Triglyceride-glucose index demonstrates reasonable performance as a screening tool but exhibits limited diagnostic utility for insulin resistance: Data from the ELSA-Brasil cohort

Abstract

Insulin resistance (IR) is a key risk factor for chronic metabolic diseases, but its laboratory diagnosis is still costly; thus, the triglyceride-glucose (TyG) index has been proposed as a surrogate method. Our aim was to provide a detailed analysis of cutoffs and test the hypothesis that the TyG index would present reasonable performance parameters for IR screening. This is a cross-sectional study with baseline data from 12,367 eligible participants of both sexes (aged 35–74 years) from the Brazilian Longitudinal Study of Adult Health. TyG correlation and agreement with the Homeostasis Model Assessment for Insulin Resistance were analyzed. Positive and negative predictive values (PV+, PV–) and likelihood ratio (LR+, LR–) were calculated. A moderate positive correlation between TyG and Homeostasis Model Assessment for Insulin Resistance was observed (Pearson r = 0.419). The area under the receiver operating characteristic curve of TyG for IR diagnosis was 0.742 and the optimal cutoff was 4.665, reaching a kappa agreement value of 0.354. For this cutoff, a PV+ of 59.3% and PV– of 76.0%, as well as an LR+ of 2.07 and LR– of 0.45 were obtained. Alternatively, because high sensitivity is desired for screening tests, selecting a lower cutoff, such as 4.505, increases the PV– to 82.1%, despite decreasing the PV+ to 50.8%. We conclude that TyG has important performance limitations for detecting IR, but that it may still be reasonably useful to help screening for IR in adults because it can be calculated from low-cost routine blood tests.