Abstract P194: Lipid Metabolic Pathways and Cardiovascular Disease Risk in the Prevención con Dieta Mediterránea (PREDIMED) Trial

Background: Metabolomics technologies can efficiently profile a large number of structurally diverse lipids, e.g., glycerophospholipids, sphingolipids, and glycerolipids, that may play differential roles in pathogenesis of cardiovascular disease (CVD). However, existing studies were focused more on discovering individual lipid metabolites for CVD risk prediction than inferring perturbed pathways responsible for the pathological processes. Hypothesis: We hypothesized that different lipid metabolic pathways, captured by network analysis, were divergently associated with CVD risk; the associations could be modified by the Mediterranean diet (MedDiet) intervention. Methods: We conducted this study in the PREDIMED trial with participants randomized to three intervention diets: MedDiet with extra-virgin olive oil, MedDiet with nuts, or a low-fat control diet. This study comprises a subcohort of 788 participants randomly selected from the PREDIMED cohort and 230 cases. The outcome was a composite endpoint of non-fatal acute myocardial infarction, non-fatal stroke and cardiovascular death. We performed network analysis using Gaussian graphic model among 200 targeted lipid metabolites and subsequent dimensionality reduction using Greedy Modularity Optimization to detect subnetworks. We calculated the subnetwork scores by summing up the products of the topological connectivity weight (representing network structure) and metabolite level and included the scores into Cox proportional hazards model with simultaneous adjustment for other subnetwork scores and covariates. The Benjamini-Hochberg procedure was applied to detect smaller subnetworks of specific interest for further investigation. Results: We detected 4 major subnetworks of lipid metabolic pathway. Most lipid metabolites with larger numbers of carbon atom and double bond clustered within a same subnetwork (subnetwork 2), while those with smaller numbers of carbon atom and double bond clustered within other 3 subnetworks (subnetworks 1, 3 and 4). The hazard ratio (HR) of CVD across quartiles of the subnetwork score 2 was 0.56 (95% CI, 0.35, 0.91, P trend =0.008) after multivariable adjustment. The HRs of CVD comparing extreme quartiles of subnetwork scores 3 and 4 were 1.85 (95% CI, 1.15, 2.97, P trend =0.02) and 1.93 (95% CI, 1.19, 3.14, P trend =0.01), respectively. The MedDiet appeared to enhance the inverse association between subnetwork score 2 and CVD ( P interaction =0.03). We detected several smaller subnetworks with functional interpretation related to CVD pathogenesis such as the ceramide pathway and the pathway including phospholipids with high unsaturation. Conclusions: Based on topological structure of lipid metabolic pathways, we detected biologically meaningful pathways. We found divergent associations between subnetworks and CVD and smaller subnetworks with functional interpretation.