Skip to content

[R-meta] Meta-analysis Hedges g - comparisons with same placebo

8 messages · P. Roberto Bakker, Gerta Ruecker, Viechtbauer Wolfgang (STAT)

Original
P. Roberto Bakker
# 
Hi, 

We are conducting a meta-analysis of RCTs between antidepressants and placebo and calculate Hedges g. 
Question:
How can we deal with studies which compare different antidepressants with the same placebo? Our concern is when we include the different antidepressant comparisons with the same placebo in the meta-analysis the placebo size will increase artificially which will lead to more precise spurious results. If our reasoning is correct, how can we deal this?

Thank you in advance, 

Roberto

Verzonden vanuit Mail voor Windows 10
Gerta Ruecker
# 
Dear Roberto,

it is a network meta-analysis what you are planning? Do you think of 
multi-arm studies including a placebo group? If so, and if it is the 
same placebo that is compared to several antidepressants, then I think 
there is no problem. The different placebo groups in different studies 
provide independent information and network meta-analysis may in fact 
increase precision (provided you have loops in your network), but this 
is exactly what network meta-analysis is intended to do.

To avoid misunderstandings: What you should not do is to use a placebo 
group in a multi-arm study twice in a pairwise meta-analysis without 
adjustment.

For an introduction, see 
http://onlinelibrary.wiley.com/doi/10.1002/jrsm.1259/full .

Best,

Gerta
P. Roberto Bakker
# 
Hi Gerta,

Thank you for your reply and suggestions.
Our meta-analysis is not a network meta-analysis but a 'regular', i.e.
pairwise. Do you have suggestion how to adjust for this?
Bw,
Roberto


Op 18 okt. 2017 10:41 a.m. schreef "Gerta Ruecker" <
ruecker at imbi.uni-freiburg.de>:

Dear Roberto,

it is a network meta-analysis what you are planning? Do you think of
multi-arm studies including a placebo group? If so, and if it is the
same placebo that is compared to several antidepressants, then I think
there is no problem. The different placebo groups in different studies
provide independent information and network meta-analysis may in fact
increase precision (provided you have loops in your network), but this
is exactly what network meta-analysis is intended to do.

To avoid misunderstandings: What you should not do is to use a placebo
group in a multi-arm study twice in a pairwise meta-analysis without
adjustment.

For an introduction, see
http://onlinelibrary.wiley.com/doi/10.1002/jrsm.1259/full .

Best,

Gerta


--

Dr. rer. nat. Gerta R?cker, Dipl.-Math.

Medical Faculty and Medical Center - University of Freiburg
Institute for Medical Biometry and Statistics

Stefan-Meier-Strasse 26, D-79104 Freiburg, Germany

Phone +49 (0)761 2036673
Fax   +49 (0)761 2036680

Mailruecker at imbi.uni-freiburg.de
Webwww.imbi.uni-freiburg.de/biom/


Am 18.10.2017 um 08:12 schrieb P. Roberto Bakker:
placebo and calculate Hedges g.
the same placebo? Our concern is when we include the different
antidepressant comparisons with the same placebo in the meta-analysis the
placebo size will increase artificially which will lead to more precise
spurious results. If our reasoning is correct, how can we deal this?
_______________________________________________
R-sig-meta-analysis mailing list
R-sig-meta-analysis at r-project.org
https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis
Gerta Ruecker
# 
As I wrote in my response, see our paper 
http://onlinelibrary.wiley.com/doi/10.1002/jrsm.1259/full 
<http://onlinelibrary.wiley.com/doi/10.1002/jrsm.1259/full> .

Best,

Gerta


Am 18.10.2017 um 13:10 schrieb P. Roberto Bakker:

  
  


  


      

    

    

      
      

        


  

        

      Viechtbauer Wolfgang (STAT)
    

    

      
      #
    

  


  

      
There is also code for computing the covariance among the standardized mean differences in such 'multitreatment' studies here:

http://www.metafor-project.org/doku.php/analyses:gleser2009

Best,
Wolfgang

-----Original Message-----
From: R-sig-meta-analysis [mailto:r-sig-meta-analysis-bounces at r-project.org] On Behalf Of Gerta Ruecker
Sent: Wednesday, 18 October, 2017 13:28
To: P. Roberto Bakker
Cc: r-sig-meta-analysis at r-project.org
Subject: Re: [R-meta] Meta-analysis Hedges g - comparisons with same placebo

As I wrote in my response, see our paper 
http://onlinelibrary.wiley.com/doi/10.1002/jrsm.1259/full 
<http://onlinelibrary.wiley.com/doi/10.1002/jrsm.1259/full> .

Best,

Gerta

Am 18.10.2017 um 13:10 schrieb P. Roberto Bakker:

            

      
      
      
    

  
  


  


      

    

    

      
      

        


  

        

      P. Roberto Bakker
    

    

      
      #
    

  


  

      
It is a nice code. Thank you.

Some questons:
1. N1 is not always n1i+n2i - how can I interpret this?

2. I am analysing standardized mean deference *change* - I suppose I can
use the syntax also, correct?

Cheers, Roberto

2017-10-19 10:26 GMT+02:00 Viechtbauer Wolfgang (SP) <
wolfgang.viechtbauer at maastrichtuniversity.nl>:

            

      
      
      
    

        

      
      
    

        

  
  


  


      

    

    

      
      

        


  

        

      Viechtbauer Wolfgang (STAT)
    

    

      
      #
    

  


  

      
1) Ni is the total sample size of a study. For multitreatment studies, this is the control group size plus the sizes of the treatment groups.

2) Unfortunately, no. The code given there is specifically for the standardized mean difference (for two or more independent groups). The equations (and hence code) will be different when the outcome measure is the difference between the standardized mean changes for the various groups; I assume you mean something along the lines of:

http://www.metafor-project.org/doku.php/analyses:morris2008

Best,
Wolfgang

-----Original Message-----
From: P. Roberto Bakker [mailto:robertobakker at gmail.com] 
Sent: Thursday, 19 October, 2017 19:25
To: Viechtbauer Wolfgang (SP)
Cc: r-sig-meta-analysis at r-project.org
Subject: Re: [R-meta] Meta-analysis Hedges g - comparisons with same placebo

It is a nice code. Thank you.?

Some questons:
1. N1 is not always n1i+n2i - how can I interpret this?

2. I am analysing standardized mean deference change?- I suppose I can use the syntax also, correct?

Cheers, Roberto

2017-10-19 10:26 GMT+02:00 Viechtbauer Wolfgang (SP) <wolfgang.viechtbauer at maastrichtuniversity.nl>:
There is also code for computing the covariance among the standardized mean differences in such 'multitreatment' studies here:

http://www.metafor-project.org/doku.php/analyses:gleser2009

Best,
Wolfgang

  
  


  


      

    

    

      
      

        


  

        

      P. Roberto Bakker
    

    

      
      #
    

  


  

      
Thank you for your reply.
Yes, I mean something along the lines of http://www.metafor-project.
org/doku.php/analyses:morris2008

2017-10-19 20:42 GMT+02:00 Viechtbauer Wolfgang (SP) <
wolfgang.viechtbauer at maastrichtuniversity.nl>: