Vascular endothelial growth factor in node-positive breast cancer patients treated with adjuvant tamoxifen

In 212 postmenopausal women with node-positive oestrogen receptor-positive (ERLBA) breast cancer subjected to radical surgery and adjuvant tamoxifen, the risk of 6-year relapse increased with increasing values of intratumoral vascular endothelial growth factor (VEGF) in patients whose tumours had a low/intermediate ERLBA content compared to patients with high-ERLBA tumours. These findings indicate that tumour progression, activated or sustained by high VEGF levels, may be counteracted in high-ERLBA cancers by tamoxifen, which in contrast fails to contrast the metastatic potential in low-ERLBA tumours.

Breast tissue is highly responsive to changes in ovarian hormone concentrations that induce a cyclic remodelling process involving epithelial, stromal and vascular components during each menstrual cycle (Vogel et al, 1981). In particular, the angiogenic turnover is regulated by oestrogens and progesterone, which modulate the expression of vascular endothelial growth factor (VEGF) in the epithelial cells of the terminal ductal -lobular units (Nakamura et al, 1999).
Although the VEGF gene promoter lacks a perfect palindromic oestrogen-responsive element, an analogue of the consensus element that functions as a classical enhancer for oestrogen receptor (ER) has been identified in the 3 0 -untranslated region (Hyder et al, 2000b).
Angiogenesis is known to represent a fundamental step in tumour progression (Hanahan and Folkman, 1996) and clinical evidence has shown that node-negative (NÀ) as well as nodepositive (N þ ) breast cancer patients with high intratumoral VEGF concentrations have a significantly shorter relapse-free survival (RFS) (Gasparini et al, 1997(Gasparini et al, , 1999Linderholm et al, 2000).
Preclinical studies on breast cancer cell lines demonstrated that, as expected, oestrogens rapidly induce VEGF expression, which is blocked by pure oestrogen antagonists (Ruohola et al, 1999;Hyder et al, 2000a).
In a series of N þ ER-positive (ER þ , by ligand-binding assay, LBA) postmenopausal women with resectable breast cancer who received adjuvant tamoxifen, we investigated the comprehensive effect on RFS of VEGF content and steroid receptor profile, evaluated in the same cytosolic fraction.

Patients
The study included postmenopausal patients with primary resectable invasive breast cancer, histologically classified as N þ , who underwent surgery at the Istituto Nazionale Tumori in Milan between March 1991 and December 1995 and received only adjuvant tamoxifen (20 mg day À1 ) for at least 2 years (median duration time, 4 years) because of their positive ER status (ER tumour concentration higher than 10 fmol mg À1 of protein). From a total of 859 N þ , ER þ postmenopausal patients, consecutive with respect to steroid receptor determination at the time of diagnosis, 289 were selected on the basis of treatment, histology (pure or mixed ductal or lobular invasive tumours) and follow-up (i.e. a minimum potential of 6 years from the date of surgery to the date of last updating of patient records).
Of the 289 eligible patients, 212 (73%) were available for VEGF evaluation. Their median age was 64 years (range, 50 -85); 88 patients (42%) were treated by mastectomy and 124 (58%) by breast-conserving surgery plus radiotherapy. All of them underwent complete axillary lymph node dissection (median number of examined nodes, 18). Most patients had one to three metastatic axillary lymph nodes (139, 66%). Small (p2 cm) and large (42 cm) tumours were equally represented (93, 48% and 99, 52%, respectively). After surgery, the patients were followed at 6month intervals during the first 5 years and at 12-month intervals thereafter; disease status was assessed by means of physical examination, chest X-ray, bone scan and abdominal sonography. Treatment failure was defined as the first documented evidence of new disease manifestations in locoregional areas (seven cases), distant sites (40 cases), or in the contralateral breast (five cases). Relapse-free survival was calculated as the time elapsed from diagnosis to the date of first recurrence or to the last clinical examination for patients without documented disease manifestation. Median follow-up for the whole series was 68 months (interquartile range, 52 -82 months).

Steroid receptor determination by LBA
Steroid receptor content was determined according to the EORTC recommendations and within national (Piffanelli et al, 1991) and international (EORTC Breast cancer co-operative group, 1980) quality control programmes by a double-labelling assay (Coradini et al, 2000), and expressed as fmol mg À1 of protein. Tumours with an ER LBA concentration higher than 10 fmol mg À1 of protein were defined as ER þ .

Vascular endothelial growth factor determination
The predominant VEGF isoform, VEGF 65 (henceforward referred to as VEGF), was measured by a quantitative enzyme immunoassay technique (Quantikine, human VEGF; R&D Systems, Minneapolis, MN, USA) as described elsewhere (Coradini et al, 2001). Concentrations were expressed as pg of VEGF protein per mg of total protein.

Statistical analysis
The overall association of VEGF level with patient age, tumour size, number of metastatic lymph nodes, ER LBA and PgR LBA content was evaluated by Spearman's rank correlation coefficient.
The effect of VEGF, ER LBA and PgR LBA content on RFS was investigated by multivariate analysis using a Cox regression model in which also the number of metastatic lymph nodes was included. All variables were considered on a continuous scale after logarithmic transformation. Null values for PgR LBA content were arbitrarily set at 1, taking a sensitivity threshold value of 2 fmol mg À1 of protein. According to a previous finding (Coradini et al, 2001), linear terms for log(ER LBA ), log(PgR LBA ) and log(VEGF) and the interaction between ER LBA and VEGF were included in the model.
The proportional hazard assumption of the Cox model was evaluated and the effect of model terms was tested as previously reported (Coradini et al, 2001).
The library written by Harrell et al (1996) was applied in some steps of the model building procedure and the SAS macro programme RELIMPCR designed by Heinze and Schemper (2001) was adopted for evaluation of the relative prognostic contribution of the covariates.

RESULTS
In this series of ER LBA -positive tumours from postmenopausal patients, the VEGF content ranged from 7 to 2186 pg mg À1 of protein with 52, 95 and 200 as the 25th, 50th and 75th percentiles, respectively; the PgR LBA concentration ranged from 1 to 2563 fmol mg À1 of protein with 40, 111 and 355 as the 25th, 50th and 75th percentiles, respectively. Vascular endothelial growth factor content did not show any significant correlation with any of the other variables considered; estimated correlation coefficients were all in the range 70.08.
In the multivariate model, the number of metastatic lymph nodes, VEGF, ER LBA and the interaction between ER LBA and VEGF were significantly related to prognosis (Table 1), whereas no significant contribution was observed for PgR LBA . As for ER LBA , VEGF and the ER LBA -VEGF interaction, the results were similar to those previously obtained in patients with N -breast cancer who did not receive systemic treatment (Coradini et al, 2001). This finding further supports the role of ER LBA in modulating the prognostic effect of VEGF and the negative interaction term indicates a decrease in the unfavourable effect of VEGF for increasing values of ER LBA .
To provide a description of the combined effect of VEGF and ER LBA , the relative hazard (RH) for increasing VEGF concentrations was plotted (Figure 1) for selected values of ER LBA (70 and 220 fmol mg À1 of protein) that were approximately equal to the 1st and third quartiles of ER LBA distribution. Owing to the interaction between ER LBA and VEGF, when the ER LBA content was set at 70 fmol mg À1 protein, the increase in the risk of disease recurrence for high VEGF values associated with low ER LBA values was evident. In fact, the estimated RH for patients with a VEGF content of 200 vs those with a VEGF content of 50 pg mg À1 protein was 1.54 (95% CI, 1.07 -2.20). Conversely, when the ER LBA content was set at 220 fmol mg À1 protein, VEGF did not show any prognostic effect and the estimated RH for the two selected values (200 and 50 pg ml À1 ) was 0.89 (95% CI, 0.54 -1.47).
Analysis of the prognostic contribution of the covariates showed that the model explained only 13.2% of the total heterogeneity of the relapse times of our case series and that, considering the relative importance of prognostic factors, 8.2% was attributable to the number of metastatic lymph nodes, whereas VEGF and ER LBA together contributed with 3.3%, and PgR only with 0.8%.

DISCUSSION
According to the recently revised treatment guidelines for early breast cancer (Goldhirsch et al, 2001), postmenopausal women with ER þ primary breast cancer should receive adjuvant hormone therapy with tamoxifen. However, a recent clinical study (Linderholm et al, 2000) provided evidence that also the intratumoral VEGF content could predict outcome following adjuvant endocrine treatment: patients with ER þ tumours, but a high VEGF expression, had a significantly shorter RFS and overall survival. In agreement with these findings, our results showed that, due to the presence of a negative interaction between ER LBA and VEGF, already observed in a series of NÀ cancers (Coradini et al, 2001), patients whose tumours had a low/ intermediate ER LBA content exhibited an increased risk of disease recurrence with increasing values of intratumoral VEGF. Conversely, the VEGF level did not show prognostic effect in patients whose tumours had a high ER LBA content. These findings suggest that in tumours characterised by a high ER LBA concentration and therefore more likely to be hormonally regulated, tumour progression, activated or sustained by VEGF, may be counteracted by the protective effect of endocrine therapy. Conversely, when a patient has a high level of VEGF and a low ER LBA concentration, tamoxifen could fail to contrast the tumour's metastatic potential and more tailored adjuvant treatment would be required including, for example, an antiangiogenic agent.