Search

Background: Androgen receptor (AR) signaling inhibitors represent the standard treatment in metastatic castration resistance prostate cancer (mCRPC) patients. However, some patients display a primary resistance, and several studies investigated the role of the AR as a predictive biomarker of response to treatment. This study is aimed to evaluate the role of AR in liquid biopsy to predict clinical outcome to AR signaling inhibitors in mCRPC patients. Methods: Six milliliters of plasma samples were collected before first-line treatment with abiraterone or enzalutamide. Circulating free DNA (cfDNA) and exosome-RNA were isolated for analysis of AR gain and AR splice variant 7 (AR-V7), respectively, by digital droplet PCR. Results: Eighty-four mCRPC patients received abiraterone (n= 40) or enzalutamide (n= 44) as first-line therapy. Twelve patients (14.3%) presented ARgain and 30 (35.7%) AR-V7+ at baseline. Median progression-free survival (PFS) and overall survival (OS) were significantly longer in AR-V7− vs AR-V7+ patients (24.3 vs 5.4 months, p< 0.0001; not reached vs 16.2 months, p= 0.0001, respectively). Patients carrying the ARgain had a median PFS of 4.8 vs 24.3 months for ARnormal patients (p< 0.0001). Median OS was significantly longer in ARnormal vs patients with ARgain (not reached vs 8.17 months, p< 0.0001). A significant correlation between AR-V7 and ARgain was observed (r= 0.28; p= 0.01). The AR gain/AR-V7 combined analysis confirmed a strong predictive effect for biomarkers combination vs patients without any AR aberration (PFS 3.8 vs 28 month, respectively; OS 6.1 vs not reached, respectively; p< 0.0001). Conclusions: The present study demonstrates that cfDNA and exosome-RNA are both a reliable source of AR variants and their combined detection in liquid biopsy predicts resistance to AR signaling inhibitors.

A trust-region algorithm is presented for finding approximate minimizers of smooth unconstrained functions whose values and derivatives are subject to random noise. It is shown that, under suitable probabilistic assumptions, the new method finds (in expectation) an ϵ-approximate minimizer of arbitrary order q≥1in at most O(ϵ−(q+1))inexact evaluations of the function and its derivatives, providing the first such result for general optimality orders. The impact of intrinsic noise limiting the validity of the assumptions is also discussed and it is shown that difficulties are unlikely to occur in the first-order version of the algorithm for sufficiently large gradients. Conversely, should these assumptions fail for specific realizations, then “degraded” optimality guarantees are shown to hold when failure occurs. These conclusions are then discussed and illustrated in the context of subsampling methods for finite-sum optimization.