relative risks, odds ratio, and hazard ratio in meta analysis.

[Fred]

Dear all,

For those three risk measurements, what we actually need in a meta
analysis?

As odds ratio and hazard ratio are the approximation to the relative
risks, but they could be adjusted in multi-variable settings.

When conducting a meta analysis, for the same disease and exposure, if
publications report those three, also their adjusted values, then what
we need in the final meta analysis?

Thanks and Regards

Fred

[Jeremy Miles]

I think you need the same one from each study, whichever you decide is
the most appropriate. (Depending on the study, you might also have
risk difference).

Jeremy

> --
> To post a new thread to MedStats, send email to MedS...@googlegroups.com .
> MedStats' home page is http://groups.google.com/group/MedStats .
> Rules: http://groups.google.com/group/MedStats/web/medstats-rules
>

--
Jeremy Miles
Psychology Research Methods Wiki: www.researchmethodsinpsychology.com

[Fred]

Thanks for the reply Jeremy.

Odds ratio could be calculate from 2X2 papers, also from multiple
logistic regression by adjusting other confounders.

If single publication reports both unadjusted and adjusted, which one
is of interest to be included in the meta analysis ??

Also odds ratio with odds ratio, relative risk with relative risk,
hazard ratio with hazard ratio, can we combine them together?

For multiple logistic regression, and Cox proportional model, each
publication may adjust difference factors or confounders, say in the
renal failure study, the proteinuria might be the risk factor cause
the renal failure. This factor maybe included in the multiple logistic
regression in all the publication, for example, but they adjusted
other factors differently. Are those odds ratio can be used in the
meta analysis??

On Aug 26, 11:44 am, Jeremy Miles <jeremy.mi...@gmail.com> wrote:
> I think you need the same one from each study, whichever you decide is
> the most appropriate.   (Depending on the study, you might also have
> risk difference).
>
> Jeremy
>

[Fred]

Sorry, 2X2 table.

[Sreenivas.V]

You can take any of the three, but the one which leads to least heterogeniety among the different studies being meta-analysed is preferable.  Also risk difference is one very important measure, as most often one can have significant results in terms of risk ratios but not so with risk differences.  Risk difference being more important clinically, the conclusions etc will be entirely different.

 

Sreenivas

 

 

 

--
To post a new thread to MedStats, send email to MedS...@googlegroups.com .

MedStats' home page is http://groups.google.com/group/MedStats .
Rules: http://groups.google.com/group/MedStats/web/medstats-rules

--
V. Sreenivas
Associate Professor
Department of Biostatistics
All India Institute of Medical Sciences
New Delhi 110029

[Pedro Emmanuel Alvarenga Americano do Brasil]

Hi stat masters,

If I got the question right, I believe the option to choose one measure or other is a exercise of considering which one is appropriate to your scenario. The interpretation ability of your result's audience should be considered too. As far as I know, functions to meta-analyze these measures (in Stata or R for example)usually  require data from 2x2 table. Thus, with the same data you may be able to try out some or all these measures.

Be careful whit the adjusted and crude effects differences of the exposure of interest in different investigations if you are dealing with observational studies.

Kind regards,

Abraço forte e que a força esteja com você,

Dr. Pedro Emmanuel A. A. do Brasil
Instituto de Pesquisa Clínica Evandro Chagas
Fundação Oswaldo Cruz
Rio de Janeiro - Brasil
Av. Brasil 4365
Tel 55 21 3865-9648
email: pedro....@ipec.fiocruz.br
email: emmanue...@gmail.com

---Apoio aos softwares livres
www.zotero.org - gerenciamento de referências bibliográficas.
www.broffice.org ou www.openoffice.org - textos, planilhas ou apresentações.
www.epidata.dk - entrada de dados.
www.r-project.org - análise de dados.
www.ubuntu.com - sistema operacional

2010/8/26 Sreenivas.V <sreevish...@gmail.com>

[Steve Simon, P.Mean Consulting]

There is no consensus in the research community about how best to
summarize results that could use an odds ratio or a relative risk (or
risk difference/number needed to treat) as the statistical measurement.

One important consideration is heterogeneity. Since the relative risk
and risk difference have restrictions on range that change depending on
the proportion improved in the control group, there is a strong chance
for heterogeneity if there is heterogeneity in the severity of illness
(which would, of course, affect the proportion improved in the control
group). This is less of a problem, potentially, for the odds ratio. I
believe there is some empirical evidence to suggest that the odds ratio
is less likely to suffer heterogeneity, but I can't remember where I saw
this.

There is general consensus, though, that if you want to summarize data
using a hazard ratio, you need access to the individual patient level
data, as the pattern of censoring and other features of the data will
have an influence over what your pooled estimate might be.

I don't think there is any precedent to pool estimates from studies
where some studies report a hazard ratio and other studies report an
odds ratio or relative risk. While there is a conceptual relationship
between these measures, they are still different enough that you should
probably report separate pooled estimates for the studies using a hazard
ratio and the studies using odds ratios/relative risks.
--
Steve Simon, Standard Disclaimer
Sign up for The Monthly Mean, the newsletter that
dares to call itself "average" at www.pmean.com/news

[Frank Harrell]

A couple of late comments -

One could say that the RR is an approximation to the odds ratio.

The odds ratio is the only measure that is capable of being constant
over the entire range of background risk.

Frank

On Aug 26, 10:21 am, "Steve Simon, P.Mean Consulting" <n...@pmean.com>
wrote:

[RogerH]

Or one could view the risk ratio and the odds ratio as approximations
to the hazard ratio or rate ratio. Rates and hazards can exceed 1,
unlike risks, so there's no constraint on the hazard ratio, unlike the
risk ratio. Hazard ratios / rate ratios can therefore be constant over
the entire range of baseline hazard / background rate.

Roger.

[Frank Harrell]

Yes I should have restricted my comment to binary outcomes.
Frank