Report of the CSM’s Expert Work Group on the Safety of Selective Serotonin Reuptake Inhibitor antidepressants

1. The launch
Here was a 235-page manuscript report, full of technical jargon and charts, along with tables, diagrams and many references to learned works. The report was presented as the output of a meticulous scientific enquiry that had lasted over eighteen months – about ten times longer than it takes to screen a new drug application.

But the report was released only to journalists, and four days after the launch, copies of the report are still not generally available. Nor is there yet any reference to it on the website of the Committee on Safety of Medicines (CSM), in whose name this review was done. Only the recommendations have been reported, so peer review will take time.

The report was launched at a press conference organised by the Department of Health (DoH), at 10.30 on the morning of Monday 6 December. None of the journalists I have spoken to (n=6) had received their (confidential) invitations from the DoH before 5.00pm on the Friday before. One of them charitably suggested that perhaps the authorities were waiting for the London Stock Exchange to close.

However, the global stock market never actually closes, and ‘news management’ probably comes closer to the mark – but yes, I have an axe to grind. Not for the first time, the DoH refused me admission to this briefing on the technically defensible grounds that I am not a member of the press. I have written extensively about the safety of SSRIs for many years, was invited to give evidence to the Working Group, and regularly brief journalists on drug policy and safety issues. I was even for some years a member of the National Union of Journalists - but yes, I no longer carry quite the right cards.

The same exclusion applied to my distinguished colleague, Dr Andrew Herxheimer. He did manage to buttonhole Professor Kent Woods to request admission – but Woods tersely denied any responsibility. He is Chief Executive of the MHRA (Medicines & Healthcare products Regulatory Agency) and was the front man that morning. The DoH is like any other supertanker or juggernaut: changing course requires teamwork and planning. Woods has been in the job for less than a year, easily long enough to appreciate the triumph of process over individual initiative and common sense.

It was a bit cold, but Andrew and I spent a productive several hours on the forecourt of Richmond House (HQ of the DoH), talking to journalists and doing a few interviews. Here we were fortunate: the journalists all trouped out of the briefing about 15 minutes after they’d all filed in. A fire alarm had gone off, so everyone evacuated the building and milled around on the pavement outside for the best part of an hour.

News management? Apart from the decision to engineer a headline response by offering the report only to the press, the presentation was organised to focus strictly on the use of antidepressants. The other two key figures on the podium had not been directly involved in the CSM Review. Their presence would have discouraged questions about the big underlying issue – the performance of the drug manufacturers and regulators themselves.

Alongside Woods was Dr Andrew Dillon, Chief Executive of the National Institute for Clinical Excellence (NICE), whose new guidelines for the treatment of depression were being published that same day. The other was Professor Louis Appleby (National Director for Mental Health), whose main concern would be about the risk of suicide and the need for effective treatment of depression:

"The CSM has delivered one of the most comprehensive reviews of a class of medicines ever to be completed and it has been painstaking work, examining evidence from literally hundreds of clinical trials. What’s important now is that their advice is put into practice. Publication of the NICE guidelines gives us the tools to do the job so that patients and prescribers can together make the best informed decisions." (DoH press release, 2004/0433)

The lack of other key players was notable. Where was Professor Gordon Duff, Chairman of the Committee on Safety of Medicines, which had supposedly initiated this review – and in whose name a warning letter to prescribers would be published that same day? Why was there no sign of Professor Ian Weller, Chairman of the CSM’s Expert Working Group whose report was being released? And what had happened to Professor Sir Alastair Breckenridge, Chairman and ‘public face’ of the MHRA and former Chair of the CSM? He was a key figure in this affair and for many years a consultant to the manufacturers of Seroxat (paroxetine), the troublesome drug that prompted this view.

In The Guardian, the redoubtable Sarah Boseley pointed also to the absence of any representative of the Royal College of Psychiatrists or the mental health charity MIND. Richard Brook, its chief executive, had resigned from the expert working group in March. Later, he would tell Panorama: "I have little confidence that the drugs they’re licensing day by day are being licensed in a way I would feel appropriate and what’s even more concerning I have very little confidence in drugs that have been regulated in the past."

So, one way and another, the headlines won the day. "Tighter warnings on antidepressants" (Reuters); "Are anti-depressants over-prescribed?" (BBC News); "GPs Told to Use Alternatives to Anti-Depressants" (Scotsman); "New guidelines for prescribing antidepressants (Medical News Today); "GPs giving Prozac to thousands who don’t need it" (The Times); "Curb handouts of the happy pills, doctors told" (Daily Mail); and "Regulators Suggest Restricted Use of Drug" (Associated Press)

To date, the regulators (and by extension the drug companies) have managed to evade the scrutiny of their performance that is now urgently overdue. Gloss and spin rule, OK? The authorities have misinformed doctors and prescribers for many years, but now seem to be blaming them. 

CM, 10 December 2004

2. Major outputs
What is the major output of this report and what of its clinical significance? The key to both is a two-page document that is neither part of the Expert Working Group (EWG) report nor formally endorsed by its members. It is a list of wording changes to the Summary of Product Characteristics (SPC) for all SSRIs, proposed by the full Committee on Safety of Medicines (CSM) on the basis of the EWG report.

The SPC is an important document. It is the small print in the implied contract between drug manufacturers and the regulator - as proxy for prescribers and patients. It formally defines what a drug should be used for, spells out the approved drug dosages and identifies known and suspected hazards. It used to be called the ‘data sheet’ but is now the SPC and applies across the EU, as ‘the label’ does in the USA.

It seems most unlikely that UK prescribers often refer to the SPC. These days, with many other sources of information, the SPC only masquerades as a set of prescribing instructions. Its overwhelming significance is as a set of rules for manufacturers – right down to the wording they must use in the tiny-print ‘health warning’ section of drug advertisements. The SPC is part of the product license and legally determines what drugs should be used for, and how manufacturers may and may not promote them.

Generally, companies write their own SPCs, which they then submit for regulatory approval – and when it matters, company lawyers fight hard for the wordings they want. In marketing terms, the wording can make or break a drug. So it is unusual – especially over a decade down the line – for the regulators to publish in advance the new SPC terms they propose. On this occasion, however, I would expect the SSRI manufacturers to comply with alacrity, once they have stopped jumping for relief and with joy.

Even before you take stock of the proposed SPC changes, this was a dismal end to the hard work of the EWG, and it seems a disgrace that the CSM allowed it. These proposed SPC changes were publicly presented as "advice" from the Committee on Medicines as a whole - supposedly a distillation and encapsulation of the findings of its expert working group. In practice, the proposed new wordings reduce the EWG’s main output to some minor small print changes - no doubt rubber-stamped by the CSM on the basis of proposals made by the MHRA.

Aside from the paucity of the SPC proposals – see below – the immediate point is that the CSM appears to be either impotent or oblivious to the limitations of the regulatory process, including the seamy side of the relationship between companies and regulators. The CSM was arguably compromised from the moment it agreed that the working group should be lumbered with impossibly restrictive terms of reference, but here the Committee missed its last opportunity to reflect on the obvious wider implications of this affair.

Individual CSM members are overwhelmingly clever and utterly decent people but, collectively, it would misrepresent them to say they ducked. As a decision-making organisation, the CSM has been squeezed out of existence – compromised by industry influence and lack of public confidence, comprehensively manipulated by the MHRA, and effectively cowed.

It seems extraordinary and more, that the companies themselves were asked to provide and analyse the main data that informed the EWG review. The EWG largely ignored the raw data; they relied instead on the clinical summaries that companies had prepared. Yet throughout this 18-month enquiry, all kinds of evidence came to light of critical omissions and opportunities missed. Is it not alarming, for example, that prescribers have been given the wrong dosage information for 15 years or more?

Part of the problem is the failure by regulators to contain a corporate culture that positively relies on cut ethical corners and games of hide-and-seek. Relationships between companies and regulators bring to mind captives/jailers and cats/mice or fox and chickens, if you prefer.

Within a few weeks of its establishment, the EWG found evidence of lack of efficacy that led it to contraindicate paroxetine (Seroxat) for depression in children – but this was from trials the manufacturer had conducted years before. Earlier this year, a 1998 company memorandum surfaced that helped to explain this. SmithKline Beecham thought:

"… it would be commercially unacceptable to include a statement that efficacy had not been demonstrated, as this would undermine the profile of paroxetine". (Medicines out of Control? p.202)

Then there was this, from a 1999 Pfizer memorandum about how to respond to a request from the US Food & Drug Administration for adverse drug reaction (ADR), presumably data for sertraline (Zoloft/Lustral), now the leading SSRI.

"The actual task of generating a list of events with a report rate for each event is a fairly easy task, the main question is do we want to provide FDA with a simple straightforward report (report 28C) or do we want to give them a complicated and confusing one (report 28)?"

The ambiguity here makes it just conceivable that the company wanted to be as helpful as possible, but the overall context suggests not. For example, another series of Pfizer memoranda (1996) described as follows the company’s position in dealing with the Norwegian drug regulators:

"We should not volunteer to describe the withdrawal symptoms, but have an agreed list prepared in case they insist" …. "We need to be careful about this, bearing in mind label harmonisation. Ideally we agreed we should keep symptoms of withdrawal out of the label."

Further evidence recently emerged suggesting that the regulators actively engage in the duck and weave. How have the SSRI manufacturers satisfied the regulators’ requirements for proof of efficacy from trials of active drug against placebo? They have submitted evidence from trials that give positive results, ignoring the ones that didn’t. Pfizer’s antidepressant, reboxetine was licensed in Europe (in 1997) on the strength of one positive trial result from the eight that were done – still not quite enough to get it a license in the USA (Lancet, 19 June 2004, 2087).

The gaping hole in the EWG report is the lack of focus on the modest efficacy of SSRI drug treatment – and the same might be said of the NICE guidelines for the treatment of depression that were informed by the EWG review. Yes, the Group did quickly establish the lack of antidepressant efficacy in children, because the data were limited and therefore manageable. But in relation to adults, the EWG seems to have inferred efficacy mainly on the grounds that not treating depression may be more hazardous than doing so. The point is debateable, depending largely on the nature of depression, and the source of information:

"Suicide is rare, even amongst people with depression" (EWG, p. 68). Alternatively, "when people with depression are left untreated, 15% will actually commit suicide" (Spokesman for Eli Lilly, the makers of Prozac (fluoxetine), quoted by BBC News, 12 December 2004). See also Medicines out of Control? p. 47 FN 2)

The EWG did not directly assess antidepressant efficacy in adults; instead, its report includes a chapter on ‘Burden of Depressive Illness and Self Harm". Nevertheless, the CSM Chairman represents this as the key conclusion: "SSRIs are effective medicines in the treatment of depression and anxiety conditions". And that is what matters, above all, in the SPC.

The new NICE guidelines for treating depression tackle the efficacy issue to the extent of recommending non-drug alternatives for mild to moderate depression – but the CSM’s proposed revisions for the SSRI Summaries of Product Characteristics (SPCs) do no such thing. Subject to additional small print warnings, the manufacturers will be able to promote SSRIs exactly as they do now – notably with ‘disease awareness’ campaigns, orchestrated by PR agencies who target the "authoritative third parties" who communicate with you and me.

The new small print warnings for SSRIs relate to the risk of suicidality and drug withdrawal reactions. To exaggerate slightly, in the interests of brevity and clarity, they merely tweak the existing wordings. They say virtually nothing that could not have been said years ago, and should have been. Moreover, the CSM has proposed no change to the SPC to reflect the EWG’s one new finding – that venlafaxine (Efexor/Effexor) should in future be prescribed only by specialists.

The proposed new SPC does, however, include one obvious ‘exclusion clause’. It is buried deep in the verbiage and all but drowned at the end of a long paragraph that begins thus: "Depression is associated with an increased risk of suicidal thoughts, self-harm and suicide (suicide-related events). This risk persists until significant remission occurs". No attempt is made to distinguish between more and less severe depression, and remission due to drug or other factors.

This exclusion clause is a fragment of text that will no doubt allow the drug regulators to say at some time in the future, that they did indeed warn of the risk of antidepressant-induced suicidal behaviour, back in 2004. The small print says: "Furthermore, there is evidence that in a small group of people, antidepressants may increase the risk of suicidal thoughts and self-harm".

But for the time being, the regulatory machine is relying mainly on the "good evidence" that points to another finding, "that there is no clear increase in the risk of suicide from SSRIs compared to other antidepressants". The lack of reference to paroxetine (Seroxat/Paxil) is conspicuous, but no justification for either relief or joy.

CM, 17 December 2004


3. Reading between the headlines

Where might this report by the Expert Working Group (EWG) rank in relation to other enquiries of threats to public health? One obvious model is the report of the Southwood Working Party (1989) into the risks of BSE (Bovine Spongiform Encephalitis – ‘mad cow disease’).

Alas, it took several years before the limitations of the Southwood Report became known. Its weaknesses were finally revealed in The BSE Enquiry report (2000), an exhaustive examination of the issues achieved through formal judicial review. This definitive review was ordered soon after the election of the Blair government in 1997; the BSE crisis was firmly identified with years of Tory rule.

The BSE enquiry proper reported, in great detail, in 2000. It found that the earlier Southwood Report, "had, necessarily, been based on very limited data" – but that this and other small print caveats had been ignored. The Southwood Working Party did in fact warn that, "if our assessment of these likelihoods are incorrect, the implications would be extremely serious". But the headline conclusion – "that it was most unlikely that BSE would have any implications for human health" – won the day:

"Unfortunately, this warning and the tentative nature of the Working Party's conclusions were not appreciated or were lost sight of. Right up to 1996 the Southwood Report was cited as if it demonstrated as a matter of scientific certainty, rather than provisional opinion, that any risk to humans from BSE was remote."

The fanfare launch of the EWG report on the safety of SSRI antidepressants underlines the risk that it will go the same way as the Southwood Report. Far from warning of the possibility that they might not have got it right, and of what might happen as a result, the EWG Chairman’s foreword invites confidence and trust: "If the Group has done its work well, this review will be recognised as the most comprehensive review of the available data so far." Only when you get into the detail of the EWG report – including most of chapter 11 – does it become clear how much of the data is either inconclusive or missing.

But, of course, very few people will get into the detail. So it seemed worth trying to find some element of the report that encapsulated the problem with the whole thing. Probably this one sentence conveys the uncertainties behind the headlines best of all:

"From the available clinical trial data, both published and unpublished, a modest increase in the risk of suicidal thoughts and self-harm for SSRIs compared with placebo cannot be ruled out, although neither can a modest benefit." (p. 112)

Why did this one sentence sweep the board? Here is a shorthand stab at it:

  • This is a fine example of the lofty obfuscation characteristic of drug regulatory communications. The statement claims authority, detachment and objectivity, but the evidence points nowhere, making the interpretation seem all the more Martian.
  • Specifically, the use of the word, ‘modest’ to describe both minor benefits and serious harms nicely illustrates the regulatory obsession with numbers, the main highway of departure from human reality and common sense. If an SSRI provides some modest benefit, it will make little difference to anyone or anything, whereas a modest increase in suicidal behaviour would tend to be devastating for those involved.
  • Under the surface, this sentence neatly explains the limits of understanding of the experts and authorities, and of the data on which they (and we) rely. Working only with "the available evidence", the EWG asserts with great authority that it cannot rule out the possibility that these drugs either do good or harm. When clinical trial data prove such blunt instruments, it seems all the more extraordinary that the CSM did not contraindicate drug use for mild depression.
  • Finally, the reference to both published and unpublished data is a reminder that good drug news always sees the light of day, while bad news often doesn’t - and that secrecy in drug regulation abets such abuse. This so-called ‘publication bias’, (cheating, more like), is now a widespread cause of suffering and ill-health.

That is no mean achievement for just one sentence, but another is a worthy runner up:

"There is some uncertainty about the true frequency of withdrawal reactions with the SSRIs" (p. 150)

  • Fifteen years down the line – and after unprecedented complaint about withdrawal problems - we finally discover that the regulators never actually required SSRI manufacturers to find out how many people might get hooked on their drugs. Was this indifference, negligence, impotence or what? Three companies – incredibly, including the manufacturers of venlafaxine – could produce no controlled clinical trials specifically designed to examine the impact of withdrawal.
  • Table 1.2 shows that the EWG examined (the manufacturers’ summaries of) 143 studies and trials for evidence of SSRI withdrawal problems that might signal drug dependence. Of these, 115 studies/trials (80%) had not been specifically designed to examine withdrawal (i.e. many patients would have stayed on the drug). Later (p. 131), we learn that, "approximately 30% of patients will experience withdrawal reactions upon stopping venlafaxine treatment, of which up to approximately 12% will be severe in nature". In other words, many thousands of Ef(f)exor users (3% to 4%) will become incapacitated when they try to stop – though there is no ‘reliable evidence’ this is so.
  • This sentence is also a reminder that, in spite of mountainous and compelling ‘anecdotal’ evidence, the regulators lost the plot for years. The measure of the failure – but no hint of it in the EWG report – was the admission in 2003 that SSRI withdrawal symptoms were very common (25%) and not rare (0.1%)

The table below provides further context for analysis of the EWG report. It gives a count of some of the most and least frequently used words (and strings) that signal the timbre of the report, and the lack of evidence (and investigation) that make the headlines seem a sham.

Word string Number of mentions and comment
"Availab(le) (ility)" 73 occurrences – almost all caveats relating to the possibility or actuality of data insufficiency
"Important(ly)" 43 references – of which three (pp. 7, 68, 186) related to EWG recommendations. See details.
"Balance(d)" 35 mentions - almost all in connection with the need to balance drug benefits and risks/harms, but without reference to the built-in bias in the reckoning of positive and negative drug effects.
"No evidence" 34 references – mainly pointing to some lack of risk that appears from the available data. See details.
"Limitation(s)" 24 occurrences - overwhelmingly references to insufficiency of data
"Further research" 15 mentions – all relating to lack of data and/or research that is needed or required
"No clear evidence" 13 references - all referring to some lack of certain evidence of risk or harm
"Fail(ed) (ure)" 11 mentions – none of them suggesting even the possibility of regulatory shortcomings.
"Rule(d) out" 8 references – all to harms that the EWG cannot completely and confidently exclude.
"Responsib(le) (ility) (ilties)" 5 references - one to family responsibilities; the other four to drug action and/or physiological response. (e.g. "Cytochrome P450 enzymes are responsible for the hepatic metabolism of many drugs.)
"Robust" 4 mentions – three of them as in, "not robust"
"Hope" 3 references - all in the short introductory statement by the chairman of the EWG (e.g. I hope the legacy of the Group will also be in looking forward, to identifying the further work necessary to establish safety, advising on study design and identifying the lessons to be learned.)
"Lesson(s)" 2 occurrences – both empty promises "to consider what lessons have been learned during the process of the review".
"Critic(s) (ical)" 1 mention – relating to the accessibility and readability of patient information leaflets.
"Vital" 1 reference – to draw attention to the EWG/MHRA’s reliance on a reassuring lack of evidence from the UK’s General Practice Research Database.
"(Un)fortunate(ly)" 1 reference – "These are the ideal trials to assess the benefit of increasing the dose but, unfortunately, they are rarely performed." (Re: Randomised non-responder trials. See 9.2.2)
Other 0 "Regret(s)" or "Mistake(n)(s)" and nothing "Essential", "imperative" or "indispensable"

In short, the truth about the benefits and harms of SSRI antidepressants remains badly obscured, and the case for an independent enquiry into drug regulation now seems all the more irresistible. We seem to have here before: "Throughout the BSE story, the approach to communication of risk was shaped by a consuming fear of provoking an irrational public scare." Mark these words on lessons to be learned, from the report of The BSE Enquiry (2000):

"Our experience over this lengthy Inquiry has led us to the firm conclusion that a policy of openness is the correct approach. When responding to public or media demand for advice, the Government must resist the temptation of attempting to appear to have all the answers in a situation of uncertainty. We believe that food scares and vaccine scares thrive on a belief that the Government is withholding information. If doubts are openly expressed and publicly explored, the public are capable of responding rationally and are more likely to accept reassurance and advice if and when it comes…"

  • "To establish credibility it is necessary to generate trust.
  • Trust can only be generated by openness.
  • Openness requires recognition of uncertainty, where it exists.
  • The importance of precautionary measures should not be played down on the grounds that the risk is unproved
  • The public should be trusted to respond rationally to openness
  • Scientific investigation of risk should be open and transparent.
  • The advice and the reasoning of advisory committees should be made public."

CM 5 January 2005

4. Conclusions?

It seems clear from the launch and the volume of spin that the report of the Expert Working Group (EWG) on the Safety of SSRI Antidepressants was never intended for peer review, nor is it fit for it:

  • its scope is impossibly limited
  • it is remarkably incomplete
  • the authorship is uncertain
  • key source data are unavailable
  • bias is unacknowledged
  • perspectives are skewed
  • and it is tortuous to read

The legacy of the EWG report will probably have little to do with the safety of SSRI antidepressant users. Its findings and recommendations – other than those relating to the need for research – still lag well behind common sense. The real significance of this report is to do with the rich evidence it gives of the inadequacy of present systems of drug control.

It will take time to unpick and mine all the data in this report, and who knows how long it will take for all the implications to sink in. One day, it will surely be obvious that that the present system of drug regulation fails, both as a health endeavour and as a democratic enterprise. In the meantime, this report explains both why the dissolution of the system is nowhere nigh and long overdue.

CM   21January 2005

5. Postscript

There is a delightful scene in the film, The Court Jester (1956), in which the eponymous Danny Kaye had to be Knighted in order to joust with the cruel Lord Griswold on (socially) equal terms. So urgent was the need to prong the Court Jester to death, that the traditionally sedate and solemn ceremony, including many tests of Knightly accomplishment, were all grotesquely speeded up, very much accompanied by courtiers chanting, "yea, verily, yea".

The rush to publish the EWG report is reminiscent of this, and yea, verily, yea seems a fair approximation of the CSM’s response to it. But this became clear only in February 2005, when the MHRA published the summary minutes of the meeting held by the Committee on Safety of Medicines on 28 November 2004. This was the last opportunity CSM members had to discuss the report before 3 December, publication day.

It was a busy meeting, lasting from 10.00am to 4.45pm. The Committee "heard a company’s oral presentation" to support a license application. They also reviewed three full drug license applications, plus an application to increase by 33% the maximum daily dose of a drug for adolescents with ‘attention deficit hyperactivity disorder’. Later, they considered unspecified "issues" relating to drug treatment of osteoporosis, and others to do with a vaccine for paediatric immunization. They also "noted" [a] a request to sell "benadryl allergy relief" over the counter; [b] a consultation letter about conflicts of interest; [c] a paper on freedom of information; and [d] the "monthly pharmacovigilance statistics".

But the main agenda item at this meeting was a presentation on the findings of the CSM Expert Working Group on the Safety of SSRIs. The operative word is "findings", because "the Committee was informed that the full report of the Expert Working Group would be circulated to CSM the following day." In other words, the Committee on Safety of Medicines discussed the output of its Expert Working Group before actually seeing the report – and only after the MHRA had produced the final version.

Critically, the CSM also seems to have rubberstamped the MHRA’s proposed changes to the legal status of SSRI antidepressants (as defined in the SPC). Committee members were presented with the MHRA’s proposals, but did not discuss them: they were "asked to comment on the draft wording in writing following the meeting". There is no sign of it, if they did.

Bear all this in mind as you read further annotations to this review. Far from being the honest product of critical, independent minds, this report seems in essence to be an account that the regulators give of themselves.

Charles Medawar
9 February 2005

This review led to a complaint of regulatory failure