Day 040 - 21 Oct 94 - Page 19
1 report is: "The re-evaluation confirmed that there was a
2 dose related trend for increased calcification and
3 epithelial hyperplasia in the renal pelvis of the female
4 F1, second generation rats".
5
6 MR. JUSTICE BELL: Keep going.
7
8 MR. RAMPTON: You must read on, please.
9
10 MR. JUSTICE BELL: Especially the statistical analysis sentence.
11
12 MR. RAMPTON: Exactly.
13 A. Yes, indeed what it says is: "Statistical analysis",
14 using the one-tailed Fisher exact test, "showed the low
15 dose (50 mg/kg) incidences of pelvic calcification and
16 hyperplasia not to be significantly different from the
17 control."
18
19 Q. Would you, please, read on right to the end, then give me
20 an answer?
21 A. May I comment on that sentence first?
22
23 Q. Yes, by all means.
24 A. What I take it that is saying is that when the
25 statistics on these groups of animals were analysed, they
26 did find statistically significant differences in the
27 higher dose groups. But for that low dose group, the
28 difference was not statistically significant, by which
29 I take it they meant, in the normal meaning of these terms,
30 that they could not be confident that there was better than
31 19 chances out of 20 that this had been a random
32 occurrence; but one of the respects in which I think we
33 have to be cautious in the interpretation of toxicological
34 data is in the ways in which measurements of statistical
35 significance are taken and interpreted. It is common and,
36 in my view, all too -----
37
38 MR. JUSTICE BELL: How do you criticise that conclusion
39 statistically in this case?
40 A. I am trying to come to that.
41
42 Q. This particular case?
43 A. Yes, in this particular case.
44
45 Q. Not how statistics may go wrong, but how did they go wrong
46 here, or do you think they may have done, in this paper?
47 A. It seems to me that we are not being given sufficient
48 information. What I would like to know is, in order to
49 make a judgment of the toxicological significance of the
50 effects that did appear at 50 milligrams per kilogram, was
51 not whether it fell short of the significance level of 0.05
52 or one in 20, but how far it fell short. If it fell only
53 very, very marginally short, then I am not comfortable with
54 taking that level of significance as an unproblematic cut
55 off point. Anything that falls short of that by however
56 little or however much deserves to be disregarded, because
57 you have -- I think it would help if I explained it in this
58 respect -----
59
60 Q. Have you read this paper?
