Abstract 16533: Non-Invasive Insulin Resistance Evaluation Using Carotid Pressure Waveforms in Framingham Heart Study

Abstract

Introduction: Insulin resistance (IR) and the resultant hyperinsulinemia not only lead to the onset of diabetes but also are shown to elevate the risk of heart failure and underlie a spectrum of cardio-metabolic pathologies in both diabetic and non-diabetes patients. Measuring IR is challenging owing to reliance on an invasive and time-consuming procedure called a glucose clamp study, which is known as the gold standard method of assessing IR. However, this method is not routinely available in most clinical settings. The Homeostasis Model Assessment of IR (HOMA-IR) is a validated tool for evaluating IR calculated based on fasting glucose and insulin concentrations obtained from fasting blood samples. Determining optimal cut-off values for HOMA-IR has been challenging since its distribution varies across different demographic characteristics and populations. Aim: This study aims to facilitate the screening of HOMA-IR by introducing a new method that utilizes a non-invasive carotid pressure waveform to assess HOMA-IR. Methods: We extracted Intrinsic Frequency Analysis-derived features from carotid pressure waveform to detect HOMA-IR using machine learning. A non-diabetic cohort from Framingham Heart Study (FHS) (n=5427, 53% female, age range 19-90) was used to develop and test the proposed method. The method was developed on 80% of the data and blindly tested on 20% . Results: Using thresholds of 0.5 and 3 for the low (carefree) and high (alarming) zones of HOMA-IR in FHS dataset, respectively, the performance of detecting high HOMA achieved an area under curve (AUC) of 0.82 on the blind test set data (n=1085) (Fig.1; ROC: receiver operating characteristic curve). Conclusion: The non-invasive nature of the proposed method facilitates the screening of metabolic disease risks based on HOMA-IR level inferences from noninvasive carotid pressure waveforms, which can be implemented in routine clinical practice using inexpensive standard hand-held tonometry devices.