Abstract
An incorrect measure of outcome in a clinical trial may invalidate the entire study, no matter how well-designed the trial is in other respects. In general, outcome can be measured at four different levels. Level 1: the biological activity of the disease (e.g. lactate level, ipsilateral stroke after carotid endarterectomy, aneurysmal rebleeding after antifibrinolytic agents); level 2: clinical manifestations (e.g. degree of weakness or aphasia, which are included in the so-called ''stroke scales''); level 3: activities of daily living (disability scales, e.g. the Barthel Scale), and level 4: the restriction of the life-style (handicap scales, e.g. the Glasgow Outcome Scale or Rankin Scale) or even the quality of life. Level 1 is easy to measure, but is not clinically relevant, whereas level 4 is highly relevant, but difficult to measure. Levels 1 and 2 are suitable only for preliminary studies which aim to identify a biological effect; they are unsuitable for studies where the result will be applied in clinical practice. In such cases, the measurement must be based on the outcome from the patient''s own perspective, which implies the use of disability scales, or preferably handicap scales. In addition, death (or at least vascular death) should always be included in the analysis. Finally, the statistical methods used should be nonparametric, as the scales represent classes and not true numbers.