Serum microRNAs hold great promise as easily accessible and measurable biomarkers of disease. In prostate cancer, serum miRNA signatures have been associated with the presence of disease as well as correlated with previously validated risk models. However, it is unclear whether miRNAs can provide independent prognostic information beyond current risk models. Here, we focus on a group of low-risk prostate cancer patients who were eligible for active surveillance, but chose surgery. A major criteria for the low risk category is a Gleason score of 6 or lower based on pre-surgical biopsy. However, a third of these patients are upgraded to Gleason 7 on post surgical pathological analysis. Both in a discovery and a validation cohort, we find that pre-surgical serum levels of miR-19, miR-345 and miR-519c-5p can help identify these patients independent of their pre-surgical age, PSA, stage, and percent biopsy involvement. A combination of the three miRNAs increased the area under a receiver operator characteristics curve from 0.77 to 0.94 (p<0.01). Also, when combined with the CAPRA risk model the miRNA signature significantly enhanced prediction of patients with Gleason 7 disease. In-situ hybridizations of matching tumors showed miR-19 upregulation in transformed versus normal-appearing tumor epithelial, but independent of tumor grade suggesting an alternative source for the increase in serum miR-19a/b levels or the release of pre-existing intracellular miR-19a/b upon progression. Together, these data show that serum miRNAs can predict relatively small steps in tumor progression improving the capacity to predict disease risk and, therefore, potentially drive clinical decisions in prostate cancer patients. It will be important to validate these findings in a larger multi-institutional study as well as with independent methodologies.