Depressed
Mood
and
the
Incidence
of
Alzheimer's
Disease
in
the
Elderly
Living
in
the
Community
D.
P.
Devanand,
MD;
Mary
Sano,
PhD;
Ming-Xin
Tang,
PhD;
Stuart
Taylor,
MD;
Barry
J.
Gurland,
MD;
David
Wilder,
PhD;
Yaakov
Stern,
PhD;
Richard
Mayeux,
MD
Background:
It
remains
unclear
whether
depression
in-
creases
the
risk
for
dementia
in
the
elderly.
We
evaluated
the
relationship
between
depressed
mood
at
baseline
and
the
incidence
of
dementia,
particularly
Alzheimer's
dis-
ease,
in
the
elderly
living
in
the
community.
Methods:
A
total
of
1070
elderly
individuals,
aged
60
years
or
older,
were
identified
as
part
of
a
registry
for
de-
mentia
in
the
Washington
Heights
community
of
North
Manhattan,
NY.
In
a
prospective,
longitudinal
design
with
follow-up
for
1
to
5
years,
annual
physician
evaluation
and
neuropsychological
testing
were
used
to
assess
lev-
els
of
cognitive
impairment
and
to
diagnose
dementia.
Depressive
symptoms
were
evaluated
with
the
17-item
Hamilton
Rating
Scale
for
Depression.
Based
on
clinical
considerations
and
a
validity
study,
a
positive
score
for
the
depressed
mood
item
was
used
in
statistical
analy-
ses.
To
confirm
the
results,
the
total
Hamilton
Rating
Scale
for
Depression
score
was
also
evaluated
as
the
"depres-
sion"
variable.
Results:
Of
the
1070
subjects,
218
met
criteria
for
de-
mentia
at
baseline
evaluation.
In
the
852
subjects
with-
out
dementia,
depressed
mood
was
more
common
in
in-
dividuals
with
greater
cognitive
impairment.
In
a
follow-up
study
of
478
of
these
subjects
without
demen-
tia
(mean\m=+-\SD,
2.54\m=+-\1.12
years
of
follow-up),
the
ef-
fect
of
baseline
depressed
mood
on
the
end-point
diag-
nosis
of
dementia
(93%
had
possible
or
probable
Alzheimer's
disease)
was
evaluated
in
a
Cox
propor-
tional
hazards
model.
Depressed
mood
at
baseline
was
associated
with
an
increased
risk
of
incident
dementia
(relative
risk,
2.94;
95%
confidence
interval,
1.76
to
4.91;
P<.001).
This
effect
remained
after
adjustment
for
age,
gender,
education,
language
of
assessment,
Blessed
Memory
Information
and
Concentration
test
scores,
and
Blessed
Functional
Activity
Scale
scores
(relative
risk,
2.05;
95%
confidence
interval,
1.16
to
3.62;
P<.02).
Similar
results
were
obtained
when
the
total
Hamilton
Rating
Scale
for
Depression
score
was
used
as
the
depression
vari-
able,
with
the
use
of
the
same
covariates
(relative
risk,
1.07
per
point
interval;
95%
confidence
interval,
1.02
to
1.11;P<.01).
Conclusions:
Depressed
mood
moderately
increased
the
risk
of
developing
dementia,
primarily
Alzheimer's
disease.
Whether
depressed
mood
is
a
very
early
mani-
festation
of
Alzheimer's
disease,
or
increases
suscepti-
bility
through
another
mechanism,
remains
to
be
determined.
(Arch
Gen
Psychiatry.
1996;53:175-182)
IN
the
elderly,
it
remains
un¬
clear
whether
depression
in¬
creases
the
risk
of
dementia.
De¬
pression
may
be
associated
with
cognitive
deficits,
and
some
pa¬
tients
with
dementia
manifest
depressive
symptoms.1"6
Approximately
10%
to
30%
of
patients
with
Alzheimer's
disease
(AD)
meet
criteria
for
major
depression.7"10
How¬
ever,
some
symptoms
used
to
diagnose
depressive
disorders,
including
psycho-
motor
change,
apathy
or
lack
of
interest,
sleep
difficulties,
and
disturbances
in
thinking
or
concentration,
frequently
occur
in
demented
patients
who
are
not
depressed.1113
These
nonspecific
symp¬
toms
may
artificially
raise
the
reported
prevalence
of
depression
in
demented
patients.
In
the
clinically
depressed
elderly,
mild
cognitive
impairment
is
not
uncom¬
mon.1415
This
cognitive
dysfunction
im¬
proves
to
varying
degrees
once
the
depres¬
sive
symptoms
subside.1617
However,
there
is
growing
evidence
that
some
of
these
clinically
depressed
patients
with
cogni¬
tive
impairment
develop
irreversible
de¬
mentia
in
2
or
more
years,16·18·19
although
there
are
some
dissenting
reports.2021
In
clinical
samples,
follow-up
studies
indi¬
cate
that
cognitively
intact
elderly
pa¬
tients
with
depression
have
only
a
slightly
higher
probability
of
developing
demen¬
tia
than
the
general
population.22·23
From
the
Gertrude
H.
Sergievsky
Center
and
the
Center
for
Alzheimer's
Disease
Research
in
New
York
City
(Drs
Devanand,
Sano,
Tang,
Taylor,
Gurland,
Wilder,
Stern,
and
Mayeux),
the
Center
for
Geriatrics
at
Columbia
University
(Drs
Gurland
and
Wilder),
the
Department
of
Biological
Psychiatry,
New
York
State
Psychiatric
Institute
(Drs
Devanand,
Stern,
and
Mayeux),
and
the
Departments
of
Psychiatry
(Drs
Devanand,
Taylor,
Gurland,
and
Wilder)
and
Neurology
(Drs
Sano,
Tang,
Stern,
and
Mayeux),
College
of
Physicians
and
Surgeons
of
Columbia
University,
New
York,
NY.
Downloaded From: http://archpsyc.jamanetwork.com/pdfaccess.ashx?url=/data/journals/psych/12604/ by a Columbia University User on 06/13/2017
SUBJECTS
AND
METHODS
SUBJECTS
All
subjects
provided
written
informed
consent
to
partici¬
pate
in
the
Washington
Heights-Inwood
and
Columbia
Ag¬
ing
Project,
a
prospective
investigation
of
AD
and
demen¬
tias
associated
with
Parkinson's
disease
and
stroke.
Informed
consent
was
obtained
from
subjects
and
appropriate
advocates,
as
defined
by
the
Columbia
University
(New
York,
NY)
institutional
review
board
that
approved
this
project.
Sources
of
potential
subjects
included
referrals
by
community-based
health
care
providers,
random
samples
from
a
commercial
list
and
a
list
obtained
from
the
Health
Care
Financing
Agency,
a
state
agency
list
of
home
care
re¬
cipients,
self-referred
volunteers,
and
solicitation
by
news
media.
Other
details
of
subject
recruitment
to
this
com¬
munity
registry
have
been
described
elsewhere.H3'
Recruit¬
ment
was
limited
to
a
geographically
defined
area
of
north¬
ern
Manhattan.
The
present
study
sample
was
selected
by
excluding
all
subjects
identified
by
a
physician
to
have
stroke
(by
his¬
tory
or
by
focal
neurologic
signs
or
by
radiologie
evidence
consistent
with
stroke),
Parkinson's
disease,
Huntington's
disease,
multiple
sclerosis,
brain
tumor,
schizophrenia,
or
mental
retardation.
Major
depression
was
not
an
exclu¬
sion
criterion.
Subjects
were
examined
at
their
residence
(home
or
nursing
home),
senior
citizen
center,
other
com¬
munity
facility,
or
in
a
research
clinic.
Follow-up
was
con¬
ducted
with
an
identical
set
of
evaluation
procedures
at
1-year
intervals
for
a
maximum
of
5
years.
DIAGNOSTIC
EVALUATION
All
evaluations
were
conducted
in
either
English
or
Span¬
ish,
based
on
the
subjects'
primary
language
and
their
opin¬
ion
of
which
language
would
yield
better
performance.
A
physician
obtained
a
semistructured
medical
history,
con¬
ducted
a
physical
and
neurologic
examination,
completed
the
modified
short
Blessed
Memory
Information
and
Con¬
centration
test
(BMIC),36
reviewed
the
Blessed
Functional
Activity
Scale
(BFAS
parts
1
and
2),34·36
and
completed
the
Clinical
Dementia
Rating.3'
Prescribed
medications
were
reviewed
and
recorded.
The
physician,
usually
a
neurolo-
gist
or
internist,
also
obtained
a
history
of
psycho
tropic
medi¬
cation
use
and
electroconvulsive
therapy
and
made
the
di¬
agnosis
of
major
depression
based
on
clinical
impression.
The
physician
used
DSM-ÍIÍ-R
criteria38
as
a
guideline,
but
there
was
no
attempt
to
establish
reliability
of
the
physi¬
cian's
diagnosis
of
major
depression.
A
research
technician
administered
a
standardized
neu-
ropsychological
test
battery
that
assessed
memory
(verbal
and
nonverbal),
orientation,
abstract
reasoning,
language,
and
visuospatial
abilities.
These
procedures
have
been
de¬
scribed
elsewhere.34
The
initial
diagnosis
of
dementia
was
based
on
a
fixed
paradigm
that
used
a
priori
cutoff
scores
on
the
neuropsy-
chological
tests
and
evidence
of
impairment
in
social
or
oc¬
cupational
function.34·35
These
cutoff
scores
on
neuropsy-
chological
tests
were
established
after
the
cognitive
effects
of
depression,
stroke,
and
Parkinson's
disease
were
taken
into
account.34
As
a
result,
the
cutoff
scores
were
conservative
enough
to
avoid
misclassification
of
these
three
conditions
as
dementia.
The
final
diagnosis
was
based
on
a
consensus
conference
between
physicians
and
neuropsychologists.
The
diagnosis
of
dementia
was
based
on
the
DSM-ÍÍÍ-R
criteria
and
required
evidence
of
memory
impairment
and
other
cog¬
nitive
deficits
based
on
a
neuropsychological
test
para¬
digm,
as
well
as
evidence
of
impairment
in
social
or
occu¬
pational
function
based
on
the
formal
functional
assessment,
elicited
history,
or
both.
After
this
final
consensus
diagno¬
sis,
subjects
who
were
rated
as
demented
were
classified
ac¬
cording
to
cause,
based
on
the
available
medical
and
neu¬
rologic
information.
The
criteria
of
the
National
Institute
of
Neurological
and
Communicative
Disorders
and
Stroke-
Alzheimer's
Disease
and
Related
Disorders
Association39
were
used
to
make
the
diagnosis
of
possible
or
probable
AD.
In
this
consensus
diagnostic
conference,
there
was
no
attempt
to
classify
depressive
diagnoses
according
to
DSM-ÍÍI-R.
Nondemented
subjects
were
classified
into
one
of
three
categories:
(1)
no
cognitive
impairment
(no
more
than
one
neuropsychological
test
score
below
cutoffs,
no
func¬
tional
impairment,
and
Clinical
Dementia
Rating
of
0);
(2)
mild
cognitive
impairment
(two
or
more
test
scores
below
cutoffs,
no
functional
impairment,
and
Clinical
De¬
mentia
Rating
of
0);
and
(3)
moderate
cognitive
impair¬
ment
(two
or
more
test
scores
below
cutoffs,
no
func¬
tional
impairment,
and
Clinical
Dementia
Rating
of
0.5).
Follow-up
evaluations
were
conducted
at
annual
in-
Few
epidemiologie
studies
have
directly
assessed
de¬
pression
as
a
risk
factor
for
dementia,
and
many
have
ex¬
cluded
depressed
individuals
during
sample
selection.24
These
studies,
mostly
of
the
case-control
type,
suggest
that
a
history
of
medically
treated
depression
is
more
com¬
mon
in
cases
of
dementia
than
in
controls,25"31
with
lim¬
ited
contrary
evidence.32
Overall,
these
studies
suggest
that
a
history
of
depression
slightly
increases
the
risk
of
dementia.24
Major
limitations
of
many
of
these
epidemiologie
studies
include
the
use
of
a
case-control
design,
reliance
on
history
of
depression
obtained
primarily
from
medi¬
cal
records
and
informant
interviews,
absence
of
direct
evaluation
of
depression,
and
low
statistical
power.33
To
our
knowledge,
there
are
no
published
longitudinal
epi¬
demiologie
studies
that
have
used
systematic
assess¬
ments
of
depression
and
cognition
to
determine
whether
depression
is
a
risk
factor
for
dementia.
In
a
prospective,
longitudinal,
population-based
study
that
used
direct
clinical
evaluation
of
elderly
subjects
iden¬
tified
from
a
community
registry
in
a
circumscribed
geographic
area,
we
evaluated
the
relationship
between
the
presence
of
depression
and
the
incidence
of
demen¬
tia,
particularly
AD.
RESULTS
BASELINE
SAMPLE
Baseline
information
was
collected
for
1070
subjects.
Their
demographic
and
clinical
features
are
described
in
Table
1.
The
ethnic
breakdown
was
30%
non-Latino
white,
41%
Latino,
and
27%
African
American,
with
1%
in
other
categories.
For
some
clinical
variables,
eg,
psy-
Downloaded From: http://archpsyc.jamanetwork.com/pdfaccess.ashx?url=/data/journals/psych/12604/ by a Columbia University User on 06/13/2017
tervals,
with
the
use
of
the
same
diagnostic
consensus
con¬
ference
procedures.
Follow-up
diagnoses
were
made
with¬
out
knowledge
oí
any
baseline
information.
At
the
diag¬
nostic
consensus
conference,
staff
were
aware
of
whether
the
subject
was
depressed
at
that
time.
However,
this
did
not
influence
the
diagnosis
of
dementia
because
the
pri¬
mary
dementia
criterion,
cognitive
deficit,
was
based
on
a
fixed
neuropsychological
test
score
paradigm.
In
addi¬
tion,
subjects
were
studied
as
part
of
a
larger
study
with
goals
unrelated
to
those
of
this
report.
DEPRESSION
Two
senior
researchers
formally
trained
and
supervised
the
interviewers
in
the
use
of
the
HRSD,
with
the
aid
of
didac¬
tic
sessions,
videotapes,
and
in-person
interviews.
The
trained
interviewer
used
a
semistructured
interview
guide
to
administer
the
17-item
HRSD
to
the
patient.40
In
de¬
mented
subjects,
additional
information
was
obtained
from
an
informant.
To
establish
reliability,
in
a
series
of
52
elderly
sub¬
jects
(32
patients
with
stroke
and
20
normal
controls;
mean
age,
71.5
years),
three
trained
raters
and
one
clinical
psy¬
chologist
conducted
interviews
in
pairs
in
which
one
rater
administered
the
scale
and the
other
rater
observed
the
in¬
terview.
Both
raters
made
independent
ratings.
Intraclass
correlation
coefficients
were
.86
for
the
depressed
mood
item
and
.96
for
the
total
HRSD
score.
In
a
separate
validity
study
of
36
elderly
subjects
(13
patients
with
Parkinson's
disease
and
23
normal
controls;
mean
age,
73.4
years),
a
trained
interviewer
administered
the
HRSD
and
a
psychiatrist
(S.T.)
independently
con¬
ducted
the
Structured
Clinical
Interview
for
DSM-ÍÍÍ-R38
to
diagnose
major
depressive
disorder.
In
these
36
subjects,
the
mean
(±SD)
HRSD
score
was
4.4±2.8.
Six
(17%)
of
these
36
subjects
met
criteria
for
major
depressive
disorder.
For
the
single
HRSD
item
of
depressed
mood
relative
to
the
Struc¬
tured
Clinical
Interview
for
DSM-ÍÍÍ-R
diagnosis
of
major
de¬
pressive
disorder,
sensitivity
was
83%
and
specificity
was
70%.
Sensitivity
and
specificity
were
not
improved
further
by
the
inclusion
of
any
other
HRSD
item,
or
by
the
total
HRSD
score
(dichotomized
as
a
10
or
<
10).
The
symptoms
of
lack
of
interest,
agitation,
insom¬
nia,
and
weight
loss
are
not
uncommon
in
AD,
and
the
diagnosis
of
major
depression
and
the
total
HRSD
score
can
be
misleading
in
such
cases.
Given
these
considerations,
and
the
results
of
the
validity
study,
the
HRSD
depressed
mood
item
alone
(dichotomized
as
present
or
absent)
was
the
main
variable
used
for
"depression"
in
statistical
analy¬
ses.
To
confirm
the
results
obtained
with
the
depressed
mood
item,
the
total
HRSD
score
was
also
evaluated
as
the
de¬
pression
variable.
Recent
evidence41·42
suggests
a
strong
association
be¬
tween
the
presence
of
the
apolipoprotein
E-e4
alíele
and
AD.
During
the
last
year
of
the
study,
blood
was
drawn
from
a
subset
of
available
subjects
who
consented
to
assess¬
ment
of
the
apolipoprotein
E
genotype.
STATISTICAL
ANALYSIS
Baseline
Evaluation
Analyses
by
2
or
two-tailed
Student's
t
tests
were
con¬
ducted
to
compare
the
demographic
and
clinical
features
in
subjects
with
and
without
dementia.
Analyses
of
vari¬
ance
and
2
analyses
were
used
to
compare
relevant
de¬
mographic
and
clinical
variables
in
the
three
subgroups
with
no,
mild,
and
moderate
cognitive
impairment.
A
logistic
regression
model
was
used
to
assess
the
independent
ef¬
fects
of
age,
education,
language,
and
depression
on
the
di¬
agnosis
of
dementia.
Follow-up
Cox
proportional
hazards
models43
were
used
to
assess
the
relative
risk
(RR)
of
incident
dementia
associated
with
de¬
pression.
The
timing
variable
was
the
duration
from
the
initial
visit
to
the
first
follow-up
time
point
at
which
the
diagnosis
of
dementia
was
made.
Follow-up
Cox
analyses
were
recalculated
with
adjustment
for
age,
gender,
educa¬
tion,
language
of
assessment
(English
or
Spanish),
BMIC
scores
(range,
0
to
28),
and
BFAS
scores
(range,
0
to
17).
In
the
sample
subset
in
which
apolipoprotein
E
geno¬
types
were
determined,
subjects
with
the
apolipoprotein
E-e4
alíele
(homozygous
or
heterozygous)
were
classified
into
one
group,
and
the
remaining
subjects
formed
the
other
group.
The
Cox
proportional
hazards
model
was
used
to
determine
if
the
association
of
depression
with
incident
de¬
mentia
remained
after
accounting
for
the
effect
of
apoli¬
poprotein
E-e4
on
disease
risk.
chiatric
history,
information
was
not
obtained,
or
could
not
be
obtained
reliably,
in
1%
to
7%
of
subjects.
Com¬
pared
with
the
852
subjects
rated
as
not
demented,
the
218
demented
patients
were
older,
had
less
education,
and
were
more
impaired
on
tests
of
cognition
(BMIC)
and
functional
ability
(BFAS).
Of
the
demented
sub¬
jects,
89%
were
given
a
diagnosis
of
possible
or
prob¬
able
AD
with
or
without
concomitant
contributory
conditions.
Compared
with
subjects
without
dementia,
the
sub¬
jects
with
dementia
more
frequently
scored
positively
on
the
HRSD
depressed
mood
item
and had
significantly
higher
HRSD
total
scores
(Table
1).
Patients
with
de¬
mentia
were
more
likely
to
be
prescribed
anxiolytic
medi¬
cations.
The
two
groups
did
not
differ
with
respect
to
a
history
of
psychiatric
disorder
or
current
or
past
treat¬
ment
with
antidepressant
medications
or
electroconvul-
sive
therapy
(Table
1).
The
specific
type
of
past
psychi¬
atric
disorder
was
not
ascertained.
A
logistic
regression
model
was
used
to
identify
the
independent
effects
of
median
age
(^74
years
vs
>74
years),
education
(<8
years
vs
>8
years),
language
(En¬
glish
or
Spanish),
and
depression
on
the
diagnosis
of
de¬
mentia
at
baseline
evaluation.
Median
age
(RR,
6.22;
95%
confidence
interval
[CI],
4.17
to9.13;P<.001)
and
edu¬
cation
(RR,
2.27;
95%
CI,
1.53
to
3.36;
P<.001)
were
strongly
associated,
but
depressed
mood
(RR,
1.34;
95%
CI,
0.96
to
1.87;
P<.1)
and
language
(RR,
1.38;
95%
CI,
0.93
to
2.04;
P=.ll)
were
not
associated
with
the
diag¬
nosis
of
dementia
at
baseline
evaluation.
The
degree
of
cognitive
impairment
was
rated
in
849
of
the
852
subjects
who
were
not
demented
at
baseline
evaluation.
Of
the
849
subjects,
452
(53%)
had
no
cog¬
nitive
impairment,
243
(29%)
had
mild
cognitive
im-
Downloaded From: http://archpsyc.jamanetwork.com/pdfaccess.ashx?url=/data/journals/psych/12604/ by a Columbia University User on 06/13/2017
pairment,
and
112
(13%)
had
moderate
cognitive
im¬
pairment.
In
these
three
subgroups,
the
demographic
and
clinical
features
were
compared
(Table
2).
There
were
no
differences
in
gender,
history
of
psychiatric
disor¬
der,
or
current
psychotropic
medication
usage.
Greater
cognitive
impairment
was
associated
with
increased
age,
lower
education,
and
Spanish
as
the
language
of
assess¬
ment.
The
presence
of
major
depression
as
determined
by
the
physician's
clinical
impression,
as
well
as
posi¬
tive
scores
for
the
depressed
mood
item
and
the
total
HRSD
score,
increased
significantly
with
greater
cogni¬
tive
impairment.
In
the
849
subjects
without
dementia,
the
sub¬
groups
with
and
without
a
positive
score
for
the
HRSD
depressed
mood
item
were
compared
on
cognitive
per¬
formance
and
functional
ability.
The
depressed
group
per¬
formed
more
poorly
on
the
BMIC
(mean±SD,
4.9±4.4
vs
4.0±4.2;
t=2.9,
P<.005)
and
showed
greater
func¬
tional
impairment
on
the
BFAS
(2.7±3.3
vs
1.5±2.6;
t=6.8,
P<.001).
FOLLOW-UP
SAMPLE
Of
the
849
subjects
rated
not
demented
at
baseline
evalu¬
ation,
at
least
one
follow-up
evaluation
was
conducted
in
478
subjects
during
a
1-
to
5-year
period
(mean±SD
follow-up,
2.54
±
1.12
years).
The
number
of
subj
ects
fol¬
lowed
up
at
each
time
point
was
as
follows:
391
at
year
1,
346
at
year
2,
216
at
year
3,
and
105
at
year
4.
Of
the
353
subjects
who
were
not
followed
up,
165
subjects
were
recruited
in
the
last
year
of
the
study,
which
has
ended.
Reasons
for
failure
to
reexamine
the
others
were
death
(n=15),
unavailability
for
follow-up
or
relocation
(n=125),
and
refusal
(n=48).
There
were
no
significant
differ¬
ences
in
gender,
age,
language,
and
depression
scores
be¬
tween
subjects
who
were
and
were
not
followed
up.
Du¬
ration
of
follow-up
was
greater
in
subjects
with
a
high
level
of
education
compared
with
those
with
a
low
level
of
education
( 2=9.2,
P<.01).
The
mean
(±SD)
num¬
ber
of
years
of
education
was
10.4±4.8
for
subjects
who
were
followed
up
and
8.9±4.7
for
subjects
who
were
not
followed
up.
The
persistence
of
depressed
mood
over
time
was
evaluated
in
the
subjects
who
were
followed
up.
One
hun¬
dred
seven
(62%)
of
the
173
subjects
with
depressed
mood
at
baseline
had
depressed
mood
at
1-year
follow-up,
and
75%
of
the
283
subjects
without
depressed
mood
at
base¬
line
did
not
have
depressed
mood
at
1-year
follow-up
( 2=62.1,
P<.001).
Similarly,
63%
of
subjects
with
de¬
pressed
mood
at
baseline
had
depressed
mood
at
2-year
follow-up,
and
73%
of
subjects
without
depressed
mood
at
baseline
did
not
have
depressed
mood
at
2-year
fol¬
low-up
( 2=42.8,
P<.001).
These
findings
suggested
that
depressed
mood
was
often
an
enduring
symptom
and
pro¬
vided
indirect
evidence
that
the
interviewers
made
con¬
sistent
ratings
during
the
study.
Of
the
478
subjects
who
were
followed
up,
61
reached
the
end-point
diagnosis
of
dementia.
These
61
subjects
comprised
36
(21%)
of
173
subjects
who
were
depressed
and
25
(9%)
of
283
subjects
who
were
not
de¬
pressed
at
baseline
evaluation
( 2=13.3,
P<.001).
At
base¬
line
evaluation,
these
36
depressed
subjects
had
higher
*BMIC
indicates
Blessed
Memory
Information
and
Concentration
Test
(range,
0
to
28);
BFAS,
Blessed
Functional
Activity
Scale
(parts
1
and
2;
range,
0
to
17);
HRSD,
Hamilton
Rating
Scale
for
Depression;
and
ECT,
electroconvulsive
therapy.
fA
test
of
significance
for
2
(categorical
variables)
or
two-tailed
t
test
(continuous
variables),
as
appropriate.
depression
(HRSD)
and
functional
impairment
(BFAS)
scores
and
were
more
likely
to
be
Latino
(Spanish-
speaking)
than
the
25
nondepressed
subjects
who
de¬
veloped
dementia
on
follow-up
(Table
3).
However,
these
two
groups
did
not
differ
in
age,
gender,
duration
from
the
baseline
visit
to
the
final
diagnosis
of
dementia,
and
BMIC
scores.
The
mean
(±SD)
HRSD
score
in
the
de¬
pressed
group
(9.0±6.1)
indicated
that
the
depressive
syndromes
experienced
by
these
subjects
were
often
mild.
On
the
HRSD
depressed
mood
item
(scored
0
[absent]
to
4
[severely
depressed]),
22
(61%)
of
these
36
de¬
pressed
subjects
received
a
score
of
1,11
subjects
(31%)
received
a
score
of
2,
three
subjects
(8%)
received
a
score
of
3,
and
none
received
a
score
of
4.
The
36
depressed
subjects
who
developed
dementia
on
follow-up
did
not
differ
significantly
in
baseline
total
HRSD
scores
(mean±SD,
9.0±6.1)
from
the
137
depressed
subjects
(7.5±4.9;
t=1.56)
who
did
not
develop
dementia
on
fol¬
low-up.
Fifty-seven
(93%)
of
these
61
incident
dementia
cases
were
classified
as
possible
(n=9)
or
probable
(n=48)
AD.
Of
the
remaining
four
incident
dementia
cases,
one
had
progressive
supranuclear
palsy.
The
other
three
inci¬
dent
dementia
cases
had
a
progressive
clinical
course
con¬
sistent
with
AD
but
had
other
conditions
(head
injury,
alcohol
dependence
with
motor
vehicle
accident,
and
mul¬
tiple
sclerosis
that
developed
after
the
initial
evalua¬
tion)
that
were
deemed
contributing
factors.
Regardless
of
dementia
subtype,
data
for
all
subjects
were
retained
in
the
main
analyses.
A
Cox
proportional
hazards
model
evaluated
the
effect
of
baseline
depressed
mood
on
the
end-point
diagnosis
of
dementia.
The
RR
for
reaching
this
end
point
was
significantly
greater
in
subjects
with
de¬
pressed
mood
at
baseline
(RR,
2.94;
95%
CI,
1.76
to
4.91;
P<.001).
The
Figure
presents
the
survival
curves
for
Downloaded From: http://archpsyc.jamanetwork.com/pdfaccess.ashx?url=/data/journals/psych/12604/ by a Columbia University User on 06/13/2017
*BMIC
indicates
Blessed
Memory
Information
and
Concentration
Test;
BFAS,
Blessed
Functional
Activity
Scale
(parts
1
and
2);
ECT,
electroconvulsive
therapy;
and
HRSD,
Hamilton
Rating
Scale
for
Depression.
tLevel
of
significance
in
analysis
of
variance
or
2,
as
appropriate.
^Diagnosis
of
major
depression
was
based
on
the
clinical
impression
of
the
physician,
usually
a
neurologist
or
internist.
*HRSD
indicates
Hamilton
Rating
Scale
for
Depression;
BMIC,
Blessed
Memory
Information
and
Concentration
Test;
and
BFAS,
Blessed
Functional
Activity
Scale
(parts
1
and
2).
f/4
test
of
significance
for
2
(categorical
variables)
or
two-tailed
t
test
(continuous
variables),
as
appropriate.
reaching
dementia
in
subjects
with
and
without
depres¬
sion
at
baseline.
The
Cox
proportional
hazards
model
was
then
recalculated
with
age,
gender,
education,
lan¬
guage,
and
BMIC
and
BFAS
scores
included
as
covari-
ates.
Baseline
depressed
mood
remained
significantly
as¬
sociated
with
the
end-point
diagnosis
of
dementia
(RR,
2.05;
95%
CI,
1.16
to
3.62;
P<.02).
The
covariates
of
age
(RR,
1.11
per
year
interval;
95%
CI,
1.07
to
1.15;P<.001),
education
(RR,
1.97;
95%
CI,
1.05
to
3.68;
P<.05),
BMIC
scores
(RR,
1.14
per
point
interval;
95%
CI,
1.08
to
1.20;
P<.001),
and
BFAS
scores
(RR,
1.14
per
point
interval;
95%
CI,
1.02
to
1.28;
P<.02)
were
significantly
associ¬
ated
with
an
increased
risk
of
developing
dementia
on
follow-up;
gender
and
language
showed
no
significant
effects.
Similar
results
were
obtained
when
the
four
non-AD
cases
were
excluded,
with
depressed
mood
sig¬
nificantly
associated
with
the
diagnosis
of
AD
on
fol¬
low-up
(RR,
1.91;
95%
CI,
1.07
to
3.42;
P=.03),
and
the
covariates
of
age,
education,
BMIC
scores,
and
BFAS
scores
showed
significant
effects.
With
the
use
of
the
same
covariates,
the
Cox
analy¬
ses
were
repeated
with
the
total
HRSD
score
as
the
de¬
pression
variable.
The
results
were
unchanged,
with
the
total
HRSD
score
significantly
associated
with
the
end-
point
diagnosis
of
dementia
(RR,
1.07
per
point
inter¬
val;
95%
CI,
1.02
to
1.11;
P<.01),
and
the
covariates
of
age,
education,
BMIC
scores,
and
BFAS
scores
showed
significant
effects.
Similar
results
were
obtained
with
the
HRSD
score
dichotomized
into
high
(^10)
or
low
(<10)
scores
(RR,
2.16;
95%
CI,
1.14
to
4.08;
P<.02),
and
the
same
covariates
remained
significantly
associ¬
ated
with
an
increased
risk
of
developing
dementia
on
follow-up.
Of
the
48
nondemented
subjects
with
moderate
cog¬
nitive
impairment
at
baseline
evaluation,
30
(62%)
met
criteria
for
dementia
on
follow-up.
One
possibility
is
that
nearly
all
these
subjects
with
moderate
cognitive
impair¬
ment
would
have
met
criteria
for
dementia
if
followed
up
long
enough.
To
control
for
the
possibility
that
the
definition
of
moderate
cognitive
impairment
in
this
study
was
synonymous
with
very
early
dementia,
the
Cox
pro¬
portional
hazards
model
was
recalculated
after
exclu¬
sion
of
these
48
subjects
with
moderate
cognitive
im¬
pairment.
The
results
were
unchanged,
with
the
exception
that
BFAS
scores
no
longer
showed
a
significant
effect.
Baseline
depression
(RR,
2.31;
95%
CI,
1.04
to
5.15;
P=.04),
age
(RR,
1.07
per
year
interval;
95%
CI,
1.07
to
1.19;
P<.001),
education
(RR,
3.85;
95%
CI,
1.62
to
9.13;
P<.05),
and
BMIC
scores
(RR,
1.13
per
point
interval;
95%
CI,
1.04
to
1.22;
P<.01)
were
significantly
associ¬
ated
with
an
increased
risk
of
developing
dementia
on
follow-up;
gender,
language,
and
BFAS
scores
showed
no
significant
effects.
In
the
subset
of
121
subjects
who
were
assessed
for
the
apolipoprotein
E
genotype,
a
Cox
proportional
haz¬
ards
model
evaluated
the
effect
of
age,
baseline
de¬
pressed
mood,
and the
presence
of
the
apolipoprotein
E-e4
Downloaded From: http://archpsyc.jamanetwork.com/pdfaccess.ashx?url=/data/journals/psych/12604/ by a Columbia University User on 06/13/2017