Inconsistencies in cardiovascular data from COX-2 inhibitor trials--is it a class effect?

Abstract

The recent editorials in the New England Journal of Medicine about inaccuracies in cardiovascular data in the publication of the VIGOR (Vioxx Gastrointestinal Outcomes Research) study1,2 has prompted us to report similar inaccuracies in the CLASS study (Celecoxib Long-term Arthritis Safety Study), which were identified during a systematic review of the cardiovascular effects of celecoxib.3 Like the VIGOR study, CLASS was designed primarily to compare the incidence of upper gastrointestinal adverse events resulting from a COX-2 specific inhibitor with those of conventional non-steroidal anti-inflammatory drugs (NSAIDs) in patients with arthritis. As with the VIGOR study, in CLASS there were also inconsistencies in the serious adverse cardiovascular events reported in the different published reports. The outcomes of CLASS have been reported in two peer-reviewed journals4,5 and in a number of documents posted on the Food and Drug Association website, of which the medical officer review is the most complete.6 We have reviewed the data from these three major publications.4-6 In September 2000 the initial 6-month data were published by Silverstein and colleagues in JAMA in which no cardiovascular deaths were reported (Table 1).4 Also in September 2000, the FDA medical officer review of celecoxib data supplied by GD Searle & Co, reported that in CLASS there were five and eight cardiovascular deaths in the celecoxib and NSAID groups, respectively, during the full 12-month study period or within the following 28 days (Table 1).6 In the subsequent 2002 American Journal of Cardiology publication of the cardiovascular events from CLASS, a total of 10 cardiovascular deaths were reported in each of the celecoxib and NSAID groups (Table 1).5 It is difficult to understand why all the cardiovascular deaths would have occurred after 6 months of treatment, and how there could be such major discrepancies between the two complete study reports for such a definite outcome measure as death. Table 1 Mortality data reported in the three CLASS (Celecoxib Longterm Arthritis Safety Study) publications Similarly, there were inconsistencies between the three sources of CLASS data in the reporting of myocardial infarction (Table 2). For example, in the publications of the entire CLASS data, one reported 26 myocardial infarctions in the celecoxib group and 15 in the NSAID group5, whereas the other reported 19 and 13 myocardial infarctions, respectively.6 Table 2 Myocardial infarction data reported in the three CLASS (Celecoxib Long-term Arthritis Safety Study) publications These findings add to the concerns previously expressed regarding the selective and partial reporting of the gastrointestinal side effects from the CLASS study and that the published CLASS study differs from the original protocol in primary outcomes, statistical analysis, trial duration and conclusions.7-10.7,8 Thus history appears to have repeated itself in two respects. First, in CLASS there were inconsistencies in the reporting of cardiovascular data. Of greater concern is the identification that in the two major safety studies of the most commonly used COX-2 specific inhibitors there were major inconsistencies in reporting of the cardiovascular events. Thus, there would seem to be a systemic problem in research conduct in this field. There are considerable public health, pharmaceutical, regulatory and clinical implications.