Exploratory analyses of candidate predictive and prognostic tissue biomarkers (BMs) in the open-label randomised phase III TANIA trial of bevacizumab (BEV) in HER2-negative locally recurrent/metastatic breast cancer (LR/mBC)

In the TANIA trial (NCT01250379), adding BEV to 2nd-line chemotherapy (CT) significantly improved 2nd-line progression-free survival (PFS, primary endpoint; hazard ratio [HR] 0.75, 95% CI 0.61–0.93; p = 0.0068) in BEV-pretreated LR/mBC. No difference in 3rd-line PFS or overall survival (OS) was observed. Post hoc analyses explored potential associations between 6 candidate BMs (based on previous results/biological hypotheses) and 2nd-line PFS and OS. Patients (pts) with BEV-pretreated HER2-negative LR/mBC were randomised to receive 2nd-line CT ± BEV until disease progression (PD), unacceptable toxicity or withdrawal. At 2nd PD, 3rd-line CT was given without BEV in the CT arm and with BEV in the BEV + CT arm. Archival primary or metastatic tumour samples collected before 2nd-line BEV from pts consenting to optional translational research were analysed by immunohistochemistry using median expression levels for each BM as the cut-off for high vs low BM subgroups. Samples (mainly primary tumour) were available from 210 (43%) of 494 randomised pts. Baseline characteristics and efficacy in these pts were broadly representative of the ITT population. High CA9 and CD31 were associated with worse prognosis but larger BEV benefit, although neither was consistently predictive across endpoints (table). Pts with high CA9 had median PFS of 4.6 vs 1.9 mo with BEV + CT vs CT, respectively; pts with high CD31 had median OS of 19.3 vs 13.0 mo, respectively. No BM was associated with a detrimental BEV effect.Tabled 1BM2nd-line PFS HR (95% CI)Interaction p-valueaOS HR (95% CI)Interaction p-valuea≤Median>Median≤Median>MedianCD31 No. of vessels1.02 (0.68–1.55)0.60 (0.39–0.93)0.1101.42 (0.88–2.27)0.60 (0.37–0.99)0.016CD31 volume fraction0.85 (0.57–1.29)0.74 (0.48–1.15)0.5851.20 (0.74–1.95)0.75 (0.47–1.22)0.179tVEGF-A1.10 (0.73–1.65)0.63 (0.41–0.95)0.0910.99 (0.63–1.55)0.86 (0.53–1.38)0.626CA9 cytoplasm H-score0.92 (0.64–1.30)0.49 (0.30–0.81)0.0640.93 (0.62–1.39)0.93 (0.54–1.60)0.878CA9 membrane H-score0.90 (0.65–1.27)0.41 (0.23–0.74)0.0170.99 (0.67–1.47)0.70 (0.38–1.27)0.258PD-L1 immune cell0.84 (0.59–1.19)0.61 (0.36–1.04)0.3571.01 (0.68–1.51)0.68 (0.37–1.26)0.196PD-L1 tumour cell0.79 (0.58–1.06)0.96 (0.17–5.38)0.7950.92 (0.65–1.29)0.36 (0.04–3.16)0.469CD8 invasive margin0.77 (0.44–1.33)1.07 (0.63–1.83)0.3570.87 (0.46–1.65)1.35 (0.73–2.48)0.313CD8 tumour centre0.61 (0.40–0.93)0.97 (0.63–1.49)0.1550.87 (0.55–1.37)1.00 (0.62–1.63)0.711CD163 invasive margin1.00 (0.55–1.80)0.96 (0.56–1.64)0.8910.76 (0.40–1.44)1.65 (0.87–3.14)0.094CD163 tumour centre1.09 (0.71–1.67)0.71 (0.47–1.06)0.2620.91 (0.57–1.45)0.93 (0.58–1.48)0.886aNot adjusted for multiplicity. CA9 = carbonic anhydrase 9; PD-L1 = programmed cell death 1 ligand 1; tVEGF-A = tumour vascular endothelial growth factor-A. Open table in a new tab aNot adjusted for multiplicity. CA9 = carbonic anhydrase 9; PD-L1 = programmed cell death 1 ligand 1; tVEGF-A = tumour vascular endothelial growth factor-A. Continued BEV significantly improved 2nd-line PFS irrespective of BM levels in LR/mBC. Exploratory analyses suggested potential predictive effects of CA9 (PFS) and CD31 (OS), suggesting involvement of tumour and endothelium characteristics in the therapeutic concept of anti-angiogenesis.