Modelos de mistura na análise da síndrome metabólica em mulheres: padrões latentes, fatores associados e relação com o diabetes mellitus tipo 2.

Abstract

Metabolic Syndrome (MS) is a complex disorder characterized by the co-occurrence of metabolic abnormalities, which has outstanding importance from an epidemiological point of view due to its high prevalence and the evidence of its association with cardiovascular diseases (CVD) and type 2 diabetes mellitus (DM2). The most recent criterion proposed for MS diagnosis establishes the presence of at least three of the following components: abdominal obesity, hyperglycemia, hypertriglyceridemia, low levels of HDL cholesterol and high systemic blood pressure. Sex-gender differences in the expression of MS have been reported in the literature. In several studies, there is evidence that women, especially after the age of 50, are more susceptible to developing the syndrome and its associated comorbidities. The most common approach in quantitative research with MS establishes the identification of cases considering a dichotomy (presence/absence). Alternatively, MS can be considered as a heterogeneous entity, with variation in grouping of its components. Thus, an interesting approach seeks to characterize MS as a construct (unobserved variable), identifying, empirically, the patterns of its occurrence based on the analysis of the response profiles of its components. This approach uses multivariate statistical methods, particularly the Mixture Models, which have recently been used in MS research. These are flexible models that allow discriminating non-observed subgroups of individuals throughout the population and evaluating the relationship between belonging to these subgroups and other variables in both cross-sectional and longitudinal studies. The present study explores the use of Mixture Models to expand the knowledge about the MS occurrence among Brazilian women, identifying distinct latent patterns of this disorder in participants of the first two waves of the Longitudinal Study of Adult Health (ELSA-Brasil). For the analyses, different LCMs were used, namely: Latent Class Analysis, Latent Class Analysis with Repeated Measures, Latent Transition Analysis and Latent Class Analysis with Distal Outcomes. At the baseline of ELSA-Brasil, MS patterns were discriminated, and the effects of sociodemographic characteristics and menopause status were examined. Using longitudinal latent approaches, shifts and transition profiles for MS were evaluated using data from the two first waves of the study. The risk for the development of type 2 diabetes mellitus was estimated according to different patterns of MS. Initially, three distinct patterns of MS were identified for the women in the study. Longitudinal analysis mainly revealed permanence of MS patterns. Individual characteristics, such as elementary or middle schooling level, and insufficient leisure-time physical activity, increased the chance to non-changing patterns over time, especially for those with greater metabolic expression. After 4 years of follow-up, the transition from lower to higher health risk patterns was more likely than the reverse transition. The risk of developing DM2 differs according to MS patterns, confirming the relevance of considering the heterogeneous expression of the syndrome. Particularly, we found high prevalence and risk for developing DM2 among the women characterized by the joint occurrence of abdominal obesity, hyperglycemia and high blood pressure. Additionally, the analyzes suggest that abdominal obesity is the main initial abnormality in triggering the syndrome in this population. These results point to the relevance of considering the heterogeneity in evaluating MS expression, its relationships with risk factors and associated diseases. Furthermore, it contributes for drawing the attention for the potential use of mixture models as analytical strategy when studying MS and other heterogeneous health outcomes.