FDA,
Merck and Vioxx: Putting Patient Safety First?
November
18 , 2004, at 10:00 a.m. in 216 Hart Senate Office Building
Gurkirpal Singh, MD
Adjunct Clinical Professor of Medicine
Department of Medicine, Division of Gastroenterology and Hepatology
Stanford University School of Medicine
Stanford, CA
and
Chief Science Officer
Institute of Clinical Outcomes Research and Education (ICORE)
175 Eleanor Drive, Woodside, CA 94062
gsingh@stanford.edu
Chairman Grassley, Senator Baucus, Senators, and Ladies and Gentlemen:
Thank you for inviting me to testify before the Senate Finance Committee. I apologize
for not appearing in person, and giving this testimony by a video conference. I am unable to
travel because exactly two weeks ago today, I had a heart attack – and before the plaintiff’s
attorneys rush out of this room to call me - no, I was not taking Vioxx.
I have been asked to review the science of Cox-2 inhibitors, the link of rofecoxib to heart attacks,
the timeline of different studies, and my own role in teaching physicians about these issues.
Hindsight is always 20/20, and I do not intend to be a Monday morning quarterback today.
Instead, I will try to highlight the learnings and knowledge that we can derive from this episode
so that early signals are not missed again with another drug. At the end of my presentation, I will
make recommendations that I believe are essential to avoid a repetition of this unfortunate
incident where millions of Americans were unknowingly subjected to serious harm.
I am a rheumatologist by clinical training with research interests and expertise in drug
safety and epidemiology. My group and I were instrumental in pointing out the risks of
painkillers such as motrin and aleve (a class of drugs called NSAIDs), identification of patients
who have a risk of serious stomach bleeding from such drugs and potential ways to avoid such
risks. I have been working in the research area of drug safety and outcomes research for almost
15 years, and have published extensively in the medical literature. I am currently working with
large public datasets such as Medicare and Medicaid to study early safety signals of medications.
I lecture medical students, residents and other physicians, both at Stanford, and in conferences
worldwide, on many of these issues.
Science of specific Cox-2 inhibitors
There are 2 enzymes in the human body – cox-1 and cox-2 (attachment 1). Cox-1
enzyme is needed for the normal functioning of stomach and platelets. Cox-2 enzyme, on the
other hand, is thought to be responsible for pain and swelling of arthritis. Traditional painkillers
such as ibuprofen (the chemical in motrin) inhibit both cox-1 and cox-2. This means that while
these drugs are effective in reducing pain, they increase the risk of stomach bleeding. A few
years ago, my colleagues and I estimated that there are over 103,000 hospitalizations and 16,500
deaths every year from the stomach bleeding complications of these drugs (1, 2). The specific
cox-2 inhibitor drugs such as Vioxx and Celebrex, were developed to inhibit only cox-2, and not
cox-1. It was hoped that these drugs would relieve pain but not have any stomach problems.
Indeed, this seems to be the case. In May 2004, I presented data that showed a significant
reduction in the number of stomach bleeds in the US after the launch of these drugs (3).
However, it is important to remember that drugs such as Vioxx do not cure arthritis – they are
used only for control of pain, and are medicines for convenience and quality-of-life improvement
rather than for savings lives or preventing disabilities. There are many other ways to effectively
control pain as well.
Heart Attacks
It is believed that most heart attacks occur when the blood vessels supplying blood to the
heart become narrowed because of cholesterol deposits (attachment 2), and a blood clot forms at
this narrowing, stopping the flow of oxygen to the heart muscle. The blood clot is formed by
cells called platelets, and it is the cox-1 enzyme in the platelets that is responsible for this
function. Aspirin destroys this enzyme in a permanent fashion and prevents blood from clotting
in the heart blood vessels, thus helping reduce the risk of heart attacks. Other painkillers such as
ibuprofen and naproxen also inhibit the enzyme in the platelets, but only temporarily and
incompletely. While it is possible that these non-aspirin painkillers may also reduce the risk of
heart attacks, this has never been shown in any randomized clinical trial, despite claims to the
contrary (4). These drugs are not used for preventing heart attacks since even if they were to be
effective, the effect of temporary and incomplete inhibition of platelet would be much less
beneficial than the complete and permanent inhibition caused by aspirin.
Vioxx and Risk of Heart Attacks
The Senate Finance Committee provided me with information on events surrounding the
approval and withdrawal of Vioxx, and the supporting documents attached to my testimony. I
have been asked to comment on this with the specific purpose of identifying key events that
should have alerted scientists and public to the potential problems with Vioxx so that a similar
problem can be avoided in the future with another drug.
Before I review the attachments, I wish to reiterate that the fundamental principle of
medicine – one that every physician swears by is - Primum, Non Nocere – First, Do No Harm. A
second principle is a careful evaluation of risk-benefit ratio of any treatment. It is easier to accept
a more serious side-effect such as heart attack in a drug that cures cancer, for example, than in
one that is used to treat skin rash.
We now know that by November of 1996, Merck scientists (5) were seriously discussing
a potential risk of Vioxx – association with heart attacks (attachment 3). At that time, it was not
known that Vioxx may itself cause heart attacks. Rather, the discussion focused on the issue that
other painkillers by inhibiting platelets may protect against heart attacks. Vioxx has no such
effect on platelets, and thus may seem to increase the risk of heart attacks in studies comparing it
to other painkillers. This was a serious concern because the entire reason for the development of
Vioxx was safety – please note, once again, that it is no more effective than older NSAIDs. If the
improved stomach safety of the drug was negated by a risk of heart attacks, patients may not be
willing to make this trade-off. Merck scientists, considered by many to be the best and brightest
in the pharmaceutical industry, were among the first to recognize this. At this point in time,
scientists should have started a public discussion about this potential trade-off, and designed
studies that would more carefully evaluate the risk-benefit ratio of the drug.
It appears from the internal Merck e-mails provided to me (attachment 4), that in early
1997, Merck scientists were exploring study designs that would exclude people who may have a
weak heart so that the heart attack problem would not be evident. The discussion also focused on
the fact that if aspirin were permitted in these trials, there may not be any significant safety
advantage of Vioxx on the stomach. On the other hand, as one scientist pointed out, if aspirin
was excluded, patients on Vioxx may have more heart attacks and this would “kill the drug”. He
also points out that in the real world, “everyone is on it”. Clinical trials should be designed to test
a drug under “real world” circumstances – on patients who are most likely to use the drug.
Clinical trials should not be designed to selectively favor one outcome over another by excluding
people similar to those who would take the drug after its approval. Certainly, clinical trials
should not be designed to put marketing needs in front of patient safety – we need to know how a
drug behaves in people who are going to take it, even if it “kills the drug”. It is better to kill a
drug than a kill a patient.
According to documents provided to me by the Senate Committee, there were many other
internal discussions within Merck on these concerns of heart attack-stomach bleed trade-offs,
although the practicing physician did not learn of any of this till many years later. In 1998, Dr.
Doug Watson, a Merck scientist presented an analysis of serious heart problems with Vioxx
compared to patients enrolled in studies of other Merck drugs. This analysis (attachment 5)
concluded that men taking Vioxx had a 28% greater risk (not statistically significant), but in
women, the risk was more than double (216%, statistically significant) compared to people not
taking any drug in other Merck studies. To the best of my knowledge, these data were never
made public. This is when a public scientific discussion of the pros and cons of the medication
should have started.
By 1999, an even more serious problem was emerging. By the time Merck had filed for
the approval of Vioxx, there were several small studies evaluating the efficacy and safety of
Vioxx in patients with pain and arthritis. None of these studies were large enough to study the
risk-benefit trade offs of stomach bleeds versus heart attacks. But a careful FDA review of
Merck’s new drug application for Vioxx, Dr. Villalba (attachment 6) noticed that
“thomboembolic events [such as heart attack and stroke] are more frequent in patients receiving
VIOXX than placebo…” [page 105]. Among 412 patients taking placebo, 1 had a cardiovascular
event (0.24%); and among the 1631 patients receiving 12.5 mg or more of VIOXX daily, 12 had a
cardiovascular event (0.74%) (6). This meant that not only did VIOXX not inhibit the platelets,
but for some reason, it was likely to promote heart attacks directly. Many scientists would
consider this three-fold difference as an early warning sign. But there were no adequate data to
make a firm conclusion one way or another. In fact, the FDA reviewer went on to point out that:
“With the available data, it is impossible to answer with complete certainty whether the risk of
cardiovascular and thromboembolic events is increased in patients on rofecoxib. A larger
database will be needed to answer this and other safety comparison questions” [page 105].
It is my opinion that at this point in time, larger and more definitive studies should have been
done before the drug was approved. After all, the drug was no more effective than any other
available pain-killer – and there were nearly 30 such drugs available in the US. Another drug
(celebrex) that had no such signal had also been available in the market for 6 months prior. A
combination of two older drugs – a pain-relieving drug such as motrin with a drug that protects
the stomach such as prilosec – is as effective and almost as safe on the stomach as Vioxx, with no
heart attack risk. There was certainly no emergent need to approve Vioxx without further studies
if there were lingering safety concerns. The trade-off of heart attacks for the rare instances of
stomach bleeds is not a reasonable one. Remember, primum non nocere – first, do no harm.
Instead, the drug was approved by the FDA in a priority review within 6 months – with no
discussion on the heart attack trade-off. The prescribing physicians remained unaware of any of
these data or discussions, till much later – with the new label change in April, 2002.
VIGOR Trial and my interaction with Merck
The VIGOR trial, which will be discussed in detail later, was the first public release of
heart attack-stomach bleed trade-off concerns. At the time VIGOR study results were
announced, I was actively involved in research and teaching in this area. Some of my medical
education lectures were sponsored by Merck and other drug companies. I was strongly in favor
of this new class of drugs, and before the VIGOR trial, was unaware of any significant heart
attack issues. The results of the VIGOR trial – a 500% increase in the risk of heart attacks with
Vioxx – stunned me. Clearly, the trade-off of 500% increase in heart attacks for a 50% reduction
in stomach bleeds did not seem attractive – at least, not without a further discussion of data.
Merck’s press release on this issue and a brief mention of the heart attack data were not enough
for me to continue to educate physicians in my lectures. I asked Merck for more detailed data,
including information on high blood pressure and heart failure rates. When I was unable to
obtain this data after multiple requests, I added a slide to my presentations that showed a man --
representing the missing data -- hiding under a blanket (attachment 7). Up until this point in time,
Merck had responded to all my requests promptly and in a scientific fashion. With VIGOR,
suddenly it was as if the Company had to think what questions to answer. I persisted in my
enquiries – and I was warned that if I continued in this fashion, there would be serious
consequences for me. I was told that Dr. Louis Sherwood, a Merck senior vice-president, and a
former Chief of Medicine at a medical school, had extensive contacts within the academia and
could make life “very difficult” for me at Stanford and outside. But as a research scientist, I felt
that it was unethical for me not to discuss my concerns in public. An open scientific debate was
important – it is only through open debate and discussion that we advance science. Dr. Sherwood
called several of my superiors at Stanford to complain (attachment 8). Subsequently, I learnt that
this was a persistent pattern of intimidation by Dr. Sherwood. Professor Fries too felt that this
suppression of scientific discussion was unethical and complained to Mr. Raymond Gilmartin
(attachment 9). Mr. Gilmartin and Mr. David Anstice took immediate action, and the threats
stopped immediately. From then onwards till today, Merck scientists and officials have treated
me and my colleagues with appropriate respect and have always shared scientific data promptly.
We have not always agreed with the interpretation of data, but to the best of my knowledge,
nothing has been hidden, suppressed or falsified by any Merck scientist since this episode. All my
requests for scientific information are handled promptly and courteously, and for this, I thank
Merck in general, and Dr. Alise Reicin in particular.
Publication of VIGOR data
Scientific publications in a medical journal are the most credible way to disseminate data
about a medication. VIGOR data was published in the New England Journal of Medicine in
November, 2000. A few weeks ago, Merck announced that the published VIGOR data was
“preliminary” and that the “final” data was presented to the FDA. In my view, and all of my
colleagues that I have consulted with, it is inappropriate to publish “preliminary” or incomplete
data without clearly stating that the data are preliminary. This is especially true if the favorable
data are complete but the unfavorable data are “preliminary” and likely to get worse. To the best
of my knowledge, the VIGOR paper did not indicate anywhere that the data were preliminary or
incomplete. Nor, did I ever see a correction or erratum indicating this fact subsequently – up
until a few weeks ago, almost 4 years later.
The VIGOR publication minimized the significance of heart attacks. While it
prominently discussed the reduction of stomach bleeds in patients taking Vioxx, it did not
mention that in spite of this, patients on Vioxx had more serious adverse events, and more
hospitalizations than patients on Naproxen. The true rates for cardiovascular thrombotic adverse
events (a prespecified study endpoint in the protocol), hypertension and congestive heart failure –
which were all higher in the Vioxx group - were not shown in the paper at all.
The FDA review of VIGOR correctly pointed out that the explanation advanced by the
authors – that naproxen reduced the risk of hear attacks – could not explain the 500% difference
between Vioxx and naproxen. The reviewers also highlighted data from many other studies
showing that this was not an isolated finding in VIGOR. However, Merck continued to claim
“favorable cardiovascular safety profile” of Vioxx in multiple press releases and Companysponsored
lectures and conferences. In September 2001, in a Warning Letter to Merck, the FDA
Division of Drug Marketing, Advertising, and Communications (DDMAC) called the press
releases claiming a “favorable cardiovascular safety profile” for VIOXX “simply
incomprehensible”, and pointed out that the naproxen explanation was merely “hypothetical”
rather than factual. These facts had previously been discussed by FDA reviewers as well (7).
.
Post-VIGOR Label Change
The VIGOR data were first made public in May 2000. However it was not until almost 2
years later that the FDA requested Merck to revise Vioxx’s product label to reflect the heart
attack risks observed in the VIGOR trial. These revisions were added to the “Precautions”
section, under “Cardiovascular Effects”, instead of being prominently displayed as a “Warning”.
While the stomach bleed safety data was added in a prominent fashion, the heart attack
information seemed to support Merck’s contention that Vioxx did not increase the risk by adding
statements such as “Because of its lack of platelet effects Vioxx is not a substitute for aspir in for
cardiovascular prophylaxis”. Was there a single physician in the world who had prescribed
Vioxx for cardiovascular prophylaxis? Why not also say “Because of its lack of anti-tumor
effect, Vioxx is not a treatment for brain cancer” or “Do not use Vioxx for erectile dysfunction or
depression”? The favorable data for Alzheimer’s disease studies was included at Merck’s
insistence, but no unfavorable data from studies such as 085 or 090 as added. Even the
Alzheimer’s disease studies data was favorably biased – while the label showed that there was no
difference in heart attacks between Vioxx and placebo in these studies, it did not mention that the
mortality rate of patients on Vioxx was almost twice that of those on placebo. Negotiations
certainly succeeded for Merck.
Many people claim that the heart attack – stomach bleed data trade off was a favorable
one, since there are many more stomach bleeds prevented than heart attacks caused by Vioxx. As
the FDA review of VIGOR data pointed out, this was simply not true (7). Attachment 9 is selfexplanatory.
No long-term safety studies
More importantly, there were no attempts to design and carry out large safety studies to
prove or disprove the link of Vioxx to heart attacks. Apparently, a 30,000 patient study had been
announced in November, 2001 but never started. Last week, New York Times reported that
Merck had considered a cardiovascular outcome study, but decided that it would send the
“wrong” marketing and public relations signal. "At present, there is no compelling marketing
need for such a study," said a slide prepared for a meeting of senior executives. "Data would not
be available during the critical period. The implied message is not favorable." It is regrettable that
scientific decisions on patient safety are influenced by perceived marketing and public relations
concerns. In my opinion, it is better to kill a drug than kill a patient.
It is important to note that the APPROVe study which conclusively proved the increased
risk of Vioxx was not a safety study – it was an efficacy study, designed to add another indication
for Vioxx treatment. It was not large enough to detect a heart attack risk – that it did find a risk
was a lucky break for patients, but this is not what it was designed to do.
The failure to conduct large long-term safety studies subjected millions of patients over 4 years
to a drug whose safety had been questioned by the FDA even before its approval. This is not the
proudest chapter in drug approval in the US.
Recommendations
What can we do to prevent this from happening again? First, we must find out exactly what went
wrong.
1. A public enquiry should be conducted by an independent group of scientists with free
access to all Merck internal documents to study all aspects of safety data surrounding
Vioxx, with a particular emphasis on (a) if earlier, better studies could have shown the
heart attack risk, (b) if such studies had indeed been suppressed by marketing and public
relations worries, and (c) if a discussion of this heart attack risk was suppressed in an
unethical fashion.
2. A public discussion of the role of FDA in approving drugs and labels. As the delay in
Vioxx label shows, the current process of labeling is one of negotiations – if the
“sponsor” does not agree with what the FDA wants, it can continue to stall or worse. It
took 2 years for the label change of Vioxx to take effect, and even then, the label change
supported mostly Merck’s position, not the one advanced by FDA’s own reviewers in
public hearings. This process needs to be fixed, if need be, by new legislation. The FDA
should be given the authority that is accorded to our judicial system – to make unilateral
decisions on issues of public health safety, without having to negotiate and reach
agreement with drug companies. The FDA should regulate the drug companies, not
collaborate or negotiate with them if there is any question of public safety.
3. The FDA approval process needs to be more open and subject to public scrutiny. Once a
drug is approved, all the data supporting such approval should be put in the public
domain. If this had been done with Vioxx, perhaps independent scientists would have
been able to spot early signals. Similarly, all clinical study data submitted to the FDA
should be available to the public after the drug is approved. Claims of “trade secrets”
should not take precedence over public health and safety. Pharmaceutical companies
should not be allowed to selectively disseminate only positive data.
4. On drugs that need further safety data, a system of conditional or time-limited approvals
should be instituted. For example, since the FDA reviewer had concerns about heart
attacks before the approval of Vioxx, but there was not enough data to decide the issue
one way or other, the FDA could have provided a conditional approval (if any) that
would have required Merck to complete large safety studies within a certain time period.
5. An independent office of drug safety which does not report to the FDA new drug
approval section should be established. Safety data on all new drug approvals must be
vetted through this office. This office should have an independent authority to conduct
safety studies on approved drugs, or require that such studies be conducted if there are
safety signals. Only then will be able to adhere to the principle of “Primum, Non
Nocere” – First, Do No Harm.
Thank you.
References
1. Singh G. Recent considerations in nonsteroidal anti-inflammatory drug gastropathy.
Am J Med 1998; 105(1B):31-38.
2. Singh G and Triadafilopoulos G. Epidemiology of NSAID-induced GI complications.
J Rheumatol 1999; 26 Suppl 26:18-24.
3. Singh G, Mithal A, Triadafilopoulos G. Decreasing hospitalizations due to complicated
gastric and duodenal ulcers in the United States: 1998-2001. Gastroenterology 2004;
126 (4 Suppl. 2): A97-98.
4. Ray WA, Stein CM, Hall K, Daugherty JR, Griffin MR. Non-steroidal antiinflammatory
drugs and risk of serious coronary heart disease: an observational cohort
study. Lancet 2002;359:118-123.
5. Muliner T. Anticipated consequences of NSAID antiplatelet effects on cardiovascular
events and effects of excluding low-dose aspirin use in the Cox-2 GI Outcomes
Megatrial. Letter of November 21, 1996 to B Friedman, A Nies, and R Spector.
6. Villalba ML. FDA Medical Officer Review of VIOXX (rofecoxib), NDA 21-042
(capsules) and NDA 21-052 (oral solution). Http://www.fda.gov/
cder/drug/infopage/vioxx/default.htm.
7. Targum SL. Consultation on NDA 21-042, S-007; Review of cardiovascular safety
database [on Vioxx or rofecoxib). FDA Memorandum, Feb 1, 2001.
Http://www.fda.gov/ohrms/dockets/ac/01/briefing/3677b2_06_cardio.doc; last accessed
on June 5, 2001.
Vioxx Info | Hearing Index | Home Page
