Plasma (p) VEGF-A and VEGFR-2 biomarker (BM) results from the BEATRICE phase III trial of bevacizumab (BEV) in triple-negative early breast cancer (BC)

Background: Several candidate BMs have been explored in randomized trials of BEV across tumor types with the aim of identifying patients (pts) deriving the most substantial benefit from BEV therapy. In phase III trials, baseline pVEGF-A and pVEGFR-2 showed potential predictive value in metastatic BC (AVADO, AVEREL), pancreatic cancer (AViTA), and gastric cancer (AVAGAST; VEGF-A only). The randomized phase III BEATRICE trial, evaluating the addition of BEV to adjuvant chemotherapy in pts with triple-negative early BC, includes a comprehensive program to identify potential BMs predicting efficacy and toxicity of BEV therapy. We report results for baseline pVEGF-A and pVEGFR-2. Methods: After selection of chemotherapy (anthracycline and/or taxane), pts with T1a-T3 operable BC were randomized 1:1 to receive 4 cycles of chemotherapy either alone or with 1 year of BEV 5 mg/kg/wk equivalent. The primary endpoint is invasive disease-free survival (IDFS). pEDTA samples were collected from consenting pts at baseline (before treatment, after surgery), during study treatment, and at relapse. Pts were dichotomized using the median baseline concentration of each marker as the cut-off between high and low cohorts; further exploratory analyses were also performed by quartile. Results: Between Dec 2007 and Mar 2010, 2591 pts were enrolled. Of these, 1273 (49%) consented to the BM study and 1178 (45%) were included in the BM-evaluable population (BEP). Overall, the BEP was representative of the ITT population except for lower proportions of Asian pts (12% vs 24%). IDFS was similar in the BEP and ITT populations. Baseline characteristics were balanced between arms in the BEP. Baseline pVEGF-A showed neither prognostic nor predictive value using the median as the cut-off, although with a third quartile (Q3) cut-off there was a more pronounced but non-significant differentiation between treatments (HR 0.92 [low] vs 0.64 [high]). High baseline pVEGFR-2 showed potential predictive value for BEV efficacy (HR 1.24 [low] vs 0.61 [high]; p=0.029). (Table presented) Conclusions: Unlike trials in metastatic BC (AVADO, AVEREL), in the adjuvant setting, baseline pVEGF-A concentration did not show predictive value for BEV efficacy with a median cut-off. However, analyses using the Q3 cut-off suggest a trend toward predictive value. High baseline pVEGFR-2 was associated with greater BEV treatment effect, consistent with previous results in AVADO and AVEREL. The impact of differing biology in the adjuvant setting, lower median VEGF-A concentration than in the metastatic setting (77.0 vs 125.0-129.1 pg/mL), and the possible influence of surgery immediately before treatment require further investigation. Additional exploratory analyses are ongoing to provide better understanding of the BEATRICE dataset and the complex biology of angiogenesis, including additional markers, changes over time, and combination signatures.