To gain insight into the factors that contribute to the more favorable prognosis associated with recurrence limited to bone in patients with breast carcinoma, the authors analyzed the number of sites of initial involvement identified on radionuclide bone scans in relation to long term outcome.
Records of 641 patients with clinical Stage I–III breast carcinoma that originally was diagnosed in 1974–1985 were reviewed. During follow-up, 295 patients (46%) experienced distant recurrence, including 116 with bone as the sole initial site of metastatic disease. Radionuclide bone scans identified the initial site(s) of recurrence in 113 of these latter 116 patients, and these studies were categorized by the number of skeletal lesions subsequently confirmed as metastases (1, 2, or ≥ 3). Survival from time of recurrence and time of original diagnosis was analyzed using Kaplan–Meier methods, and factors associated with recurrence and mortality were examined using logistic and Cox regression.
Median survival from time of recurrence was 35 months in the patients with bone-only metastases, compared with 11–26 months for all other sites of visceral recurrence exclusive of bone. Number of positive lymph nodes and estrogen receptor status were the only predictive variables for recurrence. Median survival from time of recurrence and time of original diagnosis for the 3 bone scan categories was: 1 lesion (n = 47), 53 and 86 months; 2 lesions (n = 22), 38 and 68 months; and ≥ 3 lesions (n = 44), 22 and 58 months (P < 0.0001 and P < 0.005 for 1 and 2 lesions vs. ≥ 3). In the “bone-only” group, the number of scan lesions was the strongest predictor of length of survival.
Aside from regional lymph nodes, bone is the most common distant site of metastatic breast carcinoma.1, 2 Although metastatic spread to most end organs occurs by similar mechanisms,1, 3, 4 and survival is almost always poor in patients with extensive disease involving multiple organs,2, 3 bone-only involvement with metastatic breast carcinoma appears to carry a somewhat less dire prognosis.5–9 This latter observation may be a specific reflection of a subpopulation of breast carcinoma cells with relatively indolent biologic characteristics, or relate to other host factors. However, the relation between this tumor behavior and bone as a site of distant spread has not been defined.
In a previous review of bone scintigraphic findings in breast carcinoma patients without known metastases, we documented the association between the number of new lesions and the likelihood of metastatic disease.10 Prevalence of metastatic disease increased from 11% for scans with one new lesion to 100% for those with ≥ 5 new abnormalities. In the course of that review, we noted a longer survival for patients with fewer bone metastases at initial presentation in comparison to those with early evidence of widespread bony involvement. This report specifically examines the prognostic significance of the pattern of bone recurrence in patients with breast carcinoma, comparing long term survival in patients with limited bone recurrence, multiple sites of bone involvement, and other initial sites of visceral metastatic disease.
PATIENTS AND METHODS
Six-hundred forty-one patients with newly diagnosed breast carcinoma who were recruited for several adjuvant chemotherapy protocols between 1974 and 1985 represent the population examined in this review. This is the same population from which our earlier review of the association of number of new bone scan lesions and likelihood of skeletal metastases was drawn.10 Most of the patients (n = 507; 79%) participated in a randomized trial with 4 treatment arms (cytoxan and adriamycin once every 3 weeks for 15 weeks (n = 131) or 30 weeks (n = 128); methotrexate and 5-fluorouracil (first and eighth day of each month for 8 months) alone (n = 130) or with cytoxan (first 14 days of each month) (n = 118).11–13 All patients were initial clinical American Joint Committee on Cancer Stage I–III and underwent either a lumpectomy or a simple, modified radical, or radical mastectomy with axillary lymph node dissection before protocol entry. Most patients had either a modified radical or radical mastectomy. Initial staging evaluation included bone scintigraphy, liver scintigraphy during the early stages and abdominal computerized tomography during the later stages of the protocol, routine liver biochemistry tests, and, during the last 9 years of the protocol, carcinoembryonic antigen [CEA] screening.
Patients were categorized according to the clinical, surgical, and histologic results from their initial evaluations (Table 1), and the clinical, imaging, and histopathologic data showing credible evidence of local, locoregional, or distant recurrence of breast carcinoma. Local failure was defined as recurrent disease documented at either the primary site or the skin or chest wall in direct contiguity with the location of previous surgery. Locoregional recurrence included involvement of ipsilateral regional lymph nodes, including axillary and supraclavicular groups. Involvement of other ipsilateral lymph nodes such as cervical lymph nodes or any contralateral lymph nodes was considered distant recurrence. Although lymph nodes and other sites of distant recurrence sometimes were identified and confirmed from biopsy and autopsy results, clinical presentation and diagnostic imaging findings (radiographic, scintigraphic, sonographic, computed tomography [CT], and magnetic resonance imaging [MRI]) were commonly used to characterize the sites of recurrence.
Table 1. Clinical and Staging Information
No. of Patients
AJCC: American Joint Committee on Cancer.
AJCC Clinical Stage
No. of positive lymph nodes
The site of initial presentation of distant metastatic disease was determined from a review of available imaging, biopsy, and surgical results. The initial site of distant recurrence was defined as the organ or location in which an abnormality was first identified that subsequently was confirmed by characteristic imaging or histopathologic criteria to reflect metastatic disease, even if that confirmation occurred after other sites had been identified and shown to be malignant. Metastatic involvement at 2 or more sites was considered concurrent if all were documented within a 30-day interval. Tabulation of sites of distant recurrence was performed without reference to previous local or locoregional recurrence.
For patients whose initial recurrence was in bone, the initial site or sites of involvement were determined from an intensive review of all bone scintigraphic and radiographic data. Three hundred twelve patients had bone scintigraphic studies performed at our institution, and each study from the time of breast carcinoma diagnosis was reviewed by one of the authors (A.F.J.). As previously reported, the studies performed before 1983 involved total body anterior and posterior sweeps at 3 hours after administration of 740–1110 MBq technetium-99m diphosphonate compounds, with selected spot images of sites that were abnormal on the total body images. Studies performed from 1983 onward were composed of overlapping spot images of the total body, excluding the distal extremities, performed with a large field of view gamma camera and all-purpose or high resolution collimators.10 The date of initial appearance of each new lesion was recorded, which in a few instances (n = 5) was earlier than reported on the original scan interpretation. In most such cases, a second reviewer (W.D.K. or A.D.V.) was consulted, and a final determination of the earliest date a lesion could be detected was reached by consensus. The remaining 329 patients had bone scans performed at other institutions. Studies reported as abnormal were reviewed whenever possible by one of the authors or a radiologist or nuclear medicine physician at the original hospital to confirm the earliest date of new lesion detectability. However, because it generally was not possible to perform a complete and detailed review of all bone scans of each patient followed at other institutions, an expansion of our earlier analysis of number of scan lesions versus prevalence of metastases10 was deemed impractical and was not performed.
The etiology of each new bone scan finding was determined to be either benign or malignant based on review of available correlative radiologic studies, as well as biopsy, surgery, autopsy, and clinical follow-up data. Metastatic disease at a site was considered proven by one or more of the following: 1) radiographs demonstrating a characteristic lytic or blastic lesion, with subsequent follow-up films showing either changes consistent with disease progression or healing in response to therapy; 2) follow-up scintigraphy showing increasing intensity at the site in question and/or new scan abnormalities characteristic of metastatic disease, usually associated with typical radiographic findings; 3) positive histopathology at biopsy, surgery, or autopsy; 4) persistence of initial abnormal scintigraphic and radiographic findings with progression of visceral metastatic disease at other sites. Patients with bone-only metastases and bone scintigraphy at the time of diagnosis of recurrence were further categorized relative to the number of new scan abnormalities (one, two, or three or more).
Based on all available information, patients were coded into one of the following four groups as of the time of last follow-up.
2Local or locoregional recurrence only; no evidence of distant metastatic disease.
3Distant recurrence of metastatic breast carcinoma at one or more sites.
4Censored as of a specified date based on one of the following: 1) diagnosis of breast carcinoma in the contralateral breast, unless the new malignancy could be characterized histopathologically as a metastasis; 2) development of a distinct second primary malignancy; 3) initiation of chemotherapy or hormonal therapy for presumed metastatic disease without adequate confirmation on initial or follow-up studies; 4) reported recurrence (usually from chart notes or other secondary sources) that could not be confirmed. Patients in categories 3 and 4 were censored because of potential bias to the outcome data from categorizing as recurrent individuals who might not have had metastatic disease.
For analysis of recurrence and survival data, patients in Groups 1–3 were categorized according to dates of latest follow-up or death. Among the censored patients (Group 4), those with definite recurrence from breast carcinoma before censorship were included in tabulations of recurrence, but follow-up was considered ended at the time of censorship. For those patients censored because of receiving additional therapy without clear documentation of distant breast carcinoma recurrence, date of censorship was again used as the final follow-up time, with patients considered to be without distant recurrence.
The association between clinical variables and occurrence of distant metastatic disease was examined using logistic regression analysis. Similarly, recurrence and survival data were analyzed using Kaplan–Meier actuarial methods14 and the Cox regression model using forward selection to identify predictive variables. In these analyses, both the dates of primary diagnosis and documented recurrence were used as initial time points. Median survival estimates were obtained from the 50% probability points on the Kaplan–Meier curves, and statistical comparisons of the curves were performed using the log rank test. Differences at the level of P value less than 0.05 were considered significant.
The clinical characteristics of the study population are presented in Table 1. At the time of diagnosis, patients ranged in age from 20 to 90 years (mean, 51). Most patients had Stage II disease (68%) and were postmenopausal (52%) and among those with known estrogen receptor status were positive (69%).
Seventy-eight of the original 641 patients (12%) were censored: 38 with contralateral breast primaries, 29 with other malignancies, 9 who were treated for metastatic disease without adequate confirmation, 1 with report of late recurrence that could not be confirmed, and 1 who developed a solitary bone metastasis from breast carcinoma and later had an osteosarcoma diagnosed in an area of previous radiation therapy. Only this latter patient was categorized as having a definite recurrence of the original breast carcinoma before being censored. Of the remaining 563 patients, 247 (44%) were disease free at last follow-up (mean follow-up time, 109 months), whereas 22 (4%) had local or locoregional recurrence only and 294 (52%) had distant recurrence. The only clinical variables with predictive power for the occurrence of metastatic disease were the number of positive lymph nodes and estrogen receptor status. The most common single initial site of distant recurrence was bone (Table 2), followed by lung and liver. Overall, 158 (54%) patients with distant recurrence showed evidence of bone involvement on bone scintigraphy and/or radiography at the time of recurrence. At last follow-up, 20 of the 295 patients (7%) with distant recurrence (including the one censored case) were still alive, although only 12 of these individuals had been examined within the preceding year.
Table 2. Survival from Time of Initial Distant Recurrence Relative to Site of the Recurrence
One hundred thirteen of 116 patients (97%) with bone as the sole initial site of distant recurrence had undergone bone scintigraphy at the time of that recurrence. Of these 113 patients, bone involvement initially was identified at a solitary site in 47 patients (41%), 2 new sites in 22 patients (20%), and 3 or more new locations in 44 patients (3 sites, n = 7; 4 sites, n = 2; 5 or more sites, n = 35) (39%). The remaining three patients had bone metastases identified initially on radiologic studies and did not have concurrent bone scan evaluation. These latter individuals were excluded from the subsequent analyses.
Overall, 333 patients (54%) died during the follow-up interval. Median survival from diagnosis of recurrence for patients with different sites of initial distant involvement is summarized in Table 2. Median survival ranged from 11 to 26 months after recurrence for all sites of visceral involvement exclusive of bone. Median survival for patients with initial recurrence in bone alone was 35 months. However, with categorization of bone recurrence based on presenting bone scintigraphic findings of 1, 2, or ≥ 3 new abnormalities, median survival was 53, 38, and 22 months, respectively (Fig. 1; P < 0.0001 between 1 and ≥ 3, P < 0.005 between 2 and ≥ 3 abnormalities). The median interval (between successive bone scans) during which scintigraphic evidence of metastatic disease appeared was not significantly different for patients with initial presentation of 1, 2, or ≥ 3 metastatic lesions (8, 10, and 9 months, respectively; P value not significant).
Median survival from time of diagnosis for patients with distant recurrence in bone was 79 months for those who experienced a recurrence at a solitary bone location, compared with 68 months for those with recurrence at 2 bone locations and 57 months for patients with ≥ 3 new scan abnormalities. By comparison, median survival from time of diagnosis for patients with initial recurrence in lung was 47 months (Fig. 2). The median survival from diagnosis for patients with solitary bone lesions was significantly longer (P < 0.05) than for other patterns of bone metastatic recurrence or other initial sites of visceral recurrence, with or without concurrent bone involvement.
In the bone-only recurrence group, there was no correlation between any clinical parameters or the adjuvant therapy received and the pattern of initial skeletal involvement. By Cox regression analysis, the strongest predictor of survival after recurrence was the bone scan pattern (one, two, or three or more lesions); in keeping with the results for the total study population, the number of positive lymph nodes and the estrogen receptor status also were statistically significant predictors of outcome in this subpopulation.
Thirty-three of 47 (70%) patients who initially presented with a solitary bone metastasis on scintigraphy demonstrated progression to a more characteristic pattern of diffuse skeletal involvement on subsequent scans. A similar proportion of the patients with initial recurrence at two sites (17/22; 77%) showed subsequent scintigraphic progression. The median interval for occurrence of this progression was 14 months for patients with one initial lesion and 9 months for the group with two.
Routine imaging surveillance of patients with breast carcinoma for early detection of metastatic disease has not been shown to improve long term survival.15–17 Whereas periodic evaluation with chest radiographs and radionuclide bone imaging often has been included in clinical research protocols, in large part because of the demonstrated efficacy of these modalities for detection of unsuspected metastatic disease in asymptomatic patients,16, 18–21 data from recent prospective trials do not confirm the value of such imaging in the management of clinical breast carcinoma populations.16, 17 Although the empirical rationale for early detection of metastatic disease, that the probability of successful therapy should be inversely proportional to the quantity of malignant cells present, appears reasonable,22, 23 the limitations of available hormonal, cytotoxic drug and radiation therapies for achieving prolonged control of disseminated metastatic disease may render the routine use of highly sensitive diagnostic investigations moot.16, 17 In addition, although the tendency of breast carcinoma to spread initially to certain organ systems, primarily regional and distant lymph nodes, lung, and bone, is well documented,1, 3, 4 it is less certain whether the qualitative or quantitative character of this involvement affects prognosis.
The role of imaging for detection of metastatic disease during follow-up of patients with Stage I–III breast carcinoma increasingly has been called into question during the past decade.16, 24–26 At one time, bone scintigraphy was commonly used as part of routine follow-up during the first several years after diagnosis of breast carcinoma, but the absence of a proven clinical benefit for early identification of asymptomatic metastatic disease has resulted in the discontinuation of this practice.24, 27–32 Bone scintigraphy is now generally performed only in patients with new biochemical abnormalities (e.g., elevated alkaline phosphatase, CEA), bone symptoms, or suspicious findings on other radiological studies. Use of MRI for evaluation of metastatic involvement of bone marrow in the axial skeleton also has supplanted bone scintigraphy in some evaluations of suspicious radiographic findings.33–35
Given the extensive literature on the use of skeletal scintigraphy in breast carcinoma,5, 10, 18, 27–32, 36, 37 it is somewhat surprising that the clinical significance of the pattern of metastatic involvement has received relatively limited attention. Although solitary scan lesions in cancer patients have had prevalences of metastatic disease ranging from 10% to 64% in several small series,10, 38–41 it remains unclear whether identifying bone metastases at this early stage rather than when skeletal involvement is more extensive is of value. Although the more indolent character of “bone-only” metastatic disease in breast carcinoma has been described,6–9 the specifics of pattern and extensiveness of skeletal involvement often have not been included in detailed analyses of prognosis. Although extent of initial metastatic bone involvement has been noted to be inversely correlated with prognosis, until recently these observations have only been reported in small series with limited statistical power.6, 9, 42
In the current series, patients with breast carcinoma who initially presented with metastatic disease at a solitary site in bone had median survival of 31 months longer than those with 3 or more bone metastases, and 26–41 months longer than patients with other visceral metastases. A similar advantage was maintained when the reference time was the date of initial diagnosis. These results likely reflect the biologic characteristics of the tumors rather than an effect of early diagnosis of asymptomatic disease (lead-time bias), because the latter would not be expected to produce a survival advantage from time of diagnosis.43 Also, although three-quarters of patients with solitary initial recurrence eventually progressed to diffuse bone involvement, the time interval during which this occurred was longer than that associated with conversion from a scan showing no metastases to one with three or more new lesions. The slower disease progression among patients with new solitary metastases was noted both in those who received additional chemo- and/or hormonal therapy immediately and those whose additional treatment was initiated only at a later time, usually when scan progression, new positive radiographs, or other clinical or radiologic changes served to confirm the diagnosis.
Although increasing number of positive axillary lymph nodes at diagnosis and negative estrogen receptor status were consistently associated with increased risk for breast carcinoma recurrence and poorer overall prognosis, including among patients with bone-only recurrence, the bone scintigraphic pattern of metastatic disease was an independent predictor of length of survival. In addition, neither clinical staging variables nor adjuvant therapy received were predictive of whether bone metastases would present with limited or widespread skeletal involvement. Thus, later occurrence of isolated or widespread bone metastases does not appear related to the initial extent of local or locoregional breast carcinoma involvement or the nature of initial therapy, again suggesting other features of tumor biology are involved beyond just the standard staging variables.
The numeric observations from this retrospective review have several sources of uncertainty. Although all patients were participants in adjuvant chemotherapy protocols that included routine bone scintigraphy every 6 months for 3 years and annually thereafter, the prescribed schedule of studies was not always maintained. In addition, many scans were obtained independently of the protocol schedules, based on clinical symptomatology, evidence of recurrence in nonosseous locations, or as part of evaluations after extended periods as long as 2–3 years without medical follow-up. Nevertheless, the finding that the median interval between a previous bone scan without metastatic disease and the first study showing bone involvement was essentially the same for the three categories of recurrence indicates that neither patients with limited nor disseminated metastases were overrepresented. In addition, the longer survival from the date of diagnosis for patients with solitary lesions is a finding unaffected by the frequency with which follow-up studies were obtained.
One possible explanation for the better clinical outcome for patients with solitary new bone scan findings might be that many of these new lesions were incorrectly classified as malignant.44 Most of the lesions in question were not sampled by biopsy, with diagnosis of malignancy established noninvasively. In addition, the retrospective scan review identified a few abnormalities on earlier exams primarily because of knowledge of their presence on later studies. Nevertheless, although the diagnosis of malignancy at the initial sites identified was most commonly derived from indirect imaging evidence, the criteria for interpreting the correlative imaging results were as rigorous as feasible to validate or exclude malignancy as the etiology for the solitary lesions. Among 47 solitary lesions, 13 had initial radiographs that showed bony changes consistent with lytic or blastic metastatic disease. However, initial radiographs either were normal or showed nonspecific findings for 33 solitary lesions, with only later studies showing characteristic changes of malignancy, 21 patients having subsequent scintigraphic progression at these sites (increased intensity of uptake and/or size of lesion), and the remaining 12 patients demonstrating a combination of scintigraphic progression and visceral recurrence at nonosseous sites during follow-up. This confirms our earlier observations concerning scan positive/radiograph negative metastases,10 a subject that itself might warrant further review in another series. One patient had the diagnosis of bone metastases established in the absence of definitive radiologic studies. Although there may have been individual cases in which the lesion first identified was not malignant, most cases have sufficient documentation to conclude that these lesions reflected metastatic breast carcinoma.
The current results suggest that patients with bone metastases from breast carcinoma can be subdivided into more than one prognostic category.9, 42 Although the primary routes of spread of malignant cells to bone, via the arterial circulation, the paravertebral venous plexuses, and direct invasion from adjacent soft tissue and lymph nodes, are well characterized,45, 46 the factors that determine the extent of dissemination are less certain. There remains no definite explanation for the apparent greater indolence of some tumors spreading to bone than to other distant locations.
Another issue raised by the current data is whether earlier treatment of patients with limited metastatic bone disease might result in an even greater survival advantage than was observed. Whether a discrete focus of metastatic neoplasm potentially could be eradicated, rendering the patient “disease free,” is an intriguing speculation, but one that most published survival data, based on heterogeneous populations with a range of skeletal involvement, cannot be used to address.6–9, 47 The possibility that a more aggressive approach to evaluation of solitary bone lesions might produce a desirable effect on long term outcome seems a question worthy of further examination, even though the number of patients with bone recurrence at a solitary site represented only 16% of the total number of patients with distant recurrence. However, only a prospective study in which new scan abnormalities are aggressively investigated with biopsy to establish their malignant character, followed by randomization of patients with asymptomatic metastases to therapy or no therapy arms, can determine definitively whether early diagnosis provides an additional survival advantage or merely introduces a lead-time bias without changing patient outcome.43
In summary, our retrospective review of recurrence and survival data in a large series of Stage I–III breast carcinoma patients has demonstrated that those who experienced a recurrence at a solitary bone site had a survival advantage both from the time of diagnosis of recurrence and from time of initial diagnosis of breast carcinoma. This better prognosis occurred despite the finding that most of these patients did not have confirmation of recurrence initially via other radiologic imaging and thus were not treated with additional chemotherapy, hormone, or radiation therapy at the time the new bone scan abnormality first appeared. Although the current results further support previous observations concerning the unique characteristics of bone as a site of breast carcinoma recurrence,8, 9, 42 this appears to apply only to instances of limited skeletal involvement, because survival in patients with disseminated osseous metastases at presentation was similar to that in those with initial recurrence in all other solitary visceral locations.
The authors gratefully acknowledge the consultation and intellectual input of several individuals over the long gestation period of this article: I. Craig Henderson, M.D.; Daniel Hayes, M.D.; Diane Ascoli; Becky Gelman; and Janet Andersen. The assistance of Kay Kehoe in performing a final update on the patients in this study, and Robin Sawyer in typing various versions of the manuscript, are also gratefully acknowledged. Finally, the input of the late William Kaplan, M.D., cannot be overestimated or ever adequately acknowledged; without his original inspiration and continued support over the better part of a decade, this work would never have been completed and made available to a wide audience.