Abstract A18: Metabolomic analysis of ovarian cancer risk in the Nurses’ Health Studies: Metabolite associations are more pronounced in non-serous tumors

Abstract

Abstract Background: To identify potential novel risk factors for ovarian cancer, we assessed the role of metabolites and lipids in the development of ovarian cancer. Prior research suggests that lipid synthesis and metabolism are dysregulated in ovarian tumors. Furthermore, studies show that lysophosphatidylcholines (LPC), PCs, and sphingomyelins (SM) are differentially expressed in ovarian cancer cases versus controls. We used a validated metabolomics platform to assess prediagnosis measures of these lipid classes with ovarian cancer risk and conduct a discovery-based analysis of other metabolites. Methods: LC-MS/MS targeted metabolomics were measured at the Broad Institute (Cambridge, MA, USA). 300 cases were matched to one control on age, menopause status/hormone therapy use, fasting, season, and time of day of blood draw. 348 metabolites and lipids passed QC. Metabolites were transformed using probit scores for normality. We used conditional logistic regression to identify metabolites and lipids associated with total invasive ovarian cancer risk. We used unconditional logistic regression, adjusting for matching factors, in histotype-specific analyses including all available controls, thus increasing power. All analyses adjusted for known risk factors: oral contraceptive use, parity, and tubal ligation. Results: We found 18 metabolites that were associated with risk of invasive ovarian cancer (p≤0.05). These included sphingosine, C3-DC-CH3 carnitine, choline, cytosine, triacylglycerols (TAG), C18:1 SM, lysophosphatidylinositol (LPI), mono hydroxyoctadecadienoic acid, phosphatidylethanolamine (PE), and taurine. The odds ratios ranged between 0.73 and 0.84 per 1 standard deviation (SD) increase in metabolite value. None of the metabolites remained significant after multiple testing correction (false discovery rate (FDR) corrected p≥0.85). In histotype analysis, 13 metabolites were associated with serous disease (p≤0.05). These included C3-DC-CH3 carnitine, 7 TAGs, choline, symmetric dimethylarginine (SDMA), citrulline, cytosine, and 6-deoxodolichosterone. Odds ratios ranged between 0.75 and 0.82 for 1 SD increase in the metabolite values. No metabolites remained significant after multiple testing correction (FDR-corrected p≥0.75). Interestingly, 40 metabolites were associated with non-serous ovarian cancer at p≤0.05. Nearly all were different than those found for total invasive or serous disease. These included pregnanediol-3-glucuronide, laserpitin, trihydroxyecdysone, docosahexaenoic acid, decosapentaenoic acid, niacinamide, testosterone sulfate, alanine, taurine, eicosapentaenoic acid, thiamine, 1-methyladenosine, cajaisoflavone, kurilensoside H, and multiple lipid classes (cholesterol ester (CE), PCs, PE and PC plasmalogens, TAGs, carnitines, SMs, DAG, LPI, lysophosphatidic acid (LPA), lysophosphatidylethanolamine (LPE), PE, TAGs, and diacylglycerols (DAGs)). Odds ratios for protective metabolites ranged between 0.49 and 0.74 while for harmful metabolites they ranged between 1.38 and 1.48 for 1 SD increase in the metabolite values. However, FDR-corrected p-values were ≥0.41. Conclusion: While several metabolites differed between ovarian cancer cases and controls on a nominal p-value scale, none remained significant after multiple testing correction. However, this test does not account for the high correlation between many of the metabolites. Notably, distinct markers were associated with different histotypes and more metabolites were associated with non-serous tumors, which have been associated more strongly with BMI and other metabolic risk factors. For nonserous disease, we observed associations with LPCs, PCs, and SM lipids, which were part of our a priori hypotheses, as well as sex hormone metabolites. Ongoing analyses include adjusting for BMI at age 18 and adolescent somatotype to account for early life adiposity, consideration of other multiple testing correction methods, and pathway analyses. Citation Format: Oana A. Zeleznik, Elizabeth M. Poole, Clary Clish, Heather A. Eliassen, Peter Kraft, Shelley S. Tworoger. Metabolomic analysis of ovarian cancer risk in the Nurses’ Health Studies: Metabolite associations are more pronounced in non-serous tumors. [abstract]. In: Proceedings of the AACR Conference: Addressing Critical Questions in Ovarian Cancer Research and Treatment; Oct 1-4, 2017; Pittsburgh, PA. Philadelphia (PA): AACR; Clin Cancer Res 2018;24(15_Suppl):Abstract nr A18.