Randomized controlled trials of COX-2 inhibitors: an analysis of doses used and trends over time to investigate implications for comparative safety.

BACKGROUND: Naproxen, ibuprofen and diclofenac are frequently used as comparators in randomized controlled trials (RCTs) on the safety and efficacy of cyclooxygenase (COX)-2 inhibitors. Different comparator doses may influence the results of RCTs. It has been hypothesized that RCTs of COX-2 inhibitors where different doses were administered resulted in different conclusions about the cardiovascular safety of COX-2 inhibitors. High comparator doses may let COX-2 inhibitors look better in terms of safety, while low comparator doses may result in the opposite.
OBJECTIVE: The aim of the study was to compare doses of COX-2 inhibitors and comparator drugs used in RCTs, and to investigate dose changes over time. STUDY DESIGN AND METHODS: We searched the Cochrane Central Register of Controlled Trials, The Cochrane Library for published Cochrane reviews, Clinicaltrials.gov and PubMed, for RCTs between 1995 and 2009 in which celecoxib or rofecoxib were compared with naproxen, ibuprofen or diclofenac. All articles labelled as RCTs mentioning rofecoxib or celecoxib and one or more of the comparator drugs in the title and/or abstract were included. We extracted information on doses of both non-selective NSAIDs and selective COX-2 inhibitors used in the RCTs, and study year. The Mann-Whitney test was used to compare the difference in median dose in rofecoxib and celecoxib RCTs. Linear regression was performed to evaluate trends in dosage over time. For comparisons between COX-2-inhibitors, celecoxib trials after the 2004 market withdrawal of rofecoxib were excluded.
RESULTS: Median defined daily dose (DDD) of celecoxib (2.00) was higher than the median DDD of rofecoxib (1.00; p < 0.001), whereas non-selective NSAID doses were comparable in rofecoxib (2.00) and celecoxib (2.00; p = 0.988) studies. In both groups, the non-selective NSAID doses decreased over time (B [regression coefficient] = -0.07; p = 0.28, and B = -0.054; p = 0.09, respectively). In contrast, the DDDs of rofecoxib increased slightly over time (B = 0.037; p = 0.28), whereas the celecoxib DDDs decreased over time (B = -0.081; p = 0.09). In due course, the contrasts between DDDs of COX-2 inhibitors and non-selective NSAIDs converged, both in rofecoxib and celecoxib RCTs; therefore, doses have become more comparable in recent years because of differences in steepness of two decreasing dose trends in the case of celecoxib, and opposing dose trends in the case of rofecoxib.
CONCLUSIONS: Although the dose trends over time differed for RCTs comparing rofecoxib and celecoxib with diclofenac, ibuprofen or naproxen, the results of our study do not support the hypothesis that dose trends influenced the decision to continue marketing celecoxib after the withdrawal of rofecoxib because the overall median DDD of celecoxib was substantially higher than the median DDD of rofecoxib, while non-selective NSAID DDDs were comparable.