American Journal of Epidemiology
Copyright O 1996 by The Johrra Hopkins University School of Hygiene and Public Health
All rights reserved
Vol. 144, No. 11
Printed In USA
Food and Phytochemicals, Magic Bullets and Measurement Error: A Commentary
John D. Potter
The original explanation for a lower risk of colon
cancer in the presence of a diet high in plant foods was
attributed to effects of fiber—and there are both human studies (1) and an extensive experimental literature to support this view. Nonetheless, the more interesting and consistent finding is that vegetable and fruit
intake is associated with lower risk of most epithelial
cancers (2). Colon cancer is among these cancers; the
highest quantile of intake is usually associated with an
approximate halving of risk (3).
Less is established about the possible prevention of
the precursor lesions—adenomatous polyps—by a
diet high in plant foods or their constituents. There are,
nonetheless, several papers that suggest that a greater
intake of vegetables and fruit lowers risk of polyps (4).
The paper of Tseng et al. (5), the first in this issue of
the Journal, presents some evidence of a reduced risk
of adenomatous polyps in the presence of greater
intakes of specific micronutrients. The most striking
features of these findings are the sex differences. With
the sole exception of the (unexpectedly) inverse association with dietary iron, all of the nutrients that suggest a reduced risk in women are unassociated in men
and vice versa. With additional analysis, folate appears
to be the most consistent predictor of lower risk in
women (with no evidence of interaction with alcohol
use); in men, the consistent findings are vitamin E and,
perhaps, calcium. It is not clear whether these differences speak to real differences between the sexes in
the dietary predictors of colon cancer (3, 6) or whether
what is being observed are differences in where dietary measurement error occurs across the sexes. The
markedly different p values for trend in tables 2 and 3
of the Tseng et al. article (5), for what appear to be
slopes of a similar order, are a consequence of differences between the sexes in the differences in distributions between cases and controls (M. Tseng, Department of Epidemiology, University of North Carolina
Received for publication July 1, 1996.
From the Cancer Prevention Research Program, Fred Hutchinson Cancer Research Center, Seattle, WA.
Reprint requests to Dr. Potter, Head, Cancer Prevention Research Program, Fred Hutchinson Cancer Research Center, 1124
Columbia Street, MP-702, Seattle, WA 98104-2092.
at Chapel Hill, 1996, personal communication). Again,
is this telh'ng us about sex differences or measurement
error?
The paper by Witte et al. (7), the second in this
issue, provides some additional data. The authors report two important findings in addition to supporting
the general lowering of risk of adenomatous polyps
with greater intakes of vegetables and fruit. First, they
show that some specific plant foods and plant products
such as grains, garlic, and tofu are also associated with
lower risk. Second, and most usefully, they show that
the lower risks are not explained, at least completely,
by several candidate agents—fiber, folate, /3-carotene,
and vitamin C, nutrients that were previously shown to
be inversely associated with risk in this study (8).
There are at least two interpretations of the second
finding. The conservative one is that the relevant variables are too poorly measured for the association to
become completely attenuated, even when the variables are on the same causal pathway. The other is to
conclude that there are unmeasured constituents that
also contribute to the lower risk. Given the number of
phytochemicals that could explain the association with
high intakes of plant foods (9, 10), prudence and good
science suggest the latter. The inconsistency of the
findings across studies could support either explanation. Should we interpret findings that are more consistent (e.g., folate) as indicating something closer to
the causal pathway (11), or do we just measure it
better?
Although the adenoma-carcinoma sequence is a
well-established hypothesis with both strong clinicopathologic support and evidence of specific somatic
genetic changes (12, 13), the part played by external
exposures, especially dietary constituents, in increasing or decreasing the likelihood of neoplasia is not
well understood at all. The fact that a diet high in plant
foods is associated with lowered risk of both polyps
and cancer suggests, minimally, that the relevant
agents act early in the pathway and that they reduce
the risk of polyps that are likely to progress to cancer.
Similar or different agents may, of course, act late in
the process as well. This differs from the impact of
smoking on colon carcinogenesis wherein exposure
1026
Food and Phytochemicals
appears to increase the risk of adenomatous polyps,
but these polyps do not often progress to cancer, and
smoking does not act late in the cancer process.
Vegetable and fruit intakes are measured with error.
Food tables, at best, provide average values for specific nutrients, and the levels of many phytochemicals
in many foods are not known. Because of these and
other causes, there is a marked lack of precision in
measuring specific micronutrients and phytochemicals. Because specific nutrients are then metabolized
in a variety of ways (e.g., by P450 and phase II
enzymes), doses delivered to any tissue will be determined by metabolic profiles that vary both across
populations and, by organ site, within individuals.
These problems can be overcome partially by continuing to undertake food analysis, by increasing our
understanding of error structures in dietary studies,
and by improving our capacity to measure, more directly, tissue levels of the specific nutrients and phytochemicals of interest. At some point, perhaps, it will
become possible to undertake the kind of analysis
presented by Witte et al. (7) in studies of both precursor lesions and cancer and to conclude, with confidence, which dietary constituents play what roles in
explaining the protective effects of plant foods at
various stages of carcinogenesis.
In the meantime, two hypotheses are worth entertaining. First, if we are able to identify specific protective constituents, these not only will vary with the
stage of the cancer process, but they also will differ by
sex and age and across genetic metabolic profiles; they
will certainly vary, more generally, by organ site.
Second, there will be very few magic bullets. Therefore, formulating pills to match a risk that is defined as
[cancer site by sex by metabolic profile by . . . and so
forth] will be much more problematic than encouraging a general increase in the consumption of plant
Am J Epidemiol
Vol. 144, No. 11, 1996
1027
foods. This is a step we can take right now and enjoy
the built-in redundancy of multiple agents with independent, overlapping, and perhaps interactive mechanisms. Besides, well-prepared food tastes better.
REFERENCES
1. Howe G, Benito E, Castelleto R, et al. Dietary intake of fiber
and decreased risk of cancers of the colon and rectum: evidence from the combined analysis of 13 case-control studies.
J Nad Cancer Inst 1992;84:1887-96.
2. Steinmetz K, Potter JD. A review of vegetables, fruit, and
cancer. I. Epidemiology. Cancer Causes Control 1991;2:
325-57.
3. Potter JD, Slattery ML, Bostick RM, et al. Colon cancer a
review of the epidemiology. Epidemiol Rev 1993; 15:
499-545.
4. Potter JD. Nutrition and colorectal cancer. Cancer Causes
Control 1996;7:127-46.
5. Tseng M, Murray SC, Kupper LL, et al. Micronutrients and
the risk of colorectal adenomas. Am J Epidemiol 1996;144:
1005-14.
6. McMichael AJ, Potter JD. Do intrinsic sex differences in
lower alimentary tract physiology influence the sex-specific
risks of bowel cancer and other biliary and intestinal diseases?
Am J Epidemiol 1995;118:620-7.
7. Witte JS, Longnecker MP, Bird CL, et al. Relation of vegetable, fruit, and grain consumption to colorectal adenomatous
polyps. Am J Epidemiol 1996;144:1015-25.
8. Enger S, Longnecker M, Chen M-J, et al. Dietary intake of
specific carotenoids and vitamins A, C, and E and prevalence
of colorectal adenomas. Cancer Epidemiol Biomarkers Prev
1996;5:147-53.
9. Steinmetz K, Potter JD. A review of vegetables, fruit, and
cancer. II. Mechanisms. Cancer Causes Control 1991 ;2:
427-42.
10. Jacobs D, Slavin J, Marquart L. Whole grain intake and
cancer: areview of the literature. Nutr Cancer 1995;24:221-9.
11. Giovannucci E, Stampfer MJ, Colditz GA, et al. Folate, methionine, and alcohol intake and risk of colorectal adenoma.
J Natl Cancer Inst 1993;85:875-84.
12. Fearon E, Vogelstein B. A genetic model for colorectal tumorigenesis. Cell 1990;61:759-67.
13. Markowitz S, Wang J, Myeroff L, et al. Inactivation of the
type II TGF-receptor in colon cancer cells with microsatellite
instability. Science 1995;268:1336-8.