Background and Purpose-In hypertensive stroke patients, for the same level of blood pressure control, eprosartan will be more effective than nitrendipine in reducing cerebrovascular and cardiovascular morbidity and mortality. Methods-A total of 1405 well-defined, high-risk hypertensives with cerebral event during the last 24 months (proven by cerebral computed tomography scan or nuclear magnetic resonance) were randomized to eprosartan or nitrendipine (mean follow-up 2.5 years). Primary end point was the composite of total mortality and all cardiovascular and cerebrovascular events, including all recurrent events. Results-Randomization was successful without significant differences in the baseline characteristics. Blood pressure was reduced to a comparable extent without any significant differences between the 2 groups during the whole study period (150.

Response to Letter by Larsen et al Response:It is correct that the stated protective effect of Eprosartan over Nitrendipine concerns the composite primary end points. As pointed out in the statistical section of the article, we summed up the number of primary end points and the observation time in each group to be able to calculate an event rate per group. However, because of limited space in the article, we could not state in detail that before we computed any statistical test, we ascertained that event frequencies were not different between the groups by comparing the number of persons with 0, 1, 2 or more end points. Because no heterogeneity was detected, we still believe that the test used to compare the ratio of the 2 event rates and our conclusions are correct.Larsen et al argue that it would have strengthened the studyconclusion if the design had been double-blind. This question was addressed already in the original article. 1 We believe that the PROBE design 2 might have been the key to get comparable blood pressure values in both treatment groups, which is the most important aspect to strengthen any conclusion of therapeutic effects in hypertension studies (like the MOSES study). Joachim Schrader, MDStephan

Since the time of Hippocrates patients and their physicians have been highly interested in knowing whether, and with what degree of probability "undesirable events" will occur in the course of illness--whether without, in spite of, or because of therapy. In surgery especially, the concept of "risk" is a familiar one to all concerned. The present paper describes traditional and modern methods of quantifying risk and the goals of medical risk research. Application of the results of such research in practice is discussed and illustrated by examples from publications and the authors' own clinical research.