🚀 go-pugleaf

I have been asked to advise on required sample size for a study into
multifactor risks of stroke.

In a particular group of patients (those with carotid stenosis), there
is an elevated risk of stroke (which is known). It is not sufficient to
justify surgery becuase the surgery itself has a significant risk
(which is also known). There are numerous documented additional risk
factors within this group, with various prevelances and risks that have
been estimated. None of them alone give really convincing indication
for surgery.  (or at least those that do have low sensitivity, so a lot
of stroke cases will not be detected).

It is postulated that some combination of them might yield a good
enough sensitivity and specificity to make surgery a beneficial
option.

What sort of power calculation could I do in this case to determine a
sensible size for a cohort study ?

Please e-mail me and I will summarize to the list.

Thanks very much in advance,

Luke Whitaker, St Mary's Hospital Medical School, London.
(l.whitaker@sm.ic.ac.uk or the address on this post)