Association of RAD51 with homologous recombination deficiency (HRD) and clinical outcomes in untreated triple-negative breast cancer (TNBC): analysis of the GeparSixto randomized clinical trial

Background: Current genetic and genomic tests measuring homologous recombination deficiency (HRD) show limited predictive value. This study compares the performance of an immunohistology-based RAD51 test with genetic/genomic tests to identify patients with HRD primary triple-negative breast cancer (TNBC) and evaluates its accuracy to select patients sensitive to platinum- based neoadjuvant chemotherapy (NACT). Patients and methods: This is a retrospective, blinded, biomarker analysis from the GeparSixto randomized clinical trial. TNBC patients received neoadjuvant paclitaxel plus Myocet (R)-nonpegylated liposomal doxorubicin (PM) or PM plus carboplatin (PMCb), both arms including bevacizumab. Formalin-fixed paraffin-embedded (FFPE) tumor samples were laid on tissue microarrays. RAD51, BRCA1 and gH2AX were quantified using an immunofluorescence assay. The predictive value of RAD51 was assessed by regression models. Concordance analyses were carried out between RAD51 score and tumor BRCA (tBRCA) status or genomic HRD score (Myriad myChoice (R)). Associations with pathological complete response (pCR) and survival were studied. Functional HRD was predefined as a RAD51 score <= 10% (RAD51-low). Results: Functional HRD by RAD51-low was evidenced in 81/133 tumors (61%). RAD51 identified 93% tBRCA-mutated tumors and 45% non-tBRCA mutant cases as functional HRD. The concordance between RAD51 and genomic HRD was 87% [95% confidence interval (CI) 79% to 93%]. In patients with RAD51-high tumors, pCR was similar between treatment arms [PMCb 31% versus PM 39%, odds ratio (OR) 0.71, 0.23-2.24, P = 0.56]. Patients with RAD51-low tumors benefited from PMCb (pCR 66% versus 33%, OR 3.96, 1.56-10.05, P = 0.004; interaction test P = 0.02). This benefit maintained statistical significance in the multivariate analysis. Carboplatin addition showed similar diseasefree survival in the RAD51-high [hazard ratio (HR) 0.40, log-rank P = 0.11] and RAD51-low (0.45, P = 0.11) groups. Conclusions: T