Rapid Responses to:
|
|
Rapid Responses: Rapid Responses published:
-
![[Read Rapid Response]](/icons/misc/sectionDOWN.gif)
Celecoxib causes excess cardiovascular events and illustrates the folly of allowing drug sellers too much power.
- Peter R Mansfield (11 September 2005)
-
The evidence is scientifically inconclusive
- Jeffrey Mann (12 September 2005)
-
Response to Jeffrey Mann: Clinical trials never produce certainty
- Peter R Mansfield (17 September 2005)
|
|
|||
|
Peter R Mansfield, Director, Healthy Skepticism 34 Methodist St, Willunga SA 5172, Australia
Send response to journal:
|
The perils of allowing drug sellers to control or influence research, education and promotion are illustrated by lack of knowledge of the harms of celecoxib. It has been claimed that there is not enough evidence to conclude that celecoxib causes more cardiovascular adverse effects than less selective NSAIDs.<1> To test that claim I did a quick meta-analysis of randomized controlled trials of celecoxib vs less selective NSAIDs. Methods A priori exclusion criteria were: a) comparisons with diclofenac because it is almost as COX-2 selective as celecoxib<2> and b) concomitant use of aspirin, where data available, because that could obscure a real difference. The US FDA and Clinical Study Results websites were searched. Review Manager 4.2.8 software was used for the analysis. Results I found data for 3 trials. Data for a sub-group who did not take concomitant aspirin was available for one trial: CLASS. Trial Events/N CLASS non-aspirin<3> celecoxib 28/3105, ibuprofen 6/1573 Success-1<3> celecoxib 61/8800, naproxen 4/905 Protocol No. F49-98-02-137<4> celecoxib 1/80, ketoprofen 0/90 The odds ratio for all types of cardiovascular events (including deaths) with celecoxib vs less selective NSAIDs was 2.02 (95% CI 1.06 - 3.92) Conclusions The available evidence suggests that celecoxib causes more cardiovascular events than non-selective NSAIDs but does not give a precise estimate of the magnitude of additional harm. This information was available to the manufacturer, Pfizer, in 2001 but to my knowledge has not been published previously. Governments and the public should pay for research directly via publicly accountable competitive tender rather than via over-pricing for drugs.<5> 1. Elevated cardiovascular risk with NSAIDs? NPS RADAR June 2005 www.npsradar.org.au/site.php?page=1&content=/npsradar/content/nsaids.html (Accessed 11 Sep 2005) 2. Selective COX-2 inhibitors: Are they safer NSAIDs? Therapeutics Letter 2001 Jan-Feb; (39) http://www.ti.ubc.ca/pages/letter39.htm (Accessed 11 Sep 2005) 3. Witter J. COX-2 CV Safety celecoxib. Center for Drug Evaluation and Research DAAODP/ODE V February 16, 2005 http://www.fda.gov/ohrms/dockets/ac/05/slides/2005-4090S1_07_FDA-Witter- Celebrex.ppt (Accessed 11 Sep 2005) 4. A Multicenter, Double-Blind, Placebo-Controlled Comparison, Randomized Study to Evaluate the Short-Term Efficacy and Safety of Celecoxib 100 mg BID and Ketoprofen 100 mg BID in Patients With Ankylosing Spondylitis 01000004960706\ 1.1 \ Approved\ 23- Jun- 2005 11: 03 http://www.clinicalstudyresults.org/documents/company-study_81_0.pdf (Accessed 11 Sep 2005) 5. Mansfield PR. Healthy Skepticism about drug promotion: Memorandum for the UK House of Commons Health Committee Inquiry: The Influence of the Pharmaceutical Industry. Healthy Skepticism International News 2004 Oct; 22(10-12) http://www.healthyskepticism.org/news/issue.php?id=6 (Accessed 11 Sep 2005) Competing interests: I am the Founder of Healthy Skepticism Inc which aims to improve healthy by reducing harm from misleading drug promotion. www.healthyskepticism.org |
|||
|
|
|||
|
Jeffrey Mann, Retired physician Salt Lake City, UT 84103
Send response to journal:
|
Peter Mansfield states that celecoxib causes excess cardiovascular events and he bases his conclusions on an arbitrary selection of studies where the control event rate and experimental event rate are in the region of 0.5-2% [1]. I don't believe that it is easily possible to obtain scientifically conclusive results from RCTs where the control event rate is so low -- if the sample size is so small (<10,000 patients). I used a Clinical Trial Simulator tool to demonstrate that RCTs that have such a low control event rate (0.5-2.0%) cannot generate scientifically conclusive results (high signal/noise ratio RCT with narrow 95%CIs) unless the sample size is very large (>100,000 patients) [2]. I believe that the clinical research community seems to be oblivious of the fact that low control event rate (0.5-2.0%) RCTs are generally too underpowered to produce scientifically conclusive results. In the first US court case involving Vioxx, the newspaper reporters stated that the evidence from randomised controlled trials strongly suggests that Vioxx doubles the risk of cardiovascular events. However, my analysis of the evidence suggests that the relevant rofecoxib RCTs had too low a control event rate and too low a signal/noise ratio to produce scientifically conclusive results, and the evidence can only be rationally regarded as being scientifically inconclusive [3]. Jeff Mann. MD jmannemg@earthlink.net References: 1.Peter Mansfield. Celecoxib causes excess cardiovascular events and illustrates the folly of allowing drug sellers too much power. bmj rapid response letter. 11 September 2005. Available at http://bmj.bmjjournals.com/cgi/eletters/331/7516/528#116293 2. Mann J. Quantifying the potential magnitude of chance event noise in randomised controlled trials. Available at http://jeffmann.net/soapbox/chanceevents.htm Adobe pdf version available at http://jeffmann.net/soapbox/chanceevents.pdf 3. Mann J. Questioning the scientific validity of randomised trials of COX-2 inhibitors showing an increased risk of adverse cardiovascular events. Available at http://jeffmann.net/soapbox/vioxx-cox2critique.htm Adobe pdf version available at http://jeffmann.net/soapbox/vioxx- cox2critiqueadobe.pdf Competing interests: None declared |
|||
|
|
|||
|
Peter R Mansfield, Director, Healthy Skepticism 34 Methodist St, Willunga SA 5172, Australia
Send response to journal:
|
I accept that the title of my rapid response should have had two extra words: “Celecoxib VERY LIKELY causes excess cardiovascular events and illustrates the folly of allowing drug sellers too much power”. Differences between groups in trials could always be due to chance alone. Jeffrey Mann is correct in that a trial would need a larger number of subjects than my meta-analysis to have a high power for avoiding a false negative. However that is not relevant because the issue here is the chance of my finding being a false positive. That chance is a bit less than 1 in 20 because the lower 95% confidence interval is an odds ratio of 1.06. As I stated in my rapid response, the available evidence does not provide a precise estimate of the magnitude of additional harm. Because the only available studies of celecoxib vs non-selective NSAIDs were done with subjects who were not at high risk of cardiovascular events the sample size is too small to produce a high signal to noise ratio.<1> This is indicated by the wide confidence intervals that I reported. I agree with Jeffrey Mann that it is misleading to just say that that Vioxx doubles the risk of cardiovascular events. The magnitude of harm caused by Vioxx could be much higher or lower. This problem of overlooking the uncertainty conveyed by the confidence intervals and thinking that the results in a sample closely represent the magnitude of the effect caused by a drug is common amongst doctors as well as journalists. It is known by psychologists as representativeness bias.<2> Arbitrary selection of studies would be contrary to my personal values and work with Healthy Skepticism. The 3 studies in my meta-analysis were selected by logical pre-specified criteria. I was not aware of the existence of 2 of those 3 studies when I set the criteria. I was expecting that there would not be enough data to find a statistically significant difference. I was wrong: The available evidence shows that celecoxib very likely causes more cardiovascular events than non-selective NSAIDs but does not give a precise estimate of the magnitude of additional harm. 1. Sackett DL. Why randomized controlled trials fail but needn't: 2 Failure to employ physiological statistics, or the only formula a clinician-trialist is ever likely to need (or understand!). CMAJ 2001 Oct 30;165(9):1226-37 2. Kahneman D, Slovic P, Tversky A. Judgment under uncertainty: heuristics and biases. New York: Cambridge University Press; 1982 Competing interests: I am the Founder of Healthy Skepticism Inc which aims to improve healthy by reducing harm from misleading drug promotion. www.healthyskepticism.org |
|||