Skip to main content
NCBI home page
The BMJ logoLink to The BMJ
. 2002 Jul 20;325(7356):161.

Are selective COX 2 inhibitors superior to traditional NSAIDs?

Rofecoxib did not provide unequivocal benefit over traditional NSAIDs

Brian R Budenholzer 1
PMCID: PMC1123680  PMID: 12130617

Editor—In their editorial Jüni et al say that celecoxib is no safer than diclofenac or ibuprofen and that the CLASS authors “spun” their analysis to suggest otherwise.1 They also state: “In contrast with the CLASS trial, the VIGOR trial, which was similar in design and outcomes, found an unequivocal benefit of another selective COX 2 inhibitor, rofecoxib, over traditional non-steroidal anti-inflammatory agents [NSAIDs].”2

I disagree. Although serious gastrointestinal adverse effects were less frequent in rofecoxib users than in patients with rheumatoid arthritis treated with naproxen (number needed to treat to prevent one serious upper gastrointestinal event 191; 95% confidence interval 114 to 586), rofecoxib was in fact less safe than naproxen. The published version of the VIGOR trial focused on the narrow outcome of serious gastrointestinal complications.

The US Food and Drug Administration took the unprecedented step of presenting its review of both the CLASS trial and the VIGOR trial on its website.3 Review of the complete data presented there shows that when all serious adverse events are included—not just gastrointestinal events—patients treated with naproxen had fewer serious events. Among patients treated with rofecoxib, 9.3% experienced a serious adverse event compared with 7.8% of those treated with naproxen (relative risk 0.81; 0.62 to 0.97). When all serious adverse events are counted, the number needed to harm when rofecoxib is used compared with naproxen is 66 (36 to 332).

The increased risk of serious adverse events was due to an increase of serious adverse cardiovascular events, including a 300% greater risk of myocardial infarction in those treated with rofecoxib.

The VIGOR results, examined fully, show that at least one traditional non-steroidal anti-inflammatory drug—naproxen—is unequivocally safer than rofecoxib, albeit with an increased risk of adverse events limited to the gastrointestinal tract.

Footnotes

Created by potrace 1.16, written by Peter Selinger 2001-2019 Three other responses on bmj.com (bmj.com/cgi/eletters/324/7349/1287) made similar points.

References

  • 1.Jüni P, Rutjes AWS, Dieppe PA. Are selective COX 2 inhibitors superior to traditional non steroidal anti-inflammatory drugs? BMJ. 2002;324:1287–1288. doi: 10.1136/bmj.324.7349.1287. . (1 June.) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Bombardier C, Laine L, Reicin A, Shapiro D, Burgos-Vargas R, Davis B, et al. Comparison of upper gastrointestinal toxicity of rofecoxib and naproxen in patients with rheumatoid arthritis. VIGOR Study Group. N Engl J Med. 2000;343:1520–1528. doi: 10.1056/NEJM200011233432103. [DOI] [PubMed] [Google Scholar]
  • 3. US Food and Drug Administration. www.fda.gov/ohrms/dockets/ac/01/briefing/3677b2_03_med.pdf (accesssed 9 July 2002)
BMJ. 2002 Jul 20;325(7356):161.

Pharmacia's response to editorial

G Steven Geis 1

Editor—Contrary to the assertions of Jüni et al in their editorial,1-1 the CLASS design, analyses, and outcome definitions were predefined. The CLASS authors reviewed all the data and decided that the six month analyses were most appropriate for initial publication while the Food and Drug Administration chose nine month data as most appropriate for a recent label change.1-2,1-3 Despite differing medical judgment for the time interval that best reflected the data, and contrary to the allegations in the editorial, the conclusions were similar.

CLASS was a single study using two protocols to ensure treatment blinding, but analysis of the combined results was prespecified. The protocols were similar; however, one included celecoxib 400 mg twice daily and ibuprofen 800 mg thrice daily while the other used celecoxib 400 mg twice daily and diclofenac 75 mg twice daily. Low dose aspirin was allowed, and the minimum expected duration of participation in the study was six months. The primary end point was ulcer complications (bleeding, perforation, and outlet obstruction) verified by endoscopy or contrast radiography, but analysis of symptomatic ulcers was also prespecified. Early withdrawal for an uncomplicated ulcer—that is, symptomatic ulcer—was mandated in the protocol.

The primary analysis was a comparison of ulcer complication rates between celecoxib and the combined non-steroidal anti-inflammatory group (ibuprofen and diclofenac). To control the overall alpha-level, comparisons of celecoxib with each non-steroidal agent were allowed only if the primary analysis was statistically significant. Analyses of ulcer complication risk factors—for example, use of low dose aspirin—were also preplanned. Study assumptions included (a) constant complication rates for non-steroidal anti-inflammatory drugs1-4 and (b) a rate of use of low dose aspirin of around 11%.

Ulcer complication rates were not significantly different for the two groups. However, the rate of the combined end point of symptomatic/complicated ulcers was significantly lower with celecoxib. Since the primary analysis was not significant, comparisons with the individual non-steroidal anti-inflammatory drugs were not valid.

Important design assumptions did not prove to be true. Ulcer complication rates with non-steroidal anti-inflammatory drugs decreased over time instead of remaining constant (figure). Those given non-steroidal anti-inflammatory drugs had a significantly greater withdrawal rate for symptomatic ulcers than those given celecoxib, which was most evident after the first six months (figure). Since symptomatic ulcers are precursors of ulcer complications, patients at high risk who were given non-steroidal anti-inflammatory drugs were being withdrawn more quickly than high risk patients given celecoxib. This differential withdrawal rate introduced study bias, which reduced statistical and medical validity of the analyses over time. Therefore, the CLASS oversight committees judged the six month data to be most valid and reported: “The data after six months were so confounded as to be difficult to interpret for assessing a drug-related causal gastrointestinal toxicity.”1-5

About 22% of patients took low dose aspirin, which affected the results of treatment. Differences in treatment were not significant for the six month comparison of ulcer complication rates for the all patient cohort, but for the cohort of non-aspirin users, the rate was significantly lower for celecoxib than for non-steroidal anti-inflammatory drugs. The US Food and Drug Administration noted: “The use of aspirin . . . may have obscured the ability to accurately compare the gastrointestinal safety of Celebrex to other non-steroidal anti-inflammatory drugs.”1-6 The US label describes a fourfold increase in the nine month ulcer complication rate with celecoxib plus aspirin v celecoxib alone. For the rate of symptomatic/complicated ulcers, a near threefold increase for celecoxib plus aspirin v celecoxib alone is described.1-3

Jüni et al describes CLASS as “overoptimistic,” using “post hoc changes.”1-1 However, the CLASS publication clearly acknowledges that the primary end point was not reached.1-2 There were no post hoc protocol changes, and the analyses of the longer term data, although complicated by the differential withdrawal of patients, do not differ substantially from the six month analyses.1-7

Jüni et al misrepresented CLASS.1-1 We continue to stand behind the study design, analyses, and conclusions.1-2 Furthermore, we invite any discussions that will ensure an understanding of the facts and help in clarifying the safety profile of celecoxib.

References

  • 1-1.Jüni P, Rutjes AWS, Dieppe PA. Are selective COX 2 inhibitors superior to traditional non steroidal anti-inflammatory drugs? BMJ. 2002;324:1287–1288. doi: 10.1136/bmj.324.7349.1287. . (1 June.) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 1-2.Silverstein FE, Faich G, Goldstein JL, Simon LS, Pincus T. Whelton A, et al. Gastrointestinal toxicity with celecoxib vs nonsteroidal anti-inflammatory drugs for osteoarthritis and rheumatoid arthritis: the CLASS study: a randomized controlled trial. Celecoxib Long-term Arthritis Safety Study. JAMA. 2000;284:1247–1255. doi: 10.1001/jama.284.10.1247. [DOI] [PubMed] [Google Scholar]
  • 1-3. FDA approved Celebrex approved prescribing information (7 June 2002). Available at: www.fda.gov/cder/foi/label/1998/20998lbl.pdf (accessed 11 July 2002).
  • 1-4.MacDonald TM, Morant SV, Robinson GC, Shield MJ, McGilchrist MM, Murray FE, et al. Association of upper gastrointestinal toxicity of non-steroidal anti-inflammatory drugs with continued exposure: cohort study. BMJ. 1997;315:1333–1337. doi: 10.1136/bmj.315.7119.1333. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 1-5.Silverstein F, Simon L, Faich G. Reporting of 6-month vs 12-month data in a clinical trial of celecoxib. In reply. JAMA. 2001;286:2399–2400. [PubMed] [Google Scholar]
  • 1-6. FDA talk paper. Labeling changes for arthritis drug Celebrex. Available at: www.fda.gov/bbs/topics/ANSWERS/2002/ANS01151.html (accessed 11 July 2002).
  • 1-7.Goldstein JL. Gastrointestinal (GI) event rates in the CLASS study: six months vs longer-term follow-up analyses. Gastroenterology. 2002;122 (suppl 1):A-469. for the CLASS investigators. [Google Scholar]
BMJ. 2002 Jul 20;325(7356):161.

Little is known about COX 2 inhibitors

Muhammad Mamdani 1,2, David N Juurlink 1,2, Geoffrey M Anderson 1,2

Editor—The editorial by Jüni et al focuses attention not only on the reporting of clinical trials in peer reviewed journals but also on the interpretation of available evidence.2-1 We agree more needs to be done to determine whether COX 2 inhibitors are superior to traditional non-steroidal anti-inflammatory drugs (NSAIDs) and have some suggestions of how that might be done.

Firstly, understanding of the cellular effects of the COX 2 inhibitors is evolving, and conclusions about the comparative safety of these agents based on in vitro data cited by Jüni et al may be premature. Although whole blood assays show that rofecoxib is more selective than celecoxib for COX 2,2-2 such assays have been criticised for having limited clinical relevance.2-3

Furthermore, studies of cancer cell lines indicate that celecoxib has far greater antiproliferative effects than rofecoxib, implying that celecoxib has greater COX 2 selectivity.2-4,2-5 Recent evidence also suggests, however, that celecoxib may possess unique and largely unknown COX independent characteristics over rofecoxib.2-5 The clinical implications of such differences on the gastrointestinal and cardiovascular safety of these two agents are not known. In short, this is a new class of drugs about which comparatively little is known.

Secondly, meta-analysis of existing trial data cannot overcome the important design issues associated with the existing randomised trials of selective COX 2 inhibitors, including the choice of comparison drugs and outcomes. What is needed, and perhaps what should have been ordered by licensing bodies at the outset, is a study or set of studies designed to allow direct comparisons of COX 2 agents with appropriate existing alternative treatments.

Thirdly, regardless of the debate about phase III trials, there is an important role for phase IV pharmacosurveillance studies. While the clinical findings of well designed randomised trials are awaited, phase IV studies will shed light on the impact of COX 2 inhibitors in comparison with other agents in “real world” settings; provide much needed information on rare adverse events such as gastrointestinal haemorrhage, and acute myocardial infarction that may be associated with COX 2 inhibitors; and clarify the extent to which these agents are being used in patients who are most likely to benefit.

As noted by Jüni et al, billions of dollars are being spent on COX 2 inhibitors. It seems prudent to determine whether that investment is justified.

References

  • 2-1.Jüni P, Rutjes AWS, Dieppe PA. Are selective COX 2 inhibitors superior to traditional non steroidal anti-inflammatory drugs? BMJ. 2002;324:1287–1288. doi: 10.1136/bmj.324.7349.1287. . (1 June.) [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2-2.Silverstein FE, Faich G, Goldstein JL, et al. Gastrointestinal toxicity with celecoxib vs nonsteroidal anti-inflammatory drugs for osteoarthritis and rheumatoid arthritis: the CLASS study: a randomized controlled trial. JAMA. 2000;284:1247–1255. doi: 10.1001/jama.284.10.1247. [DOI] [PubMed] [Google Scholar]
  • 2-3.Blain H, Boileau C, Lapicque F, et al. Limitation of the in vitro whole blood assay for predicting the COX selectivity of NSAIDs in clinical use. Br J Clin Pharmacol. 2002;53:255–265. doi: 10.1046/j.0306-5251.2001.01533.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2-4.Patti R, Gumired K, Reddanna P, Sutton LN, Phillips PC, Reddy CD. Overexpression of cyclooxygenase-2 (COX-2) in human primitive neuroectodermal tumors: effect of celecoxib and rofecoxib. Cancer Lett. 2002;180:13–21. doi: 10.1016/s0304-3835(02)00003-4. [DOI] [PubMed] [Google Scholar]
  • 2-5.Waskewich C, Blumenthal RD, Li H, Stein R, Goldenberg DM, Burton J. Celecoxib exhibits the greatest potency amongst cyclooxygenase (COX) inhibitors for growth inhibition of COX-2-negative hematopoietic and epithelial cell lines. Cancer Res. 2002;62:2029–2033. [PubMed] [Google Scholar]
BMJ. 2002 Jul 20;325(7356):161.

Both the CLASS and VIGOR trials support the COX 2 hypothesis

Richard R Stover 1

Editor—With respect to the recent editorial by Jüni et al on the CLASS trial,3-1 the statistical flaws in the trial's design have been well documented, and the manner in which the data were reported have been widely criticised.

Yet, the question raised in the title of this editorial seems far fetched. Clearly, the CLASS protocol was doomed at the outset by permitting aspirin use. Aspirin, after all, is the grandfather of all non-steroidal anti-inflammatory drugs (NSAIDs) with profound COX 1 inhibition. Thus, the Kaplan-Meier curves in figure 2 of the editorial do not reflect any true test of the COX 2 specific hypothesis with celecoxib.

Notwithstanding Jüni et al's protestations of the analytical presentations of the CLASS data, the incidence of ulcers with COX 2 selective treatment with celecoxib in the CLASS trial was considerably below the 2-4% per year.

The US Food and Drug Administration's labelling of celecoxib shows a Kaplan-Meier rate of complicated ulcers at nine months of 0.32% with celecoxib alone v 1.12% among patients taking aspirin with celecoxib. For rofecoxib the rate (at 10.5 months) was 0.52% v 1.22% for naproxen.3-2

Similarly, the rate of symptomatic ulcers for patients taking celecoxib was 0.78% v 2.19% for those taking apirin and celecoxib. For rofecoxib the labelling shows a rate of 1.80% v 3.87% for naproxen.

Presumably, the FDA has made the appropriate adjustments to the analysis.

It should be noted that CLASS enrolled a more elderly, at risk, population. In the CLASS trial 34.8% of patients taking celecoxib alone were aged 65 or over v 24.6% of patients taking rofecoxib in the VIGOR trial.

It would have been instructive to see what the rates were with ibuprofen and diclofenac. I agree with Jüni et al's recommendation for a meta-analysis. I think that within the CLASS trial the P value would be <0.05 if patients treated with celecoxib alone were compared with all patients who received a COX 1 inhibitor (aspirin, ibuprofen, diclofenac) using the Food and Drug Administration's methods.

Today, celecoxib, rofecoxib, and valdecoxib carry the traditional warning for non-steroidal anti-inflammatory drugs, implying that they are associated with ulcer rates of 2%-4% per year. I think that this represents mislabelling of the cruellest sort to patients and which is not supported by any scientific or clinical data. In the United States it enables managed care to deprive patients of the safer alternatives.

References

BMJ. 2002 Jul 20;325(7356):161.

Authors' reply

Peter Jüni 1,2,3, Anne W S Rutjes 1,2,3, Paul Dieppe 1,2,3

Editor—We agree with Budenholzer, who criticises our statement on the “unequivocal benefit” of rofecoxib found in the VIGOR study.4-1 We were referring to ulcer complications, and should have reported VIGOR to have found an “unequivocal gastrointestinal benefit” only. Rofecoxib's fivefold increase in myocardial infarctions observed in VIGOR is particularly worrying.4-2 Incidentally, whereas patients in the CLASS study randomised to celecoxib had similar rates of myocardial infarction to those randomised to ibuprofen, they tended to experience more myocardial infarctions than patients randomised to diclofenac (relative risk 2.21, 95% confidence interval 0.74 to 8.94, P=0.14).4-3

According to Geis, CLASS was a single study, and the JAMA article reported patients to be “randomly assigned on a 2:1:1 basis.”4-4 This is misleading. There were two trials with separate patient recruitment and randomisation procedures requiring separate analyses to preserve randomisation. None the less, the two trials were combined by simply summing them.4-4 This would be appropriate only if allocation of patients to trials was ruled by chance. However, there were highly significant differences at baseline between trials in patients' age, disease severity, ethnic group, and histories of intolerance to non-steroidal anti-inflammatory drugs and use of alcohol. The probability that these differences occurred by chance is P<10-19 .4-3 Considering this and different follow up durations and COX 2 selectivities of comparator drugs,4-5 simply summing results is inappropriate.

The figure (top) shows relative risks of comparisons discussed during the current debate. While there were some significant differences between celecoxib and ibuprofen, there were none between celecoxib and diclofenac in any of the 18 analyses. Relevant differences between trials were hidden and spurious statistical precision was implied through inappropriate pooling. Therefore, we find it remarkable that Geis misuses CLASS's analytical two step procedure,4-3 which should have been a safeguard against a type I error, as a justification for obscuring individual trial results and as a rationale to commit a type I error by inappropriately pooling secondary outcomes. Disturbingly, Pharmacia has also presented pooled results from different protocols with different comparator drugs for SUCCESS-1, the successor study to CLASS.4-6

The main argument for reporting only six month data from CLASS was that patients dropping out because of gastrointestinal adverse events/symptomatic ulcers were at increased risk of ulcer complications.4-3 However, only 11 out of 44 patients with ulcer complications in CLASS developed gastrointestinal symptoms before an ulcer complication occurred (25%) and none had a symptomatic ulcer as a precursor.4-3 Patients who experience gastrointestinal adverse events may be monitored and treated more carefully than patients without any gastrointestinal symptoms, with ulcer complications avoided particularly in patients with symptoms. It is therefore not surprising that gastrointestinal adverse events were associated with a significantly decreased risk of subsequent ulcer complications in CLASS (relative risk 0.28, 0.12 to 0.66, P=0.0018).4-3

The wide confidence intervals of comparisons in the figure (top) indicate that CLASS trials were underpowered. We therefore fully agree with Mamdani et al's suggestions about future research. A meta-analysis of individual patient data will allow adequately powered explorative comparisons between COX 2 inhibitors and traditional non-steroidal anti-inflammatory drugs. Subsequently, an industry independent long term trial is needed, allocating patients to one of four agents: celecoxib, rofecoxib, diflofenac, or naproxen.

Stover suggests unfair testing of the COX 2 hypothesis in CLASS because of aspirin use in some patients. The figure (top) shows no consistent pattern distinguishing aspirin users from non-users, suggesting that CLASS comparisons were not confounded by aspirin use. According to the results of an in vitro analysis by Warner et al summarised in the figure (bottom), celecoxib and diclofenac actually have similar COX 2 selectivity. Therefore CLASS's failure to demonstrate celecoxib's superiority may have more to do with celecoxib's shortcomings as a COX 2 inhibitor. Not surprisingly, the US Food and Drug Administration refused to change celecoxib's labelling, which still states as of 7 June 2002 that “serious gastrointestinal toxicity such as bleeding, ulceration, and perforation can occur at any time.”4-7

Footnotes

We thank Yoav Ben-Shlomo for helpful comments, and Ben Calnan for data management and entering.

References

  • 4-1.Bombardier C, Laine L, Reicin A, Shapiro D, Burgos-Vargas R, Davis B, et al. Comparison of upper gastrointestinal toxicity of rofecoxib and naproxen in patients with rheumatoid arthritis. VIGOR Study Group. N Engl J Med. 2000;343:1520–1528. doi: 10.1056/NEJM200011233432103. , 2. [DOI] [PubMed] [Google Scholar]
  • 4-2.Mukherjee D, Nissen SE, Topol EJ. Risk of cardiovascular events associated with selective COX-2 inhibitors. JAMA. 2001;286:954–959. doi: 10.1001/jama.286.8.954. [DOI] [PubMed] [Google Scholar]
  • 4-3. Jüni P, Rutjes AW, Dieppe PA. Pharmacia addresses June 1 editorial regarding CLASS study: authors' response. bmj.com 2002: bmj.com/cgi/content/full/324/7349/1287 23766 (accessed 12 July 2002).
  • 4-4.Silverstein FE, Faich G, Goldstein JL, Simon LS, Pincus T, Whelton A, et al. Gastrointestinal toxicity with celecoxib vs nonsteroidal anti-inflammatory drugs for osteoarthritis and rheumatoid arthritis: the CLASS study: a randomized controlled trial. Celecoxib Long-term Arthritis Safety Study. JAMA. 2000;284:1247–1255. doi: 10.1001/jama.284.10.1247. [DOI] [PubMed] [Google Scholar]
  • 4-5.Warner TD, Giuliano F, Vojnovic I, Bukasa A, Mitchell JA, Vane JR. Nonsteroid drug selectivities for cyclo-oxygenase-1 rather than cyclo- oxygenase-2 are associated with human gastrointestinal toxicity: a full in vitro analysis. Proc Natl Acad Sci US. 1999;96:7563–7568. doi: 10.1073/pnas.96.13.7563. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4-6. Goldstein J, Eisen G, Bensen W, Agrawal N, Singh G, Pavelka K, et al. SUCCESS in osteoarthritis (OA) trial: celecoxib significantly reduces the risk of upper gastrointestinal (UGI) hospitalizations compared to diclofenac and naproxen in 13,274 randomized patients with OA. Abstract presented at EULAR 2001, European Congress of Rheumatology Available at: www.eular.org/eular2001/AbstractsOnline.cfm [Accessed 24 June 2002].
  • 4-7. Celebrex labeling, revised 7 June 2002. Available at: www.fda.gov/cder/foi/label/2002/20998s009lbl.pdf (accessed 28 June 2002)

Articles from The BMJ are provided here courtesy of BMJ Publishing Group

RESOURCES