how to evaluate site/center effect in multicentric bioequivalence study? [General Sta­tis­tics]

posted by DavidManteigas – Portugal, 2016-06-15 20:30 (3160 d 11:10 ago) – Posting: # 16430
Views: 6,500

Hi,

Quoting E9:

"The statistical model to be adopted for the estimation and testing of treatment effects should be described in the protocol. The main treatment effect may be investigated first using a model which allows for centre differences, but does not include a term for treatment-by-centre interaction."

So imo the first step would be to include the centre effect in your model and check the significance. If the centre effect is significant, and given the nature of your endpoints, that would be a big trouble for you to explain. Being your study cross-over, maybe you should examine interactions between centre and sequence/group effect (question: was your randomization stratified by centre?). Analysis of multicentre data is dependent on the study features. For instance, if you have 50 centres with and average of 6~7 patients by centre maybe it is reasonable to include centre effects as random. However, if you have 2/3 centres with 30~40 patients per centre I would assess centre effects as fixed terms in the model and pray for non-significance :-D

It's hard to give an opinion without having a background on your study. However, I believe ICH E9 is quite clear in what you should do regarding your statistical analysis.

Complete thread:

UA Flag
Activity
 Admin contact
23,376 posts in 4,912 threads, 1,665 registered users;
151 visitors (0 registered, 151 guests [including 6 identified bots]).
Forum time: 06:40 CET (Europe/Vienna)

There are sadistic scientists who hurry to hunt down errors
instead of establishing the truth.    Marie Curie

The Bioequivalence and Bioavailability Forum is hosted by
BEBAC Ing. Helmut Schütz
HTML5