Validating Metabolic Syndrome Through Principal Component Analysis in a Medically Diverse,Realistic Cohort

Abstract

Background:

The concept of metabolic syndrome has been subject to etiological and clinical controversies in recent years. Associations among the five risk factors (obesity, hypertension, hyperglycemia, high triglyceride levels, and low high-density lipoprotein cholesterol) may help establish the validity of the concept, especially in a cohort representative of an actual population.

Methods:

We used principal component analysis (PCA) to analyze the structure of the physiological components of metabolic syndrome in 7213 patients contained in an administrative database for the Centre Hospitalier Universitaire de Sherbrooke in Sherbrooke, Quebec, a realistic cohort with diverse medical histories. We validated the results by repeating the analysis on stratified and random subgroups of patients, and on different combinations of risk factors. The first axis of the PCA was used to predict coronary heart disease (CHD) and diabetes.

Results:

The two first axes explained 53% of the variance. The first axis (33%) was associated in the expected direction with all five predictor variables, consistent with its interpretation as metabolic syndrome. The first axis was more predictive of subsequent CHD and diabetes than the formal definition of metabolic syndrome.

Conclusions:

These results suggest that the concept of metabolic syndrome accurately captures an existing underlying physiological process. A continuous indicator could be constructed to identify metabolic syndrome more accurately, thus improving risk assessment for CHD and diabetes mellitus. Metabolic syndrome can be measured well even without all five predictors. However, discrepancies with other studies suggest that our results may not be generalizable, perhaps because our cohort tends to be sicker.