We enrolled 808 older patients with isolated systolic hypertension (160 to 219/71 <95 mm Hg) to investigate whether ambulatory measurement of pulse pressure and mean pressure can refine risk stratification. The patients (> or =60 years) were randomized to nitrendipine (10 to 40 mg/day) with the possible addition of enalapril (5 to 20 mg/day) or hydrochlorothiazide (12.5 to 25 mg/day) or to matching placebos. At baseline, pulse pressure and mean pressure were determined from six conventional blood pressure (BP) readings and from 24-h ambulatory recordings. With adjustment for significant covariables, we computed mutually adjusted relative hazard rates associated with 10 mm Hg increases in pulse pressure or mean pressure. In the placebo group, the 24-h and nighttime pulse pressures consistently predicted total and cardiovascular mortality, all cardiovascular events, stroke, and cardiac events. Daytime pulse pressure predicted cardiovascular mortality, all cardiovascular end points, and stroke. The hazard rates for 10 mm Hg increases in pulse pressure ranged from 1.25 to 1.68. Conventionally measured pulse pressure predicted only cardiovascular mortality with a hazard rate of 1.35. In the active treatment group compared with the placebo patients, the relation between outcome and ambulatory pulse pressure was attenuated to a nonsignificant level. Mean pressure determined from ambulatory or conventional BP measurements was not associated with poorer prognosis. In conclusion, in older patients with isolated systolic hypertension higher pulse pressure estimated by 24-h ambulatory monitoring was a better predictor of adverse outcomes than conventional pulse pressure, whereas conventional and ambulatory mean pressures were not correlated with a worse outcome.