Introduction

Apparent continuing increases in the incidence of breast cancer have
stimulated interest in gaining a greater understanding of the development
of this disease so that it may be possible to prevent it.[1] At present,
considerable attention is being focused on the inherited genetic changes
that play important roles in perhaps 10% of cases of breast cancer among
American women. Although investigation of the functional implications of
these changes should provide useful information about broad developmental
processes in breast cancer, a more general focus on our emerging understanding
of this process is critical to addressing the ultimate public health goal
of breast cancer prevention: the reduction of the incidence of breast cancer
in the entire American population.

A broad strategy to address this goal is clear: to identify, understand,
and act broadly on manipulable causes of breast cancer. Pursuit of this
strategy requires that the following challenges be addressed:

1. Identification of (new) specific manipulable causes. For example,
   dietary factors and exercise appear to be important causes, but exactly
   which dietary constituents are important and when in the developmental
   causal sequence these constituents and exercise exert their effects are
   unknown.
2. Greater understanding of the developmental causal sequence.
   More inclusive, more comprehensive models are needed, along with targeted
   studies to explain how apparent causes (eg, lactation, alcohol) fit these
   models. This understanding is critical to manipulating usual physiology
   to prevent clinical cancers.
3. The development of sound biologic, behavioral, and social data as
   bases for broad action.

This article will explore each of these challenges in turn.

Manipulable Causes of Breast Cancer

A discussion of manipulable factors associated with the development
of breast cancer serves to introduce and emphasize the importance of hormonal
physiology in the pathogenesis of this disease and especially in its prevention.

Three Initiators

Known manipulable risk factors--in particular, three that act as initiators--are
relevant for a minority of the population. Specifically, inherited genetic
susceptibility appears to be important in approximately 10% of cases, and
thus, manipulation of these changes and their functions (when defined)
is likely to be of limited benefit in the general population. For example,
the extreme intervention of prophylactic mastectomy is acceptable only
in genetically at-risk individuals.

Radiation is a second well-described initiator of breast cancer,[2]
and limitation of exposure, particularly of the prepubertal breast through
judicious use of therapeutic radiation for malignancy,[3] is an important
preventive strategy. Again, however, this strategy can be applied to only
small numbers of women. Nevertheless, there remains the possibility that
excess sensitivity to lower doses of diagnostic radiation may be an important
initiator of breast cancer in larger numbers of women.

Finally, cigarette smoking seems to be a likely initiator.[4] For many
health reasons, major efforts to reverse the upward trend in smoking among
younger female age groups are well justified.

More Recently Identified Associations

Four more recently identified associations have potentially greater
application to populations. Lactation for total periods of more than 6
months appears to be associated with a reduced risk for premenopausal breast
cancer.[5] This association has obvious widespread application, and is
particularly intriguing because of its physiologic mechanism of action
(discussed further below).[6]

It appears likely that the basis for the associations of alcohol[7]
and exercise[8] with breast cancer is hormonal modulation, whereas the
possible basis for the association of fresh fruits and vegetables with
breast cancer[9] is uncertain. There appear to be broad health benefits
and risks of both alcohol and exercise, and thus, the specifics of how
these factors work in breast cancer are more important. The possibility
that certain dietary constituents (fruits and vegetables) are protective
is also of significant interest because of their direct applicability to
public health. Although these associations (for alcohol, exercise, and
fruits and vegetables) are not strong (twofold increased risk for high
alcohol intake and a halving of risk for exercise and for high fruit and
vegetable consumption), they are worthy of attention because they can be
altered.

Developmental Biology of Breast Cancer

There has been a long-standing emphasis on the observation that many
breast cancers in individual Western women occur in the absence of attributable
risk factors.[10] Recently, for example, Brinton suggested that only half
of the disease can be explained on the basis of widely accepted risk factors.[11]
This perspective has left the impression that little is known about the
development of breast cancer, and this impression has been reinforced by
the fact that comprehensive models have rarely been presented or discussed.

Despite these circumstances, it is becoming clearer that we do have
a broad understanding of the major physiologic factors in breast cancer,
which provides a sound basis for prevention strategies. The initiation,
promotion, progression model from skin carcinogenesis experiments appears
to be a less useful model for breast cancer than is a physiologic model.
The basic skin-derived model, however, is useful in emphasizing the multistep
nature of malignancy development and the prolonged multiyear nature of
the process.

There now appear to be two key physiologic variables in breast cancer:
breast lobular maturation[6] and hormone exposure of breast tissue.[12]
Together, these variables provide a rational model that can account for
many epidemiologic observations (Table 1)
and that can form the basis for prevention strategies. Although the importance
of lobular maturation and hormonal exposure has been recognized increasingly
over the last 20 years, there remains the perception that these variables
(and their associated risk factors) do not provide a complete or satisfying
explanation for the disease.

Perhaps the major reason for this perception is the fact that, in a
majority of breast cancer cases, no exposure can be identified that is
considered critical in initiating the malignant transformation of breast
epithelia. Inherited genetic changes, radiation, and perhaps cigarette
smoking, described earlier, may each represent such exposures or act as
initiators, but together they appear to play a role in only a minority
of cases. (Clearly, the details and extent of impact of these factors deserve
further intense study.) This unidentified exposure-cause may not be as
critical numerically as has been suggested. Some experts suggest that,
together with spontaneous mutation rates, these exposures provide adequate
explanation for observed rates.

Lobular Maturation

The first critical physiologic variable--lobular maturation--was so
recognized by Russo and Russo in animal studies.[13] The supporting observation
in humans, however, preceded this laboratory work by a decade. Based on
international data, MacMahon et al clearly laid out the powerful relationship
of age at first full-term pregnancy and breast cancer risk later in life
(Figure 1).[14] How powerful a risk factor
this age is found to be clearly depends on the referent group. In Figure
1, a relative risk of 1.0 is assigned to nulliparous women. Note that
a woman completing her first pregnancy at age 15 has approximately 25%
the risk of a similar woman who completes her first pregnancy at age 35.

Perhaps the obvious occurrence of significant hormonal perturbations
during and after pregnancy (which do have a recognizable impact on breast
cancer risk) has detracted from an appreciation of the explanation for
this strong relationship recognized by the Russos.[6,13] In elegant work
following the suggestion that lobular maturation and permanent differentiation
of breast terminal end-bud cells were critical events in susceptibility
to breast malignancy in animals, the Russos studied the process in human
tissues (Table 2).[6] The coherent picture
suggested by their work is very compelling. Four types of breast lobules
are seen in human breasts. Lobule 1, which is the dominant type in women
before a pregnancy (and in nulliparous women throughout their premenopausal
years), is an undifferentiated structure with only several ductules. Cells
in these lobules appear to exhibit increased sensitivity to malignant transformation,
and appear to be the cells of origin of the most common breast cancers
of the ductal type.

The hormonal proliferative stimulation of menstrual cycles causes some
type 1 lobule cells to begin to differentiate; prior to pregnancy, young
women have a mixture of types 1 and 2 lobules, but the majority are type
1. The profound and sustained hormonal stimulation of a full-term pregnancy
results in differentiation of the cells in most breast lobules to semidifferentiated
(lobule 2) or fully differentiated (lobule 3) states. Although lobular
structures with the appearance of types 1 and 2 are seen in postmenopausal
women, these are, in fact, terminally differentiated structures without
the sensitivity to malignant transformation of these types seen in younger
women.

This theory of lobular development clearly suggests that a full-term
pregnancy is a defining event for breast cancer risk expressed perhaps
2 decades later. It is noteworthy that during the 10 to 20 years following
a first pregnancy, breast cancer risk is first greater and then lower than
that for nulliparous women (Figure 2);
consequently, for premenopausal women, age at first full-term pregnancy
is not an obvious risk factor.[15]