Odds ratio table with forest plot

[ritupuri]

Hello,

I saw this image(attached) in the following journal:

Chromosomal Instability Portends Superior Response of Rectal Adenocarcinoma to Chemoradiation Therapy
Cancer 120(11) · June 2014.

It was performed on XLSTAT. Is it possible to do such table/image in Stata ?

Thank you,
Ritu

[Marcos Almeida]

I recommend to try to fiddle with the user-written - metan - or other SSC related to meta-analysis in Stata.

[David Fisher]

Hi Ritu,

It certainly is. There are a range of meta-analysis/forest plot commands that can be used for the purpose.

If you simply have the summary data in a similar format to that in your attached images, then you can use metan or admetan (both with the nooverall option as you are not actually pooling anything). If you wish to have greater control over the look of the plot, try forestplot. Note that both admetan and forestplot are part of the ipdmetan package (ssc describe ipdmetan), whilst metan can be downloaded separately (ssc describe metan).

If you have the original trial data ("individual participant data", IPD) rather than summary data, you could use ipdover to calculate (and plot) the odds ratios "on the fly" using logistic regression.

Best wishes,

David.


(P.S. Disclaimer: I am the author and maintainer of ipdmetan, admetan, forestplot and ipdover. metan is the traditional Stata meta-analysis command, but is now a little old and poorly-maintained. However, it does have various functionalities that my programs do not -- yet! -- especially regarding count data, as here.)

[ritupuri]

Hello Marcos and David,
I have downloaded Metan and ipmetan .
I will work on the recommendations.
Its amazing , how helpful the forum is.
I thank you all,
Ritu

[Bruce Weaver]

Hi Ritu. It looks like you're plotting odds ratios from a logistic regression model. In that case, maybe Benn Jann's coefplot (package gr0059_1 from http://www.stata-journal.com/software/sj15-1) would be useful. You can view a presentation on it here.

net install gr0059_1.pkg // To install -coefplot-

Judging from the presentation, it has some nice functionality. I couldn't tell whether it allows you to specify a log-scale on the X-axis, as in the plots you showed above.

HTH.

[ritupuri]

Hello Bruce,
Never heard about it. But on reviewing the presentation, looks like it is quite useful. I will download it and try to work with the data.
Thank you again.
Sincerely,
Ritu

[Muhammad Rashid]

Dear All

I am in desperate need of doing what Ritu has tried doing, my main issue is that I can't find the option to produce the log scale on X-axis in this programme mentioned by Bruce. if there is any other programme or to do this please help me.
PS: Ritu did you actually manage to produce your graph if so do you mind sharing your code here?

Thank you

[Yogesh Sharma]

Dear David,

I am trying to input ORs directly for the Forest plot.
Is it possible to suppress the size of the variables in the Forest plot after while using metan command?

Thanks
Yogesh

[David Fisher]

Dear Yogesh Sharma

Do you mean, to suppress the relative sizes of the boxes, based on the effect-size weights? If so, then the easiest way of achieving this, is with the forestplot option nobox. This removes the grey weighted boxes entirely; and you can change the remaining, smaller markers as you wish (including making them larger!) using pointopts().

Best wishes,
David.

[Yogesh Sharma]

Thanks David