Day 036 - 13 Oct 94 - Page 21
1 Q. "Two limitations in the ability to detect associations of
2 dietary fat and breast cancer in analytic epidemiologic
3 studies have been the narrow range of fat intake in the
4 study population and the substantial measurement error
5 associated with dietary assessment methods. Several
6 authors note that the range of fat intake in most analytic
7 studies, in which the magnitude of the dietary fat-breast
8 cancer association has tended to be small, is
9 substantially narrower than in ecologic studies, where a
10 strong association has been observed. The effect of a
11 narrower range of fat intake would be to decrease the
12 relative risk between extreme categories of fat intake in
13 the study population. If it is assumed that the relative
14 risks observed in ecologic studies represent 'truth', then
15 the underlying relative risk between high and low
16 quintiles of fat intake in a cohort study such as the
17 Nurses Health Study" -- that I will tell you is (you will
18 know it is by Willett, but it is important which one it
19 is) the one that was reported in 1987 in the New England
20 Journal of Medicine -- "in which fat intake according to
21 four 7-day food records ranged from about 32% to 44% of
22 energy, is about 1.4. In our study, mean fat intake in
23 the extreme quartiles according to the food-frequency
24 questionnaire was 26.8% and 41.2% of energy, indicative of
25 a fat intake range that is essentially comparable to that
26 in the Nurses Health Study. This suggests that the Iowa
27 study population was also not ideally suited for
28 investigating the dietary fat-breast cancer
29 association."
30 A. Yes. Perhaps what Dr. Kushi is referring to is the
31 fact that Iowa is a state underneath Minnesota, just south
32 of Minnesota, where Dr. Kushi's office is, where pork
33 production and cattle production is a principal industry
34 where a high-fat diet is particularly common, and there
35 are obviously other problems with the study that Dr. Kushi
36 is citing as well.
37
38 Q. He is telling us anyway, is he not, that his mean extremes
39 were 26.8 and 41.2 per cent?
40 A. Yes, that is what it says there.
41
42 Q. The lowest extreme, the mean lowest extreme was 26.8 per
43 cent; he did not go any lower than that?
44 A. This was very similar to Willett's study, as he says.
45
46 Q. He gives the figure for Willett's study as 32 per cent to
47 44 per cent, does he not?
48 A. Although in Willett's follow-up study he ended up at
49 27, I believe was the lowest, but, yes, you are quite
50 right.
51
52 Q. In fairness to Willett -- we are going to look at him in
53 1992 in just a moment after the short break -- "Willett
54 et al. (38)" -- we had better just see which one that is;
55 that is 1990 -- "pointed out that this argument may not be
56 valid because the same dietary assessment methods were
57 adequate to detect positive associations of fat intake
58 with colon cancer in the Nurses Health Study." Is that a
59 fair point made by Dr. Willett and his colleagues, do you
60 think, Dr. Barnard?
