Can I input my own odds ratio value and run metan?

[Christine S]

Hi all, I am running a meta analysis and though I did need to use metan to obtain weights for each of my 5 studies, I do not want to use the odds ratio values it calculated since I have calculated values that account for confounding variables. Is it possible to run metan using odds ratios that differ from what is calculated? Thanks so much, Christine

[David Fisher]

Hi Christine,

Are you looking to perform a meta-analysis of adjusted odds ratios, which you estimated yourself using the original data? If so, my program ipdmetan might be worth a look (ssc describe ipdmetan). It will pool the adjusted odds ratios to give an estimate different to the one metan would give you, and which may be the result you desire (depending on your problem!).

Otherwise, do you wish to use an alternative pooled odds ratio to the one metan gives you? If so, metan can accommodate this via the first(...) option, which is described in the help file. If you can't find it, check that you have the latest version of metan installed.

If you basically just want to create a forestplot that contains your own values (i.e. not derived from metan), my program forestplot might help (included in the ipdmetan package, ssc describe ipdmetan). It's a bit fiddly, though, so feel free to ask for further assistance in this thread if you think it will be useful.

Thanks,

David.

[Christine S]

David, I cannot thank you enough. That is exactly what I want to do - perform a meta-analysis of adjusted odds-ratios which I have already from the original data. I installed ipdmetan and will also use forestplot. My question now is how to form the correct command? I included my ORs, and lower and upper CIs in my dataset in Excel and imported this into stata. Then I used the command ipdmetan c1 no1 c2 no2, or xlab(.1,1,10) which is similar to how I ran metan the first time. I am not really sure how to incorporate my ORs and CIs in ipdmetan. Any help would be so gratefully appreciated. I uploaded a couple screenshots for reference. Sincerely, Christine

[David Fisher]

Hi Christine,

Thanks for the screenshots! OK I think I see what you want to do. You can pool the adjusted odds ratios very easily using metan as follows:

gen lnor = ln(or)
gen lnlci = ln(lci)
gen lnuci = ln(uci)
metan lnor lnlci lnuci, eform effect(OR)

However, if you estimated those adjusted odds ratios yourself using raw (i.e. patient-level) data, then you can use ipdmetan to do the whole thing in one go, as follows:

ipdmetan, study(studyname) or : logistic outcome treatment adjustment_vars

where "studyname", "outcome" and "treatment" are existing variables in your patient-level dataset, and "adjustment_vars" is a list of existing variables.

David.

[Christine S]

David,

ipdmetan worked great. Thank you again for all of your help!

Christine

[Hatty Wong]

Hi there,
I found this post during my search metan and was hoping I may get some help. I'm quite new to STATA and similar to Christine's situation, I collected information from approximately 30 articles but will need to perform separate meta-analyses as they are combination of different topic.

Some of the data only provided me with OR and I don't have CI. I'm curious as to how I should first go about setting this up in Excel in order to have the data properly imported into STATA.

Thank you!

[Rose Harrington]

Hello,

I used this code suggested above to estimate an overall effect measure from 19 studies as well as the I-square to test heterogeneity:

gen lnor = ln(or) gen lnlci = ln(lci) gen lnuci = ln(uci) metan lnor lnlci lnuci, eform effect(OR) The code worked perfectly - thank you for posting.

I would like to also use the commands metabias, metafunnel, and metatrim. From what I have read about these commands, _ES and _seES are needed, however when I run the metan code above these variables are not generated. Could you please help explain how I can run these commands? Thank you in advance for your time.

Rose

[Rami AlRifai]

Hello all,

It would be greatly appreciated if you could help on the following:

I am conducting now a meta-analysis on specific research question, it was easy for me to get a pooled crude estimate using the "metan" syntax. However, for some studies I do not have the frequency of subjects for the 2X2 table, but I do have the crude relative risk stratified by age and sex:

Age 20-<30: 1.32 (95%CI 1.12-1.91)

Age 30-<40: 2.39 (95%CI 1.92-2.89)

Age 40-<50: 0.82 (95%CI 0.72-1.3)

Age 50-<60: 3.23 (95%CI 2.89-4.22)

Male: 1.7 (95%CI 1.44-2.3)
Female: 4.2 (95%CI 3.45-6.07)

My question is: How can I get the overall RR adjusted for age and sex for the above data??

I used the above code, but it did not work, please help if you can.

[Rami AlRifai]

Hello all,

It would be greatly appreciated if you could help on the following:

I am conducting now a meta-analysis on specific research question, it was easy for me to get a pooled crude estimate using the "metan" syntax. However, for some studies I do not have the frequency of subjects for the 2X2 table for each strata, but I do have the crude relative risk stratified by age and sex:

Age 20-<30: 1.32 (95%CI 1.12-1.91)

Age 30-<40: 2.39 (95%CI 1.92-2.89)

Age 40-<50: 0.82 (95%CI 0.72-1.3)

Age 50-<60: 3.23 (95%CI 2.89-4.22)

Male: 1.7 (95%CI 1.44-2.3)
Female: 4.2 (95%CI 3.45-6.07)

My question is: How can I get the overall adjusted RR for age and sex using the this only available data??

I used the above code, but it did not work, please help if you can.