SEROXAT/PAROXETINE
‘PLACEBO SUICIDES’
CORRESPONDENCE WITH PROFESSORS ASHBY AND GUNNELL -
MEMBERS OF THE CSM EXPERT WORKING GROUP ON SSRIs
19 February 2006
Dear Professor Gunnell,
Suicides in clinical trials of paroxetine –v- placebo
I am enclosing a copy of the Freedom of Information request I made recently to the MHRA, together with their recent response and my reply. I have sent the same papers to Professor Ashby.
You will see that I have strong reservations about the results (1 suicide of paroxetine, 3 on placebo) reported both by the Expert Working Group on SSRIs and in your BMJ paper published on 19 February 2005.
I would welcome any thoughts you had on the queries I have raised and specifically I would appreciate comment on your observation: "We did not have access to individual patients' data." As you wrote, this limited the extent of your analysis. It would therefore help if you could let me know if you requested those data from the MHRA, and/or if you recommended that the MHRA obtain them, and what response you received?
Thank you for considering this request and I look forward to hearing from you.
Sincerely,
Charles Medawar
========================================================
28 February 2006
Dear Dr Medawar,
Thanks for you letter about our BMJ article and for enclosing copies of
your correspondence with the MHRA.
You asked for clarification about our comment in the paper that "We could
not conduct a meta-analysis based on individual patient data." We were
raising a methodological concern here (for further details see: Egger et al
Eds Systematic reviews in health care. BMJ Books 2001).
Our evidence synthesis was based on summary evidence for each trial (
overall numbers of patients in each arm of the trial / total number of
adverse events in each arm). Some meta-analyses - most often those
involving a smaller number of trials and where electronic records of all
the patients who were recruited to trials are still held by the triallists
- go a step further than we did and construct databases with one record for
each patient included in the trial. Such datasets are then combined across
trials - so in the case of SSRI trials with suicide as the outcome, these
would contain separate records for over 40,000 patients.
Databases such as these would include additional data on patient
characteristics such as age / gender / depression severity at recruitment /
follow-up time etc. and this allows further pooled analyses to look at
issues such as (a) time to event; (b) patient subgroups (e.g. gender
specific effects) with greater power.
You will note that in view of our observation that all three placebo
suicides in the paroxetine trials occurred after the end of the trial we
assessed the effect of excluding these deaths from the analysis. This
sensitivity analysis did not alter our overall conclusions - the number of
suicides in the trials included in our analysis were too few to give us
sufficient power to detect clinically relevant beneficial or adverse
effects.
Yours sincerely
David Gunnell and Deborah Ashby
---------------------
David Gunnell,
Department of Social Medicine
University of Bristol
Whiteladies Road
Bristol BS8 2PR
===============================================
3 March 2006Dear Professors Gunnell and Ashby,
Suicides in clinical trials of paroxetine –v- placebo
Thank you for your email reply to my letter.
I regret that you did not answer the one specific question I asked: did you ask the MHRA to obtain from the manufacturers the raw data relating to the four cases of suicide in the paroxetine versus placebo trials and, if so, what was their response? However, from your reply, and from your seemingly irrelevant allusion to the problems that would be involved in checking data on 40,000 records, I infer you made no such request, in spite of the several obvious anomalies to which I referred.
The reason I suggested the need to request the raw data, instead of relying on the manufacturer’s summaries clearly had very little to do with the need to obtain "additional data on patient characteristics such as age / gender / depression severity at recruitment / follow-up time etc". The point was to establish that the manufacturers’ summaries of data on which you and the MHRA have relied presented a true fair of events.
I was rather dismayed too that you expressed no interest or concern in the issues I raised in my FOI request to the MHRA – from which you might well have inferred that your professional focus on quantity rather than quality of data was badly misplaced. Time will tell, but I have no doubt that the MHRA and EWG were making a grave mistake in relying on the manufacturer’s summaries of data. If this proves to be the case, I hope you will then feel able to modify your paper in the BMJ
I did note that your paper included an assessment of risk when the placebo suicides from the paroxetine trials were discounted. I agree that, when these three cases are excluded, the aggregated results for the five SSRIs in your analysis do not reach statistical significance; they are merely transformed. However, your analysis excluded data for fluoxetine and, if you had included venlafaxine and mirtazepine in your analysis, I cannot imagine you would have arrived at similar conclusions for this class of drugs. (The exclusions of these drugs can hardly be justified on pharmacological grounds. The distinction between SSRI and SRNI overwhelmingly reflects market posture and seems no more justified that the distinction you made between citalopram and escitalopram).
I hope you will let me know if you do not agree with the summary in the table below. These data suggest an RR of 2.66 (90% CI: 1.07, 6.63 and 95% CI: 0.90, 7.90).
Trial drug |
Suicide on active drug |
Suicide on placebo |
| Citalopram | 1/1320 |
1/622 |
| Escitalopram | 0/2648 |
1/2088 |
| Fluvoxamine | 2/4186 |
2/3396 |
| Sertraline | 4/7169 |
0/5108 |
| Paroxetine | 1/8481 |
0/5808 |
| Venlafaxine | 4/6153 |
0/2962 |
| Mirtazepine | 5/2618 |
0/388 |
| TOTALS | 17/32,575 |
4/20,372 |
Yours sincerely,
(Mr) Charles Medawar
==================================================
17 March 2006
Dear Mr Medawar,
We were rather surprised by your recent letter. You say you regretted that
we did not respond to the one specific question you asked. We think you
have misinterpreted our reply.
In your letter of the 19th February you wrote:
"I would welcome any thoughts you had on the queries I have raised and
specifically I would appreciate comment on your observation: "We did not
have access to individual patients' data." As you wrote, this limited the
extent of your analysis. It would therefore help if you could let me know
if you requested those data from the MHRA, and/or if you recommended that
the MHRA obtain them, and what response you received?"
We interpreted your question as seeking clarification about the meaning of
our comment concerning access to individual patient data. Our response to
you clarified that we were raising a methodological issue when we raised
this point, one that is common to many meta-analyses such as ours. For our
BMJ paper we did not request specific new data from the MHRA, we extracted
relevant data reported by the MHRA in the published report of the Expert
Working Group's findings (Report of the CSM Expert Working Group on the
safety of selective serotonin reuptake inhibitors. 2004). We make this
clear in the BMJ paper.
As you are aware, as members of the Expert Working Group, we are requested
not to discuss the specific content of discussions that took place in the
meetings outside the meeting. Inquiries about this should be directed at
the MHRA - we note with interest that you are currently pursuing these
issues.
It is our belief that you have made important contributions to debates
concerning risks associated with antidepressants. The findings reported in
our paper, together with those reported in Fergusson et als accompanying
paper in the same edition of the BMJ, confirm that earlier concerns about
increased risks of suicidal behaviour associated with SSRIs were justified.
We stand by our BMJ analysis and our decision to restrict this to the SSRIs
rather than include the SNRI (Venlafaxine) and the pre-synaptic alpha-2
antagonist mirtazapine. Pharmacological advice suggests venlafaxine shares
some of the properties of the SSRIs. As pointed out in the EWG report this
is less the case with mirtazapine. Both drugs are listed in a separate
section from the SSRIs in the British National Formulary and we similarly
excluded these drugs from our previous paper presenting data on evidence of
the association of SSRI prescribing with suicide risk in children (BMJ
2004;329:34-38). The data on suicide risk in relation to these drugs is
clearly of concern. It is also a concern that the placebo controlled trials
of the SSRIs generally included too short a period of follow-up to assess
any longer term beneficial (or adverse) effects of SSRIs on suicide risk.
You are, we are sure, aware of the current debates concerning the possible
beneficial impact of rising sales of SSRIs on population suicide rates.
The numbers you report in your table are consistent with those given in the
EWG's report though we note you have excluded the suicide in the
escitalopram group occurring after treatment cessation and the 3 placebo
suicides occurring in the paroxetine trials after treatment cessation.
We do not believe it will be fruitful to pursue this correspondence with
you any further.
Yours sincerely
David Gunnell and Deborah Ashby
Suicides in clinical trials of paroxetine –v- placebo
I have been revisiting our earlier correspondence in the light of the attached letter (4 September) from Ms Sarah Walk at the MHRA. As you know, I raised a number of Freedom of Information requests with her and Dr Williams, specifically about the three so-called ‘placebo suicides’ in the paroxetine trials.
In the six months that have elapsed since you told me that you and Professor Ashby wished to end our correspondence, things have moved on. Ms Wark specifically suggests I now address these questions to you:
Q1: Were the "Narratives for
Cases of Suicides in Clinical Trials" (supplied to the MHRA by GSK in January 2004,
as Appendix 6, in response to the Seroxat Article 31 submission) made available to [a]
Professor D Ashby; [b] Professor D. Gunnell and/or [c] to any other member of the CSM
Expert Working Group (EWG) on the Safety of SSRls? If so, please explain why did the
Gunnell/Ashby review in the BMJ (19 February 2005) state: " . . . we did not have
access to individual patients' data . . ." and if they were not supplied, please
explain why not.
A: The text of the narratives of the cases in question were supplied as annexes to
assessment reports provided to the members of the Expert Working Group on SSRIs. I would
ask that you direct any questions about the BMJ review (19 February 2005) to the authors.
Q2. Had Professors Gunnell and Ashby and/or any other members of the
CSM Expert Working Group (EWG) on SSRls been informed, by this or any other documents or
information supplied to them by MHRA staff, by the time of the publication of the EWG
report (December 2004), that the three so-called 'placebo suicides' in the paroxetine
trials comprised: [a] one case improperly included because death occurred beyond the
30-day cut-off point; [b] one case in which the patient had been treated with fluoxetine,
oxazepam and clotiapine shortly before committing suicide; and [c] a third case in which
the supposition that the patient committed suicide was apparently made solely on the basis
of two telephone calls from the brother of the deceased - the first reporting this patient
had 'passed away', the second 'confirming' the patient committed suicide (by unrecorded
means)?
A: As stated above the EWG were supplied with the case narratives.
In your BMJ paper (19 February 2005), in the description of ‘Methods’ and ‘Data Sources’, you wrote that you did not have access to individual patients’ data. However, Ms Wark now states that EWG members were provided with the individual case summaries for the three so-called ‘Placebo suicides’ in the paroxetine trials. These were critically important data. Please will you confirm [a] that you did receive these individual case summaries; and [b] Whether or not you were aware, before submitting your BMJ paper, of the limitations of those data, as summarised in Q2 above, and in related correspondence.
Your paper presented analyses of relative risk of suicide that both included and excluded these three placebo suicides, but I do not understand why. Assuming you were aware of these data, did you not conclude that the evidence was ropey – and, if so, why did you include them? Why were the references to these three placebo suicides in both your paper (e.g. in Figure 1) and in the EWG report, so bland and uncritical as to dignify and validate these absurd ‘results’?
And did you decide to exclude these placebo suicides in one analysis solely for the reason you gave - that these fatalities "occurred after they had stopped treatment"? I don’t understand why this should be some binding criterion for exclusion – and if it should, why does the MHRA routinely rely on protocols for studies of active drug –v- placebo that include some post-treatment analysis? Would you confirm (or not) that your decision to exclude these data in one analysis was mainly because of the very poor quality of evidence?
I am anxious to pursue these questions, if only in the interests of better public understanding of ‘science’ and ‘scientific’ communication. In this connection, would you have any objection to my posting this correspondence on our website?
I look forward to hearing from you.
Sincerely,========================================================
21 October 2006
Dear Professors Ashby and Gunnell,
I wrote to you three weeks ago, but have had no response from you. However, the MHRA has since told me - flatly contradicting their earlier assurance - that neither you nor any other member of the Expert Working Group on SSRIs were provided with the case narratives of the three placebo suicides. I'm very sorry you were so badly let down, and I hope the Agency is too.
In my letter of 1 October (copy attached) I asked if you would have any objection to my posting our correspondence on the Social Audit website. In view of the silence, I assume not, and I'm now writing to say that I shall shortly be doing so.
Sincerely,CLICK HERE TO RETURN TO MAIN TEXT