COX 2 inhibitors, traditional NSAIDs, and the heart

BMJ 2005; 330 doi: (Published 09 June 2005) Cite this as: BMJ 2005;330:1342
  1. Peter Jüni, senior research fellow in clinical epidemiology (juni{at},
  2. Stephan Reichenbach,
  3. Matthias Egger, visiting professor of clinical epidemiology
  1. Department of Social and Preventive Medicine, University of Berne, Finkenhubelweg 11, 3012 Berne, Switzerland
  2. Department of Rheumatology, University Hospital Berne, 3010 Berne, Switzerland
  3. MRC Health Services Research Collaboration, Department of Social Medicine, University of Bristol, Bristol BS8 2PR

    Adverse event data from clinical trials must inform decision making

    These are trying times for patients with chronic musculoskeletal pain. Worrying data about the drugs they regularly use keep emerging. In September 2004 rofecoxib (Vioxx) was withdrawn by Merck after the adenomatous polyp prevention on Vioxx (APPROVe)1 trial showed an increase in major cardiovascular events in patients with a history of colorectal adenomas who were randomised to receive Vioxx, compared with those in the placebo group.w1Rofecoxib had been marketed as the non-steroidal anti-inflammatory drug (NSAID) of choice because selective inhibition of the isoform 2 of the cyclooxygenase (COX 2) enzyme made it highly effective but free from gastrointestinal toxicity.

    More unwelcome data from placebo controlled trials of rofecoxib's competitors followed: valdecoxib (Bextra, Pfizer) taken after coronary artery bypass grafting was shown to be associated with an increased incidence of cardiovascular events2; and the adenoma prevention with celecoxib (APC) trial3 reported an increased risk of cardiovascular events associated with use of celecoxib (Celebrex, Pfizer), a drug known to be less selective for COX 2 than rofecoxib or valdecoxib.4 A small increase in the risk of myocardial infarction was also observed for the highly selective lumiracoxib (Prexige, Novartis).5 No data on the cardiovascular safety of etoricoxib (Arcoxia, MSD) from large trials have been published so far, but no news is no longer good news: patients and doctors are anxious to know whether cardiotoxicity is a class effect applicable to any COX 2 inhibitor, or even to NSAIDs in general.

    In this week's BMJ two observational studies address this question. A retrospective cohort study ( ARTICLE)6 in patients with congestive heart failure found lower mortality in patients treated with celecoxib than with rofecoxib or traditional NSAIDs. A case-control study nested in a UK general practice database ( ARTICLE)7 found a similar risk of myocardial infarction for celecoxib, rofecoxib, ibuprofen and naproxen, but a somewhat higher risk with diclofenac.

    We believe that these results should be interpreted with caution. For example, the similar risk of myocardial infarction for naproxen and rofecoxib found in the case-control study7 is incompatible with the trial data8 and could be explained by confounding by indication if patients with a history of heart disease were more likely to receive naproxen than rofecoxib or other NSAIDs. The quality of the data on cardiovascular risk factors and other potential confounders was poor in both studies, and the ability to control for confounding therefore limited. For example, information on smoking was unrecorded in 13% of cases and 20% of controls in the case-control study7 and entirely unavailable in the retrospective cohort study.6

    What are the alternatives? We have argued that all unbiased data on serious adverse events from clinical trials should be made available to independent researchers and the public and analysed in a timely fashion.9 Indeed, in the case of rofecoxib, cumulative meta-analysis of clinical trial data showed that an increased risk of myocardial infarction was evident from 2000 onwards.8 Similar analyses are now required for the other COX 2 inhibitors.

    The US Food and Drug Administration (FDA) and other licensing authorities are an important source of relevant data. The FDA reviews clinical trials worldwide before approval and labelling, and again before relabelling of approved drugs. As part of the 1966 Freedom of Information Act, the agency is required to make available its reports on all drugs that are approved. Unfortunately, these reports are not as useful as they could be. We found that the criteria for including trials in reports were often unclear. For example, only 16 out of at least 27 trials of celecoxib that were performed up to 2002 in patients with musculoskeletal pain were included in the relevant reports. In any event, reporting on study characteristics and adverse events was not always transparent, and complete data on cardiovascular safety were available for only three trials. In the case of valdecoxib, we found that many pages and paragraphs had been deleted because they contained “trade secret and/or confidential information that is not disclosable” (figure).w2


    Facsimile of pages 1 and 2 of the Food and Drug Administration (FDA) statistical review and evaluation of valdecoxib.w2Publicly available at (accessed 20 May 2005)

    Surely, the protection of the public's health justifies full access to the safety data submitted by industry to the FDA and other drug licensing authorities, and mandates transparent reporting on harms, in accordance with international guidelines.10 Meta-analyses of adverse events might not resolve controversies, but will help decision making about issues such as the need for additional trials.

    Observational studies “simply cannot test definitely whether there are small to moderate risks or benefits of a class of drugs when the factors associated with prescription of a particular drug are difficult to control and perhaps even uncontrollable.”11 This statement referred to postulated adverse events of calcium antagonists in the treatment of hypertension, a controversy finally resolved by large pragmatic trials, including the seminal antihypertensive and lipid lowering treatment to prevent heart attack trial (ALLHAT).12 Such large trials might be required, ultimately, to establish the best and safest treatment for patients with musculoskeletal pain.

    Embedded ImageAdditional references w1 and w2 are on

    Primary care pp 1366, 1370


    • Competing interests None declared


    1. 1.
    2. 2.
    3. 3.
    4. 4.
    5. 5.
    6. 6.
    7. 7.
    8. 8.
    9. 9.
    10. 10.
    11. 11.
    12. 12.