Limited reproducibility of hourly blood pressure values obtained by ambulatory blood pressure monitoring: implications for studies on antihypertensive drugs