You are here

Mortality and morbidity effects of long-term exposure to low-level PM2.5, Black Carbon, NO2 and O3: an analysis of European cohorts

Principal Investigator: 
,

University of Utrecht, Netherlands

This study will investigate health effects of low levels of air pollution in Europe using pooled data from 10 ESCAPE cohorts with individual covariate information, and 6 large administrative cohorts with less detailed information; resulting in a study population of about 25 million people. See also this Program Summary of HEI's research program on low levels of air pollution. 

Funded under
Status: 
Ongoing
Abstract
Abstract for the 2018 HEI Annual Conference. Please scroll down to view the three abstracts and posters.
 

Mortality, morbidity and low-level air pollution in a pooled cohort  of 485,000 in Europe  in the ELAPSE project

Bert Brunekreef,1 Maciej Strak,1 Jie Chen,1 Marjan Tewis,1 Kees de Hoogh,2,3 Sophia Rodopoulou,4 Evi Samoli,4 Klea Katsouyanni,4,5 Gerard Hoek,1 on behalf of the ELAPSE Project Team

1 Institute for Risk Assessment Sciences, Utrecht University, the Netherlands; 2 Swiss Tropical and Public Health Institute, Basel, Switzerland; 3 University of Basel, Switzerland; 4 Medical School, University of Athens, Greece; 5 School of Population Health & Environmental Sciences, King's College London, UK

Background

Epidemiological cohort studies have consistently found associations between long-term exposure to outdoor air pollution and morbidity and mortality endpoints. Recent evaluations by the World Health Organization and the Global Burden of Disease study have suggested that these associations may  persist at very low concentrations. However, uncertainty about the shape of the concentration response function exists at these low concentrations. Within the multicentre Effects of Low-Level Air Pollution: A Study in Europe (ELAPSE), we investigate the association between long-term exposure to low concentrations, defined as less than current EU, EPA and WHO Limit Values or guidelines, of several air pollutants and morbidity and mortality endpoints.

Methods

To assess long-term residential exposure, we developed a Europe-wide hybrid land use regression models estimating annual 2010 mean concentrations of PM2.5, NO2, O3 and BC (including cold and warm season estimates for O3). The models were based on AirBase routine monitoring data and ESCAPE monitoring data (for BC) and incorporated land use and traffic data supplemented with satellite observations and dispersion model estimates as additional predictor variables. Models were applied to 100*100 m grids across Europe to allow for exposure assignment in all ELAPSE cohorts. We applied multivariate Cox proportional hazard models to investigate the association between long-term air pollution exposure and a number of morbidity and mortality endpoints in a pooled cohort (this abstract) and in several large administrative cohorts (companion abstract).

Results

We pooled data, so far, from eight European cohort studies, resulting in a study population of 400,187 subjects. We are adding three more cohorts, increasing the study population with 85,000 subjects. The average follow up time in the pooled cohort was 19 years. Average exposure to air pollution was 15.1 (SD 3.3) µg/m3 for PM2.5, 25.1 (8.1) µg/m3 for NO2, 1.5 (0.4) µg/m3 for BC, and 67.6 (6.9) µg/m3 for (annual) O3. PM2.5 was moderately correlated with NO2 (Pearson’s r 0.53) and highly correlated with BC (0.80), whereas NO2 was moderately correlated with BC (0.69). Ozone had a high negative correlation with NO2 (-0.81) and low negative correlations with the two other pollutants (-0.23 to -0.26). First results of the associations between exposure and selected endpoints will be presented at the HEI annual meeting.

Conclusions

We successfully pooled data and assigned exposure for several European cohorts. Analyses of associations between low-level air pollution and morbidity and mortality endpoints are ongoing. Details regarding statistical analyses can be found in another companion abstract.

POSTER

Poster by Brunekreef, Strak et al, 2018 Annual Conference


Mortality, morbidity and low-level air pollution in a population of 35 million in Europe – analysis of administrative cohorts in the ELAPSE project

Bert Brunekreef,1 Danielle Vienneau,2,3 Nicole Janssen,4 Massimo Stafoggia,5 Mariska Bauwelinck,6,7 Klea Katsouyanni,8,9 Evi Samoli,8 Sophia Rodopoulou,8 Maciej Strak,1 Gerard Hoek,1 on behalf of the ELAPSE Project Team

1 Institute for Risk Assessment Sciences, Utrecht University, the Netherlands; 2 Swiss Tropical and Public Health Institute, Basel, Switzerland; 3 University of Basel, Switzerland; 4 National Institute for Public Health and the Environment, Bilthoven, the Netherlands; 5 Lazio Region Health Service, Rome, Italy; 6 Free University Brussel, Brussels, Belgium; 7 Scientific Institute of Public Health, Brussels, Belgium; 8 Medical School, University of Athens, Greece; 9 School of Population Health & Environmental Sciences, King's College London, UK

Background

ELAPSE aims to investigate the adverse health effects of long-term exposure to low levels of ambient air pollution in Europe. To capture a large population and exposure contrast, we exploit large administrative cohorts geographically spread across Europe. Specifically, these are used to explore associations between exposure to NO2, PM2.5, Black Carbon and O3 and mortality as well as CVD and cancer incidence.

Methods

Europe-wide hybrid land use regression models for each pollutant were developed for 2010 (i.e. ELAPSE exposures). These were transferred to each cohort, and linked to participants on the basis of address-level geocodes. To facilitate harmonisation, the cohorts implemented the ELAPSE codebook where possible, including specified definitions for socio-economic status (SES) at different geographic scales. Each cohort further collected local air pollution exposure data, area-level SES data, and health survey data for indirect adjustment. Statistical scripts were centrally developed, on the basis of the study manual, and tested in a selection of cohorts. These apply Cox proportional hazard models with successively more detailed control for individual- and area-level confounders. Data from the administrative cohorts are being analysed locally and combined by meta-analysis.

Results

Seven very large administrative cohorts (>35 million participants) are included, with data access granted for 6 cohorts (English pending). Exposure estimates and area-level covariates have been linked to 5 cohorts (Belgian, Danish, Dutch, Rome and Swiss; Norwegian in progress). As an example, the median exposures range from 23, 16, 1.6 and 73 µg/m3, respectively for NO2, PM2.5, Black Carbon and O3, in the Swiss cohort (4.4 million adults). A series of working group and bilateral meetings have taken place to ensure harmonisation across the cohorts in particular for area-level SES definition and statistical modelling. Statistical scripts have been developed using the Rome cohort, and tested with the larger Swiss cohort. Testing has focused on aspects related to Cox models with time invariant exposure, basic confounder adjustment and the different options to control for potential differences in baseline hazard by region. Epidemiological results for a section of administrative cohorts will be presented at the HEI conference.

Conclusions

As more cohorts proceed with the analysis, further harmonisation of geographic areas for SES and model definition will be implemented to ensure the subsequent analyses (i.e., indirect adjustment, dose response modelling) are sufficiently comparable to allow meaningful meta-analyses.

POSTER

Poster by Brunekreef, Vienneau et al, 2018 Annual Conference


Development of the statistical protocol for the investigation of the health effects of long-term exposure to low air pollutant concentrations in the ELAPSE project

Bert Brunekreef,1 Klea Katsouyanni,2,3 Evangelia Samoli,2 Gerard Hoek,1 Sophia Rodopoulou,2 Maciej Strak,1 on behalf of the ELAPSE Project Statistical Group

1 Institute for Risk Assessment Sciences, Utrecht University, the Netherlands; 2 Medical School, University of Athens, Athens, Greece; 3 School of Population Health & Environmental Sciences, King's College London, London, UK

Background and objectives

ELAPSE investigates the effect of long-term exposure to low levels of PM2.5, NO2, O3 and black carbon on mortality and morbidity in European cohort studies. We present the statistical analysis protocol in this abstract.

Statistical methods

Data are pooled from 11 cohorts with detailed information at the individual level and analysed as one dataset.  Additionally 7 European administrative cohorts are included, analysed individually, with cohort-specific effect estimates consequently pooled by random-effects meta-analysis. Pooling individual data from the 11 cohorts presents a great challenge in terms of variables’ availability and harmonization. Further, area-level covariates are different between European countries as they are reported under varying geographical scales and definitions. The main exposure indices are the pollutants’ concentrations at the subject's residence estimated by modelling at specific reference years. We assess the latency of effects by using exposures 2, 4 and 6 years prior to each health event. Time-varying exposures and levels derived from the average of back extrapolated estimates and local models are also being assessed.

We apply Cox proportional hazard models with varying control for individual- and area-level covariates that are outcome-specific. In the pooled ESCAPE cohorts analysis we compare different approaches accounting for the clustering of data by cohort including correction of the standard errors, stratification and frailty models with random intercepts. We further apply multiple imputation by chained equations for main covariates such as smoking. In the administrative cohorts’ analysis we compare Cox models with and without control for large geographical areas and correction of the standard errors. In all analyses we assess model diagnostics, and heterogeneity between cohorts/areas. We use the pooled database and the UK administrative cohort, which includes information on several individual confounders, to evaluate the impact of missing covariates in the other administrative cohorts and further apply indirect adjustment correction methods previously proposed.

We apply multi-pollutant models and compare risk estimates from single and multi-pollutant models to disentangle interdependencies and pollutant-specific impact on the analysed outcomes. We evaluate effect modification patterns and investigate concentration-response by fitting fractional polynomials, spline and threshold models. Analysis are also being performed by subgroups below certain concentration levels. Finally we assess measurement error by regression calibration and comparison of effects using exposure estimates derived from 80% of the monitors for the LUR models.

Specific codes have been developed for all analyses in the R statistical package.

Discussion

Comparison of various models’ results demonstrates the robustness of our findings and provides an indirect approach to address the causality of the associations.

POSTER

Poster by Brunekreef, Katsouyanni et al, 2018 Annual Conference

PUBLICATIONS

de Hoogh K, Chen J, Gulliver J, Hoffmann B, Hertel O, Ketzel M, Bauwelinck M, van Donkelaar A, Hvidtfeldt UA, Katsouyanni K, Klompmaker J, Martin RV, Samoli E, Schwartz PE, Stafoggia M, Bellander T, Strak M, Wolf K, Vienneau D, Brunekreef B, Hoek G. Spatial PM2.5, NO2, O3 and BC models for Western Europe - Evaluation of spatiotemporal stability. Environ Int. 2018 Jul 31;120:81-92. doi: 10.1016/j.envint.2018.07.036. [Epub ahead of print]