[R-meta] Combine HR from multiple groups?

[Kathrin Kuhr]

Hi all,

our situation: In a network meta-analysis, we want to pool two nodes 
(two doses of the same treatment). Outcome of interest is overall 
survival (Hazard ratios). To include a three-arm study (arm A: treatment 
dose A; arm B: treatment dose B, arm C: placebo), we have to combine the 
two treament arms. We only have relative treatment effect, i.e. HR A vs. 
B, HR A vs. C, HR B vs. C. Is there a possibility to pool HR A vs. C and 
B vs. C? As we use the same comparator C, we cannot easily calculate a 
pooled meta-analysis estimate, right? And as we only have relative 
effects, we cannot easily pool arm A and B as for binary data or mean 
values, right? Or do we think too complicated?

We are happy for any ideas or examples.

Thanks,

Kathrin and Anne

-- 
Dr. rer. medic. Kathrin Kuhr, Dipl.-Stat.
AG Medizinische Statistik

Das IMSB ist umgezogen:

Universitätsklinikum Köln
Institut für Medizinische Statistik und Bioinformatik
Robert-Koch-Str. 10 (Gebäude 55, Gartenhaus; 2.OG, Raum 2.004)
D-50931 Köln

fon +49 (0)221 478-33410
fax +49 (0)221 478-33420
kathrin.kuhr using uni-koeln.de
http://imsieweb.uni-koeln.de/