May 26, 2015
Spiegelman & colleagues’ work strengthens the causal evidence for the health effects of chronic exposure to air pollution by eliminating uncertainty due to exposure measurement error
Opponents of stronger legislation to mitigate exposure to air pollution in the U.S. and around the world have long argued that the available data is difficult to causally interpret due to an unspecified degree of exposure measurement error. Dr. Spiegelman and colleagues in the Departments of Epidemiology, Environmental Health, and Biostatistics have addressed this critique head-on in a series of three recently published manuscripts on this topic. In the first one, first authored by former Environmental Health doctoral student Marianthi-Anna Kioumourtzoglou, all exposure validation data available in the United States that relate personal exposure to air pollution constituents to the standard measures used in the large-scale epidemiologic studies which have established the health effects were pooled. Kioumourtzoglou and her colleagues (1) found that the correlation between personal exposure to air pollution of ambient origin and the typical exposure surrogates used in epidemiology ranged between 50-60% and varied significantly by season and measures of urbanicity. Next, with colleagues Jaime Hart and Francine Laden in the EH Department, and using new statistical methods, the development of which was led by Xiaomei Liao and colleagues in the Departments of Epidemiology and Biostatistics, these data were used to explicitly correct estimates of the health effects of chronic air pollution exposure for bias due to exposure measurement error in Harvard’s longstanding Nurses’ Health Study, a prospective cohort of over 100,000 U.S. women who have been followed since 1976.(2) The authors considered the impact of measurement error in two commonly used exposure estimation methods – spatiotemporal models and EPA nearest monitor, using Kioumourtzoglou’s pooled exposure validation study. After adjusting for a large number of well-measured potentially confounding risk factors for all-cause mortality and after correcting for exposure measurement error, each 10 μg/m3 increase in exposure to PM2.5 of ambient origin was associated with a statistically significant 18%-22% increased risk of mortality, depending on which error-prone exposure assessment was used in the uncorrected analysis, nearly double the 12-13% statistically significantly increased risk observed using the mis-measured methods of exposure assessment. By eliminating exposure measurement error as a source of uncertainty in interpreting these results, Spiegelman and colleagues demonstrate that the health effects of chronic exposure to air pollution have previously been moderately to substantially under-estimated.
To put these risks into perspective, in 1988 and 2013, average PM2.5 levels in the U.S. were 35 and under 10 μg/m3, respectively, while currently in Beijing, China and Delhi, India, they are approximately 100 and 150 μg/m3, respectively.
A similar approach was taken to remove uncertainty due to exposure measurement error in a key study based in the Netherlands, the Netherlands Cohort Study of Diet and Cancer, which has examined the impact of several exposures contained in air pollution on lung cancer risk.(3) With Drs. Hart, Piet van den Brandt HSPH ’84, and other colleagues at HSPH and in the Netherlands, Spiegelman adjusted for other risk factors for lung cancer and corrected for exposure measurement error, and found elevated 19% and 37% relative risks for black smoke and PM2.5 for each 10 μg/m3 increase in exposure to each, compared to the 16-17% estimated prior to the measurement error correction. These results again demonstrate that the health effects of chronic exposure to air pollution persist after uncertainty due to exposure measurement error is eliminated, and suggest that in fact the health effects have previously been moderately to substantially under-estimated.
Spiegelman has always taken the approach that advanced statistical methodology must be accessible to have an impact. Thus, user-friendly macros to implement these methods can be obtained at http://www.hsph.harvard.edu/donna-spiegelman/software/ blinplus-macro/ and ../rrc . The pooled U.S. exposure validation study can be obtained by request at http://www.hsph.harvard.edu/pm2-5-validation-dataset/.
1 Kioumourtzoglou MA, Spiegelman D, Szpiro AA, et al. Exposure measurement error in PM2.5 health effects studies: A pooled analysis of eight personal exposure validation studies. Environ Health. 2014;13(1):2.
2 Hart JE, Liao X, Hong B, et al. The association of long-term exposure to PM2.5 on all-cause mortality in the Nurses’ Health Study and the impact of measurement-error correction. Environ Health. 2015;14(1):38.
3 Hart JE, Spiegelman D, Beelen R, et al. Long-Term Ambient Residential Traffic-Related Exposures and Measurement Error-Adjusted Risk of Incident Lung Cancer in the Netherlands Cohort Study on Diet and Cancer. Environ Health Perspect. 2015 Mar 27.
October 6, 2014
Donna Spiegelman, professor of epidemiologic methods at the Harvard T.H. Chan School of Public Health, has received a Director’s Pioneer Award from the National Institutes of Health (NIH) for $500,000/year of direct costs for the next 5 years. One of 10 researchers honored, Spiegelman is believed to be the first epidemiologist and biostatistician, and the first faculty member from a school of public health, to receive the award.
Harvard Chan School Press Release here.
Harvard Gazette coverage here.
The National Institutes of Health (NIH) Press Release here.
Association of Schools & Programs of Public Health (ASPPH) Coverage here.
Society for Epidemiologic Research (SER) Coverage here.
epimonitor Coverage here.
Amstat News Coverage here.
Harvard Chan School EpiCenter Department of Epidemiology Newsletter here.
Cambridge Chronicle & Tab here.
International Biometric Society’s Biometric Bulletin here.