either ccA (less aggressive) or ccB (more aggressive) molecular sub- types. Age-and sex-adjusted logistic regression models estimated as- sociations between sarcopenia and molecular subtype separately for obese and non-obese patients. Statistical significance was regarded as a p-value of<0.05. RESULTS: The cohort was predominantly male (77%), white (97%), and had localized disease (62%). Median age was 58.7 years (IQR: 34-86.7). Overall, 53% of patients were obese, 39% were sar- copenic, and 58% of tumors were ccB subtype. Sarcopenic patients were more likely to have ccB tumors (66.7%) compared to patients without sarcopenia (26.1%) p¼0.00008. Among patients who were not obese, aggressive ccB subtype was more common in sarcopenic (69.6%) than non-sarcopenic patients (30.8%) (p¼0.03). A similar pattern was observed among patients who were obese; aggressive ccB subtype was more common in sarcopenic (57.1%) than non-sarcopenic patients (24.2%) (p¼0.04). CONCLUSIONS: While preliminary, our findings suggest that sarcopenia is associated with aggressive ccRCC regardless of obesity and lend biologic support to the observation that sarcopenia is asso- ciated with poor prognosis. It is not clear whether sarcopenia is a cause or consequence of tumor aggressiveness. RNA-Seq analysis of tumor tissue is being carried out to explore specific mechanisms underlying these observations. Source of Funding: Chanel grant (HF) and SOAR grant (HF & AJD) and Ruth L. Kirschstein Research Service Award T32CA082088 (AS). MP72-18 ASSOCIATION OF COMPUTED TOMOGRAPHY-BASED RADIOMIC FEATURES WITH EPIGENETIC VARIATION OF CLEAR CELL RENAL CELL CARCINOMA USING DNA METHYLATION Vinay Duddalwar*, Gangning Liang, Kim Siegmund, Steven Cen, Bino Varghese, Darryl Hwang, Bhushan Desai, Mihir Desai, Manju Aron, Inderbir Gill, Los Angeles, CA INTRODUCTION AND OBJECTIVES: Current standards of tumor assessment often do not provide a comprehensive evaluation for the optimal management of renal cancer patients. While genomic and epigenomic biomarkers have shown improvements in patient risk classification, the heterogeneity of renal tumors on imaging can confound assessment. The objective was to investigate the association of contrast-enhanced computed tomography (CECT)-based radiomics obtained from clinical imaging features with DNA-methylation (DNAm) features that predict aggressive clear cell renal carcinoma (ccRCC). METHODS: In this IRB-approved retrospective study, 29 pa- tients (15 aggressive and 14 non-aggressive) with pathologically confirmed ccRCC and a pre-operative multiphase CECT were identi- fied. An abdominal radiologist supervised the segmentation of each tumor from a clinical, standard of care CT study on a dedicated work- station. We evaluated 6 different types of texture extraction techniques from every phase of the imaging study. Genome-wide DNAm profiles were generated from tumor tissue extracted from the same patients using the Infinium Methylation EPIC array. We correlated all 1495 radiomic features processed with 97 DNAm features selected based on their strength of separating non-aggressive from aggressive renal tu- mors and non-aggressive from normal kidney in 281 ccRCC from the Cancer Genome Atlas. We computed Pearson and Spearman corre- lations depending on data distribution, for all possible pairs of radiomics and DNAm features. The results were summarized within each of the 6 radiomics categories by the percent of correlation coefficients attaining nominal p<0.05 for each DNAm feature (suggestive of a positive signal) and presented in a heatmap (Figure 1). RESULTS: There were significant (p<0.05) correlations be- tween the selected epigenetic features and the radiomics features. Gray-level difference method (GLDM) 2D and 3D were shown having a strong signal in the correlation with cg12697139 and cg16973527. In general, second-order statistical texture method: GLDM was seen to have the most correlation with selected epigenetic nodes. CONCLUSIONS: To our knowledge, this is the first study to correlate epigenomic profiles with CECT based radiomics features of renal masses. We plan to validate the finding in a larger patient cohort study. Source of Funding: Whittier Foundation MP72-19 PLASMA GLYCOSAMINOGLYCAN SCORES IN RENAL CELL CARCINOMA Kyle A. Blum*, New York City, NY; Francesco Gatto, Goteberg, Sweden; Mazyar Ghannat, Alejandro Sanchez, New York City, NY; Francesca Maccari, Fabio Galeotti, Modena, Italy; James Hsieh, St. Louis, MO; Nicola Volpi, Modena, Italy; A. Ari Hakimi, New York City, NY; Jens Nielsen, Goteberg, Sweden INTRODUCTION AND OBJECTIVES: Glycosaminoglycan (GAG) levels are measurably altered in the plasma of patients with clear cell renal cell carcinoma (ccRCC). GAG scores have been used to detect ccRCC in a cohort of patients with metastatic disease with 92.7% accuracy (Gatto et al, Cell Reports, 2016). However, it is unknown if GAG scores can detect RCC in earlier stages or non-ccRCC histologies. METHODS: In this retrospective study, pre-operative plasma GAG levels from 162 RCC patients were compared to 19 healthy controls between 5/2011-2/2014. GAG scores were generated using 19 pre-specified properties and measured using capillary electrophoresis with laser induced fluorescence. GAG profile differences in RCC versus healthy controls were assessed using unsupervised clustering methods. A discovery set of 68 RCC vs. 19 healthy samples were first analyzed to update the historical GAG score. The new GAG score accuracy in RCC detection versus healthy subjects was validated using the remaining 94 RCC samples. RESULTS: Median age was similar between RCC and healthy cohorts, 60 years (IQR: 52-67) vs. 55 years (IQR: 50-60), respectively. In the RCC cohort, 113 (70%) were ccRCC and 49 (30%) non-ccRCC. RCC stage included 86 (53%) pT1, 66 (41%) pT2-3, and 12 (7%) pT4. GAG profiles in RCC clustered separately from healthy volunteers (Figure 1A). In the first discovery set (n¼68), the GAG score distin- guished RCC from healthy subjects with an AUC equal to 0.999, 94.7% specificity and 100% sensitivity at an optimal cut-off equal to 0.87 (Figure 1B). In the validation set (n¼94), the GAG score achieved an AUC equal to 0.994 (95% CI: 0.985 - 1) and 95.7% sensitivity (Figure 1B). GAG scores did not correlate with age or gender. GAG scores were elevated in all RCC samples compared to normal controls, irre- spective and uncorrelated to stage, grade or histology. CONCLUSIONS: Plasma GAG scores are measurably elevated in RCC compared to healthy individuals. It was possible to detect RCC irrespective of stage or histology with 95.7% sensitivity. GAG scores did not correlate with pathologic stage, grade, or histology. These findings suggest that GAG alterations occur early in tumor for- mation but are likely independent of progression. These findings war- rant prospective validation to assess the clinical utility of pre-operative GAG scores as biomarkers for RCC. Vol. 199, No. 4S, Supplement, Sunday, May 20, 2018 THE JOURNAL OF UROLOGY â e959