Healthy Skepticism Library item: 2218

Warning: This library includes all items relevant to health product marketing that we are aware of regardless of quality. Often we do not agree with all or part of the contents.

Publication type: news

Reporting of 6 Month vs 12 Month Data in a Clinical Trial of Celecoxib
JAMA 2001 Nov 21
http://jama.ama-assn.org/cgi/content/extract/286/19/2398

Full text:

To the Editor: The authors of the CLASS trial1 reported that celecoxib caused fewer symptomatic ulcers and ulcer complications than did diclofenac or ibuprofen at 6 months of follow-up.1 We are concerned that subsequent information from the trial, which is available on the US Food and Drug Administration (FDA) Web site,2 appears to contradict these conclusions. As described on the FDA Web site, the published CLASS trial differs from the original protocol in primary outcomes, statistical analysis, trial duration, and conclusions.2-4 In particular, the unpublished data show that by week 65, celecoxib was associated with a similar number of ulcer complications as diclofenac and ibuprofen.2

Scientists at the FDA were concerned that COX-2 inhibition would interfere with the beneficial role of COX-2 in the final stages of ulcer healing and actually increase the number of complicated ulcers that occur without symptomatic warning.5 Therefore, the sponsor specifically listed complicated ulcers (ie, perforation, obstruction, or significant bleeding) as the primary study outcomes for the CLASS trial.4 Although complicated ulcers were the primary outcome in documents submitted to the FDA, the published study also included symptomatic ulcers.1

The original study protocol called for inclusion of 40 ulcer complications in the 2 studies combined. Patients were subsequently enrolled for a range of 6 to 13 months before a total of 44 complicated ulcers occurred.4 However, the published study truncated data from both studies at 6 months. The FDA’s statistical reviewer stated that “this reviewer regards the analysis for data for the entire study period as specified in the protocol, which includes most information, the appropriate analysis.“2

The overall findings of the trial, according to the study protocol, were to be statistically significant before exploratory analyses on subgroups would be conducted. This stepwise procedure was not followed, and the authors made conclusions about the effect of celecoxib in patients taking aspirin despite the lack of significance in the primary study outcome. The FDA Web site indicates that in patients not taking aspirin, diclofenac and celecoxib appear to have similarly low gastrointestinal (GI) toxicity compared with ibuprofen. On the other hand, the data also suggest that patients taking ibuprofen have about the same GI risk regardless of whether or not they are also taking aspirin. This apparent aspirin effect may be a chance finding resulting from multiple statistical comparisons.

The results of these unpublished data may not be widely known by physicians who prescribe celecoxib. The complete study may never be published because 1 version of the study already has been. We believe that the complete findings of the CLASS trial should be made widely known.

Jennifer Berg Hrachovec, PharmD, MPH
Marc Mora, MD
Pharmacy and Therapeutics Committee
Group Health Cooperative
Seattle, Wash

1. Silverstein FE, Faich G, Goldstein JL, et al, for the Celecoxib Long-Term Arthritis Safety Study. 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. ABSTRACT | FULL TEXT | PDF | MEDLINE

2. 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 July 25, 2001.

3. Goldkind L. Gastrointestinal Highlights of the CLASS Study. Available at: http://www.fda.gov/ohrms/dockets/ac/01/slides/
3677s1_03_goldkind/. Accessibility verified October 18, 2001.

4. Witter J. Medical Officer Review. Available at: http://www.fda.gov/ohrms/dockets/ac/01/briefing/
3677b1_03_med.pdf. Accessibility verified October 18, 2001.

5. US Food and Drug Administration. Transcript of the Arthritis Advisory Committee. Available at: http://www.fda.gov/ohrms/dockets/ac/01/transcripts/3677t1.rtf. Accessibility verified October 18, 2001.

To the Editor: Cyclooxygenase-2 inhibitors such as celecoxib (Celebrex) are generally presumed to have a lower risk of causing complicated peptic ulcers and other serious GI toxicity than other nonsteroidal anti-inflammatory drugs (NSAIDs).

The full data from the CLASS trial,1 which are publicly available on the FDA Web site,2 call these beliefs into question.3 The FDA found that “For upper GI safety, and also for global safety, there does not appear to be any meaningful advantage for Celebrex.“4

However, the published report of the CLASS trial draws the opposite conclusion.1 This is because the published article represents selective reporting of the combined analysis of only the first 6 months of 2 separate protocols of longer duration: the first a 12-month trial comparing celecoxib with diclofenac, and the other a 16-month trial comparing celecoxib with ibuprofen.2

The FDA Web site contains hundreds of pages of data and analysis, but nowhere does it separately report the results as 2 distinct trials. Furthermore, the relative risks of both complicated ulcers and all serious adverse events are higher at 12-16 months than they are at 6 months.

The trend toward an increased risk of serious adverse events, particularly with celecoxib long-term therapy, is particularly concerning. The unfortunate result of the selective and partial reporting of the CLASS study is that it could mislead physicians and patients. Until there is a better understanding of the risk of serious adverse events with COX-2 selective drugs, these drugs should be prescribed with caution.

James M. Wright, MD, PhD
Thomas L. Perry, MD
Departments of Pharmacology and Therapeutics and Medicine

Ken L. Bassett, MD
Department of Family Practice

G. Keith Chambers, MD
Department of Health Care and Epidemiology
University of British Columbia
Vancouver

2. Witter J. Medical Officer Review. Available at: http://www.fda.gov/ohrms/dockets/ac/01/briefing/3677b1_03_med.pdf. Accessibility verified October 18, 2001.

3. US Food and Drug Administration. Celebrex capsules (celecoxib) NDA 20-998/S-009Medical Officer Review. 2000. Available at: http://www.fda.gov/ohrms/dockets/ac/01/briefing/3677b1.htm. Accessibility verified October 18, 2001.

4. FDA Panel finds no safety benefit for Celebrex. Scrip World Pharm News. February 9, 2001;No. 2616:19.

In Reply: Three committees of outside investigator-consultants were established to evaluate each possible GI complication, to ensure patient safety and ethical conduct, and to provide overall direction for the CLASS study. Every suspected GI complication was individually and carefully reviewed in a blinded fashion by committee members consisting of experienced gastroenterologists. The CLASS data were carefully reviewed by the members of each of the oversight committees who had full access to the 6-month data, the longer-term data, and the analyses of these data. All of these committee members agreed that the data we submitted to JAMA were valid, the analyses appropriate, and the results clinically interpretable.

In retrospect, we acknowledge that we could have avoided confusion by explaining to the JAMA editors why we chose to inform them only of the 6-month analyses, and not the longer-term data that were available to us when we submitted the manuscript. We submitted only this information because the authors believed the 6-month data were the most scientifically and clinically valid. The data after 6 months were so confounded as to be difficult to interpret for assessing a drug-related causal GI toxicity.

The problem after 6 months resulted from the differential loss of susceptible patients that occurred between the groups of the study, which confounded comparisons between the treatment groups. This is because the treatment groups had changed fundamentally after 6 months in several ways. First, it was apparent before the data were unblinded that more than 50% of the patients dropped out. Second, aspirin was used by more than 20% of patients, twice the percentage anticipated, and this likely was a confounding factor. Third, the accrual of new ulcer complications had markedly decreased after 6 months, perhaps because of depletion of susceptible patients. Fourth and most important, after the blind was broken, it became clear that there was a differential dropout rate of NSAID patients with GI intolerance or symptomatic ulcers, suggesting that those patients at greatest risk were no longer in the study. This type of informative censoring leads to a bias, which potentially invalidates statistical analysis of complicated ulcers by the log rank test.1

This all implies that the data after 6 months are not valid for examining drug-induced GI toxicities. Based on these concerns, the authors concluded that the 6-month analyses of complicated ulcers and symptomatic ulcers were less subject to bias and would reflect true drug effects and therefore should be the basis of the report of the trial. Moreover, analyses of the differences between the effects of celecoxib and the comparator NSAIDs for the longer-term follow-up data for the combined end point of complications and symptomatic ulcers, for all and non–aspirin-using patients, do not substantially differ from the 6-month analysis as shown in Tables 26 and 30 of the FDA Web site.2 We believe that the differences found between the 6-month and longer-term follow-up data for complicated ulcers (Tables 13 and 14 of the FDA Web site2) are likely due to the biases we mentioned.

In response to other issues raised by Drs Hrachovec and Mora and Dr Wright:

The study was conducted under 2 protocols that prespecified that the data would be pooled for the analyses. The primary end point was ulcer complications and the design required that the study would continue until 40 ulcer complications were accrued between the 2 trials, and also required that all patients participated for at least 6 months unless discontinued from the study due to lack of efficacy or an adverse event. The protocol did not predefine the time period for the data lock point for the analysis.

A particular regression procedure (model selection stepdown) was appropriately applied to the comparison of celecoxib to ibuprofen and diclofenac combined and then individually for GI outcomes. However, this procedure was not applied in the secondary analysis of each of the risk factors for poor outcome such as use of aspirin.

After adjusting for duration of exposure, the relative risk of all serious adverse events with celecoxib compared with NSAIDs at 12 to 16 months was comparable to the relative risk at 6 months.

Fred Silverstein, MD
Department of Medicine
University of Washington
Seattle

Lee Simon, MD
Harvard Medical School
Boston, Mass

Gerald Faich, MD
Pharmaceutical Safety Assessment, Inc.
Narberth, Pa

1. Yola M, Lucien A. Evidence of the depletion of susceptibles effect in non-experimental pharmacoepidemiologic research. J Clin Epidemiol. 1994;47:731-737. MEDLINE

2. Witter J. Medical Officer Review. Available at: http://www.fda.gov/ohrms/dockets/ac/01/briefing/3677b1_03_med.pdf. Accessibility verified October 18, 2001.

Financial Disclosure: Drs Silverstein and Faich consult for Pharmacia; and Dr Simon is on the speakers bureau for Merck and is a consultant and speaker for Pfizer and Pharmacia.

Healthy Skepticism Library item: 2218

Announcements