White I R, Chaturvedi N, McKeigue P M
Department of Epidemiology and Public Health, University College London, U.K.
Stat Med. 1994 Aug 30;13(16):1635-41. doi: 10.1002/sim.4780131604.
We consider the analysis of epidemiological studies in which the outcome of interest is a physiological disturbance (such as blood pressure) which may have been corrected by medical intervention (such as antihypertensive treatment). It is important to take account of such intervention in studying the determinants of hypertension. We recommend using the median as a measure of location and assuming that the underlying values of subjects under intervention lie above the median. Some well-established exact methods for simple problems are presented. An example is given in which the blood pressure differences between Afro-Caribbeans and Europeans are underestimated by about one-quarter if adjustment is not made for treated hypertensives.
