Presentation Authors: Alexander Haese*, Hamburg, Germany, Geert Trooskens, Sandra Steyaert, Daphne Hessels, Michael Brawer, Irvine, CA, Virginie Vlaeminck-Guillem, Lyon, France, Jack Schalken, Nijmegen, Netherlands, Jack Groskopf, Irvine, CA, Wim Van Criekinge, Ghent, Belgium
Introduction: A 2-gene, urine-based molecular test that combines mRNA biomarkers with clinical factors can risk stratify patients for high-grade prostate cancer (PCa). To ensure the generalizability of the risk score, the model was optimized and validated for patients from three European countries who were undergoing initial prostate biopsy.
Methods: The study population consisted of 1956 men suspected of PCa from The Netherlands, France and Germany who underwent initial prostate biopsies between December 2007 and December 2014. The median age was 65 years (interquartile range 60-70), median serum PSA 6.3 ng/mL (IQR 4.5-9.3) and 485/1956 (25%) had abnormal DRE results. PCa was detected in 991/1956 (51%) men biopsied: 46% ISUP grade group (GG) 1, 27% GG2, 12% GG3, 7% GG4 and 8% GG5. The prevalence of GG2-5 PCa was 27.6% (540/1956)._x000D_
Post-DRE urine was collected from all subjects prior to biopsy, and samples were stored at -700C. Urinary HOXC6 and DLX-1 mRNAs were quantified by PCR (MDxHealth, Irvine CA), and the RNA results were next combined with other risk factors (age, serum PSA level, DRE, prostate volume and family history of PCa) to determine the likelihood that subsequent biopsy would identify GG > 2 (Gleason Score > 7) PCa. _x000D_
Subjects were divided into Discovery (N=1040) and Validation (N=916) cohorts: the Dutch cohorts have been previously described*, French and German groups were evenly split by randomization. The combination of multiple risk factors was assessed by generating linear models, resulting in a continuous Risk Score. Models were optimized for maximum sensitivity and negative predictive value (NPV) for >= GG2 PCa in subjects with PSA < 10 ng/mL. Clinical performance was assessed at a fixed Risk Score cutoff of -2.8*. The PCPT Risk Calculator Version 2.0 (PCPTRC, http://myprostatecancerrisk.com, accessed October 7, 2018) was used for comparison._x000D_
*Van Neste L, Hendriks RJ, Dijkstra S, et al; Detection of High-gradeÂ Prostate CancerÂ Using a Urinary Molecular Biomarker-Based Risk Score. Eur Urol 2016 Nov;70(5):740-748.
Results: For the Discovery set, the optimal model included urinary HOXC6+DLX1, age, DRE and PSA density (serum PSA/prostate volume). For the Validation set, the AUC was 0.87, sensitivity 95%, specificity 49% and NPV 96%, whereas the PCPTRC AUC was 0.69 (0.65-0.74), with sensitivity of 95% and specificity of 16%. For the subset of patients with PSA < 10 ng/mL, the Validation set yielded an AUC of 0.82 (0.79-0.86), sensitivity 89%, specificity 53% and NPV 95%. Overall performance for the combined Discovery and Validation cohorts (all PSA levels, N=1956) was AUC 0.85 (0.83-0.87), with sensitivity and NPV of 95% and 96%, respectively.
Conclusions: The biomarker-based clinical model, optimized primarily for biopsy naive patients with serum PSA < 10 ng/mL, demonstrated high sensitivity and NPV for detection of GG2 and higher PCa across all patient groups. Clinical performance characteristics were equivalent to the published clinical validation study. These data support use of the test to help guide initial prostate biopsy decisions.
Source of Funding: MDxHealth