Pharmageddon (38 page)

Medicine today has to be based on evidence—but what is evidence? Industry has masterfully exploited ambiguities in the word and, especially, a gap that lies between data—what actually happens to a patient—and the later construction of “evidence” as to what a drug does. Where doctors are faced with the word “evidence” they for the most part believe they are dealing with data-based medicine, and as long as they think this way industry can vigorously promote their version of what happened when drugs are given under the banner of “evidence”- based medicine.

There are several ways in which the data that should be at the heart of medical evidence are hidden by the pharmaceutical industry. First, there is a drug company's outright hiding of the data that is collected so that no one gets to see it. Second, the remaining data is disguised through certain kinds of statistical models and what comes out the far end is presented as “evidence,” while the injuries that happen to you and me are degraded to the status of an anecdote. And finally companies have created a culture of data neglect, in which doctors have become blind to what is happening to the person in front of their eyes.

The issue of hiding data is not the scandal it should be in part because common perception is that drugs are made in company laboratories. This view misses something fundamental: companies make chemicals, but we are the laboratories in which modern drugs are made. Drugs are chemicals used for a social purpose—to treat conditions that we define as diseases. Drugs cannot come into being unless we as healthy volunteers and later as patients in clinical trials agree to take them to see what happens. Without our participation, there is no drug.

Our willingness to participate in these studies was borne out of a sense of civic duty in the 1950s. We participated on the understanding that taking risks might injure us but would benefit a community that included our friends, relatives, and children. We did so for free— in perhaps the greatest ever example of how a system geared around people rather than products can make much more economic sense. The system worked and extended the compass of human freedom from the many epidemics and other scourges to which our forefathers had been subject for millennia.

But the research in which we once participated has morphed from scientific studies whose data was in the public domain into company trials where the data has been sequestered. There are all sorts of protections now built into trials that weren't there in the 1950s—centering on informed consent about the possible benefits and side effects of treatment and suitable safeguards of anonymity. But we are never informed about and asked to consent to the sequestration of our data in company trials. We assume we are participating in science and that the data arising from the risks we take are available to scientists more generally, as they once were.

It would be an easy matter to remedy this—by ensuring that the consent forms for a trial tell us whether the company will sequester our data. Without this any participation in such studies is more likely to jeopardize the health and well-being of our friends, children, and communities than to widen the compass of freedom from disease because companies will parade clinical trials where injurious side effects are hidden, artfully coded, or simply eliminated as evidence that their drug could not have caused the injury, and if doctors or the courts believe them, any effort on our part to seek redress will fail at the first hurdle.

Evidence-based medicine as first conceived was a highly moral enterprise. It takes courage to subject all our preconceptions to testing, and then to treat the people who come to us for care on the basis of the data. But what passes for evidence-based medicine has now been subverted by companies into an exercise that skirts conflicting values by appealing to preselected “data” that are supposedly value neutral. We need to recover the perspective that science, far from being value free, values data and does so with a passion. “Controlled trials” that involve restricted access to data are not science, and following selected “data” can only diminish both the carers and the cared-for.

If all the data on the benefits and hazards of treatment were available, the exuberance that companies can engender by marketing agents of supposedly extraordinary efficacy and almost no risks would to some extent be curbed, and if curbed the vast profits that support the most sophisticated marketing on the planet, that conjures diseases out of vicissitudes and can reconfigure our very selves to suit its purposes, would in some measure be limited. If this happened clinical practice would have a better chance of returning to something closer to what it once was, when one person consulted another.

We might require companies to post all records from clinical trials, with suitable safeguards to protect individual patient identities, to an Internet site, before any analysis of the evidence was published. This would simply replicate for medicine the common scientific practice of journals like
Nature
or
Science
that require the relevant genetic sequence to be posted to the Internet before any claims are published.

But the simple idea of posting the data reveals a profound ambiguity in the notion of evidence-based medicine. While most people assume medicine has been data-based for the last twenty years or so, there are vanishingly few articles in journals or presentations at meetings—other than, ironically, some good case reports—that offer data. In the course of clinical trials things happen to patients, but these things are assembled and analyzed using models that come with assumptions including that if a finding is not statistically significant it essentially does not exist. It is the “results” of these analyses that are published—rather than any account of what in fact happened to patients. When it comes to results, he who controls the program giving rise to the results can control everything.
³

When we have Pierre Touery drinking ten times the lethal dose of strychnine in front of our eyes and surviving because he had previously taken activated charcoal or when a child with bacterial endocarditis gets up off her deathbed and walks after being given penicillin, we have data-based medicine. When you are put on a statin, a biphosphonate, or a psychotropic drug because in some trial that has recruited hundreds or even thousands of subjects so that a finding on a blood test or rating scale that suits some company has become statistically significant, you are being subjected to results-based medicine rather than cared for with data-based medicine.

When it comes to the hazards of drugs as reported by their makers we are invariably dealing with results rather than data. Consider the following excerpt from the deposition of Ian Hudson, then head of Global Safety at GSK:

Q:
Okay. So, your view is: It's simply impossible for SmithKline Beecham to decide whether Paxil did or did not contribute to the homicidal or suicidal behavior of any one given individual; is that your testimony?

A:
We would certainly gather all the information, but on an individual case basis it would be impossible to decide whether paroxetine caused an event or not…. It is impossible, on an individual case basis, from individual reports, to assign causality especially in a very complicated area such as this. That's why, when we have issues, we review all the available data and make a determination, on the basis of all the available data, whether there is an issue or not.

Q:
Okay. Do you believe that it is possible that Paxil has caused any person, worldwide, to commit an act of homicide or suicide?

A:
I have seen no evidence to suggest that at all.
⁴

Hudson is demonstrating the standard company approach to determining whether there is any link between an adverse effect and the company's drug. In brief, the company position is that unless an adverse event has been statistically significantly associated with the company's drug in a controlled trial, that event cannot be said to have been caused by the drug. And companies can ensure their drug
never causes anything
by making sure that trials are organized so that important adverse events cannot become statistically significant.

Taking this approach, even when faced with assessments by doctors or their own personnel that the company's drug has caused some problem, based perhaps on a temporal link between a Vioxx or a Paxil, say, and a side effect that emerges on treatment, clears when the treatment is stopped, and reappears when the treatment is restarted, the company will still not concede a link.

What, then, about the long list of side effects reported in ads for the drug or in the
Physicians' Desk Reference (PDR)?
The company view is that these have been reported but this is not the same thing as being caused. These reports are merely anecdotal unless or until they have occurred enough times in clinical trials to have their link to the drug come out as statistically significant. Drug company personnel will of course warn doctors about all proven hazards their treatment causes (conveniently, none as no trials are done to establish a linkage)—it would be inhuman to do otherwise in such a sensitive domain as medicine. But they will not, as they put it, spread rumors about their drug—that would be good neither for the company nor for the patient. In adhering to this kind of “science” company personnel—and doctors—have found a way to live with themselves that has echoes of a totalitarian regime where people have learned to live with themselves on the basis that they are obeying orders or keeping to the rule of law.

What these doctors miss is that companies are working from assumptions or results rather than from data. There may have been thousands of reports of a problem to the company as clinically significant as Pierre Touery's use of activated charcoal, but these never stand a chance of becoming statistically significant.

When John Snow, investigating the outbreak of cholera in London in 1856, found a cluster of deaths from the disease on Broad Street, he was not inhibited by considerations of statistical significance because this concept hadn't been invented. He was free to recognize a cluster. Clinical prudence dictated a course of action—remove the handle of the pump. Science mandates efforts to go behind the data in an attempt to establish what has given rise to the cluster. But if companies had controlled the pump and took an Ian Hudson approach, they would not countenance the removal of the handle or any investigation of what the mechanisms giving rise to the cluster might be—unless the findings were statistically significant, which in all likelihood at Broad Street they weren't.

This account of what happens when our deaths or injuries on treatment are hidden or disguised may also help explain why doctors neglect what is happening to the person in front of them—in a manner that seems close to the antithesis of what good care should be. Anything that happens to the patient in front of a doctor has been degraded to an anecdote rather than an event that the doctor needs to pay heed to. If the doctor, using standard methods to decide if treatment has caused a particular problem, attempts to report it to a journal, the report will be turned down on the basis that journals no longer take case reports. If the doctor tried to report it to either the company or the FDA, it disappears into a statistical black hole that can swallow thousands of such events without the slightest indigestion. In the case of Vioxx it is estimated that there were at least 30,000 heart attacks linked to this drug from the time clinical trials first showed an increase in risk to the point where it was conceded there might be a problem and Merck drew back from its marketing blitzkrieg.
⁵

The result is that, astonishingly, in the midst of this emphasis on evidence, medicine is collecting even less evidence on those harmed by treatments than it has done in the past. Although we now have the ability to register all patients with particular diagnoses or patients receiving any drug and could track all outcomes and assemble the data on what actually happens when treatments are given, there is in fact a yawning hole—through which Cora and others in the prime of life are abducted by Hades. Those who love them, who are left behind, have a right to complain to Zeus that there has been a fundamental breach here in the contract between heaven and hell.

But in addition to frustrating the giving of decent and reliable medical care, the neglect of data on the problems treatment may cause is cutting off our ability to develop new drugs. The single most fruitful source for the discovery of new drugs continues to be observations that a drug given for one purpose in some cases is doing something else. It is perhaps no coincidence that as the possibility of reporting “adverse” events has dried up, with journals unwilling to take reports on unexpected outcomes in individual cases and companies hiding effects that do not fit with the business plan for a drug, so also drug discovery has dried to a trickle.

In response to questions as to whether major medical journals should insist on access to the raw data from controlled trials, the editorial staff of the
New England Journal of Medicine (NEJM)
said in April 2009 they do not see this happening in the near future—“that would be a large step outside our role”—while at the same time maintaining that the
NEJM
offers transparency.
⁶
As things stand we might all be safer if clinical trials were published in the
New York Times
rather than the
NEJM
, as the
Times
makes some effort to check the integrity of the primary sources for its stories, whereas the
NEJM
does nothing. When a
Times
story turns out to have been invented, as in the case some years ago of a series of articles written by Jayson Blair, it is national news, and the editor's job is on the line; not so when a drug trial such as Study 329 turns out to have been bogus.

Study 329 was the 1997 trial, discussed initially in
chapter 4
, in which Paxil was found to be no more effective yet more dangerous than placebo but from which GlaxoSmithKline (GSK) decided to select and publish the good bits, concealing the fact that six times more children became suicidal on Paxil than on placebo. The ghostwritten article was published in the most influential journal in child psychiatry, the
Journal of the American Association of Child and Adolescent Psychiatry
, whose editor at the time was Mina Dulcan. Dulcan had this to say in 2004 when interviewed by Shelley Jofre of the BBC after the history of the paper had become clear, and at a time when the paper was the basis of a fraud case taken by New York State against GSK: