DEFINING AND CHARACTERIZING ENVIRONMENTAL RISKS AND BENEFITS
In theory, it should be possible to engage in rational comparative risk analyses as a means of selecting cost-effective means for the protection of the public's health and our common natural environment. At present, however, the available knowledge base is generally too limited to adequately guide risk-based actions by legislators and/or by governmental agencies to protect and/or improve the environment. What we need is a strategic plan to extend the range and depth of knowledge for risk assessment, taking advantage of the scientific and technical capabilities that are advancing so remarkably in the current era. We also need an effective means of organizing that knowledge, effectively communicating it to appropriate stakeholders, and we need processes for the identification of socially acceptable means of risk-based intervention to prevent, ameliorate, and/or to reverse environmental degradation by more efficient and effective means.
In other words, we must be careful to distinguish between what capabilities we can hope for and expect to be available in the not-too-distant future, and what current tools can provide for us now. We must also recognize that advancement and refinement of our tools for quantitatively determining risks and benefits will not just improve on their own. New research resources will need to be invested to further develop and hone these tools. With appropriate investments in risk assessment research, we can look forward to ever increasing capabilities for more quantitative risk assessments, more definitive comparative risk assessments, more definitive benefit-cost analyses, and more efficient and effective risk management options.
In recent years, as a result of my Chairmanship of various EPA Science Advisory Board (SAB) Committees (Clean Air Scientific Advisory Committee, Human Exposure Committee, Secondary Data Use Committee, Environmental Tobacco Smoke (ETS) Risk Assessment Review Committee, Dioxin Risk Assessment Review Committee), and as my membership on the Steering Committees for the SAB Reports on Future Risk, Reducing Risk, and Beyond the Horizon, my continuing participation in the SAB Advisory Council on Clean Air Act Compliance Analysis (Council) reviews of the Benefits and Costs of the Clean Air Act, and my contributions to the recently completed National Research Council's Report on "Strengthening Science and Peer Review at the EPA", I have become quite familiar with the capabilities and limitations of the predictive models and of environmental and epidemiological data bases available at EPA for risk and benefits assessments. In this regard, I have come to recognize that EPA is heavily dependent on its predictive models for exposure and risk estimation. Unfortunately, many of these models have not yet been fully validated. The adequacy of EPA's models for quantitative risk assessment is discussed in greater detail in the testimony of Dr. Philip Hopke in Panel 3.
This hearing is focused on the capabilities and limitations of current knowledge and technical means of comparative risk assessment for guiding new legislative mandates, societal choices, and individual decisions based on risk avoidance. In my remarks, I will focus on health risks associated with exposures to airborne chemicals and mixtures thereof in our communities.
In order to determine the extent of any health risk existing among the members of the population of concern resulting from the inhalation of airborne chemicals we need to know: 1) the distribution of the concentration of the agent in the air and, for airborne particulate matter (PM), the distribution of particle sizes; and 2) the unit risk factor, i.e., the number of cases and/or the extent of the adverse effects associated with a unit of exposure. For more sophisticated analyses, we may also need to know more about the population of concern, such as the distribution of ages, pre-existing diseases, pre-disposing factors for illness, such as cigarette smoking, dietary deficiencies or excesses, etc.
When basic information on ambient levels and unit risks is available, it is relatively straightforward to compute, tabulate, and compare the risks associated with the different chemicals in our community air. However, based on the experience gained in the Council Review of the Benefits and Costs of the Clean Air Act, such direct comparisons can, in practice, only be made with any quantitative reality for a handful of chemicals, i.e., the so-called criteria pollutants, whose ambient air levels are routinely monitored and for which directly measured human exposure-response relationships have been developed. For hundreds of other airborne chemicals, known collectively as hazardous air pollutants (HAPs), a.k.a. air toxics, there are neither extensive ambient air concentration data nor unit risk factors that do not intentionally err on the side of safety. This disparity has resulted from the different control philosophies built into the Clean Air Act (CAA) and maintained by the EPA as a part of its regulatory strategy. The rationale for the distinction is that criteria pollutants come from numerous and widespread sources, have relatively uniform concentrations across an airshed, require statewide and/or regional air inventories and control strategies for source categories (motor vehicles, space heating, power production, etc.) focused on the attainment of air quality standards (concentration limits) whose attainment provides protection to the public health with an adequate margin of safety. There is also a long history of routine, mostly daily measurements of criteria pollutant concentrations throughout the country.
By contrast, HAPs sources are far fewer in number and are considered to be definable point sources at fixed locations. Downwind concentrations are highly variable, and generally drop rapidly with distance from the source due to dilution into cleaner, background air. The national emission standards for hazardous air pollutants (NESHAPs) are based on technologically-based source controls and are intended to limit facility fenceline air concentrations to those that would not cause an adverse health effect to the (most exposed) individual living at the fenceline. Also, until quite recently, there has been no program for routine measurements of air toxics in our communities.
Most of the unit risk factors for air toxics are based on cancer as the health effect of primary concern. In these studies, and in studies to assess noncancer effects the data are most often derived from controlled exposures in laboratory animals at maximally tolerated levels of exposure. The translation of the results of these studies to unit risk factors relevant to humans exposed at much, much lower levels in the environment is inherently uncertain, and is approached conservatively, following the model pioneered for food and drug safety beginning in the 1930s by the Food and Drug Administration (FDA). The resulting unit risk factors are generally based on an assumption of no threshold and a linear extrapolation to zero risk at zero dose. They are generally described in terms of being 95% upperbound confidence limits, but this descriptor is undoubtedly conservative in itself.
When these conservative unit risk factors are used for the prediction of the consequences of human exposures, they are multiplied by estimates of predicted ambient air concentrations which are, themselves, in the almost universal absence of air concentration measurements, almost certainly upper bound estimates from pollutant dispersion models that apply to the most highly exposed individuals in the community.
The resulting estimates of health risk are therefore highly conservative upper bound levels. Thus, they are inherently incompatible with population impacts estimated for the more widely dispersed criteria pollutants. The margins of safety for criteria pollutants are generally less than a factor of two, rather than the multiple orders of magnitude of safety factors built into the risk assessments for air toxics.
The same considerations discussed above, i.e. the limitation of available knowledge for determining realistic unit risk levels has also made it virtually impossible for EPA to meet its Congressional mandate to determine residual risks after the imposition of technology-based controls of air toxics, as discussed in greater detail in the testimony provided to this Hearing by Dr. Philip Hopke in Panel # 3.
The highly conservative nature of unit risk factors for air toxics was well illustrated by a calculation made during work done for EPA during the preparation of the Congressionally mandated report on the Benefits and Costs of the Clean Air Act: 1970-1990. It was determined that the imposition of the vinyl chloride NESHAP had prevented 6,000 cases of cancer. Vinyl chloride is a known human carcinogen that produced a very rare tumor (angiosarcoma of the liver) in highly exposed vinyl chloride production workers. The handful of cancers observed among these workers was not large in relation to overall cancer incidence, but this particular tumor was such a rare one that even the first few cases that were observed among a group of vinyl chloride production workers were sufficient to establish a causal relation. Since the calculated cancer incidence reduction was considerably larger than the historic incidence level for this cancer, it was obvious that the benefit claimed for the imposition of the NESHAP was grossly exaggerated.
The lack of any alternative quantitative approach to the quantitative estimation of health effects due to exposure to air toxics has left EPA with no viable option for the realistic estimation of population impacts. With prodding from the SAB Council, the Agency has recognized the need to develop one. That effort is now underway, through EPA and SAB sponsorship of a first Workshop (June 22 and 23, 2000) in a series designed to address the issue directly. Extension of this initiative would lead to the development of a capability to produce more unbiased predictions of the health consequences of HAPs exposures for benefits assessments.
In the meantime, EPA needs to undertake a public education program about the essential nature of its widely distributed and commonly used unit risk factors. This is especially urgent in view of its recent initiative to support a nationwide network of routine air quality monitoring stations for a large number of representative air toxics. Pilot studies have already demonstrated the multiplication of measured levels times the current unit risk factors suggest that urban dwellers are at lifetime risks of excess cancer greater than one in a thousand. Exaggeration of risks pertaining to the general public could produce a considerable problem for EPA in its communication to the public, and could lead to a loss in its credibility.
Comparative risk analysis, as currently practiced, has other inherent limitations as well. Even when we can reasonably and reliably estimate the exposure-related numbers of cases of premature mortality, hospital admissions, other uses of medical, clinical and pharmaceutical drug resources, lost time from work or school, reduced physiological and functional capacities, we face daunting societal equity and valuation challenges in intercomparing numbers of incident cases of quite variable clinical severity and psychological impacts. For carcinogenic agents, it has become customary to expect regulations to be effective in limiting the risks of lifetime exposures to no more than one-in-ten thousand and often to less than one-in-a-million. For less dreaded diseases that also reduce lifespan, such as chronic obstructive pulmonary disease and heart attack, which also are exacerabated by air pollutant exposures, a much higher risk level has been considered acceptable by regulators and the public. By contrast, economists do not make such a drastic distinction. EPA's recent White Paper on the economic valuation of cancer mortality concluded that the economic literature did not provide a basis for a greater benefit for a prevented cancer death than for other causes of premature deaths.
The National Research Council (NRC) committee that issued its report on "Strengthening Science and Peer Review at EPA" was well aware of the current limitations of comparative risk assessment when it concluded that:
"Scientific knowledge and technical information are essential for determining which environmental problems pose important risks to human health, ecosystems, the quality of life, and the economy. We need scientific information to avoid wastefully targeting inconsequential risks while ignoring greater risks. We need such information to reduce uncertainties in environmental decision making and to help develop cost-effective strategies to reduce risk. We need science to help identify emerging and future environmental problems and to prepare for the inevitable surprises."
The quotation above provides a good part of the background that led to the key recommendations of the NRC Report regarding the management and use of science in regulatory programs, and the need for and Agency management of its own research program to fill key gaps in our current abilities to quantify risks. Focus on this need should be a priority for the recommended position of Deputy Administrator for Science in EPA. This individual should have the background and judgment essential to ensure that current risk-related knowledge is appropriately used to develop, describe, and guide scientific input into regulations, and to ensure communication of the knowledge gained by the regulatory programs in terms of further research needs for risk assessment and risk management. The Deputy Administrator for Science could also provide oversight for EPA's new Office of Information in regard to facilitating more data entry into and wider access to and usage of EPA's environmental monitoring data sets that are now seldom used for secondary data analysis and/or model validation.
The NRC Report also concluded that research on risk assessment and risk management was not only needed, but needed to be conducted by EPA, since no other federal agency had the mandate, need, or desire to conduct such research.
Finally, it should be recognized that research on risk assessment and risk management needs to be a long-term core component of EPA's research program. Core research needs stability, a feature which has not been a hallmark of EPA's Office of Research and Development (ORD). Tenure for a Presidentially selected and Senate confirmed Assistant Administrator (AA) for ORD has been three years or less, and Acting AAs for ORD have occupied the position for about half of the whole history of EPA. Thus, the NRC Report recommended that the position be changed to a six-year term-appointment, with the AA selected for expertise in both science and research management. This change would help to ensure the primacy of a longer term view of research goals focused on EPA's unique role as a regulatory agency that relies strongly on sound science to guide the formulation of its standards, guidelines, and cost-effective risk management.
The differences in EPA's current abilities to make estimates of health risks for air toxics on the one hand and estimates of benefits resulting from its successes in source controls on the other, while notable and unfortunate, are remediable, and the research needed to overcome the current deficiencies should be given a high priority. The ORD has come a long way in recent years in terms of its development and updates on its strategic plan, its inventory of science activities and capabilities through the Agency, its closer coordination with research programs in NIH, NSF, and CDC, and its shift of resources toward an extramural grant program in which EPA's research needs are met, in part, through individual investigator-initiated proposals that address critical information needs identified in Requests for Applications. It will also soon occupy new state-of-the-art research facilities in Research Triangle Park, NC that will enhance its capabilities.
In summary, our current abilities to determine residual risks of air toxics and to compare risks quantitatively are quite limited by key gaps in knowledge, and by reliance on unvalidated predictive models for exposure and for dose-response. A major part of the problem is the existence of two very different cultures of risk assessment: 1) for carcinogens; and 2) for other toxicants. Carcinogen risk assessments seldom have been based on relevant data on either low-dose exposure on human exposure-response data at concentrations anywhere near ambient levels. They require high-dose to low-dose extrapolations and generally animal-to-human extrapolations as well, using unvalidated predictive models. In the face of such a high degree of uncertainty in the output of the models, conservative assumptions are used to ensure that potency and exposures are not underestimated. Thus, yields of risk estimates are almost always far higher than the real risks. Such risk estimates cannot be fairly compared to the risks associated with criteria air pollutants, which are determined largely from the product of measured air pollutant concentrations and measured responses among humans exposed to either ambient air or to controlled exposures in chambers. Fair comparisons can only be done within the separate categories of pollutants.
Comparative risk assessment is an idea whose time is coming, and if EPA is provided with appropriate research resources to harness the new technical approaches and sophisticated research tools now emerging to fill in key knowledge gaps, it can make comparative risk assessment more useful and feasible in the not-too-distant future. If the recommendations in the NRC "Strengthening Science at EPA" report are adopted, the prospects for such advances would be greatly improved. In the meantime, the resources now dedicated to performing comparative risk assessments would be more productively employed if redirected to improving the technology for quantitative risk assessment and for filling key knowledge gaps that have been identified in the analyses already performed.
Dr. Hopke, in his testimony in the next panel will address the major knowledge gaps limiting EPA's ability to perform the residual risk assessments mandated for HAPs in the 1990 Clean Air Act Amendments, even for an industrial sector like secondary lead smelters that is relatively data-rich. In my view, we should celebrate the success of the application of the best available technology approach in greatly reducing emissions and ambient concentrations of air toxics and limit the use of quantitative risk analyses of residual risks to the screening out of deminimus risks.
Finally, I encourage the Congress to implement the legislative changes needed for the creation of the new position of Deputy Administrator for Science in EPA and for transforming the position of Assistant Administrator for Research and Development in EPA to a six-year term appointment. These changes will help ensure institutional stability and a more long-term framework for core research. I also encourage Congress to consider explicitly giving EPA a mission statement that includes the performance of a long-term research program as a means of enhancing its capabilities for effective and efficient stewardship of its environmental responsibilities.
In closing, I want to thank the Committee for inviting me to testify on these important issues related to scientific aspects of environmental risks and on opportunities to improve the practice and utility of risk assessment and risk management.