Although several prognostic factors for survival have been identified in metastatic breast cancer patients, the prognosis of individual patients frequently fails to confirm published predictions (Honig, 1996). Additional biological parameters are therefore required to gain more accurate insight into the outcome of metastatic breast cancer patients. Recent cancer research has highlighted the importance of tumour neovascularisation in the initial and metastatic growth of cancer (Hanahan and Folkman, 1996) and there is solid evidence supporting the evaluation of tumour angiogenesis for prognosis in cancer patients (Poon et al, 2001). Thus, the assessment of angiogenic factor production may allow a better estimation of survival for patients with metastatic breast cancer. The vascular endothelial growth factor (VEGF) is one of the most potent and specific angiogenic molecules. In breast cancer, VEGF expression in the primary tumour appears to be significantly correlated with relapse-free survival (Gasparini, 2000). Elevated serum levels of VEGF have been found in metastatic cancer patients. These levels have been correlated to the clinical course of the disease in several tumour types (Poon et al, 2001; Ugurel et al, 2001; Molica et al, 2002). As most serum VEGF is released from platelets during blood clotting, it has been suggested that plasma VEGF may be a better reflection of ongoing angiogenic activity, although this assumption has been challenged (Banks et al, 1998; Salven et al, 1999a; Wynendaele et al, 1999; George et al, 2000). To date, plasma VEGF has not been shown to be a better prognostic factor than serum VEGF in any solid tumour (Werther et al, 2002). Interleukin 6 (IL-6) is a pleiotropic cytokine produced by a variety of cell types, including endothelial cells and normal haematopoietic cells (Kishimoto et al, 1995). High IL-6 serum levels are correlated with shorter survival in patients with haematological malignancies, renal cell carcinoma and prostate cancer (Blay et al, 1992; Nakashima et al, 2000). Recent publications have highlighted the importance of IL-6 in the regulation of VEGF production. Interleukin 6 expression is induced by hypoxia and, in turn, is able to upregulate VEGF transcription via a specific response element present upstream of the transcription initiation site of VEGF (Cohen et al, 1996; Yan et al, 1997). Furthermore, IL-6 has been shown to correlate with platelet count and platelet VEGF content. As circulating VEGF is mostly transported by platelets, IL-6 could be regarded as an indirect angiogenic factor that facilitates the production and the distribution of VEGF to metastatic sites (Salgado et al, 1999,2002; Vermeulen et al, 1999). The present study was undertaken to investigate connections between serum VEGF, plasma VEGF and serum IL-6 levels in metastatic breast cancer patients and evaluate their potential relation to the clinical outcome. PATIENTS AND METHODS Patients Serum and matched plasma samples from 73 patients were used in this study. A serum sample without matched plasma sample was available for an additional 14 patients. All patients had a fully documented history of metastatic breast cancer and were treated in our institution between January 1995 and November 2000. Serums and EDTA plasma samples were collected after obtaining informed consent, aliquoted and stored at −80°C. The characteristics of the patients are shown in Table 1 Table 1 Patient characteristics Characteristics No. of patients Percent Entered on study 87 100 Age (years) Median 54 Range 26–86 PS 0–1 55 63 2–4 32 37 Receptor status ER Positive 58 67 ER Negative 23 26 Unknown 6 7 Adjuvant chemotherapy Received 41 47 Not received 46 53 Prior chemotherapy for metastatic disease Received 61 70 Not received 26 30 Disease-free interval 24 months 38 44 Metastatic site Liver 51 59 Lung 45 48 One or two 49 56 Three or more 38 44 LDH ⩽1 × normal 25 29 >1 × normal 62 71 PS=performance status; ER=estrogen receptor; LDH=lactate dehydrogenase. . Median age was 54 years (range 26–86). In all, 41 patients (47%) had received adjuvant chemotherapy as part of initial therapy. The median disease-free interval (from first diagnosis to recurrence) of the 87 patients was 32 months (range 0–19 years). For the 41 patients who had received adjuvant chemotherapy, there was at least a 4-month interval between the last course of adjuvant chemotherapy and blood sampling. All patients in this series had metastatic disease refractory to hormone therapy, that is, receptor-negative for both oestrogen and progesterone or progressive after hormone therapy. In all, 61 patients had previously received first-line chemotherapy for metastatic disease. Most patients (75%) had metastatic disease at multiple sites and 59% had liver involvement. Serum analysis Serum and EDTA plasma samples were kept frozen at −80°C, then thawed shortly before determination of VEGF and IL-6 levels. Commercially available immunoassay kits were used according to the manufacturers' instructions. Vascular endothelial growth factor levels were determined with the Quantikine® human VEGF immunoassay kit from R&D system (R&D System, Minneapolis, MN, USA); IL-6 levels were determined with the Human IL-6 Enzyme Immunoassay kit from Immunotech (Immunotech, Marseille, France). Both assays employ the quantitative sandwich enzyme immunoassay technique. All plasma and serum level determinations were performed in triplicate. The theoretical detection limits of the immunoassay were 5 pg ml−1 for VEGF and 8 pg ml−1 for IL-6. Statistical analysis Statistical analyses were carried out according to the procedures of the SPSS® package 8.0 (Chicago, USA, 1999). Patients' characteristics were compared using Student's t-test for quantitative variables. Correlation analysis with the Pearson test was used to study relations between continuous variables. Survival data, considered from the day of sampling to the time of death or the time of last follow-up, were assessed by means of the Kaplan–Meier method. Univariate analyses of survival times were performed using the log rank test. For this analysis, we tested three different threshold values for each of the two variables: the lowest quartile, the median and the highest quartile. In addition, we tested the detection limit for IL-6 and two threshold values previously reported as relevant for VEGF (Salven et al, 2000; Ugurel et al, 2001). The Cox proportional hazard regression model was used to identify independent prognostic factors for survival. The variables introduced in the model were prognostic factors commonly accepted for metastatic breast cancer and parameters previously identified as significantly related to survival in the univariate analysis. RESULTS Serum and plasma levels of VEGF and serum levels of IL-6 Serum VEGF was detectable in all 87 patients. The median serum VEGF level was 302 pg ml−1 (range 31–3183 pg ml−1). Confirming previously published results (Salgado et al, 1999), high serum VEGF levels were correlated with high platelet counts in our patients (r=0.6; P normal), oestrogen receptor status (present vs absent) and disease-free interval (⩽24 months vs >24 months) (Honig, 1996; Yamamoto et al, 1998). Using a model in which all these factors were maintained, IL-6 remained an additional independent prognostic factor for survival. This was confirmed on introducing serum IL-6 levels as a continuous variable (P 1 =8 pg ml−1 0.007 2.2 1.2–3.8 Prior adjuvant chemotherapy 0.002 2 1.4–4.7 PS=performance status. ): IL-6 ⩾8 pg ml−1 (P=0.007, HR=2), IL-6 ⩾13 pg ml−1 (P<0.001, HR=3.2), IL-6 ⩾30 pg ml−1 (P=0.001, HR=3) and IL-6 ⩾55 pg ml−1 (P<0.001, HR=5.2). DISCUSSION There is accumulating evidence that the serum levels of angiogenic factors have prognostic significance in human cancers of epithelial or hematological origin (Poon et al, 2001). In this study, we have evaluated the clinical significance of serum VEGF and IL-6 levels on 87 metastatic breast cancer patients. Several studies have reported that metastatic breast cancer patients have elevated serum levels of VEGF as compared with normal controls, as well as with women with primary breast cancer. Three of those studies, performed with the same immuno-assay as the one we used, reported median serum VEGF levels ranging from 186 to 368 pg ml−1, with a considerable range between minimum and maximum levels (10 to 2426 pg ml−1 in the series of Kraft et al, 1999) (Salven et al, 1999b; Adams et al, 2000). These results are consistent with our findings (median: 302 pg ml−1; range 31–3183) confirming the strong correlation between serum VEGF and platelet count (Salgado et al, 1999). In this series, the median platelet VEGF concentration, estimated from the ratio between serum VEGF and platelet count, was 1.5 pg 10−6 platelets, which is in agreement with results obtained in other series of metastatic cancer patients (Salgado et al, 1999; Salven et al, 1999a). Furthermore, we have confirmed the correlation between platelet VEGF concentration and IL-6 levels, thus placing further emphasis on the importance of IL-6 on VEGF metabolism (Vermeulen et al, 1999; Salgado et al, 2002). As most serum VEGF is released from platelets during blood clotting, plasma VEGF may be a better reflection of ongoing angiogenesis and, accordingly, a better prognostic marker for patients with cancer (Banks et al, 1998). To test this hypothesis, we assessed plasma VEGF levels in 73 of our patients for whom an EDTA plasma sample had been collected at the same time as the serum sample. We found that plasma VEGF is strongly correlated to serum VEGF and to platelet count, which is in agreement with previous findings on colorectal cancer patients (George et al, 2000; Werther et al, 2002). With regard to survival, plasma VEGF levels were not related to prognosis in this subgroup of 73 patients. The strong correlation between serum and plasma VEGF may reflect the fact that plasma VEGF levels are directly related to total blood VEGF levels and correspond to free VEGF after equilibration with platelet levels (George et al, 2000). Alternatively, we cannot rule out that the platelet VEGF released following a partial platelet degradation during plasma sampling and storage, may account for part of the EDTA plasma VEGF (Wynendaele et al, 1999). In our series, only 39% patients had detectable levels of IL-6, which is less than previously reported and may be because of the low sensitivity of the ELISA kit we used (detection limit: 8 pg ml−1) (Salgado et al, 1999; Zhang and Adachi, 1999). We could not find any correlation between serum IL-6 and platelet numbers nor between serum IL-6 and VEGF levels, and this result was maintained when the analysis was restricted to patients who had not received any chemotherapy for their metastatic disease. With regard to survival, the present results show that high serum IL-6 levels are associated with a poor overall survival, while serum and plasma VEGF levels do not significantly correlate to the outcome. Results of the multivariate analysis show that IL-6 levels remain independent variables for survival when assessed with clinical prognostic markers. No studies to date have shown any correlation between the survival of metastatic breast cancer patients and serum or plasma levels of VEGF, which is in contrast with several other publications reporting a strong correlation between VEGF expression in the tumour and prognosis in early-stage breast cancers (Gasparini, 2000; Manders et al, 2002). Different factors might account for the lack of prognostic value of VEGF in our series. Breast cancer is a heterogeneous disease with variable biological and clinical behaviours, in which clinical evolution is much less predictable than in lymphoma or metastatic renal cell carcinoma. Serum VEGF may reflect the disease status at a given time (i.e. stable or progressive) and fail to mirror important biological properties, such as sensitivity to chemotherapy or to hormone manipulation, that will ultimately be of utmost importance for prognosis (Honig, 1996). This hypothesis is supported by the work of Dirix et al (1997), who have reported that the positive association of a short tumour volume-doubling time with elevated bFGF and VEGF serum levels in advanced cancer patients is largely independent from the metastatic pattern and the extent of the disease. An alternative explanation may be the relative importance of VEGF, as compared to other angiogenic factors, in metastatic breast cancer. In a model of human breast carcinoma xenografted into nude mice, VEGF was shown to be critical for tumorigenesis, but not for growth, after the tumours had reached a certain size (Yoshiji et al, 1997). The authors hypothesised that VEGF downregulation might be balanced by bFGF, TGFα or other angiogenic factors. In metastatic breast cancer, angiogenesis may thus depend on numerous redundant angiogenic factors whose relative importance varies from one patient to another and/or during the evolution of the disease. Since IL-6 induces VEGF transcription and augments both platelet production and platelet storage of VEGF, it was formerly hypothesised that IL-6 could be regarded as an indirect angiogenic factor acting mainly on VEGF metabolism (Salgado et al, 1999; Vermeulen et al, 1999). Results presented here do not support this hypothesis. We have confirmed the relation between platelet VEGF content and IL-6, but we have failed to show any association between survival and VEGF, or between survival and platelet VEGF content. On the other hand, the serum level of IL-6 is an independent prognosis factor, which confirms results obtained by Zhang et al on a smaller series (Zhang and Adachi, 1999). This suggests that, apart from the regulatory effects of IL-6 on VEGF production, other mechanisms are responsible for the deleterious effects of high serum IL-6 levels on tumour growth, thus explaining their strong correlation with prognosis in metastatic breast cancer patients. Although IL-6 expression in early-stage breast carcinomas has been correlated with low grade, oestrogen receptor status and good prognosis (Fontanini et al, 1999; Karczewska et al, 2000), several studies have shown that IL-6 may contribute to disease progression, particularly in advanced breast cancer patients. Indeed, IL-6 can modulate steroid hormone responsiveness by increasing aromatase activity and oestradiol 17 beta-hydroxysteroid dehydrogenase activity in malignant tissue and by activating oestrogen receptor-α (Duncan et al, 1994; Singh et al, 1999; Speirs et al, 2000). IL-6 may promote cell migration by activating the mitogen-dependent protein kinase pathway, and increase chemotherapeutic resistance by inhibiting the activation of proteases involved in apoptosis (Lotem and Sachs, 1997; Badache and Hynes, 2001). Finally, IL-6 promotes osteoclast formation and inhibits dendritic cell differentiation, thus facilitating metastatic growth (Menetrier-Caux et al, 1998; Grano et al, 2000). As tumours evolve toward a metastatic phenotype and interfere with other endogenous or exogenous factors, IL-6 activity on cancer cells and their environment might actually shift from growth inhibition and differentiation to proliferation and antiapoptosis (Lu and Kerbel, 1993; Chiu et al, 1996; Badache and Hynes, 2001). This could explain the favourable prognosis associated with the presence of tumour IL-6 in early-stage breast cancer (Karczewska et al, 2000) and the poor survival associated with high serum IL-6 levels in this series of metastatic breast cancer patients. It is noteworthy that, as in early-stage breast cancer, IL-6 functions as an inhibitor of cancer cell growth in benign prostate hyperplasia (Degeorges et al, 1996), but is correlated with a poor survival in stage-D prostate cancer (Nakashima et al, 2000). Finally, as no serum acute-phase protein levels were assessed in this study, we cannot rule out that IL-6 serum levels in this population may merely be associated with a generalised inflammatory response to metastatic disease. IL-6 has been shown to be involved in the physiopathology of paraneoplastic inflammatory syndromes observed in patients with renal-cell carcinoma; a similar phenomenon might occur in metastatic breast cancer patients (Blay et al, 1997). In conclusion, this study shows that high serum IL-6 levels are independently correlated to survival in a relevant population of metastatic breast cancer patients. Our results suggest that IL-6 plays an important role in metastatic breast carcinoma progression in vivo. If confirmed on a larger, prospective study, this might prove helpful in identifying patients with a particularly poor prognosis, which is of utmost importance when planning therapy.
