Uploaded by User24675

Systematic review and meta-analysis of the prevalence of CIND in the first year post-stroke

Original research article
Systematic review and meta-analysis of
the prevalence of cognitive impairment
no dementia in the first year post-stroke
European Stroke Journal
0(0) 1–12
! European Stroke Organisation
Article reuse guidelines:
DOI: 10.1177/2396987318825484
Eithne Sexton1 , Affraic McLoughlin1, David J Williams2,
Niamh A Merriman1, Nora Donnelly3, Daniela Rohde1 ,
Anne Hickey1 , Maev-Ann Wren3 and Kathleen Bennett4
Introduction: Increasing attention is being paid to interventions for cognitive impairment (CI) post-stroke, including
for CI that does not meet dementia criteria. The aim of this paper was to conduct a systematic review and meta-analysis
of the prevalence of cognitive impairment no dementia (CIND) within one year post-stroke.
Patients and methods: Pubmed, EMBASE and PsychInfo were searched for papers published in English in 1995–2017.
Included studies were population or hospital-based cohort studies for first-ever/recurrent stroke, assessing CIND using
standardised criteria at 1–12 months post-stroke. Abstracts were screened, followed by full text review of potentially
relevant articles. Data were extracted using a standard form, and study quality was appraised using the Crowe Critical
Appraisal Tool. A pooled prevalence of CIND with 95% confidence intervals (CI) was estimated using random-effects
meta-analysis. Heterogeneity was measured using the I2 statistic.
Results: A total of 7000 abstracts were screened, followed by 1028 full text articles. Twenty-three articles were
included in the systematic review, and 21 in the meta-analysis. The pooled CIND prevalence was 38% [95% CI ¼ 32–
43%] (I2¼92.5%, p < 0.01). Study quality emerged as one source of heterogeneity. The five studies with the highest
quality scores had no heterogeneity (I2¼0%, p ¼ 0.99), with a similar pooled prevalence (39%, 95%CI ¼ 35–42%). Other
sources of heterogeneity were stroke type, inclusion of pre-stroke CI, and age at assessment time.
Discussion and conclusion: Meta-analysis of available studies indicates that in the first year post-stroke, 4 in 10
patients display a level of cognitive impairment that does not meet the criteria for dementia.
Stroke, cognition, dementia, systematic reviews
Date received: 21 September 2018; accepted: 20 December 2018
Post-stroke cognitive impairment (CI) is a common
consequence of stroke, leading to reduced quality of
life.1 Increasing attention is being paid to identify and
develop interventions that can improve cognition poststroke, and quantifying the potential benefits of such
interventions for improving cognition. However, there
is an absence of robust evidence on the prevalence of
post-stroke CI, based on rigorous systematic review
and meta-analysis.
A previous systematic review and meta-analysis estimated the prevalence of dementia in the first year
after stroke.2 In hospital-based studies that included
Department of Psychology, Royal College of Surgeons in Ireland,
Dublin, Ireland
Department of Geriatric and Stroke Medicine, Royal College of
Surgeons in Ireland, Dublin, Ireland
Social Research Division, Economic and Social Research Institute,
Dublin, Ireland
Division of Population Health Sciences, Royal College of Surgeons in
Ireland, Dublin, Ireland
Corresponding author:
Eithne Sexton, Department of Psychology, Royal College of Surgeons in
Ireland, Beaux Lane House, Lower Mercer St, Dublin 2, Ireland.
Email: [email protected]
first-ever and recurrent stroke, the pooled prevalence of
dementia was 26.5%.2 However, there has been no
meta-analysis of the prevalence of CI that does not
meet the criteria for dementia (cognitive impairment
no dementia, CIND).
Information on the prevalence of post-stroke CIND
is critical in order to alert clinicians to its frequency
among stroke patients, informing screening, diagnosis
and rehabilitation practices. It can also inform policy
and planning for services and interventions for CI, by
identifying the number of patients who could potentially benefit from such services. Prevalence estimates are
also required for robust epidemiological modelling of
post-stroke cognitive disease progression, which is necessary to evaluate the cost-effectiveness of cognitive
A significant challenge in estimating the prevalence
of post-stroke CIND is the lack of uniformity in its
definition. Studies vary as to the number of domains
that must be impaired, the number of tests per domain,
how domains are defined, and specific cut-offs used to
define impairment.3,4 Early definitions focussed on the
concept of mild cognitive impairment (MCI), primarily
framed as a pre-cursor to Alzheimer’s disease.5 This
definition often requires intact basic activities of daily
living (ADLs), which can be problematic for stroke
patients who may have ADL difficulties due to sensorimotor problems, independent of cognitive function.6
It has been recommended that requirements for intact
ADLs should be specific to instrumental ADLs
(IADLs) that are associated with cognition (e.g. managing money).3 This is consistent with the recent DSM5 criteria which emphasise impairment in IADLs as a
distinguishing factor between major and minor neurocognitive disorder. These new criteria also do not
require memory impairment, which is not always present in stroke-related dementia.7
Population or sampling characteristics, such as
study setting (hospital or community), inclusion of
recurrent stroke, or of pre-stroke dementia, can also
lead to variation in estimates of post-stroke CI prevalence.2,8,9 Exclusion of patients who have difficulty
undergoing cognitive testing (e.g. language difficulties)
may be unavoidable, but can also effect prevalence
We aimed to systematically review and meta-analyse
current evidence on the prevalence of post-stroke cognitive impairment no dementia (CIND) up to one year
post-stroke. We included any study that examined a
level of CI that did not meet the criteria for dementia.
Differences in definition were then explored in stratified
analysis, along with other factors including case-mix,
follow-up time, definition of CIND, and factors affecting selection.
European Stroke Journal 0(0)
This study was carried out according to the Preferred
Reporting Items for Systematic Reviews and MetaAnalyses Statement.11 The protocol was registered in
the International Prospective Register of Systematic
Reviews in September 2016 (registration number:
Search strategy
Three databases were searched from January 1995 to
30 September 2017: Pubmed (http://www.ncbi.nlm.nih.
gov/pubmed/), EMBASE (https://www.embase.com)
and PsycINFO (http://www.apa.org/pubs/databases/
psycinfo/index.aspx). Older studies may be less generalisable to the present-day, due to changes in stroke outcomes, casemix and treatment, and 1995 was chosen as
an arbitrary cut-point. The search strategy was developed in consultation with a librarian and a stroke specialist. Title and abstract terms for stroke, ischaemic
stroke and CI were used, in addition to relevant topic
terms (e.g. Medical Subject Headings). A sample search
strategy (for PubMed) is displayed in Supplementary
Table S1. Following removal of duplicates, each
abstract was assessed by one reviewer (ES), with a
random 45% of the independently screened by three
further reviewers (ND, NAM and DR). Any disagreements were discussed to reach consensus.
Following the abstract screening process, full texts
of potentially relevant articles were examined. If there
was doubt over inclusion of an article, it was discussed
with a further independent reviewer (KB) to reach a
consensus decision.
The reference lists of included articles were reviewed
to identify any further relevant articles. A citation database (Scopus: https://www.elsevier.com/solutions/
scopus) was also searched to identify any relevant
papers citing the included articles. Two articles that
reviewed available cohort studies on stroke and cognitive impairment were also reviewed to ensure that no
relevant study had been overlooked.13,14
Inclusion and exclusion criteria
Studies that included a population of adult stroke
patients, either mixed ischaemic and haemorrhagic, or
ischaemic stroke alone, were included. Studies that
included patients with transient ischemic attack (TIA)
were eligible if stroke patients comprised 75% of the
sample. Hospital-based studies with inclusion of consecutive eligible patients, or community-based studies
that included all eligible strokes within a defined geographical area, were included. Included studies assessed
cognitive function using a standardised test, at a
defined time post-stroke, and reported the percentage
Sexton et al.
of patients classified as having CIND. Studies that used
alternative terminology – for example, mild cognitive
impairment instead of CIND – were included.
Studies of patient populations that were unlikely to
be representative of the total population with stroke
were excluded – for example, studies that only focussed
on specific stroke sub-types, only included patients with
specific deficits or co-morbidities, studies that excluded
patients with any disability, or populations recruited
via rehabilitation or outpatient settings. Studies that
only included haemorrhagic stroke were excluded.
Studies where the follow-up time was not known (e.g.
stroke was ascertained on imaging or autopsy), or with
a broad range of follow-up times (e.g. studies of stroke
patients living in the community at varying times poststroke) were excluded. Studies that assessed cognition
less than one month after the index episode were
excluded, to account for the influence of post-stroke
delirium. Studies that excluded post-stroke dementia,
without clearly reporting the number who were excluded for that reason, were excluded from the review, to
ensure consistency in the denominator.
Studies that were published in a peer-review journal
in English were included. Commentary or review
articles, case reports and conference abstracts were
excluded. When more than one article based on the
same study sample was identified, the most complete
and/or relevant paper was identified as the reference
article, with supplementary information obtained
from the other articles. If it was not possible to establish whether a study should be included on the basis of
the published article, the authors were contacted for
further information.
Data extraction
Data were extracted using a standard form by ES, and
checked by AMc. This included the total study N, the
number with CIND and the number with dementia,
setting, population (stroke type, first-ever/recurrent
stroke, inclusion of pre-stroke dementia/CI, mean
sample age), outcome measurement (definition of
CIND, definition of dementia) and follow-up time.
Details about the definition of CIND and dementia
were extracted – e.g. whether intact ADLs were
required for a diagnosis of CIND, and whether
memory was included as a criterion for dementia.
Where sufficient information was available, rates of eligibility (the percentage of patients assessed for eligibility who were included) and response rates (the
percentage of eligible patients who participated in the
study) were calculated. The CIND prevalence rate for
each study was calculated using the total patients with
no cognitive impairment (NCI), CIND and dementia
as the denominator.
Quality assessment
The Crowe Critical Appraisal Tool was applied independently by two individuals (ES, AMc).15 Quality criteria are based on reporting quality and suitability of
study characteristics for the research questions. This
tool includes eight sections, covering areas such as
“Design” and “Data Collection”, each scored out of
five. We defined suitability criteria in advance of applying the tool, based on the specific purpose of estimating
the prevalence of CIND. The full criteria used are
available in Supplementary Table S2. If there was a
difference between the two appraisers of >2 points in
a score for a specific section, or the total score for the
study, this was discussed and any adjustments made to
the score accordingly.
Consistency in ratings was evaluated with an intraclass correlation co-efficient (ICC), based on the following criteria: <05 ¼ poor; 0.5–0.749 ¼ moderate;
0.75–0.9 ¼ good; >0.9 ¼ excellent.16 An overall quality
score was calculated for each study by averaging the
two appraiser scores. For the stratified meta-analysis,
the quality score was divided into quartiles, based on
the 25th, 50th and 75th percentiles.
A pooled prevalence rate was estimated based on all
eligible studies, using random effects meta-analysis
based on the binomial distribution. 95% confidence
intervals were calculated with the exact method, using
the metaprop command in Stata.17 Heterogeneity was
quantified using the I2 statistic. The statistical significance and magnitude of the I2 was considered, with
values over 40% considered moderate, and values
over 60% considered substantial.18 Reasons for heterogeneity were explored using stratified meta-analysis.
Per protocol,12 stratified meta-analysis was carried
out based on the following groups: casemix (stroke
type, inclusion of recurrent stroke and pre-stroke
dementia), CIND definition; and setting (hospital V
community). In addition, we also examined heterogeneity on the basis of quality score quartile and year of
data collection.
Included articles
A total of 7000 unique abstracts were identified, of
which 5971 were excluded in the initial title/abstract
review (see flow chart in Figure 1). The full texts of
1028 articles were reviewed, with 184 of these examined
in detail and of these 161 of these were excluded.
Following review of the reference lists, articles citing
included studies, and two relevant review papers,13,14
European Stroke Journal 0(0)
Figure 1. Flow chart of included studies.
three further full text articles were reviewed, none of
which were eligible for inclusion. The most frequent
reason for exclusion was that the study population
was not a general stroke population, e.g. non-stroke
participants were included, or only a specific type of
stroke patient was included. The next most frequent
reason was that CIND was not assessed – for example,
CI was measured as a continuous variable, or CI was
reported as a single category that included CIND
and dementia.
There were 23 articles included in the review. Almost
half included a mixed ischaemic and haemorrhagic
stroke population (k ¼ 11) (see Table 1). Follow-up
time ranged from 1 to 12 months, with three months
the modal follow-up time (k ¼ 12). The average age
ranged from 39 to 81. Most studies (k ¼ 15) had an
average age between 65 and 75. One study was population-based,19 and the remainder were hospital-based.
There was some consistency in definitions of dementia,
with 15/23 studies using a DSM definition of dementia.
Definitions of CIND were much more diverse, with
varying tests, domains and cut-offs used.
Exclusion criteria related to the capacity to participate in the research were common across studies,
including psychiatric or neurological disorders, language or sensory problems, lack of consciousness,
refusal to consent. Some studies had criteria beyond
these that excluded the most severely affected patients,
IS only
IS (88%) and H (12%)
IS (81%) and ICH (19%)
IS (89%) and ICH (11%)
IS (76%) and ICH (8%),
TIA (16%)
IS (93%) and H (16%)
7/40, (17.5%)
27/102, (26.5%)
14/60, (23%)
39/101, (38.6%)
19/98, (19.4%)
69/184, (37.5%)
Saini et al.22
Salvadori et al.26
Serrano et al.40
Srikanth et al.19
Stephens et al.30
Tang et al.28
Hong Kong
Madureira et al.,29
Mehrabian et al.38
Oksala et al.27
Sachdev et al.39
IS (93.5%) and H (6.5%)
32/110, (29.1%)
IS (88%), H (12%)
IS (88%) and ICH (12%)
IS (89%) and ICH (11%)
Not clear
IS (93%) and ICH (7%)
41/ 209, (19.6%)
33/88, (37.5%)
92/384, (24%)
39/222, (17.6%)
IS only
62/ 166, (36.7%)
35/ 80, (43.8%)
IS only
242/ 451, (47%)
IS (79%) and TIA (21%)
IS only
30/74, (41%)
167/ 318, (52.5%)
IS (75%) and H (25%)
101/ 220, (45.9%)
87/220, (39.5%)
IS (86.2%) and ICH (13.8)
57/143, (39.9%)
Akinyemi et al.31
Arauz et al.25
Cao et al.35
Chaudhari et al.21
Cumming et al.1
Delgado et al.32
Hobson et al.24
Ihle-Hansen et al.20
Jacquin et al.33
CIND/N, (%)
Study, Country
Table 1. Characteristics of included studies (n ¼ 23).
excluded due
as ineligible (%)
3 months
3 months
12 months
3 months
6–9 months
3–6 months
3–6 months
3 months
12 months
3 months
3 months
12 months
3 months
12 months
3 months
6 months
6–12 months
3 months
3 months
Both (20%
Both (11.8%
Both (20%
Subjective þ objective impairment
OR objective evidence of decline.
MMSE 26 and MoCA < 26,
OR 1.5 SD below mean
on 1þ domains
2SD below mean on 1þ domains
2SD below mean on 1 test, or >1SD
below mean on >1 tests
<5th %tile on 1þ domains or 5th-10th
%tile on 2þ domains
MMSE education-based cut-offs
>1SD below mean on 1þ domains
73 (7.5)
80.6 (nr)
69.3 (nr)
68.2 (14.6)
59.8 (11.4)
72.2 (9.0)
71.2 (7.7)
65.6 (5.6)
59 (12.7)
66 (16.6)
72 (12.2)
73.1 (10.9)
72 (7.5)
72.1 (13.9)
59.4 (10.9)
38.8 (8.3)
56 (17.8)
60.4 (9.5)
Mean age in
years (SD)
Impairment in 1þ domain, cut-off
not specified
<5th %tile in 1þ test, within CI limits
of 5th %tile in 1þ further test
<6th %tile in 1þ domain
CI in 2þ domains,
functional decline;
1SD below mean on 1þ domains
2SD below the mean
in 2þ domains
Dementia definition
Not specified
1þ domains impaired, cut-off
not specified
1 SD below the mean on
2þ domains
2 SD below mean on 3þ domains
1.5 SD below mean on 1þ domains
CIND criteriaa
Both (12.6%
First-ever or
recurrent stroke
Sexton et al.
IS: Ischaemic stroke; ICH: intracerebral haemorrhage; H: haemorrhagic (including ICH and SAH); TIA: transient ischaemic attack; nr: not reported; CIND: cognitive impairment no dementia; MMSE: Mini
Mental State Examination; MoCA: Montreal Cognitive Assessment; DSM-IV: Diagnostic and Statistical Manual of Mental Disorders, 4th Edition; SD: standard deviation; IQCODE: Informant Questionnaire
on Cognitive Decline in the Elderly; CAMCOG: Cambridge Cognitive Examination; CDR: Clinical Dementia Rating.
The mean used to define a cut-off relates to a mean observed in a normative population – either a study-specific control group or published normative scores.
Not explicitly reported that first-ever and recurrent included, but assume both included.
63.9 (12.4)
IS only
56/104, (53.8%)
176/353, (49.9%)
3 months
Both (18.4%
<10 %tile in 1þ domains
52.5 (nr)
1 month
MoCA<26, MMSE>¼24
71.1 (11.0)
CDR score 1–3
IS (68%), TIA (14%)
ICH (7.7%) Other 11%)
IS only
204/518, (39.4%)
3-6 months
Impairment in 1þ domain,
cut-off not specified
CDR ¼ 0.5
IS only
180/328, (54.9%)
Tang et al.23
Hong Kong
Wong et al.41
Hong Kong
Yan et al.42
Yu et al.34
3 months
Dementia definition
CIND criteriaa
First-ever or
recurrent stroke
CIND/N, (%)
Study, Country
excluded due
as ineligible (%)
Table 1. Continued
68.2 (9.9)
European Stroke Journal 0(0)
Mean age in
years (SD)
for example those with mRS > 4 or NIHSS > 4 (see
Table S4 for more details).
Study quality
The mean quality score was 30.3/40 (range: 24.8–33.5),
or 3.75/5 per section. The intra-class correlation coefficient was 0.82, which indicates good reliability.16 Quality
scores were influenced by multiple factors, with studies
performing well in some areas and poorly in others.
Almost all studies had at least one problem with their
definition of CIND or dementia – for example, a lack of
detail on specific cut-offs used20–24 or whether intact
ADLs were required for CIND.19,22–30 Many studies
reported limited information on how patients were
selected into the study, with only 10/23 studies providing
a clear flow chart.20–22,25,26,28,31–34 Poor information on
patient selection was a key distinguishing characteristic
of studies classified as low versus high quality (see
Tables S3 and S4 for more details).
Prevalence of CIND
The estimated prevalence of CIND in the first year
after stroke ranged from 17.5%35 to 54.9%.23 Two
studies had restrictive age criteria that meant the
mean age was substantially lower (age < 49, mean
age ¼ 39 years)35 or higher (age > 75 years, mean
age ¼ 80),30 and were only included in the agestratified meta-analysis (see Supplementary Figure
S1), leaving 21 studies included in the meta-analysis.
Figure 2 displays the meta-analysis of 21 studies.
The pooled prevalence was 38% (95% CI: 32–43%).
The total number of patients across studies was 4152,
and 1680 were classified as having CIND (n ¼ 1680/
4152). However, the heterogeneity was very high
(I2 ¼ 92.5%, p < 0.01). This heterogeneity was not
explained by factors related to casemix, follow-up
time, age, patient selection, or definition of CIND
and dementia (see Supplementary Figures S1 to S12).
Differences across sub-groups were hard to interpret
due to the high heterogeneity, but no substantial differences in prevalence between sub-groups were observed.
However, the heterogeneity did reduce when the
studies were stratified by quality score, with no heterogeneity observed in the highest quartile of quality (see
Figure 3). In these highest quality studies, the pooled
prevalence was 39% (95% CI 35–42%, n ¼ 285/736).
This prevalence estimate was similar to the overall estimate, only more precise.
The studies in the two highest quality quartiles (the
top 50% of quality scores) were stratified by casemix,
follow-up time, age, patient selection, and definition of
CIND and dementia (see Figures S13–S22). The analysis
by stroke type (see Figure 4) indicated that in higher
Sexton et al.
Figure 2. Pooled prevalence of CIND in the first year post-stroke.
quality studies with a study population of mixed
ischaemic and haemorrhagic stroke, the prevalence
estimate was 36% (95% CI 33–43%, n ¼ 310/844),
with moderate heterogeneity (I2 ¼ 48.5%, p ¼ 0.07).
Stratification by age of the sample suggested a higher
pooled prevalence in studies with a mean age greater
than 65 years (41%, 95% CI: 36–45%), compared with
younger than 65 years (32%, 95% CI: 24–40%),
although this difference was not statistically significant
(p ¼ 0.06) (see Figure 5). Heterogeneity was also reduced
when stratified by inclusion of pre-stroke CI and dementia (Figure S14), and when we only included high-quality
studies that used a domain-based definition of CIND
(e.g. impairment in one or more cognitive domains)
and DSM-IV criteria for dementia (Figure S20).
The meta-analysis was re-run excluding patients
diagnosed with dementia post-stroke from the denominator, to estimate the proportion of stroke survivors
without dementia who have CIND. The pooled
prevalence was 45% (95% CI: 40–51%), and again
heterogeneity was high (I2 ¼ 89.76%, p < 0.01)
(see Figure S23). The stratified analysis was similar to
the main meta-analysis, with no heterogeneity observed
in the highest quartile of quality (see Figure S24).
This meta-analysis found a pooled prevalence of CIND in
the first year after stroke of 38% (95% CI: 32–43%).
However, heterogeneity was very high, indicating that
this estimate is not consistent across studies. Sources of
heterogeneity related to casemix and study design were
explored, with no one factor explaining differences in the
estimates obtained. However, overall study quality
emerged as a significant source of heterogeneity, with no
heterogeneity in the five studies in the highest quartile of
quality score. The pooled prevalence for these five studies
was 39%, very close to the overall pooled prevalence. As
quality scores reflected multiple factors, it may have been a
combination of factors that led to variation in estimates,
rather than a single factor. High-quality studies had clearer information on patient selection, and may have been
European Stroke Journal 0(0)
Figure 3. Pooled prevalence of CIND in the first year post-stroke, stratified by quality quartile.
more representative of the overall stroke population.
Casemix appeared to play a role, as stratified analysis by
stroke type and sample age of studies in the top two quality
quartiles led to a reduction in heterogeneity. Some key
casemix factors, such as inclusion of prior stroke or
TIA, were not associated with differences in estimates.
Patients with TIA or prior stroke tended to be in a minority in these studies (<20%, see Table 1) and may not have
had a substantial influence on the overall estimate.
Study quality
Many studies relied on narrative descriptions of patient
selection into the study and follow-up, which were often
incomplete or inconsistent. This highlights the need for
clear and comprehensive charts describing the flow of
participants. Very few studies included detailed information on data collection procedures, for example use of
home visits or telephone calls, which can be an important influence on selection bias.9 There was some consistency in definitions of CI, in that most used the DSMIV criteria for dementia, and a definition of CIND based
on impairment in at least one cognitive domain.
However, there was diversity in how these definitions
were operationalised, including specific cut-offs for
impairment, the use of functional criteria and the
requirement for a memory impairment.
Strengths and limitations of the review
Strengths of this meta-analysis include a rigorous and
comprehensive database search, covering over 20 years
Sexton et al.
Figure 4. Pooled prevalence of CIND in first year post-stroke, in high and medium-high quality studies, stratified by stroke type.
IS: Ischaemic stroke, TIA: Transient Ischaemic Attack; H: Haemorrhagic stroke.
of studies; a detailed quality appraisal process; and a
thorough exploration of heterogeneity. A limitation
was the inclusion of some lower quality or less relevant
studies, such as studies did not provide sufficient detail
on study recruitment, or that included TIA patients.
However, we were concerned that strict inclusion criteria would exclude some potentially relevant studies. We
decided to take a more inclusive approach, but stratify
the meta-analysis by these factors, to identify any effect
on the results. A further limitation is that studies were
predominantly hospital-based (with only one
population-based), and higher risk patients were generally excluded due to lack of capacity to participate.
We excluded studies that focussed only on specific
stroke sub-types, such as lacunar stroke. This ensured
that the study populations were as representative as
possible of the overall general stroke population.
Lacunar stroke is a very common stroke type, and
this approach thus potentially excluded a large body
of research. However, we only identified one study
that would have been included if we had not excluded
lacunar stroke.36 This study estimated a CIND prevalence of 35%, which is within the confidence intervals
of our prevalence estimate. Similarly, a recent systematic review of MCI in lacunar stroke found a pooled
prevalence rate of 37% across four studies.37.
The quality of studies in this area could be improved by
more consistent definition of CIND across studies, and
clearer reporting of patient selection and data collection procedures. However, given the consistency
between the overall pooled estimate and the estimate
derived from the highest quality studies, further highquality prevalence studies may not alter the findings of
this review. None of the studies identified used the
DSM-5 criteria for mild neurocognitive disorders,
which is not surprising given the relative recency of
their publication.7 However, it will be interesting to
European Stroke Journal 0(0)
Figure 5. Pooled prevalence of CIND in first year post-stroke, in high and medium-high quality studies, stratified by mean age of the
study sample.
compare the results of this review with any future studies using the DSM-5 criteria.
In addition to being a useful guide for clinicians, this
prevalence estimate will be used in the StrokeCog epidemiological modelling platform. This model will apply
this estimate and those from other sources to realworld populations to estimate the number of people
with post-stroke CIND, and make projections into
the future. This will inform policy and planning for
future service provision, and facilitate costeffectiveness analysis.
Approximately 4 in 10 stroke patients who are hospitalised and capable of undergoing cognitive assessment have
CIND in the first year post-stroke. This finding is consistent across high-quality studies that include a mixed
ischaemic and haemorrhagic stroke population. The
prevalence may be lower (3 in 10) in younger age
groups. Clinicians need to be mindful of the need to
screen for post-stroke CIND, using tools such as the
MoCA, and refer for more detailed neuropsychological
assessment where necessary. However, the benefits of
screening and assessment are limited in the absence of
cost-effective interventions for this condition. The prevalence estimate identified in this study is a critical parameter for the StrokeCog epidemiological model that will
contribute to identifying such cost-effective interventions.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with
respect to the research, authorship, and/or publication of
this article.
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this
article: This work was funded by the Health Research Board
(HRB) in Ireland under Grant No. ICE-2015–1048 and
award RL-15–1579.
Sexton et al.
Ethical approval
Ethical approval was not sought for this article because this
was a systematic review of existing publications and no new
patient data was collected.
Informed consent
Informed consent was not sought for this article because this
was a systematic review of existing publications and no new
patient data was collected.
AH, KB and MAW developed the original idea. ES developed the review protocol, and was primarily responsible for
the data analysis and drafting the manuscript, under the
supervision of KB. The search strategy and inclusion criteria
were developed by ES with substantial input from DW and
AH. Abstract screening and article review was carried out by
ES, NAM, ND, DR and KB. ES and AMc conducted the
quality appraisal and data extraction. All authors commented
on and gave final approval to the final version of
the manuscript.
We acknowledge the valuable input or Prof Frances Horgan,
Prof Niall Pender and Grainne McCabe. We also acknowledge the participants and authors of the studies included in
this meta-analysis.
Eithne Sexton
Daniela Rohde
Anne Hickey
1. Cumming TB, Brodtmann A, Darby D, et al. The importance of cognition to quality of life after stroke. J
Psychosom Res 2014; 77: 374–379.
2. Pendlebury ST and Rothwell PM. Prevalence, incidence,
and factors associated with pre-stroke and post-stroke
dementia: a systematic review and meta-analysis. Lancet
Neurol 2009; 8: 1006–1018.
3. Harrison SL, Tang EYH, Keage HAD, et al. A systematic review of the definitions of vascular cognitive impairment, no dementia in cohort studies. Dement Geriatr
Cogn Disord 2016; 42: 69–79.
4. Pendlebury ST, Mariz J, Bull L, et al. Impact of different
operational definitions on mild cognitive impairment rate
and MMSE and MoCA performance in transient ischaemic attack and stroke. Cerebrovasc Dis 2013;
36: 355–362.
5. Petersen R, Doody R, Kurz A, et al. Current concepts in
mild cognitive impairment. Arch Neurol 2001;
58: 1985–1992.
6. Stephan BCM, Minett T, Terrera GM, et al. Dementia
prediction for people with stroke in populations: is mild
cognitive impairment a useful concept? Age Ageing 2015;
44: 78–83.
7. Sachdev PS, Blacker D, Blazer DG, et al. Classifying
neurocognitive disorders: the DSM-5 approach. Nat
Rev Neurol 2014; 10: 634–642.
8. Pendlebury ST, Chen PJ, Bull L, et al. Methodological
factors in determining rates of dementia in transient
ischemic attack and stroke: (I) impact of baseline selection bias. Stroke 2015; 46: 641–646.
9. Pendlebury ST, Chen PJ, Welch SJV, et al.
Methodological factors in determining risk of dementia
after transient ischemic attack and stroke: (II) Effect of
attrition on follow-up. Stroke 2015; 46: 1494–1500.
10. Pendlebury ST, Klaus SP, Thomson RJ, et al.
Methodological factors in determining risk of dementia
after transient ischemic attack and stroke: (III) applicability of cognitive tests. Stroke 2015; 46: 3067–3073.
11. Moher D, Liberati A, Tetzlaff J, et al. Preferred reporting
items for systematic reviews and meta-analyses: the
PRISMA statement. PLoS Med 2009; 6: e1000097.
12. Sexton E, Merriman N, Donnelly N, et al. Prevalence
and incidence of cognitive impairment no dementia
(CIND) in ischaemic stroke patients: a systematic
CRD42016047840; 2016.
13. Sachdev PS, Lo JW, Crawford JD, et al. STROKOG
(stroke and cognition consortium): An international consortium to examine the epidemiology, diagnosis, and
treatment of neurocognitive disorders in relation to cerebrovascular disease. Alzheimer’s Dement Diagnosis,
Assess Dis Monit 2017; 7: 11–23.
14. Dichgans M, Wardlaw J, Smith E, et al.
METACOHORTS for the study of vascular disease and
its contribution to cognitive decline and neurodegeneration: an initiative of the Joint Programme for
Neurodegenerative Disease Research. Alzheimer’s
Dement 2016; 12: 1235–1249.
15. Crowe M and Sheppard L. A general critical appraisal
tool: An evaluation of construct validity. Int J Nurs Stud
2011; 48: 1505–1516.
16. Koo TK and Li MY. A guideline of selecting and reporting intraclass correlation coefficients for reliability
research. J Chiropr Med 2016; 15: 155–163.
17. Nyaga VN, Arbyn M and Aerts M. Metaprop: A Stata
command to perform meta-analysis of binomial data.
Arch Public Heal 2014; 72: 1–10.
18. Higgins J and Green S. Cochrane handbook for systematic
reviews of interventions. Version 5. The Cochrane
Collaboration, 2011. Available from: www.handbook.
19. Srikanth VK, Anderson JFI, Donnan GA, et al. Progressive
dementia after first-ever stroke: a community-based followup study. Neurology 2004; 63: 785–792.
20. Ihle-Hansen H, Thommessen B, Bruun Wyller T, et al.
Incidence and subtypes of MCI and dementia 1 year after
first-ever stroke in patients without pre-existing cognitive
impairment. Dement Geriatr Cogn Disord 2011;
32: 401–407.
21. Chaudhari TS, Verma R, Garg RK, et al. Clinico-radiological predictors of vascular cognitive impairment (VCI)
in patients with stroke: a prospective observational study.
J Neurol Sci 2014; 340: 150–158.
22. Saini M, Tan CS, Hilal S, et al. Computer tomography
for prediction of cognitive outcomes after ischemic cerebrovascular events. J Stroke Cerebrovasc Dis 2014;
23: 1921–1927.
23. Tang WK, Chen YK, Lu JY, et al. Absence of cerebral
microbleeds predicts reversion of vascular “cognitive
impairment no dementia” in stroke. Int J Stroke 2011;
6: 498–505.
24. Hobson P and Meara J. Cognitive function and mortality
in a community-based elderly cohort of first-ever stroke
survivors and control subjects. J Stroke Cerebrovasc Dis
2010; 19: 382–387.
25. Arauz A, Rodrıguez-Agudelo Y, Sosa AL, et al. Vascular
cognitive disorders and depression after first-ever stroke:
The Fogarty-Mexico stroke cohort. Cerebrovasc Dis
2014; 38: 284–289.
26. Salvadori E, Pasi M, Poggesi A, et al. Predictive value of
MoCA in the acute phase of stroke on the diagnosis of
mid-term cognitive impairment. J Neurol 2013;
260: 2220–2227.
27. Oksala NKJ, Jokinen H, Melkas S, et al. Cognitive
impairment predicts poststroke death in long-term
follow-up. J Neurol Neurosurg Psychiatry 2009;
80: 1230–1235.
28. Tang WK, Chan SSM, Chiu HFK, et al. Frequency and
clinical determinants of poststroke cognitive impairment
in nondemented stroke patients. J Geriatr Psychiatry
Neurol 2006; 19: 65–71.
29. Madureira S, Guerreiro M and Ferro JM. Dementia and
cognitive impairment three months after stroke. Eur J
Neurol 2001; 8: 621–627.
30. Stephens S, Kenny RA, Rowan E, et al.
Neuropsychological characteristics of mild vascular cognitive impairment and dementia after stroke. Int J Geriat
Psychiatry 2004; 19: 1053–1057.
31. Akinyemi RO, Allan L, Owolabi MO, et al. Profile and
determinants of vascular cognitive impairment in African
stroke survivors: The CogFAST Nigeria Study. J Neurol
Sci 2014; 346: 241–249.
32. Delgado C, Donoso A, Orellana P, et al. Frequency and
determinants of poststroke cognitive impairment at three
European Stroke Journal 0(0)
and twelve months in Chile. Dement Geriatr Cogn Disord
2010; 29: 397–405.
Jacquin A, Binquet C, Rouaud O, et al. Post-stroke cognitive impairment: High prevalence and determining factors in a cohort of mild stroke. J Alzheimer’s Dis 2014;
40: 1029–1038.
Yu KH, Cho SJ, Oh MS, et al. Cognitive impairment
evaluated with vascular cognitive impairment harmonization standards in a multicenter prospective stroke cohort
in Korea. Stroke 2013; 44: 786–788.
Cao M, Ferrari M, Patella R, et al. Neuropsychological
findings in young-adult stroke patients. Arch Clin
Neuropsychol 2007; 22: 133–142.
Mok VCT, Wong A, Lam WWM, et al. Cognitive
impairment and functional outcome after stroke associated with small vessel disease. J Neurol Neurosurg
Psychiatry 2004; 75: 560–566.
Makin SDJ, Turpin S, Dennis MS, et al. Cognitive
impairment after lacunar stroke: systematic review and
meta-analysis of incidence, prevalence and comparison
with other stroke subtypes. J Neurol Neurosurg
Psychiatry 2013; 84: 893–900.
Mehrabian S, Raycheva M, Petrova N, et al.
Neuropsychological and neuroimaging markers in prediction of cognitive impairment after ischemic stroke: a
prospective follow-up study. Neuropsychiatr Dis Treat
2015; 11: 2711–2719.
Sachdev PS, Brodaty H, Valenzuela MJ, et al. The
neuropsychological profile of vascular cognitive impairment in stroke and TIA patients. Neurology 2004;
62: 912–919.
Serrano S, Domingo J, Rodriguez-Garcia E, et al.
Frequency of cognitive impairment without dementia in
patients with stroke: a two-year follow-up study. Stroke
2007; 38: 105–110.
Wong A, Lau AYL, Yang J, et al. Neuropsychiatric
symptom clusters in stroke and transient ischemic
attack by cognitive status and stroke subtype: frequency
and relationships with vascular lesions, brain atrophy
and amyloid. PLoS One 2016; 11: 1–12.
Yan H, Yan Z, Niu X, et al. Dl-3-n-butylphthalide can
improve the cognitive function of patients with acute
ischemic stroke: a prospective intervention study.
Neurol Res 2017; 39: 337–343.