central roles of obesity-associated dyslipidaemia, endothelial activation and cytokines in the Metabolic Syndrome--an analysis by structural equation modelling

Hypothesis: The multi-faceted components of the metabolic syndrome now include markers of inflammation and endothelial activation. Despite this growing body of epidemiological data, standard statistical methods fail to evaluate the nature of these associations adequately. In this pilot study, we hypothesize that obesity may lead to endothelial activation which is in part mediated by dyslipidaemia and proinflammatory cytokines. These factors interact to give rise to hyperinsulinaemia, hypertension and an anti-fibrinolytic state. To test this hypothesis, we used confirmatory factor analysis and structural equation modelling to fit these data to a model designed on theoretical grounds. Methods: Metabolic syndrome variables, cytokines (IL6 and TNFα), markers of inflammation and endothelial activation were measured in 107 Caucasian non-diabetic subjects aged 40-75 y. Using confirmatory factor analysis, we identified six factors to represent composite measurements of blood pressure, obesity, dyslipidaemia, hyperinsulinaemia, endothelial activation and the anti-fibrinolytic state. We fitted these variables to two separate models, one using IL-6 and the other TNFα as the cytokine, and examined the inter-relationships (path analysis) amongst these variables, based on the above hypothesis. Results: Men were centrally more obese and had increased markers of endothelial activation, inflammation and the anti-fibrinolytic state as well as hyperinsulinaemia and dyslipidaemia, compared with women. Obesity indexes (both body mass index and waist-hip ratio) were strongly associated with multiple cardiovascular risk factors. Both IL6 and TNFα were correlated with age, male gender, obesity indexes and markers of endothelial activation. Only IL-6 was associated with smoking while TNFα was correlated with hyperinsulinaemia. In the TNFα model, 61% of the obesity variance was explained by male gender, 36% of TNFα variance by age and dyslipidaemia, 43% of dyslipidaemia variance by age and obesity, 33% of hyperinsulinaemia variance by dyslipidaemia and a non-smoking state, 29% of anti-fibrinolytic state variance by hyperinsulinaemia, 65% of endothelial activation variance by TNFα, dyslipidaemia and hyperinsulinaemia, 34% of blood pressure variance by hyperinsulinaemia and endothelial activation. In the IL-6 model, we observed similar relationships except that 23% of IL6 variance was explained by smoking and age. Conclusions: Using confirmatory factor analysis and structural equatio