DRUG REGULATION ROSIGLITAZONE WHAT WENT WRONG Over 10 years after the diabetes drug rosiglitazone was approved by regulators, and despite studies on tens of thousands of people, questions remain about its cardiovascular safety. An investigation by Deborah Cohen looks at why this happened.

t was, as one Food and Drug Administration media. Ahead of the meeting, the FDA published A paucity of evidence (FDA) adviser put it, a “perfect regulatory the 765 page report circulated to panel members. Concerns were expressed early on about the storm”—a combination of problematic data, This is far removed from the secrecy shroud- paucity of evidence to support rosiglitazone’s uncertain clinical need, politics, and poor ing proceedings at Europe’s regulator, the EMA. use. According to documents obtained under drug company behaviour. The BMJ has talked to a range of experts close to the Freedom of Information Act, advisers to the INow, 10 years after its approval by regulators the European regulatory process and submitted EMA noted the lack of good evidence during its in the United States and Europe, the widely pre- a series of Freedom of Information requests to approval process. scribed blockbuster diabetes drug rosiglitazone the EMA, but we still have no clear picture of In comments sent to the EMA approval meeting may be about to fold. Two months ago, in July why, after initial rejection in October 1999, the in 1999, one expert adviser noted that without a 2010, the FDA convened a 33 member expert EMA gave market authorisation to rosiglitazone long term study with hard primary endpoints it advisory panel to decide whether it should be in July 2000 in the absence of new evidence. was not clear whether rosiglitazone would have withdrawn from the market in the light of evi- Neither have doctors and patients been told any beneficial impact on cardiovascular disease. dence that it may increase the risk of myocardial that in July the UK’s Commission on Human This adviser also questioned whether you could infarction. Earlier this year a US Senate finance Medicines— in an unanimous vote— advised put a drug on the market without these long term committee report had detailed concerns about the the Medicines and Healthcare products Regu- data and was unconvinced that rosiglitazone in paucity of evidence to support the use of rosigli- latory Agency (MHRA) combination therapy CLINICAL IMPLICATIONS tazone and about the way in which the drug was to withdraw the drug. In offered any advantages According to John Yudkin, emeritus professor evaluated and licensed.1 a statement, the MHRA over what was already of medicine at University College London and At the advisory meeting, members of the pub- has confirmed that the endocrinologist; “No new patients should be available—metformin lic heard a damning analysis of the RECORD trial, evidence now suggests started on rosiglitazone, and patients already and sulphonylurea com- commissioned by the European Medicines Agency that the risks associated taking it should be reviewed and alternative bined, or insulin. (EMA) when it approved the drug in 2000 in order with rosiglitazone out- treatments considered. Those at higher risk Another adviser to determine its safety. Millions of prescriptions weigh the benefits and of heart disease should be advised to stop pointed out that safety later and with the drug still on the market around “that it no longer has a taking the drug.” problems were evident the world, this trial and other post-marketing sur- place on the UK market.” in the data presented by veillance have failed to resolve the concerns. But a “dear doctor” letter sent to UK doctors in GlaxoSmithKline (then SmithKline Beecham) and To date, the FDA and the EMA have decided July advised doctors to “consider alternative asked the panel whether they should postpone that the drug is safe enough to stay on the mar- treatments where appropriate.”2 The MHRA approval until better data were available. ket. But the story reflects badly on almost every- said that it used the information provided by Silvio Garattini, a member of the EMA panel one involved: the regulators, the manufacturer, the Commission on Human Medicines to push in 2000 that approved the drug and director of GlaxoSmithKline, and the clinical community. It for a UK withdrawal as part of the Europe-wide the Mario Negri Institute for Pharmacological has also raised a host of questions. Why did the review by the European Medicines Agency. Research, told the BMJ that the documentation regulators accept such poor evidence on benefit Rosiglitazone is one of two available glitazones presented for approval was initially poor and that and safety for rosiglitazone? Did ­GlaxoSmithKline known to reduce blood sugar and was heralded the studies were of a relatively short duration. The mislead the regulators? Should the drug have as a much needed new approach to improving initial decision to reject the drug was overturned been licensed in the first place and should it now outcomes and long term complications, includ- despite there being no new evidence, he says. be withdrawn? Why haven’t patients in the UK ing cardiovascular disease, for people with type When rosiglitazone was approved, even clini- and Europe been made aware of the concerns 2 diabetes. Of the other glitazones, troglitazone cians who were nominally supportive of the drug about rosiglitazone’s effects? And is the current was withdrawn in 1997 in the UK and in 2000 remarked about the poor evidence base and lack drug regulatory system up to the job? in the USA because of hepatotoxicity; pioglita- of long term clinical trials.4 The FDA meeting was held in the open, in zone remains on the market as a competitor to After approval, three EMA panel members— front of a packed audience including the world’s rosiglitazone.3 whose names were redacted from the minutes

530 BMJ | 11 SEPTEMBER 2010 | VOLUME 341 DRUG REGULATION

Rosiglitazone was Silvio Garattini was on Steven Nissen, a cardiologist Janet Woodcock, head of The Commission for Human approved by the FDA in the EMA committee which at the Cleveland Clinic, the FDA’s Center for Drug Medicines, which advises the 1999. The same year it was approved the drug. published a meta-analysis in Evaluation and Research MHRA, voted unanimously rejected by the EMA, but He was concerned that the the NEJM. This drew public admitted there were to withdraw the drug in July was later approved upon long term risk/benefit was attention to the increased divisions within the FDA this year saying the risks appeal unknown risk of heart attacks about withdrawal outweigh the benefits

sent to the BMJ—remained concerned that “the “There were tremendous expectations about in other classes for treating type 2 diabetes. These long term risk/benefit of rosiglitazone is still Avandia—partly because scientifically it was drugs offer new options to health care profession- unknown and that there are several safety con- extremely interesting. It was a whole new model als who treat people with type 2 diabetes and cerns.” of the way a drug could act. It affected the body represent important advances in drug therapy,” The EMA told the BMJ that the principal safety and its energy metabolism in totally new ways it said.9 concerns at the time were that rosiglitazone that were very interesting and fascinating,” said induced weight gain, with possible serious cardio­ Gale. Was there a need? vascular effects; the induction of anaemia; and One anonymous member of the EMA commit- Before the final approval, the EMA committee that rosiglitazone raises blood lipids, although this tee that approved rosiglitazone in 2000 told the discussed whether there was an unmet need effect was of unknown clinical relevance. BMJ that he had been contacted by respected among the treatment options currently avail- members of the diabetes community to urge him able for diabetes and whether a niche indication Why did the EMA let it through? to approve the “wonder” drug, something he had could be appropriate. Minutes of the meetings Given these concerns and the lack of good evi- not experienced before. show that this suggestion won out. The commit- dence, why did the EMA approve rosiglitazone? Speaking at the European Association for the tee decided that rosiglitazone could be used in Garattini said that rosiglitazone is “an example Study of Diabetes (EASD) annual conference in combination with other oral antidiabetics as a of what happens for drugs that have large com- September 1999, diabetologists urged the use second line treatment in certain circumstances. mercial interest such as the antidiabetic drugs.” of rosiglitazone as a first line treatment for type However, Garattini, who was on the panel, When appealing against a decision not to 2 diabetes. was not convinced. “There was no need for approve their drug, Garattini says pharmaceuti- “Unlike most traditional drugs for type 2 diabe- another antidiabetic drug—there are so many cal companies bring forward opinion leaders who tes, rosiglitazone works in a novel way to reduce already that are more or less the same,” he told are obviously favourable. These paid advisers give insulin resistance, helping the body’s own insu- the BMJ. And as one adviser from the FDA said presentations to the regulators and companies lin work more effectively and offering patients at an advisory meeting, it’s for the regulators turn to them whenever an oral presentation is improved glycaemic control, as measured by to protect public health and not to equip phy- required. fasting plasma glucose. The hope is that this will sicians with a broader array of medicines for On receiving the negative opinion Jan Leschly, slow long-term deterioration,” Dr David Matthews clinical choice. chief executive officer of SmithKline Beecham, of the Oxford Diabetes and Endocrinology Centre told the press it was “a temporary setback,” add- told the meeting. The RECORD trial ing that “in the coming months we will be work- But it was this new way of working—the stimu- Confronted with weak evidence and an appeal ing with the committee to address their concerns. lation of genes that acted on more than blood glu- from GlaxoSmithKline to reconsider its decision We are confident that by the end of March we will cose—that would account for its adverse effects. to reject the drug, the EMA’s committee discussed have demonstrated Avandia’s unique benefits in This should have led to greater caution in the whether additional clinical trials should be the treatment of type 2 diabetes to the CPMP [Com- regulatory process, says Gale. required before or after marketing authorisation. mittee for Proprietary Medicinal Products].”5 In 1999 in the United States, pressure was on In the event the committee approved the drug According to Edwin Gale, a diabetologist and the regulators to fast track rosiglitazone to provide with the requirement for two additional studies adviser to European regulators, in the years before a safer alternative in the same class of drugs as tro- once the drug was on the market. The first was rosiglitazone’s approval diabetologists were also glitazone.8 study 211, a double blind trial of the effect of ros- putting pressure on the regulators and clamouring For example, in the weeks leading up to the iglitazone on cardiovascular structure and func- to use this new class of drug. Some of this clamour FDA’s initial approval meeting in 1999, the Ameri- tion in patients with type 2 diabetes and chronic was fuelled by pharmaceutical analysts touting can Diabetic Association affirmed the importance heart failure. And the second was a randomised its blockbuster potential, which at the time they of glitazones on its website. “These drugs have trial of six years’ duration with the composite said was crucial to SmithKline Beecham’s future mechanisms of action—working directly on insu- outcomes of cardiovascular hospitalisation and growth.6 7 lin resistance—that are not shared by other drugs cardiovascular mortality—the RECORD trial.10 11

BMJ | 11 SEPTEMBER 2010 | VOLUME 341 531 DRUG REGULATION

TIMELINE

1999 2000 2004 APRIL: American Diabetes JULY: Rosiglitazone given market With an increasing number of Association declares that the authorisation in Europe with people taking rosiglitazone, World drug’s properties are not shared by restrictions and with warnings on Health Organization picks up safety 2001 any others, offering new options to heart failure. GlaxoSmithKline (GSK, signals and alerts GSK FEBRUARY: healthcare professionals then SmithKline Beecham) is asked June: GSK ordered to publish FDA approve 2005 2006 MAY: Rosiglitazone approved as to conduct two post-marketing summaries of results of all its new warnings on SEPTEMBER: APRIL: FDA monotherapy by FDA with label trials: one study to look at effect on clinical trials on its website once potential for heart Internal GSK approves new precautions for use in patients with cardiovascular structure; the other a product has been launched in a failure meta-analysis warnings on risks heart failure to assess cardiovascular safety— settlement in New York finds 29% of cardiovascular OCTOBER: Rosiglitazone turned the RECORD trial. non-significant events down by EMA by 14 of 25 votes OCTOBER: Pioglitazone approved increased risk MAY: Internal GSK in Europe of ischaemic meta-analysis cardiovascular finds 31% increase events in ischaemic events

According to Bo Odlind, EMA rapporteur at the commissioning a trial after approval to resolve safety of marketed drugs—a proposition that badly time of the SmithKline Beecham’s fi rst attempt safety concerns the right approach? Garattini needs public debate and reconsideration,” Jerry to get Avandia approved in 1999, the EMA com- says that regulators request additional trials after Avorn, professor of medicine at Harvard Univer- mittee believed it was important to do a cardio- approval to overcome a stalemate when concerns sity, told the BMJ . vascular outcome study rather than one looking about toxicity exist. “It is not the best way because Labelling in Europe by the EMA was updated only at surrogate endpoints such as haemoglobin you need a long time to do the study and mean- shortly afterwards to refl ect the results of the Glax- A 1c —although he is critical of the RECORD trial’s while the drug remains on the market. By the time oSmithKline meta-analysis, with cautions about open label unblinded design. the study is fi nished the drug patent is fi nished so cardiovascular risk. 13 And Odlind isn’t the only person to be critical. there is no inconvenience to the company. This is In an internal memo, Thomas Marciniak, a drug what happened with sibutramine as well, which Sales still booming approver for the FDA, wrote: “We did not review was eventually withdrawn,” he says. Safety concerns didn’t seem to hit sales of the the protocol [of the RECORD trial] prior to study Garattini is concerned more broadly about the drug, and in early 2006 GlaxoSmithKline won implementation. If we had, we would have judged current reliance on drug companies to perform approval for a combined product of rosiglitazone it to be unacceptable” 11 post-marketing surveillance. “The EMA has never and glimepiride in both the US and Europe. 14 Ros- It seems GlaxoSmithKline knew this. A produced a document indicating the percentage iglitazone remained in wide clinical use. At the company slide show cited in a US Senate of fulfi lment of such commitments,” he says. “For end of 2006 and beginning of 2007, the sales of finance committee report noted that the the FDA, it’s about 30%.” The BMJ has asked the Avandia were up.15 16 It was GlaxoSmithKline’s RECORD trial did not provide suffi cient data to EMA to say how many companies carry out their second biggest drug, making an estimated $3bil- test for cardiovascular safety. It also noted that post-marketing surveillance commitments to be lion per year. 17 And it was outselling its rival, GlaxoSmithKline was trying to create studies to told that it has never published a comprehensive pioglitazone.18 counter the PROactive study on rival drug, piogli- report on post-marketing commitments. But a tazone (Actos) that Takeda planned to release. 1 study on the number, type, and status of post- Beginning of the end? In a statement to the BMJ , GlaxoSmithKline said authorisation studies requested by the CPMP or But in 2007, the fortunes of rosiglitazone began that this was not the case. RECORD met its pri- centrally authorised products in the year 2007 to change with the publication of a meta-analysis mary endpoint. “The study confi rmed its primary until 2010 is pending. of GlaxoSmithKline’s study reports by Steve Nis- hypothesis. It showed that cardiovascular hospi- sen and Kathy Wolski in the New England Journal talisation or cardiovascular death (which includes Agencies swamped of Medicine . 19 It claimed that rosiglitazone was heart attack, congestive heart failure, and stroke) Another fl aw in post-marketing surveillance is “associated with a signifi cant increase in the risk was not statistically different between the two that regulators sometimes fail to act on safety of myocardial infarction” compared with placebo groups after an average of 5.5 years of therapy.” information appropriately when they are given or other antidiabetic regimens. When the adverse However, the EMA told the BMJ that it acknowl- it. In 2004, with increasing numbers of people eff ects of the drug became widely known in the edged weaknesses in the trial, including a low taking rosiglitazone, signals of adverse events wider medical community, sales halved. 17 20 event rate in a high risk population of patients were picked up by WHO. They sent GlaxoSmith- Nissen’s ability to access study reports arose with diabetes, a high loss to follow-up, and the Kline an alert about cardiac disease. GlaxoSmith- out of a court case in New York. As part of a set- open label design of the study. Kline conducted a meta-analysis and confi rmed tlement with the state over GlaxoSmithKline’s “The RECORD study was designed some 10 an increase in cardiac events to the FDA and the non-disclosure of possible heightened suicide risk years ago. Since then the design of post market- EMA in 2006. among teenagers taking antidepressant paroxet- ing studies has evolved,” a spokesperson for the The FDA has been accused of sitting on the ine (Paxil) the company had to put all its recent EMA said. reports and not suffi ciently alerting the public.12 clinical studies on a website. “These data were kept from the public because Whatever the criticisms of this particular meta- Problems with post-marketing surveillance of acceptance of the proprietary nature of com- analysis were, it allowed academics to scrutinise But even without the flawed design, was panies’ trial results, even when they concern the the study summaries.

532 BMJ | 11 SEPTEMBER 2010 | VOLUME 341 DRUG REGULATION

2007 2008 2009 2010 MAY: New England Journal of Updated internal MARCH: International FEBRUARY: US Senate finance committee releases report that includes Medicine publishes meta-analysis GSK analysis Journal of Cardiology internal FDA safety report calling for drug to be withdrawn reporting 43% increased risk of finds no risk meta-analysis finds JUNE: David Graham’s study leaked to the Pharmalot blog. It is published myocardial infarction of myocardial no risk of myocardial in JAMA regardless. At the same time, another JAMA journal, Archives of JUNE : NEJM publishes interim analysis of the RECORD trial infarction or infarction Internal Medicine, publishes an updated meta-analysis by Steve Nissen JULY: FDA advisory committee finds increased cardiac other major JUNE: RECORD trial 13-14 JULY: FDA advisory committee meeting held. FDA drug approver gives ischaemic risk but votes to keep drug on market cardiovascular published in the damning verdict on the RECORD trial. Majority of committee vote either to OCTOBER: European Medicines Evaluation Agency asserts events Lancet . EMA adds withdraw the drug or restrict it severely positive benefit-risk profile, recommends new warnings for a statement to its 15 JULY MHRA meet. Commission on Human Medicines patients with ischaemic heart disease scientific information vote to withdraw rosiglitazone NOVEMBER: FDA approves new boxed warnings that document saying there TIDE trial suspended by the FDA drug may increase myocardial ischaemic events, including was no difference 19-22 JULY: EMA (left) meet to discuss rosiglitazone myocardial infarction, though evidence “inconclusive” in the number of 26 JULY: the MHRA send out “dear doctor” letter advising DECEMBER: UK Medicine and Healthcare products adjudicated primary doctors to consider alternative treatments where Regulatory Agency warns drug might be associated with endpoints between the appropriate small increased risk of cardiac ischaemia arms of the study SEPTEMBER: EMA will finalise its review

“It’s important to realise what an important GlaxoSmithKline’s own analysis showed that events to the board for adjudication, missed end- role publicly available trial results data played in 321 (14.5%) patients treated with rosiglitazone points, insuffi cient collection of information, and the rosiglitazone story. Having this information compared with 323 (14.5%) controls experi- issues with data handling. posted on the GlaxoSmithKline website made it enced either cardiovascular death or hospitali- Specifi cally, he detailed eight failures to refer possible for Steve Nissen to perform his critical sation. In essence, rosiglitazone did not cause patients for adjudication—all in those taking meta-analysis published in NEJM in May 2007, more cardiovascular problems than metformin rosiglitazone. “The eight cases weren’t the only which really ‘broke the case wide open’ on this or sulphonylureas. problem cases I found. I classifi ed the problem matter,” says Jerry Avorn, professor of medicine The EMA accepted these findings when it cases into categories and the eight were ones of at Harvard Medical School. received the fi ndings of the completed trial in ‘failure to refer for adjudication.’ Of 549 patient “Requiring the posting of clinical trial results 2009. “No diff erence in the number of adjudi- case report forms reviewed I found 70 serious on clinicaltrials.gov should help provide a warn- cated primary endpoint events for rosiglitazone problems, four to one favouring rosiglitazone,” ing system for other drug risk issues in the future, (321/2220) versus active control (323/2227) he said in an interview. as will the growth of FDA’s new Sentinel system (HR 0.99, CI 0.85-1.16) was observed” the EMA’s GlaxoSmithKline says there was no wrongdo- for post-marketing surveillance. The Avandia case scientifi c summary said.13 It took the FDA to thor- ing. In a statement to the BMJ , a GlaxoSmithKline provides compelling evidence of the vital neces- oughly analyse the trial, which raises questions spokesperson said “An inspection of the RECORD sity and importance of both of these develop- about the EMA’s ability to oversee post-marketing study by the FDA concluded that there was no ments,” he added. trials. evidence of systemic or pervasive fi ndings that The FDA had the individual case reports to do would undermine the reliability of the RECORD FDA rules allowed greater scrutiny a more thorough job, but not the resources. In data.” For those with access only to the medical lit- house statisticians could not wade through the Nevertheless, in his report Marciniak said: erature, unpicking the evidence behind ros- voluminous RECORD trial dataset—running to “While these numbers may seem small com- iglitazone has not been easy. Bristol University 1438 pages for one patient, and several hun- pared to the size of the trial, note that about 15 diabetologist, Edwin Gale—who was chair of dred pages for most of the other 4500 patients. more [myocardial infarctions, MI] in the rosiglita- the EMA’s scientifi c advisory group on diabe- But Marciniak analysed 549 patient case report zone arms are needed to change the GlaxoSmith- tes—complained in 2001 in the Lancet how lit- forms, including 278 from the rosiglitazone Kline MI results to a relative risk of 1.4 and a p tle data on rosiglitazone had been placed in the group and 271 controls. Of these 549, 100 were value of 0.042”—which would make an increase public domain. Company documentation at that a random sample and the rest had been the sub- in myocardial infarction statistically signifi cant. time cited fi ve confi dential fi les, 13 abstracts, and ject of adjudication disputes. Marciniak said it was “a huge challenge to try four papers (two of which were clinical) for ros- Marciniak found problems that cast doubt on to fi nd those few needles in the haystack,” and iglitazone.3 GlaxoSmithKline’s analysis, which he detailed it’s certainly too much for the under-resourced The regulators have access to more informa- in his damning report for the July FDA advisory European regulator. But it’s a job that’s needed, tion. Since the 1950s, FDA rules have required committee. he says. According to Odlind, unlike the FDA, the drug companies to turn over all individual His concerns were over study design, con- EMA takes a top down approach—it takes the patient case reports from their clinical studies, founding variables that could have biased the study summaries and asks the drug companies not just the statistical summaries but the reports overall fi ndings, and the conduct of the study. for more data if it sees fi t. that permit re-analysis of how each case was He considered that these limit “any reassurances Marciniak is a keen advocate of accessing raw coded. that RECORD can provide regarding the CV safety data: “You will not fi nd the truth in drug review It was the availability of these case reports of rosiglitazone.” 1 unless you dig,” he told the BMJ . “I believe the that allowed Thomas Marciniak, an FDA medi- Much of his concern hinged on data ascertain- FDA approach is better or potentially more thor- cal offi cer, to scrutinise the RECORD trial. He was ment in patients who stopped the trial or were ough than the EMA’s, but it also needs more asked in October 2009 to review the cardiovascu- lost to follow-up. Marciniak detailed 11 diff er- complete implementation. One public sugges- lar events in the trial. ent conduct problems including failure to refer tion has been to release the raw data to academic

BMJ | 11 SEPTEMBER 2010 | VOLUME 341 533 DRUG REGULATION

ЖЖWatch the BBC Panorama programme, “A risk worth taking?” (www.bbc.co.uk/programmes/b00tr25t) bmj.com archive ЖЖAnalysis: Patients and the public deserve big changes in evaluation of drugs (BMJ 2009; 338:b1025) ЖЖResearch: Adverse cardiovascular events during treatment with pioglitazone and rosiglitazone (BMJ 2009;339:b2942) ЖЖEditorial: Rosiglitazone or pioglitazone in type 2 diabetes? (BMJ 2009;339:b3076) ЖЖResearch: Association between industry affiliation and position on cardiovascular risk with rosiglitazoneBMJ ( 2010;340:c1344) ЖЖFeature: Rosiglitazone, marketing, and medical science (BMJ 2010;340:c1848)

organisations. That would be an advance, but I [were] taken out of context, Where next for diabetes believe that most of the academic organisations which therefore are incomplete drugs? don’t realise that you need not only raw compu- and misleading.” They also said While the focus has been firmly ter data files but also the case report forms and a that the “assertion that this study on rosiglitazone, what about variety of other source documents to understand informed GSK’s views about heart pioglitazone? Its manufacturer completely a study,” he said. attacks and Avandia is completely Takeda had benefited from the GlaxoSmithKline had employed an independ- unfounded.” controversy of rosiglitazone. ent statistician, as is a prerequisite for the publica- But like rosiglitazone, piogli- tion of pharmaceutical sponsored trials in some Decision time tazone is associated with an medical journals, notably JAMA. Marciniak does As Gerald Van Belle, director of the increased risk of oedema, heart not blame the statisticians for not picking up the Clinical Trials Center at Washing- failure, and bone fracture.23 flaws in the case reports, as they can work only ton University and FDA advisory Will regulators, As Professor Van Belle said, with what is given to them. panel member, put it—it was a industry, and the he doesn’t want to be sitting “perfect regulatory storm.” clinical community do a at an FDA advisory meeting Shining a spotlight on GlaxoSmithKline At the hearing in July, the FDA better job for patients in three years’ time discussing Neither the Nissen meta-analysis nor Marciniak’s panel voted that the available next time? pioglitazone. Professor Gale is digging show GlaxoSmithKline in the best light. data supported a conclusion that also concerned. In a letter to the In a Senate committee finance report published rosiglitazone increases cardiac ischaemic risk in UK regulator in 2008 seen by the BMJ, he wrote in February 2010, GlaxoSmithKline executives type 2 diabetes patients. that “there is an urgent need to determine the stood accused of focusing “on strategies to mini- But the question of what to do raised mixed safety of pioglitazone”. mize or misrepresent findings that Avandia may answers. There were five options to choose from, “Pioglitazone may or may not prove to be increase cardiovascular risk.” Internal documents ranging from the removal of the black box warn- safer than rosiglitazone. There is an urgent show that GlaxoSmithKline quickly published an ing to withdrawal—and the breakdown of the votes need for more and better data addressing this interim report of the RECORD trial to counter the would mean that it could be subject to interpreta- issue. On present evidence, its safety cannot and negative effect of Nissen’s meta-analysis.1 tion. A majority of votes recommended keeping should not be assumed,” he wrote. Takeda say The company, however, said in a statement that the drug on the market, but with more warnings that they continuously monitor the safety and the interim analysis of the RECORD study was con- or restrictions. But viewed in a different way, a efficacy of their compounds. ducted urgently to “gather additional information majority recommended either removal or severely Meanwhile other anti-diabetes drugs using about the potential risk for patients.”21 restricting access to rosiglitazone. However, they a similar pathway are in development: the To add to the company’s woes, on the day of the voted to continue the TIDE trial—a study commis- chequered history of the glitazones not having committee meeting this year the New York Times sioned by the FDA to assess rosiglitazone’s cardio- deterred manufacturers. According to reports, splashed with allegations that GlaxoSmithKline— vascular safety—which some argue the FDA should Dr Reddy’s Laboratories and Nordic Bioscience’s then known as SmithKline Beecham—had started have asked for at the outset. This trial has since partial PPAR-gamma agonist balaglitazone met a secret trial to see if rosiglitazone was safer for the been put on “clinical hold” by the FDA. its primary endpoint in its first phase III trial heart than pioglitazone.22 “Not only was Avandia Janet Woodcock, director of the FDA’s Center in patients with type 2 diabetes (reduction in no better than Actos, but the study also provided for Drug Evaluation and Research, admitted the glycated haemoglobin). It was claimed to be clear signs that it was riskier to the heart. The advisory meeting featured conflicting opinions “non-inferior” to pioglitazone. The companies company did not post the results on its Web site between two branches of the FDA. are currently in discussions with regulators and or submit them to federal drug regulators, as is Marciniak agrees: “The Avandia advisory com- hope to eventually file the drug in the European required in most cases by law,” the article alleged. mittee meeting was a battle between the “diggers” Union and the United States.24 And both Roche To date, there are no head to head trials of the [Marciniak and a few others] and the “delibera- and Metabolex have drugs in phase two trials.25 two drugs in the public domain—companies do tors,” [meaning] the rest of the FDA and most of Will regulators, industry, and the clinical not have to demonstrate added therapeutic value the committee and, I believe, the EMA.” community do a better job for patients next under European Union or United States law. Trying to gain an overall perspective of delib- time? The news story went on to quote from an inter- erations within the EMA has been far trickier. The Deborah Cohen is investigations editor, BMJ nal company memo: “‘Per Sr. Mgmt request, these BMJ attempted to speak to people who had sat on [email protected] data should not see the light of day to anyone out- panels for the MHRA and the EMA. But they were Competing interests: None declared. side of GSK,’ the corporate successor to SmithK- bound by confidentiality clauses. The EMA would Provenance and peer review: Commissioned, externally peer line.” not release the names of the members of the sci- reviewed. GlaxoSmithKline, however, say that the entific advisory group discussing rosiglitazone References are availiable on bmj.com. emails were “selectively disclosed by lawyers under the Freedom of Information Act. Secrecy Cite this as: BMJ 2010;341:c4848 seeking damages” and that “other documents also shrouds the UK’s regulatory agency. See EDITORIAL, p 513; LETTERS, 519

534 BMJ | 11 SEPTEMBER 2010 | VOLUME 341 DRUG REGULATION

COMMENTARY NICK FREEMANTLE What can we learn from the continuing regulatory focus on the thiazolidinediones? Should rosiglitazone be withdrawn case of rosiglitazone, sponsors, the 30 day outcome in a trial of substantial increases in the size from use? An answer is hampered investigators, regulators, and tenecteplase versus alteplase, and length of follow-up in diabetes by the inadequacies of the data journal editors seem to have been the ASSENT-2 investigators knew trials, and should ensure that high collected to date. In particular, content to accept truly dreadful the vital status of all but six of the rates of loss to follow-up are no regulators have tolerated loss to standards of methodological 16 949 participants randomised to longer considered acceptable. But follow-up in trials designed to quality. Until the recent RECORD that comparison.7 But for reasons what about other clinical areas? It examine the effects of rosiglitazone study,5 trials have used symptoms that are unclear, regulators have would make sense to have better on surrogate outcomes and surrogates rather than major been less critical with trials that use quality trials in all contexts, rather (haemoglobin A1c) despite relying morbidity and mortality as the surrogate endpoints. than waiting until safety concerns on these same trials to provide outcomes of interest and were This acceptance of higher arise and reacting, at which point evidence on safety (mortality sabotaged by unacceptably high rates of loss to follow-up in trials trials will be harder to run as the and morbidity). The more robust and sometimes hidden losses to where outcomes are symptoms experience of RECORD shows.5 regulatory approach required for follow-up. or surrogates makes no sense. In order to learn from our new diabetes drugs introduced This ill conceived distinction is mistakes, we must improve in late 2008 should prevent this Loss to follow-up highlighted when we use those the quality of safety data situation recurring in diabetes, but Loss to follow-up can undermine same trials of symptoms and from clinical trials on all new if we are to avoid similar problems randomisation and introduce bias, surrogates to assess safety, since healthcare interventions. in other clinical areas, we need a far reducing the reliability of the trial.6 adverse events will include serious Sponsors should not be content more widespread overhaul in the If patients leave a trial because morbidity and mortality. Even when to design and undertake poor standards of regulatory trials. of health reasons, and if there trials examine serious morbidity quality trials. Regulators should Few would disagree that large is a risk that the reasons for loss and mortality, loss to follow-up insist on high quality trials scale randomised trials provide to follow-up may differ between remains a problem. The RECORD with low loss to follow-up, the most reliable evidence on the the randomised groups, then the trial, conducted to examine and adequate examination safety of interventions.1 Through results of that trial may no longer the safety of rosiglitazone and of potential safety challenges randomisation, both known and be explained simply by chance or sponsored by GlaxoSmithKline in early in the development of new unknown biases are distributed the treatment allocation but may be the UK,5 failed to follow up survival pharmaceutical agents across between the experimental groups due in whole or in part to selection in 127 participants (2.9%). In the all clinical areas. Investigators by the play of chance.2 Because of bias. In the ADOPT trial, which context of an observed mortality should seek to keep participants random allocation of participants, examined rosiglitazone, metformin, of 6.6% this failure is challenging in trials, and journals should any observed difference between and glyburide, 40% of participants and requires a series of exploratory be more critical of the trials that groups on an outcome of interest did not complete the one year analyses examining the possible they are offered for publication. may be attributed either to the follow-up.4 The article describing implications. These result in the A trial like RECORD is clearly of play of chance or to the random the main results of the trial states conclusion that rosiglitazone legitimate public interest and treatment allocation. If chance that safety data are based on all could be associated with either an should be published and available is unlikely (for example, if the randomised patients, and yet 5% increase or a decrease in mortality for scrutiny,5 but the limitations of confidence intervals are relatively of patients had disappeared before given different assumptions on the such a study should be described narrow and the corresponding the first follow-up visit. fate of those lost to follow-up. As more carefully in order to achieve a P value small) the only plausible In the case of for the non-mortality outcomes, balanced portrayal of the facts and alternative is that the randomised the 11.7% rate of loss to follow-up their interpretation. treatment is responsible for the rosiglitazone, sponsors, in RECORD makes interpretation Nick Freemantle professor of clinical difference. investigators, regulators, very difficult, although the FDA epidemiology and biostatistics, School and journal editors of Health and Population Sciences, When there are few or no good has rightly long emphasised the University of Birmingham, Birmingham randomised trials, we have to seem to have been importance of all cause mortality as [email protected] look at other sources of evidence, content to accept truly an outcome measure.8 Competing interests: The author has but this should not have been completed the unified competing interest form dreadful standards of at www.icmje.org/coi_disclosure.pdf (available necessary with rosiglitazone. Wider change methodological quality on request from the corresponding author) More than 30 000 participants In the US, the FDA has made and declares no support from any organisation have been randomised in trials of Regulators are well aware that changes to improve the quality for the submitted work; funding for research, rosiglitazone.3 4 This should allow loss to follow-up is a big problem of data on the safety of new travel, and consulting from Novo Nordisk, Sanofi Aventis, Johnson and Johnson, and Eli Lilly; and 9 us to estimate small differences in clinical trials examining serious drugs for diabetes. Requiring a no other relationships or activities that could in event rates attributable to morbidity and mortality and that predefined upper boundary to appear to have influenced the submitted work rosiglitazone. However, not all very low losses can be achieved. exclude the possibility of excess References are availiable on bmj.com. trials are alike and, sadly, in the For example, when comparing cardiovascular risk will lead to Cite this as: BMJ 2010;341:c4812

BMJ | 11 SEPTEMBER 2010 | VOLUME 341 535 HEALTH WHITE PAPER

WHITE PAPER Join the discussions about the white paper on

Liberating the NHS: Transparency in outcomes doc2doc, BMJ Group’s global online clinical community, – a framework for the NHS A consultation on proposals at doc2doc.bmj.com/whitepaper

Commissioners doing it for themselves All general practitioners in England will soon be expected to commission the majority of health care services for their patients. Jane Cassidy talks to two groups already fulfi lling this role

hen a group of Northampton- “Managers stopped services without thinking ity is maintained are paramount. “It’s not about shire general practitioners about the implication for patients,” she said. This accountancy and contracting. Other people will started commissioning patient led to meetings between like minded GPs, which do that better than you,” he said. services, little did they know that led to Nene Commissioning. Providing education to support those involved three years later they would fi nd She is now able to save time and stress for her- in change helps smooth the way. Nene provides Wthemselves in the media spotlight. self and patients by helping design streamlined several training programmes. The publication of the government’s health projects such as the county’s end of life service. Behaviour change from some medical profes- white paper in July kick started a stream of There is now a link nurse at the local hospital sionals is also vital, say the Nene doctors. GPs will requests from journalists to the Nene Commis- to ensure whenever possible that patients die in need to ask themselves: “How does my behaviour sioning office in Northampton.1 An interview the setting they choose. This means Dr Poggi has aff ect relationships with other practices in the with Nene’s chairman, Darin Seiger, on BBC Radio only one person to liaise with instead of many. It’s locality and in the consortium?” 4’s Today programme sparked 12 000 hits to the one instance of how Nene tries to provide a more They will no longer be able to say: “I do it this organisation’s website in one day. 2 seamless service. way because I’ve always done it this way.” Now media interest is being replaced by calls Dr Seiger had similar motivations. In 2006 he The Nene GPs are used to sharing best practice, from doctors keen to shadow the GPs’ work. They was the medical director of a local GP cooperative scrutinising activity such as referral patterns to want to fi nd out more about a role all family prac- and, along with clinical colleagues, had to explain identify variations and to improve quality. Though titioners are expected to take on by 2013. to patients the eff ect of the PCT cuts. Telling cou- they stress this is done in a supportive rather than ples that NHS funding for in vitro fertilisation had fi nger pointing way. Making it work stopped was particularly painful, he recalls. The Nene group is the largest of its kind, represent- Practice based commissioning, introduced by Answering concerns ing 76 practices, more than 350 GPs, and a patient the last government, gave GPs some power to What do they think of the concern that hospitals population over 650 000—94% of the county. A design care for patients through commissioning. that fail to compete with other providers in the not for profi t social enterprise set up in 2007, it is The Nene group wanted to operate independently market will be starved of cash under the new run by a board of nine elected GPs supported by a from the local PCT as a GP led organisation. To do commissioning arrangements? small management team. this, they formed a community interest company. Over the years, a wedge has been driven Ideas for new and improved services, along A support team deals with the administrative between GPs and secondary care and these rela- with concerns, are aired at monthly meetings in side of the commissioning process, so GPs can get tionships need to be rebuilt, they say. “There have four locations county- on with the task of mak- previously been perverse incentives in the system TIPS FROM THE COMMISSIONING GPs wide. Progress reports ing clinically sensitive whereby managers see collaboration across a are fed back regularly. The • Don’t get hung up on the technical side of decisions. They say the pathway, between primary and secondary care, group’s website is bulging things—accountancy, contracts. The most PCT is now supportive. as a threat to their income. While some of these important thing at the moment is to talk about with examples of how its GP director Matthew incentives will still be present, I believe that clini- forming relationships innovative ideas are win- Davies believes com- cally led commissioning can identify and break • Ask yourself, “What will this mean for me as a ning awards and trans- GP, for my practice, my locality?” missioning can lead to a them down,” said Dr Davies. forming patient care. It • Learn the lessons of GPs who have already set feeling of empowerment But acute trusts will need to adapt too. The is easy to see the benefi ts up commissioning organisations and increased morale in white paper challenges everybody in the system for patients, but what’s in • Don’t let your PCT bounce you into making the profession. to work diff erently. it for clinicians? decisions too soon. You’ve got to sit back, talk But what about cli- The Nene doctors are obviously highly moti- How about greater job to each other, and reflect on where you want nicians anxious that vated. Dr Seiger lists endurance triathlons among satisfaction, less frustra- to go they will be expected to his hobbies. Dr Poggi has an MBA. Dr Davies has tion, and lower stress lev- • Read the white paper. It’s not going to go away become hybrid doctor- been involved in commissioning for several years. els? The move away from managers? But how will others manage such thorny issues as primary care trust (PCT) control has resulted in all “People get hung up on technical skills. I don’t confl icts of interest, performance related pay, and of these, say three of Nene’s GP directors. think there is a need for these. It’s leadership skills anxiety over bankruptcy? Nene deputy chair, Raff aella Poggi, explains that are important,” he said. Softer skills are cru- The doctors point to some troubleshooting solu- her involvement evolved from dissatisfaction cial to commissioning success. Winning hearts tions. Performance can be managed using agreed about a PCT merger four years ago that led to a and minds when redesigning services and the standards everyone can be supported to achieve. series of “blunt cuts.” ability to negotiate and infl uence to ensure qual- A group decision can be made about how to share

536 BMJ | 11 SEPTEMBER 2010 | VOLUME 341 HEALTH WHITE PAPER

Darin Seiger: Commissioning gives Matthew Davies, GP: Commissioning Dr Raffaella Poggi: Commissioning Laurence Buckman: Commissioning GPs power to design care increases morale has saved time and stress more an opportunity than a threat

quality money. Clear governance rules and trans- “We have a good relationship with the PCT by firm financial governance and legal protec- parency will prevent conflicts of interest, they say. and our monthly meetings are extremely useful. tion, he can’t see why GPs would find such a If a practice is overspending, there needs to be We’re learning the ropes from an excellent man- responsible role attractive. Anyone who takes a peer supported process in place to support com- agement team,” said Dr Mehta. up the job also needs to be given a return road missioners working alongside the local medical It is, however, a steep learning curve, he says, back to full time clinical work afterwards. committee. which makes it essential that GPs get support Laurence Buckman, chairman of the BMA’s GP The accusation of a privatisation agenda under- from managers who have already done the job. Committee, says GP commissioning could work pinning the government plan is also rejected by A doctor in the area for 13 years, he believes well to benefit patients and save money. the Nene doctors, who say their organisation is GPs are in an ideal position to influence health “We see this as much more of an opportunity entirely committed to the NHS. care. One example of a service reshaped under than a threat,” he said, although he agrees key What about concerns over time management? the new system was a areas need to be clari- How do you encourage good clinical directors to costly specialist GP service “People get hung up on fied. Among these are come forward? for diabetes where patients technical skills. I don’t potential problems if The answer, they say, lies in paying well and were failing to turn up the process is thwarted offering a professional management structure that for appointments. Non- think there is a need for by central government, allows focus on areas in which GPs have clinical attendance was picked up these. It’s leadership underfunded to the flair. They shouldn’t be expected to put in extra from conversations with point that no one can hours on a goodwill basis or get bogged down by patients and from clinic skills that are important.” commission, or the tasks that can best be done by administrators. staff. whole process goes so In Redbridge, north east London, Anil Mehta Diabetic specialist nurses now run the clinics, fast that proper governance arrangements aren’t is no less enthusiastic about commissioning, spending up to an hour with patients, talking to put in place. although he admits to putting in extra hours. them about a range of issues such as diet. Other big worries are the effect on clinical care “Initially, I thought it would be a one day a week “The immediate benefit for us is that our if practices are too busy commissioning, the risk job. I’ve ended up doing a lot more in between and chronic disease management of patients of the private sector moving in, and what hap- after hours.” improves,” said Dr Mehta. pens if the money runs out. He does eight clinics a week and is also clini- “What happens to a patient who doesn’t get “We’re very concerned about GPs being made cal director of Fairlop Polysystem, one of five what he needs? He ends up at the door of acci- the scapegoats for financial constraints or being advanced practice based commissioners run dent and emergency, perhaps attended by a jun- made to look like they’re the ones closing hospi- by GP boards. Each supports 40 000 to 70 000 ior member of the profession who tells him to go tals,” he said. patients in one geographical area. They are on back to his GP. This is a bizarre waste of time and The deadline for responses to the white paper course to control 80% of the commissioning resources.” is 11 October. All eyes will be on the govern- budget by the end of the year. He dismisses the notion that doctors shouldn’t ment’s reaction to the consultation and whether He is currently using his weekends to get to get involved in contracting decisions, saying GPs they move to clarify their radical plans. grips with population management software that have a dual role as both clinicians and custodi- Jane Cassidy is a freelance journalist, Hertfordshire allows him to scrutinise patient data in detail. ans of public money. [email protected] Each polysystem is charged with redesigning “Unless we look at where the money is being Competing interests: None declared services to meet the specific needs of its immedi- spent, we can’t possibly improve health care. Provenance and peer review: Commissioned; not ate community. Set up to allow GPs to have real A lot of wastage is never found unless you get externally peer reviewed. 1 Department of Health. Liberating the NHS: control over health services in their area, the involved.” commissioning for patients. A consultation on ­polysystems work collaboratively with a commu- However, there is uncertainty among GPs proposals. 2010. www.dh.gov.uk/prod_consum_dh/ nity panel and share local knowledge. about the white paper, he concedes. For instance, groups/dh_digitalassets/@dh/@en/documents/ digitalasset/dh_117705.pdf. In contrast to the Northamptonshire model, the who will be willing to take on the role of the 2 Nene Commissioning. www.nenecommissioning. local PCT worked closely with GPs to create and accountable officer in the new consortiums? com/. roll out the programme last year. Unless there is adequate training, backed up Cite this as: BMJ 2010;341:c4488

BMJ | 11 SEPTEMBER 2010 | VOLUME 341 537