Application of computer tomography-oriented criteria for stroke subtype classification in a prospective study

PURPOSE: To apply stroke subclassification criteria based on computer tomography (CT) to strokes in the Nurses' Health Study and to assess reliability and validity of the criteria. METHODS: Among 121,701 women aged 30-55 years at entry, subclassification criteria were applied to 1369 incident strokes which occurred from 1976 to 1994. Reproducibility of the subclassification criteria was assessed in a systematic sample of 100 strokes reviewed by two independent reviewers. As a validation, relative risks (RR) fur stroke subtypes were examined in a prospective analysis of 112,282 women aged 34 to 59 years, free of cardiovascular disease and cancer in 1980, with follow-up through 1994. RESULTS: Strokes were subclassified as follows: 226 subarachnoid hemorrhages, 135 intraparenchymal hemorrhages, 103 embolic, 217 large-artery occlusive, 309 lacunar, and 97 other thrombotic infarctions, as well as 282 strokes of undetermined type. No intercoder discrepancies were found in classification of subarachnoid hemorrhage, intraparenchymal hemorrhage, or embolic infarction, whereas there were four discrepancies relative to subclassification of thrombotic infarction, mostly due to inconsistent documentation of CT findings. In multivariate models, associations of older age (> 60 years), current smoking, history of diabetes, and high cholesterol with stroke subtypes were consistent with previous epidemiologic, clinical, or pathologic findings. CONCLUSIONS: Stroke diagnostic criteria based primarily on CT were found to have a high rate of reliability and validity. These stroke subclassification criteria may be useful in examining whether associations for various lifestyle and health behaviors differ with stroke subtypes. (C) 2000 Elsevier Science Inc. All rights reserved.