On February 12, 2019, The Mount Sinai Transdisciplinary Center on Effects of Early Environmental Exposure and the Conduits, the Mount Sinai CTSA co-sponsored a workshop on Critical Windows and Distributed Lag Regression (DLM) in collaboration with the Harvard P30 Center. There were 35 attendees that included faculty, postdoctoral fellows and data analysts. The facilitators included: Rosalind Wright (MSSM); Robert Wright (MSSM); Chris Gennings (MSSM); Brent Coull (Harvard); Ander Wilson (Colorado State); Leon Hsu (MSSM) and Mathilda Chiu (MSSM).
Recent advances in exposure science allow for very temporally dense measures of environmental exposures, such as air pollution and metals in deciduous teeth. In response, statistical methods were developed by MSSM P30 faculty in collaboration with Harvard and Colorado State faculty to bring together exposure science, biostatistics and developmental biology principles to discover, then validate the time boundaries of critical exposure windows. This DLM workshop was designed to integrate all aspects of these exposure windows so that participants would gain a deeper understanding of 1) why critical exposure windows exist, 2) the types of exposure data needed to identify windows 3) data driven statistical methods to identify windows as well as extensions that can address complex mixtures, effect modification and the overall effect of an analysis and 4) principles for interpreting results within the appropriate biological and statistical framework. The workshop emphasized the importance of timing in epidemiology studies and human development and also introduced how this methodology was first conceptualized in a transdisciplinary collaboration Drs. Wright and Coull, demonstrating the value of transdisciplinary scientific collaborations. As the temporal density and dimensions of health and environmental data keep increasing, we are now better equipped to understand data driven exposure biology methods increase, more and more application of these principles and methods are anticipated, providing new insights into the underlying mechanisms of the developmental origins of health and disease. The DLM workshop provided a hands-on experience for all participants to learn and apply DLM statistical code with exposure and health data. The workshop covered “Bayesian distributed lag model”, “Reversed distributed lag model”, and “Lagged Weighted Quantile Sum Model” in both “R” environment and “SAS” environment, inspiring participants many different approaches to solving environmental health problem with great complexity when dealing both high-resolution time series data and mixtures. After completing the workshop, participants should be able to confidently apply these analytic techniques on their own research.