Abstract for the 2019 Annual Conference
Health effects of air pollution components, noise and socio-economic status (“HERMES”)
Ole Raaschou-Nielsen1,3, Theis Lange2, Matthias Ketzel3,5, Ulla Hvidtfeldt1, Henrik Brønnum-Hansen2, Thomas Münzel4, Lise M. Frohn3, Jesper Christensen3, Ulas Im3, Ole Hertel3, Jørgen Brandt3, Mette Sørensen1,6
1Danish Cancer Society Research Center, Copenhagen, Denmark; 2University of Copenhagen, Copenhagen, Denmark; 3Aarhus University, Roskilde, Denmark; 4Johannes Gutenberg University, Mainz, Germany; 5University of Surrey, Guildford, United Kingdom, 6Roskilde University, Roskilde, Denmark
Background. Traffic-related air pollution (TRAP), traffic noise and low socio-economic status (SES) impair health, including cardiovascular disease and diabetes. However, knowledge gaps still remain including identification of the causal agent(s) in the complex TRAP, the degree of confounding or possible interaction between TRAP and traffic noise, and how socio-economic status (SES) and individual susceptibility interplay in this equation.
Objectives. The objectives of the study are (1) to identify the specific TRAPs strongest associated with myocardial infarction (MI), stroke and diabetes; (2) to disentangle how TRAP and road traffic noise interact in relation to these endpoints; (3) to investigate how SES, green spaces, co-morbidity and stress confound or interact with the associations between TRAP and road traffic noise and risk of MI, stroke and diabetes; and (4) to investigate effects of TRAP and road traffic noise in relation to cardiovascular and metabolic biomarkers.
Experimental design and results. At present (month 8 of the study) we focus on data collection and method development.
Data collection. We use data for the entire Danish population, the Diet, Cancer & Health cohort, and the DCH Next Generations cohort. We will link each individual to the nationwide Danish registries with information on residential address history, prevalent and incident MI, stroke and diabetes, vital status, indicators of stress and SES. We will present an overview of the registries and data to be combined in the HERMES database, and we will provide a status of the data collection process.
Modelling ultrafine particles (UFP). We have implemented the M7 UFP module into the Danish AirGIS dispersion modelling system. We will present the model set-up.
Development of a new statistical method to disentangle effect of correlated pollutants. This approach is based on a first step where machine learning techniques are used to capture all links between pollutants and outcomes without imposing linearity assumptions or similar. In a second step, the prediction model from the first step is used to predict events under specific pollutants scenarios. From a technical perspective, this can be viewed as using machine learning methods as the so-called Q-model when applying the G-formula, which is well-known from causal inference theory. Similar ideas have recently been implanted in the R-package Causal forest; however, this implementation cannot accommodate the time-to-event type outcomes that we analyse in HERMES. We have extended the method to also handle time-to-event outcomes.
Discussion/interpretation. The data collection is ongoing. We will present a first comparison of modelled and measured UFP concentrations at regional, urban background and street level. We will present our first experience with the performance of the new statistical model on simulated data.
Poster by Raaschou-Nielsen et al., 2019 HEI Annual Conference