Statement to the CSM working group on SSRIs

The following is a slightly edited version of the paper
submitted and presented to this enquiry in July 2003,
by Andrew Herxheimer FRCP and Charles Medawar


The terms of reference of the SSRI Working Group duck the most important issues: could this crisis have been avoided, and what might now be done to avoid similar problems in future?

For the record, our view is that the regulators themselves played a major part in creating the problems now being addressed. We believe that nothing short of a thorough overhaul of the drug safety monitoring system is essential and that comparable problems will arise in future unless and until that is done.

The present crisis has been looming for years (Teicher et al., 1992; Healy 2003; Medawar 1994, 1997). We believe the regulatory response was inadequate for three main and related reasons: over-reliance on misleading definitions and on data represented as ‘scientific’ and reliable, at times to the point of defying logic and common sense; a chronic tendency to marginalise, dismiss or ignore critical comment; and inadequate, sometimes incompetent analyses of the available data.

Because of the limited terms of the Review, we see no point in elaborating these contentions, though we would of course provide further information on request. Meanwhile, our evidence comes specifically to the Group’s terms:

1. "to consider the currently available evidence with regard to behavioural disorders, particularly suicidal behaviour, suicide attempt and suicide, and a possible causal association with SSRI's."

Central to our evidence is the attached paper (Medawar, Herxheimer et al., 2002) in which we review direct e-mail reports from patients about problems they experienced with paroxetine. The analysis drew from 1,374 emails sent to Panorama (BBC-TV, October 2002) plus 862 reports sent to the Social Audit website (January 2000 to October 2002).

We are also submitting to the review the summary of a longer paper (Medawar and Herxheimer, 2003, in press) in which we examined Anonymised Yellow Card reports (ASPPs) from health professionals of suicidal behaviour and of withdrawal reactions with paroxetine.

The analyses led us to these main conclusions:

[a] Reports from professionals and from patients described the same phenomena, but those from patients gave much richer descriptions of behaviours and feelings, and were often better at explaining the nature, meaning and consequences of adverse effects. Reports from patients have a key role to play in pharmacovigilance.

[b[ Analysis of anonymised Yellow Card reports revealed serious inadequacies both in the quality of data submitted and in the processing of reports, including lack of follow-up. Re-analysis of ASPPs underlines our concern that changes in the drug concentration in the body (both up and down) can precipitate potentially dangerous mental turmoil, a finding previously missed and in need of urgent consideration.

[c] Overall, the sparseness of the data in Yellow Card reports (on indication, severity, history of suicidal ideation) raises many questions that must be answered to arrive at a proper benefit-harm assessment. Until there are sufficient data the precautionary principle must be followed.

From these analyses and other data, we inferred that dosages are, for some patients, much too high; also that doctors too readily increase dosages (without considering instead the need to reduce them) when patients initially complain they feel worse.

Both doctors and patients need much clearer information about using these drugs, in particular about the significance of the dosage. We have been struck by the lack of coherent science-based advice on the relationship of dose and timing of doses to the various reported adverse effects. A closely related point is the lack of attention given to the major differences among individuals in the metabolism of paroxetine.

The evidence on dosage we have found most persuasive (Cohen, 2001) suggests that half of all patients in general clinical practice are prescribed excessive doses of SSRIs. Several related factors play a part in this (Herxheimer, 1991). In affective disorders, there is not much to distinguish response to placebo and active drug; both this and the pressure to differentiate products, has encouraged manufacturers to promote higher doses. Secondly, the strong preference for a convenient, ‘one size fits all’ dose strategy takes no account of marked variations in individual response, including genetic polymorphisms affecting drug metabolism. Thirdly, these factors have been reinforced by the strong regulatory tradition of approving the ‘optimum’ dosage – the one size that best fits all.

The following table (after Cohen, 2001; Wernicke et al, 1988) illustrates and underlines the point. Fluoxetine (Prozac) was approved and marketed at a starting dose of 20mg/day for everyone – the only tablet strength marketed in the UK. The table shows that more people respond to the 20mg dose than to a 5mg dose – but the choice of 20mg/day as the optimum dose also means that half of all users get four times the dose they need, exposing them to an increased risk of ADRs.

Fluoxetine dosage - mg/day

Response rate







Comparable data are urgently needed to explain the rationales for recommended doses for other SSRIs in all indications for prescription, and for the lack of emphasis on dose titration adjusted to individual response. The recommended doses for paroxetine are a case in point: depending on the indication, the recommended starting dose ranges from 10mg/day (panic disorder) to 20mg/day; the "usual" dose from 20mg/day to 40mg/day (OCD) and the maximum from 40mg/day (elderly) to 60mg/day (OCD).

We could add much more information. Our ability to do so will depend not least on the response we get to our (14 June) request to the MHRA for [a] anonymised Yellow Cards (ASPPs) classified under ‘Injury & Poisoning,’ ‘Suicide and parasuicide,’ and ‘drug withdrawal reaction’ for each of the following drugs: amitriptyline; citalopram, escitalopram, fluoxetine, imipramine, mirtazapine, sertraline and venlafaxine; and [b] a list in rank order, indicating in each case the numbers of reports received, of the top 20 drugs on the ADROIT database classified as suspected cases of [1] Non-accidental overdose; [2] Overdose NOS; [3] Suicidal ideation; [4] Suicide (accomplished); and [5] Suicide attempt.

We appreciate this is a rather voluminous request, requiring several hours work to fulfill, and have no reason to complain that these data have not yet been supplied. However, we ask the Review Team to endorse our request, even to the extent of proposing a date by which these data might be provided. Analyses of such data seem pre-requisite to fulfillment of the terms of this review.

[Note added on 25 July. The above request may now have been pre-empted by the announcement of a review of the Yellow Card system].

2. "to consider the currently available evidence on withdrawal reactions and possible dependence with SSRIs and any implications for the risk: benefit balance".

Current evidence relating to withdrawal and possible dependence on paroxetine and related antidepressants underlines the shortcomings of the MCA/CSM’s main statement about this problem (Price et al, 1996). This widely quoted paper concluded that withdrawal reactions with paroxetine were rare and predominantly mild, a badly erroneous conclusion based on misinterpretation of inadequate data. Social Audit has on several occasions, since 1998, formally requested the MCA/CSM to modify, withdraw or otherwise distance itself from these conclusions. It is time this was done.

The underlying reason for the problems many patients now face is the forced and unwarranted distinction between ‘withdrawal reactions’ and ‘dependence’. The writing has been on the wall for years:

"There is obviously some confusion about the concept of dependence ... The simplest definition of drug dependence given by WHO is 'a need for repeated doses of the drug to feel good or to avoid feeling bad' (WHO, Lexicon of alcohol and drug terms, 1994). When the patient needs to take repeated doses of the drug to avoid bad feelings caused by withdrawal reactions, the person is dependent on the drug. Those who have difficulty coming off the drug even with the help of tapered discontinuation should be regarded as dependent, unless a relapse into depression is the reason for their inability to stop the antidepressant medication…" (WHO, 1998)

Essentially the same point was emphasised at the September 2002 meeting of the WHO Expert Committee on Drug Dependence (WHO/ECDD): "Based on common sense, one would consider that an uncomfortable withdrawal syndrome will increase the risk of dependence" and "Unless studies indicate the contrary, it is highly unlikely that a serious withdrawal syndrome will not lead to drug dependence."

In the report adopted by WHO/ECDD, it was noted that, "In terms of the number of drug dependence reports received (by the Uppsala Monitoring Centre) up to mid-2002, three SSRIs are among the top 30 drugs for which drug dependence has ever been reported to the UMC database." The following table (WHO 2002) highlights how confusion in terminology has dominated perceptions of problems related to drug discontinuation. The last column (dd/ws) suggests to us that the term ‘dependence’ has been most used where withdrawal symptoms are least prominent; this can be discussed.

Drug name

Withdrawal syndrome reports (ws)

Drug dependence reports (dd)

Ratio (%)



































Currently available evidence, including copious testimony from patients, suggests that:

[a] a significant number of patients experience withdrawal symptoms so severe as to make it very hard if not impossible to stop taking SSRIs;

[b] many patients are either not warned of such problems, or have been positively reassured by doctors that no problems with dependence can arise;

[c] doctors often misinterpret withdrawal reactions as ‘relapse’; and

[d] not only abrupt but also gradual withdrawal appears to be associated with potentially dangerous psychiatric adverse effects.

In connection with the last problem [d], the Review team should know that the MHRA apparently deleted from the anonymised Yellow Card data they provided to us, the database field itemising the ‘Other Reactions’ experienced by patients who suffered during withdrawal. We specifically requested these and other data on 11 May (and since), but they have not been supplied.

It will be clear from the above that the scale of the withdrawal/dependence problem is unknown and needs to be established through systematic review. Patients now urgently need advice on how to accomplish effective withdrawal (other than by switching from one drug to another in the same general class), but reliable advice cannot be given without systematic review of what is known. This is not however an argument for delaying interim advice.

We emphasise that the measure of the dependence problem is not so much that some people experience withdrawal symptoms, but that many others continue with treatment to avoid them. We believe that the present position of many SSRIs users is comparable to that of benzodiazepine users, one generation back. The problem is largely hidden: "The probable reason is that patients abort these reactions early on because they think their original symptoms are returning, and they get back on the drug. So we rarely see the full-blown picture." (Hollister, 1977)

Two further aspects of withdrawal reactions affect the benefit-harm relationship. One is that many patients seek to get off these drugs not simply because they feel they no longer need them, but because they experience a range of nasty and often very distressing adverse effects – notably loss of libido, weight gain, mood disorders and impairment of concentration. This underlines the urgency of the need for effective advice and procedures for withdrawal.

The second aspect does not directly concern patient suffering, but is arguably the most important influence on the benefit-harm relationship of SSRIs. It is that the alleged long-term effectiveness of these drugs was established in clinical trials that discounted the existence of withdrawal effects. Briefly, they ‘proved’ long-term effectiveness by withdrawing established SSRI users from their drugs – and then counting all cases of withdrawal reactions as cases of relapse.

Since 1997, Social Audit has repeatedly drawn the attention of the MCA/CSM to this fatal flaw in the trials it approved – in particular in the trial they relied on to prove the claim of long-term efficacy of paroxetine. This trial exemplifies the problem: it was conducted jointly by a (then) CSM member and an employee of SmithKline Beecham. See the description of methods and critique of results in Medawar (1997) and in correspondence between Social Audit and the then Chairman of the CSM (Rawlins, 1998).


3. "to consider the adequacy of the current product information for SSRI's and to make proposals to strengthen this if necessary"

Substantial changes are needed to improve the current product information to both professionals and the public, but it is far beyond us to make proposals line by line. Our general view is that:

[a] the wide range of indications – sufficient to justify prescribing in so many circumstances – invites the inference that the benefit-harm relationship is a constant. Severity rather than type of symptoms should be the main determinant in the decision to prescribe or not, particularly for non-specialists.

[b] the precautionary principle should rule, in place of the prevailing tendency to assume that no ‘scientific’ evidence of risk amounts to evidence of no risk;

[c] all information directed to patients should be rigorously copy tested and shown to be generally comprehensible; information should also be ‘cascaded’ by importance and comprehensibility, to take account of the substantial minority of users who are not native English speakers or who may be less functionally literate; and

[d] product information should be basically the same for all SSRIs and related drugs. Only any clearly demonstrated qualitative differences, and quantitative and kinetic differences, as for example between paroxetine and fluoxetine, should be described and explained.

[e] these considerations lead us to ask by whom the use of SSRIs will be monitored from now on, and how.

We do, however, wish to comment specifically on one statement made in the recently (June 2003) revised SPC and PIL for Seroxat. This seems to us to epitomise the gulf between standard, approved practice in the formulation of warnings, and honest and straightforward communication. This example concerns the incidence and severity of withdrawal reactions: "In recent clinical trials which used a taper phase to discontinue treatment with paroxetine … approximately 25% of patients on paroxetine and approximately 15% of patients on placebo experienced such symptoms. That is, approximately 10% of patients in the paroxetine group experienced symptoms potentially attributable to the withdrawal of therapy when corrected for placebo."

We comment as follows:

[a] We find it astonishing that, even now, the regulators approve a statement which implies an equivalence in severity of symptoms experienced by patients withdrawn from either active drug or placebo. We realise that this is standard practice but ask the Review Team to accept it is also false and misleading – for reasons explained in Social Audit’s earlier critical review (Medawar, 1997)

[b] Less than two years ago, the manufacturers of Seroxat were producing ‘scientific’ evidence and expert testimony to defend their contention that withdrawal reactions were "very rare", with an incidence of the order of 1 in 2,000. The MCA/CSM not only accepted this, they had produced their own ‘scientific’ data to ‘prove’ much the same (Price et al, 1996). The point we emphasise here is that effective communication of the risks, from this point on, fundamentally depends on correcting the misleading information handed down over years. Something analogous to "corrective advertising" is now called for; warnings in small print will not be enough.

[c] Current evidence on the incidence of withdrawal reactions highlights the dodginess of claims made for the lasting effectiveness of these drugs (see above), and the need for re-evaluation. The urgent need for this is underlined by the fact that treatment guidelines were substantially revised in the early 1990s (leading to recommendations for higher doses and longer periods of treatment), largely on the basis of belief in the long-term effectiveness of these drugs. In our view, this is seriously in doubt and the risks involved seem great, because severity of dependence is likely to depend in part on dosage and duration of treatment.


4 and 5. "to consider the need for wider communication on the safety of SSRI’s" …. "and to advise the Committee on Safety of Medicines"

Wider communication is essential to rebuild trust in medicines information. This includes openness and full disclosure of the underlying evidence and its limitations. ASPPs should be made readily available and arguably should posted on the Internet, to encourage scrutiny and analysis.

The Committee on Safety of Medicines needs advice:

[a] on the implementation of a new approach to the categorisation and analysis of ADR reports, specifically giving more attention to patterns of events, dosage, timing, and follow up; and

[b] on the need for dose response data for all categories of effects, beneficial and harmful, and their publication. Prescribers and users need them.

Under severe pressure of time, we have referred only to a fraction of the relevant evidence accumulated over the years; see

We are convinced that nothing less than a full public enquiry is called for; we believe this should not be organised by the regulators themselves.

In the interests of public safety, we believe interim recommendations and warnings are urgently needed, to emphasise that changes in drug concentration in the body (both up and down) may precipitate potentially dangerous adverse reactions.



Cohen J.S. Overdose – The case against the drug companies, New York: Jeremy P Tarcher/Putnam, 2001.

Healy D. Lines of evidence on the risks of suicide with selective serotonin reuptake inhibitors, Psychotherapy and Psychosomatics, 2003; 72, 71-79.

Herxheimer A. How much drug in the tablet? Lancet, 1991; 337: 346-348.

Herxheimer A, Medawar C. Listening to paroxetine – making sense of Yellow Cards and user reports of ADRs, Submitted for publication, 2003. (Copy attached)

Hollister LE. Valium: a discussion of current issues. Psychosomatics, 1977, 18 (1), 44-58.

Medawar C. The Antidepressant Web – Marketing depression and making medicines work, Int. J. Risk & Safety in Medicine 1997;10: 75-126. (Copy attached)

See also: (1998-2003).

Medawar C, Herxheimer A, Bell A, Jofre S. Paroxetine, Panorama and user reporting of ADRs: consumer intelligence matters in clinical practice and post-marketing drug surveillance, Int. J. Risk & Safety in Medicine 2002; 15: 161-169.

Price J.S., Waller P.C., Wood S.M. (Medicines Control Agency), Mackay A.V.P. (Committee on Safety of Medicines). A comparison of the post-marketing safety of four selective serotonin reuptake inhibitors including the investigation of symptoms occurring on withdrawal, Br. J. Clin. Pharmacol. 1996; 42: 757-763.

Rawlins M. Letter to Social Audit Ltd, 8 May 1998, and reply from Social Audit, 10 June 1998. (Copy attached)

Teicher M.H., Glod C.A., Cole J.O. Antidepressant drugs and the emergence of suicidal tendencies, Drug Safety 1993; 8, (3): 186-212.

Wernicke JF, Dunlop SR, Dornseif BE, Bosomworth JC, Humbert M. Low-dose fluoxetine therapy for depression. Psychopharmacology Bulletin 1988: 24: 183-188.

World Health Organisation, Selective serotonin re-uptake inhibitors and withdrawal reactions, WHO Drug Information 1998; 12 (3): 136-138.

World Health Organisation. Terminology in reporting abuse-related adverse drug reactions, draft report as adopted by the WHO Expert Committee on Drug Dependence, 33rd meeting, Geneva, 20 September 2002.


Contents page
What's New?