Overdosed America (6 page)

As I got into the review of the scientific evidence, I became increasingly puzzled. The authors praised both drugs, stating that they cause “significantly fewer serious gastrointestinal adverse events than does treatment with non-selective NSAIDs.” I pulled the “Dear Healthcare Provider” letter out of my files to make sure that my recollection of its wording was accurate. It was. After going over the language very carefully, I realized that this review article had taken drug salesmanship into new territory. This supposedly authoritative review article in the
New England Journal of Medicine
seemed to be stating as fact one of the “unsubstantiated comparative claims” that the FDA had forbidden the manufacturer of Celebrex to make on its own behalf.

The review article also seemed to be trying too hard to present a favorable picture of Vioxx. The results of the Vioxx Gastrointestinal Outcomes Research (VIGOR) study, published in the NEJM in the fall of 2000, had raised the possibility of Vioxx increasing the risk of heart attack and other cardiovascular complications. The NEJM review article published in August 2001 reported that people who took Vioxx had twice as many heart attacks, strokes, and cardiovascular deaths and four times as many heart attacks as the people who took naproxen (
p
< .05 and
p
< .01, respectively). But then, rather than addressing these serious complications, the authors dismissed them with a most unusual statement: “The difference in major cardiovascular events in the VIGOR trial [of Vioxx] may reflect the
play of chance
” (italics mine) because “the number of cardiovascular events was small (less than 70).” The comment that a statistically significant finding “may reflect the play of chance” struck me as very odd. Surely the experts who wrote the review article knew that the whole purpose of doing statistics is to determine the degree of probability and the role of chance.

Anyone who has taken Statistics 101 knows that
p
values of .05 or less (
p
< .05) are considered statistically significant. In this case it means that if the VIGOR study were repeated 100 times, more than 95 of those trials would show that the people who took Vioxx had at least twice as many heart attacks, strokes, and death from any cardiovascular event than the people who took naproxen. And in more than 99 out of those 100 studies, the people who took Vioxx would have at least four times as many heart attacks as the people who took naproxen. This “play of chance” caveat seemed like a not very well camouflaged attempt at damage control. Then I noticed that the two authors of the review article had financial ties to the manufacturers of Celebrex and Vioxx, even though financial relationships between authors of review articles and drug companies were
prohibited by the
New England Journal of Medicine
at the time the article was published.
*

I went back to the NEJM article that presented the
results of the VIGOR study from the fall of 2000
. The article mentioned that the differences in cardiovascular complications between the patients who had taken Vioxx and naproxen had been measured. But, curiously, the article did not report what those differences were; purportedly they were to be included in a later article. The NEJM article claimed, “A separate analysis of these events, however, was not specified in the study design.”

I became even more suspicious when I noticed that the total number of serious gastrointestinal complications that occurred in the patients in the VIGOR study was only 53. Wait a minute. The authors of the VIGOR study (as well as the authors of the review article) declared that a total of 53 serious GI complications was the most important finding in the study, and reason alone to prescribe Vioxx instead of a far less expensive NSAID. How could the authors of the review article then dismiss the reported cardiovascular risk with Vioxx on the grounds that there were “less than 70” cardiovascular events? Moreover, as I searched through the original VIGOR article, I couldn’t find the number 70 anywhere, suggesting that the authors of the review had another source of information.

Now I was determined to find out the real story. The articles about statins and strokes had shown me how effectively the drug companies can spin their “scientific evidence.” As I was about to find out, the articles about Celebrex and Vioxx went a step further.

One might assume that anyone who was willing to go to the trouble of verifying conclusions presented in medical journals could do so. This is not the case at all. Drug companies often
keep the results of their studies secret
, even from their own researchers, on the grounds that such results are “proprietary information” of economic value. Getting to the bottom of the Celebrex and Vioxx stories would be difficult, but I had already done enough of this kind of research to know that it was usually worth poking around online.

I started with Harvard Medical School’s digital library, but couldn’t find any new research or more complete data. Next I tried the government-sponsored “Pub-Med” website, a database of several thousand scientific journals; but a comprehensive search there didn’t yield any new information either. Then I went to the FDA website to see if I could find any unpublished data the agency had used in its decision to mandate Pharmacia’s “Dear Healthcare Provider” letter. Not knowing what I was looking for, I came blindly upon reports from the FDA’s
Arthritis Advisory Committee meeting of February 7–8
, 2001, which had been held specifically to go over the FDA reviewers’ analyses of the data from the CLASS and VIGOR studies in order to
consider the manufacturers’ requests
to remove warning labels about the risk of ulcers from Celebrex and Vioxx.

As I started to go through these reports, I realized that they were the actual FDA reviewers’ internal analyses of the research data that the manufacturers of Celebrex and Vioxx had submitted to the FDA from the CLASS and VIGOR studies. I could not believe what I was seeing. The FDA reviewers’ reports contained a treasure trove of “not-yet-spun” data, revealing a very different picture of the safety of these two drugs than had been presented in the JAMA and NEJM articles.

Based on the data from the manufacturers’ own studies and the FDA reviewers’ analyses, the conclusions that had been presented in the Celebrex CLASS and Vioxx VIGOR articles seemed to be encouraging physicians to prescribe drugs that provide few benefits and possibly even cause harm. The amazing thing is that the conclusions presented in these articles were based upon exactly the same data that the manufacturers had sent to the FDA. I couldn’t believe that these data—clearly illustrating the magnitude of the scientific distortion presented in our two most respected medical journals—were publicly available and yet nobody seemed to be aware of the problem.

I proceeded to spend nights and weekends comparing the data available on the FDA website with those presented in the journal articles in the fall of 2000. Meanwhile, sales of the two drugs kept growing, bolstered by an advertising blitz featuring a still graceful former Olympic figure-skating champion and a strikingly happy older woman in an outdoor tai chi class. And my patients kept requesting and demanding that I prescribe these “better” new drugs for their aches and pains.

The
results of the CLASS study
were published in the September 13, 2000, issue of JAMA. CLASS is what’s known as a Phase 4 postapproval study, which was required by the FDA. Before any drug is approved, manufacturers have to submit data to the FDA that demonstrate the drug’s safety and effectiveness. These preapproval clinical trials are expensive, and as a result usually include a relatively small number of people. Occasionally, a drug that seems safe in preapproval trials is later found to cause unexpected side effects when used more widely. Therefore, the FDA often requires that drug companies conduct postapproval Phase 4 studies on a much larger group of people to make sure the drug is as safe and effective as it appeared to be in the earlier, smaller trials.

CLASS, which included over 8000 people with rheumatoid and osteoarthritis, compared the risk of gastrointestinal problems in people taking Celebrex with the risk in those taking ibuprofen (Motrin, Advil) and diclofenac (Voltaren). The article in JAMA concluded that Celebrex, “when used for 6 months . . . is associated with a lower incidence of clinical upper GI events than comparator NSAIDs (ibuprofen and diclofenac).” The
accompanying editorial
supported this conclusion: “The results of this important study . . . provide promising data to suggest that [Celebrex is] . . . effective in reducing, but not eliminating, the risk of symptomatic [minor] ulcers and [major] ulcer complications in the enormous number of individuals who might benefit from these drugs . . .”

There was, however, one very large problem. The
manufacturer’s original research plan
, as submitted to the FDA, had defined the duration of the CLASS study that compared Celebrex with ibuprofen as 12 months, and that of the study comparing Celebrex with diclofenac as 16 months. And, indeed, the combined study had run for a full 12 months. The authors, however, submitted only the first 6 months for the article in JAMA. Left unreported (and unmentioned) in the JAMA article were the data from the
second
6 months of the study, during which time, as shown in the data on the FDA’s website, six of the seven serious gastrointestinal complications that occurred were in patients taking Celebrex.

Pharmacia, the manufacturer of Celebrex, presented a statistical argument to the FDA justifying its omission of the data from the second half of its study. The company claimed that since a higher percentage of people taking diclofenac dropped out of the study because of minor symptoms like heartburn, the data from the second half of the study were invalid because of what is called “informed censoring.” The manufacturer argued that these dropouts would have gone on to develop serious gastrointestinal complications, and their dropping out of the study artificially minimized the risk of serious complications from taking diclofenac. The FDA flatly rejected this argument. It countered that there was no proof that the people with heartburn would have developed more serious gastrointestinal problems. Further, if minor symptoms caused people in the study to stop taking diclofenac, people in the real world would similarly stop taking the drug if it caused heartburn and would similarly protect themselves from going on to develop serious gastrointestinal complications.

The FDA’s opinion of the manufacturer’s decision to publish only half of the data from its study was clear:
“the sponsor’s presentations of 6-month data . . . are not statistically valid or supportable.”
The
FDA’s gastroenterology reviewer concluded
that the first 6 months of data—which had been presented in the JAMA article as if they were a report of the entire study—were not worthy of separate consideration: “Based on the lack of adequate rationale, these post-hoc analyses will not be further discussed or presented in this review.” Looking at the data from the entire year of the study, the FDA’s gastroenterology reviewer concluded that “the sponsor has failed to demonstrate a statistically significant lower rate” of serious GI complications in the people who took Celebrex compared with the people who took the other NSAIDs. When the reviewer looked at only the second six months of data (i.e., the data that had not been published in the JAMA article), he concluded that the risk of serious GI complications appeared to be higher in the people who took Celebrex “compared to
both
ibuprofen and diclofenac” (FDA’s underscore). This was hardly an endorsement for a drug whose only advantage (besides the convenience of a once-daily dosing) was that it caused fewer serious GI problems.

This didn’t entirely escape public notice. While I was delving into FDA data, I came across a
story that had been published in the
Washington Post
about Pharmacia’s having provided only the first half of the data to the editors of the
Journal of the American Medical Association.
“I am furious.. . . I looked like a fool,” Dr. Michael Wolfe, coauthor of the JAMA editorial supporting Celebrex, told the reporter when he learned of this deception. JAMA editor Dr. Catherine D. DeAngelis said, “I am disheartened to hear that they had those data at the time that they submitted [the manuscript] to us.. . . We are functioning on a level of trust that was, perhaps, broken.” I could find no other major media that reported this story.

The disparity between the CLASS article published in JAMA and the information in the FDA’s files by no means stopped there. The primary question that the CLASS study had been designed to answer had been changed, producing results that were far more favorable to the manufacturer. The original research design submitted to the FDA by the manufacturer of Celebrex had stated:
“The primary objective of this study
is to compare the incidence of clinically significant [major] upper gastrointestinal events . . . in patients taking Celebrex to patients taking other NSAIDs.” The term “clinically significant” refers to complications that would generally require hospitalization: active bleeding, perforation of the stomach or duodenum requiring surgery, or obstruction of the outlet of the stomach. The research plan specifically called for the less serious gastrointestinal side effects to
“be categorized and analyzed separately.”
Indeed the FDA’s gastroenterology reviewer specifically commented that the plan to identify the “truly significant” serious gastrointestinal complications alone was a
“major strength of the current study.”

But when the results of the study were published in JAMA, the incidences of major and minor gastrointestinal complications were combined. Why the change? The results of the study as originally designed failed to show that the people who took Celebrex developed significantly fewer major gastrointestinal complications than the people who took ibuprofen or diclofenac, even for just the first six months. Only by combining the minor GI symptoms with the more serious gastrointestinal complications could the article conclude that Celebrex caused a statistically significant decrease in gastrointestinal complications compared with the other NSAIDs. As noted above, when the FDA looked at the results of the CLASS study in terms of the research question that had
originally
been posed, Celebrex was not significantly safer than the other NSAIDs.