Cluster Analyses of Biomarkers Differentiate Metabolic Risk in Hispanic Adolescents of South Texas

Background: High rates of obesity impact Hispanic adolescents in S. Texas. This study examined the use of data mining methods, particularly clustering algorithms, to identify inflammatory biomarker phenotypes and their association with obesity and metabolic risk in a Hispanic adolescent population. Methods: Measures of obesity, insulin resistance and plasma levels of inflammation biomarkers in 167 subjects were utilized. Unsupervised К-means, Hierarchical cluster, and PCA analyses were performed on log-transformed and standardized biomarkers to identify groups of adolescents having similar biomarker patterns. All statistical analyses were performed using R and SAS. Results: 3 clusters were identified. Cluster 3 was characterized by the highest CRP and Leptin levels; Cluster 1 had the highest IL-8, TNF-a, MCP-1, and HGF levels; and Cluster 2 had the lowest levels of all inflammatory markers. Individuals in cluster 3 had significantly higher mean BMI levels (31.7 ± 7.37), compared to cluster 1 (23.7 ± 3.92) and cluster 2 (23.2 ± 4.15). Based on their clinical characteristics (BMI, WC, WHR, HOMA-IR, insulin, lipids, and liver enzymes), the clusters were labeled as severe risk for metabolic dysfunction (Cluster 3), moderate risk for metabolic dysfunction (Cluster 1), and normal metabolic function (Cluster 2). Statistically significant negative correlations between MCP-1, leptin levels, BMI and WC were observed. The severe metabolic risk cluster had levels of IL-8, TNF-α, MCP-1, and HGF similar or lower than the normal metabolic risk group. Conclusions: Adolescents with higher BMI and waist circumference as well as elevated lipids, liver enzymes, insulin and HOMAIR had higher odds of being classified in the severe metabolic risk cluster, characterized by elevated biomarkers Leptin and CRP. This group had dampened levels of cytokines and chemokines suggesting a compromised immune response. The moderately at-risk group was distinct from the normal group in having slightly elevated Leptin and CRP levels, as well as the highest levels of cytokines and chemokines suggesting that their inflammatory response was still functional. We conclude that levels of various obesity-related inflammatory biomarkers should be evaluated carefully to discriminate between severe and moderate risk for metabolic and immune dysfunction in adolescent Hispanic populations.