To date, men undergoing screening through the measurement of prostate-specific antigen (PSA) levels have had a significant reduction in neoplastic mortality. Because of its low specificity, however, this practice often leads to frequent, unnecessary, invasive biopsies and the diagnosis of low-grade, indolent cancer.
The National Comprehensive Cancer Network proposes a test consisting of six blood and urine biomarkers for all grades of prostate cancer, and it outperforms PSA testing. However, current practice focuses on detecting high-grade cancers. It has been hypothesized that increasing the number of biomarkers by including molecules specifically expressed in aggressive high-grade prostate cancers could improve test accuracy. Based on the identification of new genes that are overexpressed in high-grade cancers, a polymerase chain reaction (PCR) technique targeting 54 candidate markers was used to develop an optimal 18-gene test that could be used before imaging (with MRI) and biopsy and to assess whether the latter procedures are warranted.
Development Cohort
In the development cohort (n = 815; median age, 63 years), quantitative PCR (qPCR) analysis of the 54 candidate genes was performed on urine samples that had been prospectively collected before biopsy following a digital rectal examination. Patients with previously diagnosed prostate cancer, abnormal MRI results, and those who had already undergone a prostate biopsy were excluded. Participants’ PSA levels ranged from 3 to 10 ng/mL (median interquartile range [IQR], 5.6 [4.6-7.2] ng/mL). Valid qPCR results were obtained from 761 participants (93.4%). Subsequently, prostate biopsy revealed grade 2 or higher cancer in 293 participants (38.5%).
Thus, a urine test called MyProstateScore 2.0 (MPSA) was developed, with two formulations: MPSA2 and MPSA2+, depending on whether a prostate volume was considered. The final MPSA2 development model included clinical data and 17 of the most informative markers, including nine specific to cancer, which were associated with the KLK3 reference gene.
Validation and Analyses
The external validation cohort (n = 813) consisted of participants in the NCI EDRN PCA3 Evaluation trial. Valid qPCR results were obtained from 743 participants, of whom 151 (20.3%) were found to have high-grade prostate cancer.
The median MPS2 score was higher in patients with grade 2 or higher prostate cancer (0.44; IQR, 0.23-0.69) than in those with noncontributory biopsies (0.08; IQR, 0.03-0.19; P < .001) or grade 1 cancer (0.25; IQR, 0.09-0.48; P < .01).
Comparative analyses included PSA, the Prostate Cancer Prevention Trial risk calculator, the Prostate Health Index (PHI), and various previous genetic models. Decision curve analyses quantified the benefit of each biomarker studied. The 151 participants with high-grade prostate cancer had operating curve values ranging from 0.60 for PSA alone to 0.77 for PHI and 0.76 for a two-gene multiplex model. The MPSA model had values of 0.81 and 0.82 for MPSA2+. For a required sensitivity of 95%, the MPS2 model could reduce the rate of unnecessary initial biopsies in the population by 35%-42%, with an impact of 15%-30% for other tests. Among the subgroups analyzed, MPS2 models showed negative predictive values of 95%-99% for grade 2 or higher prostate cancers and 99% for grade 3 or higher tumors.