Published in:
01-08-2007 | Correspondence
Response to the letter by Williams et al.
Authors:
Guido Bertolini, Carlotta Rossi, Abramo Anghileri, Sergio Livigni, Antonio Addis, Daniele Poole
Published in:
Intensive Care Medicine
|
Issue 8/2007
Login to get access
Excerpt
Sir: The experts at Eli Lilly and Company appear concerned about whether the covariates considered in our multivariate analysis were predefined or not. We object that since observational studies often carry unforeseeable confounders, it is advisable to explore available data to find the variables needed to adjust the estimates of interest, in an effective multivariate analysis [
1]. Therefore, unlike subgroup analysis in randomised controlled trials (RCTs), it is illogical to prospectively define all confounders in observational studies. However, the covariates that turned out to be significant in our study were a logical choice suggested by the literature and were selected before performing the statistical analysis. …