THE ROFECOXIB CASE

WHAT HAVE WE LEARNT FROM

Vioxx?
In October UK patients who had cardiovascular events
while taking rofecoxib lost the right to fight Merck in the
US for compensation. But researchers and journals can still
benefit from this case if they learn from the mistakes,
write Harlan Krumholz and colleagues

Rofecoxib (Vioxx) was intro- volunteers by about half.w2 In internal emails cardiovascular risks, despite FDA concern,5
duced by Merck in 1999 as an made public through litigation,3 Merck offi- and disseminated the results to promote the
effective, safer alternative to cials sought to soften the academic authors’ drug’s cardiovascular safety to doctors in its
non-steroidal anti-inflamma- interpretation that cyclo-oxygenase-2 (COX “cardiovascular card,”6 7 a marketing device
tory drugs for the treatment 2) inhibition within the vascular endothe- cited by US Congressman Henry Waxman
of pain associated with osteoarthritis. It was lium may increase the propensity for throm- for falsely minimising cardiovascular risks8
subsequently found to increase the risk of bus formation, the basis of what became and never approved by the FDA.
cardiovascular disease and withdrawn from known as the FitzGerald hypothesis.w3 The
the worldwide market. Merck now faces academic authors changed the manuscript at The VIGOR study
legal claims from nearly 30 000 people who Merck’s request—for example, they changed In January 1999, Merck launched its larg-
had cardiovascular events while taking the “systemic biosynthesis of prostacyclin ... was est study yet of rofecoxib, the Vioxx gastro-
drug.1 The company has stated that it will decreased by [rofecoxib]” to “Cox-2 may intestinal outcomes research (VIGOR) study.
fight each case, denying liability.2 Our recent play a role in the systematic biosynthesis of The study was intended to expand the drug’s
participation in litigation at the request of prostacyclin.”3 w2 To the authors’ credit, they approved indications by showing that it
plaintiffs provided a unique opportunity to continued to investigate the effects of COX would have fewer gastrointestinal side effects
thoroughly examine and reflect on much of 2 inhibition and ultimately provided much than naproxen for the treatment of rheuma-
the accumulated court documents, research, of the basic science knowledge that clarified toid arthritis. The study of over 8000 patients
and other evidence. This story offers impor- the pathways by which rofecoxib probably was initiated without a standard operating
tant lessons about how best to promote con- leads to cardiovascular events.w4-w7 procedure for collecting information on car-
structive collaboration between academic However, despite Merck’s knowledge that diovascular events and without a cardiologist
medicine and industry. rofecoxib might increase thrombus forma- on the data safety monitoring board. Data
tion, none of the intervention studies that safety monitoring boards are independent
Early suspicion of cardiovascular risk constituted its new drug application to the committees whose purpose is to monitor
Since the early development of rofecoxib, Food and Drug Administration in 1998 were the results of an ongoing trial to ensure the
some scientists at Merck were concerned designed to evaluate cardiovascular risk. safety of trial participants.w8 The study was
that the drug might adversely affect the car- The nine studies were generally small, had designed to continue until a predetermined
diovascular system by altering the ratio of short treatment periods, enrolled patients at number of confirmed uncomplicated or
prostacyclin to thromboxane, which act in low risk of cardiovascular disease, and did complicated gastric perforations, ulcers, or
opposition, balancing blood flow and clot- not have a standardised procedure to collect bleeds had occurred.
ting.w1 A study sponsored by Merck during and adjudicate cardiovascular outcomes.4 The first non-endpoint safety analysis was
1996-7 reported that rofecoxib reduced uri- Moreover, Merck seemingly pooled data presented to the safety board in November
nary metabolites of prostacyclin in healthy from these studies and others for analysis of 1999, at which time a 79% greater risk of

120 BMJ | 20 JANUARY 2007 | VOLUME 334

 THE ROFECOXIB CASE

Vioxx in the dock: lawyer Mark Lanier

holds up a sample packet of rofecoxib as
he speaks during proceedings against
Merck in New Jersey in March 2006

contained data from an interim analysis that

had different termination dates for cardio-
vascular and gastrointestinal events (gastro-
intestinal events were counted for one month
longer than the cardiovascular events). This
highly irregular procedure was not described
in the publication and had the effect of favour-
ing the drug’s effect on gastrointestinal events
while understating the risk of cardiovascu-
lar events.w9 The published cardiovascular
risk was not accurate because three addi-
tional myocardial infarctions occurred in
the rofecoxib group in the month after the
researchers stopped counting cardiovascular
events (none had occurred in the naproxen
group). The potential harm was further mini-

MERL EVANS/AP/EMPICS
mised by a post hoc subgroup analysis based
on “indication for aspirin prophylaxis”; had
Merck included the three cases, the subgroup
analysis would have shown an increased
cardiovascular risk in both groups.w10
The publication concealed the cardio-
death or serious cardiovascular event was naproxen, it could be used to petition the vascular risk even further by presenting the
found in one treatment group compared with FDA for a new indication. However, if the hazard of myocardial infarction as if naproxen
the other (P=0.007).9 The board allowed the study raised concerns about cardiovascular was the intervention group (relative risk 0.2,
study to continue and planned to review sub- harm, the billion dollar drug franchise 0.1 to 0.7) and without reporting the absolute
group analyses in December, at which time would be threatened. The study showed number of cardiovascular events, even though
the analyses again showed higher cardio- that rofecoxib was not more effective in all other results were presented appropriately
vascular risk in one group. On this basis the relieving symptoms of rheumatoid arthritis with rofecoxib as the intervention group.w11
board recommended that an analysis plan be but did halve the risk of gastrointestinal Finally, the authors proposed a naproxen
developed to examine serious cardiovascular events. However, there was also evidence hypothesis, suggesting that rofecoxib had not
events and that the study continue until it of an increased risk of myocardial infarction been harmful but that naproxen had been
reached its gastrointestinal endpoint target (relative risk 5.00, 95% confidence interval protective, despite there being no accepted
(expected March 2000). 1.68 to 20.13). When this result was circu- evidence that naproxen had a strong cardio-
Matters were complicated by the existence lated internally at Merck, Edward Scolnick, protective effect.
of conflicts of interest among board mem- the company’s chief scientist, Merck strongly promoted
bers. According to Merck policies, the board wrote in an email to colleagues the VIGOR study, purchas-
The published VIGOR
is supposed to be independent, without finan- about the cardiovascular risk: ing nearly 1 million reprints
cial or emotional stake in the trial being mon- “It is a shame but it is a low study obscured the to circulate to doctors and
itored.10 Yet, the head of the VIGOR board incidence and it is mechanism cardiovascular risk other health professionals.
was awarded a two year consulting contract based as we worried it was. associated with The New England Journal of
two weeks before the trial ended and as the [Merck employees/consultants] rofecoxib Medicine reported problems
trial was concluding disclosed family owner- were right about the metabo- with the study in an “expres-
ship interest in Merck shares worth $70 000 lite meanings, ie, urine [prostacyclin] data.”13 sion of concern” published in 2006,w10 and the
(£37 000; €55 000).11 12 Although it is not pos- This indicates that, at the least, there were editor in chief has said that the authors “with-
sible to tell whether this financial relationship grounds for suspicion within Merck before held critical data on the cardiovascular toxi-
made any difference, the conflict of interest the VIGOR study was published that Vioxx city of Merck’s drug Vioxx.”14 Nevertheless,
was not a matter of public record at the time was associated with cardiovascular risk. none of the authors has publicly conceded
the trial was conducted or published and of error or taken responsibility for the biased
itself calls into question the independence of Obscuring the risk presentation of the study results. In fact, two
the safety board. Despite the concern articulated by Dr VIGOR authors and the head of the VIGOR
The VIGOR study had enormous financial Scolnick, the published VIGOR study board continue to collaborate on high profile
implications for Merck. If it showed rofecoxib obscured the cardiovascular risk associated research with Merck.15
to have better gastrointestinal safety than with rofecoxib in several ways. The report Except for a 2001 study published in

BMJ | 20 JANUARY 2007 | VOLUME 334 121

THE ROFECOXIB CASE

JAMA that raised questions about the safety of concern”w9 w10 and a correctionw14 and yet, there is hardly a sense of outrage in the
of rofecoxib and the validity of the naproxen publishing a methodological paperw15 and profession about the events that transpired.
hypothesis,w12 few academic researchers other related comments and editorials.w16-w24 Defenders of Merck may say that we do
publicly questioned the company before But other academic medical journals also not know how rofecoxib’s cardiovascular
its voluntary withdrawal of the drug. More- played important parts. In 2001, Circulation risk compares with that of other COX 2
over, Merck selectively targeted doctors published a pooled analysis of 23 phase IIb- inhibitors or traditional non-steroidal anti-
who raised questions about the drug, going V studies examining the association between inflammatory drugs. But the proper place of
so far as pressurising some of them through rofecoxib and cardiovascular risk. The paper these drugs in the medical armamentarium
department chairs.16 had no editorial commentary or critique,w25 is beside the point. With billions of dollars at
even though the study was coordinated inter- stake, Merck conducted the trials, stored and
Short and long term use nally at Merck, the results highly favoured analysed the data internally, paid academic
For several years, Merck continued to inves- the safety of rofecoxib, and five of the seven researchers as consultants to the investiga-
tigate other indications for rofecoxib and authors were Merck employees (the two tive teams and the safety monitoring boards,
conducted additional trials. The increased academic authors acknowledged being paid and maintained heavy involvement in the
cardiovascular risk compared with pla- consultants to Merck). Moreover, in internal writing and presentation of findings. The
cebo was reported in a 2004 analysis of the emails made public through litigation, even journals published the studies, and the aca-
adenomatous polyp prevention on Vioxx an executive scientist at Merck criticised the demic community accepted the findings
(APPROVe) study,w13 which led to the drug’s analysis, stating: “The data appears to have without expressing much concern. Nearly
withdrawal. The financial implications were been interpreted to support a preconceived 107 million prescriptions for rofecoxib
immense not only because of loss of rev- hypothesis rather than critically reviewing were dispensed in the US between 1999
enue but also because of expected litigation. the data to generate hypotheses.”17 and September 2004,21 when the drug was
The key question was when the risk became The Annals of Internal Medicine published withdrawn from the market, and none of the
manifest. If short term use was not associ- the assessment of differences between Vioxx people picking up those prescriptions had
ated with increased cardiovascular risk, Mer- and naproxen to ascertain gastrointestinal the opportunity to consider the true balance
ck’s liability would potentially be drastically tolerability and effectiveness (ADVAN- of its risks and benefits.
reduced. TAGE) study.w26 It later learnt that article was What should we do going forward?
The APPROVe authors, five of whom written by Merck without accreditation,w27 Academic medicine, industry, medical
were Merck employees and the remain- w28 contained errors in the presentation of journals, and government agencies need to
der of whom received con- cardiovascular events with come together to define a set of principles by
sulting fees from Merck, Merck selectively rofecoxib (minimising car- which we can restore faith in collaborations
asserted that the increased targeted doctors diovascular risk), and was on new treatments that can improve patient
risk became apparent only who raised questions conducted for marketing care. We might consider adopting some new
after 18 months of use.w13 about the drug, going purposes, a so called seeding approaches. Academics engaged in industry
This conclusion was based so far as pressurising trial. The journal was quick designed and sponsored studies should insist
on an analysis that was not to condemn ghostwritingw29 that the data are stored on an academic site,
prespecified and a flawed
some of them through and a full correction of the analysed by non-company investigators, and
methodological approach. department chairs errors was published recent- eventually made accessible to the public for
Merck subsequently admit- lyw30 after Merck scientists scrutiny. Several early, large clinical trials
ted that it had incorrectly described the provided an initial, but incorrect explana- of rofecoxib were not published in the aca-
statistical approach, and the New England tion.w31 Many other journals have published demic literature for years after Merck made
Journal of Medicine issued a correction indi- articles with results favourable to rofecoxib them available to the FDA,22 preventing
cating that statements regarding an increase that court documents have shown to be independent investigators from accurately
in risk after 18 months should be removed ghostwritten by scientific writing companies determining its cardiovascular risk using
from the article.w14 Again, mistakes that hired by Merck.w32-w36 meta-analysis. In addition, independent
favoured the company, with colossal eco- audits should be conducted to ensure
nomic implications, made it through the Promoting constructive collaboration that companies follow a standard-
journal peer review process to the profes- The rofecoxib case is bad news for industry, ised, prespecified protocol.
sion and the public. academics, journals, and the public. Merck Independent data and safety
was once one of the US’s most publicly monitoring boards should be
Medical journals admired companies,w37 and its behaviour mandated and their govern-
The New England Journal of Medicine has had a may not be different from that of others in ance should not be under the
prominent role in the story. It published the the pharmaceutical or biotechnology indus- control of the company.
VIGOR and APPROVe studies, respond- try. Journalists have questioned the ethics of Industry should not be
ing to their inaccuracies with “an expression industry and academic researchers.18-20 And allowed to select who

122 BMJ | 20 JANUARY 2007 | VOLUME 334

 THE ROFECOXIB CASE

serves on these boards or allowed to com- 208088, New Haven, CT 06520-8088, USA 7 Waxman HA. Merck documents show aggressive
marketing of Vioxx after studies indicated risk. US House
pensate members after their service. Joseph S Ross is instructor, Department of Geriatrics and Adult
of Representatives Committee on Government Reform,
In considering articles for publications, Development, Mount Sinai School of Medicine, New York, USA 2005. www.democrats.reform.house.gov/story.asp?ID
journals should understand Amos H Presler is research associate, =848&Issue=Prescription+Drugs.
Never Again Consulting, Attleboro, MA, 8 Waxman HA. Memo re: the marketing of Vioxx
that studies with immense In considering articles USA to physicians. Washington, DC: US House of
financial implications require for publications, Representatives, 2005.
David S Egilman is clinical associate 9 Weinblatt M. Memo to Drs. Bjorkman, Neaton, Shapiro,
a higher level of scrutiny journals should professor, Department of Bio Med Silman, and Sturrock re: Interim non-endpoint safety
than others, especially when Community Health, Brown University, analysis of VIGOR—unblinded minutes. November
understand that 18, 1999. Found in sNDA S-007: P088C: Appendix
the study is conducted by the Providence, RI, USA
company with the financial studies with immense Correspondence to: H M Krumholz 3.9.2 at p.2939-2946 (Bates No. MRK-00420015464-
MRK-00420015471. www.vioxxdocuments.com/
stake. Journals should be pre- financial implications [email protected] Documents/Krumholz_Vioxx/VigorDSMB.pdf.
pared to go beyond the usual require a higher level Contributors and sources: HMK’s 10 Merck. Medical affairs procedures and policies.
Procedure 23: collaborative research efforts and
high quality review, paying of scrutiny than others research is focused on determining
optimal clinical strategies and
megatrials. Appendix 2: Merck guidelines for data
particular attention to the pos- and safety monitoring boards. 29 Feb, 1999. Bates
identifying opportunities for Nos MRK-AFK0047772 to AFK0047791. www.
sibility of bias. Articles should be accompa- improvement in the prevention, treatment, and outcome of vioxxdocuments.com/Documents/Krumholz_Vioxx/
nied by editorials by people without financial cardiovascular disease with emphasis on older populations. Merck1999guidelines.pdf.
conflicts of interest. Moreover, ghostwriting He was responsible for the concept and writing of this article 11 Reicin AS. Letter re: financial disclosure for Merck
and is its guarantor. JSR has studied and reported on conflict and Co, Inc sponsored protocol entitled: “A double-
constitutes a false statement of authorship or a of interest in medicine. He participated in the concept and blind, randomized, stratified, parallel-group study
to assess the incidence of PUBs during chronic
false attribution of authorship, and academic design of this article and revised it for critical content. AHP treatment with MK-0966 or naproxen in patients with
researchers who sign off or “edit” original has research interests in the history of public health and rheumatoid arthritis (VIGOR).” 4 Feb, 2000. Merck.
publications or reviews written by industry law, including pharmaceutical research and marketing.
He consults on this topic at the request of plaintiffs in the
Bates Nos MRK-MEW00012 to MRK-MEW00014. www.
vioxxdocuments.com/Documents/Krumholz_Vioxx/
should be penalised unless there is full disclo- Vioxx litigation. He contributed document research and ReicinWeinblatt2000.pdf.
sure of the authorship, such as: “Representa- interpretation to this article and participated in its revision. 12 Merck and Co. Multidisciplinary strategic advisory
board for cox-2 inhibitors, consulting agreement.
tives from XYZ drafted the manuscript; the DSE has studied and reported corporate corruption of February 29, 2000. Bates Nos MRK-STI0037747 to
science. He contributed to the development of this article and
authors were responsible for critical revisions revised the work for critical content. This article arose from STI0037751. www.vioxxdocuments.com/Documents/
of the manuscript for important intellectual access to Merck documents as a result of tort litigation. Krumholz_Vioxx/WeinblattContract.pdf.
13 Scolnick EM. Email communication to Deborah Shapiro,
content.” Competing interests: HMK has research contracts with the Alise Reicin, and Alan Nies re: Vigor. 9 Mar, 2000. Bates
Even the best oversight cannot always American College of Cardiology and the Colorado Foundation No MRK-ABH0016219. www.vioxxdocuments.com/
Documents/Krumholz_Vioxx/Scolnick2000.pdf.
detect mistakes. When journals discover that for Medical Care; serves on the advisory boards of Amgen, 14 Drazen JM. Hidden data counfounds medical journal
information has been withheld or that results Alere, and UnitedHealthcare; is a subject expert for VHA; and
is editor in chief of Journal Watch Cardiology. All authors have
editors. Wall Street Journal 2006 May 19:A11.
are incorrect, they need to rapidly dissemi- been consultants at the request of plaintiffs for recent suits 15 Cannon CP, Curtis SP, FitzGerald GA, Krum H, Kaur A,
Bolognese J, et al. Cardiovascular outcomes with
nate that information and ensure that any web against Merck related to rofecoxib. etoricoxib and diclofenac in patients with osteoarthritis
search that identifies the errant manuscript and rheumatoid arthritis in the multinational etoricoxib
and diclofenac arthritis long-term (MEDAL) programme:
also identifies the correction. Authors should 1 United States Securities and Exchange Commission. a randomised comparison. Lancet 2006;368:1771-81.
Form 10-Q: quarterly report pursuant to section 13 or
sign agreements that they will notify journals 15(d) of the Securities Exchange Act of 1934 (Note 7 16 Fries JF. Letter to Raymond Gilmartin re: physician
intimidation. 9 Jan, 2001. Merck. Bates No
if such information becomes available or face to consolidated financial statements). Merck, 2006;
MRK-ABH0002204 to MRK-ABH0002207. www.
http://phx.corporate-ir.net/phoenix.zhtml?c=73184&
being blacklisted by the journal. p=irolSECText&TEXT=aHR0cDovL2NjYm4uMTBrd2l6YXJ vioxxdocuments.com/Documents/Krumholz_Vioxx/
Our system depends on putting patients’ kLmNvbS94bWwvZmlsaW5nLnht Fries2001.pdf.
bD9yZXBvPXRlbmsmaXBhZ2U9NDMxNDgxMiZkb2 17 Morrison BW. Email communication to Rhonda
interests first. Collaborations between aca- Sperling, Alise Reicin, Deborah Shapiro, et al. re: fw:
M9MSZudW09MTI=.
demics, practising doctors, industry, and 2 Merck. Vioxx (rofecoxib): frequently asked questions. for review [peer]: 2001-ms-2470 (full paper) - due
date Monday, 27 August 2001. 17 Aug, 2001. Bates
journals are essential in advancing knowl- www.vioxx.com/rofecoxib/vioxx/consumer/faq.jsp.
Nos MRK-ACF0005697 to MRK-ACF0005699. www.
edge and improving the care of patients. 3 Morrison BW. Memo to Alan Nies, Barry Gertz, Beth
Seidenberg. Letter to Ken Lasseter; manuscript draft
vioxxdocuments.com/Documents/Krumholz_Vioxx/
Trust is a necessary element of this part- Morrison2001.pdf.
for protocol 023. 18 Feb, 1998. Merck. Bates Nos MRK-
18 Meier B, Kolata G, Pollack A. Medicine fueled by
nership, but the recent events have made NJ0017794 to MRK-NJ0017822. www.vioxxdocuments.
marketing intensified trouble for pain pills. New York
com/Documents/Krumholz_Vioxx/Morrison1998.pdf
it necessary to institute proper systems that 4 Food and Drug Administration, Division of Anti- Times 2004 Dec 19:A1.
19 Meier B, Saul S. Marketing of Vioxx: how Merck played
protect the interests of patients. A renewed Inflammatory, Analgesic, and Ophthalmic Drug
game of catch-up. New York Times 2005 Feb 11:A1.
Products. HFD-550, medical officer review. Vioxx
commitment by all those involved and the (rofecoxib), NDA 21-042/052. Washington, DC: FDA,
20 Berenson A, Harris G, Meier B, Pollack A. Dangerous
institution of these systems are the only way data—retracing a medical trail. New York Times 2004
1999.
Nov 14:A1.
to extract something positive from this unfor- 5 Food and Drug Administration. FDA advisory committee 21 IMS Health. National prescription audit plus time period
briefing document NDA 21-042, s007: VIOXX
tunate affair. gastrointestinal safety. Washington, DC: 2001.
1999 to September 2004, extracted 2004. Plymouth
Meeting, PA: IMS Health, 2004.
6 Merck. Vioxx. In response to your questions: 22 Targum SL. Review of cardiovascular safety database,
Harlan M Krumholz is Harold H Hines Jr professor of medicine cardiovascular system, clinical profile in osteoarthritis rofecoxib. NDA 21-042, S-007. Washington, DC: FDA,
and epidemiology and public health, Department of Medicine, studies. Whitehouse Station NJ: Merck, 2000. www. 2001. www.fda.gov/ohrms/dockets/ac/01/briefing/
vioxxdocuments.com/Documents/Krumholz_Vioxx/ 3677b2_06_cardio.pdf.
Yale University School of Medicine, 333 Cedar Street, PO Box Merck2000CVcard.pdf. References w1-w37 are on bmj.com

BMJ | 20 JANUARY 2007 | VOLUME 334 123