Comment significant decrease in the frequency of upper gastrointestinal ulcer complications in patients not taking low-dose aspirin (64 [0· 92%] vs 14 [0· 20%]). Thus, the number needed to treat to prevent one ulcer complication in this group of low-risk patients was 139. However, in the 24% of patients taking low-dose aspirin, the benefit of lumiracoxib on ulcer complications was no longer significant (19 [0· 88%] vs 15 [0· 69%]), again confirming the results of previous studies. 2 Beyond the salutary effects on ulcer complications, the hepatotoxicity of lumiracoxib was manifest: the frequency of greater than 3-fold transaminase elevations was 2· 57%(230) for lumiracoxib compared with 0· 63%(56) for the two NSAIDs (hazard ratio 3· 97 [2· 96–5· 32]). The authors claim that the dose of lumiracoxib tested was supratherapeutic, but why a higher than anticipated dose of lumiracoxib would have been studied or whether its hepatotoxicity was dose-dependent is unclear.

TARGET quantifies lumiracoxib’s narrow benefit over two NSAIDs with a trade-off. For patients not taking aspirin, there is an absolute reduction of 0· 72% in ulcer complications, with an excess of 2· 0% of liver function test abnormalities. The putative benefit is further compromised if naproxen is the NSAID, with a 0· 17% excess of myocardial infarction. For patients taking low-dose aspirin, it is hard to justify the coxib: there is no benefit in ulcer complication reduction, but the risk of myocardial infarction and hepatotoxicity persist. The coxib field has been marked by intensive direct-toconsumer advertising in the USA, and sales of these drugs exceed US $7 billion per year. Yet it is hard to imagine the justification for this extraordinary adoption of coxibs in light of marginal efficacy, heightened risk, and excessive cost compared with traditional NSAIDs. 13, 14 A recent epidemio-