Comparison of conventional and modern methods in determination of etiology of ischemic stroke

Bacground and Purpose: Ischemic stroke is a heterogenous disease and it is the crucial issue to make a rapid and accurate assessment of the etiology not only for the treatments and interventions in acute phase, but also to settle a strategy for secondary prevention. Diagnostic investigations performed with high technology reveal that there is more than one causal explanation in almost half of the patients with ischemic stroke. Conventional classification systems differentiate several subgroups of ischemic stroke without any suggestion of a predominant etiological factor. Recently designed web-based, semi-automated systems allow a decision making algorithm to assign the most likely causative mechanism. The aim of this sudy is to establish the inter-rater reliability in determination of subtypes of ischemic stroke; using the conventional method TOAST (Trial of ORG 10172 in Acute Stroke Treatment) and the new method CCS (Causative Classification of Stroke) system. Two stroke neurologists, and two general neurologists participated. It has been assumed that, CCS would demonstrate higher inter-rater agreement in addition to lower subgroup of undetermined cases, compared to TOAST classification. Methods: Fifty ischemic stroke patients admitted consecutively to stroke unit at University Hospital of Dokuz Eylül University, included for the study. Suptype determination has been performed by study investigators using both TOAST and CCS methods, and this diagnosis has been accepted as ?gold standard?. Two stroke neurologists, and two general neurologists took part in the study to assess independently the same patients first according to TOAST, then to CCS system, through the records of patients containing detailed history, laboratory and radiological findings. Before assessment according to CCS, all raters were required to be certified by the web-based system (http://ccs.martinos.org). After the collection of all data, statistical analyses was performed using ?SPSS 15.0 for Windows?. Agreement between investigators was measured by kappa statistic. Values were interpreted as moderate (0?41?0?6), good (0?61?0?8), or excellent (0?81?1). Results: There was good agreement between two stroke neurologists and gold standard using TOAST system. Kappa values were 0.77 (%95 CI 0.60?0.94) and 0.67 (%95 CI 0.52?0.83) respectively. Same raters performed excellent agreement using CCS for 5 subtypes, Kappa values were 0.83 (%95 CI 0.68?0.99) and 0.86 (%95 CI 0.71?1.02) respectively. The agreement between two general neurologists and gold standard, using TOAST system was good. Kappa values were 0.76 (%95 CI 0.60?0.91) and 0.78 (%95 CI 0.61?0.94) respectively. According to CCS, while one rater performed excellent agreement for 5 subtypes, kappa 0.89 (%95 CI 0.73?1.05), the other performed good agreement. Kappa value was 0.70 (%95 CI 0.55?0.86). The inter-rater agreement of two stroke neurologists was good both by using TOAST and CCS. Kappa values were 0.61 (%95 CI 0.45?0.77) and 0.78 (%95 CI 0.62?0.94) respectively. The inter-rater agreement of two general neurologists was good both by using TOAST and CCS. Kappa values were 0.64 (%95 CI 0.48?0.80) and 0.75 (%95 CI 0.60?0.91) respectively. Finally, all four raters performed moderate agreement using TOAST sysytem. Kappa value was 0.59 (%95 CI 0.52?0.65). The agreement was good for CCS and kappa value was 0.75 (%95 CI 0.68?0.81). Five raters (including gold standard) performed 500 etiologic stroke subtype designations. There were 73 cases (29.2%) of undetermined group assigned by TOAST system. This category was decresed to 54 cases (21.6%) when used CCS. Conclusions: CCS improved inter-rater agreement compared to TOAST system, both for stroke neurologists and general neurologists. The proportion of undetermined group was decreased (7.6%) when used CCS. Keywords: Acut ischemic stroke, etiologic classification, CCS, TOAST