1. Toward capturing the exposome variability: Household co-exposure patterns in the LIFE Study
Jake Chung
Feb 2017

2. Questions
1. Is it possible to characterize the density of correlations of environmental exposures?
2. How are the correlations of exposures driven by shared environment (e.g., within household) vs. the individual?
Aims
1. Explore the co-exposure patterns of chemical biomarkers in couples
2. Examine the influence of familial relatedness in correlation of exposures

3. Main findings Comparing females and males:
1. Highly similar co-exposure patterns between the male and female groups
2. Higher within-class Spearman correlations for persistent organic chemicals (PCBs, PBDEs, organochlorine pesticides) than other classes (abs. median rs: vs )
Comparing couples to above:
1. Abs. median rs of all chemical classes are lower
2. Most significant drop for persistent organics with abs. median rs in range of
3. In addition, interquartile ranges of correlation of all chemical classes are lower

4. Design 501 pairs of couples 128 serum and urinary biomarkers
In 13 different chemical classes
Persistent organics, non-persistent organics, and metals etc

5. Biomarkers measured in this study

6. Method 1. Load multiply imputed chemical data to R (provided by NICHD)
2. Obtain residuals from regressing chemicals against covariates (serum total lipids or urine creatinine)
Linear model: scale(log10(biomarker+1)) ~ scale(covariate)
Adjusted for total lipids: PBDEs, PCBs, OCPs
Adjusted for creatinine: phytoestrogen, phthalates, phenols, parabens, paracetamols, metals, metalloid
Unadjusted: PFASs, cotinine, blood metals
3. Calculate rs matrices with the residuals for males, females and couples
Get final rs estimates by averaging across imputed frames (Rubin’s method)
Get final p values from permutation tests

7. Two generally positively correlated clusters
1) PCBs, OCPs, and PFASs
2) phytoestrogens, phthalates, phenols, anti-microbials, paracetamols, urine metals and metalloids
Left: females; Right: males. Interactive plots:

8. Similar co-exposure patterns between females and males
(red line represents perfect agreement)

9. Similar density curves between females and males

10. Weaker co-exposure patterns in couples!
Females
Males
Interactive plots:

11. Female chemicals are weak predictors to the chemicals in male

12. Persistent organic chemicals have higher within-class correlations
Left: females; Right: males. Higher resolution plots:

13. Lower within-class correlations in couples
Females
Males

14. Interpretation Co-exposure patterns of biomarkers were found in:
Both between and within chemical classes
With stronger within-class correlation for persistent organic chemicals (PCBs, PBDEs, OCPs)
Analytical challenges for exposome-wide association study: p value correction, multicollinearity
Damping of correlations in couple group
Chemical correlations were weaker in the couples than females and males, especially for persistent organics
Non-home environment could be a significant source of exposure variability?
Conditions to use indoor dust as surrogate of exposure assessment?

15. Higher PBDEs correlations within family members
Young children (2-8 years old; N = 67)
Parents of young children (<55 years old; N = 90),
Older adults (≥55 years old; N = 59)