Clinical trials: creating graphs for subgroup interactions
I wanted to update Statalist with an resolution to a previous query I had made here (a follow-up from the thread at http://www.statalist.org/forums/foru...p-interactions).
The goal: to use Stata to create publication-quality Figures depicting the effect on outcomes in various subgroups involved in clinical trials. (See attached figure 1 as an example.)
The resolution: If you read the linked-to thread above, you will see that David Fisher had just written a series of commands for precisely that use (amongst other things, namely to perform 2-stage inverse-variance individual participant data meta-analysis).
His Stata package is -ipdmetan- and is available by typing "ssc install ipdmetan".
I had a go using his commands -ipdover- and -forestplot-, and, thanks to his detailed help, was able to come up with an example which worked for me, and I thought I should share the results. (He has also recently updated his software, so those who already downloaded it should -adoupdate- to get his latest version.)
I have attached a sample data file and a sample .do file to the next message which will run self-contained and produce the second attached figure. (Statalist didn't seem to let me attach graphics and a .dta file at the same time.)
Although I used pre-summarized data in my example, David gives two additional, very useful examples in the help files for -ipdover- and -forestplot-. These examples show you how the commands can seamlessly be integrated with your trial's data (without pre-calculating effect sizes and confidence intervals). Only use my example if you want to have manual control over how your plot is created.
My comments:
Kind regards,
Phil
The goal: to use Stata to create publication-quality Figures depicting the effect on outcomes in various subgroups involved in clinical trials. (See attached figure 1 as an example.)
The resolution: If you read the linked-to thread above, you will see that David Fisher had just written a series of commands for precisely that use (amongst other things, namely to perform 2-stage inverse-variance individual participant data meta-analysis).
His Stata package is -ipdmetan- and is available by typing "ssc install ipdmetan".
I had a go using his commands -ipdover- and -forestplot-, and, thanks to his detailed help, was able to come up with an example which worked for me, and I thought I should share the results. (He has also recently updated his software, so those who already downloaded it should -adoupdate- to get his latest version.)
I have attached a sample data file and a sample .do file to the next message which will run self-contained and produce the second attached figure. (Statalist didn't seem to let me attach graphics and a .dta file at the same time.)
Although I used pre-summarized data in my example, David gives two additional, very useful examples in the help files for -ipdover- and -forestplot-. These examples show you how the commands can seamlessly be integrated with your trial's data (without pre-calculating effect sizes and confidence intervals). Only use my example if you want to have manual control over how your plot is created.
My comments:
- the data in my files is fictitious
- weights are entered directly. So, if you are analyzing a clinical trial with a total of 500 patients, you enter in the number of patients in each subgroup directly (the attached example should be clear)
- P values for interactions are not included in the suite of commands, so you'll need to calculate them yourself
- I include the gaudy colours only for illustrative purposes, not because I think they are attractive
Kind regards,
Phil
Comment