3.67
Language
7.30
Recall
2.08
Construction
0.71

9.43
3.00
3.87
7.19
2.11
0.69

0.113
0.072
0.440
0.378
0.838
0.748

34.94
31.24
5.41
32.55
19.88

0.471
0.470
0.997
0.146
0.513

Mattis Dementia Rating Scale (MDRS), baseline
Attention
34.70
Initiation and Perseveration
30.64
Construction
5.41
Conceptualization
33.68
Memory
20.29

Note: EOD: onset Ͻage 60; LOD: onset Նage 60. Multivariate
ANCOVA revealed no significant difference between the two groups
(EOD vs. LOD) with respect to performance on the subscales of the
MDRS and the MMSE (Wilks F[1, 184]‫ ;29.0ס‬PϾ0.05).

TABLE 6.

Chi-square comparison of the proportions of
early-onset-of-depression (EOD) vs. late-onset-ofdepression (LOD) patients with dementia as
defined by a Mattis Dementia Rating Scale (MDRS)
score Ͻ123
Onset of Depression
Total Group
(N؄191)

MDRSՆ123
MDRSϽ123

EOD
(n؄92)

LOD
(n؄99)

115 (60.2%)
76 (39.8%)

63 (68.5%)
29 (31.5%)

52 (52.5%)
47 (47.5%)

P
0.025

Note: EOD: onset Ͻage 60; LOD: onset Նage 60. Odds ratio
(OR)‫.69.1ס‬

Onset of Depression
EOD

Mini-Mental State Exam (MMSE), baseline
Orientation
9.56‫45.0ע67.9 09.0ע‬
Registration
2.99‫31.0ע89.2 11.0ע‬
Calculation
3.82‫83.1ע09.3 55.1ע‬
Language
7.24‫87.0ע04.7 48.0ע‬
Recall
2.10‫09.0ע22.2 39.0ע‬
Construction
0.70‫44.0ע57.0 84.0ע‬
Mattis Dementia Rating Scale (MDRS), baseline
Attention
34.73‫18.1ע30.53 85.2ע‬
Initiation and
Perseveration
30.88‫62.5ע04.13 27.5ע‬
Construction
5.39‫89.0ע75.5 23.1ע‬
Conceptualization 33.05‫78.4ע84.43 15.5ע‬
Memory
20.03‫16.3ע29.02 33.4ע‬

LOD

P

9.38‫01.1ע‬
2.99‫09.0ע‬
3.74‫07.1ע‬
7.10‫69.0ע‬
1.99‫49.0ע‬
0.66‫15.0ע‬

0.002
0.148
0.381
0.006
0.418
0.131

34.44‫232.0 11.3ע‬
30.38‫11.6ע‬
5.21‫65.1ע‬
31.70‫47.5ע‬
19.18‫87.4ע‬

0.318
0.090
0.001
0.009

Note: EOD: onset Ͻage 60; LOD: onset Նage 60. SD‫ס‬standard
deviation.

Am J Geriatr Psychiatry 7:2, Spring 1999

TABLE 7.

Comparison of changes in Mini-Mental State Exam
score for early-onset and late-onset (EOD and
LOD) subgroups of the dementia population
(MDRSϽ123), meanؓSD
Mini-Mental State Exam
At Baseline

Total group (EOD and LOD;
N‫)13ס‬
Early-onset depression
(EOD; n‫)51ס‬
Late-onset depression
(LOD; n‫)61ס‬

At Discharge

P

25.21‫10.3ע‬

25.50‫21.3ע‬

0.562

24.93‫76.2ע‬

25.50‫72.3ע‬

0.468

25.44‫84.3ע‬

25.38‫62.3ע‬

0.931

Note: EOD: onset Ͻage 60; LOD: onset Նage 60; MDRS‫ס‬Mattis
Dementia Rating Scale; SD‫ס‬standard deviation.

155

Depression and Dementia
DISCUSSION
A number of potential selection biases exist in regard
to the day hospital. Although Baycrest Centre’s outpatient programs, such as the Day Hospital for Depression, are nondenominational, they tend to attract primarily members of the Jewish community. The day
hospital is a tertiary-care treatment program for elderly
patients with depression and related disorders, and of
course this may further select the population. Also, the
program is intensive, with an expectation of attendance
at most of the 4-day-per-week activities.
The program is geared to those who may best benefit from multidisciplinary treatment, with a strong emphasis on various psychotherapies, and therefore individuals who are felt by the team to have cognitive
impairments that would prevent progress within the
program are excluded. Despite this exclusion, some individuals who were felt to have relatively mild dementia
have been admitted into the program; they were excluded, however, in this study (with the exception of
the regression analyses).
Other limitations of the study include its very brief
period of follow-up, reliance on clinical diagnoses (including the admitting psychiatrist’s lack of diagnosis of
dementia in subjects who were included in the study
and who clearly manifested some cognitive impairment), missing data, and post-hoc group formation. Despite these limitations, this group of Day Hospital for
Depression attenders provided clinical, cross-sectional,
and short-term follow-up data that are broadly supportive of the hypothesis that in some individuals, late-life
depression is a prodrome of dementia. In particular, the
MDRS mean score of 120.9 in the LOD group suggested
that the LOD group as a whole was scoring in the range
that is generally considered to be indicative of dementia,
although age and education contribute to the low score.
Although this study did not have a normalcomparison population, a healthy, community-dwelling
elderly population scored much better on the MDRS
than than did the LOD group in this study.27 Montgomery and Costa28 studied groups of healthy, communitydwelling elderly people (65 to 81 years old) and found
that only 3 of 85 subjects had an MDRS score of 121 or
less. When one considers the selection biases operating
at the level of the Day Hospital admission criteria (cognitively well enough to benefit from the program) and
the study’s exclusion (with the exception of the regression analyses) of individuals with a clinical diagnosis of

156

dementia or other CNS illness, these data become even
more compelling.
However, the need for analyses controlling for age
and level of education was suggested by the evidence
that MDRS subscale scores correlate with age and educational level in normal, community-dwelling subjects.27 The analyses that controlled for these factors
(Table 3) eliminated the differences between the LOD
and EOD groups, and this, of course, weakens support
for the hypothesis that the LOD group, in particular, is
made up of a large proportion of individuals with prodromal AD. However, in this sample, the LOD group
were both older (by more than 5 years (P‫ )100.0ס‬and
tended to be less well educated (P‫ ;60.0ס‬Table 1), suggesting that age, education, and age at onset of depression are related. Given that the incidence of AD varies
with age,29 it may be that another of the effects of age
on the sample is to increase the incidence of prodromal
AD, and it follows that this would have the greatest effect in the LOD group. If so, then controlling for age
would have the effect of masking the impact that a
greater frequency of prodromal dementia would have
in creating differences between the groups.
Similarly, although the lower educational level in the
LOD group would be expected to drag down the MDRS
score, it may also be predicted that the lower educational
level of the LOD group would affect those with prodromal AD, vs. the impact in the EOD group who have prodromal AD, if their higher education “compensates” for
neuronal damage.30 Hence both age and education may
have opposing effects on cognitive assessments in the
groups as a whole (i.e., they may have caused the apparent differences in MMSE and MDRS scores between
groups) vs. their effects in the subgroup of subjects who
were dementing (i.e., in these subjects the effects of age
and education are true effects that would be expected to
be masked if age and education are covaried).
Ultimately, knowledge about the diagnostic status
of the subjects (i.e., with or without dementia) would
be needed to address this issue. Because these data are
unavailable, we report both the covaried and non-covaried analyses herein. However, it is notable that the
differences on the MDRS as a function of age and education, in a population without dementia, as reported
by Schmidt et al.,27 are small. The mean MDRS score in
the group age 50–59 years was 141.8, and that in the
group age 70–79 was 140.8 (SDs approximately 3.0).
Those with 4–9 years of education had a mean MDRS
score of 140.2, and those with college or university

Am J Geriatr Psychiatry 7:2, Spring 1999

van Reekum et al.
education scored 142.3 (SD‫ 2.4ס‬and 1.7, respectively).
Hence, the effects of age and education in normal elderly
subjects are relatively small, and this suggests that the effects of age and education in the current sample are likely
to be weak unless they are associated with a third factor,
such as onset of prodromal dementia. Neither of the subgroups reported on by Schmidt et al.26 scored in the low
range seen in the LOD group in this study.
To further assess the contributions of age, education, and age at onset of depression on MDRS scores in
this study, logistic regression analyses were completed
(not tabulated; see Results). The analysis showed that
for the entire available data-set, age decile, education
level, presence of CNS disease (including dementia),
and age at onset of depression (EOD or LOD) all entered
the model. When the analysis was repeated in the absence of age and education as covariates, age-at-onset
category was the strongest predictor of MDRS status,
and the magnitude of R was similar to that found for
age in the first analysis. Finally, a third logistic regression
analysis controlling for age, using Schmidt et al.’s data,27
again showed that age at onset of depression was the
strongest predictor of MDRS status. These analyses suggest that age-at-onset independently contributes to the
likelihood of scoring below the cutoff score for dementia of 123 on the MDRS.
The nature of the cognitive impairments seen in the
LOD group, as compared with the EOD group and as
assessed by the subscales of the MMSE and MDRS
(when age and level of education were not controlled
for [Table 3]), also supports the hypothesis that the LOD
group is made up of a large proportion of individuals
with prodromal AD. These impairments are seen in individuals who do not have AD, but have subtle cognitive
impairment on testing and who are at risk for developing AD in follow-up. Several studies20,31–37 have demonstrated that impairments in memory, including impairments in orientation, are associated with
subsequent onset of AD. Similarly, data also exist showing that language and abstract reasoning (“conceptualization” on the MDRS) deficits are present in mild stages
of AD and/or predict future onset of AD.31,32,34,36,38 It is
also interesting to note that the cognitive deficits found
in this study are most suggestive of left frontal-temporal
dysfunction, and this is congruent with the left frontaltemporal hypometabolism reported in a previous positron emission tomography (PET) study of AD patients
with depression.39
The proportion of subjects who scored below the

Am J Geriatr Psychiatry 7:2, Spring 1999

MDRS cutoff for dementia, according to the manual for
the scale,26 was large, at nearly 40%, and largest in the
LOD subgroup, at nearly 50% (47.5%), which differed
from the EOD group, at 31.5%. This finding suggests
that nearly half of the LOD group are functioning at a
cognitive level that is typical for early dementia. Age and
education differences between the EOD and LOD group
are likely to contribute to the observed difference in
proportion. The OR of 1.96 suggests that those scoring
below cutoff for dementia were nearly twice as likely
to have LOD than EOD.
Finally, the data describing the effects of treatment
for depression on cognition (as assessed by pre–posttreatment MMSE scores [Table 2]) show that although
the cognitive impairment seen for the groups as a whole
improved somewhat (but at a statistically insignificant
level for the LOD group), the improvement was modest,
suggesting that not all of the impairment seen is “reversible.” This appeared to be the case to an even
greater extent for those subjects who had more severe
impairment (i.e., scored below the dementia cutoff
score on the MDRS [Table 7]), given that even smallermagnitude improvements were seen in these subjects
than were seen for the full groups in Table 2.
Interestingly, the LOD group, with initial MDRS less
than 123, actually demonstrated a very small, statistically insignificant, decline in cognitive functioning as
measured by the MMSE over the course of their treatment. Of course, the MMSE is limited in its sensitivity
and scope of cognitive assessment, and this limitation
may well have limited the ability of the data to demonstrate change. Nonetheless, the psychometric qualities of the MMSE have been reviewed by Thompson et
al.40 as being adequate for research purposes “as a brief
screening instrument to detect cognitive impairment”
(p.302), and Nelson et al.41 reviewed the evidence that
the MMSE can record cognitive improvement, when it
exists, upon treatment for depression. Hence, the lack
of change on the MMSE in the LOD group, particularly
those scoring positive for dementia on the MDRS at
admission, suggests that this group did not improve
cognitively with treatment for the depression, and this
finding is again consistent with the hypothesis that subjects in this group are already evidencing prodromal
dementia.
Overall, the results of this study, in combination
with the available literature, suggest that LOD may well
represent prodromal AD for at least some of the individuals assessed. A reasonable hypothesis is that the

157

Depression and Dementia
presence of particular cognitive impairments at a severe
enough level in those with LOD may identify a subgroup of the LOD population who are at even greater
risk of having prodromal AD. The nature of cognitive
testing done in the early studies on late-life depression
is relatively limited, and the MMSE and the MDRS are
not the equivalent of a reasoned and thorough neuropsychological assessment. Hence, another limitation of
the current data is that we do not know with any certainty which cognitive markers, when combined with
LOD, are most likely to be predictive of AD. Hence, a
reasonable approach for future studies would be to
combine the data regarding cognitive predictors of
AD20,31–38 with the data on LOD. It is likely that those
individuals with LOD and persistent memory, language,
and conceptualization impairments are at very high risk
for the development of AD. Furthermore, it may be that
this risk is even greater in those with other signs of CNS
disease13 and in those with APO-E ⑀4 alleles (especially
those homozygous for this allele).

Clearly, a prospective cohort study examining this
hypothesis will ultimately be required, but it is hoped
that the data presented here will stimulate interest in
such a study and that it will help to shape the nature of
the cohorts and the testing paradigms used. Ultimately,
it is hoped that these data will take us one step closer
to identifying high-risk populations in whom preventive
strategies designed to delay or prevent AD can be tested
efficiently and be used clinically with minimum risk and
cost.
We gratefully acknowledge the efforts of the Psychiatry Day Hospital staff at Baycrest Centre and Dr.
A. Steingart, for creating and maintaining the database used in this study.
Dr. Simard is supported by a grant from the Alzheimer’s Society of Canada and the Kunin-Lunenfeld
Applied Research Unit of Baycrest Centre. Dr. van
Reekum is supported by the Kunin-Lunenfeld Applied
Research Unit of Baycrest Centre.

References
1. Black SE: Alzheimer’s: Update on new drug treatments. Canadian
Journal of Continuing Medical Education 1994; 6:69–82
2. Schneider LS, Tariot PN: Emerging drugs for Alzheimer’s disease:
mechanisms of action and prospects for cognitive enhancing medications (review). Med Clin North Am 1994; 78:911–934
3. van Reekum R, Black SE, Conn D, et al: Cognition-enhancing drugs
in dementia: a guide to the near future. Can J Psychiatry 1997;
42:35S–50S
4. Wertheimer J: Clinical assessment of cerebro-active drugs, in Facts
and Research in Gerontology, Series IV: Dementia and Cognitive
Impairments. Edited by Fitten LJ. New York, Springer, 1994, pp
161–170
5. Davis RE, Doyle PD, Carroll RT, et al: Cholinergic therapies for
Alzheimer’s disease: palliative or disease altering? (review). Arzneimittel-Forschung 1995; 45:425–431
6. Thal LJ, Carta A, Doody R, et al: Prevention Protocols for Alzheimer
Disease. Position paper from the International Harmonization of
Dementia Drug Guidelines. Alzheimer Dis Assoc Disord 1997;
11(suppl.3):46–49
7. Wells CE: Pseudodementia. Am J Psychiatry 1979; 136:895–900
8. Caine ED: Pseudodementia: current concepts and future directions. Arch Gen Psychiatry 1981; 38:1359–1364
9. Kral VA: The relationship between senile dementia (Alzheimer
type) and depression. Can J Psychiatry 1983; 28:304–305
10. Kral VA, Emery OB: Long-term follow-up of depressive pseudodementia of the aged. Can J Psychiatry 1989; 34:445–446
11. Alexopoulos GS, Meyers BS, Young RC, et al: The course of geriatric depression with “reversible dementia”: a controlled study.
Am J Psychiatry 1993; 150:1693–1699
12. American Psychiatric Association: Diagnostic and Statistical Manual of of Mental Disorders, 3rd Edition, Revised. Washington, DC,
American Psychiatric Association, 1987
13. Reding M, Haycox J, Blass J: Depression in patients referred to a
dementia clinic. Arch Neurol 1985; 42:894–896
14. Devanand DP, Sano M, Tang M-X, et al: Depressed mood and the

158

incidence of Alzheimer’s disease in the elderly living in the community. Arch Gen Psychiatry 1996; 53:175–182
15. Jorm AF, van Duijn CM, Fratiglioni L, et al: Psychiatric history and
related exposures as risk factors for Alzheimer’s disease: a collaborative re-analysis of case-control studies. Int J Epidemiol 1991;
20:S43–S47
16. Speck CE, Kukull WA, Brenner DE, et al: History of depression as
a risk factor for Alzheimer’s disease. Epidemiology 1995; 6:366–
369
17. Alexopoulos GS, Young RC, Meyers BS: Geriatric depression: age
of onset and dementia. Biol Psychiatry 1993b; 34:141–145
18. Steffens DC, Plassman BL, Helms MJ, et al: A twin study of lateonset depression and apolipoprotein E ⑀4 as risk factors for Alzheimer’s disease. Biol Psychiatry 1997; 41:851–856
19. Steingart A, Herrmann N: Major depressive disorder in the elderly:
the relationship between age of onset and cognitive impairment.
Int J Geriatr Psychiatry 1991; 6:593–598
20. Morris JC, Storandt M, McKeel DW, et al: Cerebral amyloid deposition and diffuse plaques in “normal aging”: evidence for presymptomatic and very mild Alzheimer’s disease. Neurology 1996;
46:707–719
21. Reiman EM, Caselli RJ, Yun LS, et al: Preclinical evidence of Alzheimer’s disease in persons homozygous for the E4 allele for apolipoprotein-E. N Engl J Med 1996; 334:752–758
22. Persson G, Skoog I: Subclinical dementia: relevance of cognitive
symptoms and signs. J Geriatr Psychiatry Neurol 1992; 5:172–178
23. Folstein MR, Folstein S, McHugh PR: Mini-Mental State: a practical
method for grading the cognitive state of patients for the clinician.
J Psychiatr Res 1975; 12:189–198
24. Mattis S: Mental status examination for organic mental syndrome
in the elderly patient, in Geriatric Psychiatry. Edited by Bellak L,
Karasu TB. New York, Grune & Stratton, 1976, pp 77–121
25. Mattis S: Dementia Rating Scale, Professional Manual. Psychological Assessment Resources, Odessa, FL, 1973
26. Hamilton M: A rating scale for depression. J Neurol Neurosurg
Psychiatry 1960; 23:56–62

Am J Geriatr Psychiatry 7:2, Spring 1999

van Reekum et al.
27. Schmidt R, Freidl W, Fazekas F, et al: The Mattis Dementia Rating
Scale: normative data from 1,001 healthy volunteers. Neurology
1994; 44:964–966
28. Montgomery KM, Costa L: Neuropsychological test performance
of a normal elderly sample. Presented at the International Neuropsychological Society Meeting, Mexico City, 1983
29. Bachman DL, Wolf PA, Linn RT, et al: Incidence of dementia and
probable Alzheimer’s disease in a general population: the Framingham study. Neurology 1993; 43:515–519
30. Caramelli P, Poissant A, Gauthier S, et al: Educational level and
neuropsychological heterogeneity in dementia of the Alzheimer
type. Alzheimer Dis Assoc Disord 1997; 11:9–15
31. Greene JDW, Hodges JR, Baddeley AD: Autobiographical memory
and executive function in early dementia of Alzheimer type. Neuropsychologia 1995; 33:1647–1670
32. Masur DM, Fuld PA, Blau AD, et al: Predicting development of
dementia in the elderly with the Selective Reminding Test. J Clin
Exp Neuropsychol 1990; 12:529–538
33. Masur DM, Sliwinski M, Lipton RB, et al: Neuropsychological prediction of dementia and the absence of dementia in healthy elderly
persons. Neurology 1994; 44:1427–1432
34. Petersen RC, Smith GE, Ivnik RJ, et al: Memory function in very
early Alzheimer’s disease. Neurology 1994; 44:S67–S72

Am J Geriatr Psychiatry 7:2, Spring 1999

35. Schofield PW, Marder K, Dooneief G, et al: Association of subjective memory complaints with subsequent cognitive decline in
community-dwelling elderly individuals with baseline cognitive
impairment. Am J Psychiatry 1997; 154:609–615
36. Stern Y, Andrews H, Pittman J, et al: Diagnosis of dementia in a
heterogeneous population: development of a neuropsychological
paradigm-based diagnosis of dementia and quantified correction
for the effects of education. Arch Neurol 1992; 49:453–460
37. Tierney MC, Szalai JP, Snow WG, et al: Prediction of probable Alzheimer’s disease in memory-impaired patients: a prospective longitudinal study. Neurology 1996; 46:661–665
38. Jacobs DM, Sano M, Dooneief G, et al: Neuropsychological detection and characterization of preclinical Alzheimer’s disease. Neurology 1995; 45:957–962
39. Lopez OL, Smith G, Becker JT, et al: Depression in Alzheimer’s
disease: exploration of underlying mechanisms using positron
emission tomography (abstract). J Neuropsychiatry Clin Neurosci
1997; 9:703
40. Thompson LW, Gong V, Haskins E, et al: Assessment of depression
and dementia during the late years, in Annual Review of Gerontology and Geriatrics, Vol 7. Edited by Schaie W. New York,
Springer, 1987
41. Nelson A, Fogel BS, Faust D: Bedside cognitive screening instruments: a critical assessment. J Nerv Ment Dis 1986; 174:73–83

159