PDF: TOAST
(1993)Objective
To develop and test a standardized classification system for determining the etiologic subtype of acute ischemic stroke for use in a multicenter clinical trial.
Study Summary
• This paper describes the TOAST (Trial of Org 10172, a heparinoid, in Acute Stroke Treatment) classification system.
• It defines five subtypes of ischemic stroke.
• The system relies on clinical features and ancillary diagnostic tests, and allows for 'possible' or 'probable' diagnoses.
• A validation study on 20 patients found very high interrater agreement (95%).
• It defines five subtypes of ischemic stroke.
• The system relies on clinical features and ancillary diagnostic tests, and allows for 'possible' or 'probable' diagnoses.
• A validation study on 20 patients found very high interrater agreement (95%).
Intervention
This was a validation study of a diagnostic classification system. The parent TOAST trial was a randomized, placebo-controlled study of the low-molecular-weight heparinoid Org 10172 in patients within 24 hours of acute ischemic stroke.
Study Design
Arms: Array
Outcome
• The TOAST classification system was found to be easy to use with high interobserver reliability.
• Two independent neurologists agreed on the final etiologic diagnosis in 19 of 20 patients.
• The inclusion of ancillary diagnostic test results was critical, causing physicians to change their initial diagnosis in 35% of cases.
• Two independent neurologists agreed on the final etiologic diagnosis in 19 of 20 patients.
• The inclusion of ancillary diagnostic test results was critical, causing physicians to change their initial diagnosis in 35% of cases.
Bottom Line
The TOAST classification system is a straightforward and reliable method for categorizing ischemic stroke subtypes with high interobserver agreement. It became the most widely used stroke classification system worldwide, cited over 12,000 times, and remains the standard for clinical trials, epidemiologic studies, and clinical practice.
Major Points
- THE most widely used stroke classification system in the world — cited >12,000 times. Every major stroke trial since 1993 uses TOAST categories for enrollment, subgroup analysis, or outcome reporting.
- Five etiological subtypes: (1) Large-artery atherosclerosis (>50% stenosis or occlusion of major brain/neck artery), (2) Cardioembolism (high-risk: AF, mechanical valve, thrombus; medium-risk: PFO, MVP), (3) Small-vessel occlusion (lacune <1.5 cm, classic lacunar syndrome, no large-artery/cardiac source), (4) Stroke of other determined etiology (dissection, vasculitis, hypercoagulable, Moyamoya, etc.), (5) Stroke of undetermined etiology ('cryptogenic').
- Two-tier certainty system: 'Probable' (clinical + diagnostic studies strongly support one subtype, excluding others) vs 'Possible' (incomplete workup or competing etiologies). This distinction proved essential for trial enrollment — most trials require 'probable' classification.
- Undetermined etiology (cryptogenic) became the largest category in many cohorts (25–40%) — this gap drove decades of research into ESUS, PFO closure (RESPECT, CLOSE, DEFENSE-PFO), occult AF detection (CRYSTAL AF, EMBRACE), and aortic arch atheroma.
- Validation: 95% interrater agreement (19/20 patients classified identically by two independent neurologists). Ancillary studies changed initial clinical diagnosis in 35% of cases — underscoring the importance of comprehensive workup.
- Named after the parent trial: Trial of Org 10172 in Acute Stroke Treatment — a randomized trial of danaparoid (low-molecular-weight heparinoid) for acute stroke that was ultimately negative, but whose classification system became far more influential than the trial itself.
- Large-artery atherosclerosis criteria: >50% stenosis or occlusion of a major extracranial or intracranial artery consistent with clinical syndrome. Cortical, cerebellar, brainstem, or subcortical >1.5 cm infarct. Negative cardiac workup.
- Small-vessel occlusion (lacunar) criteria: Classic lacunar syndrome (pure motor, pure sensory, sensorimotor, ataxic hemiparesis, dysarthria-clumsy hand) with subcortical/brainstem lesion <1.5 cm. No >50% stenosis. No cardiac source.
- Later competitors: CCS (Causative Classification System), ASCO (Atherosclerosis, Small-vessel, Cardiac, Other), SSS-TOAST (Stop Stroke Study TOAST with automated algorithm) — but none fully replaced the original TOAST in clinical practice.
- Limitations recognized from the start: high rate of 'undetermined' classification (especially in patients with incomplete workup or competing etiologies), somewhat arbitrary 1.5 cm lacunar size cutoff, and no provision for multiple simultaneous etiologies (addressed by ASCO).
Study Design
- Study Type
- Validation study of a diagnostic classification system.
- Randomization
- No
- Blinding
- The two neurologists performing the classification were independent and had not participated in the writing of the criteria.
- Sample Size
- 20
- Centers
- 1
- Countries
- United States
Primary Outcome
Definition: Interphysician agreement in classifying stroke subtype using the TOAST system after all ancillary diagnostic studies were reviewed.
| Control | Intervention | HR/OR | P-value |
|---|---|---|---|
| Agreement in 19 of 20 patients (95%). | - |
Limitations & Criticisms
- High rate of 'undetermined etiology' (25–40% of strokes) — critics argue the system creates a diagnostic 'wastebasket' category that is too large to be clinically useful for guiding treatment decisions.
- Validation on only 20 patients at a single center — while 95% agreement was impressive, larger multi-center validation studies showed lower agreement rates (kappa 0.54–0.78), particularly for the undetermined category.
- No provision for multiple simultaneous etiologies — a patient with both AF and >50% carotid stenosis falls into 'undetermined,' losing information about both competing etiologies. ASCO classification later addressed this.
- 1.5 cm lacunar size cutoff is arbitrary — modern MRI shows that some lacunar-mechanism infarcts exceed 1.5 cm, and some small infarcts are embolic rather than lacunar.
- Does not account for modern diagnostic advances: prolonged cardiac monitoring (implantable loop recorders), high-resolution vessel wall MRI, genetic testing, and advanced echocardiography not available in 1993.
- Static classification — TOAST assigns a single diagnosis at one time point, but stroke etiology may be reclassified as additional testing becomes available (e.g., AF detected months later by insertable cardiac monitor).
- Binary medium- vs high-risk cardiac source distinction is oversimplified — PFO with atrial septal aneurysm was classified as 'medium risk' but later proven to warrant closure (CLOSE, RESPECT long-term, DEFENSE-PFO).
- Does not incorporate the ESUS concept (embolic stroke of undetermined source) — introduced by Hart et al. in 2014 as a refinement of the cryptogenic category, but ESUS itself failed to guide treatment in NAVIGATE ESUS and RE-SPECT ESUS.
- Geographic bias: developed and validated in a US academic center — stroke etiology distributions differ substantially in African, Asian, and Latin American populations where intracranial atherosclerosis and hematologic causes are more prevalent.
Citation
Stroke 1993;24:35-41