Lesion Location and Poststroke Depression

Systematic Review of the Methodological Limitations in the Literature

Sanjit K. Bhogal, BA(Hon); Robert Teasell, MD; Norine Foley, BASc; Mark Speechley, PhD
BackgroundIt has been hypothesized that poststroke depression (PSD) results from left hemisphere lesions. However,
attempts to systematically review the data investigating lesion location and PSD have yielded conflicting results. We
sought to investigate the methodological differences across the literature studying the relationship between lesion
location and PSD.
Summary of ReviewA MEDLINE literature search to retrieve articles investigating the association between PSD and
lesion location was performed. Information sought included source population of samples, definition of depression,
standardized measurement of stroke and depression, blinding, time since stroke onset, and study design. Odds ratios
(ORs) and 95% CIs were calculated with the use of Review Manager and MetaView statistical software. Twenty-six
original articles were reviewed. Much of the heterogeneity across studies reflected differences in methodology. The
direction of association between left hemisphere lesion location and PSD varied depending on whether patients were
sampled as inpatients (OR, 1.36; 95% CI, 1.05 to 1.76) or from the community (OR, 0.60; 95% CI, 0.39 to 0.92). Change
in the direction of association was also observed across assessment interval from the acute stroke (OR, 2.14; 95% CI,
1.50 to 3.04) to the chronic stroke (OR, 0.53; 95% CI, 0.30 to 0.93) phase. Differences in the measurement of
depression, study design, and presentations of results also may have contributed to the heterogeneity of the findings.
ConclusionsSeveral key initiatives should be addressed before future research is undertaken, including the development
of a comprehensive measure of PSD, optimal poststroke assessment intervals, and determination of a representative
population reference. (Stroke. 2004;35:794-802.)
Key Words: brain diseases cerebrovascular accident depression methods review literature

epression is a common sequela of stroke. The prevalence of clinical depression varies from 20% to 50%
among inpatients during the acute and subacute phases of
recovery.1 Poststroke depression (PSD) has been reported to
negatively affect functional and cognitive recovery25 and is
associated with social withdrawal after stroke6,7 as well as
increased mortality.8
Two theories of PSD have been espoused. The first states
that depression after stroke or after brain injury is a psychological reaction to the clinical consequences of stroke. The
second suggests that depression arises as a consequence of
the specific brain lesion and presumably subsequent changes
in neurotransmitters. Accordingly, Folstein et al9 observed
that stroke patients suffered more from depression than
equally physically impaired orthopedic patients, thereby suggesting that the brain lesion itself could influence mood.
Furthermore, differences in emotional reactions depending on
hemisphere of the infarct suggested an organic basis of
PSD.10
The association between the location of brain lesion as a
result of stroke and PSD has been the topic of much research.

However, this complex association is not well understood11

despite the large number of studies that have examined the
association. Recent attempts to systematically review studies
of lesion location and PSD have not served to clarify this
association. Pooling previous results, Robinson12 identified
specific relationships between the locations of brain injury
and the character and severity of poststroke mood disturbances. Robinson12 suggested that left anterior cerebral lesions were associated with significantly higher depression
scores than left posterior lesions. Diagnosable depressive
disorders were found in approximately 70% of stroke patients
with left frontal brain injuries. Robinson12 estimated that only
15% of the variance in depression could be explained by the
severity of intellectual impairment, physical impairment,
quality of social support, or age, whereas the site of the lesion
explained 50% of the variance.
In contrast, a systematic review by Carson et al13 found that
when data from 34 primary studies were pooled, lesion
location was not associated with depression. The pooled
relative risk of depression after a left hemisphere stroke,

Received August 13, 2003; final revision received October 29, 2003; accepted November 25, 2003.
From the Department of Physical Medicine and Rehabilitation (S.K.B., R.T., N.F.) and Epidemiology and Biostatistics (M.S.), St. Josephs Health Care
London, Parkwood Hospital (S.K.B., R.T., N.F.), London, Ontario, Canada; and University of Western Ontario (S.K.B., R.T., N.F., M.S.), London,
Ontario, Canada.
Correspondence to Sanjit K. Bhogal, c/o Robert Teasell, MD, 801 Commissioners Rd E, London, Ontario N6C 5J1, Canada. E-mail
sanjit.bhogal@sjhc.london.on.ca
2004 American Heart Association, Inc.
Stroke is available at http://www.strokeaha.org

DOI: 10.1161/01.STR.0000117237.98749.26

794

Bhogal et al
compared with a right hemisphere stroke, was 0.93 (95% CI,
0.83 to 1.62). The authors noted that restricting the analyses
to higher-quality studies or major depressive disorders did
not affect their findings. Furthermore, the results were not
affected by stratification of time between stroke and the
assessment of depression. Thus, this systematic review offered no support for the hypothesis that the risk of depression
after stroke was influenced by lesion location.
A review conducted 2 years earlier did not yield definitive
conclusion concerning stroke lesion location and the risk of
depression.14 Singh et al14 systematically selected 26 original
articles that examined lesion location and PSD, of which 13
satisfied inclusion criteria. The authors noted that all of the
studies reviewed were methodologically flawed and were not
comparable with respect to sample, timing, and analysis of
CT scan and psychiatric evaluation. As a result, Singh et al14
did not believe that they could proceed with a formal
evaluation.
The inconsistent results among these 3 reviews (1 found an
association, 1 found no association, and 1 believed that the
studies could not be compared) indicate that there are
methodological problems in the PSD literature. There is a
lack of uniformity in the definition of stroke given the various
classifications of stroke that are used (eg, definition based on
clinical consideration versus category of stroke).15 When
stroke diagnostic criteria are unclear, data collection and
selection of appropriate samples for research are also unclear.15 Likewise, there is a lack of uniformity in the definition and measurement of depression.2 To further complicate
the issue, it is often difficult to differentiate endogenous
depression from an adjustment reaction to losses suffered
after the stroke event.15 Because of this last point, it is
difficult to separately test the 2 competing theories because
the assessment of the outcome will be confounded within and
between studies.15
Given the inconsistent results across the 3 aforementioned
reviews and the methodological concerns regarding research
in PSD,14,15 in this article we review and critique the research
methodology used in studies investigating the relationship
between stroke lesion location and depression. Although
Carson et al13 (2000) examined the relative risk of developing
PSD according to time since stroke onset and source of
patients, this article provides a more thorough methodological
review of the literature in an attempt to explain the heterogeneity of results across the literature. In addition, key
research initiatives within the area of PSD are proposed.
Areas of interest include the sample population source,
definitions of stroke and depression, imaging used, timing of
diagnosis, and statistical analysis.

Methods
Study Identification and Selection
Two MEDLINE literature searches (via PubMed) were used to
identify all studies from January 1970 to March 2003 that investigated the association between stroke lesion location and depression.
Search limits included English, human, middle age: 45. The first
search included the key words depression, cerebrovascular accident
or stroke, computerized tomography or CT scan, and post stroke and
the combined exploded key words depression AND cerebrovascular
accident OR stroke AND computerized tomography OR CT scan

Lesion Location and Poststroke Depression

795

AND post stroke. The second search included the key words
depression, cerebrovascular accident or stroke, and computerized
tomography or CT scan and the combined exploded key words
depression AND cerebrovascular accident OR stroke AND computerized tomography OR CT scan.
Articles identified in the MEDLINE searches were included if the
purpose was to study the association between stroke lesion and the
development of PSD. Studies were limited to stroke survivors and
excluded studies that included nonstroke patients.
Eleven articles were retrieved from the first search, of which 1 was
a review article and 5 met inclusion criteria. The second search
yielded 34 articles, of which 2 were review articles, 3 were
duplicates from the previous search, and 18 met inclusion criteria.
Studies cited in the retrieved articles but not identified through the
MEDLINE literature search were sought, bringing the total number
of primary articles retrieved to 38.

Data Abstraction
A single reviewer extracted data from the selected studies onto a data
abstraction form. Information sought included sample population,
definition of stroke, definition of depression, standardized measurement of stroke, standardized measurement of depression, lesion
volume and lesion localization method, blind assessment of depression to CT scan and vice versa, timing, and final conclusions.

Statistical Analysis
For articles that did not report odds ratios (ORs) and/or CIs,
frequency of patients with and without depression and site affected
were recorded from the articles; the Peto OR and 95% CI were
calculated with the use of Review Manager and MetaView, a
statistical software package available from the Cochrane Collaboration Group. Differences in the frequency of depressed and nondepressed patients between left and right hemisphere strokes were
assessed by the 2 statistic with Yates correction with the use of
Statcalc, a statistical software package available through the Centers
for Disease Control. Fishers exact test was used when observed cell
counts were 5. Statistical significance was set at the 0.05 level.

Results
A total of 38 articles were retrieved. Seven articles were
excluded because they did not meet inclusion criteria.9,16 21
In 2 instances, multiple articles reported different analyses
performed on a single group of patients (References 1, 7, 24,
25 and References 26, 27). In the end, 26 original reports
were included in the review (Table).10,22 48
Ten articles observed no significant association between
lesion locations and the frequency or severity of PSD.* Two
studies cited a significant association between right hemisphere strokes and PSD33,45; 4 studies noted a significant
association between left hemisphere strokes and PSD7,28,30,41;
3 studies noted greater depression associated with left hemisphere basal ganglia lesions31,39,40; and 7 studies noted
changes in the association between lesion location and PSD
over time.10,32,36,38,46 48
Given that many studies did not provide ORs and/or
corresponding CIs, the present article calculated the OR of
developing PSD for a left hemisphere stroke (Figure 1). ORs
and/or CIs could not be calculated for all studies because of
insufficient detail in the reports of the findings. One study did
not provide the number of nondepressed patients who had
suffered either a left or right hemisphere stroke.42 Five studies
did not provide frequency counts of the number of depressed
*References 22, 23, 26, 29, 34, 35, 37, 42 44.

796

Stroke

March 2004
Articles Investigating Lesion Location and Poststroke Depression (n25)
Article
Lipsey et al

28

Prevalence
of PSD, %

Association

Onset Since Stroke

34

Left hemisphere

Within 2 weeks

Robinson et al1,7,24; Parikh et al25

47

Left hemisphere

3060 days

Sinyor et al10

23

Association changes over time

55 days

Collin et al

29

Starkstein et al30

30

No association

Undefined

31

Left hemisphere

2 months

Starkstein et al31

42

Left basal ganglia

718 days

Starkstein et al32

Not given

Association changes over time

35 days

Dam et al

33

30

Right hemisphere

23 weeks

Sharpe et al34

18

No association

45 days

House et al35

20

No association

1 month

Astrom et al36

25

Association changes over time

35 years

Agrell et al37

29

No association

Undefined

Andersen et al22

Not given

No association

Within 7 days

Burvill et al23

Not given

No association

4 months

36

Association changes over time

7 days

Herrmann et al

22

Left basal ganglia

Within 2 months

Morris et al40

24

Left basal ganglia

Within 1 month

Morris et al41

34

Left hemisphere

710 days

Angeleri et al42

29.7

No association

Within 10 days

Gainotti et al26,27

42.5

No association

37.5 months on average

Nagaraja et al43

24

No association

2 weeks and 6 months

Pohjasvaara et al44

20

No association

Within 3 weeks

MacHale et al45

50

Right hemisphere

6 months
3 months

Iacoboni et al38
39

46

Shimoda and Robinson

Not given

Association changes over time

Sato et al47

52

Association changes over time

2 weeks

Singh et al48

36

Association changes over time

3 months

and nondepressed patients with either left or right hemisphere

strokes.10,32,33,38,47
As shown in Figure 1, the 95% CIs for individual studies
were wide, and the test of heterogeneity was highly significant (275.45, P0.00001), indicating substantial variation
in OR among studies. As noted earlier, much of this variation
is believed to reflect the methodological differences of the
literature. A review of the main methodological differences
that may contribute to heterogeneity across studies follows.

Source Populations and Study Samples

The source of patients was observed to vary across studies.
Patients were selected from 2 main sources: hospital-based
patients or community-based patients.22,23,29,34,35,42,47 The
population base of the study sample has been proposed by
some investigators15,48 to influence the outcome of the association between lesion location and PSD. When studies are
examined according to population source, 2 trends emerged
(Figure 2). Studies of hospital inpatients significantly favored
depression occurring after left hemisphere stroke (OR, 1.36;
95% CI, 1.05 to 1.76), whereas studies of community-based
samples were observed to significantly favor depression after
a right hemisphere stroke (OR, 0.60; 95% CI, 0.39 to 0.92).
References 7, 10, 26, 28, 30 33, 36 41, 44 46, 48.

An interesting finding was that the test of heterogeneity

was significant across hospital inpatient studies (262.84,
P0.00001) but not across community-based studies
(21.62, P0.81). The significant heterogeneity across
hospital inpatient studies may reflect differences in the
composition of the study samples due to different healthcare
practices. For example, in Sweden 90% of all stroke patients
are admitted to the hospital; therefore, a hospital-based study
of stroke patients in this region is representative of most
stroke patients.36 However, in some regions fewer than half of
stroke patients are admitted at the time of acute episode.35
Patient characteristics also differed among studies. Several
samples were predominantly characterized by inner city
residence, low income, and black ethnicity, with approximately one half possessing less than a grade 9 education.7,30,31,41,46 On the other hand, the patients of Singh et al14
were predominantly middle-class suburban residents, and
66% of the Sharpe et al34 sample was white.
Other inconsistencies in sample composition due to inclusion criteria were also noted. Lipsey et al28 recorded patients
with a history of alcohol abuse and noted that more of these
patients were in the left hemisphere group than the right
hemisphere group (50% versus 17%). Similarly, inclusion of
patients with aphasia resulted in a higher percentage of
patients with communication disorders in the left hemisphere

Bhogal et al

Lesion Location and Poststroke Depression

797

list,10 the Composite Depression Index,10,40 the Cornell Depression Scales,39 the Montgomery-Asberg Depression Rating Scale,37,39,40,44,48 the Present State Examination,34 and the
Hospital Anxiety and Depression Scale.34 In total, 13 different measures of depression were recorded in the 26 studies.

Figure 1. OR of poststroke depression after left hemisphere

stroke (n20).

stroke group than in the right hemisphere stroke group in 2

studies.39,46 In addition, 86% of the left hemisphere patients in
the study of Shimoda and Robinson46 had a family history of
psychiatric disorder. Alcohol abuse, family history of a
psychiatric disorder, and aphasia can increase the risk of
developing depression. The fact that these factors were
overrepresented in 1 group (ie, the left hemisphere stroke
group) may partly explain why a difference in the frequency
of depression between the 2 hemisphere groups was found in
these articles.

Definition, Diagnosis, and Standard Measurements

of Depression
Assessment of Depression
The timing and location of the assessment of depression were
well described in most articles. Most of the studies conducted
on hospital-based patients explicitly stated that patients were
interviewed for depressive symptoms in a private room,
between 11 AM and 2 PM, to minimize diurnal mood variation.7,26,28,30 32,37,38 Several studies assessed patients in their
residence, including their own homes or nursing
homes.22,23,34,35,42 Collin et al29 mailed out a questionnaire
regarding depressive symptoms to stroke patients living in the
community.
The most common measures used to assess the severity of
depression were the Hamilton Depression Rating Scale
(HDRS) and the Zung Self Report Depression Questionnaire.7,14,26,30 32 Lipsey et al28 calculated an overall depression score based on the Zung questionnaire and the HDRS,
whereas Dam et al33 used the HDRS in conjunction with the
Beck Depression Inventory (BDI). Other measures used to
assess the degree of depression were the BDI alone,35,42 the
Beck Hopelessness Scale,10 the Hopkins Symptom Check-

Depressive Symptoms
Diagnosing depression according to Diagnostic and Statistical Manual of Mental Disorders (DSM) criteria was the most
common approach. Several studies diagnosed depression on
the basis of Diagnostic and Statistical Manual of Mental
Disorders, Third Edition (DSM-III) symptom criteria for
major depression and minor depression as elicited by a
modified Present State Examination.7,22,23,30 32,35,46 Nagaraja
et al43 used the HDRS to determine Diagnostic and Statistical
Manual of Mental Disorders, Revised Third Edition (DSMIII-R) diagnoses. In some cases, clinical judgment was
required to translate responses to the appropriate DSM-III
symptom criteria.
strom et al36 diagnosed depression according to the
DSM-III symptom criteria with physical disorder (stroke) on
axis III, while several studies used the DSM-III-R with the use
of a structured clinical interview designed for the DSM-IIIR.34,37,39 41,44 One study used the more recent Diagnostic and
Statistical Manual of Mental Disorders, Fourth Edition
(DSM-IV) with the Schedule for Affective Disorders and
Schizophrenia.45
Gainotti et al26 used the DSM-III-R diagnostic criteria with
the Visual Analogue Dysphoria Scale to elicit responses from
patients suffering from aphasia. The inclusion of a measure to
assess depression in patients with aphasia is unfortunately
overlooked by many studies. Primeau15 noted that because the
DSM-III criteria rely heavily on verbal responses, assessment
of these patients with the DSM-III criteria may not be
feasible, and interpretation of their responses may be incorrect. Gainotti et al26 noted that when assessment of patients
with aphasia took their communication deficit into account,
left hemisphere lesions were not a major determinant of PSD.
It has been proposed that diagnosing stroke patients with
the use of the DSM-III-R is inappropriate because depression
may be a consequence of the brain lesion per se, with
stroke-induced changes in neurotransmitter concentration in
the cerebrum.15,26 For this reason, several studies opted to
move away from DSM criteria and based diagnosis of
depression on other criteria. Two studies used the Centre for
Epidemiological StudiesDepression,41,47 2 used the Rome
Criteria for Depression,33,38 and Angeleri et al42 clinically
assessed depression using International Statistical Classification of Diseases, 10th Revision criteria.
A concern with the use of the DSM criteria is that the
diagnostic criteria included vegetative symptoms of depression, such as psychomotor retardation, fatigue, and sleep and
appetite disturbances, which may be a consequence of the
stroke event itself independent of affect.49 While psychometric measures better capture patterns of depressive symptomology and are better able to chart changes in depressive
symptomology in response to treatment, their sensitivity has
not been fully studied.49 Goldberg50 (as cited by Primeau et
al,15 1988) reported that self-report measures tend to have low

798

Stroke

March 2004

Figure 2. OR of developing poststroke

depression after left hemisphere stroke
according to patient source (n19).

specificity and poor predictive values because of their inclusion of somatic symptoms.
There has been considerable debate regarding the use of
self-report measures to diagnose and measure depression,
given the known incongruity between DSM-III diagnostic
criteria and psychometric measures.51 Dam et al33 observed
significantly higher degrees of depression in patients with
right hemisphere stroke when measured by the HDRS;
however, this difference disappeared when the BDI was used.
Dam et al33 suggested that this was caused by indifference to
the symptoms that are often seen in patients with right
hemisphere lesions. This neglect would be more pronounced
in self-rating than in an interviewer-scored scale. Gordon et
al51 noted that the HDRS alone was more congruent with the
DSM-III than were self-report measures. The combination of
self-report measures (eg, the Zung scale) with the HDRS
resulted in a discrepancy with DSM-III diagnosis of depression.51 It has been suggested that the discrepancies resulting
from sole use of self-report measures were due to underreporting of depressive symptomology compared with observer
rating.51 Gordon et al51 interpreted the findings to suggest that
either patients tend to minimize the severity of their mood
disorders or examiners are sensitive to patients behaviors.

Timing of First Interview After Stroke

Timing of first interview for depressive symptoms ranged
from the acute period7,22,30 32,36 41 to the subacute phase of

stroke rehabilitation23,35,44,45,48 to the chronic phase.34,42 One

study was noted to have staggered assessment times between
the acute and chronic stages of stroke.26
The time of assessment since stroke onset can potentially
influence study outcome. Patients interviewed during the
subacute phase may still be adjusting to their stroke experience, and depression in these patients may reflect this
transition stage. In fact, the highest rates of depression were
noted in patients assessed within the first 28 days of stroke
(Table).
When we examined studies that assessed depression within
the first 28 days of stroke, a significant association between
left hemisphere stroke and PSD was demonstrated (OR, 2.14;
95% CI, 1.50 to 3.04). However, studies that assessed
depression between 1 and 4 months observed that the association began to favor development of PSD after right
hemisphere stroke (OR, 0.93; 95% CI, 0.66 to 1.32), with a
significant association between right hemisphere stroke and
PSD at 6 months (OR, 0.53; 95% CI, 0.30 to 0.93) (Figure 3).
It is possible that the difference in association across the 3
time intervals reflects the degree of transient mood changes
related to the stroke event itself. Shimoda and Robinson46
suggested that impairments such as social isolation or visual
perceptual disorders associated with right hemisphere strokes
may play a role in the etiology of depression in the chronic
phases of stroke. Andersen et al22 suggested that this trend
may be reflective of the neuropsychological changes that may

Bhogal et al

Lesion Location and Poststroke Depression

799

Figure 3. OR of developing poststroke

depression after left hemisphere stroke
according to time since stroke onset
(n17).

follow hemisphere strokes: catastrophic reaction with left

hemisphere strokes and denial and/or neglect or aprosody
with right hemisphere lesions. Accordingly, several
follow-up studies have suggested that PSD occurring in the
chronic phase may have a different mechanism than acute or
subacute PSD.10,32,36,38,46 48

Research Design: Cohort Versus Case-Control Study

Of the 25 studies reviewed, only 13 studies stated explicitly
how groups were assigned. Seven studies grouped patients
according to lesion characteristics, thereby using a cohort
design.1,22,30,31,36,41,45,48 Three studies33,36,40 used a casecontrol design in which patients were sampled according to
depressive symptoms. Gainotti et al26 and Sato et al47 used a
cross-sectional design.
Knowledge of whether patients were grouped according to
lesion location or depression status is important to determine
the potential bias that may affect a study. For example, if
patients were grouped according to lesion characteristics,
then the findings might be biased as a result of patients with
lesions in certain stroke locations being probed more for
depressive symptoms or lack thereof. Grouping patients
according to their depressive symptoms runs the risk that the
radiologist reading the CT scan may be biased to look for
signs of lesions in one hemisphere or another.
The majority of studies ensured that the neurological
evaluators were blinded to the findings on psychopathologi-

cal examination and that CT scans were read by an investigator who was blinded to all clinical findings. Six studies
made no specific mention regarding blinding of assessors.28,33,37,40,42,44 Although blinding of both the radiologist
and psychiatrist helps to reduce interview and detection bias,
in the case of the psychiatrist, the clinical presentation (eg,
side of hemiparesis) of a stroke patient provides clues
regarding the location of the stroke. A solution would be to
ensure that all assessors involved are unaware of the studys
hypotheses and prediction. Unfortunately, none of the reviewed studies indicated whether the assessors were unaware
of the purpose of the study.

Presentation of Results
Level of Statistical Significance
Eight of the studies indicated that all analyses were performed as 2-tailed tests at the 0.05 level.34 37,39,42,45,48 The
importance of stating whether a test was performed as a
1-tailed versus 2-tailed test was highlighted by the results of
Starkstein et al,30 who did not state whether they tested
significance at a 1-tailed or 2-tailed level. A reanalysis of
their findings using a 1-tailed test demonstrated a significant
association (Yates2.78, P0.044), a result very similar to
that presented by Starkstein et al.30 However, with the
References 7, 22, 23, 26, 30 32, 34, 35, 38, 45, 46, 48.

800

Stroke

March 2004

2-tailed test the probability value becomes 0.09, which is

nonsignificant at the 0.05 level. Therefore, their alternate
hypothesis of unequal frequency of either major or minor
depression on the basis of left or right hemisphere lesion
location was not sustained according to a 2-tailed test. One
other study performed a 1-tailed test of significance.28
Adjustment for Multiple Comparisons
Another statistical error noted was the effect of multiple
comparisons on the level. Morris et al41 found no significant association in the frequency of depression between right
and left hemisphere strokes. The authors then stratified on the
basis of size of lesion and performed multiple tests without
correcting for the number of comparisons. Robinson et al7
also performed multiple tests on their study sample without
correcting their level of significance. Only 2 studies14,42 used
a Bonferroni-adjusted level for multiple testing, and neither
study found a significant association between lesion location
and PSD.
Severity Versus Frequency
A common trend noted in the results was the tendency to
interchange the outcomes of severity of depression and
frequency of depression. Robinson et al7 indicated that left
hemisphere stroke patients had greater depression scores than
right hemisphere stroke patients as indicated on psychometric
scales, whereas the difference in frequency was not reported.
Similarly, Dam et al33 measured the severity of depression but
not the frequency of depression. Herrmann et al39 and
Angeleri et al42 also stated that the severity of depression was
greater in left versus right hemisphere lesions, but when
frequency of depression was examined, there was no difference between lesions involving the left or right hemisphere.
Use of Regression Models
Regression analysis was used in 3 studies. MacHale et al45
used stepwise regression to determine the predictive power of
lesion location on the development of PSD. Similarly, Sato et
al47 performed bivariate analyses with dummy variables used
to model lesion site, and the regression coefficient for each
variable was determined via stepwise regression. Singh et al48
used multiple linear rather than logistic regression modeling
because dichotomization of the depression variables was
arbitrary.
Inappropriate analysis and presentation of results contribute to the differences in the literature. Use of inappropriate
levels of significance misleads the readers. Moreover, the
interchanging use of severity of depression and frequency of
depression makes interpretation of results difficult. Finally,
the general absence of multiple regression models in the
literature suggests that the effects of confounding variables
were not removed from summary estimates of association
between lesion location and PSD.

Discussion
There were methodological differences across the PSD literature. These differences contribute to the discrepant findings
in the association between lesion location and PSD. Much of
the heterogeneity noted across the studies appeared to be a
function of the population source. Left hemisphere lesion

location appeared to contribute to the development of PSD in

hospital inpatient population studies. Right hemisphere lesion
location appeared to contribute to the development of PSD in
community-based population studies. Differences in time
since stroke onset and the lack of a standard tool to assess
PSD also appeared to contribute to the heterogeneity of
findings. Furthermore, the inappropriate presentation of statistical analysis added to the limitations of the literature.
Even if the site and size of the brain lesion in stroke were
significantly correlated with depression, it is difficult to
determine whether depression is due to the clinical consequences of stroke or due to neurophysiological changes that
may lead to depression. At present, it is not known whether
certain types or degrees of neurological impairment and
disability are associated with more or less depression.
The association of PSD and left hemisphere damage,
although compelling when the work of 1 group (the Johns
Hopkins group) is considered, has not been adequately or
consistently confirmed.15 The association between the site of
the stroke lesion and the likelihood of developing depression
independent of the clinical consequences remains unproven.
The authors decided to focus only on hemisphere lesion
location for simplicity since the methodological differences
that arose from hemisphere CT location were great. As a
result, the authors recognize that a limitation of this review is
that other lesion variables, such as anterior-posterior lesion
location, were not examined. It is possible that these variables
may also play a role in the development of PSD.
Before further research in this area is conducted, several
key initiatives should first be undertaken. These include the
following: development of an appropriate and standardized
measure of depression; identification of a representative
population source for research purposes; use of optimal time
since stroke onset to interview for PSD; and creation of more
comprehensive predictive causal models of PSD (as opposed
to investigation of lesion characteristics alone as a predictor
of PSD).

Development of an Appropriate and Standardized

Measure of Depression
A measure of depression that is best suited for the stroke
population is required. Such a measure would take into
account the functional, cognitive, and language impairments
that often accompany stroke. Furthermore, a measure of
depression for the stroke population must be sensitive to
some of the potential vegetative symptoms of the stroke itself
(psychomotor retardation, fatigue, and sleep and appetite
disturbances) that most conventional scales classify as depressive symptoms.

Identification of a Representative Population Source

Given that the population source of a study can influence
results, the choice of entirely hospital-based samples or
entirely community-based samples will systematically skew
results. Accordingly, researchers should work together in
conducting large-scale studies that involve follow-up of all
stroke patients coming to emergency departments for treatment, regardless of whether they were admitted to the
hospital or sent back to the community for treatment.

Bhogal et al

Optimal Time After Stroke Onset to Interview

for PSD
Several authors have suggested that PSD over the long term
may have a different mechanism than acute or subacute PSD
with left-sided stroke and denial and/or neglect or aprosody
with right-sided lesions.10,32,37,46 48 Therefore, a prospective
cohort design would be best suited for studying the relationship between lesion location and PSD. As patients are
followed, data regarding their recovery and present state
should be kept current.

Creation of Predictive Causal Models of PSD

Many of the studies reviewed investigated only the association between lesion location and PSD. Further multifactorial
approaches were taken by only 3 studies.45,47,48 Given that
depression in the general population is multicausal, the study
of depression in the stroke population should be viewed as
such. On the basis of the research thus far, PSD should be
studied in terms of lesion location, functional impairment,
social dependence, intellectual impairment, and stroke onset.
Like other fields of psychiatry and psychology, the study of
PSD is riddled with methodological differences. An analysis
of these differences can be used to guide future research. One
such direction is the development of a depression measure for
the stroke population. Not only would such a measure allow
for accurate and reliable diagnosis of stroke for research
purposes, but it might also help clinicians to distinguish true
clinical depression from reactive adaptation in their patients.
In addition, the realization that depressive symptoms manifest
themselves differently during stroke recovery (acute versus
subacute versus chronic) and also across populations (hospital inpatient versus community) allows researchers and clinicians to approach the diagnosis and treatment of depression
accordingly. Finally, use of a multifactorial approach would
allow researchers to study several means by which different
stroke survivors can develop PSD. Such an approach can be
used to explain the variability in the frequency and severity of
PSD across populations and time. In addition, understanding
PSD in terms of causal complexes would allow clinicians to
monitor patients at risk of developing PSD.

Acknowledgments
This study was supported by a grant from the Ministry of Health and
Long-Term Care of Ontario, Heart and Stroke Foundation of
Ontario, and Canadian Stroke Network.

References
1. Robinson RG, Starr LB, Lipsey JR, Rao K, Price TR. A two-year
longitudinal study of post-stroke mood disorders: dynamic changes in
associated variables over the first six months of follow-up. Stroke. 1984;
15:510 517.
2. Kotila M, Numminen H, Waltimo O, Kaste M. Post-stroke depression and
functional recovery in a population-based stroke register: the Finnstroke
study. Eur J Neurol. 1999;6:309 312.
3. Loong CK, Kenneth NKC, Paulin ST. Post-stroke depression: outcome
following rehabilitation. Aust N Z J Psychiatry. 1995;29:609 614.
4. Bacher Y, Korner-Bitensky N, Mayo N, Becker R, Coopersmith H. A
longitudinal study of depression among stroke patients participating in a
rehabilitation program. Can J Rehabil. 1990;4:2737.
5. Morris PL, Raphael B, Robinson RG. Clinical depression is associated
with impaired recovery from stroke. Med J Aust. 1992;157:239 242.
6. Feible JH, Springer CJ. Depression and failure to resume social activities
after stroke. Arch Phys Med Rehabil. 1982;63:276 277.

Lesion Location and Poststroke Depression

801

7. Robinson RG, Starr LB, Kubos KL, Price TR. A two-year longitudinal
study of post-stroke mood disorders: findings during the initial evaluation. Stroke. 1983;14:736 741.
8. Morris PL, Robinson RG, Andrejewski, Samuel J, Price TR. Association
of depression with 10-year post stroke mortality. Am J Psychiatry. 1993;
150:124 129.
9. Folstein MF, Maiberger R, McHugh PR. Mood disorder as a specific
complication of stroke. J Neurol Neurosurg Psychiatry. 1977;40:
1018 1020.
10. Sinyor D, Jacques P, Kaloupek DG, Becker R, Goldenberg M,
Coopersmith H. Poststroke depression and lesion location: an attempted
replication. Brain. 1986;109(pt 3):537546.
11. Ghika-Schmid F, Bogousslavsky J. Affective disorders following stroke.
Eur Neurol. 1997;38:75 81.
12. Robinson RG. Post-stroke mood disorder. Hosp Pract. 1986;83 89.
13. Carson AJ, MacHale S, Allen Kathryn, Lawrie SM, Dennis M, House A,
et al. Depression after stroke and lesion location: a systematic review.
Lancet. 2000;356:122126.
14. Singh A, Herrmann N, Black SE. The importance of lesion location in
poststroke depression: a critical review. Can J Psychiatry. 1998;43:
921927.
15. Primeau F. Post-stroke depression: a critical review of the literature. Can
J Psychiatry. 1988;33:757765.
16. Rosse RB, Ciolino CP. Effects of cortical lesion location on psychiatric
consultation referral for depressed stroke inpatients. Int J Psychiatry Med.
1986;15:311320.
17. Rao R. Cerebrovascular disease and late life depression: an age old
association revisited. Int J Geriatr Psychiatry. 2000;15:419 433.
18. Rao R, Jackson S, Howard R. Depression in older people with mild
stroke, carotid stenosis and peripheral vascular disease: a comparison
with healthy controls. Int J Geriatr Psychiatry. 2001;16:175183.
19. Gonzalez-Torrecillas JL, Mendlewicz J, Lobo A. Effects of early
treatment of post stroke depression on neuropsychological rehabilitation.
Int Psychogeriatr. 1995;7:547560.
20. Starkstein SE, Robinson RG, Price TR. Comparison of patients with and
without poststroke major depression matched for size and location of
lesion. Arch Gen Psychiatry. 1988;45:247252.
21. Paradiso S, Robinson RG. Minor depression after stroke: an initial validation of the DSM-IV construct. Am J Geriatr Psychiatry. 1999;7:
244 251.
22. Andersen G, Vestegaard K, Ingemann-Nielsen M, Lauritzen L. Risk
factors for post-stroke depression. Acta Psychiatr Scand. 1995;92:
193198.
23. Burvill PW, Johnon GA, Chakera TMH, Stewart-Wynne EG, Andersen
CS, Janrozik KD. The place of site of lesion in aetiology of post-stroke
depression. Cerebrovasc Dis. 1996;6:208 215.
24. Robinson RG, Lipsey JR, Rao K, Price TR. Two-year longitudinal study
of post-stroke mood disorders: comparison of acute-onset with
delayed-onset depression. Am J Psychiatry. 1986;143:1238 1244.
25. Parikh RM, Lipsey JR, Robinson RG, Price TR. A two year longitudinal
study of poststroke mood disorders: prognostic factors related to one and
two year outcome. Int J Psychiatry Med. 1988;18:4556.
26. Gainotti G, Azzoni A, Gasparini F, Marra C, Razzano C. Relation of
lesion location to verbal and nonverbal mood measures in stroke patients.
Stroke. 1997;28:21452149.
27. Gainotti G, Azzoni A, Marra C. Frequency, phenomenology and
anatomical-clinical correlates of major depression post-stroke depression.
Br J Psychiatry. 1999;175:163167.
28. Lipsey JR, Robinson RG, Pearlson GD, Rao K, Price TR. Mood change
following bilateral hemisphere brain injury. Br J Psychiatry. 1983;143:
266 273.
29. Collin SS, Lincoln NB. Depression after stroke. Clin Rehabil. 1987;1:
2732.
30. Starkstein SE, Robinson RG, Price TR. Comparison of cortical and
subcortical lesions in the production of poststroke mood disorders. Brain.
1987;110(pt 4):10451059.
31. Starkstein SE, Robinson RG, Berthier ML, Parikh RM, Price TR. Differential mood changes following basal ganglia versus thalamic lesions.
Arch Neurol. 1988;45:725730.
32. Starkstein SE, Robinson RG, Honig MA, Parikh RM, Joselyn J, Price TR.
Mood changes after right hemisphere lesions. J Psychiatry. 1989;155:
79 85.
33. Dam H, Pedersen HE, Ahigren P. Depression among patients with stroke.
Acta Psychiatr Scand. 1989;80:118 124.

802

Stroke

March 2004

34. Sharpe M, Hawton K, House A, Molyneux A, Sandercock P, Bamford J,

et al. Mood disorders in long-term survivors of stroke: associations with
brain lesion location and volume. Psychol Med. 1990;20:815 828.
35. House A, Dennis M, Warlow C, Hawton K, Molyneux A. Mood disorders
after stroke and their relation to lesion location: a CT scan study. Brain.
1990;113(pt 4):11131129.
36. strom M, Adolfsson R, Asplund K. Major depression in stroke patients:
a 3-year longitudinal study. Stroke. 1993;24:976 982.
37. Agrell B, Dehlin O. Depression in stroke patients with left and right
hemisphere lesion: a study in geriatric rehabilitation in patients. Aging
Clin Exp Res. 1994;6:56.
38. Iacoboni M, Padovani A, Piero D, Lezi GL. Post stroke depression:
relationships with morphological damage and cognition over time. Int
J Neurol Sci. 1995;16:209 216.
39. Herrmann M, Bartels C, Schumacher M, Wallesch CW. Poststroke
depression: is there a pathoanatomic correlate for depression in the
postacute stage of stroke? Stroke. 1995;26:850 856.
40. Morris PL, Robinson RG, Raphael B, Hopwood MJ. Lesion location and
poststroke depression. J Neuropsychiatry Clin Neurosci. 1996;8:399403.
41. Morris PL, Robinson RG, de Carralho ML, Albert P, Wells JC, Samuels
JF, et al. Lesion characteristics and depressed mood in the Stroke Data
Bank Study. J Neuropsychiatry Clin Neurosci. 1996;8:153159.
42. Angeleri F, Angeleri VA, Foschi N, Giaquinto S, Nolfe G, Saginario A,
et al. Depression after stroke: an investigation through catamnesis. J Clin
Psychiatry. 1997;58:261265.

43. Nagaraja D, Taly AB, Chatterjee S. Post stroke mood disorders. J Assoc
Physicians India. 2003;45:179 181.
44. Pohjasvaara T, Leppavuori A, Siira I, Vataja R, Kaste M, Erkinjuntti T.
Frequency and clinical determinants of poststroke depression. Stroke.
1998;29:23112317.
45. MacHale SM, ORourke SJ, Wardlaw JM, Dennis MS. Depression and its
relation to lesion location after stroke. J Neurol Neurosurg Psychiatry.
1998;64:371374.
46. Shimoda K, Robinson RG. The relationship between poststroke
depression and lesion location in long-term follow-up. Biol Psychiatry.
1999;45:187192.
47. Sato P, Bryan RN, Fried LP. Neuroanatomic and functional correlates of
depressed mood: the Cardiovascular Health Study. Am J Epidemiol.
1999;150:919 929.
48. Singh A, Black SE, Herrmann N, Leibovitch FS, Ebert PL, Lawrence J,
et al. Functional and neuroanatomic correlations in poststroke depression:
the Sunnybrook Stroke Study. Stroke. 2000;31:637 644.
49. Swartzman LC, Gibson MC, Armstrong TL. Psychosocial considerations
in adjustment to stroke. Phys Med Rehabil. 1998;12:519 541.
50. Goldberg D. Identifying psychiatric illness among general medical
patients. BMJ. 1985;291:161162.
51. Gordon WA, Hibbard MR, Egelko S, Riley E, Simon D, Leonard D, et al.
Issues in the diagnosis of post-stroke depression. Rehabil Psychol. 1991;
63:71 87.