Probing the factor structure of metabolic syndrome in Sardinian genetic isolates

Owing to the multiplicity of the key components of metabolic syndrome (MetS), its diagnosis is very complex. The lack of a unique definition is responsible for the prevalence variability observed among studies; therefore, a definition based on continuous variables was recommended. The aim of this study was to compare competing models of the MetS factor structure for selecting the one that explains the best clustering pattern and to propose an algorithm for computing MetS as a continuous variable.

Data were from isolated Sardinian populations (n = 8102). Confirmatory factor analysis (CFA) and two-group CFA by gender were performed to evaluate the sex-specific factor structure of MetS. After selecting the best model, an algorithm was obtained using factor loadings/residual variances. The quality of the MetS score was evaluated by the receiver operating characteristics curve and the area under the curve. Cross-validation was performed to validate the score and to determine the best cut point. The best fit model was a bifactor one with a general factor (MetS) and three specific factors (f1: obesity/adiposity trait; f2: hypertension/blood pressure trait; and f3: lipid trait). Gender-specific algorithms were implemented to obtain MetS scores showing a good diagnostic performance (0.80 specificity and 0.80 sensitivity for the cut point). Furthermore, cross-validation confirmed these results.

These analyses suggested that the bifactor model was the most representative one. In addition, they provided a score and a cut point that are both clinically accessible and interpretable measures for MetS diagnosis and likely useful for evaluating the association with adverse cardiovascular disease and diabetes and for investigating the MetS genetic component.