- Email: [email protected]
Stroke subtype classification: A comparative study of ASCO and modified TOAST
Journal of the Neurological Sciences 314 (2012) 66–70
Contents lists available at SciVerse ScienceDirect
Journal of the Neurological Sciences journal homepage: www.elsevier.com/locate/jns
Stroke subtype classification: A comparative study of ASCO and modified TOAST Wan yu Shang, Jun yan Liu ⁎ Neurology Department, Third Hospital of Hebei Medical University, Shijiazhuang, Hebei, China
a r t i c l e
i n f o
Article history: Received 13 July 2011 Received in revised form 22 September 2011 Accepted 24 October 2011 Available online 25 November 2011 Keywords: Acute ischemic stroke ASCO Modified TOAST
a b s t r a c t Background and purpose: The ASCO stroke classification may be an improvement over the modified TOAST for etiological diagnoses. We aimed to compare the differences in stroke subtype classification between these two classification system. Methods: Selected for this study were 425 first-time acute ischemic stroke patients. For each, the cause of ischemic stroke was classified according to both the ASCO and modified TOAST criteria. The κ statistic and McNemar test were used to compare the similarities and differences, respectively, between the two approaches. Results: More patients were classified as having an atherosclerotic etiology under the ASCO 1 category than the modified TOAST system (60.2% vs. 57.9%; P = 0.132). There was no significant difference between the proportion of patients with undetermined etiology as defined by the ASCO 1 and the modified TOAST (15.5% vs. 16.2%; P = 0.795). Both the modified TOAST and ASCO-1 correctly identified all patients with etiology “other cause”. Agreement between the two classification systems was high in every subtype category except undetermined’ (κ > 0.81 for atherosclerosis, κ = 0.61 to κ = 0.8 for cardiac disease, and κ = 0.480 for undetermined). When ASCO-1 to ASCO 3 were applied, atherosclerosis was identified as the cause in 76.0% of patients, small artery disease in 46.4%, and cardiac disease in 11.3%. Conclusion: There is a moderately high agreement between the ASCO and modified TOAST classification schemes in all subtypes except that of “undetermined” etiology. Application of ASCO-1 did not reduce the proportion of patients ‘undetermined’ etiology compared to modified TOAST. Crown Copyright © 2011 Published by Elsevier B.V. All rights reserved.
1. Introduction A reliable and precise etiologic classification is highly important in the treatment and prognostication of ischemic stroke. The modified TOAST (Trial of ORG 10172 in Acute Stroke Treatment) classification system for determining stroke etiologies [1] improved upon the criteria of the initial TOAST system in terms of reliability and accuracy [2]. However, even the modified TOAST is based on “the most likely cause(s) of stroke”, but ignores coexisting etiologies. In addition, patients with more than two stroke causes are still classified as “undetermined etiology”. Thus a good proportion of patients fall under the undetermined category, even though the potential causes of stroke are identifiable. In the newer A–S–C–O (A for atherosclerosis, S for small vessel disease, C for cardiac source, O for other cause) Phenotypic System (ASCO) [3], every stroke patient's etiological pathology can be characterized, as well as acknowledgment made of the strength of the diagnostic evidence. Each of the four phenotypes (A, S, C, and O) is graded 1, 2, or 3, based on level of diagnostic certainty. In addition, the phenotype is given a grade of 0 if there is no evidence after thorough investigation, or a grade 9 if there was insufficient work-up to make a determination.
⁎ Corresponding author. Tel.: + 86 13313012890; fax: + 86 311 87023626. E-mail address: [email protected] (J. Liu).
It may be presumed that application of the ASCO classification would reduce the proportion of patients with undetermined etiology from that determined by the modified TOAST. A previous study which compared TOAST and ASCO showed that ASCO grade 1 did not reduce the proportion of patients with undetermined cause [4]. We wanted to further test, in a larger study of first-time acute ischemic stroke patients, the hypothesis that the ASCO would result in a smaller proportion of patients with undetermined etiology than the modified TOAST method. We also aimed to test the extent of agreement between each etiologic subtype classified under ASCO vs. the modified TOAST. 2. Methods 2.1. Patients We retrospectively analyzed data from 1375 ischemic stroke patients who were consecutively admitted to the Third Hospital of Hebei Medical University from 2007 to 2010. All the patients came from the Hebei region of China. Subsequently, we included only the first-time ischemic stroke patients from the 1375, based on the inclusion criteria. Demographic data and stroke risk factors such as age, gender, hypertension, diabetes mellitus, hyperlipidemia, current cigarette smoking, atrial fibrillation (AF), coronary artery disease (CAD), peripheral artery disease (PAD), history of transient ischemic stroke (TIA) and myocardial infarction (MI) were recorded.
0022-510X/$ – see front matter. Crown Copyright © 2011 Published by Elsevier B.V. All rights reserved. doi:10.1016/j.jns.2011.10.029
W. Shang, J. Liu / Journal of the Neurological Sciences 314 (2012) 66–70
67
Hypertension was diagnosed when patients presented with persistently elevated systolic blood pressure >140 mm Hg or diastolic >90 mm Hg 7 days after stroke onset, or the patient had a history of taking antihypertensive medication. Diabetes mellitus was defined as prescribed treatment of hypoglycemic agents, or two fasting hyperglycemia tests >7 mmol/L, or plasma glucose ≥11.1 mmol/L at 2 h into an oral glucose tolerance test. Hyperlipidemia was recorded for a patient with total cholesterol >5.0 mmol/L, or low-density lipoprotein >3.5 mmol/L, or was taking a pre-stroke cholesterol-lowering agent. Patients were considered current cigarette smokers if they currently smoked or had stopped smoking less than 7 months previous to the occurrence of stroke. TIA, MI, CAD, AF, and PAD were recorded from clinical history. We recorded the findings of routine tests including full blood count, biochemical screening, and coagulation. At least one brain imaging modality (computed tomography, magnetic resonance, or diffusionweighted) was obtained for each patient. Intracranial and extracranial arteries were imaged by either carotid duplex, transcranial Doppler ultrasound, magnetic resonance angiography or computed tomography angiography. We also recorded the electrocardiogram (ECG)/Holter findings. For patients younger than 45 years, suspected cardioembolism and stroke classification based on standard tests were difficult, and therefore special tests were carried out during hospitalization for better accuracy. Those tests included transthoracic echocardiography, digital subtraction angiography, serum levels of extractable nuclear antigen, antistreptolysin O titer, rheumatoid factor, C-reactive protein, erythrocyte sedimentation rate, serum levels of antineutrophil cytoplasmic antibody, anticardiolipin antibody, and tumor biomarkers.
possible due to insufficient work-up. There are also 3 scores for level of diagnostic evidence: A, for direct demonstration by gold-standard diagnostic tests or criteria; B, for indirect evidence or less sensitive or specific tests or criteria; and C, for weak evidence in the absence of specific tests or criteria. For example, a patient with a 90% ipsilateral symptomatic stenosis, no brain imaging, normal ECG and thrombocytosis (600,000/mm3) would be classified as A1-S9-C0-O3. Two experienced neurologists (JYL and BW) categorized all the patients using data that had been retrospectively collected. Review of the original brain images was done whenever necessary. In patients with dissonant results, both the authors re-checked the data together before asserting the final diagnosis. The ASCO and modified TOAST criteria were strictly applied.
2.2. Inclusion criteria
2.5. Statistical analyses
Patients who fulfilled the inclusion criteria were those in whom neurological deficits still existed 24 h after stroke onset and a documented acute infarct was seen on neuroimaging [5], or neurological deficits resolved within 24 h but an index stroke lesion was identified by brain imaging [6]. In addition, this was the enrollee's first-ever ischemic stroke and hospitalized within 15 days of stroke onset.
Categorical variables were calculated as percentages; continual variables as mean ± standard deviation (SD; x ± s) or median (interquartile range, IQR). The McNemar test was used to compare the probabilities of discordant categorization after a Bonferroni correction for multiple comparisons. Agreement between approaches was determined by the κ statistic, in which κ values were interpreted as moderate (0.41 to 0.6), good (0.61 to 0.8), or very good (0.81 to 1) agreement [4]. Independent proportions were compared using either the chi-square (χ 2) or Fisher's exact test, as appropriate. Multiplegroup comparisons of independent means were performed using one-way ANOVA or the Kruskal–Wallis H test; the Nemenyi test was used for multiple independent comparisons of means when necessary. P-value b 0.05 was considered statistically significant. SPSS version 13.0 was used for all statistical analyses.
2.3. Stroke subtype classification The modified TOAST system [2] classifies stroke etiology into 5 different categories: atherosclerosis (AT), small-artery disease (SAD), cardioembolism (CE), stroke of other determined etiology (SOD), and stroke of undetermined etiology (SUD). In addition, SUDs are further divided into the subcategories SUDm (two etiologies are identified), SUDu (no likely etiology was found despite extensive evaluations), and SUDi (the etiology of stroke cannot be determined due to incomplete evaluation). The modified TOAST criteria differ from that of the TOAST in that a relevant artery stenosis b50% with an unstable plaque is placed in the AT category. In addition, a single ischemic lesion occurring in a single perforating arterial territory without relevant large artery stenosis (determined by angiographic evaluation) is considered due to SAD. A patient with a small brain lesion (b2 cm) is classified as AT if a relevant and systemic arterial stenosis or occlusion is identified on angiography. However, the patient will be classified SUDu if no evidence of systemic atherosclerosis is found. Furthermore, classification with the modified TOAST rests on more test evidence than the older TOAST algorithm. In the ASCO classification [3], etiological phenotypes include atherosclerosis (A), small vessel disease (S), cardiac source (C), and other cause (O). Each phenotype is given a grade of 1 when the potential cause of the index stroke is definite, 2 when the causality is uncertain, 3 when the phenotype is present but the casualty is uncertain, 0 (zero) when the disease is completely absent, or 9 if grading is not
2.4. Methods for combined etiology of ASCO We grouped patients into ASCO categories based on the certainty of diagnosis. When any of the four phenotypes were graded as 1 (ASCO 1; cause of the index stroke is definite), irrespective of whether a full diagnostic evaluation was completed or not, these were grouped as “etiology identified”. The remainders were grouped as “undetermined etiology”, or ASCO X. The two groups were compared to the modified TOAST scheme of classification. Subsequently, ASCO 1 and ASCO 2 (cause is uncertain) were combined as an identified etiological group, a single domain. Similarly, any single domain that was not ASCO 1 or ASCO 2 (i.e., ASCO 0, 3, or 9) was recognized as a group of undetermined etiology and named ASCO Y. The comparison of the groups was then repeated as for the first comparison described above.
3. Results 3.1. Subject characteristics and risk factors Of 1375 stroke patients, 425 with first-time stroke were recruited after excluding 948 with TIA, chronic cerebral infarction, recurrent stroke, or ischemic stroke associated with subarachnoid or parenchymal hemorrhage. The median age of the study population was 65 ± 18.0 years. Sixty percent of the patients were male and were younger than females (61.4 ± 0.79 vs. 67.6 ± 0.83; P b 0.001). The prevalence of hypertension, atrial fibrillation and coronary artery disease was significantly higher in the female patients (77.6% vs. 62.4%, 14.1% vs. 5.5%, and 35.9% vs. 18.8%, respectively). Hypertension and hyperlipidemia were the most common risk factors in each subtype of the ASCO and modified TOAST system. Diabetes mellitus in AT patients was significantly higher compared to other etiologies under the modified TOAST (P = 0.009). Clinical characteristics and risk factors associated with the different phenotypes of the ASCO and modified TOAST classifications are summarized in Table 1.
68
W. Shang, J. Liu / Journal of the Neurological Sciences 314 (2012) 66–70
Table 1 Clinical characteristics and risk factors in different phenotypes of ASCO and modified TOAST classification systems. Modified TOAST
Age, x ± SD, (IQR) Male (%) Hypertension (%) Diabetes (%) Hyperlipidemia (%) CAD (%) MI (%) AF (%) TIA (%) PAD (%) Smoking (%)
ASCO
AT (n = 246)
SAD (n = 63)
CE (n = 37)
SOD (n = 10)
SUD (n = 69)
A1 (n = 256)
S1 (n = 57)
C1 (n = 36)
O1 (n = 10)
ASCO X (n = 66)
64.2 ± 11.7 63.0 69.5 39.0 47.6 21.5 1.2 0.8 5.3 2.0 44.3
59.7 ± 11.0 65.1 71.4 23.8 57.1 19.0 1.6 1.6 6.3 3.2 36.5
74.0, (19.0) 45.9 56.8 21.6 35.1 56.8 10.8 62.2 2.7 5.4 35.1
53.0 ± 19.0 50.0 40.0 0 20.0 20.0 0 0 0 0 20.0
64.3 ± 12.7 53.6 59.4 23.2 44.9 30.0 7.2 17.4 1.4 1.4 36.2
64.1 ± 11.9 62.9 68.8 37.5 49.6 23.0 2.3 1.6 5.5 2.0 44.1
57.9 ± 10.2 61.4 71.9 24.6 52.6 14.0 1.8 0 3.5 1.8 33.3
75.5 (8.0) 36.1 58.3 22.2 30.6 72.2 13.9 91.7 2.8 5.6 36.1
53.0 ± 19.0 50.0 40.0 0 20.0 20.0 0 0 0 0 20.0
63.8 ± 11.7 62.1 60.6 25.8 43.9 21.2 1.5 1.5 3.0 3.0 37.9
Total (n = 425)
65.0 (18.0) 60.0 66.4 31.8 46.8 25.6 3.1 8.9 4.5 2.4 40.5
TIA: transient ischemic attack; CAD: coronary artery disease; AF: atrial fibrillation; MI: myocardial infarction; PAD: peripheral arterial disease; ASCO X: patients with no ASCO 1 (definite) evidence in any single domain.
All 425 patients had at least one neuroimaging study: 61.9% had head CT scans, 68.5% had head T1, T2MRIs, DWI was performed in 81.4%, transcranial Doppler ultrasound was performed in 80.5%; magnetic resonance angiography in 76.0%, and carotid duplex ultrasound in 13.2%. Only 3 patients had digital subtraction angiography data. ECG/Holter monitoring was performed in 85.9% of the patients and transthoracic echocardiography in 14.4%. Biomarkers such as extractable nuclear antigen, antistreptolysin O titer, rheumatoid factor, and antineutrophil cytoplasmic antibody were performed in 44 patients (10.4%) with a suspected “other cause” subtype. 3.2. Frequencies of stroke subtype When ASCO phenotyping was applied, ASCO A1 was the most frequent subtype (256/425, 60.2%) followed by ASCO S1 (57/425, 13.4%), ASCO C1 (36/425, 8.5%) and ASCO O1 (10/425, 2.4%). Sixtysix patients were classified ASCO X (15.5%). The modified TOAST criteria resulted in placing AT as the most frequent subtype (57.9%, 246/425), followed by SUD, (69/425, 16.2%), SAD (63/425, 14.8%), CE (37/425, 8.7%) and SOD (10/425, 2.4%). Of the 69 SUD patients, 68.1%, 30.4%, and 1.4% were SUDi, SUDm, and SUDu, respectively.
of patients diagnosed with either small artery disease or cardioembolism as the primary cause were all lower in ASCO 1 than in the modified TOAST, but these difference did not meet the criterion for statistical significance (P = 0.238, P = 1.000, respectively). When the number of ASCO 1 and 2 (definite and likely causes, respectively) diagnoses were compared with the modified TOAST, there was a significant increase in the atherosclerotic and small artery disease subtypes (increases of 5.6% and 19.0%, respectively; P = 0.002). There was a significant decrease in cause undetermined patients (47.8%, P b 0.001), which remained statistically significant even after the Bonferroni correction. No statistical difference was observed between the two systems in diagnoses of cardioembolic etiology, and the same 10 patients were diagnosed in the other cause subtype by both ASCO and the modified TOAST. When comparing ASCO 1, or the combined ASCO 1 and 2, with the modified TOAST, agreement was observed in every corresponding subtype. Agreement for atherosclerotic and small artery disease etiology was very good (κ > 0.81, for all), and good for the cardiac disease subtype (κ values = 0.685 and 0.722, respectively). Agreement for cause undetermined was inferior to other subtypes (κ = 0.480), especially between ASCO Y and SUD (κ = 0.411). Differences and agreements of phenotypes between ASCO and the modified TOAST are shown in Table 2.
3.3. Differentiation and agreement of phenotypes of ASCO and modified TOAST 3.4. Additional information provided by ASCO When the classification of patients' stroke etiologies fell under ASCO 1 (i.e., the etiological pathology was definite), compared with the modified TOAST system more patients were diagnosed with atherosclerosis as the primary cause, but the difference was not significant (P = 0.132). Similarly, fewer patients were diagnosed as cause undetermined by ASCO, however difference between the two classification systems was not statistically significant (P = 0.795). The proportion
For 425 stroke patients, the combined ASCO grades 1 to 3 identified patients with potential atherosclerotic (76.0%), small artery disease (46.4%) and cardiac source (11.3%) stroke etiologies. In addition, patients with any unidentified cause due to incomplete evidence ranged from 6.4 to 18.6%. Additional information provided by ASCO is shown in Table 3.
Table 2 Differences and agreement (κ) between ASCO and modified TOAST phenotypes. Subtype
AT/ASCO A SAD/ASCO S CE/ASCO C SOD/ASCO O SUD/ASCO X/ASCO Y
Modified TOAST vs. ASCO 1
Modified TOAST vs. ASCO 1 and 2
Modified TOAST
Phenotype rate, % ASCO 1
ASCO 1 and 2
Δ
P
κ
Δ
P
κ
246/425, 57.9% 63/425, 14.8% 37/425, 8.7% 10/425, 2.4% 69/425, 16.2%
256/425, 60.2% 57/425, 13.4% 36/425, 8.5% 10/425, 2.4% 66/425, 15.5%
266/425, 62.6% 75/425, 17.6% 38/425, 8.9% 10/425, 2.4% 36/425, 8.5%
+ 4.1% − 9.5% − 2.7% 0 − 4.3%
0.132 0.238 1.000 –a 0.795
0.825 0.825 0.685 1 0.480
+ 5.6% + 19.0% + 2.7% 0 − 47.8%
0.002 0.002 1.000 –a b0.001
0.814 0.879 0.722 1 0.411
ASCO X: patients with no ASCO1 evidence in any single domain; ASCO Y: patients with no ASCO1 or ASCO2 evidence in any single domain; Δ: relative change in subtype assignment between classification systems. a The same ten patients were diagnosed by ASCO 1 and the modified TOAST, and no patients were additionally diagnosed as other causes when ASCO2 evidence was applied.
W. Shang, J. Liu / Journal of the Neurological Sciences 314 (2012) 66–70 Table 3 Frequency of different grades of ASCO phenotypes in the study population. ASCO phenotype
Disease present (ASCO 1, 2, 3)
Disease absent (ASCO 0)
Insufficient investigation (ASCO 9)
A % (n) S % (n) C % (n) O % (n)
76.0 (323/425) 46.4 (197/425) 11.3 (48/425) 2.6 (12/425)
17.4 37.2 61.6 81.3
6.4 (28/425) 16.5 (70/425) 18.6 (79/425) 15.3 (65/425)
(74/425) (158/425) (262/425) (348/425)
4. Discussion A stroke subtype classification should not only be accurate and reliable, but also precise in determining etiology. Under the modified TOAST criteria a good number of patients fall within the “undetermined cause” category. Our study, whose objective was to determine the amount of agreement between the modified TOAST and ASCO classifications, showed that the two systems are comparable except for the “undetermined” subtype. When ASCO 1 was compared with the modified TOAST, a reduction in the number of patients diagnosed as cause undetermined by ASCO X was observed, but the difference was not significant (P = 0.795). Furthermore, the strength of agreement was not as robust as in other subtypes (κ = 0.48). When ASCO grades 1 to 2 were compared with the modified TOAST, the agreement was even less robust (κ = 0.41). The SUD group included in the modified TOAST may be the possible explanation for these results. Sixtynine patients were designated as SUD by the modified TOAST, of which about one-third were assigned to the SUDm group when more than two causes were identified (30.4%, 21/69). Of the 21 patients who were classified as SUDm, the cause of stroke could be identified in all patients by ASCO 1 and 2 combined, and in 17 of the 21 by ASCO 1 alone. Thus the etiology of stroke can be better identified with the ASCO for those patients classified as SUDm under the modified TOAST criteria. The criteria for small artery disease differ between the two classification systems. Nevertheless, a very good agreement can be observed when comparing ASCO S1 or ASCO S1 to 2 with the modified TOAST (κ = 0.825 and 0.879, respectively). A small, deep infarction and a normal relevant artery are both essential for determining small artery disease by each system. The ASCO S1 criteria also emphasizes the presence of ≥1 additional signs associated with small artery disease, such as radiological evidence of leukoaraiosis, microbleeds, dilatation of the perivascular spaces, or lacunar infarcts in territories different from the index stroke. Therefore, the ASCO criteria are stricter for assignment of the small artery disease phenotype than the modified TOAST. As a result, fewer patients received the small artery disease designation ASCO S1 (Δ = −9.5%). ASCO S2 criteria include the presence of a single, deep branch artery stroke and the symptomatology of classic lacunar syndromes [3]. When we applied both ASCO S1 and S2 criteria, the proportion of patients with small artery disease increased significantly (Δ = + 19.0%, P = 0.002). Thus, the lack of specificity of ASCO S2 criteria for detecting small artery disease is unsatisfactory. When high-risk causes such as cardioembolism and atherosclerosis are present, it is reasonable to consider them as the possible etiology [2]. Therefore, stricter criteria for classifying small artery disease may be beneficial. The proportion of patients with cardioembolic causes was similar in the two systems, and good agreement was observed between ASCO C1, or ASCO C1 to 2, and the modified TOAST. The results imply that the criteria for cardioembolic causes of the two approaches are similar. However, both systems presume that the cause of stroke from a cardiac source is due to embolism, by arbitrarily purging the possibility of any other cause. Therefore, possible causes of stroke will be lost if identified solely by the modified TOAST [7], and the incidence of cardioembolism will be magnified by ASCO. Such assumptions may
69
even lead to inappropriate therapy [8]. In addition to considering embolism as the cause of a cardiac-related etiology, evaluations of heart function such as diastolic function are indispensable for a complete cardioembolic stroke phenotyping [9]. When ASCO A1 was compared with the modified TOAST, it was observed that the proportion of patients diagnosed with atherosclerotic causes was higher, but the difference did not reach statistical significance (p = 0.132), and excellent agreement was observed. However, the definition of atherosclerotic stenosis is different in each classification system. Atherosclerotic stenosis >50% is the accepted criterion in the modified TOAST, but a stenosis >70% is required by ASCO; the presence of an unstable plaque is however a criterion in both the classification systems, regardless of the stenosis grade. Therefore, an excellent agreement was observed when ASCO A1 was compared with the modified TOAST in our study. The number of patients falling within the subtype “other causes” was the same in both classification systems. Patients in this subcategory were so few that no change was observed when either ASCO O1 or ASCO O1 to 2 were compared with SOD. The additional information provided by ASCO suggests that many patients have more than two categories of potential stroke causes. Atherosclerosis was the most common potential cause (76.0%). Therefore, complete therapy for all coexisting causes should be taken seriously to reduce the risk of recurrence. Our study had several limitations. First, it is a single-center study from a single region. However, the distribution of stroke subtypes in our study was similar to a previous study conducted in China [10]. Therefore, it is likely that our study population is representative. Secondly, none of the patients underwent transesophageal echocardiography. Therefore, the frequency of cardioembolic causes might be underestimated. In addition, digital subtraction angiography was performed in very few patients, which might have led to underdetection of arterial dissection. However, other advanced imaging such as magnetic resonance angiography was used to avoid such an error [11]. 5. Conclusion Moderate to good agreement was found in every subtype category between the ASCO and improved TOAST classification systems. The characteristics of each system were useful in delineating etiology. The ASCO system was more accurate in identifying the etiology from available information, ASCO system provided a better phenotypic classification of stroke. However, most etiological diagnoses in both the classification systems are still based on clinical information and are thus presumptive. Acknowledgment We thank Medjaden Bioscience Limited for assisting in the preparation of this manuscript. References [1] Adams HP, Bendixen BH, Kappelle LJ, Biller J, Love BB, Gordon DL, et al. Classification of subtype of acute ischemic stroke definitions for use in a multicenter clinical trial. Stroke 1993;24:35–41. [2] Han SW, Kim SH, Lee JY, Chu CK, Yang JH, Shin HY, et al. A new subtype classification of ischemic stroke based on treatment and etiologic. Eur Neurol 2007;57: 96–102. [3] Amarenco P, Bogousslavsky J, Caplan LR, Donnan GA, Hennerici MG. New approach to stroke subtyping: the A–S–C–O (phenotypic) classification of stroke. Cerebrovasc Dis 2009;27:502–8. [4] Marnane M, Duggan CA, Sheehan OC, Merwick A, Hannon N, Curtin D, CrispinoO'Connell G, Kelly PJ, et al. Stroke subtype classification to mechanism-specific and undetermined categories by TOAST, A–S–C–O, and causative classification system. Stroke 2010;41:1579–86. [5] Bonita R, Stewart A, Beaglehole R. International trends in stroke mortality: 1970–1985. Stroke 1990;21:989–99.
70
W. Shang, J. Liu / Journal of the Neurological Sciences 314 (2012) 66–70
[6] Albers GW, Calpan LR, Easton JD, Fayad PB, Mohr JP, Saver JL, et al. Tansent ischemic attack proposal for a new definition. N Engl J Med November 21, 2002; Vol. 347(No. 21). [7] Landau WM, Nassisf A. Editorial comment—time to burn the TOAST. Stroke 2005;36:902–4. [8] Amarenco P. Patent foramen ovale and the risk of stroke: smoking gun guilty by association? Heart 2005;91:441–3.
[9] McAreavey D, Vidal JS, Aspelund T, Owens DS, Hughes T, Garcia M, et al. Correlation of echocardiographic findings with cerebral infarction in elderly adults: the AGES-Reykjavik study. Stroke 2010;41(10):2223–8. [10] Ma Y, Zhao X, Zhang W, Liu L, Wang Y, Fang R, et al. Homocysteine and ischemic stroke subtype: a relationship study in Chinese patients. Neurol Res 2010;32(6):636–41. [11] Oelerich M, Stögbauer F, Kurlemann G, Schul C, Schuierer G. Craniocervical artery dissection: MR imaging and MR angiographic findings. Eur Radiol 1999;9(7):1385–91.
Contents lists available at SciVerse ScienceDirect
Journal of the Neurological Sciences journal homepage: www.elsevier.com/locate/jns
Stroke subtype classification: A comparative study of ASCO and modified TOAST Wan yu Shang, Jun yan Liu ⁎ Neurology Department, Third Hospital of Hebei Medical University, Shijiazhuang, Hebei, China
a r t i c l e
i n f o
Article history: Received 13 July 2011 Received in revised form 22 September 2011 Accepted 24 October 2011 Available online 25 November 2011 Keywords: Acute ischemic stroke ASCO Modified TOAST
a b s t r a c t Background and purpose: The ASCO stroke classification may be an improvement over the modified TOAST for etiological diagnoses. We aimed to compare the differences in stroke subtype classification between these two classification system. Methods: Selected for this study were 425 first-time acute ischemic stroke patients. For each, the cause of ischemic stroke was classified according to both the ASCO and modified TOAST criteria. The κ statistic and McNemar test were used to compare the similarities and differences, respectively, between the two approaches. Results: More patients were classified as having an atherosclerotic etiology under the ASCO 1 category than the modified TOAST system (60.2% vs. 57.9%; P = 0.132). There was no significant difference between the proportion of patients with undetermined etiology as defined by the ASCO 1 and the modified TOAST (15.5% vs. 16.2%; P = 0.795). Both the modified TOAST and ASCO-1 correctly identified all patients with etiology “other cause”. Agreement between the two classification systems was high in every subtype category except undetermined’ (κ > 0.81 for atherosclerosis, κ = 0.61 to κ = 0.8 for cardiac disease, and κ = 0.480 for undetermined). When ASCO-1 to ASCO 3 were applied, atherosclerosis was identified as the cause in 76.0% of patients, small artery disease in 46.4%, and cardiac disease in 11.3%. Conclusion: There is a moderately high agreement between the ASCO and modified TOAST classification schemes in all subtypes except that of “undetermined” etiology. Application of ASCO-1 did not reduce the proportion of patients ‘undetermined’ etiology compared to modified TOAST. Crown Copyright © 2011 Published by Elsevier B.V. All rights reserved.
1. Introduction A reliable and precise etiologic classification is highly important in the treatment and prognostication of ischemic stroke. The modified TOAST (Trial of ORG 10172 in Acute Stroke Treatment) classification system for determining stroke etiologies [1] improved upon the criteria of the initial TOAST system in terms of reliability and accuracy [2]. However, even the modified TOAST is based on “the most likely cause(s) of stroke”, but ignores coexisting etiologies. In addition, patients with more than two stroke causes are still classified as “undetermined etiology”. Thus a good proportion of patients fall under the undetermined category, even though the potential causes of stroke are identifiable. In the newer A–S–C–O (A for atherosclerosis, S for small vessel disease, C for cardiac source, O for other cause) Phenotypic System (ASCO) [3], every stroke patient's etiological pathology can be characterized, as well as acknowledgment made of the strength of the diagnostic evidence. Each of the four phenotypes (A, S, C, and O) is graded 1, 2, or 3, based on level of diagnostic certainty. In addition, the phenotype is given a grade of 0 if there is no evidence after thorough investigation, or a grade 9 if there was insufficient work-up to make a determination.
⁎ Corresponding author. Tel.: + 86 13313012890; fax: + 86 311 87023626. E-mail address: [email protected] (J. Liu).
It may be presumed that application of the ASCO classification would reduce the proportion of patients with undetermined etiology from that determined by the modified TOAST. A previous study which compared TOAST and ASCO showed that ASCO grade 1 did not reduce the proportion of patients with undetermined cause [4]. We wanted to further test, in a larger study of first-time acute ischemic stroke patients, the hypothesis that the ASCO would result in a smaller proportion of patients with undetermined etiology than the modified TOAST method. We also aimed to test the extent of agreement between each etiologic subtype classified under ASCO vs. the modified TOAST. 2. Methods 2.1. Patients We retrospectively analyzed data from 1375 ischemic stroke patients who were consecutively admitted to the Third Hospital of Hebei Medical University from 2007 to 2010. All the patients came from the Hebei region of China. Subsequently, we included only the first-time ischemic stroke patients from the 1375, based on the inclusion criteria. Demographic data and stroke risk factors such as age, gender, hypertension, diabetes mellitus, hyperlipidemia, current cigarette smoking, atrial fibrillation (AF), coronary artery disease (CAD), peripheral artery disease (PAD), history of transient ischemic stroke (TIA) and myocardial infarction (MI) were recorded.
0022-510X/$ – see front matter. Crown Copyright © 2011 Published by Elsevier B.V. All rights reserved. doi:10.1016/j.jns.2011.10.029
W. Shang, J. Liu / Journal of the Neurological Sciences 314 (2012) 66–70
67
Hypertension was diagnosed when patients presented with persistently elevated systolic blood pressure >140 mm Hg or diastolic >90 mm Hg 7 days after stroke onset, or the patient had a history of taking antihypertensive medication. Diabetes mellitus was defined as prescribed treatment of hypoglycemic agents, or two fasting hyperglycemia tests >7 mmol/L, or plasma glucose ≥11.1 mmol/L at 2 h into an oral glucose tolerance test. Hyperlipidemia was recorded for a patient with total cholesterol >5.0 mmol/L, or low-density lipoprotein >3.5 mmol/L, or was taking a pre-stroke cholesterol-lowering agent. Patients were considered current cigarette smokers if they currently smoked or had stopped smoking less than 7 months previous to the occurrence of stroke. TIA, MI, CAD, AF, and PAD were recorded from clinical history. We recorded the findings of routine tests including full blood count, biochemical screening, and coagulation. At least one brain imaging modality (computed tomography, magnetic resonance, or diffusionweighted) was obtained for each patient. Intracranial and extracranial arteries were imaged by either carotid duplex, transcranial Doppler ultrasound, magnetic resonance angiography or computed tomography angiography. We also recorded the electrocardiogram (ECG)/Holter findings. For patients younger than 45 years, suspected cardioembolism and stroke classification based on standard tests were difficult, and therefore special tests were carried out during hospitalization for better accuracy. Those tests included transthoracic echocardiography, digital subtraction angiography, serum levels of extractable nuclear antigen, antistreptolysin O titer, rheumatoid factor, C-reactive protein, erythrocyte sedimentation rate, serum levels of antineutrophil cytoplasmic antibody, anticardiolipin antibody, and tumor biomarkers.
possible due to insufficient work-up. There are also 3 scores for level of diagnostic evidence: A, for direct demonstration by gold-standard diagnostic tests or criteria; B, for indirect evidence or less sensitive or specific tests or criteria; and C, for weak evidence in the absence of specific tests or criteria. For example, a patient with a 90% ipsilateral symptomatic stenosis, no brain imaging, normal ECG and thrombocytosis (600,000/mm3) would be classified as A1-S9-C0-O3. Two experienced neurologists (JYL and BW) categorized all the patients using data that had been retrospectively collected. Review of the original brain images was done whenever necessary. In patients with dissonant results, both the authors re-checked the data together before asserting the final diagnosis. The ASCO and modified TOAST criteria were strictly applied.
2.2. Inclusion criteria
2.5. Statistical analyses
Patients who fulfilled the inclusion criteria were those in whom neurological deficits still existed 24 h after stroke onset and a documented acute infarct was seen on neuroimaging [5], or neurological deficits resolved within 24 h but an index stroke lesion was identified by brain imaging [6]. In addition, this was the enrollee's first-ever ischemic stroke and hospitalized within 15 days of stroke onset.
Categorical variables were calculated as percentages; continual variables as mean ± standard deviation (SD; x ± s) or median (interquartile range, IQR). The McNemar test was used to compare the probabilities of discordant categorization after a Bonferroni correction for multiple comparisons. Agreement between approaches was determined by the κ statistic, in which κ values were interpreted as moderate (0.41 to 0.6), good (0.61 to 0.8), or very good (0.81 to 1) agreement [4]. Independent proportions were compared using either the chi-square (χ 2) or Fisher's exact test, as appropriate. Multiplegroup comparisons of independent means were performed using one-way ANOVA or the Kruskal–Wallis H test; the Nemenyi test was used for multiple independent comparisons of means when necessary. P-value b 0.05 was considered statistically significant. SPSS version 13.0 was used for all statistical analyses.
2.3. Stroke subtype classification The modified TOAST system [2] classifies stroke etiology into 5 different categories: atherosclerosis (AT), small-artery disease (SAD), cardioembolism (CE), stroke of other determined etiology (SOD), and stroke of undetermined etiology (SUD). In addition, SUDs are further divided into the subcategories SUDm (two etiologies are identified), SUDu (no likely etiology was found despite extensive evaluations), and SUDi (the etiology of stroke cannot be determined due to incomplete evaluation). The modified TOAST criteria differ from that of the TOAST in that a relevant artery stenosis b50% with an unstable plaque is placed in the AT category. In addition, a single ischemic lesion occurring in a single perforating arterial territory without relevant large artery stenosis (determined by angiographic evaluation) is considered due to SAD. A patient with a small brain lesion (b2 cm) is classified as AT if a relevant and systemic arterial stenosis or occlusion is identified on angiography. However, the patient will be classified SUDu if no evidence of systemic atherosclerosis is found. Furthermore, classification with the modified TOAST rests on more test evidence than the older TOAST algorithm. In the ASCO classification [3], etiological phenotypes include atherosclerosis (A), small vessel disease (S), cardiac source (C), and other cause (O). Each phenotype is given a grade of 1 when the potential cause of the index stroke is definite, 2 when the causality is uncertain, 3 when the phenotype is present but the casualty is uncertain, 0 (zero) when the disease is completely absent, or 9 if grading is not
2.4. Methods for combined etiology of ASCO We grouped patients into ASCO categories based on the certainty of diagnosis. When any of the four phenotypes were graded as 1 (ASCO 1; cause of the index stroke is definite), irrespective of whether a full diagnostic evaluation was completed or not, these were grouped as “etiology identified”. The remainders were grouped as “undetermined etiology”, or ASCO X. The two groups were compared to the modified TOAST scheme of classification. Subsequently, ASCO 1 and ASCO 2 (cause is uncertain) were combined as an identified etiological group, a single domain. Similarly, any single domain that was not ASCO 1 or ASCO 2 (i.e., ASCO 0, 3, or 9) was recognized as a group of undetermined etiology and named ASCO Y. The comparison of the groups was then repeated as for the first comparison described above.
3. Results 3.1. Subject characteristics and risk factors Of 1375 stroke patients, 425 with first-time stroke were recruited after excluding 948 with TIA, chronic cerebral infarction, recurrent stroke, or ischemic stroke associated with subarachnoid or parenchymal hemorrhage. The median age of the study population was 65 ± 18.0 years. Sixty percent of the patients were male and were younger than females (61.4 ± 0.79 vs. 67.6 ± 0.83; P b 0.001). The prevalence of hypertension, atrial fibrillation and coronary artery disease was significantly higher in the female patients (77.6% vs. 62.4%, 14.1% vs. 5.5%, and 35.9% vs. 18.8%, respectively). Hypertension and hyperlipidemia were the most common risk factors in each subtype of the ASCO and modified TOAST system. Diabetes mellitus in AT patients was significantly higher compared to other etiologies under the modified TOAST (P = 0.009). Clinical characteristics and risk factors associated with the different phenotypes of the ASCO and modified TOAST classifications are summarized in Table 1.
68
W. Shang, J. Liu / Journal of the Neurological Sciences 314 (2012) 66–70
Table 1 Clinical characteristics and risk factors in different phenotypes of ASCO and modified TOAST classification systems. Modified TOAST
Age, x ± SD, (IQR) Male (%) Hypertension (%) Diabetes (%) Hyperlipidemia (%) CAD (%) MI (%) AF (%) TIA (%) PAD (%) Smoking (%)
ASCO
AT (n = 246)
SAD (n = 63)
CE (n = 37)
SOD (n = 10)
SUD (n = 69)
A1 (n = 256)
S1 (n = 57)
C1 (n = 36)
O1 (n = 10)
ASCO X (n = 66)
64.2 ± 11.7 63.0 69.5 39.0 47.6 21.5 1.2 0.8 5.3 2.0 44.3
59.7 ± 11.0 65.1 71.4 23.8 57.1 19.0 1.6 1.6 6.3 3.2 36.5
74.0, (19.0) 45.9 56.8 21.6 35.1 56.8 10.8 62.2 2.7 5.4 35.1
53.0 ± 19.0 50.0 40.0 0 20.0 20.0 0 0 0 0 20.0
64.3 ± 12.7 53.6 59.4 23.2 44.9 30.0 7.2 17.4 1.4 1.4 36.2
64.1 ± 11.9 62.9 68.8 37.5 49.6 23.0 2.3 1.6 5.5 2.0 44.1
57.9 ± 10.2 61.4 71.9 24.6 52.6 14.0 1.8 0 3.5 1.8 33.3
75.5 (8.0) 36.1 58.3 22.2 30.6 72.2 13.9 91.7 2.8 5.6 36.1
53.0 ± 19.0 50.0 40.0 0 20.0 20.0 0 0 0 0 20.0
63.8 ± 11.7 62.1 60.6 25.8 43.9 21.2 1.5 1.5 3.0 3.0 37.9
Total (n = 425)
65.0 (18.0) 60.0 66.4 31.8 46.8 25.6 3.1 8.9 4.5 2.4 40.5
TIA: transient ischemic attack; CAD: coronary artery disease; AF: atrial fibrillation; MI: myocardial infarction; PAD: peripheral arterial disease; ASCO X: patients with no ASCO 1 (definite) evidence in any single domain.
All 425 patients had at least one neuroimaging study: 61.9% had head CT scans, 68.5% had head T1, T2MRIs, DWI was performed in 81.4%, transcranial Doppler ultrasound was performed in 80.5%; magnetic resonance angiography in 76.0%, and carotid duplex ultrasound in 13.2%. Only 3 patients had digital subtraction angiography data. ECG/Holter monitoring was performed in 85.9% of the patients and transthoracic echocardiography in 14.4%. Biomarkers such as extractable nuclear antigen, antistreptolysin O titer, rheumatoid factor, and antineutrophil cytoplasmic antibody were performed in 44 patients (10.4%) with a suspected “other cause” subtype. 3.2. Frequencies of stroke subtype When ASCO phenotyping was applied, ASCO A1 was the most frequent subtype (256/425, 60.2%) followed by ASCO S1 (57/425, 13.4%), ASCO C1 (36/425, 8.5%) and ASCO O1 (10/425, 2.4%). Sixtysix patients were classified ASCO X (15.5%). The modified TOAST criteria resulted in placing AT as the most frequent subtype (57.9%, 246/425), followed by SUD, (69/425, 16.2%), SAD (63/425, 14.8%), CE (37/425, 8.7%) and SOD (10/425, 2.4%). Of the 69 SUD patients, 68.1%, 30.4%, and 1.4% were SUDi, SUDm, and SUDu, respectively.
of patients diagnosed with either small artery disease or cardioembolism as the primary cause were all lower in ASCO 1 than in the modified TOAST, but these difference did not meet the criterion for statistical significance (P = 0.238, P = 1.000, respectively). When the number of ASCO 1 and 2 (definite and likely causes, respectively) diagnoses were compared with the modified TOAST, there was a significant increase in the atherosclerotic and small artery disease subtypes (increases of 5.6% and 19.0%, respectively; P = 0.002). There was a significant decrease in cause undetermined patients (47.8%, P b 0.001), which remained statistically significant even after the Bonferroni correction. No statistical difference was observed between the two systems in diagnoses of cardioembolic etiology, and the same 10 patients were diagnosed in the other cause subtype by both ASCO and the modified TOAST. When comparing ASCO 1, or the combined ASCO 1 and 2, with the modified TOAST, agreement was observed in every corresponding subtype. Agreement for atherosclerotic and small artery disease etiology was very good (κ > 0.81, for all), and good for the cardiac disease subtype (κ values = 0.685 and 0.722, respectively). Agreement for cause undetermined was inferior to other subtypes (κ = 0.480), especially between ASCO Y and SUD (κ = 0.411). Differences and agreements of phenotypes between ASCO and the modified TOAST are shown in Table 2.
3.3. Differentiation and agreement of phenotypes of ASCO and modified TOAST 3.4. Additional information provided by ASCO When the classification of patients' stroke etiologies fell under ASCO 1 (i.e., the etiological pathology was definite), compared with the modified TOAST system more patients were diagnosed with atherosclerosis as the primary cause, but the difference was not significant (P = 0.132). Similarly, fewer patients were diagnosed as cause undetermined by ASCO, however difference between the two classification systems was not statistically significant (P = 0.795). The proportion
For 425 stroke patients, the combined ASCO grades 1 to 3 identified patients with potential atherosclerotic (76.0%), small artery disease (46.4%) and cardiac source (11.3%) stroke etiologies. In addition, patients with any unidentified cause due to incomplete evidence ranged from 6.4 to 18.6%. Additional information provided by ASCO is shown in Table 3.
Table 2 Differences and agreement (κ) between ASCO and modified TOAST phenotypes. Subtype
AT/ASCO A SAD/ASCO S CE/ASCO C SOD/ASCO O SUD/ASCO X/ASCO Y
Modified TOAST vs. ASCO 1
Modified TOAST vs. ASCO 1 and 2
Modified TOAST
Phenotype rate, % ASCO 1
ASCO 1 and 2
Δ
P
κ
Δ
P
κ
246/425, 57.9% 63/425, 14.8% 37/425, 8.7% 10/425, 2.4% 69/425, 16.2%
256/425, 60.2% 57/425, 13.4% 36/425, 8.5% 10/425, 2.4% 66/425, 15.5%
266/425, 62.6% 75/425, 17.6% 38/425, 8.9% 10/425, 2.4% 36/425, 8.5%
+ 4.1% − 9.5% − 2.7% 0 − 4.3%
0.132 0.238 1.000 –a 0.795
0.825 0.825 0.685 1 0.480
+ 5.6% + 19.0% + 2.7% 0 − 47.8%
0.002 0.002 1.000 –a b0.001
0.814 0.879 0.722 1 0.411
ASCO X: patients with no ASCO1 evidence in any single domain; ASCO Y: patients with no ASCO1 or ASCO2 evidence in any single domain; Δ: relative change in subtype assignment between classification systems. a The same ten patients were diagnosed by ASCO 1 and the modified TOAST, and no patients were additionally diagnosed as other causes when ASCO2 evidence was applied.
W. Shang, J. Liu / Journal of the Neurological Sciences 314 (2012) 66–70 Table 3 Frequency of different grades of ASCO phenotypes in the study population. ASCO phenotype
Disease present (ASCO 1, 2, 3)
Disease absent (ASCO 0)
Insufficient investigation (ASCO 9)
A % (n) S % (n) C % (n) O % (n)
76.0 (323/425) 46.4 (197/425) 11.3 (48/425) 2.6 (12/425)
17.4 37.2 61.6 81.3
6.4 (28/425) 16.5 (70/425) 18.6 (79/425) 15.3 (65/425)
(74/425) (158/425) (262/425) (348/425)
4. Discussion A stroke subtype classification should not only be accurate and reliable, but also precise in determining etiology. Under the modified TOAST criteria a good number of patients fall within the “undetermined cause” category. Our study, whose objective was to determine the amount of agreement between the modified TOAST and ASCO classifications, showed that the two systems are comparable except for the “undetermined” subtype. When ASCO 1 was compared with the modified TOAST, a reduction in the number of patients diagnosed as cause undetermined by ASCO X was observed, but the difference was not significant (P = 0.795). Furthermore, the strength of agreement was not as robust as in other subtypes (κ = 0.48). When ASCO grades 1 to 2 were compared with the modified TOAST, the agreement was even less robust (κ = 0.41). The SUD group included in the modified TOAST may be the possible explanation for these results. Sixtynine patients were designated as SUD by the modified TOAST, of which about one-third were assigned to the SUDm group when more than two causes were identified (30.4%, 21/69). Of the 21 patients who were classified as SUDm, the cause of stroke could be identified in all patients by ASCO 1 and 2 combined, and in 17 of the 21 by ASCO 1 alone. Thus the etiology of stroke can be better identified with the ASCO for those patients classified as SUDm under the modified TOAST criteria. The criteria for small artery disease differ between the two classification systems. Nevertheless, a very good agreement can be observed when comparing ASCO S1 or ASCO S1 to 2 with the modified TOAST (κ = 0.825 and 0.879, respectively). A small, deep infarction and a normal relevant artery are both essential for determining small artery disease by each system. The ASCO S1 criteria also emphasizes the presence of ≥1 additional signs associated with small artery disease, such as radiological evidence of leukoaraiosis, microbleeds, dilatation of the perivascular spaces, or lacunar infarcts in territories different from the index stroke. Therefore, the ASCO criteria are stricter for assignment of the small artery disease phenotype than the modified TOAST. As a result, fewer patients received the small artery disease designation ASCO S1 (Δ = −9.5%). ASCO S2 criteria include the presence of a single, deep branch artery stroke and the symptomatology of classic lacunar syndromes [3]. When we applied both ASCO S1 and S2 criteria, the proportion of patients with small artery disease increased significantly (Δ = + 19.0%, P = 0.002). Thus, the lack of specificity of ASCO S2 criteria for detecting small artery disease is unsatisfactory. When high-risk causes such as cardioembolism and atherosclerosis are present, it is reasonable to consider them as the possible etiology [2]. Therefore, stricter criteria for classifying small artery disease may be beneficial. The proportion of patients with cardioembolic causes was similar in the two systems, and good agreement was observed between ASCO C1, or ASCO C1 to 2, and the modified TOAST. The results imply that the criteria for cardioembolic causes of the two approaches are similar. However, both systems presume that the cause of stroke from a cardiac source is due to embolism, by arbitrarily purging the possibility of any other cause. Therefore, possible causes of stroke will be lost if identified solely by the modified TOAST [7], and the incidence of cardioembolism will be magnified by ASCO. Such assumptions may
69
even lead to inappropriate therapy [8]. In addition to considering embolism as the cause of a cardiac-related etiology, evaluations of heart function such as diastolic function are indispensable for a complete cardioembolic stroke phenotyping [9]. When ASCO A1 was compared with the modified TOAST, it was observed that the proportion of patients diagnosed with atherosclerotic causes was higher, but the difference did not reach statistical significance (p = 0.132), and excellent agreement was observed. However, the definition of atherosclerotic stenosis is different in each classification system. Atherosclerotic stenosis >50% is the accepted criterion in the modified TOAST, but a stenosis >70% is required by ASCO; the presence of an unstable plaque is however a criterion in both the classification systems, regardless of the stenosis grade. Therefore, an excellent agreement was observed when ASCO A1 was compared with the modified TOAST in our study. The number of patients falling within the subtype “other causes” was the same in both classification systems. Patients in this subcategory were so few that no change was observed when either ASCO O1 or ASCO O1 to 2 were compared with SOD. The additional information provided by ASCO suggests that many patients have more than two categories of potential stroke causes. Atherosclerosis was the most common potential cause (76.0%). Therefore, complete therapy for all coexisting causes should be taken seriously to reduce the risk of recurrence. Our study had several limitations. First, it is a single-center study from a single region. However, the distribution of stroke subtypes in our study was similar to a previous study conducted in China [10]. Therefore, it is likely that our study population is representative. Secondly, none of the patients underwent transesophageal echocardiography. Therefore, the frequency of cardioembolic causes might be underestimated. In addition, digital subtraction angiography was performed in very few patients, which might have led to underdetection of arterial dissection. However, other advanced imaging such as magnetic resonance angiography was used to avoid such an error [11]. 5. Conclusion Moderate to good agreement was found in every subtype category between the ASCO and improved TOAST classification systems. The characteristics of each system were useful in delineating etiology. The ASCO system was more accurate in identifying the etiology from available information, ASCO system provided a better phenotypic classification of stroke. However, most etiological diagnoses in both the classification systems are still based on clinical information and are thus presumptive. Acknowledgment We thank Medjaden Bioscience Limited for assisting in the preparation of this manuscript. References [1] Adams HP, Bendixen BH, Kappelle LJ, Biller J, Love BB, Gordon DL, et al. Classification of subtype of acute ischemic stroke definitions for use in a multicenter clinical trial. Stroke 1993;24:35–41. [2] Han SW, Kim SH, Lee JY, Chu CK, Yang JH, Shin HY, et al. A new subtype classification of ischemic stroke based on treatment and etiologic. Eur Neurol 2007;57: 96–102. [3] Amarenco P, Bogousslavsky J, Caplan LR, Donnan GA, Hennerici MG. New approach to stroke subtyping: the A–S–C–O (phenotypic) classification of stroke. Cerebrovasc Dis 2009;27:502–8. [4] Marnane M, Duggan CA, Sheehan OC, Merwick A, Hannon N, Curtin D, CrispinoO'Connell G, Kelly PJ, et al. Stroke subtype classification to mechanism-specific and undetermined categories by TOAST, A–S–C–O, and causative classification system. Stroke 2010;41:1579–86. [5] Bonita R, Stewart A, Beaglehole R. International trends in stroke mortality: 1970–1985. Stroke 1990;21:989–99.
70
W. Shang, J. Liu / Journal of the Neurological Sciences 314 (2012) 66–70
[6] Albers GW, Calpan LR, Easton JD, Fayad PB, Mohr JP, Saver JL, et al. Tansent ischemic attack proposal for a new definition. N Engl J Med November 21, 2002; Vol. 347(No. 21). [7] Landau WM, Nassisf A. Editorial comment—time to burn the TOAST. Stroke 2005;36:902–4. [8] Amarenco P. Patent foramen ovale and the risk of stroke: smoking gun guilty by association? Heart 2005;91:441–3.
[9] McAreavey D, Vidal JS, Aspelund T, Owens DS, Hughes T, Garcia M, et al. Correlation of echocardiographic findings with cerebral infarction in elderly adults: the AGES-Reykjavik study. Stroke 2010;41(10):2223–8. [10] Ma Y, Zhao X, Zhang W, Liu L, Wang Y, Fang R, et al. Homocysteine and ischemic stroke subtype: a relationship study in Chinese patients. Neurol Res 2010;32(6):636–41. [11] Oelerich M, Stögbauer F, Kurlemann G, Schul C, Schuierer G. Craniocervical artery dissection: MR imaging and MR angiographic findings. Eur Radiol 1999;9(7):1385–91.