Appendix II
Dear Martin,
I am writing to express my concern in regard
to your draft report, "British Biotech—Reappraising
Marimastat". I have been briefed by James Noble and others
here about the meetings which have taken place over the last few
weeks between you and your colleagues at James Capel and various
members of our staff. As you know from these discussions and from
James' letter to Quintin Price of 23 August, we are concerned
about the many areas in your 20 August draft which contain inaccurate
statements or misleading information.
I have spoken to Quintin Price earlier this
week and he has told me that you intend to prepare a fresh draft
report on a different basis, concentrating on a few issues which
you consider to be key to the assessment of marimastat and its
prospects. We would be happy to discuss our own views on key issues
at further meetings with you as appropriate. They are not the
same as the issues put forward in your draft note, which I can
only assume arise because we have not been sufficiently clear
in explaining the current clinical data to you.
I do not want to belabour the point but there
are a large number of statements in the draft which are simply
untrue. I am particularly concerned to learn that this draft report
has been passed to consultants, colleagues and others for comment,
prior to its verification. It is obvious that the comments you
receive on our studies are bound to be critical or spurious if
you provide incorrect or misleading information. Please will you
therefore not use this draft report any further as a basis for
consultation.
In respect of the ten items you named as "key
issues" in the report, I strongly endorse the comments made
by James in his letter to Quintin of 23 August. To aid your understanding,
I am also sending you a few further comments. I have used your
original headings.
1. TRIAL DESIGN
The Phase II trial programme has been aimed
at defining the optimum dose for Phase III definitive studies.
It was not the intent to prove efficacy in Phase II. This will
be done by large randomised placebo or comparator drug controlled
trials in the Phase III programme that has now begun. Nevertheless,
encouraging and important data on potential efficacy of differrent
doses has come from the Phase II antigen trials. It is untrue
to say these Phase II studies "lack a control group".
All studies have included one or more groups of patients given
drugs at doses which are below the level necessary for efficacy.
These groups constitute the best possible control and have enabled
us to define precisely the magnitude of the "placebo effect"
and the level of drug likely to produce clinical benefit.
2. ANTIGEN MARKERS
The published study you refer to (Gastroenterolgy,
90: 343-349, 1986) does not show any such thing as you quote.
Steinberg's study compares the sensitivity and specificity of
CA19-9 with that of CEA in detecting cancer of the pancreas and
concludes that while both are useful, "CA19-9 used alone
is superior to CEA used alone in detecting cancer of the pancreas".
The two sequential assays of CA19-9 in 18 patients which you have
used to question our data were not taken 28 days apart. All assays
were taken less than one month apart and the table was included
in the paper to show the reliability of the assay in the assessment
of untreated patients. In reporting the results the investigators
state: "all patients who had an elevated assay remained elevated
and the three with non-elevated levels remained non-elevated on
repeat testing". This is the precise reverse of your interpretation
of their study.
3. CUT-OFF
CRITERIA
The criteria we have used for antigen evaluation
in the Phase II trials were defined prospectively by us following
discussion with expert clinical oncologists. The fact that they
are not precisely the same as those used by Dr Gordon Rustin (whose
article you cite) in analysing response to cytotoxic chemotherapy
is neither surprising nor relevant. Nevertheless, you should note
that we have been careful to classify patients as showing a "biological
effect" on marimastat treatment only if they have demonstrated
an absolute fall in serum antigen level. Patients who showed a
rise of 0-25 per cent on treatment were classified as showing
only a "partial biological effect". You will note that
this is consistent with the principle of Rustin's recommendations
although the exact cut-offs are different because marimastat is
a tumoristatic drug, unlike cytotoxics which are tumoricidal.
It is deliberately misleading to use Rustin's paper to argue that
the marimastat criteria are inappropriate.
4. DATE VARIANCE,
AND5. STATISTICAL
ANALYSIS
Again you misquote from Dr Steinberg's study
(Gastroenterology 90: 343-349, 1986). In the 37 patients with
carcinoma of the pancreas studied by his team (not 18), the mean
CA19-9 level was 11892657 units/ml. We agree that the distribution
of these data points is skewed. However, you forget that the data
we are subjecting to analysis in our statistical treatments is
not the absolute antigen levels but the percentage change in antigen
levels over 28 days. This percentage transformation largely normalises
the data which is a major reason for reporting the results as
percentage changes. Therefore it is incorrect to state that parametric
statistical tests are inappropriate. Nevertheless, to be absloutely
sure, we have analaysed the trial results using both t-tests (parametric)
and Wilcoxon tests (non-parametric) and obtained identical results,
as would be expected from data which closely approximates a normal
distribution. Please therefore would you refrain from stating
that the data are higly skewed and that we have used the wrong
tests.
6. POOLING OF
DATA AND7. DOSE
RESPONSE
As James has explained clearly, statistical
analysis of the data from individual tumour trials shows that
pooling data from these studies is valid. We have reported the
overall pooled results on the effect of marimastat treatment and
on the dose response simply because it is the obvious way of summarising
all the data. When this is done, the results stand out plainly
for all to see. You do not need to be an expert to see that at
doses of 10 mg twice a day and above, the drug is having an obvious
effect. However, it is incorrect to state or imply that individual
trial results are different. They are not—an effect of marimastat
treatment is seen consistently (highly significant in two out
of four cancer types) and a similar pattern of dose response is
seen in each.
8. CLINICAL
OUTCOMES
Please will you go back and re-read what we
have said on clinical outcome correlations. I do not think you
have understood it correctly. We have demonstrated that there
is a statistically significant correlation between the presence
of a positive antigen response to marimastat over the first 28
days of treatment and improved longer term survival in pancreatic
cancer, improved progression-free survival in ovarian cancer and
improved rates of tumour stabilisation in pancreatic cancer measured
by CT scan. The survival curves are not "part factual and
part prediction of death rate". They are the actual survival
of patients in the trials. The latter part of the curves are
not "predicted from a historical database" they are
real observations of patients recorded as alive at the times plotted.
Consequently, the significant difference in survival curves for
"good" or "poor" antigen responders to marimastat
is real. We are careful not to claim that the Kaplan Meier curves
presented can be compared with published survival curves for untreated
patients or patients treated with other drugs. That analysis must
await the results from the comparative Phase III trial of marimastat
in pancreatic cancer currently under way and other Phase III trials
being planned.
Similarly, in the CT scan data, all we are saying
is that a good antigen response to marimastat correlates with
a lower incidence of progression of disease. We are not saying,
and it would be absurd to do so, that antigen response is a better
predictor of disease status than CT scan.
9. INCONSISTENCIES
James has responded to this in full.
10. SIDE EFFECTS
I am pleased to see that you have correctly
reported the data we have provided on side effects.
I hope the above is helpful. It is important
that there is agreement on the facts of the data in order for
us to have a meaningful discussion on the potential for the drug.
Please do let me know if any of the points in James' letter or
this one are unclear. In the interests of achieving congruent
understanding we would be happy to provide further opportunities
for discussion.
Finally, let me take a step back and put our
disagreements into an overall perspective. Since British Biotech
was founded, 10 years ago, the company has insisted on the highest
standards of integrity, honesty and communication. We have never
knowingly misled an investor, employee, analyst or business partner.
I am sure that you and your colleagues at James Capel strive to
achieve similar standards. These aspirations and our relationship
are not helped if you seek to find fault and untruths in every
piece of information we give you. I would appreciate it, therefore,
if we could re-commence our relationship on a basis of mutual
respect and trust.
Please call or write to let me know how you
wish to proceed.
| 
