SEARCH

SEARCH BY CITATION

Keywords:

  • breast cancer;
  • racial traits;
  • African American;
  • Caucasian American;
  • host-tumor interaction

Abstract

  1. Top of page
  2. Abstract
  3. EPIDEMIOLOGIC PERSPECTIVE: A MIXTURE MODEL
  4. CLINICAL PERSPECTIVE: A HIERARCHICAL PROGNOSTIC MODEL
  5. LOOKING FOR A ‘UNIFYING MODEL’
  6. REFERENCES

Since the 1970s, overall age-adjusted breast cancer mortality rates in the U.S. have been higher among African American (AA) women than among Caucasian American (CA) women. The racial disparity is not fully explainable based on socioeconomic factors. Suspected biologic factors underlying this trend may be interpreted by both epidemiologic and clinical perspectives. Descriptive epidemiologic studies suggest that breast cancer may be a mixture of at least 2 main diseases and/or causal pathways. The first breast cancer is early-onset, with peak incidence near age 50 years and generally more aggressive outcome. The second breast cancer is late-onset, with peak incidence near age 70 years and more indolent course. The early-onset type of breast cancer is overrepresented among AA women compared with CA women. Clinical studies suggest that the course of breast cancer may be characterized by a common pathway through sequential dormant and active states eventually resulting in clustered appearance of clinical metastases. A balance between tumor and host traits influences the pace of the common pathway. Therefore, the recurrence risk profile of a single patient is seemingly determined by a specific mix of hierarchical prognostic factors, resulting from the unique genetic, environmental, or behavioral traits of that individual, which may be affected by race-related factors. We suggest that the components of the AA versus CA disparity not attributable to socioeconomic factors are a particular case of the more general issue of host-tumor interaction and that epidemiologic and clinical views are complementary; each is observing biologic parameters, which are not completely captured by the other. A ‘unifying hypothesis’ incorporating findings from genetics, epidemiology, and clinical studies should be aggressively pursued. Cancer 2007. © 2007 American Cancer Society.

Population-based statistics in the U.S. indicate that overall age-adjusted breast cancer mortality rates are higher among African American women (AA) than among Caucasian American women (CA), and the disparity is increasing.1 There is a mortality disadvantage of between 1.5–fold to 2.2-fold that first appeared in the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) data in the mid-1970s and has been widening ever since.2, 3

The differential access to medical care and screening and even disparities in disease management (diagnostic procedures, treatment decisions, etc) are usually considered the source of the different race-related outcomes. A recent review on this issue,4 addressing the role of patient-, provider-, and health system-level factors and discussing practical approaches to optimize them to lower the AA disadvantage, concluded that shifting policy orientation from an emphasis on rugged individualism to one valuing social justice and shared responsibility for health is needed. Socioeconomic explanations are indubitably very important, but they are not sufficient. They can hardly explain the fact that, although AA mortality is worse than CA mortality for younger women, mortality curves cross at 57 years,3 with AA women demonstrating better survival than CA women later on. Indeed, it should be assumed (and verified) that older AA women do not suffer unfavorable social conditions, but rather benefit from better conditions than CA women. Moreover, 2 studies of breast cancer survival rates among AA and CA women in the U.S. Department of Defense (DoD) healthcare system, in which medical care is available at no cost to all eligible beneficiaries, showed that AA women still have worse overall survival. However, 1 article concluded that equal access to healthcare would reduce the mortality difference5 and another reported that the racial disparity in mortality has been widening in the DoD system since approximately 1980.6 Therefore, although it is obviously essential to equalize access to healthcare, solving ethnic disparities may still require understanding and effectively addressing other biology-based racial differences. Socioeconomic and biologic factors have established associations in breast cancer that can reasonably be expected to work together to determine race-based breast cancer risk, biology, and outcome.3

When addressing the biologic side of the disparity question by attempting to integrate views from epidemiologic and clinical studies, one realizes that a major shift in our conceptual model of breast cancer is demanded. Bridging epidemiologic and clinical data suggests the need for a unifying model of the disease that is valuable beyond the AA-versus-CA disparity question. Within the new framework, it is conceivable that entirely novel ways to decrease breast cancer mortality may emerge. Such a line of investigation together with the indispensable “shift in policy orientation” may pursue better results for all women, with no prejudice from racial or ethnic factors.

EPIDEMIOLOGIC PERSPECTIVE: A MIXTURE MODEL

  1. Top of page
  2. Abstract
  3. EPIDEMIOLOGIC PERSPECTIVE: A MIXTURE MODEL
  4. CLINICAL PERSPECTIVE: A HIERARCHICAL PROGNOSTIC MODEL
  5. LOOKING FOR A ‘UNIFYING MODEL’
  6. REFERENCES

Most clinicians and investigators view breast cancer as a single (sequential or linear) long-term biologic process7–9 resulting in a disease that remains local throughout its course to one that is systemic when first detectable or later.10 An obligate epidemiologic consequence of this linear breast cancer model is a steadily rising age-specific incidence rate curve on a log rate and log age (log-log) scale.11, 12 However, unlike many solid tumors, age-specific breast cancer rates increase rapidly until age 50 years, pause, and then continue to rise at a slower pace (Fig. 1A).13, 14 The midlife pause in the age-specific rates has been termed Clemmesen's hook,15 is absent in male breast cancer, and has been attributed to hormonal changes occurring during menopause.16

thumbnail image

Figure 1. Age incidence (rates and frequency) and prognostic patterns among women with invasive breast cancer in the National Cancer Institute's (NCI) Surveillance, Epidemiology, and End Results (SEER) 13-Registry Database (1992–2003). (A) Age-specific incidence rates for all cases combined, estrogen receptor (ER)-positive, and ER-negative expression. (B) Age distributions at diagnosis (density plots) for all cases combined, ER-positive, and ER-negative expression. (C) Hazards ratios for breast cancer death for all cases combined, ER-positive, and ER-negative expression. (Reprinted with permission from Anderson WF, Matsuno RK. Breast cancer heterogeneity: a mixture of at least two main types. J Natl Cancer Inst. 2006;98:948–951. Note this figure has been slightly modified.)

Download figure to PowerPoint

Descriptive epidemiologic studies suggest that Clemmesen's menopausal hook may reflect the mixture of 2 different incidence rate curves according to age at onset and hormone dependence (Fig. 1A),7, 16–20 that have been conceptualized as juxtaposed early-onset (estrogen receptor [ER]-negative) and late-onset (ER-positive) breast cancer populations (Fig. 1B).17, 18 The first breast cancer has a peak incidence near age 50 years, whereas incidence of the second breast cancer peaks near age 70 years.

Distinct incidence patterns (rates and age distributions) (Figs. 1A and 1B) appear to be linked to unique prognostic patterns as early-onset and late-onset breast cancers display ‘aggressive’ and ‘indolent’ clinical behavior, respectively (Fig. 1C).21 Early-onset (ER-negative) annual mortality hazard rates rise to a sharp peak of approximately 7% to 8% per year approximately 2 years after initial breast cancer diagnosis, and then decline. Late-onset (ER-positive) hazards rates for death have a constant long-term rate of approximately 1.5% to 2% per year and lack a sharp 2-year peak. Falling early-onset and constant late-onset mortality hazards rates cross at approximately 7 years. Of note, incidence and prognostic patterns appear to be similar in both AA and CA women, and approximately parallel to ER-negative and ER-positive patterns (Fig. 2), respectively. Then we may hypothesize that racial disparity between AA and CA women may in part result from the occurrence that the early-onset type of breast cancer is overrepresented among the former group compared with latter.

thumbnail image

Figure 2. Age incidence (rates and frequency) and prognostic patterns among women with invasive breast cancer in the National Cancer Institute's (NCI) Surveillance, Epidemiology, and End Results (SEER) 13-Registry Database (1992–2003). (A) Age-specific incidence rates for all cases combined, AA and CA women. (B) Age distributions at diagnosis (density plots) for all cases combined, AA and CA women. (C) Hazards ratios for breast cancer death for all cases combined, AA and CA women.

Download figure to PowerPoint

AA versus CA differences in breast cancer outcome display some peculiar secular trends. Before the 1970s, there was little disparity noted in the overall mortality of AA and CA. In the late 1970s, the calendar period curves demonstrated that AA experienced a sharp increase in breast cancer mortality, whereas CA experienced a continuous decline that had begun during the 1950s, yet with a moderation in decline occurring synchronously with the AA increase. Thus, the racial disparity in breast cancer mortality, dating back to the late 1970s, appears largely attributable to calendar period trends rather than birth cohort effects.3 Such calendar period trends are likely to reflect the impact of new medical interventions, such as screening and treatments. One might speculate that the decline in calendar period mortality from the 1950s to the late 1970s, evident for all racial groups, is due to an increase in breast cancer awareness with progressive detection of smaller palpable tumors and mammography screening development. However, the sharp increase in mortality among AA in the late 1970s occurred just when the trend would have been quite the opposite.

A possible explanation of this dynamics of divergence in AA versus CA mortality may be related to the different incidence patterns (rates and age distributions) among these groups. Indeed, some investigations22–24 have described a counterintuitive excess mortality in the first 6 to 8 years after the initiation of mammographic screening (the so-called ‘mammography paradox’) in randomized trials to determine the value of early detection for women ages 40 to 49 years, with declines in mortality thereafter.23, 24 Given that AA have higher proportions of premenopausal cancers, this may potentially account for the sharp increase in calendar period slope of mortality after the introduction of mammography screening in the late 1970s. In other words, the mammography screening introduction may be considered as a probe revealing different traits in the host-disease balance in AA and CA, which are reflected by the change in mortality dynamics.

CLINICAL PERSPECTIVE: A HIERARCHICAL PROGNOSTIC MODEL

  1. Top of page
  2. Abstract
  3. EPIDEMIOLOGIC PERSPECTIVE: A MIXTURE MODEL
  4. CLINICAL PERSPECTIVE: A HIERARCHICAL PROGNOSTIC MODEL
  5. LOOKING FOR A ‘UNIFYING MODEL’
  6. REFERENCES

Because the growing AA-versus-CA mortality differences might be closely related to an increase in breast cancer screening, it is appropriate to study these differences in the light of a recently proposed model of breast cancer clinical progression that may be able to explain the mammography paradox.

Anecdotal evidence from ancient times until the present day has hinted at the possibility that, left untreated even in the advanced stage, many breast cancers do not appear to progress in a predictable and inexorable way and that trauma might provoke the outgrowth of breast cancers that until then had existed in some kind of equilibrium with the host.25 It is likely that similar observations fueled popular ‘myths’ such as “surgery provokes the tumor to spread” that is reported by breast surgeons practicing in Africa, where the prognosis of breast cancer patients is particularly poor,26 or as “cancer spreads when the air hits it” that is surprisingly widely held and believed in the U.S. by 61% of AA and 29% of CA individuals.27

Although such folk beliefs likely contribute to delays in seeking treatment and poor prognosis, they may contain some truth regarding breast cancer progression that is now being uncovered by recent research, which supports a model for breast cancer clinical progression where primary tumor removal may accelerate the appearance of distant metastasis, mainly in premenopausal women.28 According to this model, the clinical course of breast cancer is characterized by dormancy of subclinical metastases.29 The duration of this dormancy assumes specific values, resulting, at primary tumor removal, in a surgery-gated multipeaked recurrence risk pattern.30 This surgery-gated pattern, moreover, is different for premenopausal and postmenopausal patients28 (Fig. 3).

thumbnail image

Figure 3. Hazards ratio for disease recurrence (local and distant) for 517 premenopausal patients and 656 postmenopausal patients with resectable breast cancer undergoing mastectomy only as first-line treatment. The multipeaked pattern reveals the occurrence of distinct biologic states for subclinical metastases, whereas the different recurrence dynamics suggests menopausal-related tumor traits.

Download figure to PowerPoint

In animal models, dormant tumor foci in seeded tissues have been observed both as single nondividing cells31, 32 and as avascular foci that demonstrate balanced proliferation and apoptosis.33, 34 Other more complex dormant states cannot be excluded. It is also clear that the presence of a primary tumor exerts some kind of homeostatic effect on metastases,35, 36 as primary tumor removal may induce acceleration of growth in some residual cancer deposits.37, 38

In the proposed model (outlined in Fig. 4),39, 40 sequential transitions between dormant states eventually result in progressive appearance of clinical metastases, and primary tumor surgical removal results in sudden synchronization and acceleration of metastasis development by stimulating some tumor cells to proliferate and/or eliciting angiogenesis in some avascular micrometastases.35, 36, 38 The metastatic process is similar for both the premenopausal and postmenopausal states, although postmenopausal patients apparently bear a higher load of micrometastases that are insensitive to the accelerating effect of the primary tumor surgical removal.30 Confirmation of the main traits of this picture arose from the analysis of the recurrence risk for patients receiving adjuvant cyclophosphamide, methotrexate, and 5-fluorouracil, in whom the effective cytotoxic treatment targeting proliferating cells equalizes the early 4-year hazards rate pattern for both premenopausal and postmenopausal patients.41 Also, it is interesting to note that the discrete features of the temporal pattern of the recurrence risk turn into discrete features of mortality, even if death events did not simply parallel the corresponding recurrence events.42 Details of the process supporting the acceleration of metastasis by surgical resection are not yet well understood. Both the surgical trauma per se (activation of the wound-healing program) and the primary tumor removal (disruption of homeostatic equilibrium) are strongly suspected to be relevant. Angiogenesis induction by shifted balance between enhancing and inhibiting factors is considered a major, but not exclusive, molecular mechanism for dormancy interruption.33, 43

thumbnail image

Figure 4. (A) Outline of breast cancer metastasis development. Dormant tumor foci in seeded tissues may be single nondividing cells (S1) and avascular foci that show balanced proliferation and apoptosis (S2). Orderly transitions between such dormant states eventually result in progressive appearance of clinical metastases (M). (B) Primary tumor presence is able to exert some kind of homeostatic effect on metastases, thus restraining the transitions between dormant states. Its removal may stimulate tumor cells to proliferate and/or may elicit angiogenesis, thus resulting in sudden synchronization and acceleration of the metastatic process.

Download figure to PowerPoint

Whereas timing of recurrence risk peaks is rather constant for all studied patient subsets, risk levels are greatly influenced by tumor and host traits. For example, menopausal status at surgery21, 28 and ER content21 were reported to have dissimilar effects on early and late recurrence and mortality rates. Therefore, the risk profile of a single patient is seemingly determined by a unique mix of tumor-host factors controlling the ability of tumor cells to progress throughout successive dormant states. Some factors may result from the tumor-originating cell (gene profiles suggesting transformation of different cells from breast acini have been recently reported),44 whereas others may result from tumor-host relations, as proved by studies in experimental settings, in which the tumor environment is critical in tumor development and behavior.45–47

For breast cancer, we have many reasons to suspect that the host plays an important role in determining individual cancer patient outcomes. Premenopausal breast cancer is a disease in dynamic hormonal balance with the woman who bears this disease.48, 49 Breast cancer sex hormone receptor content, which predicts outcome, varies during each menstrual cycle,50 as does the labeling index of normal breast cancer epithelium51 and breast cancer cell growth factor vascular endothelial growth factor (VEGF) content,52 as well as the actual growth rate of breast cancer.53 In this context, genetic or demographic peculiarities of the host, directly or indirectly related to the ethnic group, may be relevant to breast cancer development and clinical course. In fact, luteal phase estradiol levels are reported to be higher in AA compared with CA women.54, 55 AA women undergo menarche earlier than CA women56 and also tend to cycle more rapidly than their CA counterparts.57 Each of these differences in the cycle and its hormones could contribute to outcome differences. Indeed, there are important indications that the menstrual cycle stage of resection may impact breast cancer curability.58–60 It should be stressed here that host factors are significant whoever the woman is, and that ethnicity may be considered as a factor favoring some peculiar host traits.

LOOKING FOR A ‘UNIFYING MODEL’

  1. Top of page
  2. Abstract
  3. EPIDEMIOLOGIC PERSPECTIVE: A MIXTURE MODEL
  4. CLINICAL PERSPECTIVE: A HIERARCHICAL PROGNOSTIC MODEL
  5. LOOKING FOR A ‘UNIFYING MODEL’
  6. REFERENCES

Epidemiologic studies and clinical research studies display a few apparent disagreements. In particular, they interpret multipeak hazards rate patterns for recurrence and death differently. With regard to this, it should be emphasized that epidemiologic studies have used breast cancer mortality patterns as their main endpoint, whereas clinical studies have focused more on breast cancer recurrences. Recurrence and death from breast cancer are not superimposable endpoints because death events do not simply parallel the corresponding recurrence events.42 Conversely, a common trait shared by both epidemiologic and clinical viewpoints is that breast cancer is a multifaceted disease in which postdiagnosis clinical patterns are established long before breast cancer detection, when both host and tumor jointly played a (presumably) long drama.

The ‘2 diseases’ epidemiologic model identifies 2 clinicopathologic entities, labeled as ‘aggressive’ and ‘indolent,’ with different incidence and mortality patterns.21 The definition of ‘aggressive’ and ‘indolent’ disease results from the parallel concurrence of all favorable or all unfavorable values of a variety of tumor and host factors (host race, tumor size, axillary lymph node status, tumor grade, receptor content, etc). However, other combinations of some favorable and other unfavorable traits are not uncommon, thus delineating a spectrum of ‘intermediate outcome’ disease. Despite this remark and the risk of being a classification artifact, the ‘2 diseases’ assumption is a reasonable model of epidemiologic data. Moreover, it is even supported biologically by the finding that 2 main subtypes of breast cancer (luminal A and basal-like) identified by gene expression profiling have been recently confirmed.61

Clinical studies provide evidence that recurrence risk peaks have rather constant timing for different patient subsets,28, 30, 41 thus suggesting that the disease course after primary tumor surgical removal basically follows a common pathway with well-defined steps (dormant states).28, 39, 41 Moreover, the finding that, within this common tempo, the risk levels at a certain time are influenced by tumor and host traits28, 39 suggests that the pace (time to transition between states) of the common pathway is governed by preexisting risk factors. In addition, in the clinical approach it appears that: 1) there is a hierarchy of prognostic factors (eg, axillary lymph node status nearly abolishes the importance of other factors for lymph node-positive patients), 2) some factors display associations (eg, proliferation levels correlate directly with tumor grade and inversely with steroid receptors)62, and 3) each factor may differently influence the risk level for a given event during specific spans within the disease course (eg, whereas the ER status profoundly influences the death risk during the first 4 years after surgery,21, 63 the proliferation level measured by thymidine labeling index demonstrates a definitely less significant impact [unpublished data]). Therefore, the clinical perspective suggests that risk factors (from both tumor and host traits) may be organized hierarchically as measured by their connection with ‘aggressiveness.’ Indeed, we may reasonably assume that for a given patient a unique mix of tumor and host factors may control the ability of tumor cells to progress throughout successive dormant states, eventually resulting in clinical recurrence. Equally likely, the disease course after recurrence may also be predicted by factors (not necessarily the same as above) influencing both disease progression and sensitivity to treatments.

These 2 critical spans (from tissue seeding to clinical recurrence and from clinical recurrence to death) should be prominently influenced by the risk factors having higher hierarchical rank, whereas others would acquire relevance only in the absence of the more powerful ones. In addition, it is likely that a few clusters of risk factors, including some hierarchically high-level factor, may occur at higher frequency than others and may have specific distribution within the population according to age, race, lifestyle, and so on, thus calling for the label of autonomous clinical entity. Finally, it would not even be surprising to detect that the hierarchical arrangement of risk factors may parallel corresponding gene expression patterns, such as those providing a basis for classifying human breast tumors into subtypes that express at various levels molecular traits (eg, ER, progesterone receptor [PR], HER-2/neu, and epidermal growth factor receptor [EGFR]).63–65

Therefore, within this frame of reference, the ‘2 diseases’ assumption, which is derived from population-based incidence and mortality data, captures the 2 opposite sides of the breast cancer spectrum. We suggest that the ‘mixed proportions of hierarchy-organized factors’ suggested by clinical investigations on recurrences and the ‘2 diseases’ mixture suggested by epidemiologic studies on incidence and mortality may be 2 apparently different ways to see the same phenomenon. A ‘unifying hypothesis’ incorporating findings from genetics, epidemiology, and clinical studies is at present lacking and should be aggressively pursued.

We conclude that the biologic component of the AA-versus-CA disparity question, from which we started, is seemingly a particular case of the more general question of how the host and the tumor interact. Clarifying elements for both questions may plausibly emerge from the investigation of mechanisms underlying tumor dormancy states and transitions between them. We believe that the detection of factors governing the late natural history of breast cancer will allow us to go back to tumor development traits, thus getting in touch with the disease sources. In this context, the host contribution in determining its own disease will evolve from the current vague ‘milieu’ to a more detailed picture in which race-related and other host-specific biologic differences will most likely become clearer.

REFERENCES

  1. Top of page
  2. Abstract
  3. EPIDEMIOLOGIC PERSPECTIVE: A MIXTURE MODEL
  4. CLINICAL PERSPECTIVE: A HIERARCHICAL PROGNOSTIC MODEL
  5. LOOKING FOR A ‘UNIFYING MODEL’
  6. REFERENCES