Clinical trials in hypertension: some general thoughts and some particular controversies

A common fallacy in the analysis of clinical trials is described. Correlating the differences between pairs of random numbers and the first of them produces a correlation coefficient of 0.71 with a regression equation of y = -4.5 + X. Misuses of correlation and regression in hypertension and hypolipidaemic trials are described and advice given. The author also puts on record his objection to the pedantry that surrounds the use of the analysis by intentions-to-treat method. Patients who go into such an analysis may include those who never received a single dose of the treatment to which they were randomized, those who started it but were later switched to the other treatment though still analysed in their original group, patients who could be regarded as having died from obviously nontreatment causes by any sensible and reasonable doctor, and patients later found even to have suffered the wrong illness.