View Abstract

A2802 - Advanced Data-driven Approaches To Identify Susceptibility Windows During Gestation For The Effects Of Ambient Air Pollution Upon Child Respiratory Outcomes.
Author Block: S. Bose1, B. A. Coull2, A. Wilson3, Y. Chiu4, H. Hsu5, I. Kloog6, Q. Di7, A. G. Lee8, M. Rosa9, J. Schwartz10, S. Cohen11, W. J. Morgan12, R. O. Wright13, R. J. Wright14; 1Division of Pulmonary and Critical Care Medicine, Icahn School of Medicine at Mount Sinai, New York, NY, United States, 2Department of Biostatistics, Harvard T.H. Chan School of Public Health, Boston, MA, United States, 3Department of Statistics, Icahn School of Medicine at Mount Sinai, Colorado State University, CO, United States, 44Department of Environmental Medicine and Public Health, Icahn School of Medicine at Mount Sinai, New York, NY, United States, 5Division of Pulmonary and Critical Care Medicine, Department of Environmental Medicine and Public Health, New York, NY, United States, 65Department of Geography and Environmental Development, Ben-Gurion University of the Negev, BeerSheba, Israel, 7Department of Environmental Health, Harvard T.H. Chan School of Public Health, Boston, MA, United States, 8Division of Pulmonary and Critical Care Medicine, Icahn School of Medicine at Mount Sinai, new york, NY, United States, 9Department of Environmental Medicine and Public Health, Icahn School of Medicine at Mount Sinai, New York, NY, United States, 10Dept of Envir Hlth/Dept of Epidem, Harvard Sch of Public Health, Boston, MA, United States, 11Department of Psychology, Carnegie Mellon University, Pittsburgh, PA, United States, 12Arizona Hlth Sci Ctr, Tucson, AZ, United States, 13Institute for Exposomic Research, Icahn School of Medicine at Mount Sinai, New York, NY, United States, 14Institute for Exposomic Research, Mount Sinai School of Medicine, New York, NY, United States.
Rationale: Evidence links prenatal air pollutant exposures to childhood respiratory outcomes, but the effects of exposure timing remain unclear. Prior work is limited by arbitrary classifications of exposure timing, rather than coinciding exposure with relevant embryologic processes, leading to potentially missed or biased findings.
Methods: We implemented novel statistical methods to leverage daily, highly spatially resolved ambient air pollution exposure estimates for 752 mother-child pairs from the ACCESS birth cohort to objectively identify critical sensitive windows and exposure effect heterogeneity for respiratory outcomes. First, we conducted a simulation study to demonstrate that using predefined windows, such as mean exposure over a trimester, can result in biased findings and that distributed lag models (DLM) eliminate this bias. We subsequently developed Bayesian distributed lag interaction models (BDLIMs) that give greater flexibility to assess effect heterogeneity by partitioning the DLM into two components—the location of the window, and the within-window effect—and allowing each component to vary across subgroups of interest. This allows for effect heterogeneity in only the timing of the sensitive window (with the same within-window effects), heterogeneity in the within-window effect (but the same timing of the window), or both. We applied this novel approach to determine vulnerable windows for the independent influence of prenatal airborne fine particulate matter (PM2.5) and nitrate (NO3-) exposures upon respiratory outcomes, including clinician-diagnosed asthma and lung function, assessing for effect modification by sex and/or maternal stress.
Results: Applied DLMs adjusted for child age, sex, and maternal factors identified a sensitive window at 16-25 weeks’ gestation during which PM2.5 exposure was significantly associated with asthma development by age 6 specifically among boys. BDLIMs further identified a critical exposure window of 19-21 weeks’ gestation such that boys exposed in utero to high PM2.5 and maternal stress were most vulnerable to developing asthma. Concomitant prenatal exposures to NO3- and high maternal stress were significantly associated with increased asthma risk during two distinct windows--one early at 7-19 weeks and one late at 33-40 weeks. An identified sensitive window at 6-12 weeks gestation also linked NO3- exposure with significantly reduced FEV1 in boys at age 7.
Conclusion: Data-driven approaches such as BDLIM can identify unique prenatal windows of vulnerability to specific pollutant exposures resulting in adverse child respiratory outcomes with differential effects by pollutant composition and population subgroups. Overlaying critical windows with corresponding developmental processes disrupted by exposures may provide future insight into underlying mechanisms.