Health and Air Quality—In Search of Missing Links

The unpleasantries associated with air pollution are many, ranging from smog and obvious soot on buildings, vehicles, and clothing to the more subtle damages to crops and ornamental plants and shrubs for which a variety of pollutants is suspect. As far as federal clean air legislation is concerned, however, one purpose is paramount—protection against pollution-related sickness and death.

Section 109 of the 1970 Amendments to the Clean Air Act mandates uniform, national air quality standards that allow for "an adequate margin of safety . . . requisite to protect the public health." This provision in particular and a general concern about air pollution and health stemmed from two factors: first, several very serious air pollution inversions that resulted in immediate sickness and death; and second, a less well-documented suspicion that much lower levels of pollution might be harmful if endured over long periods of time. Indeed, the search for the latter kinds of cause-and-effect relationships between environmental conditions and human health is a tenuous science: study results often conflict—or appear to—and even the most respected experts sometimes voice diametrically opposed views. The legislators who drafted the Clean Air Act recognized this and provided in Section 103 for a program of research designed to eliminate as many uncertainties as possible.

Identifying health effects

There are three primary means of identifying any adverse health effects associated with air pollution. First are laboratory animal experiments typified by the "saccharin-and-rats" studies: lab animals are exposed to a broad spectrum of air pollutants, with dosages varying widely both in terms of intensity and duration, and health effects are observed. Obvious problems arise if we wish to draw conclusions about human reactions to pollution from such studies. First, the animals might react differently to air pollution if the exposure were to take place in a more natural (less controlled) setting. Second, even if we could be sure that we understood animal reactions perfectly, it would require a major leap of faith to extrapolate directly to human beings. This is not necessarily to doubt the value of such laboratory research; indeed, it is important. But it must be interpreted and used very carefully.

The second major method of study—clinical research—bears certain resemblances to animal experiments. Here, individuals volunteering for such studies are administered controlled doses of air pollutants (well below levels considered to be lethal or otherwise harmful), and the difference between "before" and "after" health status is then taken to be some indicator of the effects of the dose on whatever bodily functions are under study. Naturally, the results from such human experiments are intuitively more appealing than those obtained from animal tests.

Clinical studies are far from conclusive, however. Again, one is confronted with the problem of artificial dosage. That is, an individual's exposure to air pollution in the experimental setting is certain to be quite different from that he or she encounters in everyday activities where factors held constant in the clinical environment in fact are anything but constant. Then, too, clinical studies typically are plagued by small numbers of subjects, thus rendering nearly nil the possibility of obtaining robust statistical estimates. Finally, and most important, clinical studies do not allow a consideration of the possible long-term or chronic effects of air pollution on health.

The third means for studying the effects of pollution on human health is epidemiological research. The major difference between epidemiological and clinical work is that the "subjects" in epidemiological research have not been exposed to air pollution in a controlled experimental setting. Rather, their exposure comes from the "real world," and individual variation in health status over time or across geographic areas is examined for some link to the different pollutant concentrations to which individuals actually are exposed. A typical epidemiological study examines two or more distinct groups of individuals, where the grouping is based on different exposures to the environmental factors being analyzed. The study design attempts to keep all other characteristics of the groups similar so that other possibly influential factors can be assumed constant. Then, statistically significant differences in health status are attributed to the differentials in air pollution or other environmental factors.

Epidemiological strengths

There are several strengths and weaknesses to the epidemiological approach. On the positive side, epidemiological studies can, but do not necessarily, escape the three problems enumerated above that plague clinical analysis: artificial dose, small numbers, and inability to identify chronic effects. In epidemiological studies, the "dose" measurements come from actual ambient air pollution monitors, which should reflect exposures typical for the individual in his or her normal sphere of activity. Next, epidemiological study samples normally are larger than clinical study groups, thus circumventing some of the statistical problems inherent in clinical work. Finally, given sufficient consideration in studying design for time lags, the possible long-term or chronic impacts of pollution on individual health status are measurable, at least theoretically, in an epidemiological framework.

. . . And weaknesses

Yet, the drawbacks in epidemiological research are by no means trivial. Data sufficient to characterize precisely any one individual's exposure to environmental contaminants—in air, water, or land—simply are not available for any study framework yet devised. Even if current air pollution readings were a good proxy for historic concentrations, the difficult problem would remain of knowing which air pollution monitor in a particular area to use to characterize exposure. After all, concentrations of sulfur dioxide, ozone, particulates, and other pollutants vary widely within one metropolitan area—thus, use of an areawide average will be misleading. Even using data from a monitor near an individual's home is not perfect: significant variations in the concentrations of many airborne substances can occur within a few meters of the pollution monitors. Moreover, important recent work indicates that indoor concentrations of some air pollutants might be a potentially greater threat to health than even the highest levels of outdoor air pollution a typical individual might expect to encounter.

Also, much epidemiological work is based on self-reported rather than medically diagnosed health status. Although this may introduce some error into dose-response estimation, the direction of the error is not clear. Some evidence suggests individuals themselves identify only about one-third of the adverse health conditions that a doctor diagnoses in a comprehensive health exam. Although perceived health status is important, actual health status is certainly no less so. To blindly equate the two can result in biased results.

A third major problem with much epidemiological work is that studies purporting to hold constant factors other than pollution rarely in fact do so. Grouping to eliminate the effects of these other factors also often eliminates much potentially relevant information. Unless statistical regression techniques are used, it is almost impossible to avoid situations where either too many individuals are eliminated from study because of "atypical" characteristics or too much intergroup variation is allowed to make plausible assumptions about lack of variation in the other factors.

Setting air quality standards

Because of these and other drawbacks, some close observers of the process by which air quality standards are set feel that epidemiological research is given relatively little weight as evidence of pollution-related health effects. Given the difficulties and uncertainties characterizing epidemiological results, some skepticism certainly is understandable. Nevertheless, for reasons described above, no study—clinical or otherwise—is without its problems: one might be concerned with over-reliance on any one kind of study in identifying health effects associated with air pollution.

Whether to give weight to one type of study or another may not seem to be a burning policy matter. But it is. In fact, the U.S. Environmental Protection Agency is preparing to propose a revision in the national air quality standard for particulates. The precise level where this standard is set hinges critically on the acceptance or rejection of several major epidemiological studies. Because of the political sensitivity of the particulate standard (the electric utility, steel, and other major industries are directly affected by it), the outcome of this standard is important. It is sure to serve as a barometer of future sentiment toward relying on epidemiological results in the establishment or revision of air quality standards.

Future research

Regardless of the disposition of the particulate standard, if the concern about the health effects of air pollution is real—and there is no reason to doubt that it is—then research must continue. Improvements in both data and methodology are essential to the progress of research in this field, and work currently under way at RFF is attempting to meet problems on both these fronts. Active research continues on analytical methods to help deal with the knotty statistical problems that arise in all epidemiological research. Also, a large data base consisting of information on 110,000 individuals is being assembled for use in air pollution studies and other kinds of epidemiological research.

The refinements being built into the data base are considerable. Air pollution exposures will be estimated more finely—in both the temporal and spatial dimensions—than is the custom in studies using similar methodologies. For example, rather than use an average of all monitors in the metropolitan area to characterize an individual's exposure, individuals will be matched with the monitor(s) nearest their dwelling. Rather than use annual average pollution data, readings during the specific period for which health information is collected will be used. Both are noteworthy improvements. In addition, detailed information on pollen and weather is included, as are several measures of health status, socioeconomic information, and some occupational characteristics. A crude measure of exposure to indoor air pollutants also will be included, as will the all-important information on individual smoking habits. Further, data on residential mobility, using a sample of individuals who have always lived in the same place, will make it possible to explore any chronic health effects associated with air pollution.

Admittedly, these improvements are not a panacea for all problems. Data on the true doses of air pollution that people get still are not available, and we have only very rough measures for indoor air pollution concentrations. Moreover, no information is available on the measures individuals take to defend themselves against air pollution (the installation of air filters, for example), or to ameliorate any ill effects that do result. Finally, we lack diagnostic confirmation of many of the adverse health effects that individuals report.

In spite of these shortcomings, environmental epidemiology is improving rapidly. Data are now available for individuals where areawide average measures of health status were used before. Confounding factors often ignored in the past—smoking, diet, exposure to indoor pollutants—now are routinely controlled for. Environmental factors other than air pollution are being taken into account. And more powerful, and appropriate, statistical techniques are being used to analyze data. Further improvements are possible, but it is clear that with time there will be available more and better information linking the health of men and women to their environment.