Efficacy, tolerability, and upper gastrointestinal safety of celecoxib for treatment of osteoarthritis and rheumatoid arthritis: systematic review of randomised controlled trials
BMJ 2002; 325 doi: https://doi.org/10.1136/bmj.325.7365.619 (Published 21 September 2002) Cite this as: BMJ 2002;325:619All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
On eradicating publication bias, and the realities of independent reviews
Systematic reviews of drugs are plagued by publication bias. Many industry trials remain unpublished years after they are completed. Most reviewers cannot gain access to these trials, or even find out that they exist. Unless co-operation and trust between industry and systematic reviewers can be developed, full results of all trials will never become available, and reviews will be biased.
We undertook this review with the requirement that the companies involved provided reports of all randomised trials that might be of relevance, and we insisted that this be backed up by a signed legal pledge. As at the point of analysis (25th May 2000) celecoxib had not been made available for evaluation outside of the company, we can reasonably claim that we included all relevant randomised evidence. We believe that this does set a new standard for reviews of pharmaceutical products, and would suggest that others involved in industry-related reviews follow our lead and take a similar approach of obtaining a "complete data pledge" to entirely eradicate the possibility of publication bias.
As a result of this agreement, we obtained copies of three currently unpublished trials that recruited 3,158 participants. This constitutes 44% of the data available for the 12-week outcomes. Results of these trials are now published in the medical literature for the first time in our review. (In fact 63% of the 12-week data were unpublished when we submitted our review to the BMJ, as two additional trials were still in press).
The magnitude of this political achievement in the current climate, its importance in reducing the potential for bias, and the benefits of providing access to previously unpublished trial information are issues that our correspondents appear to have overlooked.
We are saddened that no one has considered it appropriate to applaud industry for taking such a large step towards increasing access to full trial information.
While we as academics would like to demand that industry surrender all data and reports without condition (Jüni), in reality the process of obtaining industry data is best approached in terms of co-operation rather than hostility. In response to our request to provide all trial reports, the companies involved very reasonably insisted that they were able to check our data extraction to ensure that we were not erroneously reporting their published and unpublished trials. Hence one co-author on our review (MB) is an employee of the company, and as declared in the paper, performed duplicate data extraction and checking. The work he undertook in double checking all trial data as well as persuading the company to make all full trial reports available and sign up to the complete data pledge cannot have done anything other than increase the validity of our review. Not all industry involvement is bad, nor does it necessarily compromise independence.
Correspondents have suggested that our review is flawed (Pechlaner), has serious validity problems (Jüni), does not add relevant knowledge (Formosa), and some insinuate that our review is biased through our contract with industry. We are criticised for not convincing readers that our review is thorough and that relevant results have been sought (Formosa). These comments all appear to have overlooked the significance of our achievement in obtaining a legally binding guarantee from industry that our review includes all relevant trials, and therefore can claim to be totally free from publication bias. This is a claim that only a handful of systematic reviews will ever be able to make, and is a clear marker of the level of rigour and efforts to avoid bias that we maintained when completing this review.
Jonathan J Deeks
Centre for Statistics in Medicine
Institute of Health Sciences
Old Road, Headington
Oxford OX3 7LF
Competing interests:
see paper
Competing interests: No competing interests
Study reports: clarification, corrections, and additional sources
Clarification of data sources
The systematic review was prepared from the full company trial reports, as was stated in the paper. As the full company trial reports are not publicly available, we attempted to match them up with journal publications so that interested readers of the review could follow links to the best publicly available trial evidence. Pfizer/Pharmacia have since suggested directly to us that this may have been misleading, and that it would have been more accurate to cite the company trial reports throughout as these were used for the review and not the journal publications cited in the references.
Notably, the criticisms made about CLASS study in the main relate to selective reporting in the journal publication and not problems in the trial itself (Black, Pechlaner, Metcalfe, Jüni, Formoso). It is always important to separate issues of trial quality from those of inadequate reporting. As we stated, our review was prepared from the data presented in the full company trial report, and not the JAMA publication. The criticisms relating to the company’s selective reporting in this publication do not affect our review – all data were available to us. We will respond in detail regarding our presentation of CLASS as the issues deserve detailed discussion and presentation of further data. However, we note that the CLASS trial is only included in 2 of the 17 outcomes summarised in the review. In our opinion Black, Pechlaner and Jüni are not justified in questioning the validity of the whole review based on problems they perceive with this single trial.
Corrections of references
During the matching process we made two minor errors, as pointed out by Jüni and also by Pfizer/Pharmacia, when matching publications with trials [2][4]. While Jüni is correct in suggesting that we have confused duplicate publications of two trials[1-4], all of our analyses are unaffected as we only have made errors with the names of publications attached to trials and not the data from the trials themselves. Each of the trials we included are distinct, and were originally identified by their FDA registered protocol numbers and not their journal publications.
Further published and unpublished sources
When checking these errors, we also came across new publications for two of the trials noted in the review as being unpublished [6][7]. We have also searched the FDA website, it having been brought to our attention by the BMJ editorial by Jüni, Rutjes and Dieppe. Here we have located the original licensing submission, medical, statistical and safety assessments for Celebrex which contain data from trials 20, 21, 22, 23, 41, 54, 62 and 71, as well as data on trials with shorter follow-up and for acute pain [9]. We have also found additional publicly available information about the CLASS trial, not previously cited by Jüni [10]. We would encourage those correspondents (Jenkinson, Wright, Formoso) who wanted more data than we were able to provide in our review to consult these sources.
The table and reference list below should provide clarification as to the company trial numbers, names and references for the trials included in the review, and website addresses where the FDA assessments of these trials can be found.
Company Trial ID |
Patients |
Comparator |
Related publications |
FDA website location |
|
20 |
Bensen 1999 |
OA |
Placebo Naproxen |
(1) (2) |
(9) |
21 |
Zhao 1999 |
OA |
Placebo Naproxen |
Abstract only* |
(9) |
22 |
Simon 1999 |
RA |
Placebo Naproxen |
(3) (4) |
(9) |
23 |
Zhao 2000 |
RA |
Placebo Naproxen |
Abstract only* |
(9) |
41 |
Emery 1999 |
RA |
Diclofenac |
(5) |
(9) |
54 |
Study 054 |
OA |
Placebo Naproxen |
(6) |
(9) |
62 |
Study 062 |
OA/RA |
Naproxen |
(7) |
(9) |
71 |
Study 071 |
OA/RA |
Diclofenac Ibuprofen |
Unpublished |
(9) |
35/102 |
Silverstein 2000 |
OA/RA |
Diclofenac Ibuprofen |
(8) |
(10) |
* the abstracts do not provide adequate data on these trials for meta-analysis |
Jonathan J Deeks
Senior Medical Statistician
Centre for Statistics in Medicine
Institute of Health Sciences
Old Road, Headington
Oxford
- Bensen WG, Fiechtner JJ, McMillen JI, Zhao WW, Yu SS, Woods EM et al. Treatment of osteoarthritis with celecoxib, a cyclooxygenase-2 inhibitor: a randomized controlled trial. Mayo Clin Proc 1999;74:1095-105.
- Zhao SZ, McMillen JI, Markenson JA, Dedhiya SD, Zhao WW, Osterhaus JT et al. Evaluation of the functional status aspects of health-related quality of life of patients with osteoarthritis treated with celecoxib. Pharmacotherapy 1999;19:1269-78.
- Simon LS, Weaver AL, Graham DY, Kivitz AJ, Lipsky PE, Hubbard RC et al. Anti-inflammatory and upper gastrointestinal effects of celecoxib in rheumatoid arthritis: A randomized controlled trial. Journal of the American Medical Association. 1999; 282:1921-8.
- Zhao SZ, Fiechtner JI, Tindall EA, Dedhiya SD, Zhao WW, Osterhaus JT et al. Evaluation of health-related quality of life of rheumatoid arthritis patients treated with celecoxib. Arthritis Care and Research. 2000; 13:112-21.
- Emery P, Zeidler H, Kvien TK, Guslandi M, Naudin R, Stead H et al. Celecoxib versus diclofenac in long-term management of rheumatoid arthritis: randomised double-blind comparison. Lancet 1999;354:2106-11.
- Kivitz AJ, Moskowitz RW, Woods E, Hubbard RC, Verburg KM, Lefkowith JB, Geiss GS. Comparative efficacy and safety of celecoxib and naproxen in the treatment of osteoarthritis of the hip. J Int Med Res 2001; 29: 467-79.
- Goldstein JL, Correa P, Zhao WW, Burr AM, Hubbard RC, Verburg KM, Geis GS. Reduced incidence of gastroduodenal ulcers with celecoxib, a novel cyclooxygenase-2 inhibitor, compared to naproxen, in patients with arthritis. Am J Gastroenterol 2001; 96:1018-27.
- 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. Journal of the American Medical Association 2000;284:1247-55.
- http://www.fda.gov/cder/approval/index.htm (search for celebrex and select report posted for NDA 20-998)
- http://www.fda.gov/cder/approval/index.htm (search for celebrex and select report posted for NDA 20-998/S9)
Competing interests:
see published paper
Competing interests: Name in BMJ review
Deeks et al’ systematic review (1) evaluated 12 weeks
gastrointestinal safety of coxibs by doing an original meta-analysis, and
26 weeks gastrointestinal safety by using data from the CLASS study. In
our opinion, this evaluation does not add relevant knowledge. While 26
weeks data from the CLASS study have been already (and repeatedly)
criticised (2), the need for 12 weeks data is questionable since patients
with osteoarthritis or rheumatoid arthritis, taking celecoxib for that
long, would probably keep on taking celecoxib for longer. Therefore 12-15
months data on 7968 patients published by FDA (3) (looking at the whole
study period of CLASS) are, in our opinion, more meaningful than 12 weeks
data on 2742 patients meta-analysed by the Deeks et al. FDA data clearly
demonstrates no difference in rates of complicated ulcers among celecoxib,
ibuprofen and diclofenac long-term users.
It would have been more interesting to know about short-term safety
(1-2 weeks), considering that, in real practice, coxibs are frequently
used in short courses (4) and have also been registered for acute diseases
(management of acute pain, primary dysmenorrhea). In the province of
Modena (632,000 inhabitants, Northern Italy) we found that 78% of patients
receiving coxibs were prescribed less than 60 Defined Daily Doses in the
whole 2001, suggesting a high prevalence of “acute” users. About long-term
use, the widespread concerns about coxibs’ cardiovascular safety (5) would
have deserved specific evaluations and probably an ad hoc meta-analysis,
since related data should be generally available from study investigators.
The importance of scientific studies may be questioned not only in
terms of systematic errors, but also of the relevance of the end-points
they looked at. When a research is industry-sponsored (as in this case)
concerns are more legitimate and Authors should make strong effort to
convince readers that their evaluation is thorough and that relevant
results have been sought. For the above mentioned reasons, we feel that
the paper by Deeks et al falls short of reassuring readers. Industry
independent studies and meta-analyses should be warranted to evaluate
short-term safety of coxibs and trade-offs between gastointestinal and
cardiovascular safety of these drugs in the long term.
Giulio Formoso, Epidemiologist†
Nicola Magrini, Director†
Anna Maria Marata, Head of Drug Evaluation Unit†
and Alessandro Liberati, Professor of Clinical Biostatistics‡ and
Scientific Director†
† Center for the Evaluation of the Effectiveness of Health Care,
Modena (Italy)
‡ University of Modena and Reggio Emilia (Italy)
REFERENCES
1. Deeks JJ, Smith LA, Bradley MD. Efficacy, tolerability, and upper
gastrointestinal safety of celecoxib for treatment of osteoarthritis and
rheumatoid arthritis: systematic review of randomised controlled trials.
BMJ 2002;325:619-26
2. Juni P, Rutjes AW, Dieppe PA. Are selective COX 2 inhibitors
superior to traditional non steroidal anti-inflammatory drugs? BMJ
2002;324:1287-8
3. Witter J. Medical officer review, celecoxib. Food and Drug
Administration, September 2000 (available at
http://www.fda.gov/ohrms/dockets/ac/01/briefing/3677b1_03_med.pdf -
accessed October 29, 2002)
4.Traversa G, Bianchi C, Pisetsky F, et al. New Anti-Inflammatory
Agents in General Medicine. BEN 2000.N.6 (available at
http://www.ben.iss.it/pre_2002/giugno02/1eng.htm - accessed October 29,
2002)
5. Therapeutics Letter, Issue 41, Jan 2002 (available at
www.ti.ubc.ca/pages/letter43.htm - accessed October 29, 2002)
Competing interests:
None declared
Competing interests: No competing interests
Dr. Juni's title "Absence of evidence versus evidence of absence" of
Oct 31 deserves to be gilded, mounted, and displayed somewhere. It's
genuinely and simply gorgeous!
Competing interests:
None declared
Competing interests: No competing interests
Phil Alderson, Associate Director of the UK Cochrane Centre,
immediately spotted a problem in our rapid response [1]. We stated in the
next to last paragraph that “the best available evidence to date (…)
shows, that in the long term celecoxib is no better than diclofenac in
avoiding severe gastrointestinal complications (RR for ulcer complications
1.10 [95% CI 0.47 to 2.58]).” Naturally, we agree with Phil Alderson that
the estimate accompanied by its wide confidence intervals do not provide
evidence that celecoxib is no more beneficial than diclofenac: “absence of
evidence is not evidence of absence” [2]. In fact, the relative risk’s
wide confidence intervals are compatible with celecoxib being either
considerably better or much worse than diclofenac. Therefore, we would
like to amend our statement to “there is no evidence suggesting that in
the long term celecoxib is more beneficial than diclofenac in avoiding
severe gastrointestinal complications.”
Peter Jüni MD, Senior Research Fellow in Clinical Epidemiology
Departments of Rheumatology, and Social and Preventive Medicine,
University of Berne, 3010 Berne, Switzerland (juni@ispm.unibe.ch)
Rebekka Sterchi, Research Associate
Departments of Rheumatology and Social and Preventive Medicine, University
of Berne, 3010 Berne, Switzerland
Paul Dieppe MD, Professor of Health Services Research
MRC Health Services Research Collaboration, Department of Social Medicine,
University of Bristol, Bristol BS8 2PR
Reference List
1. Jüni P, Sterchi R, Dieppe PA. Problems compromising the review's
validity. bmj.com 2002: http://bmj.com/cgi/eletters/325/7365/619#26567
(accessed 20 October 2002).
2. Altman DG, Bland JM. Absence of evidence is not evidence of
absence. BMJ 1995;311:485.
Competing interests:
None declared
Competing interests: No competing interests
We have had an interesting and mutually informative exchange on some
of the issues raised by the recent paper one of us co-authored [1] and the
other responded to [2], and are reporting that we have reached agreement
on all substantive issues.
Despite initial concern [2], the statistical calculations are
correct. The summary statistics are consistent when analysed by
alternative approaches and we both regard them as robust. The tabular
presentation issues are minor and arise from, for example, justified
exclusion of patients in the original trial analyses who were randomised
but never received a single drug dose.
We also agree on a suggestion to improve graphical display. In
forest plots that display results of multiple analyses, for example, in
the presentation of relative reductions in endoscopically detectable
ulcers in prophylactic aspirin and non-aspirin users (Figure 3 in the
printed paper, Figure 5 in the full paper) it is preferable to scale the
size of the points across analyses in a standard way so that they reflect
absolute weights given to each study, rather than the relative weights
within each analysis. If this had been done in Figure 3/5, it would have
been more notable at first glance that the number of patients exposed to
prophylactic aspirin in the trials is low. It is not that any data in
this figure are wrong, but that the figure may at first give a misleading
impression that the data are equally distributed between aspirin and non-
aspirin groups. This impression is corrected when one considers the width
of the confidence intervals and sample sizes also stated in the figure.
We also note that it is necessary to make assumptions when performing
meta-analysis, as with most statistical analyses. Importantly, when there
is variation in event rates between trials (as with the data in Figure
3/5) one must assume that the treatment effect can be expressed by a
particular summary statistic which is consistent at all event rates.[3]
Alternative choices of statistic do not change the direction or
significance of effects for individual trials, but may lead to differences
in treatment effects, observed heterogeneity and confidence intervals for
the aggregated data.
Reductions in rates of endoscopically detectable ulcers in
participants not taking prophylactic aspirin (Figure 3/5) are
heterogeneous regardless of how the data are analysed, although all 5
trials yield estimates in the direction of benefit. In such situations it
is desirable to know why the trials give different estimates.[4] However,
with only five data points a multitude of hypotheses can be generated that
fit the data, but there is no objective way of ascertaining which is the
real cause. Further studies and analysis of individual patient data may
help.
The use of "random effects" models for analyses displaying
heterogeneity, as used in the review, ensures that the heterogeneity is
quantified and some allowance made for it.[5] While the random effects
model makes assumptions which are contentious and difficult to assess, in
the review it has resulted in estimates of effect that are more
conservative (less likely to be significant) that those obtained by using
the alternative "fixed effect" model.
In conclusion, we both agree that more data on the occurrence of
gastrointestinal events in patients combining aspirin with a Cox-2
inhibitor is required to inform the debate as there are difficulties in
knowing the importance of the observed differences in outcomes that are
only surrogate measures of safety.
Michael L Jenkinson
Chair East Kent Hospitals Trust Drugs & Therapeutics Committee
Queen Elizabeth, The Queen Mother Hospital
St. Peters Road,
Margate,
Kent CT9 4AN
michael.jenkinson@ekht.nhs.uk
Jonathan J Deeks
Senior Medical Statistician
Centre for Statistics in Medicine,
Institute of Health Sciences,
Old Road, Headington,
Oxford OX3 7LF
Jon.Deeks@cancer.org.uk
[1] Deeks JJ, Smith LA, Bradley MD. Efficacy, tolerability and upper
gastrointestinal safety of celecoxib for treatment of osteoarthritis and
rheumatoid arthritis: systematic review of randomised controlled trials.
BMJ 2002; 325: 619-23.
[2] Jenkinson ML. On presentation and analysis. BMJ Rapid Response:
7th October 2002. http://bmj.com/cgi/eletters/325/7365/619#26091
[3] Deeks JJ. Issues in the selection of a summary statistic in meta
-analysis of clinical trials with binary outcomes. Statistics in Medicine
2002; 21(11): 1575-1600.
[4] Thompson SG. Why sources of heterogeneity in meta-analysis
should be investigated. BMJ 1994; 309: 1351-5.
[5] Thompson SG, Pocock SJ. Can meta-analysis be trusted? Lancet
1991; 338:1127-30.
Competing interests:
None declared
Competing interests: No competing interests
Dear Editor
The systematic review by Deeks et al [1] suggests that the COX-2
inhibitor celecoxib has significantly improved gastrointestinal safety and
tolerability as compared with traditional non-steroidal anti-inflammatory
drugs (NSAIDS). We have several concerns regarding this review.
First, Deeks et al reported the papers by Bensen et al [2], Zhao et
al (1999) [3], Simon et al [4] and Zhao et al (2000) [5] as if they
referred to four different trials. However, the papers by Bensen et al and
Zhao et al (1999) [2;3] were merely duplicate reports of one trial, while
the papers by Simon et al and Zhao et al (2000) [4;5] reported in
duplicate on another trial. Deeks et al either included the same data more
than once or mixed up data from unpublished trials with unrelated
published papers.
Second, Deeks et al erroneously report similar relative risks (RRs)
of ulcer complications after six months observed in the two trials
included in CLASS [6]: a RR of 0.54 (95% CI 0.20 to 1.47) for Study 035
comparing celecoxib to ibuprofen, and a RR of 0.56 (95% CI 0.19 to 1.66)
for Study 102 comparing celecoxib to diclofenac. Not surprisingly, Deeks
et al’s corresponding meta-analysis showed homogeneous effects across
trials (Q=0), implying that it is perfectly reasonable to pool CLASS’ two
trials using comparator drugs of different COX-2 selectivity. However,
according to the information available on the FDA website [7;8] four
events occurred in the celecoxib group and 11 in the ibuprofen group in
Study 035, leading to a RR of 0.36 (95% CI 0.12 to 1.14), whereas seven
events occurred in the celecoxib group and nine in the diclofenac group in
Study 102, leading to a RR of 0.78 (95% CI 0.29 to 2.08) [9]. The
heterogeneity of the corresponding meta-analysis (Q=1 on 1 degree of
freedom) reflects that pooling different trials using comparator drugs of
different COX-2 selectivity may be inappropriate.
Third, Deeks et al’s justification for only considering CLASS’ six
months results, adopted from earlier invalid arguments provided by
Silverstein et al [10] and Geis et al [11], is problematic [9;12].
Admittedly, data available on the FDA website [7] suggest that there were
differences regarding patient withdrawal between celecoxib and ibuprofen
in one of CLASS’ trials (Study 035) from the beginning, indicating that
results for this trial were unreliable at all time points, including 6
months. In accordance with Deeks et al’s arguments this trial should,
therefore, have been excluded from all analyses. Contrary to Deeks et al’s
statement, data available on the FDA website [7] suggest, however, that
there were no relevant differences between celecoxib and diclofenac groups
in CLASS’ second trial (Study 102), neither for 6 months nor for the
entire follow-up duration. Unfortunately, withdrawal rates were not
reported separately for the two trials, therefore we are unable to
calculate exact RRs for withdrawal at the different time points. It is
clear, however, that the differences in treatment duration between
celecoxib and diclofenac groups reported by Deeks et al [1] merely relate
to the fact that half of the patients on celecoxib (i.e. those allocated
to celecoxib in Study 035) had a potential maximum treatment duration of
15 months, while those allocated to diclofenac in Study 102 had a
potential maximum treatment duration of only 12 months [12].
In our experience, patients suffering from osteoarthritis or
rheumatoid arthritis generally take anti-inflammatory drugs for years.
Therefore, short term results of the sort reported by Deeks et al [1] are
misleading. The best available evidence to date [9;12] shows, that on
long term celecoxib is no better than diclofenac in avoiding severe
gastrointestinal complications (RR for ulcer complications 1.10 [95% CI
0.47 to 2.58]). However, celecoxib may be about 20 times more expensive
than diclofenac.
These problems cast serious doubts on the validity of Deeks et al’s
systematic review [1]. We therefore suggest that celecoxib’s manufacturer
(Pharmacia/Pfizer) allows an industry-independent meta-analysis of all
published and unpublished randomised controlled trials by providing
unrestricted access to all available data to a team of independent
researchers.
Conflicts of interest: none
Peter Jüni MD, Senior Research Fellow in Clinical Epidemiology
Departments of Rheumatology, and Social and Preventive Medicine,
University of Berne, 3010 Berne, Switzerland (juni@ispm.unibe.ch)
Rebekka Sterchi, Research Associate
Departments of Rheumatology, and Social and Preventive Medicine,
University of Berne, 3010 Berne, Switzerland
Paul Dieppe MD, Professor of Health Services Research
MRC Health Services Research Collaboration, Department of Social Medicine,
University of Bristol, Bristol BS8 2PR
Reference List
1. Deeks JJ, Smith LA, Bradley MD. Efficacy, tolerability, and upper
gastrointestinal safety of celecoxib for treatment of osteoarthritis and
rheumatoid arthritis: systematic review of randomised controlled trials.
BMJ 2002; 325(7365):619-623.
2. Bensen WG, Fiechtner JJ, McMillen JI, Zhao WW, Yu SS, Woods EM, Hubbard
RC, Isakson PC, Verburg KM, Geis GS. Treatment of osteoarthritis with
celecoxib, a cyclooxygenase2 inhibitor: a randomized controlled trial.
Mayo Clin Proc 1999; 74:1095-1105.
3. Zhao SZ, McMillen JI, Markenson JA, Dedhiya SD, Zhao WW, Osterhaus JT,
Yu SS. Evaluation of the functional status aspects of health-related
quality of life of patients with osteoarthritis treated with celecoxib.
Pharmacotherapy 1999; 19(11):1269-1278.
4. Simon LS, Weaver AL, Graham DY, Kivitz AJ, Lipsky PE, Hubbard RC,
Isakson PC, Verburg KM, Yu SS, Zhao WW, Geis GS. Anti-inflammatory and
upper gastrointestinal effects of celecoxib in rheumatoid arthritis: a
randomized controlled trial. JAMA 1999; 282(20):1921-1928.
5. Zhao SZ, Fiechtner JJ, Tindall EA, Dedhiya SD, Zhao WW, Osterhaus JT,
Yu SS. Evaluation of health-related quality of life of rheumatoid
arthritis patients treated with celecoxib. Arthritis Care Res 2000; 13:112
-121.
6. Silverstein FE, Faich G, Goldstein JL, Simon LS, Pincus T, Whelton A,
Makuch R, Eisen G, Agrawal NM, Stenson WF, Burr AM, Zhao WW, Kent JD,
Lefkowith JB, Verburg KM, Geis GS. 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(10):1247-1255.
7. Lu HL. Statistical reviewer briefing document for the advisory
committee. Available at:
http://www.fda.gov/ohrms/dockets/ac/01/briefing/3677b1_04_stats.doc.
[Accessed December 10, 2001.]
8. Witter J. Medical officer review. Available at:
http://www.fda.gov/ohrms/dockets/ac/01/briefing/3677b1_03_med.pdf.
[Accessed December 10, 2001.]
9. Jüni P, Rutjes AW, Dieppe P. Are selective COX 2 inhibitors superior to
traditional NSAIDs? Authors' reply. BMJ 2002; 325:163-164.
10. 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.
11. Geis GS. Are selective COX 2 inhibitors superior to traditional
NSAIDs? Pharmacia's response to editorial. BMJ 2002; 325:161-162.
12. Jüni P, Rutjes AW, Dieppe PA. Gastrointestinal ulcer complications:
Are selective COX-2 inhibitors superior to traditional NSAIDs? Adequate
analysis of the CLASS trial indicates that this may not be the case. BMJ
2002; 324:1287-1288.
Competing interests: No competing interests
Deeks et al.[1] focused their systematic review on the
gastrointestinal adverse effects of celecoxib and other non-steroidal anti
-inflammatory drugs (NSAIDs), and thereby missed the opportunity to inform
physicians and patients about the overall safety of celecoxib. It should
be increasingly apparent that all NSAIDs particularly when taken long-term
cause serious adverse events and that these events are clearly not limited
to the gastrointestinal system. NSAIDs increase blood pressure, cause
renal failure precipitate congestive heart failure, accelerate joint
degradation, prevent fracture healing, etc. [2]. The critically important
observation for patients from the CLASS and VIGOR trial data is that total
serious adverse events were increased for both celecoxib and rofecoxib
[2,3]. The fact that this increase was statistically significant for
rofecoxib and not quite statistically significant for celecoxib should not
dissuade anybody from concluding that when a patient is switched from a
non-selective NSAID to a COX-2 selective NSAID, they are probably at
increased risk of experiencing a serious adverse event. In other words
COX-2 selective NSAIDs likely cause more harm than benefit as compared to
non-selective NSAIDs.
Deeks and his two co-authors, one from celecoxib’s manufacturer,
certainly have access to the serious adverse event data from all the
trials that they reviewed. It is an international regulatory requirement
that all serious adverse events are collected and reported in clinical
trials. Serious adverse events include all deaths, unscheduled
hospitalizations, prolongations of hospitalization, other life-threatening
events (eg. cancer diagnosis) or events leading to serious disability.
Before prescribing a COX-2 selective NSAID to my patients, I want to
know that the risk of total serious adverse events are decreased as
compared to non-selective NSAIDs. If, in fact, the selective NSAIDs
increase the risk (as appears to be the case), or even if they decrease
the risk of gastrointestinal serious adverse events and this is cancelled
by an equal increase in nongastrointestinal serious adverse events, I
would not prescribe them. I call upon Deeks et al to provide the serious
adverse event data from all the trials in their systematic review. The
fact that they haven’t provided it, certainly makes me suspicious that the
data do not look good for celecoxib.
Competing interests: None declared.
References:
1. Deeks JJ, Smith LA, Bradley MD. Efficacy, tolerability, and upper
gastrointestinal safety of celecoxib for treatment of osteoarthritis and
rheumatoid arthritis: a systematic review of randomized controlled trials.
BMJ. 2002;325:619-623.
2. Wright JM. The double-edged sword of COX-2 selective NSAIDs. Can Med
Assoc J. 2002;167 (in press).
3. Bassett K, Wright JM, Puil L, Perry TL Jr,, Heran B, Cole Carol,.
Cyclooxygenase-2 inhibitor update: Journal articles fail to tell the full
story. Can Fam Physician. 2002;48:1455-1460.
Competing interests: No competing interests
Celebrex’s relative GI
safety is overstated?
The
editorial by Roger Jones[1]
makes important points about the limitations of the meta-analysis by Jon Deeks
and colleagues[2] for
celecoxib.However, we also note that
the Deeks meta-analysis does not account for the 12-15-month data for the CLASS
study compiled by the FDA[3][4]
and cited by Peter Juni and colleagues’ critical editorial.[5]We have abstracted the 12-15 month CLASS
data and applied them to the Deeks analysis, and find that these give a
different picture.
Firstly,
the FDA’s analysis of CLASS is more complete than that published in JAMA[6],
which the Juni editorial criticised for not accounting for the 12-15 month
data.We believe Deeks and colleagues
offer an unconvincing explanation for limiting the analysis to the six-month
follow up for CLASS, insufficient to justify the post hoc changes in design,
outcomes and analysis.We also believe
the 12-15 month CLASS data could have been included in figure 2 of the Deeks
paper for adverse GI effects, materially affecting those results.
For
withdrawals because of both serious upper GI events and endoscopic ulcers, the
12-15 month FDA data for CLASS showed no significant reduction in risk(relative risk (RR) 0.73 (95% CI 0.50 to
1.05)), distinct from the 39% RRR for CLASS’s 6-month data in Deeks figure
4.
Combining
the 12-month CLASS GI withdrawal data with the seven RCTs in Deeks fig 2, and
then adjusting for the longer exposure experienced in CLASS (12-15 months, not
12 weeks) decreased the overall relative risk reduction (RRR) to 32%- somewhat less than that reported in Deeks
fig 2.
These
results suggest that although celecoxib still caused statistically significant
reductions in GI adverse events overall, these reductions were appreciably less
than those suggested for the seven other RCTs by Deeks fig 2.
Secondly,
Deeks et al reported no statistically significant difference between low-dose
aspirin and non-aspirin use for both endoscopic ulcers and for CLASS.However, using the 12-15-month data for
CLASS suggests that whereas non-aspirin users had a statistically significant
42% RRR, aspirin users showed no reduction in risk (RR 1.02 (0.59 -
1.74)).The difference between the
subgroups’ RRRs over the 12 months was statistically significant (p 0.03).
Taken in
entirety (combining both endoscopic ulcers with CLASS’s GI withdrawals+ulcers)[7],
the above significant differences between subgroups persist.Including the 12-15-month CLASS data gave a non-significant
28% RRR for aspirin use, compared with a 72% RRR for non-aspirin use, the
difference between RRRs being statistically significant.Extending the analysis to adjust for the
greater exposure conferred by CLASS gave again a non-significant 4% RRR
for aspirin users (exposure/variance-weighted RR 0.96 (0.63 - 1.46), versus 52%
for non-aspirin use, p <0.01.
Hence we
disagree with the implication that the benefits of celecoxib extend equally to
aspirin users, and agree with NICE’s current precautionary recommendation to
withhold celecoxib from aspirin users.
Further
details of this analysis can be found on PHARMAC’s website at www.pharmac.govt.nz publications
page.In summary, any results presented
for celecoxib suggesting favourable GI safety need careful scrutiny.
References/Endnotes
[1] Jones R. Efficacy
and safety of COX 2 inhibitors. BMJ 2002;325:607-608. http://bmj.com/cgi/content/full/325/7365/607
[2] Deeks JJ, Smith
LA, Bradley MD. Efficacy, tolerability, and upper gastrointestinal safety of
celecoxib for treatment of osteoarthritis and rheumatoid arthritis: systematic
review of randomised controlled trials. BMJ 2002; 325: 619-623
http://bmj.com/cgi/content/full/325/7365/619
[3]Witter J. Medical
officer review. www.fda.gov/ohrms/dockets/ac/01/briefing/3677b1_03_med.pdf
[4] Lu HL.
Statistical reviewer briefing document for the advisory committee. www.fda.gov/ohrms/dockets/ac/01/briefing/3677b1_04_stats.doc
[5] Juni P, Rutjes
AWS, Dieppe PA. Are selective COX-2 inhibitors superior to traditional
non-steroidal anti-inflammatory drugs? BMJ 2002; 324: 1287-1288. http://bmj.com/cgi/content/full/324/7349/1287
[6] 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 randomised
controlled trial. JAMA 2000; 284: 1247-1255.
[7] Numbers of CLASS
withdrawals were comparatively low when compared with more sensitive GDUs found
on 12-week mandatory endoscopy in the four other RCTs. Numbers of ulcers
detected by routine endoscopy at 12 weeks reported in Deeks et al figure 5 (25%
control incidence) were considerably higher than numbers of withdrawals because
of adverse GI effects for corresponding RCTs reported in Deeks figure 2 (6%)
and in CLASS (1.6%). Hence combining the two sets of data understates ulcer
burden occurring in CLASS.
Competing interests: No competing interests
Re: On presentation and analysis
On presentation and analysis
In my original rapid response to the paper by Deeks et al [1] I
commented that " This is important as this data as presented, could be
used for promotional purposes to recommend that patients on aspirin and a
NSAID should have their NSAID changed to celecoxib" [2]. There has been a
subsequent joint letter by Deeks and myself agreeing that the graphical
presentation of the data on this issue could have been improved [3]. Other
correspondence has also commented on the limitations of the data set
analysed by Deeks et al and that analysis of more long term data does not
confirm their conclusions [4]. Today at an educational meeting in my
hospital with a pharmaceutical company stand, a reprint of Deeks et al
paper was available. It was offered to another doctor in my presence. The
discussion between the pharmaceutical representative and the other doctor
did not mention any of the reservations expressed by correspondents until
I interrupted the conversation when the company representative appeared to
be encouraging the co-prescription of aspirin and celecoxib.
Michael L Jenkinson
Chair East Kent Hospitals Trust Drugs & Therapeutics Committee
Queen Elizabeth, The Queen Mother Hospital
St. Peters Road,
Margate,
Kent CT9 4AN
michael.jenkinson@ekht.nhs.uk
Competing interests:
None declared
Competing interests: No competing interests