Finally, the Harvard researchers announced they were going to give the pollution data to the Health Effects Institute (HEI), while withholding the health data. Later, under continuing pressure, Schwartz and Dockery said they would turn over the health data as well, albeit under certain stringent conditions. HEI President Dan Greenbaum says it will take two to three years to complete an evaluation.
But the EPA has hardly been persistent about trying to get the data. One agency official wrote to Commerce Committee Chairman Thomas Bliley (R-Va.) saying there was no need "for EPA to obtain the underlying data" since the studies were published in peer-reviewed journals. Browner later told Congress the same thing. But that's not what peer review is about. You can't review what you can't see; all you can do is ensure that what is in front of you is correct. If an author is selectively presenting data that fit his hypothesis, there's little a medical journal or peer reviewer can do.
HEI's Greenbaum sympathizes with Schwartz and Dockery's recalcitrance. "It's a complex database," he says, "and you need procedures to make sure only qualified investigators can look at this stuff. Some people tend to be analysts for hire." And indeed, Schwartz has told The Wall Street Journal that he doesn't want to provide the information to "industry thugs," by which he means scientists who receive industry money.
But those who have been able to review his data have raised questions. "Going back through Schwartz's results, we found he didn't report everything, only positive results" linking pollution to health problems, says Roth, adding that "[h]e always uses a different analysis in different cities." For example? "Well, one factor is how many days of weather you're going to count before the day of death or hospital admission--the previous five days, the previous six days, or whatever. What he does is when he goes to different cities, he uses a different lag." Roth adds, "Had he been consistent, he would have found insignificant results in some of his cities."
Moolgavkar says he also sees such problems in Schwartz's work. But it's not just Schwartz who does this, he says. "If you look at various PM studies, some will use a lag of two days, some one day or some even three or four days. To me that's not consistency. To me, it means you have to go through contortions to get a specific result."
Death and Witchcraft
Such contortions are not a challenge for the EPA. The agency has ostensibly compiled all of the relevant data on PM into what is called the Criteria Document. If you want a copy, bring a forklift: It runs to some 2,400 pages. But this poundage masks a vital little fact: Only a handful of Browner's 86 studies contain direct measurements of fine particles. The rest estimate fine particles as a percentage of the total or don't discuss fine particles at all. Harvard's Dockery kindly sent REASON a list of other studies with direct PM2.5 measurements. With the addition of a subsequent study, the total stands at a baker's dozen.
Of the four studies looking for premature deaths, two shared the same database, and only one unequivocally found a significant association between such particles and premature death among nonsmokers and persons without occupation exposure: A 1995 study with Pope as lead author that relied on American Cancer Society health data and is known as ACS II. Among the most important of the PM studies, it appeared in The American Journal of Respiratory and Critical Care Medicine, published by the American Thoracic Society, a unit of the ALA.
Though it failed to adjust for weather (not even temperature), the biggest problem with ACS II is that the PM2.5 data were collected from 1979 to 1981, but the death data covered the years 1983 through 1989. "Did the PM2.5 level change after this?" asks Utah's Lyon. "Did they clean up the air? Arden [Pope] doesn't know, and nobody else knows."
The study assumes that either the PM2.5 level remained the same in all the cities, or that if it changed, it changed equally in all the cities. Neither assumption could possibly have been true. Indeed, Brookhaven's Lipfert himself noted in a 1995 paper that during this period the dirtiest cities could have been expected to be cleaning up their air at a much quicker pace than cities already in compliance with federal law.
Of the four premature death PM2.5 studies, only one besides ACS II showed increased deaths among nonsmokers. It found statistically significant associations between particle increases and increases in death in three cities, and no associations in the three others it looked at. One of the negative cities, Steubenville, also had the largest increase in fine particles. But because the data set from one of the significant cities (Boston) was so large, this made the association in all the cities combined statistically significant.
How about symptom studies? According to last year's "Gambling with Health II," an ALA report complaining that the EPA was too soft on particulates, "A number of studies have linked fine particle pollution (PM2.5) with increased prevalence of coughs and bronchitis, with especially severe effects on children with asthma." Really? Of the nine PM2.5 symptoms studies, four showed no association between fine particles and hospitalizations or any symptoms measured. Five showed an association between fine particles and some of the symptoms measured.
"We've just arbitrarily decided PM2.5 is the villain," says Lyon. "This comes as close to witchcraft as anything I've seen. This isn't science. This is a firmly held warm and fuzzy belief: `I know these things because I feel them.'"
Lipfert and EPRI's Wyzga reviewed 30 published studies in the December 1995 Journal of the Air & Waste Management Association, all they could find by that date, examining associations between deaths and particle pollution. Some used PM10 as the measure, others PM2.5. They then compared the two sets of studies. Their conclusion was that, if anything, PM10 was associated with a slightly higher risk, though there couldn't be a determination of statistical significance since not enough studies were involved.
The EPA's own Criteria Document summarizes the results of the Harvard Six Cities Study, including a table comparing the relative risks for increased death in adults. For PM10 or PM15, it is a mean increase of 42 percent. For PM2.5, it is 31 percent. A mean increase is a comparison between zero pollution and the actual amount, as opposed to, say, comparing one polluted city with another. In other words, if you got rid of all the pollution, health would not change much.
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