Then Lyon and his colleagues compared Utah County with Salt Lake County. The adjacent counties suffer the same air inversions, and while Utah County has the steel plant, Salt Lake County has a copper smelting plant that puts out more particulates. In any case, says Lyon, "There is a substantial air mixing between the two counties." What did he find? "We essentially found no association in Salt Lake County to PM10," says Lyon. "There was simply no effect."
Lyon says that "if the relationship is causal" between particulates and hospitalizations, "you'd expect it to be pretty consistent. Yet we have found an association that in some years isn't present, and is inconsistent when comparing two counties that share the same air. So just how causal can this thing be?" Others have come to the same conclusion, among them Patricia Styer and others at the National Institute of Statistical Sciences (NISS), who studied both Salt Lake County and Cook County, Illinois, in 1995 and concluded in Environmental Health Perspectives that "[t]he reported effects of particulates on mortality are unconfirmed."
But to the EPA, environmental groups, and the media, Pope remains infallible. Along with Schwartz and another Harvard researcher, Douglas Dockery, Pope has become one of the Particle Hunter Triumvirate. Indeed, when you look at all the American studies that find particles causing health associations, these three names keep popping up as the authors.
Little Things Mean a Lot
In Birmingham, Alabama, Schwartz found that deaths went up when PM10 went up. In a separate paper, he showed that increases in PM10 and hospital admissions among the elderly were linked. Jonathan Samet of Johns Hopkins University (who, unlike Schwartz, has medical credentials) did an evaluation of these studies for the Health Effects Institute and confirmed Schwartz's findings. Thus, there are three different papers showing that PM10 levels in present-day Birmingham are high enough to cause harm, even death. Case closed? Not quite.
Because swings in temperature can hospitalize and kill, good particulate studies take temperature into account. The Birmingham studies did this. But what they didn't account for is one of Birmingham's most distinctive features--summer humidity that wraps you up in a hot, sticky blanket of moist air. Researchers at NISS in North Carolina, an area known for its own humidity, did account for it when they did a follow-up study of Birmingham. In an as-yet-unpublished report funded by the EPA, NISS factored in humidity changes and found that both for illness and death "the PM10 effect is not statistically significant."
It may be surprising that a factor like humidity can invalidate the results of such a study. But it's less so if you understand that one of the most consistent features of particulate studies is that they are anything but "robust." When researchers do find an increase in deaths or illness, it's always a tiny one. "The relative increase in total mortality and morbidity associated with a 50 percent increase in air particulates is not large," admitted CASAC member and University of North Carolina epidemiologist Carl Shy in his congressional testimony backing the EPA's proposed standard, "being on the order of 5 to 10 percent above [that] on days with the lowest concentrations." But he added that because this was spread across such a large population, you could be talking about a lot of sick and dying people. That misses the point, which is that when you have just a tiny increase, any little mistake in your methodology could account for it.
In other contexts, the public health community argues that such increases are meaningless. After one study suggested that women who had undergone induced abortions had a 50 percent higher breast cancer risk, an editorial in the Journal of the National Cancer Institute warned that such a difference "is small in epidemiologic terms and severely challenges our ability to distinguish if it reflects cause and effect or if it simply reflects bias." It was a typical reaction to that study and, though politically biased, correct.
But with particulates, a mere 5 percent to 10 percent apparent increase in risk is enough to justify foisting new regulations costing hundreds of billions of dollars on the public. The only federal agency that has ever used such a tiny apparent increase as a basis for regulations was, in fact, the same EPA. That was when the agency decided that a 19 percent increased risk was solid enough to institute sweeping regulations against secondhand cigarette smoke.
Indeed, the particulate studies provide a textbook example of why small increases in risk may mean nothing. Merely focusing on different periods of time can completely alter a study's results. Neil Roth of Roth Associates in Rockville, Maryland, looked for a pollution health hazard in Birmingham using later data than Schwartz did. "Ours went from the late '80s into the early '90s, while his were from 1985 through 1988," says Roth. "We did thousands of different analyses on both the hospital admissions and the mortality data," he says, and "found the overwhelming majority of results were not statistically significant. Of those that were, half were positive but the other half were negative. So, in short, we found no evidence of health hazards from particulates in Birmingham."
This happens repeatedly. Schwartz gave Stanford statisticians his data showing a significant association between hospital admissions and particulate pollution in Detroit. When the statisticians analyzed them the same way he did, they found the same thing. But when they incorporated the potential influence of the day of the week into the model (because people tend to go to hospitals on certain days), particulate matter was no longer a significant factor. Similarly, when Dr. Suresh Moolgavkar, of the Fred Hutchinson Cancer Research Center in Seattle, and other researchers looked at Schwartz and Dockery's work correlating deaths with particulates in Steubenville, Ohio--one of America's most polluted cities--they found no such connection. When Schwartz looked at Minneapolis-St. Paul, he found that particle increases caused more old people to go to hospitals. When Moolgavkar and associates looked, they found the strongest association with ozone.
A Tale of Six Cities
The so-called Harvard Six Cities Study, in which Dockery, Pope, and others compared air pollution and premature deaths in six American cities, is one of the most important weapons in the EPA's PM2.5 advocacy arsenal. Published in the December 9, 1993, New England Journal of Medicine, the study found that Steubenville, with the most air pollution and most particulate pollution of the cities, had a 26 percent higher mortality rate than Portage, Wisconsin, the cleanest city.
But what the researchers also found--and what none of the particle pursuers ever talk about--was that among nonsmokers there was no statistically significant difference in deaths between these two cities. Moreover, there was none if you excluded persons with occupational exposures to "gases, fumes, or dust." It was only when including smokers, former smokers, and persons with occupational exposures that the researchers were able to get significant findings. You can see this in the study's own Table 3.
But in the text, Dockery and his associates state, "Although cigarette smoking and other risk factors were associated with mortality, our estimates of pollution-related mortality were not significantly affected by the inclusion or exclusion of these variables in the models." How do they figure that? In the study's Table 4, they abandon the "nonsmokers" category, replacing it with a new one that blurs the distinctions between the "nonsmoker," "current smoker," and "former smoker" categories. Voilà! They have now converted the entire Six Cities Study from negative to positive. Do that in your doctoral dissertation, and you could be in serious trouble. Do it in a medical journal, and you're the hero of the EPA.
Editor's Note: We invite comments and request that they be civil and on-topic. We do not moderate or assume any responsibility for comments, which are owned by the readers who post them. Comments do not represent the views of Reason.com or Reason Foundation. We reserve the right to delete any comment for any reason at any time.
nfl jerseys|11.16.10 @ 2:07AM|#
mcyjfg