From: Newbower, Ronald S.
Date: 31 December 1998  14:18
Re: Response to Charles Medawar

Dear Mr. Medawar;

I appreciate your concern, and your explication.

I have reviewed your letter and forwarded it to Drs. Rosenbaum and Fava for comments. For convenience, they have replied by inserting their comments in italicized form within a copy of your letter, following each point raised. That commentary is attached

It is my judgement that your letter, its points, and their replies, are now fully within the domain of scientific discourse and interaction, with the areas of agreement and of disagreement that characterizes such discourse. However, I see no further issues for which I or the institution need serve as an intermediary or as a judge. The areas of disagreement which have now been laid out do not have the characteristics of issues which invoke institutional policy. Please understand, I am not judging the validity of scientific claims or counterclaims, nor is that my role as an institutional official. I am distinguishing those legitimate areas of debate from the kinds of institutional issues alluded to in your earlier correspondance, and which we have investigated in some detail.

As you stated in your reply to me, we have put substantial effort into a careful assessment of the issues of potential violations of MGH policy, and have concluded that there is no evidence in front of us of that sort of concern. This in no way diminishes the right you have to disagree with the scientific findings. It is simply to state our conclusion that they were arrived at in a manner that appears to be in compliance with the high standards of the institution, and that disagreements about those findings should therefore be handled in the normal fashion employed for scientific disagreements --- interaction between scientists or observers of science, either directly or through the scientific literature.

I am glad that we have been able to help as an institution by being responsive to your concerns.

Thank you for your understanding.

Yours truly,

Ronald S. Newbower, Ph.D.
Sr. Vice President for Research and Technology
Massachusetts General Hospital
Attachment: copy of your letter, with the authors' reponses
interspersed in the text, and marked in bold and italics.

Dear Drs Rosenbaum and Fava,

Many thanks for your comments on our website review of the four Lilly-sponsored trials of fluoxetine. They were of much interest and I appreciate the time and trouble you took.


I entirely accept that your claim to authorship of the paper in Biological Psychiatry is well established - albeit for reasons which make even more questionable the initial publication of your work in the name of two Lilly employees, one year in advance of your definitive report.

There has been only one publication; we do not consider a 12 minute oral presentation on an interim report a publication, even if the meeting syllabus contained an abstract of that presentation.

As their names were not appended to your final report, I would still question their qualifications as authors, within the meaning of NEJM guidelines, also the degree of control you have over work published under your names.

Your statement is incorrect, as the APA abstract (page 118 of the New Research Abstract book) included only one co-author that was not included in the final Biological Psychiatry manuscript. The exclusion of that author (Michael Wilson, M.S.) is simply due to the fact that, although he had helped putting together the abstract submission, he was assigned to other projects within Eli Lilly and Company (he now works with the osteoporosis team) and he had no input on the development of the final manuscript published in Biological Psychiatry. The inclusion of the other Eli Lilly and Company employees as co-authors of the Biological Psychiatry publication simply reflects their involvement in the coordination of the study and in the writing of the manuscript (we had several conference calls between Boston and Indianapolis to refine the manuscript). We can assure you that from the beginning of the process that led to the publication of the report we were free to say or add whatever we felt appropriate as we were fully in charge of the publication.

I had not appreciated that the trial design was so much of your making and, had I done so, I would have acknowledged more generously your acumen in approaching Eli Lilly for sponsorship. Given your hypothesis that SSRIs with shorter half-lives would be associated with more severe and frequent withdrawal symptoms, it does seem surprising the company was not immediately tempted by your trial design - which also clearly reflects your view that there was little or no risk of withdrawal symptoms with fluoxetine. But then why did the people at Lilly not see this as a gift?

Again, despite your belief otherwise, research means taking a risk of testing, rather than just asserting, one's beliefs.

You wrote that the company at first resisted, "because of uncertainty that the SSRIs would differentiate" - but they would certainly have been aware of the potential advantage of the long half-life, and must have noted the difference between paroxetine and fluoxetine withdrawal symptoms. I wonder if perhaps the company was worried because of what had happened with the benzodiazepines (BDZs), shortly before - and of course I also wonder if this did not concern you.

We were certainly not concerned with this issue as we were proposing to do the study. Dr. Rosenbaum was and is an internationally recognized authority on benzodiazepine therapy and we can assure you there was no impression of correspondance between the two issues.

Dependence on diazepam (c.f. fluoxetine) went unnoticed for over 20 years, and was ultimately only inferred because of its cross-tolerance with shorter-acting BDZs, such as lorazepam (c.f. paroxetine) or triazolam (c.f. venlafaxine). For Lilly to draw attention to withdrawal symptoms in another SSRI (which fluoxetine would quell) might have seemed risky. Acknowledging your point (17) that withdrawal phenomena are a commonplace with centrally-acting drugs, it may be that the experts at Eli Lilly felt it might be tempting providence to press the point - at least before the new definition of "dependence" had sunk in (DSM-IV, 1994).

Until then, the company might well have supposed that SSRIs - including fluoxetine - would turn out to be every bit as much drugs of dependence as BDZs were defined to be, until the 1990s. The reasoning is explained in The Antidepressant Web, (section 3.5), but the essential point is this:

"Historically, long-term, high-dose, physiological dependence has been called addiction, a term that implies recreational use. In recent years, however, it has become apparent that physiological adaptation develops and discontinuance symptoms can appear after regular daily therapeutic dose administration ... in some cases after a few days or weeks of administration. Since therapeutic prescribing is clearly not recreational abuse, the term dependence is preferred to addiction, and the abstinence syndrome is called a discontinuance syndrome" (American Psychiatric Association, Task Force on Benzodiazepine Dependency. Benzodiazepine Dependence, Toxicity, and Abuse. Washington DC: APA, 1990).

With respect to the risk of a sedative-hypnotic "abstinence" syndrome or similarities between antidepressant use and BZD dependence, we believe your position is rather idiosyncratic and has not been shared by clinical researchers in the field, whether academic, federal or industry based. So, it was not likely the company's concern. As we described in our prior response, consequences of abrupt discontinuation of many classes of medicines are commonplace and generally call for taper when discontinuing. This does not imply dose escalation, drug seeking, or impairment or other issues associated with dependence on drugs of potential abuse.

True, I had not realised that what I identified as trials 2, 3 and 4 were researched so long before your own study (No 1). It would have been helpful if this had been explained in the published papers but, yes, it does explain why the DESS scale you developed was used in your study alone. However, this doesn't alter the essential proposition - that your conclusions, published in 1998, were based on two very different research methodologies. A so-called "gold standard" methodology was used to evaluate withdrawal reactions to paroxetine and sertraline, but not fluoxetine.

Yes, prospective hypothesis testing is superior to testing hypotheses with existing data-sets, but the latter can still offer data of interest to the field.

I will not labour the point (unless you request it) but you make several comments that indicate you either have not read or understood the basis of my critique. Instead, let us come to the nub of the differences between us - your assertions at point 10: "There was not evidence of an early fluoxetine discontinuation syndrome in our published study …(in Biological Psychiatry) ... Patients on fluoxetine did not have an increase in discontinuation emergent signs and symptoms …".

This is quite some denial. It goes far beyond what you claim in the study itself - eg "the incidence of an SSRI 'discontinuation syndrome' observed in fluoxetine-treated patients was significantly lower than the pooled incidence for sertraline and paroxetine-treated patients (14%, 60%, 66% respectively; p < .001)".

You are selective in your citation from the study, and fail to note that there was no increase in DESS events in the fluoxetine group overall after discontinuation. Within those data, some had an increase in DESS symptoms (14% on fluoxetine vs 60% and 66% on comparators); you are eager to conclude the smaller number is evidence of a discontinuation syndrome rather than just week to week variation (noise in the administration of the assessment instruments) in which patients have certain symptoms. We don't believe our data supports your assertion. This much variation is seen in those who simply remain on treatment.

Clearly, there was evidence of a fluoxetine withdrawal syndrome - but only in a few patients - as one might expect with such a trial design. But do you honestly doubt that you would have seen increased numbers of patients having problems with fluoxetine withdrawal, if the study had been extended in time?

We suspect that if fluoxetine was discontinued permanently, we might pick up a signal of mild dizziness in 5% or 6% of patients at 4-6 weeks..

In the circumstances, it seems neither becoming nor persuasive to argue for the superiority of fluoxetine by reference to baseline calculations of mean numbers of DESS reported by different groups. As the trial design positively invites lower numbers of DESS reports with fluoxetine, the thing to focus on would be the characteristics (and timing) of withdrawal events, where seen. Table 2 does this, and it shows beyond doubt the predominance of depressive symptoms, immediately following drug withdrawal.

One problem with your argument is your assumption that any symptom reported is discontinuation related, and does not account for the presence of these symptoms in those who do not discontinue.

Nevertheless, in your comments (point two) - as in the abstract of your paper - the greater emphasis is on "symptoms such as nausea, dizziness, fatigue, myalgia which would typically be attributed to the 'flu' or to some other medical illness". This focus seems inappropriate, and likely to lead to confusion - not least because the very attenuated period of withdrawal from fluoxetine would mean that symptoms of acute distress were less pronounced than with SSRIs with shorter half-lives. They would be, wouldn't they?

If we understand your point here, you are restating our hypothesis.

But what is the rationale for concluding that instances of depression following drug withdrawal amount to relapse, nothing more? You say "clinicians know" these things, and that it would be wrong to infer withdrawal problems from "an increase in depression rating at one point in time in a small number of patients in treatment for depression". Again, you point to small numbers - and again I refer to the trial design that makes the numbers small.

Your paper also fails to explain satisfactorily why large numbers of people get depressed during drug withdrawal. From the detail of the paper, it is clear that depression is a major "discontinuation symptom", but you otherwise point only in the direction of "re-emergence of depressive symptoms" or "relapse". Would not the timing of events point to anything but this interpretation?

From the above, we take your point to be that depression is not a recurrent disorder, or at least that for those who have been on antidepressants, the return of depression after stopping treatment represents discontinuation rather than relapse. To test your hypothesis, we should randomize patients to drug or placebo treatment, then slowly taper treatment and follow subjects over time to compare relapse rates.  

" … withdrawal syndromes developing within a few days of withdrawal cannot be attributed to a relapse of the disorder for which the antidepressant was first prescribed, because this would take several weeks to appear … Awareness of the possibility helps to avoid misinterpreting new symptoms after withdrawal as evidence of relapse" (DTB (Anon), Problems when withdrawing antidepressives, Drug & Ther. Bull., 1986, 21 April, 24, 29-30).

"The withdrawal of antidepressants can produce changes in mood, appetite and sleep that are apt to be incorrectly misinterpreted as indicating a depressive relapse ... The probability of depressive relapse is low in the days and weeks after the discontinuation of antidepressants, and the cumulative probability of relapse increases as a function of time when the patient is medication free ... In contrast the frequency of antidepressant withdrawal symptoms is high in the first 2 to 14 days following the last dose." (Dilsaver S.C. , Antidepressant withdrawal syndromes: phenomenology and pathophysiology, Acta Psychiatr. Scand., 1989, 79, 113-117).

Perhaps your colleagues at MGH would find it unexceptional, but it does seem to me extraordinary that the abstract of your paper gives barely a hint that about one-third of the patients on paroxetine and sertraline got depressed within a week of withdrawal. Given cross-tolerance between fluoxetine and these drugs, why does your paper not consider this crucial point? Again, the answer seems to be to do with the trial design. Very simply, the protocol stipulated that anyone who experienced a significant "breakthrough of depressive symptoms" should be counted as someone who did better on the drug:

"To provide additional clinical context, a depressive relapse was defined as an increase of 8 points or more in the HDRS28 score and a total score of 16 or higher." (emphasis added)

If this is really depressive relapse, then we have a novel treatment for depression that works within hours of starting medication, since these new symptoms melt with restoration of medication.

This whole issue has nothing to do with me wanting to see a level playing field; I'm simply concerned to see the risk defined for what it is. For reasons given, I believe there is a real risk of dependence, and lack of long-term effectiveness; I'm sure we agree this deserves to be honestly discussed. I'm especially concerned about fluoxetine, because I think the risk is better disguised than with paroxetine and other competing drugs. I can see the advantage of attenuated withdrawal, but I also see some additional risks.

It is unclear to us what are the additional risks that you see. Furthermore, if fluoxetine was in fact associated with a significant risk for discontinuation-emergent symptoms, we would expect to see many more clinicans observe this and report it in the literature, as there are countless reports with shorter acting agents.

In short, I believe your comments substantially reinforce the thrust of my argument, if not my apparent credibility as a researcher of finer points. Perhaps you would like to reconsider your position?

As is evident from our reply, your beliefs do not alter our view of the data.

In turn, I have revised my preliminary view, which was not to pursue this matter through Dr Newbower's office at MGH. In view of the risks to public health, and the basis on which they arise, it seems appropriate I should. I shall welcome your comments and look forward to hearing from you.

We are not sure we follow why your concern about public health would lead you to insert an extra step in our receiving your comments, but however you wish to contact us is okay.

Yours sincerely,
Charles Medawar
and, Sincerely yours,
Jerrold Rosenbaum, M.D. and Maurizio Fava, M.D.


Contents page
What's New?