Skip navigation

Coxibs and cardiovascular harm in arthritis


Clinical bottom line

Meta-analyses of randomised trials, and large randomised trials, show no convincing evidence for increased heart attacks, strokes, or cardiovascular death in patients treated with coxibs for arthritis. NNH estimates for an APTC endpoint were much larger (better) than those seen in cancer prevention trials with coxibs or aspirin.


MR Weir et al. Selective COX-2 inhibition and cardiovascular effects: a review of the rofecoxib development program. American heart Journal 2003 146: 591-604.

WB White et al. Cardiovascular thrombotic events in arthritis trials of the cyclooxygenase-2 inhibitor celecoxib. American Journal of cardiology 2003 92: 411-418.

WB White et al. Effects of the cyclooxygenase-2 specific inhibitor valdecoxib versus nonsteroidal anti-inflammatory agents and placebo on cardiovascular thrombotic events in patients with arthritis. American Journal of Therapeutics 2004 11: 244-250.

S Curtis et al. Cardiovascular safety summary associated with the etoricoxib development programme. Arthritis & Rheumatism 2003 48 Suppl, S616.

ME Farkouh et al. Comparison of lumiracoxib with naproxen and ibuprofen in the therapeutic arthritis research and gastrointestinal event trial (TARGET), cardiovascular outcomes: randomised controlled trial. Lancet 2004 364: 675-684.


Whether coxibs cause cardiovascular adverse events (myocardial infarction, stroke, unstable angina, sudden cardiac death) as opposed to congestive heart failure (a known effect of NSAIDs and coxibs) is a hot topic. Most RCTs and high-quality observational studies find no increase in individual studies.

But some studies do find an increase, albeit with small numbers of events. Plausible mechanisms have been proposed for Cox-2 inhibitors to cause thrombosis [1] and animal experiments can demonstrate increases firm platelet adhesion in arterioles, with faster occlusion of microvessels [2], as least in hamsters.

So this is an appropriate time to examine the evidence. There are a number of issues that make this a very complicated area, even for randomised trials with tens of thousands of patients enrolled:


The evidence we have comes from four analyses of clinical trial data of coxibs, and from one large randomised trial. Briefly, the details are:


None of the five reviews found any overall difference between coxib or NSAID for thrombotic events, or between coxib and placebo. All the reviews used analyses based on patient-years of exposure. Some of the analyses looked at different NSAIDs, especially naproxen or other non-naproxen NSAIDs, and some looked at patients taking or not taking aspirin. There was no significant difference for any of these, and the number of APTC events per 100 patient years for coxib compared with placebo and NSAID are shown in Table 1.

Table 1: APTC events with coxib, placebo, and NSAID

Events per 100 patient years
Percent naproxen
about 50

These analyses all take into account the propensity in some studies for patients taking coxibs to discontinue less often than those taking NSAIDs. This means that those on coxibs have more time on the drug, and can therefore accumulate more, possibly unrelated, adverse events. So all the meta-analyses have been corrected for duration.

If we take the worse case, and just take the crude results form the meta-analyses and trial, we can sum the overall risk for all coxibs, and all doses, in arthritis trials (plus some in Alzheimer's disease). These are shown in Table 2. For the comparison with placebo there were about 30,000 patients, in mainly short-term studies, with no significant difference. For the comparison with NSAIDs with over 80,000 patients, in shorter and longer term studies, there was a bare statistical significance, with an NNH of over 1,000 at the point estimate, but with a confidence interval that includes no effect.

Table 2: Overall results for all coxibs versus all NSAIDs, and all placebo

Events with
Crude RR
(95% CI)
Crude NNH
(95% CI)
All coxibs vs placebo (four coxibs)
0.9 (0.7 to 1.2)
-560 (4,200 to -260 )
All coxibs versus NSAIDs (five coxibs)
1.2 (1.02 to 1.5)
1060 (520 to -40,000)



Some will see these data as there being no smoke without fire, while others will wonder whether this is all smoke and mirrors emanating from different agendas. The overall picture from cardiovascular events observed in randomised trials is relatively benign, with no difference between coxibs and placebo, and at worst a small difference between coxibs and NSAIDs and point estimate for NNH for an APTC event of about 1000. This contrasts with studies in colorectal polyp prevention studies, where the NNH for one APTC event over three years was 70. It is difficult to reconcile this order of magnitude difference.

Points of view, and questions, abound:

If pain were well treated, this would be less of a problem. But pain is badly treated, it affects a huge proportion of elderly people. What we really need is more research in how better to deliver the effective analgesics we have, in the most effective and the safest way. Pragmatic rather than explanatory trials. The problem isn't what intervention is best, but rather how best to use the interventions available to best effect. Agonising about safety of coxibs might be seen as a distraction from that goal.


  1. GA Fitzgerald. Cox-2 and beyond: approaches to prostaglandin inhibition in human disease. Nature Reviews: Drug Discover 2003 2: 879-890.
  2. MA Buerkle et al. Selective inhibition of cyclooxygenase-2 enhances platelet adhesion in hamster arterioles in vivo. Circulation 2004 110: 2053-2059.
  3. E Ott et al. Efficacy and safety of the cyclooxygenase 2 inhibitors parecoxib and valdecoxib in patients undergoing coronary artery bypass surgery. Journal of Thoracic and cardiovascular Surgery 2003 125: 1481-1492.
  4. N Nussmeier et al. Complications of the COX-2 inhibitors parecoxib and valdecoxib after cardiac surgery. NEJM 2005 352:1081-1091.