Steven Nissen and Robert Califf: A Conversation About Rosiglitazone

August 21, 2007

Editor's Note:

The May 2007 online publication of the meta-analysis "Effect of Rosiglitazone on the Risk of Myocardial Infarction and Death From Cardiovascular Causes," in The New England Journal of Medicine (NEJM) launched a controversy that has been difficult to navigate. In an attempt to clarify the debate, our colleagues at organized a videotaped conversation between Steven E. Nissen, MD, MACC, Medical Director, Cleveland Clinic Cardiovascular Coordinating Center, Cleveland, Ohio -- and co-author of the NEJM article -- and Robert M. Califf, MD, Professor of Medicine, Vice-Chancellor for Clinical Research, and Director, Translational Medicine Institute, Duke University, Durham, North Carolina. Drs. Nissen and Cardiff review the meta-analysis and the US Food and Drug Administration (FDA) advisory panel hearing, as well as discuss the wide-ranging implications of the ongoing clinical and political debate. It is a thoughtful, collegial conversation that helps parse this controversy and what it means for clinical practice and public policy.

Kristin M. Richardson
Editorial Director
Medscape Diabetes & Endocrinology

Dr. Califf: Hello and welcome to the This is Rob Califf from Duke, and I'm here with Steve Nissen from the Cleveland Clinic. We're really delighted to have an opportunity to have a dialogue about Avandia [rosiglitazone] and its recent events. Steve, how are you doing?

Dr. Nissen: I'm doing great, Rob.

Dr. Califf: I thought, Steve, it would be fun for us, being mutually interested in not only the specific issues with Avandia but the more general issues, to perhaps cover 3 topics. The first would be the specifics of Avandia: What is the story with the drug; what happened with it? The second might be the policies related to the regulation of drugs and the development of drugs. And then finally we might cover the broader issues of the safety of drugs, particularly the cardiovascular safety and how we think things should change looking at the future.

Dr. Nissen: I look forward to that. All 3 topics are very interesting.

Dr. Califf: That's great. Now you were obviously very involved, almost from the beginning of the public proceedings. Maybe it would be good for you to give a summary of what your understanding is of the current status of Avandia, which, as we've discussed before, obviously is a very complicated drug.

Dr. Nissen: Well, these drugs are all complicated. The TZDs [thiazolidinediones], of which Avandia is one, are PPAR [peroxisome proliferator-activated receptor]-gammas, and the PPAR-gamma drugs are turning on and off a large number of genes. We don't know what most of those genes do, and that, in fact, may represent a real problem with the class, since we've now seen a large number of these drugs actually either not make it in development or be withdrawn because of toxicity. The status of Avandia today, post the FDA advisory panel meeting, is that the panel voted by a vote of 20 to 3 that they agreed that rosiglitazone, or Avandia, increases the risk of ischemic cardiovascular complications. The 3 meta-analyses that were published all showed it.[1,2] The increase in risk was in the range of 30% to 40%, and obviously this is of great public health concern for a drug that's being used to treat diabetes. The majority of diabetics, as I think we all know, die of cardiovascular complications.

What we're waiting for now is for the FDA to actually act. The FDA has to now come up with a label change that reflects the increased risk for rosiglitazone, and we haven't seen that label language yet.

Dr. Califf: Well, Steve, this is pretty confusing to the average cardiologist like me. I don't get to round that much these days, but I'm out on the floor today rounding, talking to interns and residents and fellows. And this is a class of drugs, at least if you listen to the scientists talk about, that has all these tremendous effects. It's good for HDL [high-density lipoprotein] cholesterol; it lowers blood pressure; obviously the glucose gets better -- a lot of reasons why it should be a great drug. And if there is a cardiovascular risk that was so obvious, surely that would have been picked up early on in the drug development program.

Dr. Nissen: Well, it's interesting. It actually was picked up early in the drug development program. If you actually go back to 1999, when the drug is approved, and look at the studies that were done in around 2500 patients, preapproval, that were reviewed by the FDA advisory panel, what you see is approximately a relative risk of 1.8 for myocardial infarction or other myocardial ischemic events in the registration package. And it was actually of sufficient concern that the FDA reviewer said that the drug, if it were to be approved, should only be approved with the requirement that a large safety study be performed immediately post approval. So the signals have been there for a long time.

In addition, now nearly 2 years ago, the company that makes rosiglitazone, GlaxoSmithKline, we now know -- although we were not aware of this at the time we did our analysis -- the company had taken 42 clinical trials, all the randomized clinical trials done with the drug, and had performed their own integrated analysis. It reported to the FDA in August of 2005 that their analysis suggested a 31% increase in the risk of ischemic cardiovascular complications. So this problem that many people heard about after our New England Journal of Medicine manuscript was published on May 21, it actually goes back a long way.

Dr. Califf: So, Steve, what you're telling me is that in your view there actually were signals early on, but I think you'd agree it is kind of confusing because in a sense there was a public statement in terms of a company-done meta-analysis on this issue, but it didn't get into the consciousness of most prescribers, and didn't appear in the label as a risk of myocardial infarction. Is that correct?

Dr. Nissen: That's right. It's actually one of the most unusual situations that I've certainly ever been involved with. There was a lawsuit by then-New York Attorney General Eliot Spitzer in 2004 that was settled out of court with GlaxoSmithKline, and part of that settlement required the company to disclose the results of all clinical trials, not just rosiglitazone, but every study that they did, something that many of us have advocated across the entire industry. And so it turns out we now know that sometime -- we don't know exactly when, perhaps in late 2006 -- the company did in fact post on a Web site the results of their meta-analysis showing a 31% increase in risk. We do know from FDA documents that the company told the FDA in August of 2005, and again a year later, that they were seeing very strong evidence in their own meta-analysis of an increased risk of myocardial infarction and other ischemic events. But for whatever reasons, neither the company nor the FDA revealed that in a public way.

Dr. Califf: Is there a mechanism for this increased risk that you're describing? I think most physicians have been trying to think that you can predict what's going to happen from treatment with the drug by understanding the mechanism.

Dr. Nissen: Unfortunately, as I think you know all too well, Rob, understanding the mechanism of how a drug works has never proven to be a very reliable predictor of what its benefits would be. If I had a dollar for every good hypothesis that suggested that a particular mechanism of action would lead to a health outcomes benefit that turned out not to be true, I would be a very wealthy man. We've seen it over and over again. The fact that a drug lowers blood sugar does not necessarily mean that it lowers the risk of the complications of diabetes. And, in fact, we've gone down this pathway now for decades, and many people believe that using blood sugar reduction as a surrogate outcome for approval of these drugs is a very questionable approach. I was actually pleased to see recently that Cliff Rosen, who chaired the FDA advisory panel, took a very strong public statement on this issue in a perspective he wrote in The New England Journal of Medicine.[2]

Dr. Califf: I've been involved, as you know, particularly in the last 4 or 5 years, in a lot of collaborative clinical trials with diabetes specialists, as you have also. So let me just take you down the argument that I've been through multiple times about this and just see where you come down on it. I know I'm asking questions; I'm glad to say where I come down also. When you talk about health benefits, Steve, you would agree that lowering blood sugar is a health benefit because completely untreated diabetes leads to ketoacidosis, and as interns, particularly in city hospitals, we would see this outcome. It's a pretty tragic outcome. So lowering blood sugar does have health benefits in people with diabetes, right?

Dr. Nissen: Well, first of all, ketoacidosis, as you know, is rare in type 2 diabetes, which is what we're talking about here. But certainly hyperosmolar coma and other pretty serious complications can occur due to extreme levels of hyperglycemia -- no question about that. So there are certainly benefits to lowering blood sugar. The question that has always remained is: "What about the complications of diabetes?" "What about microvascular and macrovascular disease?" And that's the question that comes up in the rosiglitazone debate: "Is there proven evidence that lowering blood sugar reduces macrovascular complications, and is there proven evidence that it lowers microvascular complications?"

I've reviewed the literature also, and I think the evidence is reasonably good that lowering blood sugar reduces microvascular complications. But the evidence is not at all clear for macrovascular complications.

Dr. Califf: Well, I guess where I come down on this is I do believe that lowering blood sugar, in and of itself, has a health benefit relative to nothing. But the fact is we have a lot of different ways of lowering blood sugar. So from my perspective, that's relatively a moot point at this juncture.

Dr. Nissen: I think you have to look at this in the totality. Lowering blood sugar is a good thing, and you can always do that with insulin. That's always been something that's in the armamentarium for type 1 and type 2 diabetes. But you would never consider a drug to be a benefit if it lowered blood sugar but, in the case of Rezulin [troglitazone], caused liver failure, or in the case of phenformin, caused fatal lactic acidosis. It's about what is the downside to lowering blood sugar. Are there side effects? Are there other adverse effects to the drugs?

It reminds me a lot of the COX-2 [cyclooxygenase-2] debate. COX-2 inhibitors were very good at relieving pain. That's a health benefit. But they increased the risk of myocardial infarction, and that was a hazard.

And so my view of drug development is: It's always about balancing benefits and risks. What happened here is a drug without any really proven benefits got on the market, stayed on the market for 8 years without any definitive trial looking for major benefits on macrovascular or microvascular complications, and we learned very late in its development that it increased the risk of myocardial ischemia.

Dr. Califf: You're not saying this is unusual, are you? Aren't there a lot of drugs on the market that haven't had definitive cardiovascular outcomes measured? I hope you're also not saying that a diabetes treatment drug would have to reduce macrovascular complications. Wouldn't a neutral effect be just fine?

Dr. Nissen: I would be fine if you knew that a drug was, at worst, neutral. You like, however, to believe that drugs for diabetes would reduce the likelihood of the most serious complication of diabetes. And let's be very clear about this. Between two thirds and three fourths of diabetics die of cardiovascular disease. One of the goals of diabetes treatment should be to reduce that disease that's likely to kill diabetics. Unfortunately, we haven't had very compelling evidence for any of the classes of drugs, and, in fact, there's been some evidence to the contrary. And so we've got this paradox here of having a lot of drugs that lower blood sugar. The best we can say about most of them is they're neutral, although there are certainly some data to suggest that metformin is protective. Then along comes a drug that is increasing risk by 30% or 40%. And I have to point out that the magnitude of the estimated effect of rosiglitazone is not small.

In the PROVE-IT [Pravastatin or Atorvastatin Evaluation and Infection Therapy] trial, we adopted intensive statin therapy because it lowered morbidity and mortality in an ACS [acute coronary syndromes] trial by 16%, an intensive statin regimen vs a moderate statin regimen. Here we have an effect that is probably in the range of 30% to 40%. That is really a very serious public health concern.

Dr. Califf: So just to say where I think I am on this issue, and it is, I think, a little bit complicated, like most of these things. I think lowering blood sugar is a good thing, everything else being equal. I think it's pretty clear that lowering blood sugar chronically reduces microvascular complications, complications really that are blood sugar-related directly like neuropathy, nephropathy, retinopathy, etc. But one can engage in a lengthy debate about that, and I don't think we need to do it today because most of those trials have been based on intermediate measures and not hard clinical outcomes, like actually becoming neuropathic or losing renal function completely. I think a neutral effect on cardiovascular outcomes would be plenty to say that a drug should be on the market and would be beneficial, because we have other ways of lowering mortality and reducing vascular events in diabetes.

Dr. Nissen: I would absolutely agree with that. But let me ask you a rhetorical question, and it's a tough one. If you were sitting there in 1999, and if you had been a member of that advisory panel, and somebody presented you data from 2500 patients in a series of reasonably well-performed, randomized, controlled trials, and the relative risk of myocardial infarction or myocardial ischemia was 1.8, compared to traditional antidiabetic therapies, would you ask for more data or would you vote to approve the drug?

I went back and looked at rosiglitazone, and I must tell you, I would have almost certainly, based upon what I saw in that packet from 1999, said it's not approvable until they demonstrate better evidence of cardiovascular safety.

Dr. Califf: You're saying that putting yourself retroactively into a 1999 frame of mind, in other words, the state of knowledge at that time and the way people were thinking. . . Let me just say, I agree with you, and I think I was arguing at that time. But I also want to clarify one thing you said. You said relative risk of 1.8. At that point in time that was not what we would call statistically significant. Is that correct?

Dr. Nissen: It was not statistically significant, and it can't be. There were not enough events. You have to understand how these drugs are developed. The studies that are used to prove new diabetic agents deliberately avoid the sickest patients. They avoid patients with cardiovascular disease. And so when I went back and looked at it, and my recollection is there were something like 36 MIs [myocardial infarctions] in 1 group and 9 in the other, but there were different denominators. So when you got done with all of it, you had a relative risk of about 1.8. But the signal was going in the wrong direction. I will point out to you that that came up for me a year or 2 later on an FDA advisory panel for yet another drug, in this case nesiritide, where I did vote against approval because the compass needles were pointing in the wrong direction, and I thought that, just as I think for diabetes drugs, drugs to treat heart failure, you want to make sure that they're at the very least neutral on things like morbidity and mortality. If they relieve symptoms, that's great, but you don't want to push the hard endpoints in the wrong direction.

Dr. Califf: I think our point of agreement here is, the criteria for saying that a drug is effective need to be more stringent than saying that there's a risk that should merit more study. In other words, a nonsignificant trend in the wrong direction of that magnitude early on, if it were known, I would agree would be a reason to say, "hold on; let's get the definitive evidence." It may be bad luck; it may be within the play of chance, but why take that risk on the safety side?

Now, on the effectiveness side, you wouldn't argue that with a nonsignificant trend in a good direction that you would declare victory at that point.

Dr. Nissen: Absolutely. But let me tell you that there was something unusual going on, and we have to talk about it. Here was the problem. Rezulin, or troglitazone, had been on the market for several years. It has been withdrawn from the market in virtually every other country except for the United States, but yet another FDA advisory panel had recommended keeping Rezulin on the market, and the FDA chose not to act. When rosiglitazone and pioglitazone were approved, the FDA felt, well, there's alternatives in the class, and they pulled troglitazone. Many of us believe that there was a lot of pressure by the agency and by the clinical community to approve rosiglitazone because people wanted an alternative to troglitazone, which had by then been clearly associated with hepatic failure. This shows how hazardous drug development is. You trade a 1 in 40,000 or 1 in 20,000 risk of liver failure for a drug of unknown risks, and it turns out those unknown risks are far more serious, far more grave, than the risks of the drug that it was designed to replace.

Dr. Califf: So we're going to ultimately both say that there is no shortcut. The only way to solve the issue for any drug is to study it in a relevant population for a relevant period of time. But before we do that I just want to make sure that I understand your view on this issue that's been controversial in the literature in the last 2 months, which is, while I agree with you that a trend in the wrong direction for safety, particularly one that's persistent, should be reason to put a pause on things and try to get the definitive data, are you saying that it's proven that the risk is there, that it's significant, and outside of the play of chance? Because other people, most notably Diamond and Sanjay Kaul,[3] the usual naysayers in the literature, have an article that says when they reanalyzed the data using Bayesian techniques they don't get the same answer that you did. The trend still goes in the same direction, so I'm not negating the point that there's an issue.

Dr. Nissen: Let me take that one on directly. I have a belief about how one should do a meta-analysis, and I want to point out to you what we really know here. First of all, the first point is that when we do a meta-analysis, the statistician and I sit down in advance, and we decide what the criteria should be for inclusions of studies. We said 6 months or longer duration, for very good reasons. We wanted randomized controlled trials. We had a series of criteria, and we prespecified the method of analysis. Anybody that wants can come along later on and try 5 other methods of analyzing the data until they find one that shows what they want. That wasn't what we did. We prespecified the Peto odds ratios, and interestingly enough, I have a note from Richard Peto that gets actually a little bit worse hazard than we do. He used a little different way of doing meta-groups.

The bottom line is, it was there; it's been there very clearly, and we know we were right because when a patient-level meta-analysis was done by the FDA based upon the more robust time-to-event data that we really wanted but weren't able to get from GlaxoSmithKline, they got exactly the same risk. We got 1.43, and the FDA reports 1.4 using the patient-level data. So I see the Kaul and Diamond analysis as data mining in the worst way, and not contributory. The company's patient-level data show a statistically significant effect. The FDA's patient-level data show a statistically significant effect, and our study-level data show a statistically significant effect, and they all show a 30% to 40% increase in risk. Let me tell you, this drug is in fact capable of increasing myocardial ischemic events substantially in diabetic patients.

Dr. Califf: But, Steve, you're so brimming with confidence. You know me. I'm a less confident kind of a guy to begin with, I guess. Because I don't know. I mean, a lot of other people besides Diamond and Kaul have looked at this. There are a couple of NIH [National Institutes of Health] data-monitoring committees that have looked at all of the available data and have ongoing trials, the results of which, for obvious reasons, we're not privy to.

Dr. Nissen: Well, again, if you read carefully, Dr. Nabel's letter to the editor to The Lancet,[4] she points out that each of those committees has made a decision that they will decide on continuing the trial based upon the internal events within that trial specifically. And I hope everyone realizes that none of those trials are even remotely powered to answer the question. The one trial that is closer to being powered to answer the question, the RECORD trial [Rosiglitazone Evaluated for Cardiac Outcomes and Regulation of Glycemia in Diabetes], shows a relative risk of 1.23 for myocardial infarction midway through the trial. So again the compass needle is pointing in the wrong direction yet again. And if you add the RECORD trial to our meta-analysis, you get even a stronger signal, statistically, for harm.

Dr. Califf: Steve, you just raised 2 really important issues for those of us interested in clinical trials in general. One is, and this I think does deserve some clarification from the NIH, you're saying that based on Dr. Nabel's letter, that these data-monitoring committees are not considering the external data in their decision making, which I, personally, my view of data-monitoring committees is, if that's the case, that would be a major error.

Dr. Nissen: Well, you know, again, I think we'd have to look at each of the committees. They have, of course, their own governance process. But my understanding is that they are looking at the trial as it is, as the data are evolving. But it's more profound than that, Rob. Because if you look at the power calculations, let me point out that the FDA safety officers, when they presented to the advisory committee meeting on the 30th of July, calculated the power to detect a 1.4-fold increased risk of myocardial infarction, in the largest ongoing trial, which is RECORD, that the power is less than 5%. And so anybody that wants to be reassured because there hasn't been the stopping of those ongoing trials is deluding themselves about what those trials are really capable of showing.

Dr. Califf: Well, Steve, from my view you've launched sort of a major Scud missile here on this issue because I would -- and I do think this deserves following up because I wouldn't presume for a millisecond that it would be wise for a data-monitoring committee to ignore external data. I think your point about the lack of power within each of those studies is very well taken. But if a committee was wisely considering all data that it had access to, including external data, I'm sure there must have been some interesting discussions.

I can't resist also asking you, since I haven't had a chance to do this before, what do you think about what the RECORD study did of publishing their interim results?[5] That's been very controversial in the clinical research world. Personally I think it's a very bad idea and a bad precedent to do that.

Dr. Nissen: I thought it was a terrible precedent. And I thought it was done to support the company's contention that the drug didn't have a risk. So it's all the wrong reasons to do it. Now there was a justification given in the manuscript that they thought it was better to do that than to have patients in the trial drop out of the study because of the increase in risk. But, frankly, I think you can make a case for stopping RECORD on the basis of futility. In fact, using conditional power for the primary endpoint, the study is way under 50% powered for the primary endpoint, which is death plus hospitalization, let alone the harder clinical endpoints that we really care about. And so, in fact, it's an exercise in futility, and I think that argument was made by the FDA reviewers in the documents presented to the advisory committee.

Dr. Califf: I'm muddled and more concerned about all the uncertainties, except we definitely agree there's a signal here. It's a cause of great concern. Your degree of certainty that the signal is real and of a certain size is greater than mine. You've raised a couple of other issues here that -- we have a short period of time, we could probably talk for many hours -- but to really get to the core of what's going on here in a more general sense, the clinical world needs to be very involved with and concerned about, you said several times that GSK [GlaxoSmithKline] had data; they had concerns. In one way or another it was sort of made public but not completely. The GSK folks, and I think this is correct, say that they did everything within the rules of the game, in other words, that they fulfilled all the commitments they made to the FDA and EMEA [European Agency for the Evaluation of Medicinal Products]. They worked with the regulatory agencies all the way along. And of course the rules at the time, which you and I both vehemently disagree with, were that you didn't have to make all your human experiments public. That, thank goodness, has changed now, so we won't have to deal with that problem in the future, I don't think. What are you really saying here about the relationships with the FDA, the thought leaders, and industry?

Dr. Nissen: I'm very troubled, Rob, by the fact that when a drug company analyzes essentially every clinical trial they've conducted with a drug and concludes that there is a 31% higher risk of myocardial ischemia, and tells the FDA that, first in August of '05 and again in late '06, and nothing happens, something went terribly, terribly wrong. Now I must point out that there are a lot of issues about whether everything right was done. And I hope everyone is aware of the fact that what the company did after they saw this very clear signal for excess cardiovascular events, is they commissioned an observational study, which was relatively recently published with UnitedHealthcare,[6] that they said showed no evidence of increased risk. But there was something wrong with the study. And what was wrong with the study is that they omitted all the patients that had been receiving pioglitazone. So they took out of the analysis the comparison of rosiglitazone and pioglitazone. That missing data was published last week[7] and shows a statistically significant difference in cardiovascular outcomes between pioglitazone and rosiglitazone, and that was omitted from the publication and from the submission to the FDA, and is not there on the GSK Web site. So I don't think you can argue that we had all of the information.

Dr. Califf: So a naive person would say it's pretty straightforward. You've got a drug that's treating an epidemic disease. You've got a company with plenty of money. You have very knowledgeable thought leaders in diabetes. Why didn't they just do the right studies from the beginning? Why has this gone so wrong?

Dr. Nissen: Well, it's a terrible problem. I think if you read the approval letter that says for approval the requisite is that they do a study to look at safety of rosiglitazone, and the study is even named, it's called the ADOPT study. And when the study was actually finally executed, they didn't adjudicate cardiovascular events in the study; they didn't power it for those events, even though there was a signal in the registration package. And the primary endpoint became glycemic control durability. So the postmarketing study that was agreed to turned out to be a completely different study,[8] and here we are 8 years later.

The one thing I will agree with you about, we might disagree about the strength of the evidence, but the better evidence would have been to have a large prospective, well-powered, randomized study with adjudicated cardiovascular endpoints to end the debate. We're not going to have that study; we're not going to have that study any time during the life cycle of this drug because the RECORD study, mandated by the European regulatory agencies, is also underpowered. So now we have to make decisions as clinicians based upon conditions of great uncertainty, and that's unfortunate.

Dr. Califf: So, Steve, why do you think so many diabetologists don't think the way you do? I've had a number of them tell me that Dr. Nissen thinks that every drug has to have a mega-trial and this will bankrupt the clinical development world, and there will be no new diabetes drugs. You'd agree with me that there's a lot of undertone of that you're just cardiovascular-centric. I, by the way, agree with you about what needs to be done. I actually think there should be a major study, but you've been embroiled in this a bit, and been in numerous venues. This is mostly a cardiologist audience on Are we just enthralled with cardiology so we're disregarding the needs of people with diabetes? What's going on?

Dr. Nissen: Absolutely not. Look, as I said several times, you've got a disorder, a cardiovascular disease that is far and away the leading cause of death in diabetes. You've got signals early on that a drug is increasing the risks of ischemic cardiovascular disease. Do we need a mega-trial for every drug? Well, we could argue that point. I happen to think that proving a health outcomes benefit for a drug is a very important mission. And when a drug is selling $3.4 billion a year in sales, I don't think it's too much to ask that some of those profits be used to develop an evidence base for the benefits, or lack of them, of the drug. And in this case, it should have been done.

Now it's interesting to know that the new DPP [dipeptidyl peptidase]-IV inhibitors, I understand, the companies are actually looking at doing dedicated cardiovascular outcomes trials. And hopefully one of the things we've done with this publication is we've raised the bar. And we ought to set the bar high enough because we've got to treat patients. I, to this day, don't always know which of the diabetes drugs to rely upon, particularly in patients with high cardiovascular risk. I'm not going to know that answer unless somebody does the studies.

Dr. Califf: Do you think the NIH is doing its part to push this along? Is this something that should be totally on the docket of the drug companies? Aren't these studies all biased in one way or another? Shouldn't the NIH be devoting more of its energy to this?

Dr. Nissen: You're preaching to the choir. I think that the NIH needs to do more clinical outcomes studies, and we need more independent trials. Short of that, we need to do something that you and I have spent much of our lives doing, which is running independent trials run through independent academic coordinating centers where the database is made available to investigators and where there is an independent contractual right to publish the results. But one of the things about the database that we used to evaluate rosiglitazone is only a handful of the studies were ever published. The vast majority of the data that we used for our meta-analysis we only had available to us because the company was forced to disclose it. If it were any other drug, we wouldn't have been able to even do the analysis, and that's really a shame. That's because so many of the studies are being done by the company with a CRO [contract research organization] without an independent steering committee, without an independent academic coordinating center, and that's got to change, even if the NIH doesn't step up to the plate.

Dr. Califf: We're definitely in agreement there. Let's turn to the public policy, which this is getting right into, and I think we can do this quickly and call it a day. It's been a fascinating discussion. You've explained a lot of things that I had wondered about in your thinking because we hadn't had a chance to talk about this.

The last issue I want to talk about is the FDA and what you think should be done about that, because I've spent a lot of time with the FDA for many years and have great admiration for the mission of the FDA. But before we get to that, I think a lot of cardiologists probably don't know that there's a major FDA reauthorization, which has things in it that are not directly FDA-related, and some other bills in Congress that would affect these issues about transparency of human experimentation.

You've been, as I understand it, directly involved with the bills. What do you think is going to happen with those bills in terms of requirements for registering and publishing clinical trials?

Dr. Nissen: Well, first of all, the bills are different, and I worked on both the Senate and the House side with staffers of both political parties on the bills. Senators Enzi [Mike Enzi, Republican, Wyoming] and Kennedy [Edward Kennedy, Democrat, Massachusetts] on the Senate side and Waxman [Henry Waxman, Democrat, California] and Markey [Edward Markey, Democrat, Massachusetts] on the House side. The House bill differs from the Senate bill in that it currently requires mandatory disclosure of the results of clinical trials. Now I will tell you I was an ardent advocate for that during the process of developing the bill, and worked with staffers on the House side to help make sure that provision stayed in the bill. It's not in the Senate version. It's under the gun now whether, in fact, it's going to stay in when the Conference Committee gets done. I hope some of you heard the interview with Representative Ed Markey on The New England Journal of Medicine Web site,[9] where he articulated, I think, very well. Let me tell you what the principle is from my point of view.

I believe that if you ask human subjects to participate in an experiment that the results of that experiment belong to the public. That it is not acceptable to do such an experiment and then to deep-six the results. And I want that to be a matter of law. I think if it's not a matter of law, we will continue to see negative publication bias where the studies that get published are the ones that show favorable effects of drugs. Things have gotten out of balance, Rob. We hear about the efficacy of drugs through media, on television, through advertising, detail people, CME, but we need to see all of the data in order for us to make good clinical judgments. And I'm fearful that it may not stay in the final bill that gets passed by the Conference Committee.

Dr. Califf: I hope it does. The way I've approached this, as I think you've heard before, is in multiple venues I've asked the question: "Will everyone in the room who's in favor of secret human experimentation please raise your hands?" And so far I haven't really had anyone -- well, maybe one exception, but I don't want to reveal who that was -- say they're in favor of that. When you couch it in terms that don't really say what's going on, like "should all clinical trials be published?" people have views. But if you really put it in front of people what the issue is, that is, when we go through an IRB [Institutional Review Board] and we get permission to do a human experiment, by law we have to certify that we understand the definition of research, which is something that's done to create generalizable knowledge. And generalizable knowledge to me can't be present if the result is kept from the public if it's negative. So we're completely together on that, and I hope something is done to make it a requirement. Because I think then companies will do what they're required to do, and it will change.

Dr. Nissen: Well, they argue that the data is proprietary. And they have gotten a lot of traction in the Congress. The companies have said through their lobbyists and so on that when they spend the money to run a clinical trial that they own those results and they should be free, in a free enterprise system, to do whatever they want with those results. And if they don't like them and they want to put them somewhere other than in the public domain, that should be their right.

Dr. Califf: Well, you'd agree with me though that while there are lobbyists arguing that, there are many good people in companies who feel quite differently and would like to have permission, as you say, with regard to their corporate boards, who are always looking after optimal profitability, they'd like to have that happen. So the way I've thought about that is, if you're on multiple corporate boards, let's say you're on a pharmaceutical board and a shoe company board, to my knowledge, if Nike did a study of Nike shoes and it looked bad, they have no obligation to publish that their shoes didn't look good. Pharmaceutical companies obviously have a different mandate because the experiments are not done on shoes; they're done on human beings. But the way the rules are right now, the fiduciary responsibility of the corporate board is to not do things that damage the profitability of the company, and that leads, I think, to a very distorted perspective on the ethics that are involved.

Dr. Nissen: Let me agree with you and say that I work with some wonderful people in industry, who absolutely get it and have a public health mission just as we do. But, unfortunately, there have been too many examples where data that we needed to see in order to make good decisions for patients was buried. That's why, in fact, Eliot Spitzer sued GlaxoSmithKline. It was over the risk of suicide in pediatric use of SSRI [selective serotonin reuptake inhibitor] antidepressants. And that's why we got a consent decree that allowed us to look into the rosiglitazone situation. We've got to have that kind of transparency. I'm very, very concerned about whether this will stay in the bill, and I would urge everybody to talk to their congressman/congresswoman and really try to get people on board on this one.

Dr. Califf: I'm really with you there, and my belief is, although in the past the argument has been made about trade secrets and so on and so forth, but when you get into the human experiment arena, if we level the playing field so all companies are working under the same rules, then the best products will win out because the public will have better information and people will make better decisions about what to do.

So, Steve, in the last minute or two, let's turn to what should happen with the FDA. I think you and I both love the mission of the FDA. I think it's the final line of defense for the American public. In work that I've been privileged to be doing over the last year or two with a subcommittee of the science board of the FDA, we've been looking at the internal structures. I think a lot of people don't realize that the food supply, although it's become a little bit more obvious now with some of the problems that have recently occurred with China, has the same issues as pharmaceuticals and devices. That is, the FDA is under tremendous pressure; it's understaffed; it's underfunded. Congress keeps adding new mandates without providing the funding to do what needs to be done. But listening to what you said, it seems like you think there's a malaise within the FDA, which is even above and beyond the understaffing.

Dr. Nissen: The FDA is clearly understaffed and they're underpaid. And I know many fine, dedicated public servants within the FDA. But I think the FDA has lost its way. And I'm not sure when things got offtrack. I believe that part of what took the FDA offtrack was the Prescription Drug User Fee Act, or PDUFA, in which industry fees are used as a principal source of funding for the regulation of drugs. But somehow or other we found ourselves in a situation where the FDA has repeatedly failed to pursue these signals until it became very obvious and the consequences were very grave. What you saw at the Avandia advisory board hearing, for the first time in history, was the internal dispute that went on in the FDA exploded into the public domain. You saw the Office of New Drugs working very hard to convince the committee not to act on the rosiglitazone case. And you had, for the first time, Gerald Del Pan, the head of the Office of Surveillance and Epidemiology, and David Graham publicly call for the removal of a drug from the market. That has been played out now over many, many years within the FDA. There is an imbalance of power within the agency where most of the power resides in the Office of New Drugs, and it's very hard to get regulators, who approved a drug, to look back and say, "Oh my God, we let the genie out of the bottle here, now let's put it back in." And so, I, for the life of me, cannot understand why, 2 years after a company informed the agency of a risk as serious as myocardial infarction in diabetic patients for a very commonly used drug, nothing happened until we published an article in The New England Journal of Medicine. Something went terribly wrong. It's not the first time, and we better figure out how to fix it because there are a lot of lives on the line.

Dr. Califf: Well, we definitely need to fix the problems in the FDA, and I'm a little stuck here too, Steve, I've got to say, because looking across multiple divisions in the FDA, as I've had the chance to do because of my job at the DCRI [Duke Clinical Research Institute], which involves more noncardiovascular than cardiovascular trials now, it looks sort of like the FDA typically gets aligned with the thought leaders in the discipline that they're working in. And you would agree with me that our stance from a cardiovascular perspective is a little different than most other medical specialties about the need for broad outcomes assessment. Although I think the world is coming along to our side.

Dr. Nissen: Oh, I agree. Cardiovascular medicine has been way ahead of the curve on that, and I think there's all kinds of competing issues going on. I think you know about some of the issues that have occurred in the division that handles antibiotics, where severely underpowered, noninferiority trials were used to give label claims, and where the drugs turned out, well, in this case, Ketek [telithromycin] to have serious problems. And so it's a very complicated situation, and the rigor differs greatly across different divisions of the agency. There isn't consistency. Some people have advocated, Senators Dodd [Christopher Dodd, Democrat, Connecticut] and Grassley [Chuck Grassley, Republican, Iowa] have a bill that would create a separate center for postmarketing evaluation and research, and to give more authority to this postmarketing surveillance center. And I've come around to believe that they were right in this vision. That unless we create balance between those people who primarily are concerned with bringing new drugs to market and those people that are primarily concerned with safety from a public health point of view, we've got to create a proper balance between those 2 missions. And if we don't create that balance, we're going to have move Vioxxs, more Avandias, more Keteks, and the like.

Dr. Califf: Well, you and I disagree on that one, but I understand why you'd think that way because we agree that we need the balance. But to me that's trying to patch a problem that's more systemic and needs to be fixed. If we don't take this agency, which is regulating over 25% of our economy, has responsibility for our food supply, our agricultural products, and the drugs and devices that we use, and start putting serious money into it, and demand that we have a nonpolitical scientific base for what it does, we are going to see more catastrophes along the lines of what we saw in the past with thalidomide and even going back to the origins of the FDA, "the horse named Jim," whose illness at the time that tetanus toxin was being made led to the deaths of some children. They really got the FDA started. I think we're going to see some pretty major catastrophes if we don't repair the problem.

So I can tolerate easily differences of opinion about exactly how to do it, but fundamentally this is not something that can be fixed with a patch; it really is systemic, in my view.

Dr. Nissen: Yes, and it's also about leadership. I think we need to get very strong leadership at the top, not make this quite such a political agency, but more of a public health agency, which I think we've gotten away from. I think the FDA lost tremendous credibility, internally and externally, when, for example, it ignored 2 of its own advisory panels that recommended approval of Plan B [emergency contraceptive] because the White House didn't like the idea. And you can't run an agency that's protecting public health in that kind of politicized way.

Dr. Califf: So, Steve, last question/prediction. Avandia, by vote of the committee, is still on the market. The NIH studies are ongoing. It has a stronger label. What do you think is going to happen in the next few years?

Dr. Nissen: We're not going to get an answer that's definitive. There's going to be a series of observational studies, as there were for the COX inhibitors, and some will point to a risk, some will point to a lack of a risk. Being observational studies, all the confounders will make it very hard to interpret the information. The ongoing studies are not adequately powered to answer the question. I think the drug limps along; it gets used much less commonly than it is used previously. People will tend to use other agents, including pioglitazone, which very interestingly does not appear to share the same risk, and we'll not get a definitive answer to the question. I would be very surprised if we got a definitive answer. It would take a relative risk on the order of 1.6-2.0 for there to actually be adequate power in any of the trials to answer the question.

So what do we have? What we have is the totality of data to date. And in 8 years on the market, if you take all the clinical trials and combine them, as we did, as the FDA did, and as the company did, you see a 34% to 40% increase in the risk of ischemic events. Physicians will have to look at that and decide, under those circumstances, do they or do they not believe that the benefits of the drug outweigh the risks. But I do not think we will see regulatory action. If what the agency wants is a definitive, large-scale, multicenter trial, they're not going to get it.

Let me make one more point that has come up several times. Can you use a meta-analysis to decide that a drug should not be marketed? That's what the FDA did recently with Zelnorm [tegaserod]. There was a meta-analysis that showed an increased risk for myocardial infarction that resulted in the drug being removed from the market. Often it's all you've got, and then we better decide whether it's enough or not. If it's not enough, then we better do large-scale trials for every drug to answer these questions because otherwise that's all we're going to have is meta-analysis.

Dr. Califf: Well, my prediction, Steve, and this is unusual for me to be so optimistic, but I think the community is going to get its act together, it's going to demand that some schema be put together by which, prospectively, the previous data and the new data, including, hopefully, some new trials that are focused on high-risk patients, will get done and will answer the question.

Dr. Nissen: Well, Rob, let me give you the downside. The drug's exclusivity period is going to end in just a few years, so there's an issue of funding, and to accumulate enough events in an outcomes trial, as you know, you've been the master of these trials for many decades, it takes years and it take thousands of patients. And I don't think that it's likely that the company, late in the life cycle, in a drug, which is already shrunken in market share by at least half, is going to have the wherewithal to do it. And so unless the NIH does it, I just don't see it happening. I think the drug limps along and goes generic and becomes one of those drugs that we never get the definitive answer about.

Dr. Califf: Well, my final retort, and I'll let you have the last word on that, is that I'm involved in several now. One of them, actually, you brought up, nesiritide, where there is no financial justification for doing the outcome trial, but there's an ethical obligation to the company to resolve the uncertainty. So if GSK is saying they don't think there's a risk here which is significant enough to tell patients not to take the drug, I think it's hard to argue that GSK doesn't have the financial means to resolve the uncertainty.

Dr. Nissen: Well, that's another question that I want to just have the final word on. Some people have said, "Okay, these meta-analyses, they're just not enough to pull a drug off the market." There's 2 questions I want to leave you with. Given what we now know about rosiglitazone, would this drug be approvable? Would anybody listening, watching, approve this drug as a new entity with all the signals that we see for increased cardiovascular harm? And if the answer is "no," then that has certain implications. I just think that the other question to be asked is, "If we don't have adequate data to answer the question, whose fault is that?" Is it the fault of those of us that did the meta-analysis? A lot of people would like to say, "Gosh, you created all this turmoil, Nissen, when you published this meta-analysis." Whose obligation was it to do ADOPT in a way that would answer the question? It was the company's obligation. And so it's very hard to criticize an analysis that takes the totality of what we know about the drug and puts it out in the public domain when the company has never done a trial that would actually answer the question. And that's where the problem is. Eight years into the use of this drug we don't have a definitive trial.

Dr. Califf: Well, Steve, I guess you don't buy my typical argument at this point, which is that the real problem is with the medical professional community that we're talking to today because companies know that their main customer is the doctor who writes the prescription, not the patient. The patient is the ultimate customer, and as you said, many people within companies are really deeply steeped in the public health mission of improving health through pharmacological therapy, which no one would argue against. But in the end, if the clinical professional community demanded to have the evidence, the companies would produce it, wouldn't they?

Dr. Nissen: They would. But we're asking physicians to behave very differently. Let's be clear. We are bombarded by messages about the efficacy of drugs.

I will tell you a quick little story. When rosiglitazone first came out, I was very struck by the 18.6% increase in LDL [low-density lipoprotein] cholesterol reported in the label. So I went by the booth, the company's display at one of our national meetings, and I said, "I'm worried about this. Raising LDL cholesterol nearly 20% in diabetics makes me worry." And the representative, who had been dutifully coached, looked at me and said, "Oh, but it's that large buoyant LDL. It's good LDL; it's not bad LDL." You and I might see through that argument. We're asking a very high standard to ask our colleagues to look at that and say, "That's nonsense. Raising LDL is not a good thing; I should worry about this for this drug."

Dr. Califf: Well, Steve, I want to thank you for engaging in the discussion. It's been enlightening to me to hear the way you're thinking about it. Hopefully you've gotten a little bit out of my ideas, and I look forward to seeing what happens now.

Dr. Nissen: Well, I did enjoy it as well. And I want to say to everybody, I know I expressed a great deal of passion about this; there's a reason. And the reason is, if I'm right about this, and I believe that I am, this represents one of the most serious public health reversals we've seen in a very long time.

I would rather be wrong, but I think the signals here are all pointing in one direction: that the drug has this hazard and obviously it's a very, very serious problem.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.