Annex H
Memorandum from Andrew Millar at British
Biotech
TO: KEITH
MCCULLAGH
15 MAY 1997
INTERIM CLINICAL
DATA
Following on from our discussions on Tuesday
and Wednesday in which I disclosed to you that from November of
last year I was in possession of clinical data that had an important
potential bearing on the company's direction, I write to confirm
the following:
Marimastat:
During November, it became apparent to me from
routine surveillance of serious adverse events that the majority
of deaths in the pancreatic cancer study (128) were occurring
on marimastat. This was not superficially surprising as 75 per
cent of patients receive marimastat compared to 25 per cent on
gemcitabine, on the other hand, because the study was not blinded
between marimastat and gemcitabine I asked the project teams in
the US and UK to report to me the actual death rates on both treatments
in order to investigate any possible safety concerns. I attach
the memo from Dr Henrik Rasmussen dated 26 November 1996. At that
time there were 56 patients recruited on marimastat, 16 of whom
had died and 18 on gemcitabine, one of whom had died. This obviously
raised some concerns. Henrik, Dr Lloyd Curtis, Dr Terry Rugg and
I discussed this in detail. Firstly we observed that numbers were
very small indeed and no significant conclusions could be drawn
at that time, even though the direction of the trends caused concern.
We concluded that the deaths on marimastat were occurring at a
very short time after entry into the study, before any possible
therapeutic effect and perhaps patients would only show a possible
benefit after taking the agent for longer periods. We had no evidence
from any other studies at that time that there was a particular
safety concern with marimastat and we all agreed that the study
should be continued with the situation under close review.
I informed Dr Peter Lewis of the situation at
the first opportunity, I cannot pinpoint the date but it would
have been either at an unscheduled meeting before 9 December or
else at our weekly meeting which was held on 9 December. He agreed
with my conclusions and indicated that he would take appropriate
cautious action from the company's point of view which was one
of my concerns. I gave him this bit of information at the same
time as the information on lexipafant (see below). He instructed
me not to tell you and to keep the information between ourselves.
He told me that we would both be for the "high jump"
if it was disclosed that I had knowledge of this clinical data.
I disagreed with this view, but recognised his authority and trusted
him to undertake appropriate action.
I watched the situation closely and decided
that I needed to look at the different dose groups on marimastat.
I directed Dr Alan Cornish to unblind the study and I attach the
worksheets for 31 January, 12 March and 29 April. I informed Peter
of each of these analyses by way on an unfiled hand written note
either at our weekly meetings or at other unscheduled convenient
opportunities. As you can see from the worksheets, the situation
with regard to deaths became no worse and equalised during the
first few months of this year. Currently there are 8/26 (30 per
cent) deaths on gemcitabine and 26/95 (27 per cent) deaths on
marimastat and I have just discontinued reviewing the situation.
At each of my meetings with Peter, he impressed upon me his view
of the importance of witholding this information and the lexipafant
data from you and gave various reasons for this. While the situation
became less concerning from a clinical point of view, I became
suspicious as to why he was so insistent, particularly as the
company was following a very optimistic course.
Lexipafant:
In November of last year, it became apparent
that the event rate in the US pancreatitis study 215 was lower
than we had expected with deaths running well below 10 per cent.
At that time, we had fully analysed the UK data which showed a
consistent but not absolutely compelling drug effect on mortality,
in that the significant analysis was stratified. On the other
hand, all the clinicians involved in the study had reviewed the
data and agreed that there was strong evidence of a therapeutic
effect in reducing mortality. It seemed essential to know whether
lexipafant was performing similarly in the high ITU usage USA
clinical situation compared to the UK. I directed Alan Cornish
to inform me of the distribution of deaths at that time. We had
very incomplete information in house with regard to dosing. I
attach his worksheets of 27 November and 31 January. On 27 November
the numbers were clearly too small to make any judgements, but
you can see that there were four deaths at 100 mg/24 hours, two
deaths at 10 mg/24 hours and one death on placebo, which represents
a trend inconsistent with great optimism. There were five deaths
for which we could not ascertain the dosage allocation. As more
information became available as recorded on the worksheets of
31 January and beyond, the death rates became 12 at 10 mg/24hours,
seven at 100 mg/24 hours and five on placebo. Clearly too small
to draw any conclusions, but not consistent with the probable
treatment effect we had observed in the UK study. I informed Peter
Lewis of these data at meetings held on a weekly basis between
us with the response which I indicate above. It was and remains
my judgement that these data were material to the direction of
the company and I trusted Peter to take appropriate action with
regard to corporate matters.
In undertaking all of these data reviews I kept
all blinded information absolutely secret from the investigators,
project physicans, CRAs, statisticians and databasers.
|
