Panel Discussion 2 •What types of epidemiologic studies and exposure information are best suited for determining gene, environment, gene-gene, and gene-environment contributions to health?
•How should environmental monitoring, biological monitoring, and traditional interviews be integrated to optimize characterization of real-world, lifetime exposures and the influence of lifestyle factors (diet, activity, stress)?
•Which biological samples and monitoring frequency are most suitable for characterizing the exposome?
To Pool or Not to Pool: That Is the QuestionEnrique F. Schisterman, PhD
Epidemiology Branch – DESPR- NICHD
Methodological Constraints with Biomarkers of Chemical Exposures
– Which marker to choose in large scale studies and how?
• Cost a big concern!
– How to account for the Instrument sensitivity or LOD?
What is pooling?
• Physically combining several individual specimens, to create a single mixed sample.
• Pooled samples are the average of the individual specimens.
1
2 p
Cost Savings Alternative 1: Random Sample of Biospecimens
RANDOM SAMPLE:
Randomly select 20 samples
FULL DATA
N = 40 Individual Biospecimens
Suppose we have N = 40 individual biospecimens, but we can only afford to test n = 20.
Cost Savings Alternative 2: Pooling Biospecimens
POOLED DATA:
40 samples in groups of 2
FULL DATA
N = 40 Individual Biospecimens
Suppose we have N = 40 individual biospecimens, but we can only afford to test n = 20.
Reporting of biomarker data below a lower threshold
ID Z1 3.12 1.53 8.44 0.85 5.46 3.27 2.08 5.89 13.410 2.511 1.912 6.1
•Reporting threshold is equal to 2.2
ID Z1 3.12 ND3 8.44 ND5 5.46 3.27 ND8 5.89 13.410 2.511 ND12 6.1
Report values < threshold as ‘not detected’
ID Z1 3.12 1.13 8.44 1.15 5.46 3.27 1.18 5.89 13.410 2.511 1.112 6.1
Report values < threshold as one half the value of the threshold
Effect of Pooling on Markers Affected by an LOD
Un-pooled
0
0.2
0.4
0.6
0.8
1
-4 -3 -2 -1 0 1 2 3 4
Un-Pooled
0
0.1
0.2
0.3
0.4
0.5
0.6
0 1 2 3 4 5 6 7 8 9 10
Pooled
0
0.2
0.4
0.6
0.8
1
-4 -3 -2 -1 0 1 2 3 4
Pooled
0
0.1
0.2
0.3
0.4
0.5
0.6
0 1 2 3 4 5 6 7 8 9 10
Summary of Cost Savings: Pooling Biospecimens
• Advantages– Gain efficiency
• Pooling is applied to more participants’ samples while doing fewer assays
• Physical application of Central Limit Theorem– Reduce cost– Volume
• Disadvantages– Pooling process may introduce variability– Requires additional effort and organization– No individual data
Panel Discussion 2
Study designs to evaluate exposure, intermediate endpoint, and genetic
biomarkersNat Rothman, MD, MPH, MHS
Occupational & Environmental Epidemiology BranchDivision of Cancer Epidemiology and Genetics,
NCI, NIH, DHHS
Matrix of study designs and biomarkers
(Rothman, Schulte & Stewart, CEBP, 1995)
Cohort
Case-Control
XXXXXCross-Sectional
Transitional
Laboratory
TumorEarly Disease
Altered Structure/ Function
SusceptibilityEarly Biologic Effects
Internal Dose/BED
ExternalExposure
Matrix of study designs and biomarkers
Cohort
X+/-X+/-XCase-Control
XXXXXCross-Sectional
Transitional
Laboratory
TumorEarly Disease
Altered Structure/ Function
SusceptibilityEarly Biologic Effects
Internal Dose/BED
ExternalExposure
Matrix of study designs and biomarkers
+/-XXXXXXCohort
X+/-X+/-XCase-Control
XXXXXCross-Sectional
Transitional
Laboratory
TumorEarly Disease
Altered Structure/ Function
SusceptibilityEarly Biologic Effects
Internal Dose/BED
ExternalExposure
China Benzene Studies: Integration of cross-sectional
biomarker studies and a cohort study
Cross-sectional studies: 1992/2000
290 exposed workersEvaluate
external/internal exposure, intermediate effects, genetic susceptibility for hematotoxicty
Cohort Study: Follow-up period from
Matrix of study designs and biomarkers
+/-XXXXXXCohort
X+/-X+/-XCase-Control
XXXXXCross-Sectional
Transitional
Laboratory
TumorEarly Disease
Altered Structure/ Function
SusceptibilityEarly Biologic Effects
Internal Dose/BED
ExternalExposure
Panel Discussion 2
Selected biomarkers by specimen, volume, and assay
Factor Biomarker Serum Vol. TechnologyA) Diet folic acid serum 75 µl immuno assay
vitamin D (25-OH) 25 µl immuno-assayselenium 75 µl fluorimetricGPx 50 µl autoanalyzervitamin C serum 100 µl HPLCcarotenoids, vit. E, vit. A 200 µl HPLCGST alpha 20 µl ELISA(phenethyl-)isothiocyanates 0.5 ml LC-MSpolyphenolics 0.5 ml HPLC-ECD (NMR?)fatty acids 200 µl GC-FID
B) Inflammation CRP 20 µl ELISAIL-6 100 µl ELISATNFa 200 µl ELISAIL-1ß, IL-1-RA ELISA
C) Peroxidation cholesterol oxidation products + phytosterols 250 µl GC-MSFRAP serum 25 µl colorimetricMDA 100 µl HPLCF2a-isoprostanes 1 ml ELISA, GC-MS
D) Infection HPV 5 µl antibody assaysChlamydia pneumoniae serum 5 µl antibody assays
E) Smoking cotinin serum 20 µl immuno assay
Study design:
Ideal is nested case-control with repeat samples, plus mixed designs such as ad hoc collections/resampling in subsets of the same population to calibrate measurements (EnviroGenoMarkers)
Ideally urines and blood should be collected
Improve exposure assessment partly with GIS
Do more pooled analyses of existing datasets on biomarkers (e.g.DNA adducts within ECNIS)
Create a repository of study results and validation studies (MEC in ECNIS)
Panel Discussion 2
Potential methodologies for measuring predictive power of
exposome• Population-based studies with nested
design– Naturalistic “Burst design” (Nesselroade, 1991)
• Repeated daily assessments within prospective longitudinal design (Tier 2—exp & response)
– Standardized lab stressor assessments (Tier 3)
• Women of childbearing age, transmission of risk