Previous genetic epidemiology studies of complex diseases have generally searched for important genetic variants through univariate association studies and therefore have not provided adequate insight into the integrated effects of genotypic and phenotypic determinants of the complex disease state. The purpose of this research was to take an integrative, multivariate approach to exploring phenotypic and genotypic factors related to risk for disease. We have researched metabolic syndrome, a highly prevalent and complex disorder indicative of increased risk for future cardiovascular disease in a sample of 1325 individuals from the relatively isolated island population of Hvar, Croatia. This population reports rates of metabolic syndrome, obesity, and hypertension similar to those of the United States, in spite of a healthier, more active lifestyle. The motivation behind this research was threefold: first, to improve our understanding of the metabolic syndrome trait network; second, to identify genetic variants associated with broad component traits of metabolic syndrome; and finally, to integrate the effects of single-trait genome-wide association variants into the metabolic syndrome network.
To identify genetic variants associated with broad component traits of metabolic syndrome, we performed genome-wide association studies (GWAS) of eight principal components built from 18 biochemical and anthropometric measures. Through this analysis, we did not uncover any signals that reached GWAS-level significance, did not detect many well-documented genetic associations, and did not uncover genes in pathways with known connections to metabolic traits. We surmise that during the formation of principal components, the variation extracted from the original traits may not have been meaningful in relation to genetic determinants. In addition, the precision of original traits may become somewhat lost in the components, reducing signal strength of variants to GWAS insignificance.
We also implemented structural equation modeling to improve our ability to accurately describe the direction and integration of significant variants uncovered through univariate GWAS, individual metabolic parameters involved metabolic syndrome and related metabolic disorders. Through this analysis we determined that obesity, dyslipidemia, hypertension, and hyperuricemia were tightly integrated in the metabolic syndrome network and were useful in the prediction of metabolic diseases, including coronary heart disease, type-2 diabetes, stroke, and gout. Glucose control, on the other hand, was relatively unconnected to the metabolic syndrome network and was associated only with obesity and type-2 diabetes. In general, genetic variants included in the structural equation model were associated only with their original GWAS trait, although several variants showed significant impact beyond their original trait, underscoring the importance of assessing the impact of genetic variants on an integrated complex system. Finally, metabolic syndrome has previously been considered an indicator of increased risk for metabolic disease; in our study, however, metabolic syndrome had little association with downstream metabolic diseases, indicating that its utility as a disease predictor may be limited.