Introduction:
To date, there are limited data on the potential role of proteomic biomarkers to predict future cardiovascular (CV) events among patients with type 2 diabetes mellitus (DM).
Hypothesis:
Specific protein biomarkers will be predictive of major adverse CV events (MACE) and incident heart failure hospitalization (HFH) among patients with DM.
Methods:
Using the Olink aptamer-based platform, we performed proteomic profiling (>700 proteins) on 440 paired cases and matched controls from placebo-assigned participants in the Trial Evaluating Cardiovascular Outcomes with Sitagliptin (TECOS). Cases were defined as having met the primary composite outcome of MACE or incident HFH and matched to controls on baseline prevalent heart failure, coronary artery disease, BMI, hemoglobin A1C, creatinine, low-density lipoprotein cholesterol, fasting status and ejection fraction. Conditional logistic regression was used to determine the association between log-transformed relative protein expression and incident MACE or HFH. False-discovery-rate (FDR) was used to adjust for multiple comparisons.
Results:
We identified three specific proteins that were significantly associated with prevalent MACE or HFH: METRNL, Notch 3, and ROR1 (OR 2.1, 1.6, 1.7 and q-value 0.01, 0.02, and 0.05 respectively) (Figure 1). METRNL, in particular, performed similarly to the established biomarker NT-proBNP (Figure 1). When MACE and HFH were analyzed separately, METRNL, in particular, remained strongly associated with both outcomes (OR 2.0, p<0.001 and OR 2.7, p=0.05).
Conclusions:
Three novel protein biomarkers, in particular METRNL (a circulating adipokine that regulates insulin-sensitivity), may identify diabetic patients at high risk for subsequent HF and MACE. Additional studies are needed to replicate these findings and uncover the biologic mechanism linking adipokine signaling and heart failure.
1485P Identification of novel protein biomarkers for FOLFIRINOX-based chemotherapy response in advanced pancreatic adenocarcinoma using patient omics and Bayesian AI
