Impact of using actual versus arbitrary awake and sleep times on analyses of 24-hour ambulatory blood pressure.