Dataset Information

COVARIATE-ADAPTIVE CLUSTERING OF EXPOSURES FOR AIR POLLUTION EPIDEMIOLOGY COHORTS.

ABSTRACT: Cohort studies in air pollution epidemiology aim to establish associations between health outcomes and air pollution exposures. Statistical analysis of such associations is complicated by the multivariate nature of the pollutant exposure data as well as the spatial misalignment that arises from the fact that exposure data are collected at regulatory monitoring network locations distinct from cohort locations. We present a novel clustering approach for addressing this challenge. Specifically, we present a method that uses geographic covariate information to cluster multi-pollutant observations and predict cluster membership at cohort locations. Our predictive k-means procedure identifies centers using a mixture model and is followed by multi-class spatial prediction. In simulations, we demonstrate that predictive k-means can reduce misclassification error by over 50% compared to ordinary k-means, with minimal loss in cluster representativeness. The improved prediction accuracy results in large gains of 30% or more in power for detecting effect modification by cluster in a simulated health analysis. In an analysis of the NIEHS Sister Study cohort using predictive k-means, we find that the association between systolic blood pressure (SBP) and long-term fine particulate matter (PM2.5) exposure varies significantly between different clusters of PM2.5 component profiles. Our cluster-based analysis shows that for subjects assigned to a cluster located in the Midwestern U.S., a 10 ?g/m3 difference in exposure is associated with 4.37 mmHg (95% CI, 2.38, 6.35) higher SBP.

SUBMITTER: Keller JP

PROVIDER: S-EPMC5448716 | biostudies-literature | 2017 Mar

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

COVARIATE-ADAPTIVE CLUSTERING OF EXPOSURES FOR AIR POLLUTION EPIDEMIOLOGY COHORTS.

Keller Joshua P JP Drton Mathias M Larson Timothy T Kaufman Joel D JD Sandler Dale P DP Szpiro Adam A AA

The annals of applied statistics 20170301 1

Cohort studies in air pollution epidemiology aim to establish associations between health outcomes and air pollution exposures. Statistical analysis of such associations is complicated by the multivariate nature of the pollutant exposure data as well as the spatial misalignment that arises from the fact that exposure data are collected at regulatory monitoring network locations distinct from cohort locations. We present a novel clustering approach for addressing this challenge. Specifically, we ...[more]

PMID: 28572869

Similar Datasets

Project description:Few epidemiologic studies have evaluated the effects of air pollution on the risk of Parkinson disease (PD).We investigated the associations of long-term residential concentrations of ambient particulate matter (PM) < 10 ?m in diameter (PM10) and < 2.5 ?m in diameter (PM2.5) and nitrogen dioxide (NO2) in relation to PD risk.Our nested case-control analysis included 1,556 self-reported physician-diagnosed PD cases identified between 1995 and 2006 and 3,313 controls frequency-matched on age, sex, and race. We geocoded home addresses reported in 1995-1996 and estimated the average ambient concentrations of PM10, PM2.5, and NO2 using a national fine-scale geostatistical model incorporating roadway information and other geographic covariates. Air pollutant exposures were analyzed as both quintiles and continuous variables, adjusting for matching variables and potential confounders.We observed no statistically significant overall association between PM or NO2 exposures and PD risk. However, in preplanned subgroup analyses, a higher risk of PD was associated with higher exposure to PM10 (ORQ5 vs. Q1 = 1.65; 95% CI: 1.11, 2.45; p-trend = 0.02) among women, and with higher exposure to PM2.5 (ORQ5 vs. Q1 = 1.29; 95% CI: 0.94, 1.76; p-trend = 0.04) among never smokers. In post hoc analyses among female never smokers, both PM2.5 (ORQ5 vs. Q1 = 1.79; 95% CI: 1.01, 3.17; p-trend = 0.05) and PM10 (ORQ5 vs. Q1 = 2.34; 95% CI: 1.29, 4.26; p-trend = 0.01) showed positive associations with PD risk. Analyses based on continuous exposure variables generally showed similar but nonsignificant associations.Overall, we found limited evidence for an association between exposures to ambient PM10, PM2.5, or NO2 and PD risk. The suggestive evidence that exposures to PM2.5 and PM10 may increase PD risk among female never smokers warrants further investigation. Citation: Liu R, Young MT, Chen JC, Kaufman JD, Chen H. 2016. Ambient air pollution exposures and risk of Parkinson disease. Environ Health Perspect 124:1759-1765;?http://dx.doi.org/10.1289/EHP135.

Project description:BACKGROUND:Residential proximity to major roadways, and prenatal exposures to particulate matter <2.5??m (PM2.5) and ozone (O3) are linked to poor fetal outcomes but their relationship with childhood development is unclear. OBJECTIVES:We investigated whether proximity to major roadways, or prenatal and early-life exposures to PM2.5 and O3 increase the risk of early developmental delays. STUDY DESIGN:Prospective cohort. SETTINGS:New York State excluding New York City. PARTICIPANTS:4089 singletons and 1016 twins born between 2008 and 2010. EXPOSURES:Proximity to major roadway was calculated using road network data from the NY Department of Transportation. Concentrations of PM2.5 and O3 estimated by the Environmental Protection Agency Downscaler models were spatiotemporally linked to each child's prenatal and early-life addresses incorporating residential history, and locations of maternal work and day-care. OUTCOMES:Parents reported their children's development at ages 8, 12, 18, 24, 30 and 36 months in five domains using the Ages and Stages Questionnaire. Generalized mixed models estimated the relative risk (RR) and 95% CI for failing any developmental domain per 10 units increase in PM2.5 and O3, and for those living <1000?m away from a major roadway compared to those living further. Models adjusted for potential confounders. RESULTS:Compared to those >1000?m away from a major roadway, those resided 50-100?m [RR: 2.12 (1.00-4.52)] and 100-500?m [RR: 2.07 (1.02-4.22)] away had twice the risk of failing the communication domain. Prenatal exposures to both PM2.5 and ozone during various pregnancy windows had weak but significant associations with failing any developmental domain with effects ranging from 1.6% to 2.7% for a 10??g/m3 increase in PM2.5 and 0.7%-1.7% for a 10?ppb increase in ozone. Average daily postnatal ozone exposure was positively associated with failing the overall screening by 8 months [3.3% (1.1%-5.5%)], 12 months [17.7% (10.4%-25.5%)], and 30 months [7.6%, (1.3%-14.3%)]. Findings were mixed for postnatal PM2.5 exposures. CONCLUSIONS:In this prospective cohort study, proximity to major roadway and prenatal/early-life exposures to PM2.5 and O3 were associated with developmental delays. While awaiting larger studies with personal air pollution assessment, efforts to minimize air pollution exposures during critical developmental windows may be warranted.

Project description:Short-term exposure to air pollution has adverse effects among patients with asthma, but whether long-term exposure to air pollution is a cause of adult-onset asthma is unclear.We aimed to investigate the association between air pollution and adult onset asthma.Asthma incidence was prospectively assessed in six European cohorts. Exposures studied were annual average concentrations at home addresses for nitrogen oxides assessed for 23,704 participants (including 1,257 incident cases) and particulate matter (PM) assessed for 17,909 participants through ESCAPE land-use regression models and traffic exposure indicators. Meta-analyses of cohort-specific logistic regression on asthma incidence were performed. Models were adjusted for age, sex, overweight, education, and smoking and included city/area within each cohort as a random effect.In this longitudinal analysis, asthma incidence was positively, but not significantly, associated with all exposure metrics, except for PMcoarse. Positive associations of borderline significance were observed for nitrogen dioxide [adjusted odds ratio (OR) = 1.10; 95% CI: 0.99, 1.21 per 10 ?g/m3; p = 0.10] and nitrogen oxides (adjusted OR = 1.04; 95% CI: 0.99, 1.08 per 20 ?g/m3; p = 0.08). Nonsignificant positive associations were estimated for PM10 (adjusted OR = 1.04; 95% CI: 0.88, 1.23 per 10 ?g/m3), PM2.5 (adjusted OR = 1.04; 95% CI: 0.88, 1.23 per 5 ?g/m3), PM2.5absorbance (adjusted OR = 1.06; 95% CI: 0.95, 1.19 per 10-5/m), traffic load (adjusted OR = 1.10; 95% CI: 0.93, 1.30 per 4 million vehicles × meters/day on major roads in a 100-m buffer), and traffic intensity (adjusted OR = 1.10; 95% CI: 0.93, 1.30 per 5,000 vehicles/day on the nearest road). A nonsignificant negative association was estimated for PMcoarse (adjusted OR = 0.98; 95% CI: 0.87, 1.14 per 5 ?g/m3).Results suggest a deleterious effect of ambient air pollution on asthma incidence in adults. Further research with improved personal-level exposure assessment (vs. residential exposure assessment only) and phenotypic characterization is needed.

Dataset Information

COVARIATE-ADAPTIVE CLUSTERING OF EXPOSURES FOR AIR POLLUTION EPIDEMIOLOGY COHORTS.

Publications

COVARIATE-ADAPTIVE CLUSTERING OF EXPOSURES FOR AIR POLLUTION EPIDEMIOLOGY COHORTS.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets