st: Meta-epidemiological analysis (RORs)

[Spyros Papageorgiou <snpapage@gmail.com>]

Dear All,

I would appreciate any comment on the following:

I want to calculate ratios of odds ratios (ROR) for my meta-analyses
cohort with random-effects meta-regression as proposed by Sterne et al
(2002) in Statist Med. I am using Stata 10.0 for Windows.

I think I ‘ve worked it out relative good with the metareg command
(with K-H modification) and tried to compare the output by the results
of manually calculating fixed the RORs, but I am not 100% sure.

I am having two kinds of binary only data (some meta-analyses give the
first, some the second): (i) raw 2x2 data and (ii) calculated ORs.

(i) For the raw data I use:
metan exn exsample ctn ctsample, or log random by(design) nograph
metareg _ES design, wsse(_seES)

with „exn, exsample, ctn, ctsample“ being the events and sample fort
he two groups, „design“ being the trial characteristic I assess and
_ES, _seES being the automatic Stata inputs for effect size and
standard error

(ii) For the calculated ORs I use:

metareg logor design, wsse(selogor)

with the log odds ratios and their standard errors respectively.

I then get the logRORs and their selogROR that are given in the
metareg output from all meta-analyses and pool them using.

metan logror selogror,random eform rfdist to get the pooled RORs with
their 95% CIs and 95% PIs in the forest plot.

Do you think this approach is ok?

Thanks in advance

Spyros
snpapage@gmail.com

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/