This article reviews published stroke subtype classification systems and offers rules and a basis for a new way to subtype stroke patients. Stroke subtyping can have different purposes, e.g. describing patients’ characteristics in a clinical trial, grouping patients in an epidemiological study, careful phenotyping of patients in a genetic study, and classifying patients for therapeutic decision-making in daily practice. The classification should distinguish between ischemic and hemorrhagic stroke, subarachnoid hemorrhage, cerebral venous thrombosis, and spinal cord stroke. Regarding the 4 main categories of etiologies of ischemic stroke (i.e. atherothrombotic, small vessel disease, cardioembolic, and other causes), the classification should reflect the most likely etiology without neglecting the vascular conditions that are also found (e.g. evidence of small vessel disease in the presence of severe large vessel obstructions). Phenotypes of large cohorts can also be characterized by surrogate markers or intermediate phenotypes (e.g. presence of internal carotid artery plaque, intima-media thickness of the common carotid artery, leukoaraiosis, microbleeds, or multiple lacunae). Parallel classifications (i.e. surrogate markers) may serve as within-study abnormalities to support research findings.

Stroke is a heterogeneous disease with more than 150 known causes. Most registries have failed to identify a definite cause in 25–39% of patients, depending on the quality, completeness, and rapidity of the work-up. This group of strokes of unknown causes (the so-called ‘cryptogenic strokes’, a term popular among neurologists but perhaps unnecessarily cryptic to students, patients, and most nonstroke physicians) should be a major focus for future clinical research.

Subtyping ischemic stroke can have different purposes, e.g. describing patients’ characteristics in a clinical trial, grouping patients in an epidemiological study, careful phenotyping of patients in a genetic study, and classifying patients for therapeutic decision-making in daily practice. Most epidemiological studies have failed to find a link between blood-cholesterol levels and stroke, but studies that looked at some subtyping (ischemic vs. hemorrhagic stroke) did find such a relation. Other large epidemiological studies have found a strong association between carotid stenosis and blood cholesterol levels. Although carotid stenosis is a cause of stroke, serum cholesterol levels have not been associated with atherothrombotic stroke because no epidemiologic studies have properly looked for this association in this particular subtype of ischemic stroke.

In addition to subtyping ischemic stroke, depending on the question being asked in clinical research, it can be useful to further characterize the cohort by intermediate phenotypes (e.g. presence of internal carotid artery plaque, intima-media thickness of the common carotid artery, or presence of leukoaraiosis, microbleeds, or multiple lacunae on magnetic resonance imaging). Since all of these subtyping approaches have their own biases, using different approaches in parallel (i.e. ischemic stroke subtyping and intermediate phenotype) may serve as a within-study method of replication to support research findings.

Stroke Data Bank Subtype Classification

Derived from the Harvard Stroke Registry classification [1], the National Institute of Neurological Disorders and Stroke (NINDS) Stroke Data Bank recognized 5 major groups: brain hemorrhages; brain infarctions, and among them atherothrombotic and tandem arterial pathological abnormalities; cardioembolic stroke; lacunar stroke; and stroke from rare causes or undetermined etiology (table 1) [2].

Table 1

Stroke subtype classifications: Stroke Data Bank, Lausanne Stroke Registry, TOAST and GÉNIC

Stroke subtype classifications: Stroke Data Bank, Lausanne Stroke Registry, TOAST and GÉNIC
Stroke subtype classifications: Stroke Data Bank, Lausanne Stroke Registry, TOAST and GÉNIC

Weaknesses of this classification:

– The very restrictive definition for atherothrombotic stroke (the investigators selected only patients with ≥90% stenosis), resulting in an evident underestimation of the overall burden of atherothrombotic disease (9% of the entire cohort were in the atherothrombotic group).

– The wide definition for lacunar strokes without asking for either assessment of intracranial arteries or additional old lacunar-type infarcts or small-vessel-related parenchymal abnormalities on brain imaging. Modern MRI with diffusion-weighted imaging was not used at the time.

– The evident overestimation of the group of ‘undetermined’ etiology (39% of the cohort), not only because the newer diagnostic tools were not available at that time (e.g. transesophageal echocardiography, TEE, magnetic resonance angiography, MRA, duplex ultrasound examination, and transcranial Doppler), but also because of the restrictive definition of atherothrombotic causes and the inclusion in this group of patients with tandem lesions.

Strength of this classification:

– Only patients with atherothrombosis likely to be causally related to stroke could be included in the group classified as having atherothrombotic stroke, with a large group of other individuals classified as having ‘stroke of unknown cause’. This approach could be the best option in the search of new causes of stroke among patients with no known cause or with disease not causally related to the stroke event.

Oxfordshire Community Stroke Project Subtype Classification

The classification from the Oxfordshire Community Stroke Project (OCSP) was proposed to characterize this population-based epidemiological study [3, 4]. The investigators had to cope with the quality and cost of the work-up available at that time in the UK. Transient ischemic attacks (TIA) and strokes were detected and investigated by general practitioners. The classification was based on clinical findings only. Computed tomography (CT) scanning was the best investigational test performed, but assessment of extra- and intracranial arteries and precise cardiac work-up were not available. This likely led the OCSP investigators to classify patients according to the extent and site of brain infarction on clinical grounds alone (table 2).

Table 2

OCSP classification

OCSP classification
OCSP classification

Weaknesses of this classification:

– The extent and site of the brain infarct is unlikely to be specific to a particular stroke etiology.

– Patients classified as having a lacunar infarct may have a missed M1 stenosis or a cardiac source of embolism.

– This classification should no longer be used in the 21st century to investigate potential risk factors or causes of stroke.

Strengths of this classification:

– Patients are easy to classify into groups based on clinical grounds and CT scanning, which is done in almost all patients.

– The outcome of stroke is driven strongly by the severity of the stroke, which is well reflected in this classification, without addressing the cause of the stroke (i.e. without a complete work-up).

Trial of ORG 10172 in Acute Stroke Treatment Subtype Classification

Since 1993, most clinical researchers used the classification proposed by the Trial of ORG 10172 in Acute Stroke Treatment (TOAST) investigators. The original purpose of this classification was to better characterize the TOAST cohort of patients in order to investigate the potential efficacy of danaparoid in various stroke subtypes [5]. The TOAST investigators defined 11 categories of stroke, which were further collapsed into 5 groups (table 1). Only these 5 groups were used in further clinical research [6]. Consequently, use of the TOAST classification can mask the paucity of clinical investigations on the one hand, and does not account for a very precise work-up on the other. This is the necessary goal in the database of a multicentre trial such as TOAST, but may not systematically apply to all other types of clinical research.

Despite this, the TOAST classification system has become the most widely used in recent literature, most often in studies that did not investigate the efficacy of new acute stroke treatments, such as genetic association studies, evaluations of new potential risk factors or causes of stroke, epidemiologic studies, etc. Consequently, we question the accuracy and utility of the TOAST classification for research goals other than acute-phase treatment trials.

Weaknesses of this classification:

– Small vessel (lacunar) stroke is defined by the clinical syndrome and the size of the infarct. Consequently, a single small deep infarct due to M1 middle cerebral artery atherosclerotic stenosis could be classified by researchers as small vessel disease if an appropriate M1 middle cerebral artery evaluation has not been performed.

– Cardiac sources of embolism are further classified as high or medium risk. Consequently, a patient with V3 vertebral artery dissection ignored by a late negative or incomplete work-up and with a patent foramen ovale (PFO) can be classified by the clinical researcher as a ‘cardiac embolism’.

– Stroke of undetermined cause is the most heterogeneous group in the TOAST classification system, as in the Stroke Data Bank, with patients having at least 2 definite potential causes grouped with those with an entirely negative work-up, and with patients with documented atherosclerotic disease who did not reach the 50% stenosis limit required to qualify for atherothrombotic stroke. Thus, when looking at genetic-association studies or potential risk factors such as serum cholesterol or a potential new cause such as PFO or antiphospholipid antibodies, clinical researchers introduced a major bias.

– This classification system could also flaw the medical decision-making process. For example, the TOAST classification allows many clinicians to oversize the ‘stroke of undetermined etiology’ group (table 1), into which patients with small vessel disease, patients with a ‘missed’ diagnosis of dissection (by ignorance or by insufficient or late work-up), and patients with several risk factors or documented atherosclerotic disease that does not fit into the ‘atherothrombotic’ group can fall (table 1). Oversizing this group may lead, for example, to inappropriately close thousands of innocent PFO [7].

Strengths of this classification:

– The reliability of the classification has been improved by the use of a computerized algorithm, and variability can be addressed by having 2 raters classify each case. [8] This does not, however, overcome the criticisms outlined above.

– Another algorithm has been proposed by adding 3 subcategories to the 11 groups (i.e. evident, probable, or possible), allowing the authors to decrease the size of the group of undetermined cause from 40 to 4% with a ĸ value of 0.90 versus 0.78 for the original TOAST classification [9]. However, one could question whether having only 4% of patients in the unknown-cause group by ‘forcing’ the classification of other patients into the 11 categories by accepting a weak level of evidence is really of help when the research purpose is to phenotype the cohort for genetic studies or to find new potential causes of stroke among the group of unknown cause.

Many other classifications have been proposed, such as those from the Lausanne Stroke Registry and the Étude du profil Génétique de l’Infarctus Cérébral (GÉNIC) study (table 1) [10, 11]. In the Lausanne Stroke Registry, the group with stroke of unknown cause was one of the least important, widening the atherothrombotic group by the inclusion of patients with stenosis <50% or plaques or with at least 2 risk factors. This actually accounted for the true prevalence of atherothrombotic disease from a preventive treatment point of view, but also likely overestimated the atherothrombotic causative relationship with the brain infarct. The GÉNIC classification also tended to overestimate the atherothrombotic group by including patients with any carotid stenosis >30% in this group. This approach was justified by the need to identify all patients with atherothrombosis for a genetic phenotyping purpose, irrespective of stroke cause, provided that there were no cardiac sources of embolism, no suspicion of lacunar stroke, and no rare causes of stroke.

A stroke subtype classification should be useful both in daily clinical practice and in epidemiological and genetic studies, randomized acute clinical trials, and prevention studies of various types (e.g. including the hemorrhagic aspects).

(1) The classification should first distinguish between ischemic and hemorrhagic stroke, subarachnoid hemorrhage, cerebral venous thrombosis, and spinal cord stroke (table 3).

Table 3

Stroke subtypes

Stroke subtypes
Stroke subtypes

(2) Regarding the 4 main categories of ischemic stroke (i.e. atherothrombotic, cardioembolic, small vessel disease, and other causes), the classification should identify the most likely etiology(ies) without neglecting mixed phenotypes (e.g. evidence of small vessel disease in the presence of severe symptomatic atherosclerotic stenosis).

(3) The classification should be based on patient’s medical history, physical examination, and diagnostic tests performed in good time.

What Is a Minimal Diagnostic Evaluation?

– Assessment of the main risk factors: blood pressure (in the sitting position after 10 minutes of rest, electronic measurement) or a personal history of chronic blood-pressure-lowering treatment, lipid profile, tobacco smoking (current or stopped within the previous 6 months), diabetes mellitus, weight, height, waist circumference, physical exercise versus sedentary lifestyle, and family history of vascular disease; personal history of coronary intervention, acute coronary syndrome or myocardial infarction; and personal history of atrial fibrillation.

– Blood tests: hematocrit, platelets, red and white cell counts, and prothrombin time.

– Electrocardiogram

– Assessment of extracranial arteries (either by carotid ultrasound examination or MRA, or CT angiography, or X-ray angiography).

– Assessment of intracranial arteries (either by transcranial Doppler, and/or MRA, and/or CT angiography, and/or X-ray angiography, and/or high-resolution MRI [12, 13]).

– Specific etiologies:

• Suspicion of endocarditis needs emergent hemocultures and comprehensive echocardiographic examination;

• Suspicion of aortic dissection needs emergent thoracic CT or TEE;

• Suspicion of cerebral artery dissection needs expert ultrasound examination, MRA or X-ray angiography, and fat-saturated MRI in the axial sections to show the hematoma in the arterial wall. This imaging should be performed within 15 days of symptom onset. A late work-up can be normal because the dissection has spontaneously resolved with disappearance of the hematoma in the arterial wall.

When and How to Evaluate Cardiac Cavities and Wall

– Suspicion of intracardiac thrombus;

– Possibility of intracardiac mass;

– Search for endocarditis, either bacterial or nonbacterial;

– Search for akinetic or aneurismal ventricle;

– Prosthetic valve or clinical suspicion of valvular disease;

– Search for endomyocardial fibrosis;

– Diagnostic tools include transthoracic and transesophageal echocardiography, cardiac CT or cardiac MRI.

When and How to Evaluate the Aorta

– The thoracic aorta should be evaluated when the cause is unknown (patients not falling into groups 1.1 to 1.5 according to table 3 or patients in group 1.6);

– TEE remains the gold standard; CT angiography and high-resolution MRI are still investigational;

– TEE should be performed by a trained echocardiographer who will look specifically for aortic atheroma during the examination.

When Should TEE Be Performed?

– When there is a need to assess the right and left atrial cavities;

– When searching for an atrial septal aneurysm;

– When there is a need to assess the thoracic aorta;

– At times, in addition to TEE, a cardiac CT or MRI can help search for a cardiac pathology.

When and How Should a Coronary Artery Disease Be Considered?

– Patients at high risk of coronary artery disease, such as those with a history of chest pain, diabetes, or documented atherosclerosis in the cerebral arteries;

– Exercise T1-201 or dipyridamole myocardial scintigraphy;

– CT coronary angiography is still under evaluation;

– X-ray coronary angiography is indicated in patients with a positive myocardial scintigraphy or in patients with acute coronary syndrome.

How Precise Should the Search for Atrial Arrhythmia Be?

– Continuous monitoring during the acute phase of stroke using a monitor – remote telemetry is helpful when available;

– Holter recording in patients with palpitations;

– Assessment of atrial vulnerability should only be investigational.

For Whom Should a Complete Evaluation of Hemostasis Be Performed?

– For patients with a family history of thrombophilia;

– For young patients with stroke of unknown cause (group 1.6);

– For patients with suspected cancer-related thrombophilia;

– For patients with associated deep vein thrombosis or pulmonary embolism (mainly if repeated and erratic events);

– Also in patients with recurrent brain embolism in the presence of atrial fibrillation and an international normalized ratio in the therapeutic range.

When to Look for Other Causes

– After quick elimination of atherosclerosis, cardiac sources of embolism, and small vessel disease;

– In stroke patients with unusual clinical features such as fever, systemic inflammation, skin changes, other organ(s) involvement (e.g. pleurisy, splenomegaly, kidney involvement, uveitis), meningitis, and epilepsy.

How Rapid Should the Work-Up Be?

– A late work-up can be normal in a patient with dissection. Here, the patient is falsely classified as having a stroke of unknown cause. Finding a PFO in this patient, or an antiphospholipid antibody concentration of 40 GPL units, could then be misleading and prompt inadequate long-term treatment. This problem is illustrated by the following case study: a 32-year-old woman had dense right-sided hemiplegia at 5:00 a.m. A CT scan showed early signs of a large left pan-hemispheric infarction. At 8:30 am, carotid ultrasound showed left internal carotid artery occlusion. At 9:00 am, TEE showed a large clot moving in the aortic arch at the level of the left common carotid artery. A follow-up TEE was performed 48 h later and the aortic arch was normal, with no evidence remaining of the earlier clot [Amarenco, P., and Cohen, A.A., unpublished data].

– This patient illustrates the importance of rapid work-up. This young woman also had a PFO; if the first TEE had been performed 8 or more days later, many clinicians would have falsely concluded that the PFO had caused the stroke and proposed closing it.

– In both young and elderly patients with stroke of no known cause, every effort should be made to perform a complete cardiac assessment as quickly as possible, and no later than 3–8 days after stroke onset.

Examples are carotid plaque or intima-media thickness on ultrasound examination and small vessel disease-related parenchymal abnormalities on MRI.

– Patients have an assessment of the internal carotid artery plaque with ultrasound, or a measurement of the common carotid artery intima-media thickness. With such an approach, the entire cohort (either clinical or epidemiological) can be characterized and analyzed with regard to the presence of atherosclerosis, irrespective of stroke subtype.

– Patients have MRI assessment of the brain parenchyma for leukoaraiosis, dilatation of perivascular spaces, multiple lacunae, and microbleeds; altogether these MRI abnormalities have been associated with small vessel disease. Looking at new risk factors for, or associations with, small vessel disease could either focus on ischemic stroke subtyping (i.e. group 3 of the classification – small vessel disease) or on small vessel disease-related parenchymal abnormalities on MRI with 1, 2, 3, or 4 of these abnormalities, or both.

The authors thank the International Expert Opinion Committee who reviewed the early draft of the classification and gave valuable comments and suggestions. However, the views of the article are not necessarily those of the experts listed: H.P. Adams, G.W. Albers, O. Benavente, J. Broderick, H. Chabriat, A. Chamorro, H.C. Diener, J. Ferro, L.B. Goldstein, W. Hacke, G.J. Hankey, D.F. Hanley, A. Hillis, S.C. Johnston, M. Kaste, S.J. Kittner, G.L. Lenzi, S.R. Levine, D. Leys, J.P. Mohr, H. Markus, J.L. Mas, B. Norrving, J.M. Orgogozo, P.M. Rothwell, R.L. Sacco, J.L. Saver, S. Uchiyama, S. Warach, R.J. Wityk, L.K.S. Wong.

Mohr JP, Caplan LR, Melski JW, Goldstein RJ, Duncan GW, Kistler JP, Pessin MS, Bleich HL: The Harvard Cooperative Stroke Registry: a prospective registry. Neurology 1978;28:754–762.
Sacco RL, Ellenberg JH, Mohr JP, Tatemichi TK, Hier DB, Price TR, Wolf PA: Infarcts of undetermined cause: the NINCDS Stroke Data Bank. Ann Neurol 1989;25:382–390.
Bamford J, Sandercock PA, Dennis MS, Burn J, Warlow CP: Classification and natural history of clinically identifiable subtypes of brain infarction. Lancet 1991;337:1521–1526.
Lindley RI, Warlow CP, Wardlaw JM, Dennis MS, Slattery J, Sandercock PA: Interobserver reliability of a clinical classification of acute cerebral infarction. Stroke 1993;24:1801–1804.
Adams HP, Bendixen BH, Kappelle LJ, Biller J, Love BB, Gordon DL, Marsh EE: Classification of subtype of acute ischemic stroke: definitions for use in a multicenter clinical trial. Stroke 1993;24:35–41.
Landau WM, Nassief A: Time to burn the TOAST. Stroke 2005;36:902–904.
Amarenco P: Patent foramen ovale and stroke: smoking gun or guilty by association? Heart 2005;91:441–443.
Goldstein LB, Jones MR, Matchar DB, Lloyd JE, Hoff J, Chilukuri V, Armstrong SB, Horner RD: Improving the reliability of stroke subgroup classification using the Trial of Org 10172 in Acute Stroke Treatment (TOAST) criteria. Stroke 2001;32:1091–1097.
Ay H, Furie KL, Singhal A, Smith WS, Sorensen AG, Koroshetz WJ: An evidence-based causative classification system for acute ischemic stroke. Ann Neurol 2005;58:688–697.
Bogousslavsky J, Van Melle G, Regli F: The Lausanne Stroke Registry: analysis of 1,000 consecutive patients with first stroke. Stroke 1988;19:1083–1092.
Touboul PJ, Elbaz A, Koller C, Lucas C, Adraï V, Chédru F, Amarenco P, GÉNIC Investigators: Common carotid artery intima-media thickness and ischemic stroke subtypes: the GÉNIC case-control study. Circulation 2000;102:313–318.
Klein IF, Lavallée PC, Schouman-Claeys E, Amarenco P: High-resolution MRI identifies basilar artery plaques in paramedian pontine infarct. Neurology 2005;64:551–552.
Klein IF, Lavallée PC, Touboul PJ, Schouman-Claeys E, Amarenco P: In vivo human middle cerebral artery wall imaging using high-resolution magnetic resonance imaging. Neurology 2006;67:327–329.
Copyright / Drug Dosage / Disclaimer
Copyright: All rights reserved. No part of this publication may be translated into other languages, reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording, microcopying, or by any information storage and retrieval system, without permission in writing from the publisher.
Drug Dosage: The authors and the publisher have exerted every effort to ensure that drug selection and dosage set forth in this text are in accord with current recommendations and practice at the time of publication. However, in view of ongoing research, changes in government regulations, and the constant flow of information relating to drug therapy and drug reactions, the reader is urged to check the package insert for each drug for any changes in indications and dosage and for added warnings and precautions. This is particularly important when the recommended agent is a new and/or infrequently employed drug.
Disclaimer: The statements, opinions and data contained in this publication are solely those of the individual authors and contributors and not of the publishers and the editor(s). The appearance of advertisements or/and product references in the publication is not a warranty, endorsement, or approval of the products or services advertised or of their effectiveness, quality or safety. The publisher and the editor(s) disclaim responsibility for any injury to persons or property resulting from any ideas, methods, instructions or products referred to in the content or advertisements.