The independent prognostic significance of circadian blood pressure (BP) changes is unclear. We investigated the association between circadian BP changes and cardiovascular risk among elderly-treated hypertensive patients. The occurrence of a composite end point (that is, stroke, coronary events, heart failure and peripheral revascularization) was evaluated among 1191 elderly-treated hypertensive patients (age range 60-90 years). According to the nighttime change and the morning surge (MS) of systolic BP, subjects were divided into groups of dippers with a normal or high MS (DNMS and DHMS, respectively), non-dippers (ND), reverse dippers (RD) and extreme dippers with a normal or high MS (EDNMS and EDHMS, respectively). During the follow-up (9.1±4.9 years, range 0.4-20 years), 392 events occurred. The event rate was 3.63 per 100 patient-years. After adjustment for various covariates, including 24-h BP, the DHMS (hazard ratio (HR) 1.49, 95% confidence interval (CI) 1.02-2.16, P=0.04), ND (HR 1.71, 95% CI 1.28-2.27, P=0.0001), RD (HR 2.05, 95% CI 1.44-2.93, P=0.0001) and EDHMS (HR 3.40, 95% CI 1.96-5.90, P=0.001) were at higher cardiovascular risk than the DNMS. The population attributable risk was 0.6, 7.1, 7.3 and 1.4% for the DHMS, ND, RD and EDHMS, respectively. In elderly-treated hypertensive patients, circadian BP changes were independently associated with increased cardiovascular risk. At the patient level, the highest risk was observed among the EDHMS, followed by the RD, ND and DHMS. At the population level, the highest risk was observed among the RD, followed by the ND, EDHMS and DHMS.