Challenges in the Use of DNA Repair Deficiency As a Biomarker in Breast Cancer

The development of drug-biomarker combinations holds promise for tailoring treatment for patients with cancer on the basis of the biology of the tumor exemplified by estrogen receptor expression and blockade and human epidermal growth factor receptor 2 targeting. Approximately 85% of breast cancers have a target for therapy, but triple-negative breast cancers (TNBCs) do not. Although trials have focused on identifying the most efficacious anticancer drug (regimen) for this subtype, it may be that predictive biomarkers are required to guide treatment choices because of the intrinsic heterogeneity of TNBCs. Molecular characterizations have demonstrated a strong association between TNBCs and BRCA1 mutations. Preclinical mechanistic insight indicated that tumor cells with a defect in BRCA1 have impaired homologous recombination (HR), the only error-free pathway of repair of interstrand crosslinks. Cells with impaired HR display sensitivity to agents that induce such lesions. In the general breast cancer population, platinum agents that cause interstrand crosslinks are not preferred over other treatment regimens. However, given the possibility that TNBCs are enriched for such a targetable HR deficiency (HRD), platinum has attracted renewed interest as has been shown in recently published articles. Furthermore, TNBCs may be enriched for BRCA1 mutations and also for a larger group of nonmutated tumors that exhibit HRD. The use of platinum in TNBCs with biomarker analyses has been investigated in the two articles that accompany this editorial. Telli et al neoadjuvantly treated patients with triple-negative or BRCA1/ 2-associated breast cancer with a regimen containing gemcitabine, carboplatin, and iniparib in the PrECOG 0105 trial. Overall, a pathologic complete remission rate of 36% and acceptable toxicity of the regimen were reported. Isakoff et al conducted a phase II trial with cisplatin or carboplatin in patients with metastatic TNBC (TBCRC009; Platinum for Triple-Negative Metastatic Breast Cancer and Evaluation of p63/p73 as a Biomarker of Response). The overall response rate was 25%. Both studies reproduce previous findings of efficacy of a platinum-based regimen or platinum alone in TNBC. However, only a subset of patients in these trials derived clinical benefit. Therefore, secondary investigations may help identify biomarkers that associate with a preferential benefit from use of a platinum regimen, which may help guide therapy decisions. Both studies report their analysis of BRCA germline mutation status as well as potential biomarkers for assessing HRD, specifically the HRD loss of heterozygosity (HRD-LOH) score and the HRD large-scale transition (HRD-LST) score. Both markers evaluate so-called genomic scars as signatures of aberrations that can be observed from singlenucleotide polymorphism array–based sequencing or similar technologies and that are associated with defects in error-free repair of interstrand crosslinks. The HRD-LOH score counts the numbers of LOH regions larger than 15 MB but smaller than chromosome size, whereas the HRD-LST score counts the number of breaks between adjacent segments of at least 10 MB. Cutoffs were trained to best separate tumors that had a known deleterious aberration in BRCA1/2 from those that did not. In both studies, patients with a BRCA germline mutation had higher response rates than the general cohort, and the HRD scores highly correlated with BRCA1/2 germline mutations and response rates. In addition, Telli et al identified BRCA1 methylation in nonmutated tumors with high HRD-LOH score, which suggests that these scores identify a subgroup of patients who could benefit from therapy that targets their HRD. Unfortunately, single-arm phase II studies do not allow proper evaluation of a biomarker as a predictor of response to therapy, and they offer limited insight on how to exploit the observed association with outcome of BRCA-associated biomarkers (Fig 1). Thus, the studies by Telli et al and Isakoff et al join several other articles published over the last decade that have reported on the potential promise of BRCA-associated HRD biomarkers but did not provide definitive data on clinical utility. It has been established that the optimal biomarker trial design requires biomarker-positive and -negative patients and patients who have been treated with and without the therapy of interest to evaluate both prognostic and predictive signals. If not included in the phase II design, such a control group could be added. Hence, the lack of a control arm in these two new studies is indeed a major limitation of their findings. In contrast, the TNT trial (Triple Negative Breast Cancer Trial, recently reported as an abstract) was able to more appropriately evaluate the use of HRD biomarkers, at least in patients with advanced disease. In that trial, patients with locally advanced or metastatic TNBC were randomly assigned to therapy with carboplatin at area under the [concentration-time] curve 6 or standardof-care docetaxel. Of great interest, among the 376 patients in that study, 29 were known to be BRCA1 or BRCA2 mutation carriers (including 13 who had estrogen receptor–positive, HER2-negative disease). Prespecified analyses specifically addressed populations JOURNAL OF CLINICAL ONCOLOGY E D I T O R I A L VOLUME 33 NUMBER 17 JUNE 1