meta analysis and forest plot for hazard ratio

[didem okmen]

Hi everyone,

I'm new in STAtA and I'm studying for patient survival or disease progression is association with TRIM29 expression in enrolled studies and There is only 2 table in the paper in the below link. I coud'nt find another data set about this. There is a lot of forest plot for HR. I don't know how HR is calculated and forest plot. What it's commands? Can you help me for this?
Thank you.

[David Fisher]

Hi Didem,

Are you asking for help with understanding the meaning and derivation of a hazard ratio? Or are you asking how to produce a forest plot such as those in the paper?

If the latter, then the command used by the authors of the paper is undoubtedly "metan". It is a user-written program which must be downloaded and installed. This can be done by typing "ssc install metan" at the Stata command line. How you proceed from there depends on how your data are stored and what options you want ... there is documentation which can be accessed by typing "help metan" at the Stata command line.

One other thing: "metan" is the most popular Stata command for meta-analysis, but it is no longer maintained and the last update was 8 years ago. As part of a larger suite of meta-analysis programs (focussing particularly on individual patient data (IPD) meta-analysis), I have written an alternative command called "admetan". It can be installed by typing "ssc install ipdmetan" (NOTE: *not* ssc install admetan!) at the Stata command line. Again, documentation can be accessed by typing "help admetan" once installed.

Best wishes,

David.

[didem okmen]

Hi David,
Actually, latter I asked you. Do you have an example for meta analysis and HR.

I coudn't find correct example. And as you tell, I installed this command.

[David Fisher]

Hi Didem,

OK. Let's work with Figure 2C of the TRIM29 paper, as it's the smallest.

First of all, we need to enter the data into Stata. The code below is one way of doing it:

* Example generated by -dataex-. To install: ssc install dataex
clear
input str14 study float(hr lci uci)
"Harris 2015a"   .024 .003 .17
"Harris 2015b"   .154 .027 .909
"Fristrup 2013a"   .6  .39  .93
"Fristrup 2013b"  .79  .51 1.25
end

Next, since a hazard ratio is measured on the ratio (exponential) scale, we need to take logarithms before combining in a meta-analysis:

gen double loghr = ln(hr)
gen double loglci = ln(lci)
gen double loguci = ln(uci)

Finally, we can run the meta-analysis and (re-) produce the forest plot:

admetan loghr loglci loguci, hr re study(study) forestplot(xlabel(.01 .1 1 10))

Note that the weights are not exactly the same as in Figure 2C, probably because the authors had access to more accurate data for the hazard ratios.

Best wishes,

David.

[didem okmen]

Thank you David, I applied the commands.
I found your paper about "Two-stage individual participant data meta-analysis and generalized forest plots", but I need a lot of source codes. Because I have to apply Sensitivitiy analysis of effect of individual studies on the pooled HRs. (Meta analysis random effects estimates (exponential form)). My punishment never ends

[Dathan Byonanebye]

Hi David, This is very helpful, is there a way i can suppress the Overall on the forest plot from not showing. I only want to show a forest plot of individual studies. Can i also add the refernce covariate and set its HR as 1? Cofidence intervals 1 to 1?

Thanks

Dathan Byonanebye