How the Safety of Drugs was Destroyed

Emer Cooke, the CEO of the European Medicines Agency (EMA) is Irish. This is Dublin’s famous Halfpenny Bridge which for some reason features in an article outlining the history of the phrase Tail wags Dog which seems appropriate for what followed.

I wrote to Emer recently about a bizarre letter I received from Kinapse – A Brush with EMA – who run EMA’s medical literature review service and had come across an article of mine on the SSRI induced suicides of Romain Schmitt and Samuel Morgan. I have had a response from Juan Garcia Burgos, head of EMA’s patient engagement programe with Emer copied in.

Dear Juan

Thank you for your email. I think EMA are missing the bigger picture, as a consequence, I suspect, of applying algorithms to its processes.  Putting everything neatly into flowcharts, with a box for every function, and expecting the system to deliver a rational verdict is an understandable mistake.  But as even children in school once knew – To err is human but to really foul things up you need a computer.

Companies are adept at hiding problems the way EMA are now, perhaps accidentally, hiding them. Faced with evidence Paroxetine causes Autistic Spectrum Disorders, GSK set up teratology, pharmacology, epidemiology, pharmacovigilance, and psychology groups, each tasked to review the evidence from their area of expertise. The overall picture is compelling, but no-one group contained slam dunk evidence. In addition, likely for “the sake of objectivity”, the work was coordinated by a person with no background in these issues, who, when asked whether the work-groups considered studies reporting on developmental delay on SSRIs rather than autistic spectrum disorders specifically, responded: “I have no expertise to decide whether we should do that or not, my brief was to get the groups to review any published articles that had the term ASD in them”.

Forgive me for developing this point as Zygmunt Baumann did. Noting the documents showing German engineers designed railway carriages where people had standing room only to have drainage systems that would eliminate urine and feces rather than let them accumulate during a long journey, he indicates that a focus on specific tasks let these men avoid the bigger question.

A lack of effective pharmacovigilance now kills people on an industrial scale. Your email gives no hint EMA can spot the reasons for this or the need for big-picture thinking.

BMJ, Beasley and the Destruction of Pharmacovigilance

A decade after EMA settled in London, FDA set up MedWatch, and US companies began to encourage the public and doctors, who had been reporting adverse events to them, to report events to FDA instead. You mention conflict of interest and I’d imagine many would figure regulators are less conflicted than companies. Companies, however, had been under a legal obligation to establish cause and effect. Regulators aren’t.

Up to 2000, companies established cause and effect the way doctors once did and still should, looking at factors like challenge and dechallenge, reviewing a person’s medical history, checking for alternate explanations. Lawyers advised companies to stop this. Unless a randomized controlled trial (RCT) showed a statistically significant excess of an event on treatment, the company should simply record that reports had been received.

This allowed Ian Hudson, Chief Safety Officer for GSK, and later head of Britain’s MHRA, to testify under oath that GSK’s view was that Paroxetine has no side effects.  Doctors and the public now misread events in drug labels as events companies link to their drug. In this 2001 lawsuit, a jury exercising common sense found against GSK in this 2001 lawsuit.  Despite this, Hudson’s view ended up ensconced at the top of MHRA, and one imagines EMA and FDA.

Hudson’s view was based on an approach Eli Lilly took when defending Prozac against charges it causes suicidal acts.  This defense, in a 1991 Beasley et al article in the BMJ, hinged on a maneuver that was statistically illiterate.

1. Statistics should only be used on carefully collected primary outcome data and not on poorly collected data.
2. There was an excess of suicidal events on Prozac compared to placebo, which Lilly dismissed as not statistically significant and declared was, therefore, simply not happening.
3. In addition, Lilly fiddled the books, as other companies did, to add suicidal events to the placebo figures that did not happen on placebo, in breach of FDA (and likely EMA) regulations.

GSK adopted Lilly’s approach. This underpins GSK claims that paroxetine has no side effects.  All companies now do this for all drug side effects. This was done with regulatory connivance or at least tolerance and continues even though regulators have conceded some breaches of regulations and are flying in the face of statistical orthodoxy.

The Lilly/GSK approach justifies the strategy of simply reporting events to regulators knowing regulators are unlikely to establish cause and effect. For most adverse events, cause and effect cannot be established without interviewing the affected parties. For some events like birth defects, the issues are different but even here it is difficult to have confidence in current regulatory processes and women who once would not take drugs in pregnancy are now swallowing more and more – because drugs like antidepressants come badged with a ‘This Works’ label that misses key safety data.

Lilly’s 1991 article had another significant consequence. Company propaganda painted RCTs as the science of cause and effect for adverse events, and case reports in major medical journals, even when using scientifically supported methods to establish cause and effect, were branded as anecdotes rather than science.

Chilling Doctors

This had a chilling effect on medical journals, whose lawyers became agitated at the thought the journal might be sued for publishing claims that could not be supported.  Between them in 1978, BMJ, Lancet, Annals of Internal Medicine, JAMA and NEJM, had 335 articles reporting adverse events. BMJ alone had roughly 110 articles – two per issue of the journal. Today, they have none.

Not only does BMJ have no reports of adverse events, the journal seems scared to publish anything that even hints at possible cause and effect. Journalists now write articles for BMJ rather than doctors; a journalist almost has to consult medical experts – one at least on either side of the argument. This is very like what companies and EMA do when they assess problems in bits. It leads to an impression of possible signals, but no-one takes responsibility and companies are brazen enough to claim BMJ or EMA’s review of the issue, despite hints of a possible signal, gives them a clean bill of health.

Allied to the change in medical journals, Drugs Bulletins, which doctors in many countries used to get routinely and for free, stopped coming in the 1990s. These openly laid out the likely adverse events linked to new drugs like the SSRIs, such as dependence and other problems. But Drugs Bulletins have disappeared behind pay walls now and shy away from analyzing specific cases, perhaps even more so after Astellas took a legal action against Prescrire.  Even without pursuing these legal actions the whole way, companies produce a chilling effect.

Instead, doctors are now notified routinely, and for free, about the latest guidelines, which emphasize the benefits of treatment and rarely contain anything about the hazards. In the case of treatments involving drugs, these guidelines are based on a ghostwritten literature, some of which is fraudulent.  With antidepressants, over 50% of trials submitted to regulators have been negative but up to half of these published as positive, like Study 329, a negative trial of paroxetine in adolescent depression published as positive, were built into guidelines. We now have a crisis in children’s mental health – how could it be otherwise with such a comprehensive mismatch between what the medical literature claims and what actually happened in the trials?

In many cases companies have told regulators that a study was negative, but the company published it as positive. This is fraud and, in not speaking up about incorrect articles, FDA in this case, and likely EMA, has been complicit in fraud. While I appreciate it is not EMA’s job to police the medical literature, I imagine there are far more articles than just this one where EMA has been effectively complicit in fraud.

This might be something for your medical literature review service to consider doing but I do not think Kinapse would be the right company to partner with for this purpose.

It is difficult to think of a better update on the Biblical straining at a gnat but swallowing a camel than this – on one hand we have Kinapse straining at details of the weight of two young men whose treatment with SSRIs induced suicide, while EMA swallows huge chunks of a fraudulent medical literature which has airbrushed away the treatment inducing suicide effects of these drugs.

I approached Andrew Dillon, when he was CEO of the NICE Guidelines, making these points. He did not deny them but asked what NICE was supposed to do about it.  I guess EMA’s response will be the same – you await some leader to require you to change your approach.

It takes a Medical Community….

You state that

pharmacovigilance should not be seen as replacing everyday clinical practice but rather an activity supporting it. It is the job of medical doctors to ensure safe clinical practice for their patients.

From 1950 to the 1990s, doctors, not regulators, were the major contributors to pharmacovigilance.  Regulators made a minimal contribution before the 1990s and overall they possibly make a negative contribution now.

Just as It takes a village to raise a child so It takes a medical community to do pharmacovigilance.  Since 1990, that community has been destroyed.

Doctors are isolated in individual offices.  None of their journals report on the adverse events of the new drugs they might be using. They have heard on all sides that their judgements about adverse events are likely to be mistaken. Or they may infer that if they make these judgements, they will get sued. If even a medical journal, like BMJ, whom you might have thought could get the backing of the entire medical profession is scared, can you expect an individual doctor to have a more substantial backbone?  Tell me how a doctor is supposed to ensure safe clinical practice in these circumstances?

On this score, EMA seem to have missed perhaps the most important point in the article Kinapse are supposed to have reviewed for you.  Dr, Adams, Samuel Morgan’s doctor, who appears to think that citalopram caused Sam’s suicide, was told by his medical insurer not to mention this at Samuel’s inquest. Medical insurers routinely give this advice to doctors at inquests. This is a major interference with effective pharmacovigilance.  It has been going on for decades and likely in all European countries. What are EMA doing about it?

Dr Adams no longer has a medical community to support him in doing the right thing by his patient and the patient’s surviving family.

Romain Schmitt’s parents have found that there is no medical community to support them. Three decades ago, his parents would have been able to get a doctor to testify in a case where their son got worse when started on a low dose of a drug, worse again after an increase, deteriorated further after another increase, got much worse after a third increase, and died when another drug was added to the mix. Whatever the drug, and whatever the cause of death, an adolescent school graduate, knowing nothing about medicine, would be able to tell you this drug likely caused this death. French doctors close to universally say they do not have the expertise to give a view on a case like this.  Something is seriously wrong here because, if true, this suggests they should not be practicing medicine.

If I were the doctor who had prescribed paroxetine to Romain, after recovering from the embarrassment of such a mistake, I would take a legal action against the company and regulators who had kept me in the dark about these risks. Companies who commissioned a fraudulent ghostwritten literature and regulators who knew the published articles built into guidelines were fake but said nothing. These actions contributed to the destruction of the community of colleagues I depend on to help me keep my patients safe.

Pharmacovigilance is not about Rare Events

You mention that:

pharmacovigilance plays an important role in detecting adverse effects which do not show up during randomised controlled trials (such as rare or delayed side effects)

And that:

it is the job of the regulator to protect public health and to ensure that medicines are safe and effective in the population overall. It is also the regulator’s role to ensure that the benefits and risks of a medicine are always accurately reflected in its product information so that it can support healthcare professionals in their daily clinical practice.

These statements are wrong.

First, EMA regulates company claims. It has nothing to do with the practice of medicine, other than in so far as constraining company claims might create a better climate for safe practice.  Far from constraining company claims, EMA seems to me to be adding fuel to company fires.

Second, ensuring that medicines are safe and effective in the population overall is essentially aiming at an average risk benefit ratio.  There is no such thing.  Individual patients are the only people who can decide on benefit and risk balances.

I have several patients with OCD who do much better smoking nicotine than on the SSRIs which are approved for OCD. The only person who can decide on what the risk benefit ratio looks like in this case is a person who has done poorly on the SSRI, who then takes nicotine, and knowing the risks of nicotine but also the benefits obtained in his case, opts to forego SSRIs and take nicotine.  Just so you know, Arvid Carlsson, Nobel Prize winner, creator of the SSRIs, ran a trial of nicotine in OCD and showed just this – that it was a better treatment for some patients than SSRIs.

Third, how can EMA presume to be able to say anything about risk-benefit ratios?  You say pharmacovigilance has a place in detecting rare side effects that are not picked up in RCTs.

You operate on the basis of the reports companies write about their trials, which often pick an effect of a drug on the basis that money can be made from it, an effect that is much less common and less important than other effects. For instance, the effect of an SSRI on a depression rating scale score is minimal, with less than one in ten people showing any specific benefit, whereas the genitals and love making of close to 100% of those taking the drug are affected within 30 minutes of a first tablet.

The depression ‘benefit’ only comes about in people who have a mild condition that clears spontaneously without treatment. SSRIs are ineffective in severe depression. The label also misleads in that these minimal effects reflect an average change in rating scale scores, that tell a doctor nothing about how the treat the person in front of them. Current labels give the impression an antidepressant will work when it is quite likely not to work. There is nothing in the label that warns doctors or patients about this.

I am more concerned about the safety aspect. Mildly depressed patients have no genital symptoms or difficulties making love linked to their condition, prior to going on treatment. Companies knew this before running their treatment trials, where doctors like me were told not to ask about sex, leaving companies able to submit reports to regulators and publish medical articles claiming less than 5% of us have sexual issues on treatment, which clear on stopping treatment when we now know this is not the case.

Millions of Europeans now have difficulties in making love, either because they cannot get off treatment or, if they do stop, they find their sexual nightmare deepens, into a permanent problem for which we have no treatment – now called PSSD, post SSRI sexual dysfunction.

The sexual problems on SSRIs are their most common effects and barely appear in the label, certainly not in a manner that contributes to people taking these drugs with informed consent. Whose role is it, if not pharmacovigilance, to tackle these not rare or delayed effects?

Ten years after Prozac launched, after a flood of reports and taking it seems standard medical approaches to causality, Lilly included a mention of enduring sexual dysfunction in its US label. It took a further decade and the efforts of the PSSD community to get EMA to include a vanishingly small mention of this problem in drug labels – exempting the still on patent vortioxetine from this requirement.

Patients still get laughed at by doctors if they describe their problem, even detained for their crazy ideas if they mention PSSD, or they may get switched to vortioxetine which unquestionably causes PSSD.

There is a political dimension to this. In many European countries, Britain and the US, reproductive rates have fallen well below replacement rates. They would be catastrophically lower again were it not for the fact that up to one third of births are to mothers born outside the country they give birth in.  These other communities have been much less likely to use SSRIs. At a time when immigration has become a hot issue, if noticed, the fact that a lack of effective pharmacovigilance may be contributing to a significant replacement of native populations may become a hot political potato.

I have relegated your Syneos points, which I regard as a sideshow to Appendix 1.

Summary

To summarize the key points. Some of those who framed the 1962 FDA Act were worried enough to stress their aim to regulate pharmaceutical company claims and not the practice of medicine. The concerns may have arisen in part because, since 1951, new treatments have been uniquely available on prescription-only. For pharmaceutical companies doctors, not patients, are the consumers and companies work to understand doctors better than they understand themselves.

New drugs were on prescription, as companies said, to avoid having to put an entire medical course on the label.  Doctors offered an extended label. Politicians and the public viewed medical professionals as leading on safety and knowing more about drugs than the bureaucrats regulating medicines. Leadership is not a box in the flowcharts outlining bureaucratic functioning.

The 1962 concerns were prescient. The pharmaceutical industry now regulates medicine, through agencies like EMA and FDA, rather than society regulating the pharmaceutical industry.

Change will need something like one of or a combination of these options:

1. Removing the prescription-only status of many but not all drugs.
2. BMJ alerting doctors to the mistake it made with the Beasley et al 1991 article and supporting a return to effective pharmacovigilance.
3. Doctors otherwise realizing they are likely to go out of business if they do not recognize their treatments are in general at least as harmful as nicotine or alcohol.
4. EMA offering a view on the advice of medical insurers to a doctor like Dr Adams in the Morgan case.
5. EMA making it clear that regulators (and this includes CDC) cannot substitute for doctors when it comes to pharmacovigilance.
6. EMA coming clean on the publication of articles on medicines that claim a trial showed the drug worked well and was safe when EMA knows this is not true and the company knows it is not true.
7. EMA and other regulators acknowledging that we stopped requiring drugs to demonstrate effectiveness in the sense that most people take that word to mean, namely saving lives or reducing disability, several decades ago.

Otherwise, the Tail will continue to wag the Dog.

Yours

David Healy MD

cc. Emer Cooke