![IconGems-Print](images/IconGems-Print.png)
474 – Infectious Disease, Environmental Epidemiology, and Diet
Application of Principal Components Analysis to Urine Metal and Metalloid Exposures in the National Health and Nutrition Examination Survey (NHANES) Data
Po-Yung Cheng
Centers for Disease Control and Prevention
Robert L. Jones
Centers for Disease Control and Prevention
Kathleen L. Caldwell
Centers for Disease Control and Prevention
Human exposure to metals is an important public health issue because even low levels of exposure are associated with adverse health effects. NHANES measures metal exposures in the U.S. population; however, few researchers have examined metal co-exposures and the effects of potential interaction. To explore potential relationships between metal co-exposures, we applied principal components analysis (PCA) to 15 urine metals and metalloids in the NHANES 2013-14 national survey data. Through PCA, principal components (PC) are created from the original variables in an attempt to summarize complex data. Each PC has an eigenvalue, which indicates the variance for each PC in the data. A higher eigenvalue represents a greater variance. The first eigenvalue obtained from this analysis explained 47% of variation; the second explained 8%; the third explained 7%. The first principal component (PC1) was strongly correlated with all elements (correlation efficient (r) range: 0.56 – 0.84) except manganese. In contrast, PC2 was correlated with manganese (0.55), arsenic (-0.55) and inorganic-related arsenic species (-0.47). PC3 was correlated with barium (-0.62) and strontium (-0.48). Certain demographic characteristics, including lower income level, Asian ethnicity, female sex, and elderly, were associated with higher scores (90th percentile and above) for PC1. In contrast, non-Hispanic whites with lower income level were associated with higher scores for PC2 and elderly smokers with ethnicities of non-Hispanic blacks, non-Hispanic Asians, and other Hispanics were associated with higher scores for PC3.