Professional Documents
Culture Documents
Background and Purpose—Silent brain infarction (SBI) on magnetic resonance imaging has been proposed as a subclinical
risk marker for future symptomatic stroke. We performed a systematic review and meta-analysis to summarize the
association between magnetic resonance imaging–defined SBI and future stroke risk.
Methods—We searched the medical literature to identify cohort studies involving adults with SBI detected by magnetic
resonance imaging who were subsequently followed up for incident clinically defined stroke. Study data and quality
assessment were recorded in duplicate with disagreements in data extraction resolved by a third reader. Strength association
between magnetic resonance imaging–detected SBI and future symptomatic stroke was measured by an hazard ratio.
Results—The meta-analysis included 13 studies (14 764 subjects) with a mean follow-up ranging from 25.7 to 174 months. SBI
predicted the occurrence of stroke with a random effects crude relative risk of 2.94 (95% confidence interval, 2.24–3.86, P<0.001;
Q=39.65, P<0.001). In the 8 studies of 10 427 subjects providing hazard ratio adjusted for cardiovascular risk factors, SBI was
an independent predictor of incident stroke (hazard ratio, 2.08 [95% confidence interval, 1.69–2.56; P<0.001]; Q=8.99; P=0.25).
In a subgroup analysis pooling 9483 stroke-free individuals from large population-based studies, SBI was present in ≈18% of
participants and remained a strong predictor of future stroke (hazard ratio, 2.06 [95% confidence interval, 1.64–2.59]; P<0.01).
Conclusions—SBI is present in ≈1 in 5 stroke-free older adults and is associated with a 2-fold increased risk of future
stroke. Future studies of in-depth stroke risk evaluations and intensive prevention measures are warranted in patients
with clinically unrecognized radiologically evident brain infarctions. (Stroke. 2016;47:719-725. DOI: 10.1161/
STROKEAHA.115.011889.)
Key Words: brain infarction ◼ infarction ◼ magnetic resonance imaging ◼ risk factors ◼ stroke
Continuing medical education (CME) credit is available for this article. Go to http://cme.ahajournals.org to take the quiz.
Received October 20, 2015; final revision received December 24, 2015; accepted December 29, 2015.
From the Department of Radiology (A.G., C.F.), Feil Family Brain and Mind Research Institute (A.G., G.G., H.K.), Department of Healthcare Policy and
Research (A.E.G.), Samuel J. Wood Library and C.V. Starr Biomedical Information Center (D.D.), Department of Neurology (H.K.), Weill Cornell Medical
College, New York, NY; Department of Neurology, Columbia University Medical Center, New York, NY (J.G.); Department of Neurology, University
of Miami Miller School of Medicine, FL (C.W.); Department of Biostatistics, Boston University School of Public Health, MA (A.S.B.); Department of
Neurology, Boston University School of Medicine, MA (S.S.); and Department of Health Policy and Management, Harvard T.H. Chan School of Public
Health, Boston, MA (A.P.).
Presented in part at the International Stroke Conference of the American Heart Association, Los Angeles, CA, February 17–19, 2016.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.
115.011889/-/DC1.
Correspondence to Ajay Gupta, MD, 525 E 68th St, Starr 8A, Box 141, New York, NY 10065. E-mail ajg9004@med.cornell.edu
© 2016 American Heart Association, Inc.
Stroke is available at http://stroke.ahajournals.org DOI: 10.1161/STROKEAHA.115.011889
719
Downloaded from http://stroke.ahajournals.org/ at VA MED CTR BOISE on February 28, 2016
720 Stroke March 2016
and expense of trials that would otherwise have relied on stroke Because no standardized tool exists to assess the risk of bias in
as a primary outcome measure.7 Although there have been many observational studies, we adapted the bias assessment criteria adapt-
ed from recently published meta-analyses focused on SBI in patients
individual studies describing the predictive value of SBI, rely-
with atrial fibrillation10 and imaging biomarkers of stroke risk.11,12 A
ing on single-study samples results in wide confidence inter- total of 11 questions were generated to evaluate potential selection,
vals (CIs) for risk estimates. For these reasons, we performed detection, misclassification, reporting, attrition, and confounding
a systematic review and meta-analysis evaluating whether MRI bias13 (Table I in the online-only Data Supplement). Risks of bias
detection of SBI is a predictor of subsequent stroke. questions were assessed by 2 readers, with disagreements in assess-
ment resolved by a third tie-breaking evaluator.
Methods
We performed this study following the guidelines recommended by the Data Synthesis and Analysis
Meta-Analysis of Observational Studies in Epidemiology (MOOSE) We estimated the prevalence of SBI in the included studies and the
group8 and the Preferred Reporting Items for Systematic Reviews and total person-years of follow-up in each study. Meta-analyses of the
Meta-Analyses statement.9 We also prospectively registered our study individual study crude RRs (ie, RR of stroke in the presence of an
protocol on the International Prospective Register of Systematic SBI) and covariate-adjusted HRs were conducted with the use of
Reviews (PROSPERO registration number CRD42014007016). StatsDirect statistical software (version 2.7.9; 7/9/2012 StatsDirect
Ltd, Cheshire, England). Each pooled risk ratio was calculated using
a random effects (DerSimonian–Laird) model,14 and forest plots were
Data Sources and Searches generated to display the individual study risk ratios and the pooled
A research librarian performed comprehensive searches from data- risk ratio. We performed all analyses using a random effects model
base inception on April 3, 2015, in Ovid MEDLINE, Ovid Embase, based on the conservative assumption that included studies did not
and the Cochrane Library. An English language filter was not applied. have exactly the same effect size given the potential for heterogeneity
The first search was conducted in Ovid MEDLINE. Subject head- between studies in terms of sample size, subject characteristics, and
ings and keywords were adapted for the other databases. Additional testing methods. To assess the combinability of the risk ratios, we
records were identified by using the cited by and view references fea- calculated the P value from the Cochrane Q statistical heterogene-
tures in Scopus on April 17, 2015 (search methodology details are ity test. The results of each study were expressed as a risk ratio with
given in Methods section in the online-only Data Supplement). a 95% CI. For each meta-analysis, the presence of publication bias
was evaluated through a Begg–Mazumdar rank correlation test. All P
values <0.05 were considered statistically significant.
Study Selection We performed subgroup analyses limited to the following patient
We included only studies with MRI characterization of SBI in sub- samples: (1) stroke-free participants recruited from population-based
jects subsequently followed up for the development of future clini- or community-dwelling samples; (2) stroke-free participants in stud-
cally overt stroke. Specific inclusion criteria were (1) studies of adult ies with inclusion criteria including at least 1 known stroke risk factor
subjects (aged >18 years); (2) at least 100 subjects; (3) MRI deter- (such as hypertension, diabetes mellitus, atrial fibrillation, coronary
mination of SBI as lesions measuring ≥3 mm with differentiation of artery disease, hyperlipidemia, or chronic kidney disease); and (3)
SBI from leukoaraiosis; (4) mean follow-up of >12 months after brain patients with a history of documented previous stroke followed up
MRI; and (5) clinical ascertainment of stroke during follow-up. In for recurrent stroke.
cases where the methods for detecting SBI or outcome data were not
clear in the article, we attempted to contact the corresponding author
for additional details. Furthermore, if test data from a cohort were Role of Funding Source
published more than once, only the article with the largest person- This study received no external funding.
years of follow-up was included to minimize the analysis of duplicate
or overlapping samples.
Results
Data Extraction and Quality Assessment Study Selection
A single investigator read the title and abstract of all references pro-
duced by our database search. After preliminary articles were short- We screened a total of 1654 titles and abstracts from which
listed as potentially eligible, the articles were read in their entirety by 2 we identified 13 articles that met all inclusion criteria for the
readers to determine eligibility, with disagreements resolved by consen- systematic review.15–27 One study27 that met our inclusion cri-
sus. Data were extracted from articles meeting the inclusion criteria by teria was published after our search was completed during the
a team of 2 readers using a prespecified data collection template, with data extraction phase but was included as it provided follow-up
disagreements in data extraction resolved by a third tie-breaking reader.
Study characteristics that were extracted included the first author of cohort data for an article initially included in our original liter-
the study, study design (prospective or not), major study inclusion cri- ature search.28 Study selection steps are summarized in Figure I
teria, country of the study, total number of subjects, mean follow-up, in the online-only Data Supplement. A crude RR expressing
SBI status at baseline, and the prevalence of stroke risk factors in the the association between SBI and incident stroke was calculable
studied populations, including age, sex, hypertension, diabetes melli-
in all 13 studies from raw data. The most commonly provided
tus, chronic kidney disease, atrial fibrillation, coronary artery disease,
hyperlipidemia, and smoking history. Additional study extraction fo- adjusted risk metric was the HR, with 8 studies15,17,19,20,22,23,25,27
cused on SBI status and stroke outcomes during follow-up, including providing covariate-adjusted HRs of SBI as a predictor of
the number of subjects with and without SBI at baseline, the number of incident stroke. One study provided a covariate-adjusted RR
strokes (all strokes, ischemic strokes, and hemorrhagic strokes) that oc- of SBI as a predictor of stroke,16 and 2 studies18,24 provided
curred during follow-up in both SBI-positive and SBI-negative groups,
the covariate-adjusted RR measure (hazard ratio [HR] or odds ratio) covariate-adjusted or age- and sex-matched odds ratios.
relating SBI and future stroke, and the specific vascular risk factors for
which the RR measures were adjusted in multivariate analyses. Data Qualitative Study Characteristics
were also extracted about the imaging definitions of SBI in each study,
including MRI magnet field strength, MRI slice section thickness, MRI Of the 13 articles meeting inclusion criteria (Table II in the
section gap, SBI size classification, SBI MRI signal characteristics, and online-only Data Supplement), all except one22 were prospec-
means of differentiating SBI from perivascular spaces. tive studies. One study was conducted as an international
Figure. Forest plots of the association between magnetic resonance imaging-determined silent brain infarction and future stroke in all
included studies.15–27 Meta-analysis performed using a random effects model, with crude relative risks pooled in (A) and adjusted-hazard
ratios pooled in (B). Squares represent point estimates for the effect sizes. The size of the squares is proportional to the inverse of the
variance of the estimate. Diamond represents the pooled estimate, and the horizontal lines represent the 95% confidence interval.
multicenter study,24 5 were in Japan,16–18,21,26 4 in the United studies were focused on evaluation of stroke in younger patients
States,15,19,25,27 and 1 each in Finland,20 Canada,22 and the with mean ages of 39.422 and 40.0 years.20 Subjects in all stud-
Netherlands.23 Eleven studies15–19,21,23–27 included cohorts with ies were followed up for at least 25.7 months (range, 25.7–174
mean or median ages >50 years (range, 56.0–75.2), whereas 2 months) for ascertainment of clinically defined stroke.
Definitions of SBI and Stroke Outcomes Association Between SBI and Future Stroke:
All studies defined SBI as lesions ≥3 mm that were hyperin- Adjusted HRs
tense on T2-weighted images (details are given in Table III in Eight of the included studies provided adjusted HRs describ-
the online-only Data Supplement). One study provided addi- ing the strength of association between SBI and future stroke.
tional separate analysis of putative vascular lesions <3 mm In the analysis of these 8 studies (10 427 subjects), there was
that were too small to definitely characterize as SBI.27 All a mean follow-up ranging from 25.7 to 174 months (mean,
studies described the use of additional MRI pulse sequences ≈88.8), yielding a total of 85 001 person-years of follow-up.
to differentiate SBI from adjacent white-matter leukoaraio- There was neither significant heterogeneity (Q=8.99; P=0.25)
sis, with most studies relying on T1-weighted hypointensity nor significant publication bias (Kendall τ score, 0.29; P=0.40)
as a feature suggestive of SBI rather than nonspecific white- present in this analysis. We found a significant positive rela-
matter leukoaraiosis. There were variable methods used to tionship between the presence of SBI and the risk of stroke
distinguish SBI from dilated perivascular spaces, including with a random effects adjusted HR of 2.08 (95% CI, 1.69–
the presence of hyperintensity on T2-weighted fluid-atten- 2.56; P<0.001; Figure [B]). Of the total study sample, 1962
uated inversion recovery images,18,20–22,24,26 proton density subjects (18.8%) had a positive MRI test for SBI, whereas
hyperintensity,15,18,27 and lesion morphology/location.19,25 8465 (81.2%) had a negative test for SBI.
MRI magnet field strengths varied from 0.2 to 3.0 T, with
most studies using a 1.5-T scanner. Most studies used clini- Subgroup Meta-Analysis Results
cally based definitions of ischemic stroke based on a focal Statistically significant random effects crude RR (Table 1)
neurological deficit lasting >24 hours and without evidence and adjusted HR (Table 2) were preserved in the following
of hemorrhage on brain imaging (details are given in Table subgroup analyses: (1) stroke-free patients recruited from
IV in the online-only Data Supplement). A majority of stud- population-based or community-dwelling studies; (2) stroke-
ies used a combination of hospital and outpatient medical free participants in studies with inclusion criteria requiring
records and telephone interviews to ascertain stroke outcome at least 1 known stroke risk factor, including hypertension,16
events. chronic renal disease requiring hemodialysis,17 diabetes mel-
litus,21 or >1 cardiovascular risk factor26; and (3) patients pre-
Association Between SBI and Future Stroke: Crude senting with first-time symptomatic acute stroke followed up
RRs for recurrent stroke in whom clinically SBI was evident on
We were able to obtain sufficient raw data to calculate a baseline imaging performed for the initial acute stroke diag-
crude RR for future stroke in the presence of SBI for each nostic evaluation. Of the studies providing an adjusted HR,
of the included 13 studies (Table V in the online-only Data the risk of stroke in the presence of SBI conferred ≈2-fold
Supplement). In this analysis of 13 studies (14 764 subjects), increased risk of future stroke in both stroke-free patients
there was a mean follow-up ranging from 25.7 to 174 months (HR, 2.06; 95% CI, 1.64–2.59) and in patients with previous
(mean, ≈76.0), yielding a total of 108 360 person-years of stroke (HR, 2.00; 95% CI, 1.08–3.71). On the basis of the 5
follow-up. There was statistically significant heterogene- studies15,19,23,25,27 with adjusted HRs derived from population-
ity (Q=39.65; P<0.001), but no significant publication bias based studies of 9483 stroke-free individuals with mean or
(Kendall τ score, 0.26; P=0.25) was present in this analy- median study ages between 62 and 76, the prevalence of SBI
sis. We found a significant positive relationship between the was 18.7%.
presence of SBI and the risk of stroke with a random effects
crude RR of 2.94 (95% CI, 2.24–3.86; P<0.001; Figure [A]). Assessment of the Quality of the Included Studies
Of the total study sample, 3007 subjects (20.4%) had a posi- The results from the quality assessment questionnaire are
tive MRI test for SBI, whereas 11 757 (79.6%) had a nega- shown in Table VI in the online-only Data Supplement. All
tive test for SBI. included studies involved >1 investigator independently
Table 1. Meta-Analysis Summary for Subgroup Analyses of Crude RR of Stroke in Patients With Magnetic Resonance Imaging–
Defined SBI
Test of
Heterogeneity Publication Bias*
Total Person- SBI
No. of Years of Prevalence, Pooled RR Cochran Q,
Subgroup Studies Follow-Up % (95% CI) RR, P Value P Value Kendall τ P Value
Stroke-free individuals recruited from population- 6 95 449.2 17.7 2.81 (2.40–3.28) <0.01 <0.01 0.07 0.99
based or community-dwelling studies15,18,19,23,25,28
Stroke-free individuals with at least 1 known stroke 4 5813.4 38.1 4.54 (2.76–7.45) <0.01 0.50 0.33 0.75
risk factor16,17,21,26
Known history of stroke followed up for stroke 3 7097.6 27.5 1.99 (1.30–3.04) <0.01 0.21 … …
recurrence20,22,24
CI indicates confidence interval; RR, relative risk; and SBI, silent brain infarction.
*Publication bias test only performed on >3 studies.
Table 2. Meta-Analysis Summary for Subgroup Analyses of Risk Factor–Adjusted Hazard Ratios of Stroke in Patients With
Magnetic Resonance Imaging–Defined SBI
Test of
Heterogeneity Publication Bias*
Total Person-
No. of Years of SBI Pooled HR Cochran Q,
Subgroup Studies Follow-Up Prevalence, % (95% CI) HR, P Value P Value Kendall τ P Value
Stroke-free individuals recruited from population-based 5 78 540 18.7 2.06 (1.64–2.59) <0.01 0.24 0 0.82
or community-dwelling samples15,19,23,25,27
Known history of stroke followed up for stroke 2 6062.6 16.2 2.00 (1.08–3.71) 0.03 0.24 … …
recurrence20,22
CI indicates confidence interval; HR, hazard ratio; and SBI, silent brain infarction.
*Publication bias test only performed on >3 studies.
evaluating MRI studies for the presence of SBI. In all but 2 Few guidelines exist on how to manage patients with inci-
studies,16,17 perivascular spaces were systematically differ- dentally found SBI. In contrast, professional societies publish
entiated from SBI. In 7 of the 13 studies,15,18,19,23–26 the risk detailed guidelines on secondary stroke prevention in patients
of selection bias was minimized by either random selec- with clinically overt stroke.29 Given that the likelihood of a
tion of subjects or recruitment from a community-dwelling brain lesion presenting clinically depends more on the elo-
population. Six studies explicitly reported that investiga- quence of the brain region affected rather than the underly-
tors were blinded to SBI status when determining stroke ing pathogenesis of the lesion,30,31 the distinction between
outcomes.15,19,20,23,25,27 All but 2 studies21,26 corrected for silent and overt brain ischemia may need reexamination. In
covariate risk factors in their analysis of the association this meta-analysis, we found that SBI confers a similar rela-
between SBI and incident stroke and provided an adjusted tive degree of risk of future stroke as a previous history of
disease association measure, such as a HR or odds ratio. clinically overt stroke in certain settings. In patients with atrial
Five studies15,19,23,25,27 had the lowest overall risk of bias, fibrillation, for example, a previous history of stroke is associ-
with each of these studies demonstrating low potential for ated with an HR of 2.3 for recurrent stroke,32 comparable with
bias in all questions but one. All 5 of these studies were the magnitude of risk we determined in our meta-analysis for
of stroke-free individuals recruited from large population- the first-time symptomatic stroke after SBI. Such consider-
based studies, including the Cardiovascular Health Study ations raise the question of whether SBI should be considered
(CHS),15 the Rotterdam Scan Study,23 the Framingham as a manifestation of cerebrovascular disease that requires
Offspring Cohort Study,19 the Northern Manhattan Study intensive evaluation and secondary prevention, as in clinically
(NOMAS),25 and the Atherosclerosis Risk in Communities overt stroke, rather than simply being considered an incidental
Study (ARIC).27 finding.
Our study has revealed some limitations about the MRI
Discussion techniques used to detect SBI in the existing body of literature
SBI is often incidentally detected in patients during MRI eval- summarizing stroke risk after SBI. First, although there was
uation. In this systematic review and meta-analysis of stud- general agreement on the MRI features of SBI, including size
ies involving >14 000 subjects and >100 000 person-years of (≥3 mm, T2 hyperintensity distinguishable from leukoaraio-
follow-up, we found that the presence of SBI conferred an sis), we found variability in the methods used to determine the
≈3-fold increased risk of subsequent stroke. After adjustment presence of SBI, including MRI magnet field strength (range,
for potentially confounding vascular risk factors, the presence 0.2–3.0 T) and specific MRI sequences obtained for evalua-
of an SBI was associated with ≈2-fold higher risk of future tion. Most of the included studies (12 of 13) did not perform
stroke. This approximate doubling of the independent risk MRI on higher field strength 3-T magnets, which are becom-
of incident stroke was seen in both stroke-free patients and ing increasingly common and are more likely to allow detec-
patients with previous stroke. Approximately 1 in 5 stroke- tion of small lesions. The prevalence and predictive value
free individuals in these studies had SBI. of SBI detected on these higher field MRI magnets require
Our results are particularly applicable to stroke-free, com- additional study. Furthermore, the methods used to distinguish
munity-dwelling individuals aged >60 years who are inciden- SBI from perivascular spaces, a potential mimic on imaging
tally found to have an SBI on brain MRI. In our subgroup studies, were variably reported and raise the possibility that
analysis of 5 large, diverse population-based studies (CHS, some perivascular spaces were misclassified as SBI. Such
ARIC, NOMAS, Rotterdam Scan Study, and the Framingham misclassification, however, may not significantly affect our
Offspring Cohort Study), the presence of an SBI conferred a study’s overall conclusions given recent data suggesting that
≈2-fold increase in incident stroke risk after adjustment for perivascular spaces are a manifestation of cerebral small ves-
vascular risk factors. These 5 studies had the lowest risk of sel disease33 and that small brain lesions <3 mm that are too
bias in our assessment and pooled together >9400 subjects small to definitely distinguish from small perivascular spaces
totaling ≈78 000 person-years of follow-up, thereby increas- are still predictive of incident stroke.27 Nonetheless, our study
ing our confidence in the conclusion from this analysis. suggests that further standardization of SBI definitions would
allow for the more widespread use of SBI as an imaging risk Northern Manhattan Study (NOMAS). Drs Seshadri and Beiser report
biomarker.6,34 grant support (NIH NS017950 and AG0008122) for the Framingham
Cohort Study. Dr Kamel reports grant support from the National
Several additional limitations of our study are important to Institute of Neurological Disorders and Stroke (K23NS082367).
consider. First, we performed one of our meta-analyses using a
crude RR because the covariate risk factors varied across stud-
ies. However, the relatively higher quality studies included in
Disclosures
None.
our analysis provided adjusted HRs, and although there were
some differences in the specific risk factors for which each study
performed statistical correction, it seems unlikely that these References
1. Mozaffarian D, Benjamin EJ, Go AS, Arnett DK, Blaha MJ, Cushman
relatively minor interstudy differences affected our overall con- M, et al; American Heart Association Statistics Committee and Stroke
clusions. Second, given the variability in length of follow-up Statistics Subcommittee. Heart disease and stroke statistics–2015
between studies, the existing data are not amenable to the calcu- update: a report from the American Heart Association. Circulation.
lation of absolute stroke rates. Third, there was variable report- 2015;131:e29–e322. doi: 10.1161/CIR.0000000000000152.
2. Fisher CM. Lacunes: small, deep cerebral infarcts. Neurology.
ing of stroke subtypes (ischemic versus hemorrhagic) in studies. 1965;15:774–784.
It would seem likely that SBI would more strongly predict isch- 3. Longstreth WT Jr, Dulberg C, Manolio TA, Lewis MR, Beauchamp NJ
emic rather than hemorrhage stroke risk, although further work Jr, O’Leary D, et al. Incidence, manifestations, and predictors of brain
infarcts defined by serial cranial magnetic resonance imaging in the
is warranted to investigate this issue. Fourth, it is unclear in a
elderly: the Cardiovascular Health Study. Stroke. 2002;33:2376–2382.
majority of studies, especially the large population-based stud- 4. Vermeer SE, Prins ND, den Heijer T, Hofman A, Koudstaal PJ, Breteler
ies, to what extent imaging confirmation of stroke outcomes MM. Silent brain infarcts and the risk of dementia and cognitive decline.
occurred. A more standardized definition of stroke outcomes N Engl J Med. 2003;348:1215–1222. doi: 10.1056/NEJMoa022066.
5. Bokura H, Kobayashi S, Yamaguchi S, Iijima K, Nagai A, Toyoda
using modern imaging-based criteria would be helpful in future G, et al. Silent brain infarction and subcortical white matter lesions
studies of SBI and stroke risk. Fifth, stroke prevention measures increase the risk of stroke and mortality: a prospective cohort
have substantially decreased the incidence of stroke1 since the study. J Stroke Cerebrovasc Dis. 2006;15:57–63. doi: 10.1016/j.
jstrokecerebrovasdis.2005.11.001.
time many of the cohorts in this study were recruited (some >20
6. Fanning JP, Wesley AJ, Wong AA, Fraser JF. Emerging spectra of
years ago), possibly limiting the validity of using these data to silent brain infarction. Stroke. 2014;45:3461–3471. doi: 10.1161/
extrapolate stroke risk in present-day populations. Despite this STROKEAHA.114.005919.
reduction in absolute stroke risk in the general population, we 7. Vickrey BG, Brott TG, Koroshetz WJ; Stroke Research Priorities
Meeting Steering Committee and the National Advisory Neurological
think that the longer term trends in stroke incidence are unlikely Disorders and Stroke Council; National Institute of Neurological
to substantially affect our study’s main conclusions given that Disorders and Stroke. Research priority setting: a summary of the 2012
our focus was on RR. Furthermore, to our knowledge, there are NINDS Stroke Planning Meeting Report. Stroke. 2013;44:2338–2342.
no data to suggest that the recent reductions in stroke incidence doi: 10.1161/STROKEAHA.113.001196.
8. Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D,
would be different in patients with or without SBI. Finally, the et al. Meta-analysis of observational studies in epidemiology: a proposal
RR and HR we determined in our meta-analyses are probably for reporting. Meta-analysis Of Observational Studies in Epidemiology
most applicable to older adults given that the majority of par- (MOOSE) group. JAMA. 2000;283:2008–2012.
9. Liberati A, Altman DG, Tetzlaff J, Mulrow C, Gøtzsche PC, Ioannidis JP,
ticipants in these studies were middle aged or elderly. Further et al. The PRISMA statement for reporting systematic reviews and meta-
studies are warranted to assess whether the future stroke risk analyses of studies that evaluate health care interventions: explanation
conferred by SBI in young adults is substantially different com- and elaboration. Ann Intern Med. 2009;151:W65–W94.
pared with relatively older individuals. 10. Kalantarian S, Ay H, Gollub RL, Lee H, Retzepi K, Mansour M, et al.
Association between atrial fibrillation and silent cerebral infarctions: a
In summary, our systematic review and meta-analysis sug- systematic review and meta-analysis. Ann Intern Med. 2014;161:650–
gests that SBI is significantly associated with an increased 658. doi: 10.7326/M14-0538.
risk of symptomatic stroke. Our study specifically suggests 11. Gupta A, Kesavabhotla K, Baradaran H, Kamel H, Pandya A, Giambrone
AE, et al. Plaque echolucency and stroke risk in asymptomatic carotid
that ≈1 in 5 older, stroke-free adults may be harboring an SBI,
stenosis: a systematic review and meta-analysis. Stroke. 2015;46:91–97.
which may in turn confer more than double the risk of future doi: 10.1161/STROKEAHA.114.006091.
first-time stroke. These data indicate a need for future studies 12. Gupta A, Baradaran H, Schweitzer AD, Kamel H, Pandya A, Delgado
of in-depth stroke risk evaluations and intensive prevention D, et al. Carotid plaque MRI and stroke risk: a systematic review
and meta-analysis. Stroke. 2013;44:3071–3077. doi: 10.1161/
measures, including lifestyle modification and proven medi- STROKEAHA.113.002551.
cal therapies for stroke reduction, in patients with these com- 13. Higgins JPT. Chapter 8: Assessing risk of bias in included studies. In:
monly discovered brain lesions. Green S, ed. Cochrane Handbook for Systematic Reviews of Interventions
Version 5.1.0 [Updated March 2011]. The Cochrane Collaboration;
2011. http://handbook.cochrane.org/.
Acknowledgments 14. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin
We thank Drs Jukka Putaala, Toshitaka Umemura, M. Arfan Ikram, Trials. 1986;7:177–188.
15. Bernick C, Kuller L, Dulberg C, Longstreth WT Jr, Manolio T,
Marileen L. Portegies, Beverly G. Windham, Michael E. Griswold,
Beauchamp N, et al; Cardiovascular Health Study Collaborative Reseach
and Wanmei Wang for providing both unpublished data and clarifica- Group. Silent MRI infarcts and the risk of future stroke: the cardiovascu-
tion of published data used in this meta-analysis. lar health study. Neurology. 2001;57:1222–1229.
16. Kario K, Shimada K, Schwartz JE, Matsuo T, Hoshide S, Pickering
TG. Silent and clinically overt stroke in older Japanese subjects
Sources of Funding with white-coat and sustained hypertension. J Am Coll Cardiol.
Dr Gupta reports support in part by the Foundation of the American 2001;38:238–245.
Society of Neuroradiology Scholar Award. Dr Wright reports grants 17. Naganuma T, Uchida J, Tsuchida K, Takemoto Y, Tatsumi S,
support (National Institutes of Health [NIH] NS 29993) for the Sugimura K, et al. Silent cerebral infarction predicts vascular events
in hemodialysis patients. Kidney Int. 2005;67:2434–2439. doi: 26. Miwa K, Tanaka M, Okazaki S, Furukado S, Sakaguchi M, Mochizuki H,
10.1111/j.1523-1755.2005.00351.x. et al. Association between interleukin-6 levels and first-ever cerebrovas-
18. Bokura H, Kobayashi S, Yamaguchi S, Iijima K, Nagai A, Toyoda cular events in patients with vascular risk factors. Arterioscler Thromb
G, et al. Silent brain infarction and subcortical white matter lesions Vasc Biol. 2013;33:400–405. doi: 10.1161/ATVBAHA.112.300350.
increase the risk of stroke and mortality: a prospective cohort 27. Windham BG, Deere B, Griswold ME, Wang W, Bezerra DC, Shibata
study. J Stroke Cerebrovasc Dis. 2006;15:57–63. doi: 10.1016/j. D, et al. Small brain lesions and incident stroke and mortality: a cohort
jstrokecerebrovasdis.2005.11.001. study. Ann Intern Med. 2015;163:22–31. doi: 10.7326/M14-2057.
19. Debette S, Beiser A, DeCarli C, Au R, Himali JJ, Kelly-Hayes M, 28. Windham BG, Griswold ME, Shibata D, Penman A, Catellier DJ,
et al. Association of MRI markers of vascular brain injury with inci- Mosley TH Jr. Covert neurological symptoms associated with
dent stroke, mild cognitive impairment, dementia, and mortality: the silent infarcts from midlife to older age: the Atherosclerosis Risk
Framingham Offspring Study. Stroke. 2010;41:600–606. doi: 10.1161/ in Communities study. Stroke. 2012;43:1218–1223. doi: 10.1161/
STROKEAHA.109.570044. STROKEAHA.111.643379.
20. Putaala J, Haapaniemi E, Kurkinen M, Salonen O, Kaste M, Tatlisumak 29. Kernan WN, Ovbiagele B, Black HR, Bravata DM, Chimowitz MI,
T. Silent brain infarcts, leukoaraiosis, and long-term prognosis in young Ezekowitz MD, et al; American Heart Association Stroke Council,
ischemic stroke patients. Neurology. 2011;76:1742–1749. doi: 10.1212/ Council on Cardiovascular and Stroke Nursing, Council on Clinical
WNL.0b013e31821a44ad. Cardiology, and Council on Peripheral Vascular Disease. Guidelines for
21. Umemura T, Kawamura T, Umegaki H, Mashita S, Kanai A, Sakakibara T, the prevention of stroke in patients with stroke and transient ischemic
et al. Endothelial and inflammatory markers in relation to progression attack: a guideline for healthcare professionals from the American Heart
of ischaemic cerebral small-vessel disease and cognitive impairment: Association/American Stroke Association. Stroke. 2014;45:2160–2236.
a 6-year longitudinal study in patients with type 2 diabetes mellitus. doi: 10.1161/STR.0000000000000024.
J Neurol Neurosurg Psychiatry. 2011;82:1186–1194. doi: 10.1136/ 30. Leary MC, Saver JL. Annual incidence of first silent stroke in the United
jnnp.2010.217380. States: a preliminary estimate. Cerebrovasc Dis. 2003;16:280–285.
22. Gioia LC, Tollard É, Dubuc V, Lanthier S, Deschaintre Y, Chagnon M, 31. Brott T, Tomsick T, Feinberg W, Johnson C, Biller J, Broderick J,
et al. Silent ischemic lesions in young adults with first stroke are asso- et al. Baseline silent cerebral infarction in the Asymptomatic Carotid
ciated with recurrent stroke. Neurology. 2012;79:1208–1214. doi: Atherosclerosis Study. Stroke. 1994;25:1122–1129.
10.1212/WNL.0b013e31826aacac. 32. Diener HC, Eikelboom J, Connolly SJ, Joyner CD, Hart RG, Lip GY, et
23. Poels MM, Steyerberg EW, Wieberdink RG, Hofman A, Koudstaal PJ, al; AVERROES Steering Committee and Investigators. Apixaban versus
Ikram MA, et al. Assessment of cerebral small vessel disease predicts aspirin in patients with atrial fibrillation and previous stroke or transient
individual stroke risk. J Neurol Neurosurg Psychiatry. 2012;83:1174– ischaemic attack: a predefined subgroup analysis from AVERROES,
1179. doi: 10.1136/jnnp-2012-302381. a randomised trial. Lancet Neurol. 2012;11:225–231. doi: 10.1016/
24. Weber R, Weimar C, Wanke I, Möller-Hartmann C, Gizewski ER, S1474-4422(12)70017-0.
Blatchford J, et al; PRoFESS Imaging Substudy Group. Risk of recurrent 33. Potter GM, Doubal FN, Jackson CA, Chappell FM, Sudlow CL, Dennis
stroke in patients with silent brain infarction in the Prevention Regimen MS, et al. Enlarged perivascular spaces and cerebral small vessel disease.
for Effectively Avoiding Second Strokes (PRoFESS) imaging substudy. Int J Stroke. 2015;10:376–381. doi: 10.1111/ijs.12054.
Stroke. 2012;43:350–355. doi: 10.1161/STROKEAHA.111.631739. 34. Wardlaw JM, Smith EE, Biessels GJ, Cordonnier C, Fazekas F, Frayne
25. Di Tullio MR, Jin Z, Russo C, Elkind MS, Rundek T, Yoshita M, et al. R, et al; STandards for ReportIng Vascular changes on nEuroimaging
Patent foramen ovale, subclinical cerebrovascular disease, and ischemic (STRIVE v1). Neuroimaging standards for research into small vessel
stroke in a population-based cohort. J Am Coll Cardiol. 2013;62:35–41. disease and its contribution to ageing and neurodegeneration. Lancet
doi: 10.1016/j.jacc.2013.03.064. Neurol. 2013;12:822–838. doi: 10.1016/S1474-4422(13)70124-8.
The online version of this article, along with updated information and services, is located on the
World Wide Web at:
http://stroke.ahajournals.org/content/47/3/719
Permissions: Requests for permissions to reproduce figures, tables, or portions of articles originally published
in Stroke can be obtained via RightsLink, a service of the Copyright Clearance Center, not the Editorial Office.
Once the online version of the published article for which permission is being requested is located, click
Request Permissions in the middle column of the Web page under Services. Further information about this
process is available in the Permissions and Rights Question and Answer document.
Ovid MEDLINE(R) In-Process & Other Non-Indexed Citations and Ovid MEDLINE(R) 1946 to Present - Search ran 04/03/2015
The English language filter was not applied. The first search was conducted in Ovid MEDLINE. Subject headings and key words were adapted for the other databases.
Supplemental Table I: Risk of Bias Questions to Assess the Quality of Included Studies
stroke-free subjects,
prospective Cardiovascular
1 Bernick 20011 randomly sampled, United States 50.4 SBI positive 923 76* 543 (58.8) 526 (57) 123 (13.3) No data 22 (2.4) 216 (23.4) 208.4† 1.2‡
cohort Health Study
population-based cohort
SBI negative 2401 73.1* 1437 (59.9) 1118 (46.6) 317 (13.2) No data 65 (2.7) 411 (17.1) 207.1† 0.5‡
SBI negative 60 49.2+/-12.8 17(28.3) 51 (85) 10 (16.7) 60 (100) No data 3 (5) 4 (6.6) 8 (13.3)
SBI negative 2304 57.1 +/- 7.1 1069 (46.4) 894 (38.8) 221 (9.6) No data 14 (0.6) No data 733 (31.8) 740 (32.1)
community-based,
Framingham
prospective prospective study, offspring of
5 Debette 20105 Offspring Cohort United States 67.2 SBI positive 253 65+/-9 126 (49.8) 146 (57.7) 40 (15.8) No data No data No data No data 33 (13.0)
cohort original Framingham Heart
Study
Study participants
SBI negative 1975 62 +/- 9 923 (46.7) 770 (39.0) 215 (10.9) No data No data No data No data 234 (12.0)
SBI negative 144 62.7 +/-8.1 102 (53.7) 96 (50.5) 144 (100) No data No data No data 95 (50) 38 (20.0)
SBI negative 122 38.2+/-8.7 54(43.8) 32 (26.2) 12 (9.8) No data No data 5 (4.1) 74 (60.7) 56 (45.9)
stroke-free subjects,
prospective The Rotterdam Total sample with SBI
9 Poels 2012§9 randomly sampled, Netherlands 120 210 || 72+/-7.4 522 (51.8) 332 (33.0) 66 (6.6) No data 28 (2.8) 96 (9.5) No data 176 (17.5)
cohort Scan Study status known
population-based cohort
prospective subjects with recent ischemic Prevention
Multicenter
cohort data stroke within 120 days of Regimen for
trial (35
10 Weber 201210 from a study, SBI positive and Effectively 30 SBI positive 207 66.2 +/-8.5 57 (27.5) 162 (78.2) 67 (32.3) No data 8 (3.8) No data 99 (47.8) 135
countries or
randomized negative patients matched by Avoiding Second
regions)
controlled trial age and sex Strokes trial
SBI negative 207 66.2 +/-8.5 57 (27.5) 155 (74.7) 63 (30.4) No data 6 (2.9) No data 111 (53.8) 130
prospective stroke-free subjects, Northern Total sample with SBI
11 Di Tullio 2013 ||, §11 United States 85.2 1287 71+/-9 779 (61) 934 (73) 291 (23) No data 31 (2.4) 160 (12.4) 504 (39) 119 (9)
cohort population-based cohort Manhattan Study status known
Osaka Follow-up
stroke-free, ambulatory
prospective Study for Carotid
12 Miwa 201312 subjects with >1 Japan 57.6 SBI positive 130 68.8+/-8.6 237 (51) 315(68) 79 (17) No data No data No data 246(53) 70 (15)
cohort Atherosclerosis,
cardiovascular risk factor
part 2
SBI negative 334 68.8+/-8.6 237 (51) 315(68) 79 (17) No data No data No data 246(53) 70 (15)
The
Atherosclerosis
prospective stroke-free subjects,
13 Windham 2015¶13 Risk in United States 174 SBI positive 220 64.0+/-4.5 134 (61) 143 (65) 50 (23) No data No data 17 (8) No data 57 (26)
cohort population-based cohort
Communities
Study
SBI negative 1611 62.1+/-4.5 972 (60) 722 (45) 264 (17) No data No data 82 (5) No data 277 (17)
§cohort characteristics refer to entire patient sample as data stratified by SBI status is not available
|| Di Tullio et al. presented SBI results and stroke outcomes for a subset of those subjects enrolled in the MRI substudy of the Northern Manhattan Study. The study characteristics in this table for this study were obtained after direct correspondence with study authors who were preparing a study focused on SBI and incident stroke risk
at the time of data extraction.
¶Windham et al. studied small brain lesions less than 3 mm as putative vascular lesions, in addition to lesions greater than 3 mm which were considered to be SBI. The data extracted from this study, including clinical characteristics, are focused on SBI defined>3mm versus no lesions.
Supplemental Table III: Silent Brain Infarction Definitions
Section
Study Study First Author Magnet Field Section Size
Thickness SBI MRI Signal Characteristics Means of differentiating SBI from perivascular spaces Means of detecting SBI in patients with documented prior stroke
Number and Year Strength Gap (mm) classification
(mm)
brighter lesions on spin density and T2 sequences than normal gray matter (for
0.35 and 1.5
1 Bernick 20011 5 mm 0 3 mm or greater cortical and deep grey matter); brighter at spin density and T1 hypointense (for spin density brigthness used to distinguish SBI from perivascular spaces NA
Tesla
white matter)
not low signal intensity area on T1-weighted images that was also visible as a
2 Kario 20012 1.5 Tesla 7.8 to 8.0 mm 3 to 15 mm not specified NA
specified hyperintense lesion on T2-weighted images
focal area on both T1 and T2-weighted images that was visible as low-intensity
not
3 Naganuma 20053 1.5 Tesla 10 mm >3 mm areas on T1 weighted image and as high signal intensity area on T2 weighted not specified NA
specified
images.
0.15, 0.2 and 1.5 not focal hyperintensity lesion on T2WI corresponding to a hypointensity lesion on proton density weighted or FLAIR images used to distinguish infarcts
4 Bokura 20064 10 mm, 7 mm >3 mm NA
Tesla specified T1WI from dilated perivascular spaces
focal hyperintense lesion on T2-weighted images and/or fluid-attenuated Two study investigators defined SBI on the baseline imaging performed for
inversion recovery with no corresponding symptoms in the clinical history of the acute ischemic stroke as chronic lesions with no corresponding symptoms in the
not hyperintensity of T2-FLAIR images used to distinguish SBI from dilated
10 Weber 201210 Not specified not specified ≥ 3 mm patient that could be attributed to the lesion; SBI were distinguished from clinical history of the patient that could be attributed the presumed SBI;
specified perivascular space
nonspecific subcortical and periventricular white matter lesions by the presence Information about symptoms of the qualifying ischemic stroke was collected
of a corresponding hypointense lesion on T1-weighted images using baseline case report forms
(1) CSF density on the subtraction image and (2) if the stroke was in the basal
FLAIR=3 mm;
11 Di Tullio 201311 1.5 Tesla 0 mm ≥ 3 mm ganglia area, distinct separation from the circle of Willis vessels and perivascular lesion morphology used to distinguish SBI from perivascular spaces NA
T1=1.3 mm
spaces.
MRI = magnetic resonance imaging; SBI = silent brain infarction; FLAIR = fluid attenuated inversion recovery; NA = not applicable
Supplemental Table IV: Definitions of stroke
evidence of brain infarction or embolism with neurological deficit intermittent review of medical records and telephone interviews confirmed by physician
Kario 20012 parenchymal or subarachnoid hemorrhage at brain imaging
lasting more than 24 hours caring for the patient at the time of the event
review of medical records with diagnosis made by attending physician blinded to study
Naganuma 20053 standard clinical criteria, not further specified standard clinical criteria, not further specified
purpose; family telephone interview in events that happened out of hospital
annual questionnaire sent to the patient's home and, if needed, telephone interview and
Bokura 20064 not specified not specified
interview of attending physician
an acute disturbance of focal neurological dysfunction with symptoms an acute disturbance of focal neurological dysfunction with symptoms outpatient periodic visits and, in case of missed appointment follow-up, telephone
Miwa 201312 lasting >24 hours (or resulting in earlier death) and thought to be a lasting >24 hours (or resulting in earlier death) and thought to be a result interview; original medical records were also reviewed to determine the occurrence of
result of either cerebral infarction of either cerebral hemorrhage. stroke; all possible events were audited independently by 3 physicians.
Magnitude of
Unknown
Number of All Adjusted All
Number of Unknown or or Adjusted
Subjects Strokes Ischemic All Ischemic Hemorrhagic Stroke Risk
Subjects with Hemorrhagic Unclassifed Unclassifed All Adjusted Adjusted All
Study Mean Follow- without MRI in SBI Strokes in Strokes Strokes in Strokes in Metric (HR
Study First Author and Year MRI evidence Strokes in Stroke Stroke Stroke Adjustments for which covariates All Stroke Stroke p-
Number Up (Mo) evidence of Positive SBI in no SBI no SBI no SBI or OR) in
of SBI at SBI Group Subtype in Subtype in risk 95% CI value
SBI at Test Group Group Group Group presence of
baseline SBI Group no SBI metric
baseline Group SBI at
Group
baseline
age, sex, systolic blood pressure, diastolic blood
pressure, atrial fibrillation, left ventricular hypertrophy,
1 Bernick 20011 50.4 923 2401 67 57 10 92 74 13 5 HR 1.52 1.10-2.10 0.051
fasting insulin, common carotid artery wall thickness,
myocardial infarction history
not not not specified, adjusted relative risks and 95% CI were
2 Kario 20012 42 282 303 38 not available 7 not available RR 4.63 2.04-10.5 0.003
available available calculated using Cox regression analysis
MRI = magnetic resonance imaging; SBI = silent brain infarction; CI = confidence interval; HR = hazard ratio; OR = odds ratio;
Supplemental Table VI: Results of Risk of Bias Questions to Assess the Quality of Included Studies
Bernick 20011 + + + + + + + + + - +
Kario 20012 - + - + + + - - + + +
Naganuma 20053 - + + + + + - - + + +
Bokura 20064 - - + + - + - + - + +
Debette 20105 + + + + + + + + + - +
Putaala 20116 - + + + - + + + + + +
Umemura 20117 - + - + + + - + - + -
Gioia 20128 - + + - + + + + - + +
Poels 20129 + + + + + + - + + + +
Weber 201210 - + - + + + + + + + +
Di Tullio 201311 + + - + + + + + + + +
Miwa 201312 - - - + + + - + + - -
Windham 201513 + + + + + + - + + + +
*Minimal (+) if studies controlled for 4 of the following 6 potential stroke risk factors which are potential confounders: age, sex, hypertension, diabetes mellitus, coronary
artery disease, and smoking history and hyperlipidemia by including these variables in the multivariate model or by ensuring that patients with and without SBI were
similar or matched on these variables. Relatively higher risk (-) if studies did not provide adjusted risk metric demonstrating strength of association between SBI and
incident stroke.
Supplemental Figure I. Study selection flow diagram.
1. Bernick C, Kuller L, Dulberg C, Longstreth Jr WT, Manolio T, Beauchamp N, et al. Silent mri infarcts and the risk of future stroke: The cardiovascular health study. Neurology. 2001;57:1222-1229
2. Kario K, Shimada K, Schwartz JE, Matsuo T, Hoshide S, Pickering TG. Silent and clinically overt stroke in older japanese subjects with white-coat and sustained hypertension. Journal of the American College of Cardiology. 2001;38:238-
245
3. Naganuma T, Uchida J, Tsuchida K, Takemoto Y, Tatsumi S, Sugimura K, et al. Silent cerebral infarction predicts vascular events in hemodialysis patients. Kidney international. 2005;67:2434-2439
4. Bokura H, Kobayashi S, Yamaguchi S, Iijima K, Nagai A, Toyoda G, et al. Silent brain infarction and subcortical white matter lesions increase the risk of stroke and mortality: A prospective cohort study. J Stroke Cerebrovasc Dis.
2006;15:57-63
5. Debette S, Beiser A, DeCarli C, Au R, Himali JJ, Kelly-Hayes M, et al. Association of mri markers of vascular brain injury with incident stroke, mild cognitive impairment, dementia, and mortality: The framingham offspring study. Stroke; a
journal of cerebral circulation. 2010;41:600-606
6. Putaala J, Haapaniemi E, Kurkinen M, Salonen O, Kaste M, Tatlisumak T. Silent brain infarcts, leukoaraiosis, and long-term prognosis in young ischemic stroke patients. Neurology. 2011;76:1742-1749
7. Umemura T, Kawamura T, Umegaki H, Mashita S, Kanai A, Sakakibara T, et al. Endothelial and inflammatory markers in relation to progression of ischaemic cerebral small-vessel disease and cognitive impairment: A 6-year longitudinal
study in patients with type 2 diabetes mellitus. J Neurol Neurosurg Psychiatry. 2011;82:1186-1194
8. Gioia LC, Tollard E, Dubuc V, Lanthier S, Deschaintre Y, Chagnon M, et al. Silent ischemic lesions in young adults with first stroke are associated with recurrent stroke. Neurology. 2012;79:1208-1214
9. Poels MM, Steyerberg EW, Wieberdink RG, Hofman A, Koudstaal PJ, Ikram MA, et al. Assessment of cerebral small vessel disease predicts individual stroke risk. J Neurol Neurosurg Psychiatry. 2012;83:1174-1179
10. Weber R, Weimar C, Wanke I, Möller-Hartmann C, Gizewski ER, Blatchford J, et al. Risk of recurrent stroke in patients with silent brain infarction in the prevention regimen for effectively avoiding second strokes (profess) imaging
substudy. Stroke; a journal of cerebral circulation. 2012;43:350-355
11. Di Tullio MR, Jin Z, Russo C, Elkind MS, Rundek T, Yoshita M, et al. Patent foramen ovale, subclinical cerebrovascular disease, and ischemic stroke in a population-based cohort. Journal of the American College of Cardiology.
2013;62:35-41
12. Miwa K, Tanaka M, Okazaki S, Furukado S, Sakaguchi M, Mochizuki H, et al. Association between interleukin-6 levels and first-ever cerebrovascular events in patients with vascular risk factors. Arterioscler Thromb Vasc Biol.
2013;33:400-405
13. Windham BG, Deere B, Griswold ME, Wang W, Bezerra DC, Shibata D, et al. Small brain lesions and incident stroke and mortality: A cohort study. Ann Intern Med. 2015;163:22-31