Ben Goldacre, The Guardian, Saturday 14 August 2010
This week the drug company AstraZeneca paid out £125m to settle a class action. Over 17,500 patients claim the company withheld information showing that schizophrenia drug quetiapine (tradename Seroquel) might cause diabetes. Why do companies pay out money before cases get to court?
One interesting feature of litigation is that various documents enter the public domain. This is how we know about the tobacco industry’s evil plans to target children, or the fake academic journal that Elsevier created for Merck’s marketing department, and so on.
One of the most revealing single documents ever to come out of a drug company came from an earlier quetiapine case, now part of a large online archive: it is an email from John Tumas, Publications Manager at Astrazeneca. In it, he helpfully admits that they do everything I say drug companies do.
“Please allow me to join the fray”, John begins, in response to a colleague. “There has been a precedent set regarding ‘cherry picking’ of data.” Cherry picking is where you report only flattering data, and ignore or bury data you don’t like. The ears of lawyers will prick up at any use of the word “bury” in relation to drug company data, as it implies something deliberate, but luckily John uses this word himself. The precedents set on cherry picking, he explains, are “the recent Velligan presentations of cognitive function data from Trial 15 (one of the buried trials).”
Trial 15 is a legendary trial commissioned by AstraZeneca. Patients with schizophrenia who were in remission were randomly assigned to receive either AstraZenece’s quetiapine, or a cheap old-fashioned drug called haloperidol. After a year, the patients on Seroquel were doing worse: they had more relapses – a real world outcome that really matters – and worse ratings on various symptom scales. These negative findings were left unpublished: to use John’s word, they were “buried”.
But in among all these important negative findings, on a few measures of “cognitive functioning” – an attention task, a verbal memory test, and so on – Seroquel did better. This finding alone was published – by Velligan et al in 2002 – the fact that patients on Seroquel had worse outcomes for their schizophrenia was not. And Velligan’s paper was not a side show: it went on to become a highly influential piece of work, with over a hundred other academic research papers citing it. Many researchers can only dream of pubishing such a well cited piece of work.
Meanwhile study 15 also found that patients on Seroquel gained, on average, 5kg a year. This plainly put them at increased risk of diabetes, which is what Astrazeneca are now paying to settle on (and in any case, 5kg weight gain is a serious side effect in itself).
Drugs have side effect, and quetiapine is not unique: psychiatric drugs in particular can do more good than harm, overall, but many also have serious, common side effects. When that is the case, it’s especially important that doctors and patients know all the risks, so that sensible and informed trade-offs can be made.
This is just the first paragraph of one email over 100 in the last quetiapine case. It is endless: in exhibit 13 Richard Lawrence sends an internal memo to colleagues: “Lisa has done a great smoke and mirrors job” on trial 15, and so on.
It is also, remarkably, unremarkable. The pharmaceutical industry’s behaviour has collapsed into farce. Doctors and academics – who should feel optimism at working with them to develop new treatments – feel nausea instead, knowing there are only informal and ad hoc systems to deal with buried data, and these systems have clearly failed.
In 2005 the International Committee of Medical Journal Editors put their foot down, and said their journals would only publish trials that were fully registered before they started, which should make trials that went missing much easier to spot. Several years later, as we saw in this column, less than half of all the trials that ICMJE editors published were adequately registered, and more than a quarter weren’t registered at all.
After the New York attorney sued GlaxoSmithKline over their “illegal and deceptive” reporting of the risks of their antidepressant paroxetine (tradename Seroxat), GSK agreed to publish all trial data on a website: but this was only for that one company, and even there it clearly did little to fix the problem. We saw last month that GSK and the FDA sat on data showing that rosiglitazone (tradename Avandia) increased the risk of heart problems, several years after this agreement.
This week in the open access academic journal Trials, two researchers set out some ideas around international law to try and fix this gaping hole, and stop companies hiding data that doctors need to make safe decisions for their patients. They don’t go far enough. Any company withholding trial data harms patients. I can’t understand why any ethics committee would allow any company currently witholding data to confuct further experiments on people. I can’t see why the state doesn’t impose crippling fines. I hope it’s because politicians don’t understand the scale of the harms.