Environmental Factor, June 2010, National Institute of Environmental Health Sciences
Taking Air Pollution Studies to the Next Level
By Eddy Ball
With a boost from NIEHS support, Harvard University biostatistician Francesca Dominici, Ph.D., is laying the groundwork for developing novel models to help solve one of the persistent mysteries of public health - the effects of environmental exposures to complex mixtures.
Speaking to an at-capacity audience at NIEHS April 29, Dominici(http://www.hsph.harvard.edu/faculty/francesca-dominici/) outlined the ways her work with two massive national databases covering 119 counties in the United States may help researchers pinpoint the health risks from the multiple individual components found in particulate air pollution mixtures. Dominici's talk, hosted by NIEHS Health Science Administrator Caroline Dilworth, Ph.D., drew epidemiologists, biostatisticians, and other scientists from NIEHS, as well as regulatory scientists from the neighboring U.S. Environmental Protection Agency (EPA), which also funds her research.
A professor of biostatistics at Harvard School of Public Health, Dominici focuses on developing statistical methods for integrating and analyzing large heterogeneous datasets to estimate health risks and evaluate the health impacts of environmental regulations. She is an internationally recognized leader in statistical methodology and in the estimation of the health effects of air pollution, with several fruitful collaborations with NIEHS-funded scientists, including Outstanding New Environmental Scientist award winner Michelle Bell, Ph.D. (see story(https://factor.niehs.nih.gov/2009/october/science-sameday.cfm)) Dominici was chosen to be the leading biostatistician in several Committees of the Institute of Medicine of the National Academies.
Air pollution poses complex health risks
Dominici began her talk by reminding her audience, "We breathe a mixture." Although researchers, especially regulatory scientists, have typically looked at single components in efforts to link a specific variable with a health effect, she said exposure is actually much more complex and involves multiple components with multiple end points. In addition, within mixtures there may be additive, agonistic, and synergistic effects that could also influence how each component impacts health endpoints in different combinations.
By understanding multi-pollutant behavior more completely, Dominici explained, scientists can more effectively estimate risks, influence policies for controlling harmful air pollution, and design compliance strategies to reduce health effects of particulates. If she and her colleagues are successful, they may also help resolve some of the contradictory outcomes of earlier single-component research tied to single endpoints and better understand geographical differences in air pollution and health effects.
Statistical size and power
Dominici will be working with two enormous data sets that she has used in previous studies. One, the most recent version of the National Mortality, Morbidity, and Air Pollution Study (NMMAPS), contains daily mortality, air pollution, and weather data on conditions in 108 U.S. cities from 1987 through 2006. The other, the Medicare Cohort Air Pollution Study (MCAPS), contains information on billing claims for everyone over 65 in 204 U.S. counties from 1999 through 2008. According to Dominici, the data in the two studies amounts to several terabytes of information she can mine for association with the seven major components of the more than fifty chemicals that constitute fine particulate matter (PM2.5) by means of new hierarchical models, as well as novel methods to account for adjustment uncertainty in effect estimation and to check regression models for data in space and time.
As she neared the end of her presentation, Dominici acknowledged the challenges researchers face with this monumental task, such as uncertainty of selection cofounders and identification of sources for components of the mixtures. Still, she insisted, answering the question of how PM2.5 components modify short- and long-term effects on mortality and morbidity is the key to translating 50 years of air pollution research into effective public health interventions.
"By identifying the toxicity of some specific agents in the mixture," Dominici said, "we will guide the development of hypotheses on biological mechanisms of action that can be tested in experimental models, and we will better inform public policy."