2
$\begingroup$

I've looked both in SO and others (Borenshtein's) and have yet to find an answer. I have some studies that provide an adjusted OR estimate, some HR, and then some with adjusted RR. I can subgroup and report them based on treatment effect, but I am also interested in bundling them. My colleague suggested treating RR as HR and then converting to OR to RR (formula available in the cochrane handbook - a solution I'm not really satisfied with, since RR != HR, and that also requires me to extrapolate baseline risk from univariate data. In addition I'll end up describing with RR although almost all studies use a different measure.

However I couldn't find an appropriate method/ function for HR -> OR.

Can someone point me in the direction? This describes converting log-hazard to probability and then back, but going the other way around again I don't have baseline hazard.

$\endgroup$
  • $\begingroup$ I think you already have there a manageable approach: convert OR to RR using control event rate, and then assume RR are a reasonable approximation of HR. I am not aware of any other sounder approach. $\endgroup$ – Joe_74 May 19 '16 at 7:52
  • $\begingroup$ I thought it was desirable to describe OR. Besides, things get weird when one approximating HR as RR as face value, e.g- getting confidence interval beyond 100% risk. $\endgroup$ – Ozeuss May 19 '16 at 14:41
  • $\begingroup$ Possible duplicate of stats.stackexchange.com/questions/130237/… $\endgroup$ – mdewey Aug 18 '18 at 16:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.