Pharmageddon

by David Healy

  The figures for suicides, cholesterol, and diabetes were all buried in reports submitted by the company to the FDA. Even furnished with these reports to the regulator, it is difficult to establish what the true figures are and a good deal of data seems to be missing.24 But NICE and TMAP didn’t have the data and didn’t even have these reports that were submitted to the regulator—they were working only from the published evidence. Based on a thorough assessment of the publications alone, NICE came to the conclusion the newer antipsychotics were no better than older agents. But the published evidence still suggested the new drugs provided a better quality of life and a lower burden of side effects than the older drugs, whereas the raw data point to just the opposite conclusions.

  Against this background, NICE also had to manage a dynamic situation. First, how would clinicians and patient lobby groups, who had been bombarded for years with hundreds of publications extolling the virtues of Zyprexa and Risperdal and claiming these drugs liberated patients from some of the terrifying problems caused by the older agents, respond to a recommendation from NICE to use older drugs—had they chosen to give it? The 234 Zyprexa publications and further hundreds from the other new antipsychotics (Risperdal, Seroquel, Abilify, Geodon) played a great part in generating this pressure. Some of the panelists may privately have thought the older drugs were as good as the new ones, but it was difficult to offer evidence for this point of view, especially since they had no access to some of the most telling data. If NICE had come down favoring the older drugs, company-sponsored patient groups, told they should have the older drugs, would likely holler rationing, and even use this supposed rationing as an argument for desocializing healthcare. Second, just as journals do not publish articles critical of the pharmaceutical industry for fear of a legal action, so also NICE knew it stood to be dragged into a legal action if it came to a decision that was not based on published evidence. And since then, in the case of guidelines for Alzheimer’s disease, it has found itself sued even though its decision is based on the published evidence.

  Rumor has it that NICE was also faced with a British government that was in receipt of communications from several pharmaceutical companies threatening to pull out of the UK if the guidelines were not favorable to its products.25

  The NICE guidelines came out in 2002. Three years later two large independent studies, one American and one European, were published showing that older antipsychotics were as effective and tolerable as any of the newer agents, and superior to some of them.26 But if doctors wanted to follow the evidence and prescribe one of the older agents, they would have found a series of guidelines standing in their way, as these are only updated periodically.

  THE GREATEST DIVIDE IN ALL OF MEDICINE

  Having been invited to a number of guideline meetings, I had a chance in 1997 to convene one. As the secretary of the British Association for Psychopharmacology, I organized the first consensus conference to look at issues surrounding the prescription of psychotropic drugs to children. The growing number of prescriptions being written for ADHD (attention deficit hyperactivity disorder) had triggered the meeting, but on the day, the treatment of depression in children was the primary focus of attention.27

  There was an important difference between this and the pediatric guideline meetings that came later. In 1997, except for ADHD, there were few published clinical trials. Furthermore when it came to depression, the clinical wisdom as of 1996 was that unhappiness in childhood was not the same as depression in adults—it was not something for which pills were the accepted answer. As a result when authoring the final document, the premium was on treating the child rather than the condition. Clinicians were recommended to lay out all the treatment options—drug and nondrug—for patients and their parents and if the first treatment didn’t seem to be working they were advised to switch to alternate treatments even if not among those the doctor preferred. This was guidance rather than a guideline.

  One feature of the meeting became intriguing later on. I had invited all panelists and a number of pharmaceutical companies. SmithKline Beecham were present as were a number of the clinical investigators for Study 329, SmithKline’s trial of Paxil in depressed children outlined in the previous chapter. This study had been completed at the time the guidelines were written but I didn’t know about it and possibly very few others did and there was not a single mention of any Paxil study on the day of the meeting.

  Two years later, in 1999, TMAP issued guidelines endorsing the use of SSRIs in children who were depressed.28 By this time a trial of Prozac in children had been reported and it was known that several other trials were underway. In 2002, the FDA endorsed Prozac for treating depression in children. The FDA had also issued a tentative approval to GlaxoSmithKline for the use of Paxil in children and was likely to do so for Zoloft. An article that had appeared in Newsweek to coincide with World Mental Health day in 2002 claimed there were three million depressed adolescents in the United States, who were supposedly at substantially increased risk of career failure, divorce, alcoholism or other substance misuse, and suicide, all of which could, according to the Newsweek article, be averted by the new SSRIs just about to be approved.29 There was no hint here that unhappiness in childhood might be different than adult depression. The thrust of the article was that a failure to treat with medication would be equivalent to failing to give an antibiotic to a child with a life-threatening infection.

  When the FDA approved Prozac, Paxil, and Zoloft for use in adults in the early 1990s, they noted that the drugs were likely to be used to treat children and encouraged companies to run studies to establish the safety of the drugs in children. Sales of SSRIs for children had been creeping up steadily through the 1990s on the back of over seventy published “open studies” of these medicines—all claiming the drugs were marvelous. Open studies are ones in which a doctor knows what the drug is and the patient may be told as well. They invariably report positive results for a drug, but companies cannot use this kind of study to get marketing approval from FDA; they can only use randomized studies.

  Because there were so few good studies for any drugs in children, in 1998 the FDA Modernization Act (FDAMA) offered pharmaceutical companies a six-month patent extension for a drug if they submitted studies that examined safety issues in children. They didn’t have to prove safety. They just had to test for it. If the drugs showed hazards, the company still received the patent extension but would have to incorporate the hazard information in the label.30 This offer of patent extension gave the companies a hefty incentive to submit studies to the FDA on the effect of their drugs on children. A six-month patent extension for a Paxil or Zoloft meant easily over $1 billion in additional revenues. And there was every chance that the FDA would miss the problems.

  As a result, in 2003 when NICE set about drawing up a guideline on the treatment of childhood depression, six randomized trials of SSRIs in children had been published.31 The new guidelines were set to endorse the use of Prozac and other SSRIs for children.32 The use of these drugs was increasing rapidly in Europe and this endorsement would likely have opened a floodgate.

  In the case of Prozac there were two Lilly trials. Graham Emslie from Texas, who had participated in drawing up the TMAP guidelines for children, was involved in both. In clinical trials, it is customary to specify a primary measure of the success of treatment—such as the score on a particular rating scale or blood test—and if the drug fails to beat placebo on this measure, the trial is considered negative. On this basis, the first Emslie study, which started in 1990 but was only published in 1998, was a negative study even though the published article claimed it was a positive study.33

  A second study published in 2002 was also negative. After the first week of the study, all children who had a bad reaction to Prozac or a good response to placebo were excluded.34 It is common for a company to load the dice in its favor by excluding anyone who responds to placebo in the initial phase of the trial, but it was almost unheard of at the time to take the e
xtra step and exclude patients who reacted poorly to the experimental drug during the first week of their exposure to it. If they dropped out of the study, they should be counted as dropouts for adverse events, not eliminated from the study calculations entirely. This novel tactic has since been increasingly copied in company trials of drugs for asthma, hypertension, and other conditions.

  In the case of Paxil, the key study and the largest of the SSRI trials was Study 329, which as outlined in chapter 4 was a negative study that Sally Laden of Scientific Therapeutics Information transformed into an article demonstrating the remarkable efficacy and safety of Paxil. In addition to Study 329, Study 377 had also been undertaken in the 1990s but remained unpublished. Two further studies were presented at academic meetings in 2002; both claimed that Paxil was safe and

  effective.35

  The third of the major SSRIs was Zoloft. The FDA requires two controlled studies to let a drug on the market. Pfizer ran two studies. In each Zoloft failed to beat placebo.36 Just as with Study 329, these studies were ghostwritten. In this case they were published in the Journal of the American Medical Association and in the process transformed into one positive study—Zoloft was deemed effective and well tolerated. The design of these trials did not encourage the detection of any problems resulting from Zoloft, but even so, compared to children on placebo, there was a doubling of the rate of behavioral problems, including suicidality and aggression, in children on Zoloft and a tripling of the dropout rate for side effects.37

  By 2003, then, a series of articles all claimed the SSRIs worked, and so an impending endorsement by NICE did not seem surprising. GlaxoSmithKline had applied to the British regulator (the Medicine and Healthcare Products Regulatory Agency, MHRA) for a license to market Paxil for childhood depression. But in October 2002 and Spring 2003 two BBC investigative journalism programs had questioned the benefits of Paxil.38 Astonishingly, MHRA turned down GlaxoSmithKline’s application to license Paxil, and in support of their move took the unprecedented step of posting on its website the details of fifteen controlled trials on antidepressants undertaken by a number of companies in pursuit of a license for treating pediatric depression. Depending on the way one reads the studies, between twelve and fourteen of these fifteen studies suggested the drug being tested didn’t work and overall the studies showed a doubling of suicidal acts on the drugs compared to placebo. It was clear from these posted studies that there were yet further data that GlaxoSmithKline had not sent to the regulator.39

  NICE was faced with two problems. First, they worked from the published data but the MHRA posting made it clear there were many more studies. Of the at least fifteen studies conducted, only six had been published. The unpublished studies were all negative. The second problem was that even the published Paxil, Prozac, and Zoloft studies, it was now clear, had been manipulated so that essentially negative studies were transformed into positive studies, hiding the fact the drugs didn’t work and masking the problems of treatment. These revelations led researchers from NICE to pen an award-winning editorial in the Lancet, lamenting “depressing research.”40 This pointed to the impossible position any guidelines agency was in if companies withheld trials and distorted the data to the extent that had happened in the case of the pediatric antidepressant trials. It was left unsaid, but the position for doctors whose legitimate concerns about giving drugs like Paxil to children might conflict with the guidelines, had they been instituted, would have been even worse. The position for the children would of course have been worst of all.

  For a brief moment, some of those in NICE who had gone through this crisis toyed with the idea of insisting that the status of any evidence that came from company trials be downgraded. Up to this point, the rules of evidence-based medicine had been that the results of clinical trials trumped everything else. Now it had become clear that companies were selective in what trial data they released, and thus company data appeared to be worth a lot less than had previously been assumed.41

  But NICE dropped the idea of downgrading company trials. Could they be sure that a rule made on the back of the issue of antidepressants for children would hold water when it came to, say, trials of antihypertensives or analgesics or drugs for osteoporosis? If company evidence were to be downgraded, to what rung in the ladder should it be relegated—above or below the opinions of experts? Just how worthless was company evidence? And having dragged pharmaceutical companies into trials by insisting upon their necessity in order to gain a license, was this really the time to give them a slap in the face? Far better, surely, to work to improve company trials.

  There had, moreover, apparently been one positive outcome of the debacle. The Paxil data that MHRA made public confirmed the message of an internal GlaxoSmithKline memo that had come to light in the BBC investigations: that Study 329 had shown Paxil was not effective, so only the good bits of the data would be published. At hearings the FDA held in February 2004 on prescribing antidepressants for children I made it public. It found its way from there to the offices of New York’s Attorney General, who sued GlaxoSmithKline for fraud, and as part of the settlement, the company agreed to register all its trials on the web.

  The idea of a clinical trial register took off. Journals indicated that they would in future only publish accounts of trials that had been registered with a central trial register beforehand and been given a unique identifier. Such an identifier would have made it easier to establish that only four trials underpinned the 234 publications on Zyprexa. But clinical trial registers and Glaxo’s posting to its website do nothing to change the basic problem, which is that companies still do not made the raw data from these trials available.

  NICE finally did issue a guideline on pediatric depression in 2004: they recommended against using SSRIs as a treatment. In 2004, the FDA held a further regulatory hearing in September to follow up the February hearing. These hearings on antidepressants and suicidality in children led to the highest level of warning, a black box warning, being put on the drugs indicating that they might trigger suicidality. The FDA meanwhile did not license Paxil or Zoloft or other antidepressants for use in children.

  Far from this being a case of all’s well that ends well, however, the use of antidepressants in children shows how far our scientific standards have slipped and how this impinges on our ability to care for some of the most vulnerable people there are. These studies of antidepressants in children offer the greatest known divide in medicine between what published reports in the scientific literature say on the one side and what the raw data in fact show, but there is no reason to think this problem doesn’t extend to other treatments in other areas, from drugs for osteoporosis to treatments for asthma, female sexual dysfunction, PTSD, or other disorders. There was another landmark also—this was the only known case where all of the published studies were ghostwritten or company written.

  The fifteen controlled studies of these new antidepressants should stand as a celebrated example of what controlled trials are there for—to stop bandwagons in their tracks. But instead, between ghostwriting and selective publication of the data, companies have turned controlled trials into their primary means to turbocharge sales. The published papers endorsing the use of Paxil, Prozac, and Zoloft remain in print in the best journals and continue to fuel a boom in off-label sales of these drugs to children.42 There have been efforts to get Study 329 retracted but these have failed.43 It continues to be built into guidelines supporting the use of antidepressants for children.

  Erick Turner, formerly a reviewer with the FDA, has demonstrated that a third of the studies undertaken to get current antidepressants on the market for adults remain unpublished but even more worryingly a third of those published were studies the FDA regarded as negative but, like Study 329, companies published as positive.44

  In other areas of medicine, where the problem has not been forced out into the open, companies can use their published studies to capture guidelines as they had almost done in the case of antidepressants given to
children and, as we shall see in the next section, they continue to do in other domains. But even when the guideline is not captured, such studies and their publications transform the way we view things. In the case of antidepressants and children, for example, there is no longer any appreciation that childhood unhappiness might be anything different from adult depression. Someone attempting to express such a view today would find it difficult to get acceptance in anything other than a marginal journal.

  THE CAPTURE OF THE BIPOLAR GUIDELINES

  Classic manic-depressive illness, which typically leads to periods of hospitalization, was and still is rare. The recent invention of bipolar disorder obscures this but reveals much about how companies capture guidelines. When patients with the classic illness were admitted to hospital, either manic or depressed, they were typically too ill to be recruited for a controlled trial. This is not as problematic for good clinical care, including care that involves pharmaceuticals, as it may sound, however. Clinical trials rarely lead to discovery of any new drugs. Chlorpromazine, for instance, the first of the antipsychotics, was discovered in the early 1950s in Paris as a treatment for mania—but it was discovered because it made such an obvious difference to the patient in front of the doctor’s eyes, not because a clinical trial showed it had an effect.45 For the ensuing forty years, no one in medicine saw a need to conduct a trial for something as obviously beneficial as giving these antipsychotics to manic patients.

  That there had been no randomized trial data for these older drugs for mania opened up a golden opportunity for pharmaceutical companies to push these older drugs out of the market, when in the 1990s the companies came out with a series of new, albeit, as it turned out, no more effective and actually more hazardous antipsychotics. The way forward led through treatment guidelines.