Search tips
Search criteria 


Logo of annoncLink to Publisher's site
Ann Oncol. 2008 December; 19(12): 2012–2019.
Published online 2008 July 17. doi:  10.1093/annonc/mdn424
PMCID: PMC2733115

Prognostic factors in 1038 women with metastatic breast cancer


Background: Treatment of metastatic breast cancer (MBC) remains palliative. Patients with MBC represent a heterogeneous group whose prognosis and outcome may be dependent on host factors. The purpose of the present study was dual: first, to draw up a list of factors easily available in everyday clinical practice requiring no sophisticated or costly methods and second, to provide results from a large cohort of women who underwent diagnostic and treatment at a single institution.

Patients and methods: From 1975 to 2005, a total of 1038 women with MBC during their follow-up were included in this retrospective analysis. Patients were subsequently assigned to five groups according to the period of metastatic diagnosis.

Results: It is shown that age at initial diagnosis, hormonal receptor status and site of metastasis are the most relevant prognostic factors for predicting survival from the time of metastastic occurrence. It is also shown that a metastasis-free interval is an easily and immediately available multifactorial prognostic index reflecting the multiparametric variability of the disease.

Conclusion: These fundamental observations may assist physicians in evaluating the survival potential of patients and in directing them toward the appropriate therapeutic decision.

Keywords: MBC, prognostic factors, survival


Treatment of metastatic breast cancer (MBC) remains palliative despite recent advances in the management of this disease. Survival time for patients with MBC varies greatly. Hence, once metastasis is detected, median survival ranges between 24 and 30 months [1, 2]. Thus, patients with MBC represent a heterogeneous group whose prognosis and clinical course may be dependent on host factors. Strategies for managing MBC have been developed and have been approved as conventional treatment schedules for MBC. Consequently, taxanes and aromatases inhibitors were introduced during the last decade, and new drugs have been introduced more recently: gemcitabine, liposomal doxorubicin, oral vinorelbine or 5-fluorouracil, antibody anti-Her-2/Neu (trastuzumab), anti-VEGF (bevacizumab) or tyrosine kinase inhibitor (lapatinib) [3]. A recent consensus paper on medical treatment of MBC presented an update of clinical and biological prognostic factors associated with clinical outcomes [4]. These factors may influence the choice of treatment and include performance status, sites of recurrent disease, number of disease sites, hormone receptor status, Her-2/Neu status, disease-free interval, prior adjuvant therapy and prior therapy for MBC [4]. Thus, improved evaluation of the individual profile of prognostic factors may be helpful when choosing a customized therapeutic strategy. For instance, hormonotherapy is indicated for positive hormonal receptor (HR) patients with long disease-free intervals and bone metastasis while chemotherapy is preferentially administered in patients with visceral metastases and aggressive disease or hormonoresistant tumors.

The purpose of the present study was dual: first, to draw up a list of factors that are easily available in everyday clinical practice and requiring no sophisticated or costly methods. Secondly, to provide results from a single institution with a large number of patients in whom diagnostic procedures, surgical procedures and treatment were homogeneous in time. A total of 1038 women who had developed MBC during their follow-up and were treated in our cancer center over a 30-year period (from January 1975 to January 2005) were included in this study.

patients and methods

study population

From 1975 to 2005, 4958 consecutive breast cancer patients with no evidence of distant metastasis at the time of primary diagnosis were followed in our cancer center in France (Centre Antoine Lacassagne, Nice) and were entered in a large database. One thousand and thirty-eight patients (21%) subsequently developed metastatic disease from January 1980 to January 2005 and were included in this analysis. Main patient characteristics, treatment and outcomes were included prospectively in the hospital database. The breast cancer institutional review board approved and validated all data entered in the database. Patients were subsequently assigned to five groups according to the period of metastatic diagnosis: the first, second, third, fourth and fifth groups included, respectively, patients from 01/01/1980 to 12/31/1984, 01/01/1985 to 12/31/1989, 01/01/1990 to 12/31/1994, 01/01/1995 to 12/31/1999 and 01/01/2000 to 12/31/2004. Patient age was taken at the time of the initial diagnosis of breast cancer. Adjuvant treatment strategy was established at a multidisciplinary tumor board including at least a medical oncologist, a surgeon, a radiation oncologist, a pathologist and a radiologist.

After initial diagnosis showing no evidence of metastasis, patients had routine follow-up including clinical examinations, laboratory tests, and an annual mammography during the first 5 years. Sites of metastasis were explored only when physical examination of the patient indicated a suspicion of metastasis. In these cases, metastasis sites were documented using X-rays, liver ultrasound, bone scan and/or computed tomography or magnetic resonance imaging. In this study, we took into consideration only the initial metastatic sites that could be single or multiple. Sites of metastasis were divided into six groups: skin and nodes, lung, liver, bone, brain and multiple. Adjuvant therapy was classified as adjuvant chemotherapy yes/no or hormonal therapy yes/no. Information concerning local relapses was also recorded. Tumors were characterized according to the 2003 tumor–node–metastasis breast classification [5] including the size of the primary tumor in millimeters (statistical analyses were carried out using size grouping in pT1 ≤20 mm, pT2 20–50 mm, pT3 >50 mm) and the absolute number of positive lymph nodes (statistical analyses were carried out using three categories: no positive node, less then four positive nodes, four or more positive nodes). Median ratio between the number of nodes resected and the number of positive nodes were 4.0 (range 1.0–39.0). The Scarff Bloom Richardson (SBR) [6] modified by Elston and Ellis [7] histological grading (SBR1, SBR2, SBR3) was also recorded. Up to 1989, HR status (estrogen and progesterone) was measured by ligand-binding assay [8]. Subsequently, measurements were obtained by cytosolic immunoassay using the Abbott Kit [9]; these two methods had been previously compared and gave concordant results [10]. After 1997, HR status was measured using the immunohistochemistry method. The reliability of this method as compared with the other two was evaluated in our cancer center (data not shown). The cut-off level for ligand-binding assay and immunoassay was ≥10 fmol/mg cytosol protein and ≥10% for immunohistochemistry. Her2/neu overexpression was evaluated by immunohistochemistry (3+ level). In our institution, routine determination of this parameter only began in 2000 and thus few data were available.


Of the 4958 consecutive patients included in the study and showing no evidence of metastatic disease at the time of initial diagnosis, all had undergone tumor resection with axillary lymph node dissection: 43% patients had mastectomy, 53% had tumorectomy and 4% patients had subcutaneous mastectomy. Locoregional treatment by radiotherapy associated with surgery was given to 77% of patients. Adjuvant chemotherapy was administered in 16% of patients, adjuvant hormonotherapy was given to 28% of patients, while 16% of patients received both hormonotherapy and chemotherapy as adjuvant treatment. The choice of therapy was always made according to specific guidelines used at the time of patient treatment. Patients receiving neo-adjuvant chemotherapy as initial treatment were not included in this study.


All categorical data were described using numbers and percentages. Quantitative data were presented using median and range or mean and standard deviation. Censored data were described using Kaplan–Meier estimation including number of patients, number of events, median survival and a 95% confidence interval (CI). When no information was available, status was coded as missing data. Statistical analyses were two sided and carried out using R-2.5.0 for Windows.

univariate analysis

Statistical comparisons were carried out using χ2 test or Fisher’s exact test for categorical data and log-rank test for censored data. Smoothing splines were used to predict death risk versus metastasis-free interval (MFI).

multivariate analysis

Multivariate analysis was carried out by creating a Cox proportional hazards model. Choice of the final model was made performing backward stepwise model selection by exact Akaike information criterion (AIC). All variables associated with P <0.10 on univariate analysis were included in the model. Colinearity between variables was evaluated using the ‘r’ Pearson correlation coefficient between all variables entered in the model. If r >0.30, one of the two variables is considered as redundant and must be removed from the model. This is the case for adjuvant chemotherapy and a number of positive lymph nodes (r = 0.64). Only adjuvant chemotherapy was introduced into the model. Proportional hazards were tested for all entered variables using graphical (Schoenfeld residuals, log–log plot of cumulative hazard) and statistical methods. Covariates with nonproportional effect were tested as standard adjustment covariates and also tested into the model as stratification factors to confirm the results. The search for interaction was automated using the R-2.5.0 step AIC procedure. Interaction was considered to be significant if P <0.01.

definition of censored data

The MFI, the interval between first diagnosis of breast cancer and first distant metastasis, was divided into three periods: <24 months, ≥24 months and <60 months, ≥60 months.

Metastasis survival was the interval between first distant metastasis and death due to cancer. Follow-up was limited to 60 months for all patients entered in the study. If death was not due to cancer or if the patient was lost to follow-up, data were censored at the date of their last known contact or 60 months after metastatic occurrence.


patient characteristics

A total of 1038 patients who presented a metastatic occurrence were analyzed in the study. Median follow-up after recurrence was 60 months for patients presenting a metastatic disease during the period 1980–2000 and 41 months for the last follow-up period (2000–2005). A description of the patient study is given in Table 1. Median age at initial diagnosis of the primary tumor was 57.8 years with 734 patients (70%) aged >50 years. Median MFI was 46.5 months (range 1.0–324.1 months). A majority of patients (75.5%) developed metastasis >2 years after the initial diagnosis. The location of metastatic occurrence was mainly in bone (38.9%). Two hundred and twenty-eight (22%) patients also had a local recurrence. Median-specific survival after metastatic occurrence was 23.1 months (95% CI 21.6–26.0 months). We observed 778 deaths among 1038 patients. Her2/neu receptor status was available in 195 patients. HR status (estrogen and progesterone) was determined in 949 patients. Distribution of these receptors is summarized in Table 2. Of the 794 patients with positive HR status, >30% (273 patients) received adjuvant hormonotherapy.

Table 1.
Patient characteristics
Table 2.
Estrogen and progesterone receptors distribution

univariate analysis

Univariate analysis (Table 3) for specific survival after metastatic occurrence pinpoints the following parameters as significant prognostic factors: metastatic diagnosis period, site of metastasis, MFI, age at diagnosis, number of positive lymph nodes, size of the primary tumor, SBR grade, HR status and presence of adjuvant chemotherapy. Women had a significantly worse survival if metastatic location was brain, multiple or liver, if metastatic diagnosis period was far from 2005 and if metastatic-free interval was shorter. Interestingly, we found an inverse relationship between death risk and MFI duration (Figure 1). In addition, older age (≥50 years), number of positive lymph nodes (>0), size of the primary tumor (>20 mm), SBR grade (>1), negative HR status and treatment with adjuvant chemotherapy were also associated with poor survival. In this univariate analysis, adjuvant hormonotherapy was not a significant prognostic indicator (P = 0.13), but became significant only when HR status (P = 0.02) was taken into account. Her2/neu status was not relevant because of the limited available data. Two patient characteristics differed regarding the period of metastatic diagnosis (data not shown): MFI was higher for the more recent period with 14%, 31%, 40%, 50% and 59% of patients having MFI ≥60 months, respectively, for the five successive periods (P < 10−6). Also, the frequency of hormonotherapy treatment was higher during the recent period with 15%, 22%, 24%, 38% and 54% of patients treated with hormonotherapy, respectively, for the five successive periods (P < 10−6). Figures 2 and and33 show survival curves in HR-positive or -negative patients according to the treatment period. In HR+ patients, survival was significantly different according to the treatment period (P = 0.0086). In contrast, there was no difference in survival for HR− women according to the treatment period (P = 0.65). Whatever the treatment period, median survival time was >17 months in HR+ patients, and < 10 months in HR− patients. Figures 4 and and55 depict survival in HR+ and HR− patients according to the site of metastasis. When considering HR+ patients, median survivals were >12 months except for brain metastasis (4 months) and multiple metastases (9 months). In contrast, regarding HR− patients, all median survivals were <12 months except for bone metastases (21 months). We also examined whether initial HR status could influence the future site of metastatic occurrence. We found that there were more bone recurrences in women with a positive hormonal status [odds ratio (95% CI) 2.95 (1.97–4.42), P < 10−3, chi-square test] and more brain [odds ratio (95% CI) 3.84 (1.77–8.35), P = 0.002, Fisher’s exact test] or multiple recurrences [odds ratio (95% CI) 1.76 (1.13–2.73), P = 0.02, chi-square test] in women with negative HR.

Table 3.
Univariate survival analysis
Figure 1.
Plot of the fitted spline function for log (hazard ratio) of metastasis-free interval (MFI) (dashed lines show the 95% confidence intervals). Risk of death declines with increasing MFI.
Figure 2.
Survival from the time of diagnosis of metastatic disease for patients with hormonal receptor positive (HR+) according to the period of diagnosis of the first metastasis recurrence.
Figure 3.
Survival from the time of diagnosis of metastatic disease for patients with hormonal receptor negative (HR−) according to the period of diagnosis of the first metastasis recurrence.
Figure 4.
Survival from the time of diagnosis of metastatic disease for patients with hormonal receptor positive (HR+) according to the localization of the first metastasis recurrence.
Figure 5.
Survival from the time of diagnosis of metastatic disease for patients with hormonal receptor positive (HR−) according to the localization of the first metastasis recurrence.

multivariate analysis

A multivariate analysis was carried out including the site of metastasis, patient age, HR status, treatment by chemotherapy, tumor size, SBR grade, the metastatic diagnosis period and the MFI (Table 4). Brain, multiple or liver sites of metastasis, age >50 years, negative HR, the presence of chemotherapy, high SBR grade and large tumor size were found to be unfavorable independent prognostic factors able to predict specific survival following the first metastasic occurrence. In contrast, metastatic diagnosis period and MFI appeared to have no influence on survival after recurrence according to the multivariate Cox model. Statistical analysis gave strong evidence of nonproportionality for HR status. Hence, this variable was also incorporated into the model as a stratification factor. Nevertheless, results were comparable when including this variable as stratification factor and thus confirmed the validity of the model. Interaction analysis indicated that there was a synergistic effect on survival when combining tumor size and HR status (P = 0.004) giving a poorer prognosis for negative HR associated with large tumor size.

Table 4.
Multivariate analyses

We analyzed more particularly the impact of HR status by performing a Cox model analysis in patients with negative HR and in patients with positive HRs. When considering positive HR status patients, we found that site of metastasis with bone (hazard ratio = 1.83, 95% CI 1.31–2.56), lung (hazard ratio = 2.06, 95% CI 1.41–2.99), liver (hazard ratio = 4.57, 95% CI 3.10–6.73), multiple (hazard ratio = 5.51, 95% CI 3.73–8.152), brain (hazard ratio = 18.17, 95% CI 8.84–37.34) and older age (hazard ratio = 1.57, 95% CI 1.27–1.92), presence of chemotherapy (hazard ratio = 1.43, 95% CI 1.18–1.72) and higher SBR (hazard ratio = 1.28, 95% CI 1.04–1.57) remained poor independent prognostic factors, while tumor size was no longer significant. When considering negative HR status patients, we noted that site of metastasis with brain (hazard ratio = 6.76, 95% CI 2.80–16.31), multiple (hazard ratio = 3.10, 95% CI 1.55–6.23), liver (hazard ratio = 9.11, 95% CI 3.82–21.73) and large tumor size (hazard ratio = 2.02, 95% CI 1.26–3.24) were the only prognostic factors that remained poor independent prognostic factors.


Studies on prognostic factors in patients with MBC vary considerably and are sometimes contradictory with respect to the selection of patients, availability of biological and clinical parameters, patients’ follow-up and methods of analysis. In the present analysis of 1038 patients treated in our institution from 1980 to 2005, we have shown that hormonal status receptor, site of metastasis, adjuvant chemotherapy, patient age, size of primary tumor and SBR grade constituted independent prognostic factors which were significant in multivariate analysis, while metastatic diagnosis period and MFI were significant only in the univariate analysis.

Among these factors, the site of metastasis seems to be the most significant independent prognostic factor. As previously described, multiple or visceral site of metastasis seems to be predictor of poor-specific survival with a median survival not exceeding 22 months, while nonvisceral sites are associated with better-specific survival with a median survival of >33 months [11, 12]. Patients with metastatic bone disease were associated with a relatively better survival [1315] and bone is the most frequently reported site of metastasis with 40% in our study and >30% described in other reports [11, 16]. Other studies have already investigated the relationship between the site of metastasis and survival [17, 18] but few of them have included enough patients to allow a powerful multivariate analysis to be carried out. It is interesting to note that these observations were more significant when considering hormonal status. Thus, whatever the metastasis site, survival after metastatic recurrence for HR+ patients was better than for HR− patients. Some groups have suggested a relationship between HR status and the site of recurrence. Our study confirms this trend and shows that positive HRs were more likely to recur in bone while negative receptor status occurred more often in brain and multiple sites.

As previously reported by Andre et al. [19], univariate analysis in the present study has shown that survival of breast cancer patients developing metastases was improved in the time period from 1980 to 2005 with median survival ranging from 16 months for the first metastatic diagnosis period (1980–1985) to 31 months for the last period (2000–2005). This result, however, was not confirmed by a multivariate analysis indicating that the survival improvement appears to be more closely linked to other independent prognostic factors studied (site of metastasis, age, HR status, previous chemotherapy and tumor size) than to time period. The present study also demonstrates that several types of therapeutic management have improved the survival of metastatic HR+ breast cancer patients while the improvement of HR− patients has remained unchanged for the past 25 years. It is generally accepted that young age at diagnosis is associated with more aggressive disease and relative poor survival from diagnosis. Risk of recurrence, however, is significantly increased in older women [20]. In our study, women aged >50 years had significantly lower survival rates. One can explain this observation more precisely: postmenopausal patients were found to have a lower response rate to chemotherapy. The increase in side-effects inducing dose reduction and loss of efficiency in older women may also explain this observation.

Some studies have already considered adjuvant chemotherapy or axillary lymph node involvement at first diagnosis as survival prognostic factors following first recurrence. Results and conclusions differed [11, 18, 2124]. In our population and in most of the others, adjuvant chemotherapy and axillary lymph node status were correlated and administration of adjuvant chemotherapy was more frequent in patients with positive lymph node status. It is difficult to distinguish the respective influence of each of these two factors. Nevertheless, adjuvant chemotherapy appears to be an unfavorable independent prognostic factor for survival.

Some studies have shown that early recurrence is an independent predictor of survival and that desease-free interval (DFI) is one of the strongest prognostic factors reflecting the aggressiveness of the disease [17, 18, 22, 23, 25, 26]. In the present study, disease-free interval appeared to be related to survival in the univariate analysis, but when performing a multivariate Cox model, the other previously described parameters were more powerful for predicting specific survival. Nevertheless, it appears that MFI may be considered an easily and immediately available multifactorial prognostic index reflecting the multiparametric variability of the disease.

In summary, this study on a particularly large number of patients with breast carcinoma shows that prognostic factors of the initial primary tumor and metastastic disease are associated with specific survival after recurrence. These findings are consistent with the hypothesis that the intrinsic tumor biology of the primary tumor plays a critical role in determining outcome following recurrence [22]. The aim of the present study was dual: first, to draw up a list of factors easily available in usual clinical practice requiring no sophisticated or costly methods. Secondly, to provide results from a large cohort of women who underwent diagnostic and treatment at a single institution. It is presently shown that age at initial diagnosis, HR status, site of metastasis and DFI are the most relevant prognostic factors for predicting survival from the time of metastatic occurrence. These three fundamental factors may enable physicians to evaluate more easily the survival potential at individual level and guide them in their therapeutic decision.


Sandra Ruitort (PharmD) and George Morgan (English teacher) provided editorial assistance in the preparation of the article.


1. Norton L. Metastatic breast cancer. Length and quality of life. N Engl J Med. 1991;325:1370–1371. [PubMed]
2. Mosconi P, Colozza M, De Laurentiis M, et al. Survival, quality of life and breast cancer. Ann Oncol. 2001;12(Suppl 3):S15–S19. [PubMed]
3. Nicolini A, Giardino R, Carpi A, et al. Metastatic breast cancer: an updating. Biomed Pharmacother. 2006;60:548–556. [PubMed]
4. Beslija S, Bonneterre J, Burstein H, et al. Second consensus on medical treatment of metastatic breast cancer. Ann Oncol. 2007;18:215–225. [PubMed]
5. Singletary SE, Allred C, Ashley P, et al. Revision of the American Joint Committee on Cancer staging system for breast cancer. J Clin Oncol. 2002;20:3628–3636. [PubMed]
6. Bloom HJ, Richardson WW. Histological grading and prognosis in breast cancer; a study of 1409 cases of which 359 have been followed for 15 years. Br J Cancer. 1957;11:359–377. [PMC free article] [PubMed]
7. Elston CW, Ellis IO. Pathological prognostic factors in breast cancer. I. The value of histological grade in breast cancer: experience from a large study with long-term follow-up. Histopathology. 1991;19:403–410. [PubMed]
8. Milano G, Moll JL, Formento JL, et al. Simultaneous micro measurement of steroid receptors in breast cancer. Br J Cancer. 1983;48:579–584. [PMC free article] [PubMed]
9. Romain S, Formento JL, Guirou O, et al. Determination of oestrogen receptors by enzyme immunoassay. Technical differences between laboratories and their consequences. Eur J Cancer. 1994;30A:740–746. [PubMed]
10. Blankenstein MA. Comparison of ligand binding assay and enzyme immunoassay of oestrogen receptor in human breast cancer cytosols. Experience of the E.O.R.T.C. Receptor Group. Breast Cancer Res Treat. 1990;17:91–98. [PubMed]
11. Solomayer EF, Diel IJ, Meyberg GC, et al. Metastatic breast cancer: clinical course, prognosis and therapy related to the first site of metastasis. Breast Cancer Res Treat. 2000;59:271–278. [PubMed]
12. Cutler SJ, Ardyce JA, Taylor SG., III Classification of patients with disseminated cancer of the breast. Cancer. 1969;24:861–869. [PubMed]
13. Smalley RV, Scogna DM, Malmud LS. Advanced breast cancer with bone-only metastases: a chemotherapeutically responsive pattern of metastases. Am J Clin Oncol. 1982;5:161–166. [PubMed]
14. Coleman RE, Smith P, Rubens RD. Clinical course and prognostic factors following bone recurrence from breast cancer. Br J Cancer. 1998;77:336–340. [PMC free article] [PubMed]
15. Robain M, Pierga JY, Jouve M, et al. Predictive factors of response to first-line chemotherapy in 1426 women with metastatic breast cancer. Eur J Cancer. 2000;36:2301–2312. [PubMed]
16. Swenerton KD, Legha SS, Smith T, et al. Prognostic factors in metastatic breast cancer treated with combination chemotherapy. Cancer Res. 1979;39:1552–1562. [PubMed]
17. Yamamoto N, Watanabe T, Katsumata N, et al. Construction and validation of a practical prognostic index for patients with metastatic breast cancer. J Clin Oncol. 1998;16:2401–2408. [PubMed]
18. Ryberg M, Nielsen D, Osterlind K, et al. Prognostic factors and long-term survival in 585 patients with metastatic breast cancer treated with epirubicin-based chemotherapy. Ann Oncol. 2001;12:81–87. [PubMed]
19. Andre F, Slimane K, Bachelot T, et al. Breast cancer with synchronous metastases: trends in survival during a 14-year period. J Clin Oncol. 2004;22:3302–3308. [PubMed]
20. De la Rochefordiere A, Asselain B, Campana F, et al. Age as prognostic factor in premenopausal breast carcinoma. Lancet. 1993;341:1039–1043. [PubMed]
21. Clark GM, Sledge GW, Jr., Osborne CK, et al. Survival from first recurrence: relative importance of prognostic factors in 1,015 breast cancer patients. J Clin Oncol. 1987;5:55–61. [PubMed]
22. Chang J, Clark GM, Allred DC, et al. Survival of patients with metastatic breast carcinoma: importance of prognostic markers of the primary tumor. Cancer. 2003;97:545–553. [PubMed]
23. Insa A, Lluch A, Prosper F, et al. Prognostic factors predicting survival from first recurrence in patients with metastatic breast cancer: analysis of 439 patients. Breast Cancer Res Treat. 1999;56:67–78. [PubMed]
24. Falkson G, Gelman R, Falkson CI, et al. Factors predicting for response, time to treatment failure, and survival in women with metastatic breast cancer treated with DAVTH: a prospective Eastern Cooperative Oncology Group study. J Clin Oncol. 1991;9:2153–2161. [PubMed]
25. Vogel CL, Azevedo S, Hilsenbeck S, et al. Survival after first recurrence of breast cancer. The Miami experience. Cancer. 1992;70:129–135. [PubMed]
26. Hortobagyi GN, Smith TL, Legha SS, et al. Multivariate analysis of prognostic factors in metastatic breast cancer. J Clin Oncol. 1983;1:776–786. [PubMed]

Articles from Annals of Oncology are provided here courtesy of Oxford University Press