Predictive and Prognostic Biomarker Identification in a Large Cohort of Androgen Receptor-Positive Salivary Duct Carcinoma Patients Scheduled for Combined Androgen Blockade

SIMPLE SUMMARY: Androgen receptor signaling seems pivotal in aggressive salivary duct carcinoma. However, androgen deprivation therapy (ADT) frequently fails, emphasizing the need for biomarkers to predict treatment response. Here, the activities of several possible tumor-driving pathways were quantified and related to clinical outcome in a large cohort of SDC patients. Our results indicated that AR pathway activity and SRD5A1 expression levels can be used to predict response to ADT, which could prevent overtreatment in clinical practice. ABSTRACT: Patients suffering from recurrent or metastatic (R/M) salivary duct carcinoma (SDC) are often treated with combined androgen blockade (CAB). However, CAB frequently fails, resulting in a worse prognosis. Therefore, biomarkers that can predict treatment failure are urgently needed. mRNA from 76 R/M androgen receptor (AR)-positive SDC patients treated with leuprorelin acetate combined with bicalutamide was extracted from pre-treatment tumor specimens. AR, Notch, MAPK, TGFβ, estrogen receptor (ER), Hedgehog (HH), and PI3K signaling pathway activity scores (PAS) were determined based on the expression levels of target genes. Additionally, 5-alpha reductase type 1 (SRD5A1) expression was determined. These markers were related to clinical benefit (complete/partial response or stable disease ≥6 months) and progression-free and overall survival (PFS/OS). SRD5A1 expression had the highest general predictive value for clinical benefit and positive predictive value (PPV: 85.7%). AR PAS had the highest negative predictive value (NPV: 93.3%). The fitting of a multivariable model led to the identification of SRD5A1, TGFβ, and Notch PAS as the most predictive combination. High AR, high Notch, high ER, low HH PAS, and high SRD5A1 expression were also of prognostic importance regarding PFS and SRD5A1 expression levels for OS. AR, Notch PAS, and SRD5A1 expression have the potential to predict the clinical benefit of CAB treatment in SDC patients. SRD5A1 expression can identify patients that will and AR PAS patients that will not experience clinical benefit (85.7% and 93.3% for PPV and NPV, respectively). The predictive potential of SRD5A1 expression forms a rational basis for including SRD5A1-inhibitors in SDC patients’ treatment.