Rheumatology Advance Access originally published online on August 27, 2007
Rheumatology 2007 46(10):1623-1624; doi:10.1093/rheumatology/kem216
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
COX-2 inhibitor
Nordic Cochrane Centre, Copenhagen, Denmark and 1Department of Clinical Pharmacology, Gentofte Hospital, Copenhagen, Denmark
Correspondence to: Peter C. Gøtzsche, Nordic Cochrane Centre, Copenhagen, Denmark. E-mail: pcg{at}cochrane.dk
SIR, We have serious doubts about the pooled data from two studies comparing etoricoxib with celecoxib and placebo reported by Bingham and colleagues [1]. First, the title indicates that the two studies were non-inferiority studies. However, the study design does not suggest that the studies were designed as non-inferiority studies, e.g. it would be highly unusual to give the power as 87% and not 90% if a study is planned prospectively. Furthermore, non-inferiority studies are only recommended and necessary when it is unethical to use a placebo, but both studies had a placebo control. We, therefore, suspect that these studies were not designed as non-inferiority studies (and one cannot ‘do’ a non-inferiority study retrospectively, of course). Second, it is unclear what constituted a study and how many arms each study had. The two studies are described as double-blind, which suggests there were three arms in each study, but Fig. 2 in the study shows four arms as there were two placebo groups (which is hard to understand for a double-dummy study), whereas Table 1 in the study shows only three arms. Third, while results for efficacy were pooled, results for harms were not. Finally, it is indicated four times in the paper that ‘etoricoxib was at least as effective as celecoxib’. This is nonsense for three reasons. First, from a non-inferiority study one can only conclude that one drug was not worse than the other by the amount specified in the protocol as the non-inferiority interval [2]. Second, this amount, 10 mm on the VAS, was almost as great as the difference to placebo in the three pain domains (11.1, 10.2 and 11.5 mm). It is grossly inappropriate to choose a non-inferiority interval of similar size as the effect of active drug over placebo [2] as the ‘non-inferior’ drug may then not be any better than placebo. Third, it is also a logical error. If a drug is ‘at least as effective’ as another, it means that it cannot be worse, which means that the entire confidence interval lies outside the non-inferiority interval, corresponding to P = 0.05 in a classical test for superiority, which means that the new drug is actually better than the control, which it was not.
The authors have declared no conflicts of interest.
 |
References |
|---|
 Top References |
|---|
- Bingham CO, Sebba AI, Rubin BR, et al. Efficacy and safety of etoricoxib 30 mg and celecoxib 200 mg in the treatment of osteoarthritis in two identically designed, randomized, placebo-controlled, non-inferiority studies. Rheumatology (2007) 46:496–507.
[Abstract/Free Full Text] - Gøtzsche PC. Lessons from and cautions about noninferiority and equivalence randomized trials. JAMA (2006) 295:1171–3.
[Free Full Text]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||