Working Package 13: Exposure and health association studies

The data necessary for comprehensive  study of the associations  between environmental exposures and health perturbations including early biological events  that can be associated to disease phenotypes as observed in the populations studied in stream 5 will be collected and clustered  after preliminary statistical analysis to remove any potential bias in either the exposure  data collection or the effect observations. The datasets will be controlled for completeness  and robustness, and will be clustered in commensurate  groups to facilitate their statistical  treatment (to be done in Task 13.1). This task is critical due to the very large number of data of different attributes that comprise  the input to the association  models  developed  in Task 13.2.

Internal doses will be coupled to health impacts  on the local population through advanced  statistical methods to derive the dose–response  functions which account for differences in exposure patterns, susceptibility differences and inter-individual variation in health response. The approach  starts from the biomarker values measured in different biological matrices  (urine and peripheral blood) to estimate  through the application of the lifetime generic  PBBK  model the biological effective dose in the target tissue, which is consistent  with the biomarker level measured. To estimate the health impact we will use a multi-faceted statistical approach  based on survey-weighted  logistic multivariate regression  adjusted for different covariates  (age, sex, socio-economic status (SES), smoking etc.) linking internal doses with health effects  or intermediate biological events that can be associated to health perturbations  through pathway  analysis considering the interdependence of the covariates (using as metric an analogy of the “linkage disequilibrium” metric used in genome-wide association studies).  Multilevel models  will be applied to take into account  the European dimension. Logic regression, an adaptive classification and regression  procedure,  will be also used in any setting with binary predictors, when the interaction of these covariates is of primary interest. In addition, causal diagrams (directed acyclic graphs – DAG) and Bayesian inference-based  modelling using state-of-the-art computational tools available to the HEALS team  will be used to conceptualize  confounding and identify the minimal sufficient adjustment set out of the large datasets  derived from the HBM and exposure analyses. This will help us identify early effect biomarkers  for improved intervention and prevention. The work will include hypothesis  testing in relation of observed contaminations in nested case-control  approaches  for selected health endpoints. The health impact estimates  will be then confronted with the observed  adverse health effects in the local population refining thus the epidemiological  observations  currently available.  Particular  attention will be given to effects  on susceptible population groups such as children (including neonates)  and the elderly.