Day 256 - 04 Jun 96 - Page 44
1 between the intake of animal fat and breast cancer in
2 China, according to this passage and the rest of your
3 study, is weak, is it not?
4 A. It is significant.
5
6 Q. But how significant?
7 A. It is quite significant. It is statistically -- there
8 is less than one chance in 20 that this is a chance
9 observation.
10
11 Q. Can I ask a question about that. Do you use relative risks
12 in expressing the magnitude of risk?
13 A. In certain kinds of analyses yes.
14
15 Q. Would you use them for this kind of work?
16 A. I have forgotten exactly, this paper here, whether we
17 did or did not. This was actually a paper -- this was
18 referring to a paper that was done by my colleagues at
19 Oxford here in England, as well as in London. When we do
20 basically all case control studies, you are comparing case
21 versus controls, then of course we use relative risk as the
22 sort of outcome, but not when we are doing sort of ecologic
23 comparison; we do not use that term. But in this case
24 I think we did use relative risk, but I cannot -----
25
26 Q. This is a slight red herring, Professor. Do you have a
27 knowledge of case control studies and the way that their
28 results are expressed?
29 A. Yes, I do.
30
31 Q. Do you agree that, at any rate, so far as a case control
32 study is concerned, a relative risk of less than about
33 threefold is not worth bothering about, save as the genesis
34 of a hypothesis?
35 A. No, I do not agree. For example, if we have a relative
36 risk of 1.6 or 7, it does not sound like much, or even 2 or
37 so, it does not sound like much, but if it is statistically
38 significant and repeated and repeated -- replicated, if you
39 will -- in other words, we have confidence that it really
40 is one and a half or two-fold or so, if it is a disease
41 that is common, then the total number of cases of course
42 can be quite significant. So we have to confess that even
43 that relative risk is significant.
44
45 Q. My fault entirely. I meant for a single study, properly
46 conducted with a sound methodology, a relative risk which
47 came out at less than about 2 would not be significant?
48 A. Probably so, yes.
49
50 Q. How far in the other direction do you go, if you take the
51 standard or basis as 1, how far in the other direction do
52 you go, for example, in a case control study, before you
53 say there is a negative or inverse association?
54 A. Those two words mean something different; negative and
55 inverse mean something very different.
56
57 Q. I am sorry. Use inverse.
58 A. OK.
59
60 Q. Protected, if you like?
