Systolic blood pressure (SBP) and pulse pressure (PP) have been identified in western industrialized countries as major predictors of cardiovascular events in the elderly on the basis of measurements taken at a single visit. Considering the wide variability of blood pressure (BP) in older people, this study set out to assess the prognostic significance of measurements of SBP and PP taken over several months according to a monitoring scheme mimicking routine care. A total of 444 Italian general practitioners enrolled a cohort of 3858 unselected elderly outpatients and followed them up for 10 years. BP was recorded at recruitment, 1 week later and at quarterly visits during the first year. The average BP of these six visits was used to define the patient's BP status. During the 10-year follow-up, 1561 participants died, 709 from cardiovascular diseases. Proportional hazard regression analysis, adjusted for all main prognostic factors including antihypertensive treatment, showed that for each 10-mmHg increment in SBP and PP there were, respectively, 5 and 9% increases in risk for total mortality (TM) and 9 and 13% increases in risk for cardiovascular mortality (CVM) (all P < 0.01). However, including both SBP and PP in the model, only PP showed an independent, significant relationship with TM and CVM. In conclusion, prognostic information based on repeated measurements of PP is stronger than that given by SBP and consequently should be recommended in the definition of cardiovascular risk in the elderly.