We have tested the concept that fewer patients are needed in trials of antihypertensive treatment if blood pressure is measured by ambulatory monitoring rather than by conventional sphygmomanometry. 233 patients (> or = 60 years old) with isolated systolic hypertension were randomly allocated placebo (n = 119) or active treatment (n = 114). Blood pressure measurements were compared by Wilcoxon's test and blood pressure profiles by ANOVA. With either method of measurement, the same number of patients (40 in each treatment group) was required to show a reduction after 1 year in clinic (13/8 mm Hg) or average blood pressure over 24 h (9/5 mm Hg). To detect that the decrease in systolic pressure was not steadily maintained through the day, 40 patients in each treatment group were needed for blood pressure profiles made up of 4-hourly or 2-hourly means and 60 for profiles of 1-hourly means. For diastolic pressure, the corresponding numbers were 80, 100, and more than the number of available patients, respectively. We conclude that parallel-group trials focusing on the average blood pressure over 24 h, rather than on conventionally measured blood pressure, cannot economise on sample size. Moreover, trials studying the full course of blood pressure throughout the day, require more--not fewer--patients than studies of only the conventional or average 24 h blood pressure.