|
Advertisement |
|
|
||
 |
|
|||||
|
|||||
|
 |
||||||
|
||||||
|
||||||
|
© 2000 American Society for Clinical Oncology Clinical Progression of Breast Cancer Malignant Behavior: What to Expect and When to Expect itFrom the Department of Radiation and Cellular Oncology, Center for Advanced Medicine, University of Chicago, Chicago, IL. Address reprint requests to Samuel Hellman, MD, Department of Radiation and Cellular Oncology, 5758 South Maryland Ave, MC 9001, Chicago, IL 60637; email s-hellman{at}uchicago.edu
PURPOSE: Seemingly localized breast cancer is a heterogeneous mix of truly localized cancers and cancers with occult metastases. Our purpose is to determine the parameters of metastatic proclivity for the different clinical presentations of operable breast cancer and to present quantitative prognostic information useful to both doctors and patients. PATIENTS AND METHODS: A series of regionally treated breast cancer patients was analyzed to determine the likelihood and time of the appearance of clinical metastases for different clinical subgroups. Patients operated on at the University of Chicago from 1927 to 1987 for clinically regionally localized breast cancer, who received no systemic therapy as a part of their initial treatment, were included. Overall survival and distant disease-free survival in this mature series are analyzed. RESULTS: Metastagenicity, the metastatic proclivity of a tumor, increases with both tumor size and nodal involvement. This is also true for virulence, which is the rate at which these metastases appear. Each clinical group has a cured population, even those with extensive nodal involvement. A table provides a tool for determining the proportion of risk expended in each clinical group as a function of the distant disease-free survival. Whereas the likelihood of metastasis increases with tumor size and nodal involvement, the time to their appearance decreases. CONCLUSIONS: Breast cancer metastagenicity and virulence are heterogeneous even within clinically similar groups of operable breast cancer patients. Tumor progression is correlated with increasing tumor size and nodal involvement. Markers are needed to identify individual tumor virulence and metastagenicity.
ALTHOUGH MUCH HAS been written about breast cancer management, there has been less study of the natural history of regionally treated disease to guide in the rational design of multimodality treatment. Curves of survival, disease-free survival, and recurrence have been reported, but the dynamics of the expression of the metastatic phenotype need elucidating. If, as we shall present, the kinetics of metastatic proclivity differ as a function of size and nodal involvement, this variation (unless considered) will confound or confuse the interpretation of clinical studies. Previously, we have described and separated these tumor properties into two functions: metastagenicity (M) and virulence (V), where M is the ultimate likelihood of a tumor developing distant metastases and V is the rate at which this will occur.1-3 Understanding these functions will allow us to answer the clinical questions most important for determining individual patient care, such as: What is the eventual likelihood of distant disease developing in patients who have had treatment directed only to the primary tumor and regional lymph nodes? If the disease is to metastasize, when will this be clinically evident? When has this risk abated? And how will these kinetics be modified by adjuvant systemic treatment? Although this last question is not addressed in this paper, the natural history of regionally treated breast cancer without systemic therapy is a necessary prerequisite for answering this question. To best address the questions presented above, a series of patients who received only local and regional treatment and who were observed long enough for full expression of the natural history of the disease was required. This cohort must be of sufficient size so as to allow analysis and comparison of clinically relevant subgroups. In most of the studies currently reported, adjuvant hormonal manipulation or chemotherapy is administered to all but the most favorable patients, potentially obscuring the natural history of regionally treated disease. We have been fortunate in being able to study a series collected and observed by Dr. Donald Ferguson that includes a large number of patients treated only regionally. In this report, we analyze both M and V for the various clinical subgroups and use the results to answer those relevant clinical questions. We recognize, that even within the clinical groups created using tumor size and lymph node involvement, there is a spectrum of disease behavior, and thus, the data reflect an average for any clinical presentation. Even T1N0 breast cancer has a 15% to 25% likelihood of distant metastases, and so this stage must be a mixture of phenotypes.4-6 Using molecular markers, we have begun to analyze the spectra of malignancy within the various clinical groupings, correlating them with V and M to determine individual tumor phenotypes.2,7,8 However, until molecular markers can be used to predict M and V for individual patients, we must continue to use the best current indicators of prognosis, size and nodal involvement, and accept the variation within these clinical groupings. In a previous report, S.H. has described some of the concepts of survival curve analysis.9 That article emphasizes the need for a log-linear analysis to evaluate rates of mortality or, as in this study, rates of the first appearance of distant metastases. This current study expands these concepts to measure the M and V of different clinical groups. It also presents answers to the clinical questions concerning recurrence likelihood and pace of disease. Both the patient and the doctor want to know what to expect and when to expect it.
The records of 2,136 patients diagnosed with local-regional breast cancer who underwent mastectomy at the University of Chicago Hospitals between 1927 and 1987 were reviewed. One hundred seventy-seven patients also received systemic adjuvant chemotherapy, 299 had some form of endocrine manipulation, and 50 patients received both chemotherapy and endocrine therapy. To eliminate the confounding effects of systemic therapy, we have excluded these patients from the analysis. In addition, 90 patients in whom the axillary node status was unknown are also excluded. Follow-up was obtained from hospital records, contact with the patients’ families, physicians, and tumor registries. The majority of the patients were observed every 3 months during the first year after diagnosis and yearly thereafter. The median follow-up of the living patients is 145 months. Pathologic information was obtained from the original reports. The tumor size was determined by gross measurements of the excised lesion or by the largest tumor diameter on the histologic section. Tumor size is unknown in three patients. The median age of the patients is 55 years (range, 21 to 91 years). The patient and tumor characteristics are listed in Table 1. Further details of the overall patient population in this database were previously described.6,10-12
Overall survival (OS) is calculated from the date of mastectomy to death by any cause. If death is unknown, the patients are censored at the last known follow-up as alive. Distant disease-free survival (DDFS) was defined as the elapsed time from mastectomy to distant disease recurrence or death. Both distant disease recurrence and breast cancer–related death were considered events for DDFS calculations, and the patients were censored for death caused by intercurrent disease. The cause of death was unknown in 60 of the patients (3.8%) who were censored. Actuarial OS and DDFS curves were calculated using Kaplan-Meier calculations.13 All time intervals are calculated from the data of mastectomy. To study the effects of tumor size and the extent of nodal involvement on the rate of metastatic expression, the data for the different subgroups of tumor size and nodal involvement are presented in a log-linear graphic format. The initial slope can be considered a measure of V, whereas the plateau is used to ascertain M (100-plateau) ÷ 100. A visual fit is used, and we have defined the plateau by at least three annual points of similar value.
When no nodes are involved and the tumors are less than 2 cm in diameter, V is most protracted, and the plateau is not achieved until 22 years after treatment (Fig. 1A). This plateau is reduced from 80% to 64% and is reached sooner when one to three nodes are involved; thus, M has increased from .20 to .36 (Fig. 1B). V is also increased as measured by the initial slope, which increases from 1.5% of patients developing distant metastases per year to 5%. If four or more nodes are affected, M increases dramatically to .81, and V increases to 9% per year (Fig. 1C). Qualitatively similar but quantitatively greater effects of node involvement are found for tumors larger than 2 cm (Fig. 1D, 1E, and 1F). When no nodes are involved, these larger tumors increased M and V compared with smaller tumors. Like the smaller tumors, V and M increase with node number and, in all cases, are greater than observed for the smaller tumors.
In Table 2, we compare the effects of increasing tumor size on M and V. When there are no nodal metastases, tumors with diameters greater than 2 cm have twice the M and three times the V of  2 cm tumors. Similar results are seen when one to three nodes are present, but, when four or more nodes are involved, the difference is only in V. M is the same in either case, but even with small tumors, only about one fifth of the patients remain metastasis-free.
Increasing extent of nodal involvement causes a doubling in M when one to three nodes are involved compared with those patients without nodal involvement (Table 2). A similar further increase in M is seen when the nodal involvement increases to four or more nodes in the small tumors. Because M is so great in the larger tumors, only a modest further increase is possible. V increases by a factor of three with minimal nodal involvement and is further doubled with more extensive nodal disease for small tumors. As with M, the increase in V occurs with larger tumors once there is any node involvement but continues to increase with further nodal metastases. Considerations of V and M can be useful in separating patients into more homogeneous groups to predict prognosis, design treatment, and compare results (ie, as a part of clinical staging). Figure 2 plots V on the ordinate and M on the abscissa for the different clinical groups studied. A sigmoid relationship is present, with four groups discernible. The most favorable group are those patients with tumors less than 2 cm without nodal involvement. This group is currently designated as stage I in the American Joint Committee on Cancer (AJCC) breast cancer staging system. Of further interest is the point presented for very small tumors (T1a,b N0), which appears even lower on the sigmoid curve. These tumors are usually found only by mammography.
Rising along the sigmoid curve after T1N0 is a group that includes both tumors larger than 2 cm without nodal involvement and T1 tumors with one to three nodal metastases. Both of these are included in AJCC stage II(a). Significantly further up the sigmoid curve are those tumors greater than 2 cm with one to three nodes involved. These are also within AJCC stage II, although in the II(b) subgroup. Stage II even includes the most ominous group of operable tumors, those with four or more nodes involved. These can be II(a) or II(b) depending on tumor size. Thus, the current AJCC stage II is especially heterogeneous. The conflating of different prognostic groups when planning therapy or making treatment comparisons can by very misleading. Not included in Fig 2 are T1N4+ because this is an almost empty category with only 25 patients alive after 5 years. To summarize, we identify four different prognostic groups of operable breast cancer: (1) T1N0 (stage I); (2) T1N1-3 and T2-3N0; (3) T2-3N1-3; and (4) T1-3N4+ (the latter three are all included in stage II).
The OS and DDFS in patients with zero, one to three, and
The practical prognostic questions stated in the beginning of this article concern the likelihood of the appearance of distant metastases in patients who are disease-free for varying intervals. Table 3 presents the subsequent likelihood of distant failure for patients disease-free at 2 years. For example, for T1N0 patients free of disease at 2 years, there is a 5% failure rate by 5 years, 10% by 10 years, and further failure in the second decade of an additional 7%. Obviously, one can also use this table for other disease-free survival intervals; a T1N0 patient disease-free at 10 years has a 3% (13% to 10%) failure likelihood at 15 years and a similar failure rate between 15 and 20 years.
Another way to present these data is to calculate the percent of the risk expended after different times for the various clinical groups (Table 4). Again, to use T1N0 as an example, by 2 years only 15% of the risk has been expended, and even by 10 years, 35% of the risk of metastases still remained. In sharp contrast, tumors greater than 2 cm with four or more nodes involved, although having a far worse prognoses, have expended almost half (45%) of the risk of metastases in the first 2 years and 90% by 10 years. In the node-positive patients, the risk of recurrence is small after 10 years (5% to 10%), whereas in the node-negative group, 30% of the risk still remains.
In the Halstedean view of breast cancer pathogenesis, tumor spreads from the primary site either by direct permeation or through the lymphatics to lymph nodes and then to distant sites.14,15 This orderliness implies the success of a strategy of comprehensive contiguous primary and regional treatment. Although this strategy has resulted in many long-term survivors even in the most favorable group (stage I), 15% to 20% of patients develop distant metastases.4-6 Because of these distant failures despite negative nodes and satisfactory regional treatment, the Halstedean hypothesis has been called into question. An alternative has been advocated by Keynes,16 Crile,17 and, most eloquently, by Fisher18; the latter augmenting his argument with laboratory studies. This systemic hypothesis posits that tumor natural history is determined before clinical presentation. If metastasis are to occur, they have disseminated before detection. Lymph node involvement is said to serve only as a marker of metastatic likelihood but not as a source of distant metastases. Both the Halsted and the systemic hypotheses are inconsistent with the clinical data reported here and elsewhere.1,4-6,10,11 Not only is aggressive regional treatment not always successful in preventing distant disease in patients without nodal involvement, but many patients with regional lymph node involvement are cured by this therapy. Early detection by mammography improves survival by about 30%.19-22 This gives us an estimate of the proportion of patients with metastatic disease whose spread occurred during the time between mammographic and clinical detectability.1,3 Because the mean diameter of mammographically detected tumors in the two-county Swedish trial was 1.4 cm and the mean diameter of tumors clinically detected was 2.2 cm, then 30% of the tumors must have metastasized during this four-fold increase in tumor volume. Consistent with this finding are the low values for V and M for tumors less than 1 cm in diameter. The concept that breast cancers comprise a heterogeneous spectrum of malignant proclivities has been introduced and the evidence for it discussed.1,3 The spectrum hypothesis suggests that breast cancer is a mixture of tumors of different metastatic capabilities. The mechanism for this spectrum is malignant progression using evolutionary principles.23 The clinical data presented here reveal that with increasing tumor size and nodal metastases this spectrum becomes weighted toward greater malignant capacity, but each clinical group is itself heterogeneous. The increase in V and M seen with increasing size and nodal involvement quantifies this progression of malignancy. Although it is difficult to determine the exact values, a clear trend is apparent. Within each group, a plateau is reached dividing metastatic from nonmetastatic tumors. We have begun to try to identify the malignant potential within clinical groups using molecular markers of tumor heterogeneity, such as CD34 staining vessels7,24 and antibodies to nm23 and proliferating cell nuclear antigen.2,8 Other markers are currently being studied using the paraffin sections from this series.
In an earlier report, we showed no difference in 20-year survival, whether or not there was limited nodal involvement (one to three nodes), if the primary tumor was less than 2 cm in diameter.1 That report was of the entire series, including patients receiving systemic therapy, whereas this series includes only those patients treated regionally without adjuvant therapy. Thus, the difference may be because of the salutary effect of hormonal manipulation or chemotherapy. In the current analysis, any nodal involvement is of prognostic importance, but small tumors ( There have been several previous reports analyzing long-term survival in breast cancer.4,10,25 This report extends the recommendations for survival curve evaluation of Harris and Hellman9 by using log-linear analysis to determine not only the force of mortality but also the appearance of initial distant metastases. When studying OS, two slopes were identified. The first was assumed to reflect the force of mortality of the most malignant tumors. The cause of the second slope is multifactorial, with both competing mortality and a cohort with less malignant disease being possible contributors. Fox,26 using semi-logarithmic methods, analyzed National Cancer Institute relative survival data to show the second more gradual slope, which he interpreted to indicate the presence of a group of patients that continue to die of breast cancer but at a slower rate without a plateau of cured patients. Breast cancer treatment was believed never to result in cure. All patients would die of breast cancer if they did not succumb to intercurrent disease. In contrast, Tubiana et al27 also analyzed breast cancer metastatic appearance and demonstrated cured patients in each clinical group. The method of analysis was different than ours, and M and V were not quantified; however, many of the concepts reported here have been informed by those reports. In our study, the force of mortality in the OS curve is quite similar to V (the rate of appearance of distant metastases), but competing mortality obscures M when survival serves as the end point. Brinkley and Haybittle28 have reported that even after the survival curve became parallel to that of the age-adjusted peer population, there were breast cancer deaths. Our analysis is different from theirs because we study DDFS directly. Using this end point, the curves become flat, consistent with a cured population in each of the clinical groups. The plateau should not be interpreted to mean there are no late failures. We recognize that there are late metastases occurring beyond the plateau, but these comprise only 1.3% of the failures and 3.7% of the 20-year survivors. In addition to the points on the initial slope and on the plateau, there are intermediate points seen deviating from the slope but not on the plateau. These can be explained by the existence of some tumors with a more protracted V in any group. The extent of adherence to the initial slope gives some measure of the homogeneity of V within any group. Some intermediate values will occur with any analysis of a mixture of two populations as the proportion in each population changes. The semi-log presentation emphasizes the limitations of using 5-year survival as an indicator of outcome. For the clinical groups studied, 5-year data reflect V, 10-year results are a mixture of M and V of differing proportion depending on the clinical group, and 20-year survival is required for evaluation of M. There is a difference between the proportion of risk remaining and the absolute likelihood of metastases occurring after a defined period of disease-free survival. T1N0 cancers do not demonstrate half of their metastatic potential until between 5 and 10 years, but because M is only 20% in this group, only 10% of the patients surviving for this period will subsequently suffer metastases. In contrast, N4+ patients have experienced about half of their metastatic likelihood by 2 years and 90% by 10 years. However, because the absolute likelihood of patients developing distant metastases is so great, about 50% of those surviving disease-free at 10 years will eventually develop metastases (Table 4). That four or more positive nodes confers such a different prognosis gives support for the experimental aggressive therapy, such as bone marrow transplantation, currently being studied in this group. We conclude from this study that tumors progress in both aspects of malignancy, M and V, with increasing size and nodal involvement. This is a powerful argument for early diagnosis and effective regional treatment. Early diagnosis finds cancer with limited malignant capabilities. Effective regional treatment prevents the persistence of disease able to increase in malignant capacity. It is also apparent from these studies that all clinical groups of operable breast cancer have a very long natural history and the frequently used parameter of success, the 5-year survival (or even metastasis-free survival), measures only V. Despite this long natural history, patients and their doctors can get meaningful prognostic information from the duration of metastasis-free survival and the M and V measured for the particular clinical group. The results are optimistic because, despite some late relapses, many breast cancers are curable by regional treatment. Because our current staging system does not allow their identification, molecular markers will be very important to separate those patients cured by regional treatment from those requiring systemic therapy. Until such markers are available, we must use clinical data to determine prognosis and direct therapy.
1. Hellman S: Karnofsky Memorial Lecture: Natural history of small breast cancers. J Clin Oncol 12:2229-2234, 1994 2. Heimann R, Ferguson D, Recant WM, et al: Breast cancer metastatic phenotype as predicted by histologic tumor markers. Sci Am 4:224-229, 1997 3. Heimann R, Hellman S: Aging, progression, and phenotype in breast cancer. J Clin Oncol 16:2686-2692, 1998[Abstract] 4. Rosen PR, Groshen S, Saigo PE, et al: A long-term follow-up study of survival in stage I (T1N0M0) and stage II (T1N1M0) breast carcinoma. J Clin Oncol 7:355-366, 1989[Abstract]
5.
Rosen PP, Groshen S, Kinne DW, et al: Factors influencing prognosis in node-negative breast carcinoma: Analysis of 767 T1N0M0/T2N0M0 patients with long-term follow-up. J Clin Oncol 11:2090-2100, 1993 6. Quiet CA, Ferguson DJ, Weichselbaum RR, et al: Natural history of node-negative breast cancer: A study of 826 patients with long-term follow up. J Clin Oncol 13:1144-1151, 1995[Abstract]
7.
Heimann R, Ferguson D, Powers C, et al: Angiogenesis is a predictor of long-term survival for patients with node-negative breast cancer. J Natl Cancer Inst 88:1764-1769, 1996
8.
Heimann R, Ferguson D, Hellman S: The relationship of nm23 and angiogenesis to the metastatic propensity of node-negative breast cancer. Cancer Res 58:2766-2771, 1998 9. Harris JR, Hellman S: Observation on survival curve analysis with particular reference to breast cancer treatment. Cancer 57:925-928, 1986[Medline] 10. Ferguson DJ, Meier P, Karisson T, et al: Staging of breast cancer and survival rates: An assessment based on 50 years of experience with radical mastectomy. JAMA 248:1337-1341, 1982[Abstract] 11. Quiet CA, Ferguson DJ, Weichselbaum RR, et al: Natural history of node-positive breast cancer: The curability of small cancers with limited number of positive nodes. Oncol 14:3105-3111, 1996 12. Heimann R, Ferguson D, Powers C, et al: Race and clinical outcome in breast cancer in a series with long follow-up evaluation. Clin Oncol 15:2329-2337, 1997 13. Kaplan EL, Meier P: Nonparametric estimation from incomplete observations. J Am Stat Assoc 53:457-481, 1958 14. Halsted WS: The results of operation for the cure of cancer of the breast performed at the Johns Hopkins Hospital from June 1889 to January 1894. Bull 4:297-319, 1895 15. Halsted WS: The results of radical operations for the cure of carcinoma of the breast. Ann Surg 46:1-6, 1907 16. Keynes G: Carcinoma of the breast, the unorthodox view. Proc Cardiff M Soc, 40 1954 17. Crile GJ: Biological Considerations of the Treatment of Breast Cancer. Springfield, IL,Thomas, 1967
18.
Fisher B: Laboratory and clinical research in breast cancer: A personal adventure—The David Karnovsky Memorial Lecture. Cancer Res 40:3863-3874, 1980
19.
Chu KC, Smart CR, Tarone RE: Analysis of breast cancer mortality and stage distribution by age for the Health Insurance Plan clinical trial. J Natl Cancer Inst 80:1125-1132, 1988 20. Miller A, Baines CJ, To T, et al: Canadian national breast screening study: Breast cancer detection and death rates among women aged 50 to 59 years. Can Med Assoc J 147:1477-1488, 1992[Abstract] 21. Tabar L, Fagenberg G, Duffy SW, et al: Update of the Swedish two-county program of mammographic screening for breast cancer. Radiol Clin North Am 30:37-59, 1992 22. Nystrom L, Rutqvist LE, Wall S, et al: Breast cancer screening with mammography: Overview of Swedish randomized trials. Lancet 341:973-978, 1993[Medline] 23. Hellman S: Darwin’s clinical relevance. Cancer 79:2275-2281, 1997[Medline] 24. Heimann R, Ferguson D, Gray S, et al: Assessment of intratumoral vascularization (angiogenesis) in breast cancer prognosis. Breast Cancer Res Treat 52:147-158, 1998[Medline] 25. Adair F, Berg J, Joubert L, et al: Long-term follow-up of breast cancer patients: The 30-year report. Cancer 33:1145-1150, 1974[Medline] 26. Fox MS: On the diagnosis and treatment of breast cancer. JAMA 241:489-494, 1979[Abstract] 27. Koscielny S, Tubiana M, Le MG, et al: Breast cancer: Relationship between the size of the primary tumour and the probability of metastatic spread. Br J Cancer 49:709-715, 1984[Medline] 28. Brinkley D, Haybittle JL: The curability of breast cancer. Lancet 2:95-97, 1975[Medline] Submitted February 12, 1999; accepted September 9, 1999. This article has been cited by other articles:
|
|||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
|||||||||||
|
|||||||||||
|
|
Copyright © 2000 by the American Society of Clinical Oncology, Online ISSN: 1527-7755. Print ISSN: 0732-183X
|
TOP
ABSTRACT
INTRODUCTION
four nodes or in patients with > 2-cm tumor and (D) no nodes; (E) one to three nodes ; and (F) 
