The cost of dichotomising continuous variables

Grateful thanks to Professors Altman and Royston for such a clear
exposition of the reasons for treating continuous variables as such. Might
they have gone a little further and suggested that symmetrical distribution
about the median suggests that one is dealing with association within a
single population and not two or more populations with different
relationship with the the independent variable, therefore dichtomisation
is absolutely wrong in these circumstances?

Back to Peart and Pickering and whether what was called then benign
essential hypertension should be considered a disease?

Competing interests:
none