Introduction and Summary
EPA has proposed lowering the level of the National Ambient Air Quality Standard (NAAQS) for PM2.5. Under EPA’s proposal, the 24-hour PM2.5 standard would be lowered from 65 micrograms per cubic meter (ug/m3) down to 35 ug/m3, while keeping the annual standard at its current level of 15 ug/m3. If adopted, the proposed standard would double the fraction of PM2.5 monitors that violate federal PM2.5 standards, resulting in a substantial increase in both the stringency of federal PM2.5 standards and the difficulty of achieving regional PM2.5 attainment.
Based on pressure from EPA’s Clean Air Scientific Advisory Committee (CASAC) and from environmental groups and newspaper editorial boards, EPA will presumably also consider adopting a 24-hour and/or annual standard of even greater stringency than the current proposed rule. Instead, EPA should scrap its proposed rule to lower the PM2.5 NAAQS and should keep the PM2.5 NAAQS at their current levels.
As demonstrated in this comment letter, and contrary to EPA’s and CASAC’s claims, the current NAAQS already protect public health "with an adequate margin of safety," as required by the Clean Air Act (CAA). Indeed, standards even less stringent than the current PM2.5 NAAQS would also protect public health.
Health experts justify the current annual and 24-hour federal PM2.5 standards on the basis of circumstantial evidence. The evidence comes from observational epidemiology studies that report small residual correlations between particulate matter (PM) and premature mortality after controlling for confounders. The implicit assumption about these studies is that after controls are added for known or expected confounders, any residual correlation between air pollution levels and health outcomes represents a genuine causal linkage.
Thousands of observational air pollution health studies have been published in the literature. Proponents of tougher PM2.5 regulation consider the large number of published studies as evidence for the certainty and seriousness of harm from PM2.5. However, experience has shown that observational studies with non-randomly-selected subjects tend to overstate the effects that the researchers are hoping to find. As demonstrated in this and other areas of observational epidemiology, residual confounding, publication bias, and model-selection bias all contribute to the creation of apparent risks from contemporary, historically low levels of PM2.5 where no risk is likely to exist in fact. Implementing a technique over and over again doesn’t increase the robustness or certainty of the evidence for harm from PM2.5 when the technique itself is fundamentally unable to perform as advertised.
Randomized, controlled studies of PM and mortality could provide more definitive evidence of PM2.5 risks. Such studies can not, of course, be performed on humans. However, animal studies have failed to provide evidence that PM causes premature death, even at concentrations much higher than ever occur in ambient air. Studies of less adverse PM2.5 effects have been performed with human volunteers, but these studies have provided little or no evidence for harm at contemporary PM2.5 levels.
EPA and CASAC have ignored the evidence against the validity of observational epidemiology as a tool for assessing PM2.5 risks. And they have marshaled evidence selectively so as to create an appearance of greater and more certain risks from PM2.5 than is warranted by the weight of the evidence.
Given that current PM2.5 standards are sufficiently stringent to protect Americans with an adequate margin of safety, EPA should reject its proposed PM2.5 NAAQS rule and leave the current PM2.5 standards in place.
*Correction, April 28, 2006: There is a numerical error here. The actual decline was from 6.9% for 1982-89 down to 1.9% for 1990-98, for a 73% decline in the PM2.5-mortality risk. At this effect size, the PM2.5-mortality relationship would be statistically insignificant.
1. Based on national PM2.5 monitoring data for 2002-2004 downloaded from EPA at