While active surveillance (AS) is the preferred management strategy for low risk prostate cancer, frequent clinical monitoring and invasive biopsy are associated with costs, discomfort and overtreatment among a group of men at minimal risk of prostate cancer-related mortality. Non-invasive markers may help guide de-escalation strategies, though none are consistently associated with risk of biopsy Gleason grade group (GG) upgrading. Our group previously determined that circulating sphingolipid levels may be associated with biopsy GG upgrading among men on AS. We describe the development and validation of a sphingolipid-based biomarker and its performance with established clinical factors in this population. Men diagnosed with low- or intermediate-risk prostate cancer in one of two AS cohorts (CANARY PASS [N=544] and MDACC [N=238]) were followed for Gleason GG upgrading after diagnostic and confirmatory biopsy. Plasmas collected prior to confirmatory biopsy were used in mass spectrometry-based lipidomic assays, and sphingolipid metabolite levels were quantified. We leveraged machine learning in 8 different models to develop a panel of sphingolipids that was associated with biopsy GG upgrading, and the sphingolipid panel was set. Multivariable cox proportional hazards models were used to determine association between sphingolipid panel and risk of biopsy GG upgrading. Cut points were then established using a combination of panel score and clinical factors. Median age was 67 (IQR 58-70) in the CANARY PASS cohort and 63 (IQR 58-68) in the MDACC cohort. 85.5% and 90.0% of patients had GG1 disease, respectively. Median follow-up was 2.1 years (1.4-4.5) and 3.5 years (1.0-5.1) and 98/544 (18%) and 33/238 (14%) had GG upgrading, respectively. A panel of sphingolipids generated using a neural network achieved the best performance, and was independently associated with GG upgrade on multivariable models in the PASS cohort (HR of 1.33, 95% CI 1.05-1.70 per StDev increase) and the MDACC validation cohort (HR 2.51, 95% CI: 1.42-2.4 per unit StDev increase). When combined with clinical factors PSA density and core biopsy positivity, tertile-based cutoffs demonstrated that high risk group stratification was associated with biopsy GG upgrade when compared to the low risk group in both cohorts (HR 3.17, 95% CI 1.84-5.46; and, HR 9.70, 95% CI 2.89-32.5, respectively). The sphingolipid panel is independently associated with GG biopsy upgrading among men in two independent AS cohorts who have previously undergone diagnostic and confirmatory biopsy. Use of the sphingolipid panel, along with cut offs that include important clinical factors, may allow for safe de-escalation strategies among men on AS. Future studies are needed to further validate these data and incorporate other clinical factors such as MRI results. [ABSTRACT FROM AUTHOR]