2
$\begingroup$

I am analyzing data from 17 studies (25 data points) with mixed measures of association (OR, RR, HR). Most associations are less than 2.0, with four from 2.0 to 3.1. Only one is less than 1.0 (not significant).

My options are fairly limited. Will consider doing stratified meta-analysis.

Are there some examples where the investigators did the meta-analysis using mixed-type measures of association?

Thanks for your help.

$\endgroup$
2
  • $\begingroup$ I think this has already been asked and answered here stats.stackexchange.com/questions/15897/… $\endgroup$ – mdewey Jan 19 '17 at 13:01
  • $\begingroup$ I saw that question and answer, and will probably end up taking one or more of those approaches. But does anyone know of a publication where OR, RR, HR were used and how it was justified? Thanks. $\endgroup$ – David Cowan Jan 23 '17 at 21:11
0
$\begingroup$

The approach I reccommend is to convert odds ratios into relative risks using the control group event rate as baseline, and then pool relative risks and hazard ratios together.

You may stratify the analysis to show consistency distinguishing relative risks and hazard ratios, though.

$\endgroup$

Not the answer you're looking for? Browse other questions tagged or ask your own question.