Construction of A Metabolic Syndrome Scale and Joint Phenotype Heritability Using Item Response Theory Models

Abstract

The metabolic syndrome is composed of a series of phenotypes that jointly contribute for cardiovascular disease and diabetes. Studies of its heritability are conducted analyzing it individually for each symptom, sometimes using dichotomized variables corresponding to defining criteria adopted by some health authority. In this work, we construct a metabolic syndrome scale using Item Response Theory (IRT) models for the dichotomized phenotypes that takes all measured phenotypes into consideration, allowing informative comparison between individuals. A welcome consequence of the IRT framework is that we derive a straightforward joint heritability parameter that can be easily interpretable as the heritability of the latent trait representing the individual metabolic syndrome level. Parameter estimation was conducted under a Bayesian statistics framework, using Markov Chain Monte Carlo (MCMC) type algorithms implemented as a script in the R Statistical Software. Simulation studies were conducted to evaluate recovery of the phenotype location and metabolic syndrome heritability parameters, indicating adequate recovery. The metabolic syndrome is first validated in comparison with a Factor Analysis, obtaining a high correlation between both constructs and then applied to a sample of 1.666 individuals obtained in the Baependi Heart Study described in Oliveira et al. (2008, BMC Medical Genetics 32 (9))