[How reproducible is blood pressure measurement? Intra- and interindividual blood pressure variability in routine practice].

Considering the reproducibility of blood pressure readings the epidemiologist and the physician in his clinical work take two different points of view. Studying drug effects the epidemiologist is interested in significant differences of blood pressures in a study population. The physician however, who estimates the effect of his prescription given to an individual patient has to compare single blood pressure readings in single individuals. In a pharmacological study we accept without any doubt the need of certain numbers of investigations to define a difference of 10 mm Hg as significant. In clinical routine however we behave quite differently and judge single readings intuitively as equal or different, independent of any statistical considerations. Not only the doctor but also the scientific societies with the highest reputation share the same point of view. In their recommendations they accept an arithmetic mean of two blood pressure readings as sufficient to judge blood pressures in individuals. In order to discuss this topic primarily from the doctors point, we analysed repeated blood pressure readings of 21 volunteers who tested 20 different blood pressure devices. We conclude, that the ongoing practice of judging the blood pressures of the patients meets only the needs of the epidemiologist. Reproducibility of single readings or mean values of two readings however are insufficient to draw satisfactory conclusions from the distributions of blood pressure in single individuals. Due to the short and long term blood pressure variability increasing the numbers of readings within a single day does not improve reproducibility sufficiently. It seems that the amount of blood pressure readings which are necessary to improve reproducibility of "the patients blood pressures" should be increased and taken over several days. If this is true, daily self-recordings should be the most promising approach.