Articles Cerebral microbleeds and stroke risk after ischaemic stroke or transient ischaemic attack: a pooled analysis of individual patient data from cohort studies Duncan Wilson, Gareth Ambler, Keon-Joo Lee, Jae-Sung Lim, Masayuki Shiozawa, Masatoshi Koga, Linxin Li, Caroline Lovelock, Hugues Chabriat, Michael Hennerici, Yuen Kwun Wong, Henry Ka Fung Mak, Luis Prats-Sánchez, Alejandro Martínez-Domeño, Shigeru Inamura, Kazuhisa Yoshifuji, Ethem Murat Arsava, Solveig Horstmann, Jan Purrucker, Bonnie Yin Ka Lam, Adrian Wong, Young Dae Kim, Tae-Jin Song, Maarten Schrooten, Robin Lemmens, Sebastian Eppinger, Thomas Gattringer, Ender Uysal, Zeynep Tanriverdi, Natan M Bornstein, Einor Ben Assayag, Hen Hallevi, Jun Tanaka, Hideo Hara, Shelagh B Coutts, Lisa Hert, Alexandros Polymeris, David J Seiffge, Philippe Lyrer, Ale Algra, Jaap Kappelle, Rustam Al-Shahi Salman, Hans R Jäger, Gregory Y H Lip, Heinrich P Mattle, Leonidas D Panos, Jean-Louis Mas, Laurence Legrand, Christopher Karayiannis, Thanh Phan, Sarah Gunkel, Nicolas Christ, Jill Abrigo, Thomas Leung, Winnie Chu, Francesca Chappell, Stephen Makin, Derek Hayden, David J Williams, M Eline Kooi, Dianne H K van Dam-Nolen, Carmen Barbato, Simone Browning, Kim Wiegertjes, Anil M Tuladhar, Noortje Maaijwee, Christine Guevarra, Chathuri Yatawara, Anne-Marie Mendyk, Christine Delmaire, Sebastian Köhler, Robert van Oostenbrugge, Ying Zhou, Chao Xu, Saima Hilal, Bibek Gyanwali, Christopher Chen, Min Lou, Julie Staals, Régis Bordet, Nagaendran Kandiah, Frank-Erik de Leeuw, Robert Simister, Aad van der Lugt, Peter J Kelly, Joanna M Wardlaw, Yannie Soo, Felix Fluri, Velandai Srikanth, David Calvet, Simon Jung, Vincent I H Kwa, Stefan T Engelter, Nils Peters, Eric E Smith, Yusuke Yakushiji, Dilek Necioglu Orken, Franz Fazekas, Vincent Thijs, Ji Hoe Heo, Vincent Mok, Roland Veltkamp, Hakan Ay, Toshio Imaizumi, Beatriz Gomez-Anson, Kui Kai Lau, Eric Jouvent, Peter M Rothwell, Kazunori Toyoda, Hee-Joon Bae, Joan Marti-Fabregas, David J Werring, on behalf of the Microbleeds International Collaborative Network* Summary Background Cerebral microbleeds are a neuroimaging biomarker of stroke risk. A crucial clinical question is whether Lancet Neurol 2019; 18: 653–65 cerebral microbleeds indicate patients with recent ischaemic stroke or transient ischaemic attack in whom the rate of Published Online future intracranial haemorrhage is likely to exceed that of recurrent ischaemic stroke when treated with antithrombotic May 23, 2019 http://dx.doi.org/10.1016/ drugs. We therefore aimed to establish whether a large burden of cerebral microbleeds or particular anatomical S1474-4422(19)30197-8 patterns of cerebral microbleeds can identify ischaemic stroke or transient ischaemic attack patients at higher absolute This online publication has risk of intracranial haemorrhage than ischaemic stroke. been corrected. The corrected version first appeared at Methods We did a pooled analysis of individual patient data from cohort studies in adults with recent ischaemic thelancet.com/neurology on stroke or transient ischaemic attack. Cohorts were eligible for inclusion if they prospectively recruited adult July 12, 2019, with subsequent corrections on January 3, 2020 participants with ischaemic stroke or transient ischaemic attack; included at least 50 participants; collected data on See Comment page 619 stroke events over at least 3 months follow-up; used an appropriate MRI sequence that is sensitive to magnetic *Collaborators are listed in the susceptibility; and documented the number and anatomical distribution of cerebral microbleeds reliably using appendix consensus criteria and validated scales. Our prespecified primary outcomes were a composite of any symptomatic Stroke Research Centre, intracranial haemorrhage or ischaemic stroke, symptomatic intracranial haemorrhage, and symptomatic ischaemic Department of Brain Repair stroke. We registered this study with the PROSPERO international prospective register of systematic reviews, number and Rehabilitation, UCL Queen CRD42016036602. Square Institute of Neurology, London, UK (D Wilson PhD, D J Seiffge MD, C Barbato MD, Findings Between Jan 1, 1996, and Dec 1, 2018, we identified 344 studies. After exclusions for ineligibility or declined S Browning BSc, R Simister PhD, requests for inclusion, 20 322 patients from 38 cohorts (over 35 225 patient-years of follow-up; median 1∙34 years D J Werring PhD); National [IQR 0·19–2·44]) were included in our analyses. The adjusted hazard ratio [aHR] comparing patients with cerebral Hospital for Neurology and Neurosurgery, London UK microbleeds to those without was 1∙35 (95% CI 1∙20–1∙50) for the composite outcome of intracranial haemorrhage (D Wilson, D J Seiffge, C Barbato, and ischaemic stroke; 2∙45 (1∙82–3∙29) for intracranial haemorrhage and 1∙23 (1∙08–1∙40) for ischaemic stroke. The S Browning, R Simister); aHR increased with increasing cerebral microbleed burden for intracranial haemorrhage but this effect was less New Zealand Brain Research marked for ischaemic stroke (for five or more cerebral microbleeds, aHR 4∙55 [95% CI 3∙08–6∙72] for intracranial Institute, Christchurch, New Zealand (D Wilson); haemorrhage vs 1∙47 [1∙19–1∙80] for ischaemic stroke; for ten or more cerebral microbleeds, aHR 5∙52 [3∙36–9∙05] vs Department of Statistical 1∙43 [1∙07–1∙91]; and for ≥20 cerebral microbleeds, aHR 8∙61 [4∙69–15∙81] vs 1∙86 [1∙23–2∙82]). However, irrespective Science, University College of cerebral microbleed anatomical distribution or burden, the rate of ischaemic stroke exceeded that of intracranial London, London, UK haemorrhage (for ten or more cerebral microbleeds, 64 ischaemic strokes [95% CI 48–84] per 1000 patient-years vs (G Ambler PhD); Department of Neurology, Seoul National 27 intracranial haemorrhages [17–41] per 1000 patient-years; and for ≥20 cerebral microbleeds, 73 ischaemic strokes University Bundang Hospital, [46–108] per 1000 patient-years vs 39 intracranial haemorrhages [21–67] per 1000 patient-years). Seoul National University School of Medicine, Seongnam, Interpretation In patients with recent ischaemic stroke or transient ischaemic attack, cerebral microbleeds are South Korea (K-J Lee MD, H-J Bae MD); Department of associated with a greater relative hazard (aHR) for subsequent intracranial haemorrhage than for ischaemic stroke, Neurology, Hallym University but the absolute risk of ischaemic stroke is higher than that of intracranial haemorrhage, regardless of cerebral Sacred Heart Hospital, Anyang, microbleed presence, antomical distribution, or burden. South Korea (J-S Lim MD); www.thelancet.com/neurology Vol 18 July 2019 653 Articles Department of Cerebrovascular Funding British Heart Foundation and UK Stroke Association. Medicine, National Cerebral and Cardiovascular Center, Copyright © 2019 The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 Suita, Osaka, Japan (M Shiozawa MD, M Koga PhD, license. K Toyoda MD); Centre for Prevention of Stroke and Introduction cause ischaemic stroke, and might result from non- Dementia, University of Oxford, Oxford, UK A central challenge in stroke prevention after ischaemic haemorrhagic mechanisms.10–13 In ischae­mic stroke co­ (L Li DPhil, C Lovelock FRACP, stroke or transient ischaemic attack is to predict the risk horts, cerebral microbleeds are associ­ ated with the P M Rothwell FMedSci); of intracranial haemorrhage and to differentiate this from risks of both subsequent intracranial haem­ orrhage Assistance Publique Hôpitaux the risk of recurrent ischaemic stroke in patients treated and recurrent ischaemic stroke.14–28 As the number of de Paris, Lariboisière Hospital, Department of Neurology, with antithrombotic therapy—usually antiplate­let drugs cerebral microbleeds increases, the risk of intracranial Paris, France (H Chabriat MD, or, in patients with atrial fibrillation, oral anti­coagulants.1 haem­orrhage seems to rise more steeply than that of E Jouvent MD); Département Cerebral microbleeds are a radiological finding of small ischae­mic stroke, and having five or more cerebral micro­ Hospitalo-Universtaire (<10 mm), hypointense (black), ovoid or rounded regions bleeds has been reported to be associated with similar NeuroVasc, University Paris Diderot, and INSERM U1141, on T2*-weighted gradient-recalled echo (GRE) or suscepti­ absolute risks of intracranial haemorrhage and ischaemic Paris, France (H Chabriat, bility-weighted imaging (SWI).2 Cerebral micro­ bleeds stroke.28,29 E Jouvent); Department of mostly correspond pathologically to haemosiderin-laden Because previous studies had small sample sizes and Neurology, macrophages close to arterioles affected by small vessel few intracranial haemorrhage outcome events, they could Universitätsmedizin Mannheim, University of diseases;3,4 strictly lobar cerebral micro­ bleeds suggest not reliably answer the important clinical question of Heidelberg, Mannheim, cere­bral amyloid angiopathy (CAA), whereas deep whether many cerebral microbleeds, or patterns (distrib­ Germany (M Hennerici MD); patterns probably indicate arterioloscler­osis and mixed utions) of cerebral microbleeds, indicate a higher risk Division of Neurology, patterns probably indicate mixed pathologies.5–8 Cerebral of intra­cran­ial haemorrhage than of recurrent ischaemic Department of Medicine (Y K Wong MSc, K K Lau PhD) micro­ bleeds might result from red blood cell leakage stroke. We established the Microbleeds International and Department of Diagnostic from arterioles and capillaries, raising clinical concerns Collaborative Network30 to undertake large-scale pooled Radiology (H K F Mak MD), that they herald an increased risk of potentially devastat­ analyses of prospective observational cohort studies. We The University of Hong Kong, ing intracranial haem­ orrhage, particularly in patients tested the hypothesis that a large burden of cerebral Hong Kong; Department of Neurology, Hospital de la Santa treated with antithrombotic drugs.9 However, cerebral microbleeds, or their anatomical patterns, can identify Creu i Sant Pau, Biomedical micro­bleeds signal small vessel diseases that can also ischaemic stroke or transient ischaemic attack patients at Research Institute, Barcelona, Spain (L Prats-Sánchez MD, A Martínez-Domeño MD, Research in context J Marti-Fabregas PhD); Department of Neurosurgery, Evidence before this study Added value of this study Kushiro City General Hospital, Kushiro, Japan (S Inamura MD, We searched Medline and EMBASE from Jan 1, 1996, Our pooled analysis of individual data from 20 322 patients K Yoshifuji PhD, T Imaizumi MD); to Dec 1, 2018 (search strategy: “cerebral adj2 micro*” OR “CMB” shows that regardless of cerebral microbleed burden and Departments of Neurology and OR “microbleed.mp” AND [“stroke.mp” OR “stroke/” OR distribution (ie, mixed, deep, or lobar), or the type of Radiology, Massachusetts “intracerebral h?emorr*” OR “intracranial h?emorr*” OR antithrombotic treatment received (oral anticoagulants or General Hospital, Harvard Medical School, Boston MA, “isch?emic stroke” OR “isch?emic infarct*”]) for studies in English antiplatelet therapy), the absolute rate of ischaemic stroke is USA (E M Arsava MD, H Ay MD); that included patients with ischaemic stroke or transient consistently substantially higher than that of intracranial Department of Neurology, ischaemic attack in whom the presence and anatomical haemorrhage. By contrast with previous studies, the large Heidelberg University Hospital, distribution of cerebral microbleeds were measured at baseline, number of participants provided more precise estimates of Heidelberg, Germany (S Horstmann MD, J Purrucker with at least 90 days of follow-up. An aggregate level stroke recurrence rates and risks, while inclusion of individual MD); Therese Pei Fong Chow meta-analysis (n=5068) showed that cerebral microbleeds patient data allowed adjustment for potential confounding Research Centre for Prevention were associated with both intracranial haemorrhage factors. Our study adds new data for patients with many of Dementia, Gerald Choa (risk ratio [RR] 3·8 [95% CI 3·5–11·4]) and ischaemic stroke (eg, ≥20) cerebral microbleeds, which cause the most clinical Neuroscience Centre, Lui Che Woo Institute of Innovative (RR 1·8 [1·4–2·5]); this pooled analysis, and another study in concern regarding intracranial bleeding. Medicine, Department of two cohorts (one including 1003 mainly Chinese participants and Implications of all the available evidence Medicine and Therapeutics, the other including 1080 mainly white participants) reported that The Chinese University of Hong Although cerebral microbleeds can inform regarding the hazard five or more cerebral microbleeds were associated with similar Kong, Hong Kong for intracranial haemorrhage in patients with recent ischaemic (B Y K Lam PhD, A Wong PhD, absolute risks of intracranial haemorrhage and ischaemic stroke. stroke or transient ischaemic attack treated with antithrombotic V Mok MD); Department of However, small sample sizes and few intracranial haemorrhage drugs, the absolute risk of ischaemic stroke is much higher than Neurology, Yonsei University outcome events in previous studies did not provide enough College of Medicine, Seoul, that of intracranial haemorrhage, regardless of cerebral statistical power and precision to establish whether a large South Korea (Y D Kim MD, microbleed presence, burden, or pattern. The available evidence J H Heo MD); Department of cerebral microbleed burden or distribution pattern is associated does not support witholding antithrombotic treatment because Neurology, Ewha Womans with a higher absolute risk of intracranial haemorrhage than of cerebral microbleeds, but to definitively answer this question University College of Medicine, ischaemic stroke in patients with recent ischaemic stroke or Seoul, South Korea requires data from randomised controlled trials. transient ischaemic attack treated with antithrombotic drugs. (T J Song PhD); Center for Brain 654 www.thelancet.com/neurology Vol 18 July 2019 Articles higher absolute risk of intracranial haemorrhage than factor profiles between patients with and without cerebral and Disease Research, VIB, ischaemic stroke. microbleeds and between patients with and without Leuven, Belgium (M Schrooten MD); outcome events using the Mann-Whitney test if not Experimental Neurology and Methods normally distributed or the t test if normally distributed; Leuven Institute for Study design we compared categorical variables between groups with Neuroscience and Disease, For this pooled analysis of individual patient data, we the χ² test or Fisher’s exact test. We censored patients at the Katholieke Universiteit Leuven, University of Leuven, identified cohorts by searching Medline and EMBASE last available follow-up (truncated to 5 years) or at the time Laboratory of Neurobiology, (search terms “cerebral adj2 micro*” OR “CMB” OR of the prespecified outcome event. When a patient had Leuven, Belgium “microbleed.mp” AND “stroke.mp” OR “stroke/” OR multiple events of the same type, we censored follow-up at (R Lemmens PhD); Department “intracerebral h?emorr*” OR “intracranial h?emorr*” OR the first event. We calculated absolute event rates per 1000 of Neurology, Medical University of Graz, Graz, Austria “isch?emic stroke” OR “isch?emic infarct*”), clinical trial patient-years for primary outcomes in patients with and (S Eppinger MD, T Gattringer MD, databases (clinicaltrials.gov and strokecenter.org), and without cerebral micro­bleeds. We assessed the proportional F Fazekas MD); Department of scientific meeting abstracts. We invited members of the hazards assumption through visual inspection of (log–log) Neurology, Demiroglu Bilim METACOHORTS consortium;31 an international database plots of log cumulative hazard against time and tested for a University, Istanbul, Turkey (E Uysal MD, Z Tanriverdi MD, of more than 90 studies of small vessel disease, includ­ non-zero slope in a regression of scaled Schoenfeld D N Orken MD); Department of ing 660 000 patients. Two authors (DW and DJWe) residuals against time. We calculated univariate Kaplan- Neurology, Tel-Aviv Sourasky independently did the search and reviewed all titles and Meier survival probabilities in patients with and without Medical Center, Tel-Aviv, Israel; abstracts; they also did an independent risk of bias cerebral microbleeds to estimate event rates and used the Sackler Faculty of Medicine, Tel-Aviv University, Tel-Aviv, assessment for all included studies. Cohorts were elig­ log-rank test to compare groups. We did multivariable Cox Israel (N M Bornstein MD, ible for inclusion if they prospectively recruited adult regression adjusting for the follow­ ing prognostic and E B Assayag MD, H Hallevi MD); participants with ischaemic stroke or transient ischaemic confounding variables (selected by consensus based on Division of Neurology, attack; included at least 50 participants; collected data on availability, biological plausibility, and known associations Department of Internal Medicine, Saga University stroke events over at least 3 months follow-up; used an with cerebral microbleeds and out­comes): age, sex, pre­ Faculty of Medicine, appropriate MRI sequence that is sensitive to magnetic sentation with transient ischaemic attack or ischaemic Nabeshima, Saga, Japan susceptibility (GRE or SWI); and documented the number stroke, history of hypertension, pre­vious stroke, known (J Tanaka MD, H Hara PhD, Y Yakushiji PhD); Calgary Stroke and anatomical distribution of cerebral microbleeds atrial fibrillation, antithrombotic use after index event, and Program, Department of reliably using consensus criteria and validated scales. type of MRI sequence used to detect cerebral microbleeds Clinical Neurosciences, Each patient was only included in one cohort. We assessed (T2*-weighted GRE or SWI). We investigated the effect of Radiology and Community all studies for risk of bias (including selection bias) and predefined cerebral microbleed burden categories (one, Health Sciences, Hotchkiss Brain Institute, University of quality using the Cochrane Collaboration tool.32 All cohorts two to four, five or more, ten or more, and 20 or more). Calgary, Calgary, AB, Canada obtained ethical approval as required by local regula­tions When investigating cerebral microbleed distribution, we (S B Coutts MD, E E Smith MD); to allow data sharing. All data reviewed by the co- adjusted for number of cerebral microbleeds. We added a Department of Neurology and ordinating centre was fully anonymised. The project shared frailty term33 to account for patients being nested in Stroke Centre, University Hospital Basel and University was approved by the Health Research Authority of the individual studies (thus potentially having correlated data). of Basel, Basel, Switzerland UK (REC reference: 8/HRA/0188). The Microbleeds We performed subanalyses for patients treated with oral (L Hert MD, A Polymeris MD, International Collaborative Network protocol and statisti­ anticoagulants and antiplatelet drugs and added interaction D J Seiffge, P Lyrer MD, cal analysis plan were registered with PROSPERO on terms between antithrombotic therapy and presence of S T Engelter MD, N Peters MD); Julius Centre for Health April 5, 2016 (CRD42016036602). cerebral microbleeds. We categorised ethnicity (when Sciences and Primary Care available) as white or Asian (Japanese, Chinese, Malays, (A Algra MD) and Department Outcomes Indian, Pakistani, or Korean) to investigate the interaction of Neurology and Our prespecified primary outcomes were a composite of be­tween ethnicity and cerebral microbleed presence. We Neurosurgery, Utrecht Stroke Centre (A Algra, J Kappelle MD), any symptomatic intracranial haemorrhage (confirmed performed two prespecified sensitivity analyses: the first University Medical Center radiologically, including subdural, extradural, and sub­ exploring time-varying risks within the Cox model to Utrecht and Utrecht University, arachnoid haemorrhage, and excluding intracranial haem­ investigate later events (beyond the first year) accounting Utrecht, Netherlands; Centre orrhages attributed to intravenous thrombolysis or trauma) for death as a competing risk (using the Fine-Gray sub­ for Clinical Brain Sciences (R Al-Shahi Salman PhD), or ischaemic stroke (acute or subacute neuro­logical symp­ distribution hazard model), calculating subdistribution Edinburgh Imaging toms lasting >24 h and attributed to cerebral ischaemia, hazard ratios (sHRs); and the second, a two-stage indi­ (F Chappell PhD, J M Wardlaw diagnosed clinically, with or without radio­ logical con­ vidual-patient meta-analysis to quantify between-study MD), and UK Dementia firmation); symptomatic intracranial haem­ orrhage; and heterogeneity using the inverse-variance method (which Institute at the University of Edinburgh (F Chappell, symptomatic ischaemic stroke. Secondary outcome events fits a separate survival model for each cohort then pools J M Wardlaw), School of Clinical were death (all cause) and vascular death. All events were and displays estimates in a forest plot). We did three post- Sciences, University of adjudicated according to individual cohort protocols. hoc analyses as follows: (1) we added white matter Edinburgh, Edinburgh, UK; hyperintensities (another common marker of cerebral Lysholm Department of Neuroradiology and the Statistical analysis small vessel disease, rated using the Fazekas scale34 and Neuroradiological Academic As per our prespecified protocol, a single dataset was considered severe if rated two or greater in the peri­ Unit, Department of Brain created by combining individual participant data from the ventricular of deep white matter) into our multivariable Repair and Rehabilitation, UCL 38 cohorts. We compared baseline demographic and risk model; (2) we included only intracerebral haemor­rhage, Institute of Neurology and the www.thelancet.com/neurology Vol 18 July 2019 655 Articles National Hospital for convexity subarachnoid haem­orrhage, and subdural haem­ The incidence of all composite events in patients with Neurology and Neurosurgery, or­rhage, because these bleeding events are the most likely any cerebral microbleed was 59 per 1000 patient-years London, UK (H R Jäger FRCR); to be associated with cerebral microbleeds; and (3) we (95% CI 54–64) compared with 35 per 1000 patient- Liverpool Centre for Cardiovascular Science, investigated the interaction between cerebral micro­bleeds years (33–38) in those without cerebral microbleeds, an University of Liverpool and and age (<80 years or ≥80 years). In sensitivity analyses, if absolute increased incidence of 24 per 1000 patient-years Liverpool Heart and Chest data for a variable of interest was not sufficiently available (21–26; table 2). The aHR for a composite event be­ Hospital, Liverpool, UK (G Y H Lip MD); Aalborg in a cohort, the cohort was excluded. We did all statistical came larger with increased cere­bral microbleed burden Thrombosis Research Unit, analysis using STATA, version 15. (figure 2, table 2; ptrend<0·0001). aHRs were similar across Department of Clinical different cerebral microbleed anatomical distributions Medicine, Aalborg University, Role of the funding source (table 2). Aalborg, Denmark (G Y H Lip); Department of Diagnostic and The funder of the study had no role in the study design, 189 patients had a symptomatic intracranial haemorrhage Interventional Neuroradiology data collection, data analysis, or data interpretation, or over 32 847 patient-years of follow-up (151 intracerebral and Department of Neurology writing of the report. The corresponding author had haemorrhages, 31 subdural haemorrhages, eight sub­ Inselspital, University Hospital full access to all the data in the study and had final arachnoid haemorrhages [ four of which were cortical], and Bern, University of Bern, Bern, Switzerland (D J Seiffge, responsibility for the decision to submit for publication. three extradural haemorrhages; four patients had more H P Mattle MD, L D Panos MD, than one type of intracranial haemorrhage). The incidence S Jung MD); Department of Results of intracranial haemorrhage was 12 per 1000 patient-years Neurology (J L Mas MD, Between Jan 1, 1996, and Dec 1, 2018, we identified and (95% CI 10–14) in those with cerebral microbleeds D Calvet MD) and Department of Neuroradiology screened 344 records (325 from database search and compared with 4 per 1000 patient-years (3–5) in those (L Legrand MD), Sainte-Anne 19 from other sources; figure 1). 263 records were excluded without cerebral microbleeds, an absolute increased Hospital, Paris Descartes because they were not full-text articles, and then a further incidence of 8 per 1000 patient-years (7–9; table 2). The rate University, INSERM U1266, 29 full-text articles were excluded because they did not of intracranial haemorrhage increased with increasing Paris, France; Peninsula Clinical School, Peninsula Health meet study inclusion criteria. The remaining 52 studies cerebral microbleed burden, but was consistently lower (C Karayiannis MD, were included in our qualitative analyses, but 14 of these than the rate of ischaemic stroke (table 2). The aHR for V Srikanth PhD) and Stroke and were excluded from the meta-analysis because they did not symptomatic intracranial haemorrhage was 2∙45 (95% CI Ageing Research Group, School respond to requests for individual patient data or declined 1∙82–3∙29) for patients with cerebral microbleeds versus of Clinical Sciences at Monash Health (T Phan MD), Monash to join the collaboration (reasons included a lack of those without, and became larger with increased cerebral University, Melbourne, VIC, resources or because of data sharing policies). From the microbleed burden (ptrend <0·0001; figure 2; table 2); Australia; Department of 38 remaining cohorts (23 published and 15 unpublished aHRs did not significantly differ between different cere­ Neurology, University Hospital studies), we included 20 322 participants (table 1). Although bral microbleed anatomical distributions. Patients with of Würzburg, Josef-Schneider Strasse 11, Würzburg, Germany more than half of participants and outcome events came multiple strictly lobar cerebral microbleeds (fulfilling (S Gunkel MD, N Christ MD, from the six largest cohorts, no major risk of bias was the Boston criteria5 for probable CAA) did not have a F Fluri MD); Department of detected for any included cohort (appendix). The mean age significantly higher aHR for symptomatic intracranial Imaging and Interventional of participants was 70 years (SD 13); 8593 (42%) of the haemorrhage than those without multiple strictly lobar Radiology (J Abrigo MD, W Chu MD) and Department of 20 322 were women. Cerebral micro­bleeds were present in cerebral microbleeds (1∙29 [95% CI 0∙60–2∙77]; table 2). Medicine and Therapeutics 5649 (28%) patients (appendix), including 2415 (12%) No interaction was detected be­tween cerebral microbleeds (T Leung MD, Y Soo MD), Prince with one cerebral microbleed, 1990 (10%) with two to and antiplatelet medi­cation (pinteraction=0∙358), oral anticoag­ of Wales Hospital, The Chinese four cerebral microbleeds, and 1244 (6%) with five or ulants (pinteraction=0∙717), or combined oral anticoagulants University of Hong Kong, Ma Liu Shui, Hong Kong; more cerebral microbleeds. Over the 35 225 patient- and anti­platelet medication (pinteraction=0∙163) for intracranial Institute of Cardiovascular and years of follow-up (median 1·34 years [IQR 0·19–2·44]), haemorrhage risk. Medical Science, University of 1474 composite events occurred: 189 intracranial haem­ 1113 patients had a symptomatic ischaemic stroke over Glasgow, Glasgow, UK (S Makin PhD); orrhages; 1113 ischae­mic strokes; and 172 compo­site events 32 293 patient-years of follow-up. The incidence of The Neurovascular Research of unknown type from one cohort of 3355 participants, symptomatic ischaemic stroke in patients with cerebral Unit and Health Research which did not subclassify composite out­comes as intra­ microbleeds was 46 per 1000 patient-years (95% CI 42–51) Board, Stroke Clinical Trials cranial haemorrhage or ischaemic stroke. Char­acteristics compared with 30 per 1000 patient-years (28–33) in Network Ireland, University College Dublin, Dublin, Ireland between patients with and without events are in the those without, with an absolute increased incidence of (D Hayden MD, P J Kelly MD); appendix. Visual assessment of the log-log plots and the 16 per 1000 patient-years (14–18; table 2). The rate of Department of Geriatric and results of testing the Schoenfeld residuals suggest that the ischaemic stroke became greater with an increasing Stroke Medicine, Royal College proportional hazards assump­tion was not violated in any burden of cerebral microbleeds, and for each burden of Surgeons in Ireland, Dublin, Ireland and Beaumont Hospital of the following analyses. category substantially exceeded the rate of intracranial Dublin, Ireland The composite outcome of any intracranial haemorrhage haemorrhage (table 2). The aHR for symptomatic (D J Williams PhD); Department or ischaemic stroke (aHR 1·35 [95% CI 1·20–1·50], ischaemic stroke was 1∙23 (95% CI 1∙08–1∙40) for patients of Radiology and Nuclear p<0∙0001; log-rank test), symptomatic intracranial haem­ with cerebral microbleeds versus those with­out, and the Medicine (M E Kooi PhD) and Department of Neurology, orrhage (2·45 [1·82–3·29], p<0∙0001), and symptomatic aHR became larger with in­creasing cere­bral micro­bleed CARIM School for ischaemic stroke (1·23 [1·08–1·40], p<0∙0001) were more burden (ptrend=0·0053; figure 2; table 2). Cerebral micro­ Cardiovascular Diseases frequent in patients with cerebral microbleeds than those bleed anotomical distribution had little effect on ischae­ (R van Oostenbrugge MD, without (figure 2; appendix). mic stroke risk (table 2). No interaction was detected 656 www.thelancet.com/neurology Vol 18 July 2019 Articles between cerebral microbleeds and antiplatelet medication (pinteraction=0·943) or oral anti­coagulants (pinteraction=0·408) for 325 records identified through database searching ischaemic stroke risk, but there was weak evidence for an 19 additional records idenitified through interaction between cerebral microbleeds and combined other sources 10 from METACOHORTS use of oral anticoagulants and antiplatelet medication 8 studies idenitified through emails to (pinteraction=0·047). centres with previous cerebral There were 2148 deaths, 484 of which were due microbleed publications 1 study identified through conferences to vascular causes. In multivariable analyses, cerebral microbleed presence was not associated with all-cause death (aHR 1∙03 [95% CI 0∙94–1∙12]) or vascular death 344 records screened (aHR 0∙97 [0∙79–1∙19]). No interaction was detected between cerebral microbleeds and ethnicity (n=15 123; 263 records excluded because not appropriate to study aims 6743 white and 8380 Asian) for the risks of the composite 81 full-text articles assessed for eligibility outcome of intracranial haem­orrhage or ischaemic stroke (pinteraction=0∙707); intracran­ial haemorrhage (pinteraction=0∙537); 29 full-text articles excluded or ischaemic stroke (pinteraction=0∙654). No interaction 3 population-based study 5 ineligible patient populations was detected between cerebral microbleed and older 6 cross-sectional studies age (4376 patients older than 80 years) for the risk 6 reviews or letters of the composite outcome (pinteraction=0∙538); intracranial 3 retrospective studies 2 case reports haemorrhage (pinteraction=0∙219); or ischaemic stroke 4 outcome measures not appropriate (pinteraction=0∙286). Using a two-stage meta-analysis, the estimated risks 52 studies included in qualitative synthesis associated with cerebral microbleed presence were con­ 14 studies excluded because they did not respond or did sistent with our main model for the composite outcome not have resources to join collaboration (heterogeneity [I²=31·7%]; intracranial haemorrhage 38 studies included in pooled analysis [I²=0%]; and ischaemic stroke [²=24·2%]; appendix). 23 cohorts, including 10 235 patients, provided ratings Figure 1: Study selection profile for white matter hyperintensities, which were moderate to severe (Fazekas grade ≥2) in 3105 (30%) patients. Including white matter hyperintensities in multivariable 696 ischaemic strokes were recorded over 12 873 patient- J Staals MD), Maastricht models did not substantially change the aHR associated years of follow-up within the first year and 417 symptomatic University Medical Centre, Maastricht, Netherlands; with the presence of cerebral microbleeds for the com­ ischaemic strokes during 30 447 patient-years of follow-up Comprehensive Stroke Service, posite outcome (aHR 1.30 [95% CI 1.12–1.52]); intracranial after the first year. In patients with cerebral microbleeds, University College London haemorrhage (aHR 2.44 [1.68–3.53]); or for ischaemic the rate of symptomatic ischaemic stroke within the first Hospitals NHS Trust, London, stroke (aHR 1.16 [0.98–1.37]). year was 70 (95% CI 62–80), then 18 (15–21) after the UK (C Barbato, S Browning, R Simister); Department of In our sensitivity analysis including only intracerebral, first year. Neurology, Donders Institute convexity subarachnoid, and subdural intracranial haem­ Accounting for death as a competing risk, we found no for Brain, Cognition and orrhages, 183 patients had a sympto­ matic intracranial evidence for a change in risk over time associated with Behaviour, Donders Centre for haemorrhage over 32 847 patient-years of follow-up. The cerebral microbleed presence for intracranial haemor­ Medical Neuroscience, Radboud University Medical aHR for symptomatic intracranial haemorrhage was rhage (sHR 4∙96 [95% CI 3∙18–7∙74] at day 0 vs 4∙81 Center, Nijmegen, Netherlands 2∙59 (95% CI 1∙91–3∙50) for patients with cerebral [3∙15–7∙35] after 1 year) or ischaemic stroke (sHR 1∙46 (K Wiegertjes MD, microbleeds versus patients without, and became larger [1∙23–1∙73] at day 0 vs 1∙49 [1∙27–1∙75] after 1 year). A M Tuladhar MD, with increasing burden. Compared with no cerebral micro­ In those treated with oral anticoagulants after their F-E de Leeuw MD); Lucerne State Hospital; Switzerland bleeds, aHRs were 1∙92 (95% CI 1∙25–2∙94) for one cere­ index ischaemic stroke or transient ischaemic attack Center for Neurology and bral micro­bleed; 2·02 (1∙30–3∙16) for two to four cerebral (n=7737; vitamin K antagonist=5253, non-vitamin K Neurorehabilitation, Luzern, microbleeds; 4∙88 (3∙29–7∙25) for five or more cere­ oral anticoagulant=2484), 91 intracranial haemorrhages Switzerland (N Maaijwee PhD); bral microbleeds; 5∙87 (3∙56–9∙66) for ten or more occurred over 13 942 patient-years of follow-up, and Department of Neurology, National Neuroscience cerebral micro­bleeds; and 9∙32 (5∙06–17∙16) for 20 or more 384 ischaemic strokes occurred over 13 737 patient- Institute, Singapore, Singapore cerebral micro­bleeds. These results are consistent with our years of follow-up. For patients with cerebral micro­ (C Guevarra MD, C Yatawara PhD, primary findings. bleeds, the rate of intracranial haemorrhage was N Kandiah FRCP); University of There were 102 symptomatic intracranial haemorrhages 12 per 1000 patient-years (95% CI 9–16); the rate of Lille, Inserm, CHU de Lille, Degenerative and vascular over 12 794 patient-years of follow-up within the first year, ischaemic stroke was 32 per 1000 patient-years (26–39; cognitive disorders U1171, and 87 over 31 059 patient-years of follow-up after the first table 3). The rate of ischaemic stroke was much higher Lille, France (A-M Mendyk MD, year. In patients with cerebral microbleeds, the rate of than that of intra­ cranial haemorrhage for all cere­ C Delmaire MD, R Bordet MD); Department of Psychiatry and intracranial haemorrhage was 18 per 1000 patient-years bral microbleed burden and anatomical distribution Neuropsychology, School for (95% CI 14–23) within the first year, and 5 per 1000 patient- categories; the aHR for intra­ cranial haemorrhage for Mental Health and years (3–6) after the first year. patients with cerebral micro­ bleeds (vs those without) Neuroscience, Maastricht www.thelancet.com/neurology Vol 18 July 2019 657 658 Articles M Lou PhD); Memory Aging and See Online for appendix php?RecordID=36602 display_record. crd.york.ac.uk/PROSPERO/ analysis plan see https://www. For the protocol and statistical d.werring@ucl.ac.uk UK Neurology, London WC1B 5EH, UCL Queen Square Institute of Brain Repair and Rehabilitation, Research Centre, Department of Prof David J Werring, Stroke Correspondence to: A van der Lugt MD) (D H K van Dam-Nolen MD, Netherlands Medical Centre, Rotterdam, Medical Centre, University Nuclear Medicine, Erasmus Department of Radiology and (B Gomez-Anson PhD); and Barcelona, Spain Universitat Autonoma, Santa Creu i Sant Pau, Neuroradiology, Hospital (R Veltkamp); Unit of Heidelberg, Germany University Hospital, of Neurology, Heidelberg (R Veltkamp MD); Department London, London, UK Medicine, Imperial College (V Thijs); Department of Stroke Hospitals Leuven, Belgium Neurosciences, University Australia; Department of Austin Health, Melbourne, VIC, Department of Neurology, Australia (V Thijs MD); Melbourne, Melbourne, VIC, Health, University of Neuroscience and Mental Florey Institute of N Peters); Stroke Division, Switzerland (S T Engelter, University of Basel, Basel, Medicine Felix Platter, Department of Geriatric Neurorehabilitation, University Neurology and Netherlands (V I H Kwa MD); Gasthuis, Amsterdam, Neurology, Onze Lieve Vrouwe C Chen FRCP); Department of (S Hilal PhD, B Gyanwali MD, Singapore, Singapore University Health System, Cognition Centre, National (Y Zhou PhD, C Xu MD, Medicine, Hangzhou, China Zhejiang University, School of The 2nd Affiliated Hospital of Department of Neurology, Netherlands (S Köhler PhD); Total Taking oral Transient Mean University, Maastricht, Proportion Hyper- Atrial Previous Ischaemic Any Susceptibility- Median follow- Patients Participants Participants partici­- anticoagu­- ischaemic age women tension fibrillation stroke heart cerebral weighted up, days (IQR) with with with pants lants attack (SD), disease microbleed imaging composite intracranial ischaemic years events haemorrhage stroke events events CROMIS-227 1490 1436 (96%) 238 (16%) 76 (10) 631 (42%) 930 (63%) 1490 148 243 (16%) 311 (21%) 0 774 (705–974) 70 (5%) 14 (1%) 56 (4%) (100%) (10%) HBS 660 114 (17%) 60 (9%) 69 (15) 289 (44%) 498 (75%) 194 (29%) 156 138 (21%) 98 (15%) 2 (<1%) 90 (90–90) 4 (1%) 0 4 (1%) (24%) IPAAC-Warfarin35 182 173 (95%) 27 (15%) 73 (9) 84 (46%) 158 (87%) 179 (98%) 36 (20%) 17 (9%) 68 (37%) 182 (100%) 738 (191–812) 7 (4%) 1 (1%) 6 (3%) Bern36 392 74 (19%) 0 68 (14) 169 (43%) 249 (64%) 142 (46%) 59 (15%) 69 (18%) 90 (23%) 392 (100%) 93 (5–106) 16 (4%) 0 16 (4%) CU-STRIDE37 536 24 (4%) 81 (15%) 67 (11) 227 (42%) 373 (70%) 38 (7%) 80 (15%) 35 (7%) 124 (23%) 238 (44%) 524 (472–557) 17 (3%) 2 (<1%) 15 (3%) TABASCO38 436 33 (8%) 213 (28%) 67 (9) 184 (43%) 250 (59%) 33 (8%) 0 60 (14%) 64 (15%) 0 1825 57 (13%) 0 57 (13%) (1164–1825) Graz 460 78 (17%) 48 (10%) 67 (13) 179 (39%) 359 (78%) 115 102 95 (21%) 88 (19%) 0 117 (87–973) 65 (14%) 13 (3%) 54 (12%) (25%) (22%) PERFORM-MRI39 1056 0 127 (12%) 68 (8) 370 (35%) 887 (84%) 16 (2%) 120 69 (7%) 381 (36%) 0 774 (701–1042) 104 (10%) 10 (1%) 94 (9%) (11%) PARISK40 228 0 127 (56%) 71 (9) 67 (29%) 156 (68%) 0 66 50 (22%) 61 (27%) 0 786 (757–819) 10 (4%) 0 10 (4%) (29%) SAMURAI 1103 1039 (94%) 45 (4%) 78 (10) 480 (44%) 1027 (93%) 1103 246 101 (9%) 265 (24%) 817 (74%) 723 (758–818) 82 (7%) 10 (1%) 72 (7%) NVAF41 (100%) (22%) RUNDMC42 179 19 (11%) 89 (50%) 65 (9) 63 (35%) 145 (81%) 18 (10%) 47 (26%) 31 (17%) 35 (20%) 0 1825 25 (14%) 2 (1%) 23 (13%) (1825–1825) Wuerzburg 358 122 (34%) 77 (22%) 71 (13) 158 (44%) 287 (80%) 105 (29%) 91 (25%) 38 (11%) 87 (24%) 160 (45%) 95 (89–103) 22 (6%) 1 (<1%) 21 (6%) Monash Stroke43 359 356 (99%) 52 (15%) 75 (11) 173 (48%) 285 (79%) 359 101 122 (34%) 154 (43%) 339 (94%) 530 (280–898) 14 (4%) 7 (2%) 9 (3%) (100%) (28%) Basel TIA18 192 33 (17%) 192 69 (13) 73 (38%) 137 (71%) 26 (14%) 13 (7%) 38 (20%) 21 (11%) 0 90 (90–90) 26 (14%) 0 26 (14%) (100%) Yonsei44 504 487 (97%) 28 (6%) 70 (11) 288 (57%) 392 504 101 109 (22%) 155 (31%) 0 849 (393–1398) 56 (11%) 7 (1%) 49 (10%) (78%) (100%) (20%) SNUBH Stroke 3355 625 (19%) 368 (11%) 67 (13) 1347 (40%) 2324 630 (19%) 487 284 (8%) 1166 (35%) 1 (<1%) 355 (340–365) 172 (5%) NA NA Cohort45,46 (69%) (15%) BIOSTROKE/ 260 73 (28%) 160 (62%) 68 (13) 95 (37%) 150 (59%) 77 (31%) 21 (8%) 57 (22%) 24 (9%) 0 90 (90–365) 14 (5%) 0 14 (5%) TIA47 Kushiro City48 784 63 (8%) 0 72 (11) 330 (42%) 498 (64%) 104 (13%) 142 89 (11%) 320 (41%) 0 1008 (105–1825) 139 (18%) 22 (3%) 119 (15%) (18%) Soo49 81 81 (100%) 16 (20%) 72 (9) 40 (49%) 56 (69%) 81 (100%) 25 (31%) 8 (10%) 24 (30%) 71 (88%) 737 (641–794) 8 (10%) 3 (4%) 5 (6%) CASPER50 135 18 (13%) 0 66 (11) 39 (29%) 96 (71%) 13 (10%) 10 (7%) 29 (21%) 79 (59%) 135 (100%) 453 (444–465) 3 (2%) 0 3 (2%) HERO51 937 933 (>99%) 122 (13%) 78 (7) 488 (52%) 694 (74%) 468 (50%)* 248 146 (16%) 248 (26%) 0 737 (641–794) 49 (5%) 18 (2%) 32 (3%) (27%) (Table 1 continues on next page) www.thelancet.com/neurology Vol 18 July 2019 Total Taking oral Transient Mean Proportion Hyper- Atrial Previous Ischaemic Any Susceptibility- Median follow- Patients Participants Participants partici- anticoagu- ischaemic age women tension fibrillation stroke heart cerebral weighted up, days (IQR) with with with pants lants attack (SD), disease microbleed imaging composite intracranial ischaemic years events haemorrhage stroke events events (Continued from previous page) HAGAKURE 426 157 (37%) 35 (8%) 74 (13) 174 (41%) 320 (76%) 135 (32%) 76 (18%) 45 (11%) 158 (37%) 39 (9%) 748 (350–1040) 34 (8%) 9 (2%) 25 (6%) Leuven14 487 133 (27%) 133 (27%) 72 (9) 192 (39%) 313 (64%) 103 (21%) 61 (13%) 112 (23%) 129 (26%) 0 804 (686–968) 36 (7%) 4 (1%) 32 (7%) NOACISP 306 286 (93%) 30 (10%) 73 (19) 139 (45%) 240 (79%) 306 60 83 (27%) 87 (28%) 300 (98%) 735 (417–836) 28 (9%) 10 (3%) 19 (6%) www.thelancet.com/neurology Vol 18 July 2019 (100%) (20%) Min Lou52 126 14 (11%) 0 65 (13) 46 (37%) 94 (75%) 25 (20%) 18 (14%) 4 (3%) 42 (33%) 126 (100%) 92 (87–218) 2 (2%) 0 2 (2%) MICRO21 397 40 (10%) 362 (91%) 65 (12) 165 (42%) 218 (55%) 30 (8%) 35 (9%) 24 (6%) 72 (18%) 0 1212 (579–1825) 30 (8%) 11 (3%) 21 (5%) Orken53 454 454 (100%) 20 (4%) 72 (12) 233 (51%) 258 (79%) 296 (65%) 123 79 (32%)† 134 (30%) 250 (55%) 575 (228–1825) 11 (2%) 3 (1%) 8 (2%) (27%) CATCH54 416 67 (16%) 173 (42%) 67 (14) 164 (39%) 226 (54%) 27 (6%) 0 NA 65 (16%) 0 88 (80–100) 14 (3%) 1 (<1%) 13 (3%) MSS255 263 24 (9%) 0 67 (12) 109 (41%) 190 (72%) 25 (10%) 32 (12%) 53 (20%) 44 (17%) 251 (95%) 368 (253–403) 31 (12%) 0 31 (12%) Sainte-Anne 385 302 (78%) 0 80 (11) 204 (53%) 277 (72%) 358 61 (16%) 72 (19%) 99 (26%) 0 440 (163–733) 25 (6%) 5 (1%) 23 (6%) (Paris) (100%) STROKDEM 181 48 (27%) 0 64 (13) 69 (38%) 100 (55%) 12 (7%) 20 (11%) 17 (9%) 24 (13%) 0 1150 (420–1820) 17 (9%) 0 17 (9%) Singapore 45 15 (33%) 0 67 (10) 13 (29%) 34 (76%) 11 (24%) 3 (7%) 4 (9%) 25 (56%) 45 (100%) 1057 (703–1199) 6 (13%) 0 6 (13%) (Chen) FUTURE Study 19 0 7 (37%) 44 (6) 10 (53%) 8 (42%) 0 0 0 1 (5%) 19 (100%) 164 (131–242) 4 (21%) 0 4 (21%) Heidelberg56 650 119 (18%) 109 (17%) 64 (14) 240 (37%) 496 (76%) 115 (18%) 107 NA 155 (24%) 650 (100%) 1534 (1271–1825) 34 (5%) 4 (1%) 30 (5%) (17%) NNI 184 32 (17%) 0 58 (11) 57 (31%) 143 (78%) 28 (15%) 27 (15%) NA 50 (27%) 0 251 (86–477) 0 0 0 OXVASC29 1080 118 (11%) 572 (52%) 68 (14) 514 (48%) 588 (55%) 167 (15%) 201 146 (13%) 157 (15%) 0 1271 (681–1825) 90 (8%) 11 (1%) 79 (7%) (19%) HKU29 1003 104 (10%) 0 69 (12) 402 (40%) 657 (66%) 130 (13%) 116 92 (9%) 450 (45%) 1003 (100%) 1005 (599–1549) 112 (11%) 20 (2%) 92 (9%) (12%) SIGNaL 213 43 (20%) 22 (10%) 72 (14) 88 (41%) 152 (71%) 67 (32%) 60 49 (23%) 94 (44%) 144 (68%) 225 (202–249) 27 (13%) 1 (<1%) 26 (12%) (28%) Total 20 322 7737/20 319 3443/ 70 (13) 8593/ 14 365/ 7557/ 3299/ 2608/ 5649/ 5164/20 284 534 (243–928) 1461/ 189/16 967 1113/16 967 (38%) 20 311 20 314 20 271 20 207 20 290 18 842 20 322 (25%) 20 322 (7%) (1%) (7%) (17%) (42%) (71%) (37%) (16%) (14%) (28%) Data are n (%) or n/N (%) unless otherwise stated. Studies without references are unpublished. FUTURE study=Follow-Up of Transient ischemic attack and stroke patients and Unelucidated Risk factor Evaluation study. HAGAKURE=Hypertension, Amyloid, and aGe Associated Kaleidoscopic brain lesions on CT/MRI Undertaken with stroke REgistry. HBS=Heart Brain Interactions Study. NNI=National Neuroscience Institute, Singapore. NOACISP=Novel Oral Anticoagulants in Stroke Patients, Basel; NCT02353585. SIGNaL=Stroke Investigation in North and Central London. STROKDEM=Study of Factors Influencing Post-stroke Dementia. *Denominator for this result is 932. †Denominator for this result is 250. Table 1: Demographics, risk factors, and outcome events for each cohort Articles 659 Articles 100 Composite outcome 100 Intracranial haemorrhage 100 Ischaemic stroke Cerebral microbleeds absent aHR 2·45 (95% Cl 1·82–3·29) Cerebral microbleeds present 30 30 30 Cumulative risk (%) 20 20 20 10 10 10 aHR 1·35 (95% Cl 1·20–1·50) aHR 1·23 (95% Cl 1·08–1·40) 0 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 Number at risk (number censored) Absent 14 436 8644 5758 2787 1852 1228 12 232 8228 5888 2877 1921 1278 12248 8130 5762 2799 1863 1236 (0) (5208) (2669) (2898) (899) (586) (0) (3964) (2318) (2995) (949) (637) (0) (3663) (2212) (2905) (906) (593) Present 5587 3222 1974 864 497 310 4417 2958 2013 900 529 333 4422 2927 1972 866 499 314 (0) (2001) (1156) (1067) (345) (166) (0) (1397) (928) (1106) (365) (190) (0) (1254) (883) (1070) (351) (170) 100 One: aHR 1·35 (95% Cl 1·20–1·50) 100 One: aHR 1·87 (95% Cl 1·23–2·84) 100 One: aHR 1·14 (95% Cl 0·94–1·37) Two to four aHR 1·25 (95% Cl 1·06–1·47) Two to four aHR 1·89 (95% Cl 1·22–2·93) Two to four aHR 1·17 (95% Cl 0·97–1·42) Five or more: aHR 1·74 (95% Cl 1·46–2·06) Five or more: aHR 4·55 (95% Cl 3·08–6·72) Five or more: aHR 1·47 (95% Cl 1·19–1·80) 30 Absent 30 30 Cumulative risk (%) One Two to four 20 Five or more 20 20 10 10 10 0 0 0 0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5 Number at risk Time from enrolment (years) Time from enrolment (years) Time from enrolment (years) (number censored) Absent 14 436 8644 5758 2787 1852 1228 12 232 8228 5888 2877 1921 1278 12 248 8130 5762 2799 1863 1236 (0) (5208) (2669) (2898) (899) (586) (0) (3964) (2318) (2995) (949) (637) (0) (3663) (2212) (2905) (906) (593) One 2385 1409 894 398 234 165 1860 1291 909 413 247 174 1861 1281 892 399 235 168 (0) (843) (486) (482) (156) (63) (0) (548) (379) (495) (165) (69) (0) (493) (364) (480) (157) (65) Two to four 1973 1113 675 292 162 85 1551 1014 687 305 173 96 1552 1002 675 293 162 85 (0) (745) (402) (369) (121) (67) (0) (520) (322) (381) (129) (76) (0) (474) (298) (369) (125) (68) Five or more 1229 700 405 174 101 60 1006 653 417 182 109 63 1009 644 405 174 102 61 (0) (413) (268) (216) (68) (36) (0) (329) (227) (230) (71) (45) (0) (287) (221) (221) (69) (37) Figure 2: Kaplan-Meier estimates for the primary outcomes in all patients (n=20 322) rose more steeply than that of ischaemic stroke with aHRs for intracranial haemorrhage and ischaemic stroke increasing cerebral microbleed burden. Mixed and deep in patients with versus without cerebral microbleeds were cerebral microbleed distributions had similar aHRs similar to those in the full cohort, with little variation for intracranial haem­ orrhage, but patients with lobar according to cerebral microbleed anatomical distribution cerebral microbleeds had a lower risk of intracranial (appendix). haemorrhage (table 3). Cerebral micro­bleeds were not Compared with patients who received antithrombotic significantly associated with ischaemic stroke risk. We treatment (oral anticoagulants or antiplatelets), those not found no evidence of an interaction between oral treated with antithrombotic drugs (n=1065) were older anticoagulants type (vita­min K antagonist vs direct oral (mean age 72 years [SD 14] for those not treated with anticoagulant) and cere­ bral micro­ bleed presence for antithrombotic drugs vs 70 years [SD 13] for those treated intracranial haemorrhage (pinteraction=0∙4) or ischaemic with antithrombotic drugs), a greater proportion were stroke (pinteraction=0∙61). women (46% vs 42%), more had ischaemic stroke In patients treated with antiplatelet drugs only (91% vs 83%), more had a previous intracranial haemor­ (n=11 520), 93 intracranial haemorrhages occurred over rhage (6% vs 2%), more had atrial fibrillation (44% vs 37%), 18 059 patient-years of follow-up and 664 ischaemic fewer had been taking regular antiplatelet drugs before strokes occurred over 17 731 patient-years of follow-up. the qualifying event (27% vs 34%), and more had been The rate of ischaemic stroke remained higher than that taking regular oral anticoagulants before the qualifying of intracranial haemorrhage for all cere­bral microbleed event (13% vs 8%). No difference in the prevalence of burden and anatomical distribution categories (appendix); cerebral microbleeds was observed based on receiving 660 www.thelancet.com/neurology Vol 18 July 2019 Articles Composite of intracranial haemorrhage Symptomatic intracranial haemorrhage Symptomatic ischaemic stroke and ischaemic stroke (n=16 447 for multivariable model) (n=16 464 for multivariable model) (n=19 816 for multivariable model) Rate, per 1000 Absolute rate Adjusted hazard Rate, Absolute rate Adjusted hazard Rate, per 1000 Absolute rate Adjusted hazard patient-years* increase, ratio per 1000 increase, ratio patient-years increase, ratio per 1000 patient-years per 1000 per 1000 patient-years patient-years patient-years None 35 (33–38) ·· 1 (ref) 4 (3–5) ·· 1 (ref) 30 (28–33) ·· 1 (ref) Any 59 (54–64) 24 (21–26) 1∙35 (1∙20–1∙50) 12 (10–14) 8 (7–9) 2∙45 (1∙82–3∙29) 46 (42–51) 16 (14–18) 1∙23 (1∙08–1∙40) One 46 (40–53) 11 (7–15) 1∙21 (1∙03–1∙42) 8 (5–12) 4 (2–7) 1∙87 (1∙23–2∙84) 37 (31–44) 7 (3–11) 1∙14 (0∙94–1∙37) Number Two to four 58 (50–67) 23 (17–29) 1∙25 (1∙06–1∙47) 9 (6–14) 5 (3–9) 1∙89 (1∙22–2∙93) 48 (40–56) 18 (12–23) 1∙17 (0∙97–1∙42) Five or more† 85 (73–99) 50 (40–61) 1∙74 (1∙46–2∙06) 23 (16–31) 19 (13–26) 4∙55 (3∙08–6∙72) 64 (53–77) 34 (25–43) 1∙47 (1∙19–1∙80) Ten or more† 91 (73–113) 56 (40–75) 1∙82 (1∙44–2∙29) 27 (17–41) 23 (14–36) 5∙52 (3∙36–9∙05) 64 (48–84) 34 (20–51) 1∙43 (1∙07–1∙91) 20 or more† 118 (86–160) 83 (53–122) 2∙61 (1·90–3∙57) 39 (21–67) 35 (18–62) 8∙61 (4∙69–15∙81) 73 (46–108) 43 (18–75) 1∙86 (1∙23–2∙82) Anatomical distribution Mixed 80 (68–94) 45 (35–56) 1∙28 (1∙06–1∙54) 20 (14–28) 16 (11–23) 2∙38 (1∙55–3∙65) 60 (49–73) 30 (21–40) 1∙12 (0∙88–1∙41) Deep 73 (65–82) 38 (32–44) 1∙29 (1∙12–1∙48) 17 (13–22) 13 (10–17) 2∙57 (1∙78–3∙70) 57 (49–66) 27 (21–33) 1∙14 (0∙96–1∙36) Lobar 60 (53–67) 25 (20–29) 1∙22 (1∙06–1∙41) 13 (9–16) 9 (6–9) 1∙87 (1∙29–2∙71) 48 (42–56) 18 (14–23) 1∙17 (0∙99–1∙40) Probable cerebral 55 (40–73) 20 (7–35) 1∙21 (0∙90–1∙64) 9 (4–18) 5 (1–13) 1∙29 (0∙60–2∙77) 48 (34–66) 18 (6–33) 1∙31 (0∙94–1∙83) amyloid angiopathy Ranges in brackets are 95% CIs. Cerebral microbleed location hazard ratios are versus patients without cerebral microbleeds in each location and are adjusted for cerebral microbleed number and our prespecified variables. *Number of patients and time at risk are shown in the appendix. †Overlapping categories. Table 2: Rate and risk of outcome events according to number (burden) and anatomical distribution of baseline cerebral microbleeds in all patients (n=20 322) antithrom­botic treatment (29% vs 28%). In those not inclusion of individual patient data allowed adjustment treated with any antithrombotic drugs, five had intra­ for potential confounding factors. Our study also adds cranial haemor­rhages over 846 patient-years and 65 had new data for the important subgroups of patients with ischaemic strokes over 825 patient-years. The aHRs many (eg, ≥20) cerebral microbleeds, which cause the associated with cerebral microbleed presence were 1·10 most clinical concern and could not be addressed by any (95% CI 0∙17–7∙34) for intracranial haemorrhage and of the previously published meta-analyses. The association 1∙51 (0∙87–2∙65) for ischaemic stroke. of cerebral microbleeds with a consistently higher rate of ischaemic stroke than intracranial haemorrhage Discussion suggests that cerebral micro­ bleeds are a marker for Our large-scale pooled analysis of individual patient data cerebral small vessel diseases that can cause not only confirms that, in patients with recent ischaemic stroke or intracranial haemorrhage, but also ischaemic stroke. transient ischaemic attack treated with antithrombotic Although it has been inferred that cerebral microbleeds drugs, cerebral microbleeds are associated with the are a marker of direct extravasation of red blood cells from subsequent risks of symptomatic intracranial haem­ arterioles and capillaries dam­ aged by bleeding-prone orrhage and ischaemic stroke; as cerebral microbleed arteriopathies, alternative non-haemorrhagic mechanisms burden increases, the relative risk (aHR) of intracranial include ischaemia-mediated iron store release by oligo­ haemorrhage rises more steeply than that of ischaemic dendrocytes10 or phagocytosis of red cell microemboli stroke. Our most important new finding is that, regardless into the perivascular space.11 A report of haemorrhagic of cerebral microbleed burden and distribution (ie, mixed, transformation of small acute micro­in­farcts into cerebral deep, or lobar), or the type of antithrombotic treatment microbleeds provides direct evidence that cerebral micro­ received (oral anticoagulants or antiplatelet therapy), bleeds can result from ischaemic mech­anisms.13 These the absolute risk of ischaemic stroke is consistently sub­ varied mechanisms under­ lying cerebral micro­ bleeds stantially higher than that of intracranial haemorrhage. might explain why even patients at the highest risk of As well as confirming the association between cerebral intracranial haemorrhage still have a higher absolute risk microbleeds and both recurrent ischaemic stroke and of ischaemic stroke. Moreover, patients with cerebral symptomatic intracranial haemorrhage found in smaller micro­bleeds often have multiple vascular risk factors, so cohorts of patients with ischaemic stroke and transient are at risk of not only small vessel ischaemic stroke but ischaemic attack treated with antiplatelet drugs28 or also other ischaemic stroke sub­ types.58 Patients with oral anticoagulants,27,57,35 the large number of partici­ cerebral microbleeds usually also have white matter pants has improved the precision of our estimates of hyper­intensities, which are associ­ated with the risk of stroke recurrence rates and relative hazards, while the recurrent stroke, death, and poor functional outcome after www.thelancet.com/neurology Vol 18 July 2019 661 Articles Composite of intracranial haemorrhage and Symptomatic intracranial haemorrhage Symptomatic ischaemic stroke ischaemic stroke (n=7582 for multivariable model) (n=6942 for multivariable model) (n=6958 in multivariable models) Rate, per 1000 Absolute rate Adjusted hazard Rate, Absolute rate Adjusted hazard Rate, per 1000 Absolute rate Adjusted hazard patient-years* increase, ratio per 1000 increase, ratio patient-years increase, ratio per 1000 patient-years per 1000 per 1000 patient-years patient-years patient-years None 31 (28 to 35) ·· 1 (ref) 5 (3 to 6) ·· 1 (ref) 27 (23 to 30) ·· 1 (ref) Any 46 (39 to 53) 15 (11 to 18) 1∙30 (1∙07 to 1∙57) 12 (9 to 16) 7 (6 to 10) 2∙49 (1∙64 to 3∙79) 32 (26 to 39) 5 (3 to 9) 1∙07 (0∙86 to 1∙35) One 38 (30 to 49) 7 (2 to 14) 1∙19 (0·91 to 1∙56) 10 (6 to 17) 5 (3 to 11) 2∙15 (1∙23 to 3∙75) 26 (19 to 35) –1 (–4 to 5) 0∙96 (0∙69 to 1∙33) Number Two to four 47 (36 to 60) 16 (8 to 25) 1∙23 (0∙93 to 1∙62) 11 (6 to 19) 6 (3 to 13) 2∙22 (1∙21 to 4∙06) 36 (26 to 48) 11 (3 to 18) 1∙10 (0∙80 to 1∙52) Five or more† 62 (45 to 84) 31 (17 to 49) 1∙69 (1∙22 to 2∙35) 20 (11 to 34) 15 (8 to 28) 3∙91 (2∙08 to 7∙34) 40 (26 to 59) 13 (3 to 29) 1∙27 (0∙84 to 1∙91) Ten or more† 75 (46 to 116) 44 (18 to 81) 2∙15 (1∙35 to 3∙43) 23 (8 to 50) 18 (5 to 44) 4∙63 (1∙92 to 11∙22) 46 (24 to 81) 19 (1 to 51) 1∙52 (0∙84 to 2∙67) Anatomical distribution Mixed 58 (42 to 77) 27 (14 to 42) 1∙43 (1∙02 to 2∙00) 15 (7 to 26) 10 (4 to 20) 2∙21 (1∙09 to 4∙47) 42 (29 to 60) 15 (6 to 30) 1∙28 (0∙85 to 1∙94) Deep 52 (42 to 63) 21 (14 to 28) 1∙43 (1∙11 to 1∙84) 14 (9 to 21) 9 (6 to 15) 2∙71 (1∙61 to 4∙59) 35 (27 to 46) 8 (4 to 16) 1∙16 (0∙85 to 1∙59) Lobar 41 (32 to 51) 10 (4 to 16) 1∙13 (0∙87 to 1∙47) 10 (6 to 16) 5 (3 to 10) 1∙63 (0∙94 to 2∙83) 29 (22 to 38) 2 (–1 to 8) 1∙00 (0∙73 to 1∙38) Probable cerebral 27 (13 to 47) –4 (–15 to 12) 0∙76 (0∙41 to 1∙39) 10 (3 to 25) 5 (0 to 19) 1∙29 (0∙47 to 3∙57) 17 (7 to 35) –10 (–16 to 5) 0∙64 (0∙30 to 1∙37) amyloid angiopathy Ranges in brackets are 95% CIs. Cerebral microbleed location hazard ratios are versus patients without cerebral microbleeds in each location and are adjusted for cerebral microbleed number and our prespecified variables. *Number of patients and time at risk are shown in the appendix. †Overlapping categories. Table 3: Rate and risk of outcome events according to baseline cerebral microbleeds in patients treated with oral anticoagulants with or without antiplatelet drugs (n=7737) ischaemic stroke59 and might also contribute to the meta-analysis,28 but our much larger individual participant increased risk of ischaemic stroke associated with cerebral sample size allowed us to investi­ gate high cerebral microbleeds. microbleed burdens (five or more, ten or more, and 20 or We found no evidence that a strictly lobar pattern of more) with adjustment for confounders and greater cerebral microbleeds (fulfilling the Boston criteria for statistical precision and power. probable CAA,5 causing clinical concern for intracranial The comparatively low frequency of symptomatic intra­ bleeding risk35) is associated with the risk of intracranial cranial haemorrhage after ischaemic stroke or transient haemorrhage or ischaemic stroke. These findings might ischaemic attack and the consistently higher risk of reflect low diagnostic accuracy when using cerebral recurrent ischaemic stroke make randomised controlled microbleeds for diagnosis of CAA in patients without trials of antithrombotic treatment (themselves proven in intracerebral haemorrhage or dementia,60 rather than a large randomised trials) guided by cerebral microbleeds true absence of any association of CAA with intracranial challenging. However, ongoing and future randomised haemorrhage. Furthermore, the aHRs for intracranial controlled trials should provide further insights. The MRI haemorrhage associated with lobar cerebral microbleeds substudy in the RESTART trial61 of antiplatelet therapy (compared with patients without lobar cerebral micro­ after intracerebral haemorrhage excluded all but a very bleeds [including none]) were closer to those associated modest harmful effect of antiplatelet therapy on recurrent with deep or mixed cerebral microbleeds (compared with intracerebral haemorrhage in the presence of cerebral patients without deep or mixed cerebral microbleeds microbleeds, but also illustrates how very large sample [including none]). sizes are probably required to identify statistically signifi­ Our results differ from some previous observations in cant interactions in smaller cerebral microbleed subgroups smaller cohorts. First, in contrast to a smaller two-centre in current (eg, the MRI substudy of NAVIGATE ESUS study,29 we did not find that the risk of intracranial [NCT02313909]) and future randomised controlled trials. haemorrhage approached the risk of ischaemic stroke Nevertheless, our large collaborative pooled analysis pro­ after 1 year. Rather, we found that the rate of ischaemic vides the best available evidence on the associations stroke was consistently higher than that of intracranial of cerebral microbleeds with subsequent intracranial haemorrhage, and the aHRs associated with cerebral haemorrhage and ischaemic stroke after ischaemic stroke microbleeds for both ischaemic stroke and intracranial or transient ischaemic attack. haemorrhage remained stable over time. Second, our We included data from a worldwide collaborative data indicate a smaller increase in the relative risk of network, making our results globally generalisable. The intracranial haemorrhage for patients with five or more large individual patient dataset provides high statistical cerebral microbleeds than reported in a previous smaller power and precision for risk estimates, allowing us to 662 www.thelancet.com/neurology Vol 18 July 2019 Articles explore associations with several clinically important Declaration of interests primary outcomes, while adjusting for important prog­ MK reports grants from the Ministry of Health, Labour and Welfare, Japan, and from the National Cerebral and Cardiovascular Center during nostic variables to minimise confounding. Included co­ the conduct of the study; and speaker honoraria from Bayer Yakuhin, horts used validated rating instru­ ments for cerebral Daiichi-Sankyo Company, and Bristol-Myers Squibb (BMS)/Pfizer. micro­bleeds, and we adjusted for the use of different MRI HC reports participation in the steering committee for a clinical trial sequences (T2* GRE or SWI) to detect cerebral micro­ supported by Servier and was a consultant for Hovid Inc. EMA reports personal fees from Pfizer, Boehringer Ingelheim, Nutricia, Abbott, and bleeds, which accounts for the higher sensitivity of SWI for Sanofi, outside the submitted work. JP reports personal fees from detecting cerebral microbleeds compared with T2* GRE.62 Boehringer Ingelheim and Akcea and personal fees and non-financial We followed a published statistical analysis plan and support from Pfizer outside the submitted work. EBA reports grants from confirmed our findings in a two-stage meta-analysis, US–Israel Bi-national Science Foundation, The American Federation for Aging Research, and The Israeli Chief Scientist, Ministry of Health, during indicating the robustness of our results. the conduct of the study. SBC reports grants from the Canadian Institute In terms of limitations, our observational design has of Health Research and a Pfizer Cardiovascular award during the potential for selection bias and confounding of antithrom­ conduct of the study. DJS reports other funding from Bayer and from botic therapy by indication or unmeasured physician BMS/Pfizer outside the submitted work. PL reports other funding from Daiichi-Sankyo, Bayer, and Boehringer Ingelheim, outside the submitted factors; thus, the relative hazards (aHRs) for intracranial work. RA-SS reports grants from the British Heart Foundation, The Stroke haemor­rhage and ischaemic stroke must be interpreted Association, and Chest Heart & Stroke Scotland outside the submitted with caution. To definitively establish whether cerebral work. GYHL reports consultancy for Bayer/Janssen, BMS/Pfizer, micro­ bleeds modify the net clinical benefit of anti­ Biotronik, Medtronic, Boehringer Ingelheim, Microlife, and Daiichi- Sankyo; and speaker honoraria from Bayer, BMS/Pfizer, Medtronic, thrombotic drugs would require a randomised controlled Boehringer Ingelheim, Microlife, Roche, and Daiichi-Sankyo. HPM trial. Many of the included studies did not formally reports personal fees from Neuravi/Cerenovus, Medtronic, Bayer, adjudicate events. The requirement for MRI-suitable Daiichi-Sankyo, and Servier outside the submitted work. DH reports patients probably led to the inclusion of less severe strokes grants from University College Dublin Newman Fellowship supported Bayer during the conduct of the study. MEK reports grants from the than an unselected population. Even with the many indi­ Center for Translational Molecular Medicine during the conduct of the vidual patients included, we could not precisely estimate study. AMT reports grants from the Dutch Heart Foundation during the risks associated with an extremely large number of conduct of the study. AvdL reports grants from the Center for Translation Molecular Medicine and Dutch Heart Foundation during the conduct of cerebral microbleeds (eg, ≥50), but such patients are very the study. JMW reports grants from Wellcome Trust, Chest Heart Stroke rare in clinical practice. Although we adjusted for known Scotland, and Row Fogo Charitable Trust during the conduct of the study. prognostic variables, residual confounding secondary to YS reports a grant from Health and Medical Research Fund. VIHK reports unknown or uncontrolled factors such as stroke mech­ grants from the Netherlands Heart Foundation (grant 2001B071) during the conduct of the study. STE reports grants from Daiichi-Sankyo, Bayer, anism could still have affected our results. Furthermore, Pfizer, and Swiss Heart Foundation during the conduct of the study; other we were unable to include some candidate variables in our funding from Daiichi-Sankyo, Mindmaze, and Stago; and grants from the multivariable models because they were not sufficiently Swiss National Science Foundation outside the submitted work. widely available across all participating cohorts (eg, white NP reports other funding from Daiichi-Sankyo, Bayer, and Boehringer Ingelheim outside the submitted work. EES reports personal fees from matter hyper­ intensities, MRI field strength, diabetes, Portola Pharmaceuticals and Alnylam Pharmaceuticals outside the ischaemic heart disease, renal function, and statin use submitted work. VT reports personal fees and non-financial support from on discharge). Our analyses did not formally assess net Boehringer Ingelheim and personal fees from Bayer, Pfizer/BMS, and clinical benefit, accounting for the greater severity of intra­ Amgen and Medtronic outside the submitted work. RV reports grants and personal fees from Bayer, Boehringer Ingelheim, BMS, Daiichi-Sankyo, cranial haemor­rhage compared with recurrent ischaemic and Medtronic; and personal fees from Morphosys and Amgen outside the stroke. submitted work. HA reports grants from National Institutes of Health In summary, our large-scale pooled analysis in patients during the conduct of the study. PMR reports personal fees from Bayer with recent ischaemic stroke or transient ischaemic attack outside the submitted work. KT reports personal fees from Daiichi-Sankyo, Bayer Yakuhin, BMS, and Nippon Boehringer Ingelheim outside the found that the absolute risk of ischaemic stroke is con­ submitted work. DJWe reports personal fees from Bayer outside the sistently higher than that of intracranial haemorrhage, submitted work. All other authors declare no competing interests. regardless of the number or anatomical distribution Acknowledgments of cerebral microbleeds. However, cerebral microbleeds Funding for the included cohort studies was provided by the British Heart are associated with a greater relative hazard (aHR) for Foundation, Stroke Association, UCLH National Institute of Health intra­cranial haemorrhage than ischaemic stroke; further Research (NIHR) Biomedical Research Centre, Wellcome Trust, Health Research Board Ireland, NIHR Biomedical Research Centre (Oxford, UK), studies are needed to establish the usefulness of neuro­ Canadian Institutes of Health Research, Pfizer Cardiovascular imaging biomarkers, including cerebral microbleeds, Research award, Basel Stroke Funds, Science Funds Rehabilitation in improving risk pre­ diction scores for intracranial Felix-Platter-Hospital, Neurology Research Pool University Hospital Basel, haemorrhage and ischaemic stroke. Bayer AG, Fondo de Investigaciones Sanitarias Instituto de Salud Carlos III (FI12/00296; RETICS INVICTUS PLUS RD16/0019/0010; FEDER), Contributors Imperial College London NIHR Biomedical Research Centre, Dutch Heart DJWe, DW, GA, and JM-F drafted the initial protocol, which was Foundation, Servier, Association de Recherche en Neurologie Vasculaire reviewed with critical revisions and approval by all authors. DW and GA and RHU TRT_cSVD (ANR-16-RHUS-004), Vidi innovational grant from did the statistical analysis. DW, DJW, and GA wrote the first draft of the The Netherlands ZonMw, Chest Heart Stroke Scotland, Medical Research manuscript. All authors contributed to data acquisition, management, Council, Fondation Leducq, The Row Fogo Charitable Trust, National and brain imaging analyses. All authors contributed to critical revision Institute of Health (USA), Adriana van Rinsum-Ponsen Stichting, Japan of the manuscript and approved the final manuscript for submission. Agency for Medical Research and Development (AMED), Ministry of www.thelancet.com/neurology Vol 18 July 2019 663 Articles Health, Labour and Welfare (Japan), and National Cerebral and 22 Lim JS, Hong KS, Kim GM, et al. Cerebral microbleeds and early Cardiovascular Center, Health and Medical Research Grant, Singapore recurrent stroke after transient ischemic attack: results from the National Medical Research Council, and Dutch Heart Foundation. Korean Transient Ischemic Attack Expression Registry. JAMA Neurol 2015; 72: 301–08. References 23 Mok VC, Lau AY, Wong A, et al. Long-term prognosis of Chinese 1 Lip G, Freedman B, De Caterina R, Potpara TS. Stroke prevention in patients with a lacunar infarct associated with small vessel disease: atrial fibrillation: past, present and future. Comparing the guidelines a five-year longitudinal study. Int J Stroke 2009; 4: 81–88. and practical decision-making. Thromb Haemost 2017; 117: 1230–39. 24 Naka H, Nomura E, Takahashi T, et al. Combinations of the 2 Greenberg SM, Vernooij MW, Cordonnier C, et al. presence or absence of cerebral microbleeds and advanced white Cerebral microbleeds: a guide to detection and interpretation. matter hyperintensity as predictors of subsequent stroke types. Lancet Neurol 2009; 8: 165–74. AJNR Am J Neuroradiol 2006; 27: 830–35. 3 Fazekas F, Kleinert R, Roob G, et al. Histopathologic analysis of 25 Ueno H, Naka H, Ohshita T, et al. Association between cerebral foci of signal loss on gradient-echo T2*-weighted MR images in microbleeds on T2*-weighted MR images and recurrent patients with spontaneous intracerebral hemorrhage: evidence of hemorrhagic stroke in patients treated with warfarin following microangiopathy-related microbleeds. AJNR Am J Neuroradiol 1999; ischemic stroke. AJNR Am J Neuroradiol 2008; 29: 1483–86. 20: 637–42. 26 Gregoire SM, Brown MM, Kallis C, Jäger HR, Yousry TA, Werring DJ. 4 Shoamanesh A, Kwok CS, Benavente O. Cerebral microbleeds: MRI detection of new microbleeds in patients with ischemic stroke: histopathological correlation of neuroimaging. Cerebrovasc Dis 2011; five-year cohort follow-up study. Stroke 2010; 41: 184–86. 32: 528–34. 27 Wilson D, Ambler G, Shakeshaft C, et al, and the CROMIS-2 5 Knudsen KA, Rosand J, Karluk D, Greenberg SM. Clinical diagnosis collaborators. Cerebral microbleeds and intracranial haemorrhage risk of cerebral amyloid angiopathy: validation of the Boston criteria. in patients anticoagulated for atrial fibrillation after acute ischaemic Neurology 2001; 56: 537–39. stroke or transient ischaemic attack (CROMIS-2): a multicentre 6 Fisher CM. Hypertensive cerebral hemorrhage. Demonstration observational cohort study. Lancet Neurol 2018; 17: 539–47. of the source of bleeding. J Neuropathol Exp Neurol 2003; 62: 104–07. 28 Wilson D, Charidimou A, Ambler G, et al. Recurrent stroke risk 7 Pasi M, Charidimou A, Boulouis G, et al. Mixed-location cerebral and cerebral microbleed burden in ischemic stroke and TIA: hemorrhage/microbleeds: Underlying microangiopathy and a meta-analysis. Neurology 2016; 87: 1501–10. recurrence risk. Neurology 2018; 90: e119–26. 29 Lau KK, Lovelock CE, Li L, et al. Antiplatelet treatment after 8 Tsai HH, Pasi M, Tsai LK, et al. Microangiopathy underlying transient ischemic attack and ischemic stroke in patients with mixed-location intracerebral hemorrhages/microbleeds: A PiB-PET cerebral microbleeds in 2 large cohorts and an updated systematic study. Neurology 2019; 92: e774–81. review. Stroke 2018; 49: 1434–42. 9 Charidimou A, Shakeshaft C, Werring DJ. Cerebral microbleeds on 30 Microbleeds International Collaborative Network. Worldwide magnetic resonance imaging and anticoagulant-associated collaboration in the Microbleeds International Collaborative intracerebral hemorrhage risk. Front Neurol 2012; 3: 133. Network. Lancet Neurol 2016; 15: 1113–14. 10 Janaway BM, Simpson JE, Hoggard N, et al, and the MRC Cognitive 31 Dichgans M, Wardlaw J, Smith E, et al. METACOHORTS for the Function and Ageing Neuropathology Study. Brain haemosiderin in study of vascular disease and its contribution to cognitive decline older people: pathological evidence for an ischaemic origin of and neurodegeneration: an initiative of the Joint Programme for magnetic resonance imaging (MRI) microbleeds. Neurodegenerative Disease Research. Alzheimers Dement 2016; Neuropathol Appl Neurobiol 2014; 40: 258–69. 12: 1235–49. 11 Grutzendler J, Murikinati S, Hiner B, et al. Angiophagy prevents 32 Cochrane. Tool to assess risk of bias in cohort studies. early embolus washout but recanalizes microvessels through http://methods.cochrane.org/sites/methods.cochrane.org.bias/ embolus extravasation. Sci Transl Med 2014; 6: 226ra31. files/public/uploads/Tool%20to%20Assess%20Risk%20of %20 12 Tanskanen M, Mäkelä M, Myllykangas L, et al. Prevalence and Bias%20in%20Cohort%20Studies.pdf (accessed May 14, 2019). severity of cerebral amyloid angiopathy: a population-based study 33 Collett D. Modelling survival data in medical research. 3rd edn. on very elderly Finns (Vantaa 85+). Neuropathol Appl Neurobiol 2012; London: Chapman and Hall, 2015. 38: 329–36. 34 Fazekas F, Chawluk JB, Alavi A, Hurtig HI, Zimmerman RA. 13 Ito AO, Shindo A, Ii Y, et al. Small cortical infarcts transformed to MR signal abnormalities at 1.5 T in Alzheimer’s dementia and lobar cerebral microbleeds: A Case Series. J Stroke Cerebrovasc Dis normal aging. AJR Am J Roentgenol 1987; 149: 351–56. 2019; 28: e30–32. 35 Soo Y, Abrigo JM, Leung KT, et al, and the IPAAC study group. 14 Thijs V, Lemmens R, Schoofs C, et al. Microbleeds and the risk of Risk of intracerebral haemorrhage in Chinese patients with atrial recurrent stroke. Stroke 2010; 41: 2005–09. fibrillation on warfarin with cerebral microbleeds: 15 Boulanger JM, Coutts SB, Eliasziw M, et al, and the VISION Study the IPAAC-Warfarin study. J Neurol Neurosurg Psychiatry 2019; Group. Cerebral microhemorrhages predict new disabling or fatal 90: 428–35. strokes in patients with acute ischemic stroke or transient ischemic 36 Gratz PP, El-Koussy M, Hsieh K, et al. Preexisting cerebral attack. Stroke 2006; 37: 911–14. microbleeds on susceptibility-weighted magnetic resonance imaging 16 Bokura H, Saika R, Yamaguchi T, et al. Microbleeds are associated and post-thrombolysis bleeding risk in 392 patients. Stroke 2014; with subsequent hemorrhagic and ischemic stroke in healthy 45: 1684–88. elderly individuals. Stroke 2011; 42: 1867–71. 37 Yang J, Wong A, Wang Z, et al. Risk factors for incident dementia 17 Akoudad S, Ikram MA, Koudstaal PJ, et al. Cerebral microbleeds after stroke and transient ischemic attack. Alzheimers Dement 2015; are associated with the progression of ischemic vascular lesions. 11: 16–23. Cerebrovasc Dis 2014; 37: 382–88. 38 Ben Assayag E, Korczyn AD, Giladi N, et al. Predictors for 18 Fluri F, Jax F, Amort M, et al. Significance of microbleeds in poststroke outcomes: the Tel Aviv Brain Acute Stroke Cohort patients with transient ischaemic attack. Eur J Neurol 2012; (TABASCO) study protocol. Int J Stroke 2012; 7: 341–47. 19: 522–24. 39 Chabriat H, Maeder P, Gass A, Michel P, Bracoud L, Hennerici M. 19 Imaizumi T, Inamura S, Kohama I, Yoshifuji K, Nomura T, Results of the PERFORM magnetic resonance imaging study. Komatsu K. Antithrombotic drug uses and deep intracerebral J Neurol 2013; 260: 3071–76. hemorrhages in stroke patients with deep cerebral microbleeds. 40 Truijman MT, Kooi ME, van Dijk AC, et al. Plaque At RISK J Stroke Cerebrovasc Dis 2013; 22: 869–75. (PARISK): prospective multicenter study to improve diagnosis of 20 Jeon SB, Kang DW, Cho AH, et al. Initial microbleeds at MR high-risk carotid plaques. Int J Stroke 2014; 9: 747–54. imaging can predict recurrent intracerebral hemorrhage. J Neurol 41 Yoshimura S, Koga M, Sato S, et al. Two-year outcomes of 2007; 254: 508–12. anticoagulation for acute ischemic stroke with nonvalvular atrial 21 Kwa VI, Algra A, Brundel M, Bouvy W, Kappelle LJ, and the MICRO fibrillation—SAMURAI-NVAF Study. Circ J 2018; 82: 1935–42. Study Group. Microbleeds as a predictor of intracerebral 42 van Leijsen EMC, van Uden IWM, Ghafoorian M, et al. haemorrhage and ischaemic stroke after a TIA or minor ischaemic Nonlinear temporal dynamics of cerebral small vessel disease: stroke: a cohort study. BMJ Open 2013; 3: e002575. The RUN DMC study. Neurology 2017; 89: 1569–77. 664 www.thelancet.com/neurology Vol 18 July 2019 Articles 43 Karayiannis C, Soufan C, Chandra RV, et al. Prevalence of brain 54 Zerna C, Modi J, Bilston L, Shoamanesh A, Coutts SB, Smith EE. MRI markers of hemorrhagic risk in patients with stroke and atrial Cerebral microbleeds and cortical superficial siderosis in patients fibrillation. Front Neurol 2016; 7: 151. presenting with minor cerebrovascular events. Stroke 2016; 44 Song TJ, Kim J, Song D, et al. Association of cerebral microbleeds 47: 2236–41. with mortality in stroke patients having atrial fibrillation. Neurology 55 Wardlaw J, Makin S, Valdés Hernández M, et al. Blood-brain barrier 2014; 83: 1308–15. failure as a core mechanism in cerebral small vessel disease and 45 Jeong HG, Kim BJ, Yang MH, Han MK, Bae HJ. dementia: evidence from a cohort study. Alzheimers Dement 2017; Neuroimaging markers for early neurologic deterioration in single 13: 634–43. small subcortical infarction. Stroke 2015; 46: 687–91. 56 Horstmann S, Möhlenbruch M, Wegele C, et al. Prevalence of atrial 46 Kim BJ, Park JM, Kang K, et al. Case characteristics, hyperacute fibrillation and association of previous antithrombotic treatment in treatment, and outcome information from the clinical research patients with cerebral microbleeds. Eur J Neurol 2015; 22: 1355–62. center for stroke-fifth division registry in South Korea. J Stroke 2015; 57 Charidimou A, Karayiannis C, Song TJ, et al, and the International 17: 38–53. META-MICROBLEEDS Initiative. Brain microbleeds, 47 Callaly E, Ni Chroinin D, Hannon N, et al. Rates, predictors, and anticoagulation, and hemorrhage risk: meta-analysis in stroke outcomes of early and late recurrence after stroke: the north Dublin patients with AF. Neurology 2017; 89: 2317–26. population stroke study. Stroke 2016; 47: 244–46. 58 Song TJ, Kim J, Lee HS, et al. The frequency of cerebral 48 Imaizumi T, Inamura S, Nomura T. The severities of white matter microbleeds increases with CHADS(2) scores in stroke patients lesions possibly influence the recurrences of several stroke types. with non-valvular atrial fibrillation. Eur J Neurol 2013; 20: 502–08. J Stroke Cerebrovasc Dis 2014; 23: 1897–902. 59 Georgakis MK, Duering M, Wardlaw JM, Dichgans M. WMH and 49 Soo Y, Abrigo J, Leung KT, et al. Correlation of non-vitamin K long-term outcomes in ischemic stroke: a systematic review and antagonist oral anticoagulant exposure and cerebral microbleeds in meta-analysis. Neurology 2019; 92: e1298–308. Chinese patients with atrial fibrillation. J Neurol Neurosurg Psychiatry 60 Martinez-Ramirez S, Romero JR, Shoamanesh A, et al. 2018; 89: 680–86. Diagnostic value of lobar microbleeds in individuals without 50 Douven E, Schievink SHJ, Verhey FRJ, et al. The cognition and intracerebral hemorrhage. Alzheimers Dement 2015; 11: 1480–88. affect after stroke—a prospective evaluation of risks (CASPER) 61 Al-Shahi Salman R, Minks D, Mitra D, et al. Effects of antiplatelet study: rationale and design. BMC Neurol 2016; 16: 1–11. therapy on stroke risk by brain imaging features of intracerebral 51 Martí-Fàbregas J, Medrano-Martorell S, Merino E, et al. MRI predicts haemorrhage and cerebral small vessel diseases: sub-group analyses intracranial hemorrhage in patients who receive long-term oral of the RESTART randomised, open-label trial. Lancet Neurol 2019, anticoagulation. Neurology 2019; published online April 19. published online May 22. http://dx.doi.org/10.1016/ DOI:10.1212/WNL.0000000000007532. S1474-4422(19)30184-X. 52 Yan S, Jin X, Zhang X, Zhang S, Liebeskind DS, Lou M. 62 Shams S, Martola J, Cavallin L, et al. SWI or T2*: which MRI Extensive cerebral microbleeds predict parenchymal haemorrhage sequence to use in the detection of cerebral microbleeds? and poor outcome after intravenous thrombolysis. The Karolinska Imaging Dementia Study. AJNR Am J Neuroradiol J Neurol Neurosurg Psychiatry 2015; 86: 1267–72. 2015; 36: 1089–95. 53 Orken DN, Uysal E, Timer E, Kuloglu-Pazarcı N, Mumcu S, Forta H. New cerebral microbleeds in ischemic stroke patients on warfarin treatment: two-year follow-up. Clin Neurol Neurosurg 2013; 115: 1682–85. www.thelancet.com/neurology Vol 18 July 2019 665