Examination of Witnesses (Questions 20
WEDNESDAY 15 NOVEMBER 2000
20. This paper, she says, was referred to referees
and five of the six referees agreed that points raised are fundamentally
correct. "The editor concluded that I would find myself `considerably
outgunned' . . .". If you are going to shoot her down in
that kind of fashion, you are going to have to use more firepower
than that because the essence of her argument is that the triplet
study is based on a reactive strategy, that badgers are only killed
after they have been associated with a herd breakdown. That is
unlikely to show an overwhelming impact on TB prevalence because,
as the Krebs Report points out, TB control can only be partially
effective because herd rates will already have occurred, and new
badgers will recolonise since only those in the immediate vicinity
are going to be culled. In other words, it is crucial to be able
to detect small effects, and the studyshe feelscannot
(Dr Donnelly) It is obviously a matter of degree.
We have to talk about what we are talking about in terms of a
small effect. Certainly we would have much less power to detect
differences of, say, 5 per cent. You are very right in saying
that the reactive strategy, because badgers are only culled after
there is a breakdown, is probably going to produce less of an
effect on incidence than the proactive, but we do not actually
know that. The idea behind the proactive strategy is to remove
the badgers associated with the particular breakdown. Although
we would expect to have immigration occur after that, if the reactive
strategy has removed the infected badgers and the badgers coming
in are not infected
21. You do not know that.
(Dr Donnelly) No, we do not know. That is why we are
doing the trial. We have structured this to test the two different
policies, and we are not prejudging it to say how they will do.
However, we have examined the data that we can, and I think it
might be better to talk to the editor of the journal to talk about
what the referees actually said, because I certainly cannot say.
22. If you were a referee you would not have
accepted the article?
(Professor Bourne) I think it is difficult for us
because we are aware that this paper has been submitted to a number
of journals, and as far as we know it has been rejected by those
journals. I do not think it is one journal, I think it is a number.
The fact of the matter is that this paper has not been published.
Nonetheless, we have taken it, as a group, very seriously and
seriously considered the contents of that paper. We believe we
have robust answers to it.
23. She is right, surely, about the weakness
of the reactive strategy?
(Professor Bourne) She may well be. There are unknowns
about this. This is the whole point of doing the trial. I think
the important thing for you to remember is that the direct comparison
between survey only and proactive treatment is the guts of the
trial. The reactive areas were included as a potential future
policy option. The data we will get from the reactive aspect of
the trial, in terms of scientific data with respect to answering
basic questions, will be less powerful than that from the proactive
area. It is primarily looking to see if it is a viable policy
24. In the memorandum that you submitted to
us you have actually made the point in paragraph 8 that in one
triplet you were unhappy with the culling effectiveness in the
proactive area, and I just wondered if you could explain what
effect that would have on the power. Does it delay its effectiveness,
or does it mean in practice that we have lost the triplet?
(Professor Bourne) Let us consider that in two parts.
The triplet that you are referring to was one that was culled
in January which ecologists recognise as being an inappropriate
time of the year for badger culling. I think the data we have
would support that. There were a number of reasons why we caught
a low number of badgers in that particular triplet. One was, we
believe, the time of the year and the relative inefficiency of
the trapping. I think that was confirmed by our own audit following
the trapping that we normally carry out a month after trapping
has ended. The other reason, of course, could be that there had
been much interference with badgers in that area through illegal
activity or previous badger removal activity. Again, we are uncertain
about that. We are aware that trapping efficiency was not as great
there as in other triplets, and we ascribe that to the time of
25. I was less interested in the reason for
it as to what its effects might be.
(Professor Bourne) Absolutely. Perhaps Christl could
come in here.
(Dr Donnelly) I suppose we also need to stress that
there will be additional culling in this proactive area. The hope
would be that these additional badgersif they were not
actually picked up in this initial cull in Januarywill
be picked up subsequently. When we are actually looking at the
effectiveness of the proactive or the reactive strategy, they
are all looking at the actual, in practice, effectiveness. It
does not actually do us a lot of good to figure out what the effectiveness
would be of a perfect strategy where all badgers were instantly
removed, because in making policy recommendations that will never
actually happen. So what we are trying to do is get a real live
test of how such a strategy would be implemented and, given its
implementation, try to cull badgers but within the constraints
of welfare concerns, and see what the effect of that is. That
is what we are trialing. If the culling does not get every single
infected badger it will not be 100 per cent effective. However,
we are seeing what effect we can get putting the best of our resources
and the best efforts into it.
26. You still have not answered my question,
with respect. My question was: what effect does it have on your
data? I understand the point that you are making that there may
be a statistical fluctuation in the effectiveness of the policy,
and it may sometimes mean that you in fact modify your practices
in order to make it effective, and you learn the lesson. I am
really interested. Does this mean that you have lost some data
for a period of time, or does it mean that the data is effective
for sometime downstream in the way that the analysis takes place?
(Dr Woodroffe) I can respond to that, if you wish.
We do not yet know the effectiveness of the cull in that area,
in part because the follow-up cull has not been carried out. That
will help us to distinguish between the numbers of badgers culled
in that area being low because of an ineffective cull or simply
because there are not many badgers there. Certainly the activity
on the ground in field science work did indicate rather low levels
of badger activity in the area. So we certainly expected the cull
there to be lower than it had been in other areas. It was lower
than we expected. Subsequent culls will help us to determine the
effectiveness of the first cull, in a sense, and I think the worst
case scenario would be that any effects of badger culling on cattle
TB would be decelerated. So the onset of any reduction you might
see could be delayed, but not negated.
27. Let us assume that it was not effective,
for the moment. If it was not effective, does that mean that your
data is undermined in terms of using the trial for a short period
of time, or does the effect continue through
(Dr Woodroffe) I would argue a short period of time.
28. It is the equivalent of losing some of your
data over a short period of time in five years.
(Dr Donnelly) It may be that we get a less effective
cull in the first six months until whenever we have the follow-up
cull that would actually be more effective in picking up these
(Professor Bourne) I do not think it is a question
of losing data. We accept that because we cannot cull at the most
appropriate period in the year we will get variation in culling
efficiency across the 10 triplets. What is important in this particular
triplet is that we do not return to that same triplet at the same
time of year and that we do return at a more appropriate time
of the year to improve trapping efficiency.
29. I do not think I am making myself clear,
Professor. What I am really trying to understand
(Professor Bourne) I think I understand the question.
30. The triplet is one of a modest number which
is important to the trial. If a mistake is made, how robust is
the experiment to the cull not being effective in one year or
two years? Does it mean that you throw out the data?
(Professor Bourne) I think you have put your finger
on it. It is a robust trial. You are referring to the 10 trial
areas as being a modest number. We believe it is not a modest
number, it is an appropriate number. We expect variations between
those trial areas, between triplets, so consequently the strength
of the initial data will vary between those triplets, but given
that we are returning to proactively cull these over a five-year
period, we would expect this to even out over that five-year period.
(Dr Donnelly) I do not think there is any reason to
think that ineffective culling in a first proactive cullif
that is in fact what took place (and, as Rosie has said, we will
not actually know until we get back into the area) would affect
the impact of incidence in the first year.
31. Can I ask a related question in my mind,
which is the power in terms of data? There is this paper we have
had which says that your calculation of your power may be wrong,
and I would be interested in the consequences of that. Referring
to previous Members of this Committee, when I was not on it, I
think you said they totally misunderstood the role of the power
calculations in your experiment. I think it is important that
we do understand it now. If you prove to be wrong and the counter-view
was correct, that in fact the power calculation was wrong, as
the data came in would you simply be able to correct the matter
by going for another year, or expanding the breadth of the experiment,
or would it be a fatal flaw?
(Professor Bourne) It would be appropriate, Christl,
if you could respond to the relevance of the power calculation.
(Dr Donnelly) Yes. In looking at the power calculation,
yes, it depends on a number of factors, but the key one is the
total number of breakdowns to the extent that you get violation
of the assumption of independence in the breakdowns that are happening,
as Fiona has notedalthough it is a matter of degrees how
much impact she thinks, we think and the independent auditor thinks
that would have. The power could probably be increased by increasing
the number of triplet years, but the power depends, in the end,
on the data that is actually generated. So we will be able to
look at this as time goes on. We have done our best to analyse
all the data that is available retrospectively. We just do not
know what the incidence will be in the underlying rate in the
next five years, and we do not know what the effect of the culling
32. I have not got an answer to my question,
which is if Fiona Mathews was right, in terms of the way the data
started to come out, and your analysis of the data indicated that
she was right, would the implication be that you then have to
extend the period of the trial or would it mean that there was
a fatal flaw? Would you be able to get more data simply by extending
it for a year?
(Dr Donnelly) Yes.
33. How soon would you know that you are right
and Fiona Mathews is wrong, in terms of data coming in?
(Professor Bourne) The information I get from statisticians
as Chairman of the group is that the initial power calculations
should be seen as purely indicative, determining the design of
the trials. The precision achieved in the trial will be determined
by the data obtained totally independently of the correctness
of the power calculations.
34. I appreciate that, I understand that. I
was simply asking roughly how long it will take before you will
know from the data coming in whether that initial calculation
of the power (and, therefore, how long this trial will have to
go on for) was correct?
(Professor Bourne) It does depend, of course, on the
strength of the data we get with respect to the quantitative involvement
of the badger.
(Dr Donnelly) We would certainly want to see data
from all the triplets before we actually had the best information.
So it would be the end of 2002 before we would be certain of having
herd test information on all herds that were enrolled in the trial.
It is the time until we finish the last proactive cull, and then
an additional year from that point until all the herd tests are
35. There is a high incidence of TB in Wales
but it is not included in the test. Why are there no triplets
(Professor Bourne) I had a couple of discussions with
the members of the Welsh AssemblyI have forgotten her name.
The Agricultural Minister, Christine Gwyther.
36. Sadly, she is no more.
(Professor Bourne) The situation was that given that
the wildlife unit resource was focused in Cornwall and in Aston
Down in Gloucestershire, it would have been difficult for us to
have used the Gloucester resource in two distant sites, one of
which could well have been the western counties of Walesin
Pembrokeshireand the other would have been in Staffordshire.
Sites were selected on the basis of the previous cattle breakdowns,
and we were quite keen to include new areas that have developed
a high number of breakdowns in the past three years. Certainly,
Staffs/Derby fitted that category. It was also useful to us in
that there had been no previous badger removal operations, so
it is quite different to some of the other areas that we are working
in. It was also made clear to me by Christine Gwyther that while
she would not welcome us with open arms to work in Wales, she
would consider us doing a project there, a triplet there, if we
felt it scientifically beneficial to do so. Frankly, we did not,
so we are not there.
37. Why did you not think that was useful? Do
you feel that the evidence you get from elsewhere is unquestionably
going to be applicable to the Welsh environment?
(Professor Bourne) I think so. There are features
of the Welsh badger removal operations which were particularly
interesting to us, namely the low incidence of TB in the badger
population which is recorded in Wales. As I mentioned, pragmatically
and logistically, we could not work in two distant areas using
the same wildlife team and stick to the timetable that we had
insisted that the wildlife unit worked to. So it was a question
of either working in one distant site, Staffordshire, or another
distant site, Wales. Also, I am bound to say we were influenced
by reports we did have of the likelihood of a large level and
high level of illegal killings going on in the western parts of
Wales. It was not a major influence, the major influence was the
other area I have mentioned.
(Dr Woodroffe) To reassure you, perhaps I could add
that the triplets that have been selected do represent quite a
wide range of habitat types and landscape types, so there should
be some generality to the trials across a wide array of environmental
conditions and badger densities. We hope it can be extrapolated
38. So you feel it can be extrapolated to Wales.
I was interested to learn that we have a healthier brand of badger
in Wales as well.
(Professor Bourne) It could be infected badgers die
39. Moving on to the second question: you said
there was about 80 per cent co-operation with the trials. How
do you measure that 80 per cent? Is that the number of farmers
(Professor Bourne) Farmer co-operation in the trial
areas has been around 80 per cent. Often it is higher but I do
not think it has been very much lower. It is based upon the identification
of landowners within trial areas by MAFF, who circulate each landowner
by letter asking them if they will become involved in the trial.
On that basis, they give permission either for surveying, surveying
and culling or no co-operation at all. The non-co-operation level
has been consistently around about 6 or 8 per cent. There is variation
between those who will agree to culling and surveying and those
who will survey only. So the figures are derived from a direct
approach by MAFF staff to landowners.