SCIENTIFIC REDUCTIONISM 285
and 17% increase with increased milk intake.^46 The Harvard researchers
therefore summarized in 2002 a more recent group of studies, this time
including eight large prospective studies where dietary information was
thought to be more reliable and where a much larger group of women
was included. The researchers concluded:
We found no significant association between intake of meat or
dairy products and risk of breast cancer.2 6Most people would say, "Well, that's it. There is no convincing evi-
dence that meat and dairy foods are associated with breast cancer risk."
But let's take another look at this supposedly more sophisticated analy-
sis.
All eight of these studies represented diets that had a high proportion
of animal-based foods. In effect, each study in this pool was subject to
the same flaw from which the Nurses' Health Study suffered. It makes
no sense, and does no good, to combine them. In spite of there being
351,041 women and 7,379 breast cancer cases in this mega-database,
these results cannot detect the true effect of diets rich in meat and dairy
on breast cancer risk. This would be true even if there were a few mil-
lion subjects in the study. Like the Nurses' Health Study, these studies
all involved typical Western diets highly skewed toward the consump-
tion of animal-based foods, where people are tinkering with the intake
of only one nutrient or one food at a time. Every study failed to take
into account a broader range of dietary choiceS-including those which
demonstrated positive effects on breast cancer risk in the past.
IGNORING MY CRITIQUEOnce, after reading a publication on animal protein and heart disease
in the Nurses' Health Study,9 I published a critique^47 summarizing some
of the same points that I am making in this chapter, including the in-
ability of the Nurses' Health Study to advance our understanding of
the original international correlation studies. They responded, and our
exchange is as follows.
First, my comment:
Within a dietary range [so rich in animal-based foods 1, it makes no
sense to me that it is possible to reliably detect the so-called inde-
pendent associations of the individual constituents of this group
when it can be expected that they share the same disease outcomes