Abstract 16083: Urinary Proteome and its Application to Predict Cardiovascular Risk in Patients With Stable Coronary Heart Disease
Introduction: Factors that lead to atherosclerosis can cause systemic effects, including impaired kidney function, potentially creating distinct urine protein signatures prognostic of cardiovascular (CV) risk. Urine as a medium for CV biomarkers eliminates the need for venipuncture and blood processing. To date, information is lacking about the nature of the urinary proteome and whether it is predictive of CV outcomes in patients with coronary heart disease (CHD).
Hypothesis: Assessment of the urinary proteome in patients with CHD will uncover a subset of proteins which are prognostic of CV outcomes.
Methods: We measured 4316 proteins by a modified-aptamer assay in 24-h urine collections from 818 participants in the Heart and Soul study with stable CHD, with up to 11.1 years follow-up and kidney function that ranged from normal to moderately impaired. A protein-based normalization process that corrects for physiologic variation in urine concentration was developed from urine dilution experiments. A multivariable CV risk model was generated by LASSO, followed by backwards-selection to create a parsimonious model.
Results: Of the 4316 proteins assayed in urine, 2315 had acceptable dilution linearity. Of these, 608 proteins were prognostic of a composite CV outcome of myocardial infarction, stroke, heart failure and death in univariate Cox proportional hazards models (p < 0.05, Bonferroni corrected). Preliminary multivariate models were generated: a 7 protein model for 4-year CV outcome had a C-statistic [95% CI] of 0.69 [0.66, 0.72] compared with 0.62 [0.60, 0.65] for urine creatinine:microalbumin, 0.61 [0.58,0.63] for the Framingham secondary risk model, and was close to the C-statistic of 0.73 [0.71, 0.76] for a 9 protein model in blood plasma from the same individuals.
Conclusions: In patients with stable CHD and no severe CKD, assay of 4316 proteins showed that the urine proteome consists of at least 2315 proteins, of which 608 are prognostic of CV risk. A CV risk prediction model based on urinary proteins performs nearly as well as a model based on plasma. Further research will determine the generalizability of these findings to other populations.
Author Disclosures: T. Hraha: Employment; Significant; SomaLogic. K. DeLisle: Employment; Significant; SomaLogic. R. Ostroff: Employment; Significant; SomaLogic. S. Williams: Employment; Significant; SomaLogic. P. Ganz: None.
- © 2016 by American Heart Association, Inc.