Application of cusums to ambulatory blood pressure data: a simple statistical technique for detecting trends over time.

Ambulatory blood pressure measurement monitoring has become an increasingly important method of investigation in hypertension [ I ] and automated non-invasive devices which allow frequent measurements of ambulatory blood pressure and the pulse rate over 24 h are now available. The accuracy [ 2 ] and the reproducibility [3] of ambulatory measurements have been documented.'However, data from ambulatory blood pressure monitoring are characterized by wide scatter, due to random variabiiity, short-term fluctuations with posture, physical and mental activity and long-term variability with seasonal [4], dietary [5,6] and hormonal [7] changes. Superimposed upon the distribution of data there is also the influence of diurnal blood pressure variation. It is difficult to detect early trends when data collected at regular time intervals show wide scatter. The calculation of cumulative sums ('cusums') is a simple statistical techluque which allows early and precise detection of trends in data of this nature. Although described in the medical literature over a decade ago [8,9], this technique is still uncommon in clinical medicine. It is particularly appropriate for the analysis of data derived from 24 h ambulatory blood pressure monitoring. The cusums technique consists of the selection of an arbitrary reference value, such as the mean of daytime blood pressure, which is then subtracted from each point in succession. The successive daiations of each data point from the reference value are then added cumulatively, i.e. the first to the second, the sum of these to the third and so on. The 'cumulative sums' derived in this manner are then plotted against time with the ambulatory blood pressure data. The reference value chosen is typically the mean of an initial series of obsetvations. However, the overall mean o r any other clinically relevant reference point may b e used.