In my letter of 14 May, I said I would come back to you on the Montgomery & Dunbar paper. You say it provides good evidence and I say it does not. The issue here is not only whether the long-term effectiveness of antidepressants is more real or illusory, but whether and to what extent the CSM and MCA require evidence put before them to be scientifically sound.

I cannot at this stage know whether the issue of possible conflicts of interests is also relevant, as I have not received the extracts from the Register of members' interests I requested the CSM to disclose in my letter to you of 12 March. Three months on, I have yet to be informed even of any reason why the Committee thinks it necessary to keep this information secret. This is one reason why I am now sending a copy of this letter to the Committee on Standards in Public Life - albeit with a covering letter emphasising they should take no action which would in any way impede this correspondence with the CSM/MCA.

Another reason is that you give no indication of any uncertainty in your mind about the reliability of this study, notwithstanding the concerns I have been trying to spell out to the authorities over the past year. I feel deeply concerned that you have now thrown the CSM's weight fully behind the science in this paper. But since you have done so, I believe the onus is now very much on you to offer good evidence that it is safe to rely on this study, and that its principal conclusion is not misleading.

In the remainder of this letter I have set out the main reasons for asking you now to provide at least a skeleton argument to explain the reasons for your confidence in this study. I have also explained in outline why I believe you have yet to provide credible evidence even of a backbone argument and I am now writing to invite you to do so.

To try to explain my position clearly, I have raised questions and sketched comments under each of the four sentences in the full statement in your letter. These points are not exhaustive and I may later elaborate on them, with comments in the electronic margins of the version of this letter to be published on our Internet site. The site is now getting over 2,000 visits per month, so there is clearly some interest in the issues. I would expect some, including both the authors of this paper and the manufacturers of paroxetine, to want to know more about the arguments you put forward.

I am copying this letter also to the MCA (Mr Alder) to underline that the questions in this letter should all be treated as formal Code requests. This also means that the 20-working-day clock (target response time) has begun to tick. If it were to tick a little faster, with the Committee (CSPL) now in the frame, that would seem no bad thing. This is meant to be an outer limit, but has relentlessly been exceeded.

In your letter you wrote this:

"I do not accept your opinion that the conclusions of the Montgomery & Dunbar paper are invalid because the authors discounted the possibility of withdrawal symptoms, and because they would have therefore interpreted withdrawal symptoms as relapse. In the Methods section it is stated that the principal outcome was withdrawal from the study because of the reappearance of depression (as support by various criteria). You may be unaware that, in most cases symptoms associated with paroxetine withdrawal are clearly distinguishable from those of the underlying depressive illness. The study was a randomised double-blind trial which provides good evidence for the efficacy of paroxetine in preventing relapse and recurrence of depression over the period of the study."

On this basis, the CSM/MCA have also approved the data sheet claim that "Long-term treatment with Seroxat has shown that antidepressant efficacy is maintained for periods of at least one year". I assume this study provides pretty much the sole basis for this claim, since any other study which reached such optimistic conclusions would surely have been published. However, please will you let me know if, at the time of licensing, the CSM/MCA had any other significant evidence to support this claim?

The central conundrum is this: when paroxetine was licensed, the MCA/CSM required a recommendation for gradual drug withdrawal - yet you seem happy to rely on a study in which all patients were abruptly withdrawn from paroxetine. I cannot begin to understand why you would expect the average prescriber to look out for withdrawal symptoms, and try to taper the doses of antidepressants given to patients - but then waive this requirement for the study investigators, one of them a member of the CSM. It does not help that this man had extensive interests with pharmaceutical companies and that you will not explain his role in the approval of this drug.

The abrupt withdrawal used in this study would have increased the incidence and severity of withdrawal reactions. Therefore I cannot understand why you conclude that it doesn't matter that the authors didn't even note the possibility of withdrawal symptoms, let alone record any attempt (or criteria used) to distinguish them from the symptoms of relapse?

This seems all the more incredible, because you seek to persuade me that symptoms of antidepressant withdrawal and depressive relapse are usually clearly distinguishable. If this were so, why is that it the recommendation for gradual withdrawal first appeared some 30 years after the introduction of antidepressants? And why do many doctors (and perhaps most GPs) still not recognise that antidepressant withdrawal reactions even exist? If the difference were anything like as obvious evidence as you suggest, how could this possibly be?

1. "I do not accept your opinion that the conclusions of the Montgomery & Dunbar paper are invalid because the authors discounted the possibility of withdrawal symptoms, and because they would have therefore interpreted withdrawal symptoms as relapse."

This statement clearly implies that the CSM/MCA would today consider such a study gave acceptable evidence of long-term efficacy. Please will you confirm this?

Please will you explain specifically the basis on which you can accept evidence from a study that reported a zero incidence of withdrawal problems, in 67 patients, given that all published studies "specifically designed to assess discontinuation symptoms consistently report high rates" (Haddad, 1997). None of the five studies relating to paroxetine reported an incidence of withdrawal reactions below 30%, though most involved gradual withdrawal after treatment lasting for much less than one year:

I accept that these studies have limitations - though nothing to compare with those in the MCA/CSM's main assessment. Your conclusion that withdrawal symptoms are rare has been based mainly, and quite inappropriately, on low levels of spontaneous reporting. The incidence of 0.03 reports per 100 paroxetine prescriptions suggests an estimate of the order of 1,000-times lower than the estimates made above. Your confidence in the Montgomery & Dunbar study is no doubt fortified by this.

2. "In the Methods section it is stated that the principal outcome was withdrawal from the study because of the reappearance of depression (as support by various criteria)."

This point seems essentially irrelevant, if not misleading, because it is premised on your belief that there is little scope for confusing paroxetine withdrawal symptoms with a relapse of depression. See below. You are quite isolated in this view, even from the opinions of Haddad, Lejoyeux and Young, to whose work you specifically directed me, in your letter of 8 May.

Both Haddad and Young have recently reported surveys in which high proportions of prescribers have been found to be unaware even of the existence of antidepressant withdrawal reactions. Haddad was involved in a survey of 200 psychiatrists of different nationalities which found that about half were unaware that SSRIs could cause withdrawal symptoms (Gillespie et al, 1995). Young & Currie (1997) reported that, among respondents to a survey in North East England, 14/50 (28%) psychiatrists and 37/53 GPs (70%) were unaware that patients might experience antidepressant withdrawal symptoms. These findings underline the risks warned of by Lejoyeux et al (1996):

"It is possible that many patients undergoing antidepressant withdrawal and classified as having a depressive relapse are, in fact, experiencing unrecognised withdrawal symptoms" … "The antidepressant withdrawal syndrome is a frequent and often under-diagnosed phenomenon" … "Withdrawal from antidepressants can produce transient changes in mood, affect, appetite and sleep, which might be incorrectly diagnosed as indicating that the patient is having a depressive relapse".

As I mentioned in my letter of 14 May, Haddad, Lejoyeux and Young were also recently involved in a small, closed, sponsored symposium on SSRI withdrawal problems. It is also clear from the published papers and comments which emerged from this that there was general concern about the risks you seem to discount. For example:

The SSRI withdrawal syndrome "can also be mistaken for new physical illness or the return of the original depression. Misdiagnosing symptoms may lead to costly, unnecessary testing and treatment" (Schatzberg, 1997)

"Our purpose is to encourage clinicians to think about the possibility of discontinuation symptoms and not to misdiagnose patients who come in to report fatigue as having a recurrence of depression" (Zajecka, 1997)

"My colleagues and I see the flu-like symptoms frequently during SSRI discontinuation but seldom hear about the intensification of an affective disorder, which I think is common but seldom reported because patients are likely to attribute it to the absence of treatment rather than to discontinuation". (Rosenbaum, 1997)

These comments echo many similar warnings in the literature. In my paper, I draw particular attention to the advice given in two major reviews: "awareness of the possibility (of withdrawal reactions) helps to avoid misinterpreting new symptoms after withdrawal as evidence of relapse" (Drug & Ther Bull, 1986); and "The withdrawal of antidepressants can produce changes in mood appetite and sleep that are apt to be incorrectly misinterpreted as indicating a depressive relapse" (Dilsaver 1989)

Your view that the symptoms of withdrawal and relapse are usually clearly distinguishable is not supported by the evidence, much of which long precedes the introduction of data sheet advice on the need for gradual drug withdrawal. Indeed, the quality of some of the advice given many years ago seems much more helpful than the kind of data sheet warnings published today. Would you not agree?

"The withdrawal syndrome complicates the evaluation of patients after drug discontinuation since both patients and physicians often interpret the onset of symptoms as an upsurge of 'anxiety' related to incipient relapse, and resume treatment with the gratifying subsidence of the 'anxiety'. This may cause both patients and physicians to overvalue the importance of the medication to the patient's stability" (J.C. Kramer, D.F. Klein, M. Fink, Withdrawal symptoms following discontinuation of imipramine therapy, Am. J Psychiatry, 1961, 118, 549-550).

"Side effects - Miscellaneous: Occasional withdrawal symptoms following abrupt discontinuation of treatment: nausea, vomiting, abdominal pain, diarrhoea, insomnia, headache, nervousness and anxiety" (Imipramine/Tofranil data sheet, 1998/99)

3. "You may be unaware that, in most cases symptoms associated with paroxetine withdrawal are clearly distinguishable from those of the underlying depressive illness."

Your view that the Montgomery & Dunbar paper provides good evidence for the long-term efficacy of paroxetine pretty much stands or falls on this point. However, you seem to be playing with words here, inviting me to believe that if, in theory, the symptoms of relapse and withdrawal can be clearly distinguished from each other, in practice, they usually are. This seems to me potentially very misleading.

• What matters is not whether symptoms of withdrawal "are clearly distinguishable" from those of relapse - but whether, in practice, the distinction between the two is actually made. Do you have any evidence to suggest that prescribers either have in the past, or do now, usually clearly distinguish between the two?