Further memorandum submitted by Professor
TB IN CATTLE
1. This note follows up my earlier submission
to the EFRA Committee for their session of 24 October, and includes
responses to some issues raised in that session.
I begin, at the request of their Chairman John
Bourne, with a brief audit of the ISG's Response to Tuberculosis
in Cattle and Badgers: a Report by the Chief Scientific Adviser.
2. Having considered carefully both the
Chief Scientific Adviser's Report, and the ISG's Response, I am
happy to endorse the ISG's main points, as set out in paragraphs
ISG3-5, and their more detailed response as set out in subsequent
paragraphs, including the key aspect of terms of reference (ISG10-11),
and the six points concerning the interpretation of scientific
data, statistical analyses, and modelling results (ISG12-27).
3. I trust it is not necessary to go into
all the detailed arguments, since I have already set out my endorsement
of the ISG's statistical and modelling work and its conclusions
in my audit of the ISG's final report, published at the same time
as that report in June 2007 (copy attached).
Among the conclusions of my audit (section 8 therein), I commended
their conduct and analysis of the RBCT as a very thorough and
sure-footed demonstration of good science aiding public policy
in a complex and controversial situation.
4. The ISG's Response covers most of the
six or seven specific criticisms to which I alluded briefly in
my earlier comments submitted to the EFRA Committee. I will therefore
comment here only on a few specific matters that are not so covered,
5. First, I stand by my general comment
that the Chief Scientific Adviser's Report is inexpert and unbalanced.
6. I am sorry that the Chief Scientific
Adviser has interpreted my accusation of lack of expertise as
"attacks on personalities", which it was not. My point
was that his report is mainly concerned with detailed discussion
of complex statistical modelling and data analysis issues, yet
his group lacked expertise in these areas. At the session of the
EFRA Committee on 24 October, Sir David neatly confirmed my criticism
in attempting to rebut it, when he set out the very considerable
expertise of his panel in quite different areas (ecologist and
badger expert; immunologist and microbiologist; epidemiologist;
expert in veterinary medicine and animal clinical studies, specifically
in the area of animal TB; developer of one of the critical tests
for cattle TB). [see EFRA Committee Proceedings, Q384]
It is worth noting in this context that the
group that previously reviewed the ISG's work in 2004, chaired
by Professor Godfray, did include an eminent applied statistician,
Professor Robert Curnow.
7. Since the Godfray Report was referred
to several times in Professor Woolhouse's evidence, including
restating that report's downplaying of the purpose of the RBCT
(see Q371) and it's likely usefulness (see Q380), I should like
to make available to the committee my rebuttal of those points
(please see Comments on Godfray et al, 2004, attached).
8. As regards balance, the CSA's Report
generally emphasises only points that downplay the detrimental
effect of culling. A clear example of this one-sided view is described
in the ISG's paragraph ISG15.
9. In my earlier comments I described the
CSA group's epidemiological analysis as muddled in its discussion
of the basic concept of R0. To explain: when as here we have disease
in two interacting populations, here one=cattle, two=badgers,
we have four parameters. For clarity, we should use four different
symbols to refer to thesefor example, R12 to denote the
number of secondary cases in badgers caused by one infected cowrather
than using the same symbol R0 for all of them as in Annex 3. The
overall R0 for the 2-population situation is a function of all
four parameters (to be precise, .
From this little bit of mathematics it is easy to
show that the statement in paragraph 11 of the CSA's report"If
R0 is below one for all the transmission routes, levels of infection
will decline substantially"is wrong; all four can
be below 1, and yet R0 >1, so that the disease will persist.
[For example, when each
Rij = 0.6, R0 = 1.2.]
The CSA's group's misunderstanding of R0 is
presumably what led to their faulty interpretation of some of
the ISG's modelling work (see ISG21).
10. A rather different point is that in
this part of the discussion they implicitly include an assumption,
which is more difficult to spot surrounded by mathematics, namely
that badger removal will reduce at least the badger to badger
transmission parameter (R22) and thus the incidence of the disease
in badgers. Unfortunately, the evidence is that this assumption
is untrue (see ISG25-26).
11. I should like to put on record my agreement
with Sir David King's preference for avoiding appearing over-precise
in the presentation of statistical results (Annex 1)a point
I have adhered to in my own reports on the ISG's work.
12. To finish with two general comments.
First, the EFRA Committee were right to ask Sir David King why
he did not discuss with the ISG his disagreements over their analysis
and conclusions before publishing his report. One cannot help
feeling that more light and less heat might have been generated
had he done so.
13. Finally, the single most important difference
between the ISG's and the CSA's conclusions stems from the latter's
too narrow remit, as discussed in ISG10-11. One might rather have
expected the CSA to have set a wider remit, so as to consider
options for improved control of TB other than badger culling.
The key point here is that a discussion of science
that excludes the practical and economic aspects cannot be adequate
as a basis for government action. The Chief Scientific Adviser's
Report therefore cannot provide justification for it's strong
and unqualified policy recommendations.
Professor Denis Mollison
Former Statistical Auditor to the RBCT
Mollison, D (2004) Comments on the Independent
Scientific Review of the Randomised Badger Culling Trial and Associated
Epidemiological Research by Godfray et al, http://www.defra.gov.uk/animalh/tb/isg/pdf/audit.pdf
Mollison, D (2007) Statistical aspects of the Randomised
Badger Culling Trials: Final Report, June 2007, www.defra.gov.uk/animalh/tb/pdf/statsaudit
11 Not printed. Back
Not printed. Back