TCS Daily


EPA's Faith-Based Pollution Standards

By Joel Schwartz - January 13, 2006 12:00 AM

Editor's note: This is the second of two parts on EPA's new particulate matter standards (the first of which can be found here).

There is no question that high levels of air pollution can kill. About 4,000 Londoners died during the infamous five-day "London Fog" of December 1952, when soot and sulfur dioxide soared to levels tens of times greater than the highest levels experienced in the United States today, and visibility dropped to less than 20 feet. [[1]] The question now is whether current, historically low levels of air pollution can also be deadly.

The Clinton administration adopted the nation's first standards for fine particulate matter (PM2.5) in 1997. The Environmental Protection Agency claims its new more stringent standard will save thousands of additional lives each year. [[2]] On the other hand, environmentalists and many air pollution researchers claim EPA is killing thousands of people by not clamping down even further.[[3]] As I detail below, both views are mistaken.

EPA based its annual PM2.5 standard mainly on the American Cancer Society (ACS) study, which followed more than 500,000 Americans in 50 cities from 1982 to 1998.[[4]] In their most recent report, the ACS researchers concluded that each 10 ug/m3 increase in long-term PM2.5 levels is associated with a 4 percent increase in risk of death.[[5]]

But the detailed ACS results are biologically implausible. For example, the ACS study reported that PM2.5 apparently kills men, but not women; those with no more than a high school degree, but not those with at least some college; and those who said they were moderately active, but not the very active or the sedentary.[[6]]

When migration rates into and out of various cities over time were added to the analysis, the apparent effect of PM2.5 disappeared. [[7]] Cities that lost population during the 1980s—Midwest "rust belt" cities—also had higher average PM2.5 levels. People left these economically depressed areas in search of work in more prosperous regions. But people who work and have the wherewithal to migrate also tend to be healthier than the average person. Hence, what appeared to be an effect of PM2.5 was more likely the result of differential migration.

Migration was just one of several confounding factors that diminished or erased the apparent harm from PM2.5, but that were not accounted for by the ACS researchers. The Harvard Six Cities study, another cohort study cited in support of PM-mortality claims, suffers from similar problems.[[8]]

A third long-term study of PM2.5 and mortality has been ignored by proponents of PM2.5 regulation, probably because it didn't find any harm from PM2.5 exposure. The study followed 50,000 male veterans with high blood pressure for 21 years. [[9]] These men's poor cardiovascular health should have made them more susceptible than the average person to any pollution-related health effects, so the study provides particularly strong evidence against a PM2.5-mortality link.

These long-term studies were based on PM levels during the 1970s and 1980s, which in many cities were two or three times greater than EPA's 15 ug/m3 annual PM2.5 standard. Thus, long-term PM2.5 exposure is unlikely to be killing people today even in areas that exceed the annual PM2.5 standard.

While there are only three major studies of long-term PM effects, literally thousands of studies have examined the association between daily fluctuations in PM levels and health outcomes. These studies typically report that daily fluctuations in PM levels are associated a few tenths of a percent increase in daily deaths. But these studies suffer from their own set of challenges.

One key problem is publication bias—the tendency for researchers and journal editors to publish studies that find an air pollution-health association rather than studies that fail to find such an association.[[10]] The National Morbidity and Mortality Air Pollution Study (NMMAPS) has revealed how large the effect of publication bias can be.

NMMAPS does not suffer from publication bias, because it evaluated the association of air pollution and mortality in 95 cities using the same statistical methods, and all the results were published. Based on NMMAPS, the apparent average risk from PM was only about one-third the typical level reported in single-city studies. Furthermore, for about one-third of the NMMAPS cities, higher daily PM levels were associated with a lower risk of death. [[11]]

Proponents of the study have made much of the fact that pooling the results from all 95 cities gives a statistically significant PM effect. But it isn't clear that a pooled result has any meaning when the individual city results suggest, implausibly, that PM protects people in some places and kills them in others. Furthermore, it turns out that removing just two outlier cities from the dataset reduces the pooled result to statistical insignificance. [[12]]

Another concern related to publication bias is known as model-selection bias or data mining. When researchers model the health effects of air pollution, in order to avoid spurious results they must control for potential confounding factors like weather and season. And since the effects of air pollution (or weather) might not show up until, say, a day or two after exposure, models need to include exposures over several days.

Once all potential combinations of variables are accounted for, there are literally millions of plausible statistical models relating air pollution to health outcomes, and no objective way to choose among them. Under these circumstances, researchers have a tendency to select those models that give the largest or most statistically significant effects. [[13]] But such a procedure risks turning up chance correlations even if no real cause-effect relationships exist. As two air pollution epidemiologists recently cautioned:

"Estimation of very weak associations in the presence of measurement error and strong confounding is inherently challenging. In this situation, prudent epidemiologists should recognize that residual bias can dominate their results. Because the possible mechanisms of action and their latencies are uncertain, the biologically correct models are unknown. This model selection problem is exacerbated by the common practice of screening multiple analyses and then selectively reporting only a few important results." [[14]]

A recent study concluded that the air pollution-mortality association disappears once model-selection bias is accounted for. [[15]]

Toxicology studies also fail to corroborate the epidemiological claims. Laboratory studies with both animals and human volunteers provide little reason to believe PM2.5 is killing people at the low levels found in ambient air. [[16]]

Thus, the entire PM2.5 regulatory enterprise rests fundamentally on the results of small and inconsistent statistical associations that are likely the spurious result of publication and model-selection biases. The result is unwarranted public fear, an ever-expanding regulatory state, and large amounts of Americans' income squandered on minute or perhaps non-existent risks.


Joel Schwartz is a visiting fellow at the American Enterprise Institute.



[1] I. M. Goklany, Clearing the Air: The Real Story of the War on Air Pollution (Washington, DC: Cato, 1999).

[2] EPA's press release and related materials can be downloaded here: http://yosemite.epa.gov/opa/admpress.nsf/4d84d5d9a719de8c85257018005467c2/1e5d3c6f081ac7ea852570de0050ae2b!OpenDocument.

[3] See, for example, http://www.npr.org/templates/story/story.php?storyId=5064021 and http://www.cleanairwatchpressroom.blogspot.com/.

[4] C. A. Pope, 3rd, M. J. Thun, M. M. Namboodiri et al., "Particulate Air Pollution as a Predictor of Mortality in a Prospective Study of U.S. Adults," American Journal of Respiratory and Critical Care Medicine 151 (1995): 669-74.

[5] C. A. Pope, 3rd, R. T. Burnett, M. J. Thun et al., "Lung Cancer, Cardiopulmonary Mortality, and Long-Term Exposure to Fine Particulate Air Pollution," Journal of the American Medical Association 287 (2002): 1132-41.

[6] D. Krewski, R. T. Burnett, M. S. Goldberg et al., Reanalysis of the Harvard Six Cities Study and the American Cancer Society Study of Particulate Air Pollution and Mortality (Cambridge, MA: Health Effects Institute, July 2000); Pope, Burnett, Thun et al., "Lung Cancer, Cardiopulmonary Mortality, and Long-Term Exposure to Fine Particulate Air Pollution."

[7] Krewski, Burnett, Goldberg et al., Reanalysis of the Harvard Six Cities Study and the American Cancer Society Study of Particulate Air Pollution and Mortality.

[8] Ibid.; F. W. Lipfert, "Estimating Air Pollution-Mortality Risks from Cross-Sectional Studies: Prospective vs. Ecologic Study Designs," Health and Regulatory Issues, Proceedings of the International Specialty Conference, Air and Waste Management Association, 1995; F. W. Lipfert, "Commentary on the HEI Reanalysis of the Harvard Six Cities Study and the American Cancer Society Study of Particulate Air Pollution and Mortality," Journal of Toxicology and Environmental Health, Part A 66 (2003): 1705-14; S. H. Moolgavkar, "A Review and Critique of the EPA's Rationale for a Fine Particle Standard," Regulatory Toxicology and Pharmacology 42 (2005): 123-44; J. Schwartz, Particulate Air Pollution: Weighing the Risks (Washington, DC: Competitive Enterprise Institute, April 2003), http://www.cei.org/pdf/3452.pdf.

[9] F. W. Lipfert, H. M. Perry, J. P. Miller et al., "The Washington University-EPRI Veterans' Cohort Mortality Study," Inhalation Toxicology 12 (suppl. 4) (2000): 41-73.

[10] Publication bias is a well-documented problem in a range of disciplines. See, for example, H. Anderson, R. Atkinson, J. Peacock et al., Meta-Analysis of Time-Series Studies and Panel Studies of Particulate Matter (PM) and Ozone (World Health Organization, 2004), www.euro.who.int/document/e82792.pdf; V. M. Montori, M. Smieja and G. H. Guyatt, "Publication Bias: A Brief Review for Clinicians," Mayo Clinic Proceedings 75 (2000): 1284-8; A. Thornton and P. Lee, "Publication Bias in Meta-Analysis: Its Causes and Consequences," Journal of Clinical Epidemiology 53 (2000): 207-16.

[11] F. Dominici, A. McDermott, M. Daniels et al., Revised Analyses of the National Morbidity, Mortality, and Air Pollution Study, Part II (Boston: Health Effects Institute, May 2003).

[12] Moolgavkar, "A Review and Critique of the EPA's Rationale for a Fine Particle Standard."

[13] Anderson, Atkinson, Peacock et al., Meta-Analysis of Time-Series Studies and Panel Studies of Particulate Matter (PM) and Ozone; G. Koop and L. Tole, "Measuring the Health Effects of Air Pollution: To What Extent Can We Really Say That People Are Dying from Bad Air?" Journal of Environmental Economics and Management 47 (2004): 30-54; T. Lumley and L. Sheppard, "Time Series Analyses of Air Pollution and Health: Straining at Gnats and Swallowing Camels?" Epidemiology 14 (2003): 13-4; Moolgavkar, "A Review and Critique of the EPA's Rationale for a Fine Particle Standard."

[14] Lumley and Sheppard, "Time Series Analyses of Air Pollution and Health: Straining at Gnats and Swallowing Camels?"

[15] Koop and Tole, "Measuring the Health Effects of Air Pollution: To What Extent Can We Really Say That People Are Dying from Bad Air?"

[16] L. C. Green and S. R. Armstrong, "Particulate Matter in Ambient Air and Mortality: Toxicologic Perspectives," Regulatory Toxicology and Pharmacology 38 (2003): 326-35; Moolgavkar, "A Review and Critique of the EPA's Rationale for a Fine Particle Standard."

Categories:
|

TCS Daily Archives