National Academies Press: OpenBook

Air Pollution, the Automobile, and Public Health (1988)

Chapter: Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations

« Previous: Part III: Biological Effects
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 389
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 390
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 391
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 392
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 393
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 394
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 395
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 396
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 397
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 398
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 399
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 400
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 401
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 402
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 403
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 404
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 405
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 406
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 407
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 408
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 409
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 410
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 411
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 412
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 413
Suggested Citation:"Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations." National Research Council. 1988. Air Pollution, the Automobile, and Public Health. Washington, DC: The National Academies Press. doi: 10.17226/1033.
×
Page 414

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Epidemiologic Studies of Effects of Oxidant Exposure on Human Populations EDDY A. BRESNITZ KATHLEEN M. REST The Medical College of Pennsylvania Role of Epidemiology in Air Pollution Research / 390 Prior and Ongoing Studies Designs, Findings, and Problems / 390 Cross-Sectional Studies of Pulmonary Function and Symptoms / 391 Prospective Cohort Studies / 399 Retrospective Studies / 400 Generic Issues / 400 Exposure Assessment / 400 Effects Assessment / 401 Specific Issues and Study Approaches / 404 Chronic Obstructive Pulmonary Disease / 404 Asthma/Bronchial Hyperresponders / 405 Respiratory Infection / 407 Attributable Risk / 408 Summary / 409 Summary of Research Recommendations / 409 Specific Research Issues / 409 Generic Issues / 411 Air Pollution, the Automobile, and Public Health. @) 1988 by the Health Effects Institute. National Academy Press, Washington, D.C. 389

390 Epidemiologic Studies of Oxidant Exposure Role of Epidemiology in Air Pollution Research Epidemiology is the primary research dis- cipline that allows investigators to examine the long-term effects of air pollution on public health. Controlled human studies in exposure chambers generate useful hypoth- eses and allow maximum quantification of dose/response relationships, but they can only address acute effects and short-term changes in functional parameters. The as- sessment of long-term exposures and chronic health effects in humans must nec- essarily fall to the epidemiologist. In epide- miologic investigations, causal relation- ships between exposure and effect are usually plausibly inferred by the strength of the association, the consistency of data, the specificity of results, the temporality of observations, the demonstration of a bio- logical gradient, and the plausibility and coherence of results (Hill 1965~. In air pollution research, animal, cham- ber, and epidemiologic studies have served to investigate a variety of health outcomes. Mortality studies have often followed on the heels of major air pollution episodes. Morbidity studies have looked at respira- tory as well as nonrespiratory end points. The former include acute effects such as asthma and infection, chronic effects such as chronic obstructive pulmonary disease (COPD), and long-term effects such as lung cancer and accelerated decline in lung function. Research on nonrespiratory ef- fects has usually focused on nonoxidant exposures (for example, on lead, organic solvents, and carbon monoxide), and has investigated neurotoxic effects, heart dis- ease, and leukemia. When epidemiology is used to study these outcomes in air pollution research, several technical and methodological issues arise. These include selection of appropriate study design and study population, assess- ment of exposure, definition and assess- ment of adverse health effects, control of bias and confounding variables, and analy- sis of data. Many studies done to date have been flawed by their method of addressing these difficult problems, leaving their re- sults and conclusions open to question. Yet the weight of the aggregate evidence sug- gests that air pollutants do cause adverse health effects at certain levels, and that further epidemiologic studies are needed to quantify dose/response and to assess the consequences of long-term exposure to air pollutants at low levels. The use of epidemiology in air pollution research is examined in this chapter. The focus of the chapter is on photochemical oxidants, mainly ozone (03) and nitrogen dioxide (NO2), and their effects on the respiratory system, exclusive of lung can- cer. A review of selected studies illustrates how investigators have addressed impor- tant technical and methodological issues, and suggests how such issues might be better addressed in the future. Important scientific and methodological knowledge gaps are then identified, and several approaches for closing these gaps are recommended. Prior and Ongoing Studies Designs, Findings, and Problems Researchers have made considerably more progress in studying the acute effects of exposure to short-term, high levels of O3 and NO2 than in studying the effects of long-term, low-level oxidant exposure, or repeated episodic oxidant exposure at peak levels. The contribution of these exposure patterns to such respiratory diseases as COPD, asthma, and pulmonary infection remains a research gap. The effect of oxi- dants on the normal rate of decline in pulmonary function with age is also an unresolved question. The studies required to fill these gaps are difficult to accomplish because of several factors: the generally low level of ambient (outdoor) exposures in the United States; the relative infrequency of some chronic respiratory diseases; the multifactorial na- ture of these chronic respiratory diseases; the lack of sensitivity and specificity of some of the tools used to detect physiolog- ical dysfunction or disease; the poor char- acterization of the actual composition of pollutants over time; the difficulty in esti

Eddy A. Bresnitz and Kathleen M. Rest 391 mating biologically effective dose; and the ethical problems inherent in conducting controlled studies of long-term exposures. Epidemiologic observational studies to date have generally suffered from one or more of these problems, precluding definitive, quantitative conclusions about causal rela- tionships between exposure to oxidants and adverse health effects. In table 1, the most commonly used types of epidemiologic study design are described. The cross-sectional study has been the most frequently used observa- tional design in epidemiologic studies of air pollution. Studies with this design deter- mine the prevalence of particular health outcomes in a population at one or more points in time and correlate them with some concurrent measure of exposure. The risk for disease development (incidence) and the temporal sequence of exposure and disease cannot always be determined in this type of study. Although cross-sectional studies have been useful for generating hypotheses, comparing health outcomes in communities exposed to high and low lev- els of air pollution, and studying the acute effects of short-term exposures, they have not been very helpful in assessing dose/re- sponse effects of long-term exposure. A cross-sectional study may be repeated in populations over time to assess trends in the prevalence of specific outcomes of in- terest. A study done this way is called a secular trend or time-series analysis. In air pollution research, the outcomes under study may be correlated with serial mea- surements of ambient air pollutants. Paral- lel changes in the prevalence of symptoms or mean pulmonary function and average air pollutant levels would suggest an association between the exposure and the outcomes. A closer examination of several cross-sec- tional studies that attempt to assess the rela- tionship between exposure to oxidants and respiratory morbidity illustrates the prob- lems and limitations of this type of study. Cross-Sectional Studies of Pulmonary Function and Symptoms Healthy Populations. Shy and coworkers (1970a) compared the ventilatory function of 987 second-grade children who lived in Chattanooga, Tennessee, in four geo- graphic areas that varied in their average 24-hr levels of ambient NO2 and particu- lates (one high-NO2 area, one high-partic- ulate area, and two control areas). Mea- surements were taken from stationary air pollutant monitoring sites during the 65- day study period. Assessment of mean weekly height- adjusted forced expiratory volume in the first 0.75 sec (FEVo 75) of the spirogram, averaged over each month of the study for children of the same sex, showed that the FEVo 75 was statistically significantly lower in the high-NO2 area compared to the two control areas in both months of the study. However, the differences were very small and clinically not very significant. This study has several methodological problems. First, the investigators did not adjust the results for parental smoking and indoor sources of NO2 which may have been associated both with decreased lung function in the children (Speizer et al. 1980) and levels of outdoor NO2. That is, paren- tal smoking may have confounded this association. Second, the levels of total sus- pended particulates (TSP) and sulfates (S04) were highest in the high-NO2 area. These other air pollutants may interact with NO2 or confound the relationship between lung function and the oxidant levels. Finally, the technique used to mea- sure NO2 was subsequently shown to be invalid. Thus, technical as well as method- ological problems affected the results of the study. In the second phase of the study, Shy and coworkers (1970b) investigated the trend in the incidence of acute respiratory illness among all families in the study having a child in the second grade. A total of 4,043 study subjects, comprising 871 families of the children who had participated in the ventilatory testing portion of the study, agreed to participate in this phase. Temporal variations of self-reported res- piratory illness rates were similar in each study area, and rates among smokers did not differ from rates in nonsmokers. Abso- lute illness rates over the entire study pe- riod were consistently and significantly

392 Epidemiologic Studies of Oxidant Exposure Table 1. Epidemiologic Research Designs Study Type Description Advantages Disadvantages Cross-sectional · Examines presence or (also called absence of exposure and prevalence effects at same point in study) time. · Sometimes repeated observations will be made over time and combined to compare trends in the population. · Is essentially the first phase of any cohort study. Cohort · Individuals (exposed and unexposed) are selected for observation and followed over time. · Study proceeds from suspected cause to effect (i.e., subjects selected on basis of exposure). · Data may be collected prospectively (i.e., outcome and exposure data unknown at beginning of study but collected as study proceeds). OR · Data may be collected retrospectively (i.e., exposures and outcome have already occurred at beginning of study and data are historical in nature). Case-control · Individuals with particular diseases or conditions (cases) are selected and compared to individuals without the disease or condition (controls). They are compared with respect to the exposure of interest. · Study proceeds from effect to cause (i.e., subjects selected on basis of disease). · Exposure data are collected retrospectively. · Relatively easy and quick to · Does not permit cause/effect perform. inferences. · Especially suited for · Individuals who studying subclinical health demonstrate an effect or effects for which records do have a disease (prevalent not exist, and for studying cases) may not be representative of all individuals who have same effect/disease. · Temporal relationship between exposure and effect may be difficult to ascertain. effects that can be quantitated (e.g., pulmonary function) and that can vary over time. · Also useful for studying relatively frequent diseases that have long duration. · Generates useful etiologic hypotheses for analytical research. · Excellent method for studying effects of rare exposures. · Permits observation of multiple effects. · Quality control in data collection is more manageable. · Yields incidence rates as well as relative risk. · Permits observation of entire exposure to disease state continuum. · Excellent method for studying rare diseases with long latency periods. · Relatively quick and Inexpensive. · Requires small number of subjects. · Existing records may be available for use. · Permits investigation of multiple causal factors. · Requires long follow-up period. · Very costly. · Requires large number of subjects. · Problems with attrition and loss to follow-up. · Changes in exposure, lifestyle factors during course of study may make findings irrelevant or difficult to interpret. · Exposure data may not exist, may be inadequate, or may be available only through recall. · Validation of exposure data may be difficult or impossible. · Selection of proper control group may be difficult. · Control of extraneous variables may be incomplete. · Incidence rates cannot be ascertained. higher in all family segments in the high- NO2 areas compared to the two control areas, especially during the interinfluenza periods. The differences in illness rates be- tween the high-particulate area versus the two control areas were smaller and less consistent. There was no consistent differ- ence in the severity of illnesses among the

Eddy A. Bresnitz and Kathleen M. Rest different areas, nor was there a dose/re- sponse effect on the rates of illnesses ob- served in the subjects whose children at- tended the three schools in the high-NO2 area. In addition to the study design weak- nesses discussed earlier, the major problem in this part of the study was the failure to relate symptoms with incidence of true infection. Self-reporting of disease was ac- cepted as evidence of respiratory illness. Although the investigators assessed the se- verity of the illnesses through follow-up telephone interviews, there was no sero- epidemiologic evidence of infection. More- over, the investigators were unable to dis- criminate between the effects of short-term peak levels and long-term averages in pol- lutant exposures, nor did they adjust for differences in indoor exposure to NO2. Also, the absolute differences in rates were small despite being statistically significant. The authors concluded that exposure to higher levels of ambient NOR and to par- ticulates (to a lesser degree) increased the incidence of acute respiratory illnesses. It is not possible to conclude that infections was the cause of these illnesses in this secular trend analysis. Cohen et al. (1972) studied 441 non- smoking, middle-aged Seventh Day Ad- ventists living in either a high- or a low- oxidant pollution area in- California. Air pollutant concentrations were measured at four sampling stations in each area; only total oxidants were measured. The mean percent daily maximum hourly concentra- tion of oxidants in the two areas varied, but the annual mean value of oxidants was identical in both areas. Of the study sub- jects, 76 percent completed spirometry and maximum expiratory flow-volume curves and 97 percent completed a Medical Re- search Council questionnaire assessing symptoms. There were no differences in age, socioeconomic status, occupation, eth- nic background, and length of residence in the two different areas. Maximum hourly averages of exposure to pollutants other than oxidants between the two groups were also the same. There was no signifi- cant difference from expected values in either respiratory symptoms or pulmonary 393 function adjusted for age, gender, and height in either group. This cross-sectional study did well in adjusting for some variables that may con- found or bias any apparent association be- tween the level of ambient oxidants and the health outcome of interest. Unfortunately, the differences in the pulmonary function measurements were small and the likeli- hood of detecting statistically significant differences in outcomes was correspond- ingly diminished. Moreover, the overall prevalence of chronic bronchitis was less than two percent. Limiting the study to nonsmokers, however, strengthened the conclusions about lack of an effect on pre- viously healthy individuals. It did not ad- dress the question of a possible increased risk of morbidity secondary to low-level, long-term exposures among individuals with or without preexisting respiratory dis ease. Other, more recent cross-sectional stud- ies have investigated the effects of short- term, low levels of O3 exposure on differ- ences in pulmonary function in children (Lippmann et al. 1983) and exercising adults (Selwyn et al. 1985~. These studies found a significant correlation between one or more measures of pulmonary function derived from spirometry and some mea- sure of maximum O3 concentration. There was a significant negative correla- tion between a single day's peak 1-fur O3 concentration and the peak expiratory flow rate in 39 healthy children between the ages of 7 and 13 (Lippmann et al. 1983~. The O3 level was less than 0.1 parts per million (ppm) on most days (28 of 32) of the study. The magnitude of the relationship, how- ever, may have been underestimated be- cause of a selection bias; that is, children who had positive histories of respiratory illness and did not participate in spirometric tests on four or more days were not in- cluded in the analysis. A study of spirometry performed in 24 healthy adult runners before and after a 3-mile run correlated changes in airflow during 28 separate runs with backside mea- surements of a time-weighted average of maximum O3 concentrations over 15-min periods (Selwyn et al. 1985~. Increasing O3

394 Epidemiologic Studies of Oxidant Exposure concentrations were statistically signifi- cantly correlated with greater pre/post run differences in the volume of air expired during the first second (FEY, ~ and the forced midexpiratory flow (FEF2~75%) Although these findings suggest a direct relationship between increasing biological dose and pulmonary function, this associa- tion did not remain statistically significant after adjustment for relative humidity. The University of California at Los Angeles (UCLA) population study (Deters et al. 1979, 1981, 1982; Rokaw et al. 1980) is a longitudinal investigation of COPD that examines the relationship between long-term exposure to photochemical oxi- dants and the prevalence of pulmonary function decrements and respiratory symp- toms. The initial survey in this (or any) longitudinal study can be analyzed as a cross-sectional study. One phase of the study (Deters et al. 1981) compared 3,192 white adult residents of Lancaster, California, a low-pollution area, to 2,369 residents of Glendora, Cali- fornia, a high-oxidant area. The age, gen- der, race, and income distributions were similar in the two cities. Daily maximum hourly concentrations of specific air pollu- tants were measured at stationary air mon- itoring stations. The air monitoring station used to estimate exposure in Glendora was three times further from the center of the community than the one used in Lancaster. Pulmonary function tests included spi- rometry and the single-breath nitrogen test. The National Heart, Lung, and Blood Institute respiratory questionnaire was used to assess the prevalence of symptoms. Test- ing in the two areas was separated by a three-year interval. The results were reported by smoking status and were age-adjusted to the 1970 U.S. population census. The prevalence of cough, sputum production, and wheezing was higher in the high-pollution area com- pared to the low-pollution area among never smokers and smokers alike, although the differences were smaller among smok- ers. Although many of the differences were statistically significant, the absolute differ- ences were small. The mean percents predicted for the forced vital capacity (FVC), the FEY,, and the FEF25_7s% were minimally lower in Glendora men, but there were no di~er- ences between women in the two areas. Differences in mean values were greater between smoking strata in both areas. The prevalence of study subjects whose FIVE and FVC were less than 50 percent pre- dicted was higher in Glendora residents, in male and female smokers and nonsmokers. Tests primarily associated with small air- way function showed little or no differ- ences between the study subjects in two areas. This study has several limitations. First, the measurement of exposure most likely underestimated the true exposure in the high-oxidant area because of the distance of the monitoring site from the community. Second, pulmonary function testing and . . . . . . questionnaire admlnlstratlon were not per- formed concurrently in the two areas be- cause of financial constraints. Changes in smoking habits could have affected per- formance of pulmonary function tests and thus the study's findings. However, the investigators stratified and compared the study populations on smoking habit. Third, questionnaire responses were not tested for reliability or validity. Fourth, there was no adjustment for potential con- founders such as occupation and indoor air pollutants, or for effect modifiers such as daily activity. Finally, all levels of pollut- ants were elevated in the Glendora area. Therefore, it is not possible to specify the pollutants) that best explains the differ- ences between the two groups, even if those differences could be attributed solely to air pollutants and not to unanalyzed confounders. The authors' conclusion that long-term exposure to high concentrations of photochemical oxidants, NO2, and SO4 results in respiratory impairment is tem- pered by these limitations. Confounders and Elect Modifiers. The previous discussion refers to several poten- tial study confounders such as parental smoking and levels of other pollutants. A confounder is an extraneous variable that is a risk factor for the disease or symptoms being studied and is associated with the exposure of interest, but is not a conse

Eddy A. Bresnitz and Kathleen M. Rest 395 quence of that exposure. The effect of a confounding variable may be to mask the underlying association or to explain par- tially (or wholly) the apparent effects of the exposure of interest. For example, the prevalence of smokers in a given geo- graphic area may vary with the degree of air pollution in that area (for example, more blue-collar workers in an industrial area). In this case, smoking is a confounder; it is a risk factor- for COPD, and it is associated with the levels of air pollution. There are two general classes of potential confounders that must be considered in air pollution research: environmental con- founders such as other air pollutants, occu- pational exposures, smoking, and meteoro- logical variables; and personal confounders such as allergies and respiratory infections. The validity of any study is affected by the appropriate control of confounders, either in the design phase or the analysis phase of the study. The investigators in the Chattanooga studies did not consider the effect of paren- tal smoking and indoor sources of NO2 on pulmonary function. This was a serious . . . . . Omission In view ot recent epic ence sug . . gest1ng an association between exposure to sidestream smoke as well as to NO2 from gas cooking and respiratory symptoms or impaired pulmonary function (Keller et al. 1979a,b; Tager et al. 1979; Speizer et al. 1980; Comstock et al. 1981; Ware et al. 1984). An effect modifier is a variable that does not, by itself, cause the effect under study but modifies the effect of the risk factor~s) under study. Use of medications, socioeco- nomic status, and activity levels may affect either the extent of exposure or the expres- sion of dysfunction. Failure to account for these variables may lead to underestimation of exposure or disease or both. Although confounders, effect modifiers, and other potential sources of bias must be considered in all epidemiologic studies, specific variables are not uniformly impor- tant across all studies. For example, smok- ing and socioeconomic status may be im- portant confounders or effect modifiers in prospective or retrospective cohort studies of the relationship between oxidant expo sure and COPD, but they may be essen- tially immaterial in cross-sectional studies that correlate oxidant air pollution data with hospital admissions (Bates and Sizto 1983~. Sensitive Populations. The impact of am- bient oxidant exposure on respiratory mor- bidity in sensitive populations is potentially greater than the impact on healthy popula- tions. High-risk groups include people with one or more of the following: (1) specific diseases such as asthma, COPD, bron- chitis, allergies, and sensitive airways; (2) specific exposures such as cigarette smoke and toxic fumes; and (3) biological factors such as age and.~-1-proteinase inhibitor de- flclency. Several cross-sectional studies of asth- matics have evaluated the association of oxidant exposure and the frequency and severity of asthma attacks. Table 2 summa- rizes those studies. Many of these studies suffer from the same methodological prob- lems encountered in some of the previously cited studies on healthy populations. The studies are especially difficult to compare in view of the variability or lack of definition for an attack of bronchospasm. (See also Bromberg, this volume. ~ However, the study by Holguin et al. (1985) deserves special comment in view of its overall excellence. Holguin and coworkers studied 42 med- ically stable, nonsmoking, well-defined asthmatics without other pulmonary dis- eases in Houston during periods of high photochemical oxidant levels in 1981. Sub- jects maintained 12-hr logs, twice daily recording symptoms, location, activity, and use of medications. The maximum value of self-administered peak flow deter- minations was recorded for each 12-hr shift. Quality control was maintained by pre- study instruction of the participants and weekly review of the daily diaries and cali- bration of the peak-flow meters by trained technicians. Exposures for 03, NO2, pollen, temper- ature, and relative humidity were deter- mined by air monitoring stations located less than 2.5 miles from all study subjects. An hourly exposure estimate was calcu

396 Epidemiologic Studies of Oxidant Exposure Table 2. Studies of Aggravation of Existing Asthma by Photochemical Oxidant Pollution Av. Max. Conc. Range (ppm) Pollutant Study Description Results and Comments Reference 0.13a Oxidants 0.01-0.37 O3 Not reported O3 0.0000.235 O3 ( Table continued next page.) Daily records of times of onset and severity of asthma attacks of 137 asthmatics residing and working in Pasadena, California, between September 3 and December 9, 1956; daily maximum hourly average oxidant levels (KI) from LA-APCD. Daily diaries for symptoms and medication of 45 asthmatics (aged 7-72 yr) residing in Los Angeles from July 1974 to June 1975; daily average concentration of 03, NO, NO2, SO2, and CO by LA- APCD within the subjects' residential zone. Daily log for symptoms, medication, and hospital visitation of 80 children with asthma (aged 8-15 yr) in the Chicago area during 197~75; air quality data on SO2, CO, PM; partial data for 03, pollen, and climate. Daily symptom rates in 82 asthmatic and allergic patients compared to 192 healthy telephone company employees in New Haven, Connecticut, from July to September 1976; average maximum hourly levels of O3 and average daily values for SO2, TSP, SO42-, pollen, and weather were monitored within 0.8 km of where the subjects were recruited. Of the 3,435 attacks reported, Schoettlin <5% were associated with and smog, and most of these Landau occurred in the same (1961) individuals; time-lagged correlations were lower than concurrent correlations; mean number of patients having attacks on days when oxidant levels were >0.25 ppm was significantly higher than days when levels were <0.25 ppm. No significant relationship Kurata et al. between pollutants and (1976) asthma symptoms; increased number of attacks at >0.28 ppm in a very small number of subjects; other factors such as animal dander and other pollutants may be important. Bad weather and high levels of SO2, CO, and PM exerted a minor influence on asthma, accounting for only 5-15% of the total variance; high levels of O3 increased both the frequency and the severity of asthmatic attacks; pollen density during fall and winter temperature variations had no influence; no exposure data given for quantitative treatment. Maximum oxidants were associated with increased daily prevalence rates for cough, and eye and nose . . . . 1rr1tat1on 1n leavy smo hers and patients with predisposing illnesses; pH of particulate was also associated with eye, nose, and throat irritation, whereas suspended sulfates were not associated with any symptoms. Questionable exposure Khan (1977) Zagraniski et al. (1979)

Eddy A. Bresnitz and Kathleen M. Rest 397 Table 2. Continued Av. Max. Conc. Range (ppm) 0.03-0. 15 Oxidants 0.07-0.39 ~O3 Pollutant Study Description Statistical analysis (repeated- measures design) of CHESS data on daily attack rates for juvenile and adult asthmatics residing in six Los Angeles area communities for 3=week periods (May-December) during 1972-75; daily maximum hourly averages for oxidants (KI) were monitored by LA-APCDs, 24-hr averages for TSP, RSP, SOx, NOX, SO2, and NO2 were monitored by the EPA, and meteorological conditions were monitored within 1 to 8 miles of homes in each community. Emergency room visits and hospital admissions for children with asthma symptoms during periods of high and low air pollution in Los Angeles from August 1979 to January 1980; daily maximum hourly concentrations of 03, S02, NO, NO2, HCs, and COH; weekly maximum hourly concentrations of SO42- and TSP; biweekly allergens and daily meterological variables from regional monitoring stations. Results and Comments Reference assessment, use of prevalence rather than incidence data, and lack of correction for dropout rates limit the usefulness of this study for developing quantitative exposure/ response relationships. Daily asthma attack rates increased on days with high oxidant and particulate levels and on cool days; presence of attack on the preceding day, day of week, and day of study were highly significant predictors of an attack. Questionable exposure assessment including lack of control for medication use, pollen counts, respiratory infections, and other pollutants and possible reporting biases limit the usefulness of this study for developing quantitative exposure/response relationships. Asthma positively correlated with COH, HCs, NO2, and allergens on same day and negatively correlated with O3 and SO2; asthma positively correlated with NO2 on days 2 and 3 after exposure; correlations were stronger on day 2 for most variables; nonsignificant correlations for SO42- and TSP. No indication of increased symptoms or medication use during high- pollution period; however, peak flow decreased (no differentiation of pollutants). Factor analysis suggested possible synergism between NO, NO2, RH, and wind speed; 03, S02, and temperature; and allergens and wind speed. Presence of ( Table continued next page.) Whittemore and Korn (1980) Richards et al. (1981)

398 Epidemiologic Studies of Oxidant Exposure Table 2. Continued Av. Max. Conc. Range (ppm) Pollutant Study Description Results and Comments Reference 0.03-0.12 O3 0.02 - 0.16 O3 Admissions to 79 acute-care hospitals in southern Ontario for the months of January, February, July, and August in 1974, 1976-78. Hourly average concentrations of particulate COH, 03, SO2, NO2, and daily temperature from 15 air sampling stations within the region. Fifty-one asthmatics (aged 7- 55 yr) exposed to ambient air from May to October 1981 in Houston, Texas; attack status determined each 12-hr period using log forms; individual environmental exposure estimates for the 12-hr period (1-hr max used for health data analysis) from regression model describing relationship of fixed-site monitors (within 2.5 miles) . . to microenvironment; multiple logistic regression model of Whittemore and Korn (1980) used for 42 subjects along with data on 03, NO2, temperature, humidity, pollen, and attack status. confounding variables, lack of definitive diagnoses for asthma, and questionable exposure assessment limit the quantitative interpretation of this study. Excess respiratory admissions associated with S02, 03, and temperature during July and August with 2t and 48-hr lag; only temperature was associated with excess respiratory admissions and total hospital admissions for January and February. Lack of sufficient exposure analysis limits the quantitative use of this study. Increased probability of asthma attack associated with increased O3 and decreased temperature (and previous asthma attack); definition of asthma was specific for each individual; good control of confounders except for pollen levels. Quality control of questionnaires was excellent; peak flow changes were measured but not reported; personal . . monitors consistent .y indicated underestimates of exposures determined by fixed-site monitors. Bates and Sizto (1983) Holguin et al. (1985) a Represents average maximum hourly concentrations during low (minimum) and high (maximum) pollution periods. NOTE: CHESS = Community Health Environmental Surveillance System; COH = coefficients of haze; HC = hydrocarbon; KI = potassium iodine method; LA-APCD = Los Angeles Air Pollution Control District; PM = particulate matter; RH = relative humidity; RSP = respirable suspended Articulates; TSP = total suspended Articulates. SOURCE: Adapted from U.S. Environmental Protection Agency 1986.

Eddy A. Bresnitz and Kathleen M. Rest 399 lated for each subject for each hour of every 12-hr period, and the 12-hr exposure esti- mate was defined as the maximum hourly estimate during the period. Time spent indoors was not included in the estimate. O3 and NO2 were the only exposure vari- ables that covaried significantly during the daytime. The study developed individual attack definitions after all health and activity data had been collected but before the analysis of air monitoring data began. Attack defini- tions included self-reported symptoms, a decrease in expiratory peak flow, and an increase in the use of asthma-specific med- ication (Holguin et al. 1983~. A logistic regression model (Whittemore and Korn 1980) was used to estimate the effect of the measured variables on the risk of an attack. The probability of an attack was signifi- cantly associated with an attack on the previous day, increasing O3 levels, and decreasing temperature. NO2, pollen count, and relative humidity had little effect on the risk of an attack. The average am- bient O3 level in the 12-hr periods during which the study subjects experienced at- tacks was less than 0.05 ppm in all in- stances. The strength of this study lies in its excellent assessment of individual exposure and its effects, its careful definition of an asthmatic attack unique for each individual, and the quality of its statistical analysis. Nevertheless, that study and the other cross-sectional studies discussed here only suggest associations between increased levels of ambient oxidants and decrements in pulmonary function, higher prevalence of respiratory symptoms, and more frequent asthma attacks. They cannot and do not indicate whether these effects are tempo- rary, have any long-term clinical signifi- cance, or lead to increased morbidity with time. Longitudinal cohort studies are best suited to assess these issues. Prospective Cohort Studies The prospective cohort method identifies study subjects on the basis of individual exposure and then follows them forward in time to assess disease development in ex posed and unexposed groups. The ability to identify exposure prior to disease devel- opment makes this design theoretically more powerful than cross-sectional studies in assessing causality in disease/exposure relationships. The magnitude of the risk factor is measured by obtaining the ratio of the disease incidence rates in the exposed and unexposed groups (the relative risk). The cohort design is particularly good for studying many outcome measures simulta- neously. Prospective cohort studies allow for better quality and control of data col- lection. The disadvantages of the cohort design include potentially long follow-up periods with the attendant problems of loss to follow-up and migration out of the geo- graphic area; expense of maintaining fol- low-up; changing levels of exposure; and the low incidence of certain diseases, such as COPD, necessitating large groups of study subjects. The Six-Cities study (Ferris et al. 1979) is an ongoing cohort study of the respiratory health elects of air pollution on children and adults. The cities in the study were selected on the basis of their historical levels of air pollution. They represent a range of ambient SO2 and particulate exposure lev- els above and below the current ambient air quality standard. The main goal of the study is to assess the health effects of long- term exposure to these pollutants. The investigators are collecting data on SO2, N02, 03, TSP, and mass respirable parti- cles by use of personal monitors (where possible) as well as central site monitors. The published reports to date have fo- cused on the health effects of ambient SO2 and TSP. The Six-Cities study has sup- ported earlier findings by Spengler and coworkers (1979) that indoor sources of NO2 contribute more to total NO2 expo- sure than outdoor, ambient levels. Ferris and coworkers have not yet published their analyses of the respiratory health effects of ambient oxidants. However, the study may provide important information on the ef- fects of oxidants on the rate of decline in pulmonary function and on the prevalence of symptoms in smokers and nonsmokers if the cities vary in their level of oxidants.

400 Epidemiologic Studies of Oxidant Exposure That study and the UCLA population studies are essentially the only ongoing prospective cohort studies in air pollution research. The UCLA study discussed pre- viously should provide additional longitu- dinal information about accelerated changes in pulmonary function among long-term residents of the high-pollution areas compared to the low-pollution areas. Repeat pulmonary function tests in the participants from Glendora and Lancaster will permit comparison of changes in de- cline of pulmonary function. An acceler- ated rate of decline in the high-pollutant community will suggest an effect of long- term, increased levels of air pollutants. Separating the effects of oxidants from the effects of other pollutants may be difficult. Recommendation 1. A cohort study of the effects of long-term exposure to oxidants on respiratory morbidity in chil- dren, normal adults, and sensitive popula- tions should be done if the Six-Cities study and the UCLA population studies cannot assess these issues. Biological Markers. Use of early markers of lung injury would shorten considerably the follow-up period necessary to assess the development of disease in any cohort study. A shorter follow-up period would decrease the problems of loss to follow-up, . . . . migration, cost, investigator turnover, ant . low incidence of disease. Several investiga- tors have proposed measuring hydroxy- proline (Yanagisawa et al. 1986), antipro- teinase fragments Johnson and Travis 1979), elastin-derived peptides (Kucich et al. 1984), and antibodies to collagen (Mi- chaeli and Fudenberg 1974) as early mark- ers of emphysema. Cohort studies that assessed cumulative exposure or delivered dose, followed subjects for an increase in one or more of these markers as evidence of early disease, and controlled for potential confounders such as smoking and occupa- tion could yield valuable information on threshold levels and dose/response relation- ships between exposure to O3 and NO2 and the development of emphysema. Un- fortunately, to date there is no evidence that establishes these substances as valid markers of early injury or disease. In fact, these substances may only be markers of exposure and reflect the activity of adaptive physiological processes. Retrospective Studies Two other study designs may be useful in studying the relationship of respiratory morbidity to oxidant exposure: the histor- ical (retrospective) cohort study and the case-control study. In both designs, infor- mation on confounders and past exposures of interest is ascertained by a variety of methods including direct interviews, ab- straction of medical records, and the use of existing data bases. These designs are par- ticularly good for studying rare diseases and chronic diseases with long latency pe- riods. They can be performed relatively quickly and often with minimal cost. Moreover, these methods can identify more than one risk factor in the same set of data. Their major limitation is determining the validity and reliability of the informa- tion on past exposures. Both designs have been used to study and imply causal rela- tionships between disease and exposure to drugs, infectious agents, chemicals, and smoking. Neither design has been used in air pollution research, most likely because of the difficulty of accurately ascertaining past exposures to pollutants. ~ Recommendation 2. Future epidemi- ologic studies of air pollution should in- clude the use of retrospective methods to investigate the relationship between oxi- dant exposure and respiratory morbidity if more accurate methods of assessing past cumulative exposure can be developed. Generic Issues Exposure Assessment Estimating exposure, prospectively as well as retrospectively, is a difficult task. In most epidemiologic studies of air pollution, exposure has been measured by stationary air pollution monitors put into place for public health and regulatory purposes, not

Eddy A. Bresnitz and Kathleen M. Rest 401 for research purposes. The distance of these monitoring stations from the population or area under study has varied a great deal. Area Monitors. A few studies have in- cluded better-placed area monitors to sup- plement and make more precise the data obtained from regional monitoring stations (Kagawa and Toyama 1975; Holguin et al. 1985; Selwyn et al. 1985~. Well-placed area monitors can yield a more precise assess- ment of exposure, especially when supple- mented with time-activity data. Personal Monitors. Elsewhere in this vol- ume, Sexton and Ryan review the status of exposure assessment in air pollution re- search, and suggest several approaches for improving the quality of this assessment process. For the most part, their recom- mendations focus on the relationship be- tween personal monitoring and ambient air quality measurements. Because there are no sensitive biological markers of exposure to NO2 and O3 as there are for carbon mon- oxide and lead, personal monitors currently provide the most accurate assessment of exposure. Investigators should consider using per- sonal monitors in all future cohort studies of air pollution as is being done in the Six-Cities study (Ferris et al. 1983~. Results from personal monitoring should be corre- lated with exposure assessment determined by questionnaires, stationary samplers in the home (for NO2), and measurements from air monitoring stations. Only a sam- ple of the subjects need personal monitors when the size of the study groups is large. By adjusting the results from personal monitors for time and activity patterns as determined by personal activity logs, in- vestigators can more accurately assess total personal exposure and better correlate this measure with early markers of effect. This information should provide more accurate assessment of dose/response effects at low levels of exposures. Separate assessments for average total exposure and peak expo- sure to oxidants should also be made. · Recommendation 3. Future epidemi- ologic studies should use personal monitors and a greater number of better-placed area monitors to assess true exposure in at least a sample of the individuals under study. Exposure Questionnaires. Standardized questionnaires for exposure assessment would be useful in ensuring comparability of data across studies. They should collect qualitative and quantitative information on indoor and outdoor activities, long-term residence, all potential confounders (per- sonal as well as environmental), and effect modifiers. Each question should be tested for reliability and validity in different pop- ulations at risk. Measures of reliability and validity would allow for adjustment of re- sponses among different subgroups where needed. A standardized exposure question- naire should be validated against estimates of exposure determined by personal and well-placed environmental monitors. A questionnaire that correlated well with other measures would obviate the need to assess actual exposure in every individual. Recommendation 4. Valid and reliable questionnaires that help assess personal expo- sure to air pollutants should be developed and used in future epidemiologic studies. Elects Assessment Questionnaires and Diaries. Question- naires have been the mainstay of assessing symptoms in epidemiologic air pollution studies. Variations of questionnaires devel- oped by the Medical Research Council (MRC) (1960) and the Division of Lung Disease in the National Heart, Lung, and Blood Institute (Ferris 1978) have been the most frequently used. Although many of the questions in these instruments have been tested for reliability and validity, most epidemiologic studies of air pollution have not actually assessed the accuracy of the responses that these questions have elicited in their own studies. Validity could be as- sessed by comparing symptoms recorded by the patient with responses to questions posed to the patient's spouse or a close family member. The use of diaries to collect health ef- fects data may be preferable to periodic use

402 Epidemiologic Studies of Oxidant Exposure of questionnaires or personal interviews. Methodological studies of the use of diaries to collect health data indicate their potential for collecting accurate information (Na- tional Center for Health Services Research 1980~. A 197~74 Survey Research Labora- tory study (University of Illinois) found that households of low as well as high educational levels were willing and able to use diaries, that they reported 14 percent more medical events than did households during recall interviews, that a ledger for- mat was superior to a journal format, and that compensated households reported higher levels of events than uncompensated households (National Center for Health Services Research 1980~. Future epidemiologic studies could use diaries to collect data on specific symp- toms, days of restricted activity, physician and emergency room visits, purchase/use of medicine, or days the respondent felt ill but performed usual activities. Statistical analysis of data obtained through diaries may be problematic if the instrument or instructions use a rating system that de- pends on the previous day's response (for example, better than yesterday). If such a problem is not eliminated by the design of the diary, special care will be required during data analysis. As with other data collection methods, validation studies should precede widespread use. · Recommendation 5. Valid and reli- able questionnaires or diaries that assess health effects should be further developed to facilitate epidemiologic studies in air pollution. Ongoing cohort studies (Six- Cities, UCLA) could be used to develop and test such data collection instruments. Spirometry and Single-Breath Tests. Spirometry is the most practical test to assess pulmonary function and the one most often used in epidemiologic studies. Pulmonary function changes from pre- dicted values are reflected in several param- eters, including average absolute differ ences in flow rates and lung volumes, variation in rates of decline of flow rates, and increases in airway resistance and air- way response to bronchoconstrictors. Spirometry is a well-standard~zed test tor determining changes in rates of expiratory flow. Its ease of administration, reliability, and validity favor its continued use as a tool for measuring functional impairment of the lung. The forced expiratory volume in one second (FIVE) has a low coefficient of variation. The forced expiratory flow rate between 25 and 75 percent of the vital capacity (FEF2~75%) is derived from the forced vital capacity and is a good measure of obstruction in small airways. The FEF2~75% varies widely among individuals and therefore may be overly sensitive in detecting minor degrees of airflow obstruc- tion if the lower limit of normal is defined as the conventional 80 percent predicted. In a cohort study, however, the rate of decline would be less affected by this variability, and the measure may have better predictive value for early COPD. The single-breath nitrogen or oxygen test has also been used in epidemiologic studies to assess small airway function. The measures that are derived from this test have high coefficients of variation. The test is somewhat more difficult to administer than spirometry, but it may be more sen- sitive than the FEF2~75% in detecting small airways obstruction in its reversible stage, prior to the development of COPD. How- ever, evidence of small airways disease does not yet have predictive value for the devel- opment of COPD. Becklake and Permutt (1979) have indicated that measures of small airway function, derived either from Spirometry or the single-breath oxygen test, may eventually be useful for predict- ing the development of COPD in smokers. It will be important to assess these tests in nonsmokers who are long-term residents of high- and low-oxidant environments. The UCLA population study and the Six- City study may yield information on changes in small airway function that are associated with varying ambient oxidant levels. ~ Recommendation 6. Longitudinal co- hort studies should assess small airway function in nonsmokers. If residents of high-oxidant environments have decreases in small airway function compared to resi .. . .. . ~

Eddy A. Bresnitz and Kathleen M. Rest 403 dents of low-oxidant environments, the cohorts should be followed for a sufficient amount of time to assess the development of COPD. Bronchial Provocation Testing. In addi- tion to using changes in spirometric values as outcome measures, some investigators have focused on bronchial reactivity as a possible predictor of progression to COPD among sensitive individuals. Barter and Campbell (1976) have theorized that, among smokers, individuals with sensitive airways are at special risk for progression to COPD, specifically chronic bronchitis and emphysema. Methacholine challenge is of- ten used to assess bronchial sensitivity; histamine and cold air challenge have also been used. In a methacholine challenge test, airflow and airway conductance are mea- sured serially after the inhalation of pro gresslvely Increasing concentrations of methacholine, a bronchoconstrictor. An in- dividual is classified as being sensitive if airflow or conductance falls a specified per- cent after the inhalation of a relatively low concentration of methacholine. This test serves as the gold standard for diagnosing bronchial hyperreactivity. (For additional discussion, see Bromberg, this volume.) Recently, Chan-Yeung and coworkers (1986) surveyed an occupational cohort of 1,392 males for respiratory symptoms and the degree of hyperresponsiveness to methacholine inhalation. The prevalence of bronchial hyperresponsiveness was 12 per- cent among 694 asymptomatic individuals with normal lung function. Among the symptomatic participants, hyperresponders were more than three times as likely to have chronic bronchitis than the nonhyper- responders. Asthmatics may also be at greater risk for accelerated decline in lung function. A study comparing asthmatic smokers and asthmatic nonsmokers to normal subjects without a history of smoking or respiratory illness, showed a significant correlation be- tween rate of decline of FEW and the level of bronchial responsiveness to histamine (Woolcock et al. 1986~. Smoking and asthma were independent risk factors for an accelerated decline in FIVE. It is possible that sensitive individuals without clinical asthma, exposed repetitively to high peak levels or to chronically elevated low levels of oxidants, are at higher risk for develop- ing an accelerated decline in lung function and COPD. Recommendation 7. Small panels of individuals with sensitive airways should be studied carefully to assess whether oxi- dants are risk factors for various outcome measures of respiratory morbidity. Data Bases. The utility of extant data bases for collecting data on the health ef- fects of air pollution remains relatively un- explored. There are several population- based data bases that, if modified, may provide useful information on the effects of oxidants on health. Data collected by the National Health and Nutrition Examina- tion Survey, the National Health Interview Survey, health maintenance organizations, preferred provider organizations, and in- surance companies are some examples of potentially useful data sources. A survey of existing data bases is needed to determine which are most likely to pro- duce useful information on the health ef- fects of air pollutants. Potentially useful data bases would probably require modifi- cation to ensure collection of information on potential confounders. Some data bases may serve as sampling frames for longitu- dinal, prospective cohort studies needed to test specific hypotheses associating expo- sure to oxidants with a specific health out- come (for example, development of early COPD). · Recommendation 8. Extant data bases should be explored for their use in studying the effects of air pollution on health. Hospital and Physician Records. Data on health service utilization have been used as indices of morbidity in populations. In air pollution research, hospital and physician records may be helpful in evaluating threshold effects and the effects of exposure On sensitive groups. The use of hospital statistics as an indi

404 Epidemiologic Studies of Oxidant Exposure rect measure of community morbidity has several limitations. Statistics such as num- ber of visits to the emergency room, num- ber of admissions, and length of stay can be affected by illness behavior, the demo- graphic characteristics of the population in the catchment area, the availability of beds, and changes in the definitions of disease or coding practices (Bennett 1981~. Most of these problems also apply to data based on outpatient visits to physicians' offices. As a result, comparison of illness rates based on encounters with the health care system can be problematic. The lack of a gold standard for diagnosing many respiratory illnesses makes the interpretation of studies that utilize these measures all the more difficult. Nevertheless, several investigators have utilized hospital data to assess the effects of air pollutants on respiratory morbidity (Sterling et al. 1966, 1967; Bates and Sizto 1983~. Bates (1985) has argued that despite limitations on hospital data, systematic di- agnostic bias would be unlikely in studies that involved large populations over long periods of time and many different institu- tions and physicians. With such caveats, hospital data may be used efficiently and appropriately, especially in cross-sectional studies. Specific Issues ant! Stucly Approaches A number of questions remain unanswered in oxidant air pollution research. Still lack- ing is quantitative information on the threshold levels of effects, dose/response relationships, effects on sensitive groups, and the quantification of disease burden on public health. In particular, the effect of long-term oxidant exposure on human health remains a significant and unan- swered question. Does such exposure in- crease the risk of or exacerbate existing COPD, bronchial hyperresponsiveness (for example, asthma), or pulmonary infection? Several approaches may be useful in exam- ining these questions. Each approach will have its strengths and weaknesses; no one study by itself will produce conclusive ev idence of causality. Nevertheless, the con- sistency of findings from studies done in different populations, during different peri- ods of time, and using different study de- signs would allow valid causal inference. Chronic Obstructive Pulmonary Disease Cohort Methods. The two ongoing co- hort studies (Six-Cities and UCLA) are collecting morbidity data based on pulmo- nary symptoms and pulmonary function test results. These and any future cohort studies should look carefully at nonsmok- ing young adults who are long-term resi- dents (more than 1~20 years) of the spec- ified geographic areas to assess the relationship between accelerated decline in lung function and exposure to oxidants. Exposure should be assessed periodically by using personal monitors and time-ac- tivity logs in at least a sample of the study population. Information on confounders and effect modifiers should be elicited pe- riodically; any differences in the prevalence of these variables between the exposure groups must be adjusted in the analysis. Information from questionnaires will re- quire reliability and validity testing. To assess the future clinical significance in any statistically increased rates of decline in high-exposure groups, prolonged follow- up of these cohorts will be necessary. Retrospective Methods. A case-control study comparing individuals with a specific res- piratory illness (for example, chronic bron- chitis) to healthy controls for estimated measures of exposure collected retrospec- tively might yield dose/response and thresh- old response information on the effects of oxidants on the disease under study. One of the data bases mentioned earlier could be useful in assessing the role of oxidants (or other air pollutants) in the etiology of COPD, using a case-control study of nonsmokers with COPD. For example, most states now have large data bases consisting of all patients enrolled in Medicaid. Individual information such as age, gender, race, county, diagnoses, drugs dispensed, procedures, and place of visit are

Eddy A. Bresnitz and Kathleen M. Rest 405 recorded in these data bases. Use of these data would allow comparisons of patients who have a recorded diagnosis of COPD, chronic bronchitis, or emphysema to con- trols without these diagnoses for county of residence. Air pollution monitoring data in each county should provide a qualitative esti- mate of exposure. Personal contact with the medical providers for selected cases and controls would be necessary to estimate the prevalence of smoking in each group and the length of residence in each area. If the COPD group lived in higher air pollution areas than the control group, the study would suggest air pollution as a risk factor for the development of COPD. Arkansas and Ohio are two states worthy of study, given the high mortality rates for chronic bronchitis in certain portions of each state (National Research Council 1985~. Recommendadon 9. Investigators should do case-control studies using existing health service data bases to assess the relationship between specific diseases and long-term residence in high-pollution areas. Cross-Sectional Methods. Cross-sectional studies and secular trend analyses using the Medicaid data bases could also assess the prevalence of chronic respiratory disease and, perhaps, the frequency of various acute res- piratory diseases in counties with different average levels of air pollutants. A rough estimate of a dose/response relationship may be evident in this type of analysis, which would also give information on other effects of air pollution exposure. The diagnoses re- corded by the physician in the office or hospital would be used as the outcome vari- able of interest. Cross-sectional studies can also be used to study the effects of point-source emis- sions on respiratory health. New roads and traffic patterns as well as industrialization of previously rural areas may produce new _ areas of high air pollution. If possible, pulmonary function measurements and morbidity surveys should be performed before and after the development of a new point source of oxidant pollutants. Changes in the mean levels and predicted rates of decline in pulmonary function or in the prev- alence of morbidity outcomes would impli- cate the point source as a causal factor. A new point source may develop in Spring Hills, Tennessee, where the con- struction of a model automobile manufac- turing plant is planned. General Motors expects to employ approximately 6, 000 workers at the new plant. Construction and operation of the plant is expected to in- crease and change traffic patterns in the community. The study outlined in the pre- ceding paragraph could identify changes in pulmonary function or morbidity out- comes in relation to specific areas of high traffic volume or congestion. To strengthen the study, a nearby com- munity that is unaffected by changing traf- fic patterns can serve as an unexposed con- trol group. The two areas can then be compared on various measures of respira- tory morbidity in the community (hospi . . . . ~ ta Batons, VlSltS to emergency rooms, anc so on) and personal morbidity (symptoms, pulmonary function). Such a study would be quasi experimen- tal in that both communities would have a similar initial exposure to air pollutants, but a different level of exposure after the construction of the plant. Recommendation 10. New point- source emissions should be identified and their effects on air pollution and respiratory morbidity should be studied longitudi- nally. Asthma/Bronchial Hyperresponders Future epidemiologic studies of air pollu- tion should include those groups at high risk for developing adverse effects. Study- ing high-risk groups will enhance our abil- ity to detect increases in the risk of exacer- bating preexisting diseases in populations exposed to low ambient levels of pollutants. Cohort Methods. The cohort design would be appropriate for assessing the ef- fects of oxidant exposure on the morbidity of sensitive and normal individuals as de- termined by methacholine challenge. The National Heart, Lung, and Blood Institute

406 Epidemiologic Studies of Oxidant Exposure recently funded a large multicenter ran- domized clinical trial in smokers to assess whether smoking cessation will effectively modify the course of COPD at a stage when mild dysfunction has already oc- curred (Gunby 1984~. A subsample of smokers with hyperreactive airways will also be followed over a period of five years. This study is being organized in eight cities across the United States and may be a good source of hyperreactive patients who are exposed to different ambient levels of air pollutants. Retrospective Methods. Alternatively, a case-control design can be used to compare the past oxidant exposure of bronchial hy- perresponders and nonresponders. Assum- ing development of a valid questionnaire for past exposures, higher exposure levels in sensitive individuals would suggest that . . . . . . OX1C .ants elt per increase sensitivity or exac- erbate the sensitive state. If oxidants have one or both of these effects, exposure to oxidants may contribute to the increased prevalence of chronic bronchitis in hy- perresponders and to the accelerated de- cline in lung function in asthmatics noted above. Cross-Sectional Methods. The study de- sign used by Holguin and coworkers (1985) to study asthmatics should be applied to other sensitive groups, such as people with COPD, who may be at increased risk for the exacerbation of symptoms. The design could be strengthened by including healthy people in the study group and by using personal monitors to better assess actual exposures to NO2. Studies should focus on smokers as well as nonsmokers with sensi- tive airways as determined by metha- choline challenge testing. Recommendation 11. A variety of de- sign methods should be used to study non- smoking and smoking individuals with bronchial hyperresponsiveness to assess whether long-term residence in high-oxi- dant environments causes accelerated rates of decline in pulmonary function, more frequent encounters with the health care system, and increased severity of pulmo- nary symptoms. Healthy Cohort Effect. Investigators must be cautious in comparing morbidity rates in groups of high-risk individuals living in dif- ferent air pollution environments. Standard- ized morbidity rates may be equal in popu- lations with extremes in their exposure to oxidants, yet the high-oxidant environment may still be a risk factor for exacerbation of disease that is not evident because of a selec- tion bias. This bias is analogous to the "healthy worker effect" seen in occupational epidemiologic studies where disease rates in occupational cohorts appear artificially low because of the "healthiness" of working pop- ulations (Monson 1980; Sterling and Weinkam 1986~. This may be termed the "healthy cohort" effect in epidemiologic air pollution studies. For example, asthmatics who are highly susceptible to oxidants and other pollutants may migrate out of a high-air-pollution environment, leaving behind those asth- matics the "healthy cohort" who may tolerate high exposures. Comparison of the remaining asthmatics (or a representative sample) with a panel of asthmatics in a low-exposure environment may show sim- ilar asthma attack rates that suggest no increased risk in high-pollution areas. Clearly, the selection of relatively healthy asthmatics is a form of selection bias that may obscure the risk of exposure to envi- ronmental pollutants. The magnitude of the healthy cohort effect is substantial and affects all age groups. For example, an analysis of data from the Household Interview Survey of the National Center for Health Statistics showed that employed individuals had lower standardized morbidity rates of chronic respiratory ailments compared to unemployed people (Sterling and Weinkam 1985~. Unemployed males were four times as likely to have chronic respiratory ail- ments as employed males. Although the magnitude of the healthy person effect is unlikely to be so large in morbidity studies on air pollutants, the size of the effect should be estimated in all air pollution

Eddy A. Bresnitz and Kathleen M. Rest \ 407 studies. This may be done by asking the study subjects whether they have ever moved for health reasons. · Recommendation 12. All epidemio- logic studies should analyze the outcome measure of interest by stratifying groups on their length of residence in the study areas. Respiratory Infection Pennington (this volume) indicated that there is little experimental evidence of a relationship between automotive emissions and the risk of respiratory infections. To date, there have been no epidemiologic studies assessing the incidence or severity of infection in populations exposed to high oxidant levels. Several studies have looked at the prevalence of symptoms that may Indicate either Injection or a direct Irritant effect of oxidants (Shy et al. 1970a,b; Pearl- man et al. 1971; Speizer et al. 1980; Com- stock et al. 1981~. Investigators have not used either sputum cultures or serology to assess the presence of true infection. A recent review examined the relation- ship between respiratory illness in child- hood and chronic airflow obstruction in adulthood (Samet et al. 1983~. The authors decided that results from different studies on this question conflicted and that the data to date were inconclusive. However, the studies cited in the review support the conclusion that a history compatible with lower respiratory infection is associated with impaired pulmonary function in chil- dren. This association may be causal or noncausal and may be mediated by a host factor such as atopy (allergy that is proba- bly hereditary) or an environmental factor such as parental smoking or oxidant expo sure. Epidemiologic investigation of the rela- tionship between the risk of respiratory infection and exposure to oxidants would be difficult to conduct. Selection of the organisms to study would be the first of several difficult decisions. Even limiting the study to viral organisms would not neces- sarily narrow the spectrum of tests needed to assess true infection. Antibody levels for Table 3. Sample Size Requirements for a Cohort Study in a Low-Oxidant Area Relative Frequency Total Risk of Disease Sample Size 2 0.5 0.4 0.2 0.1 0.05 0.5 0.4 0.2 0.1 0.05 0.2 0.1 0.05 0.01 104 176 534 1,254 2,690 18 40 148 362 794 40 112 254 1,404 NOTE: c' = 0.05 (1-tail), ,l3 = 0.20. several viral antigens would be needed to ensure complete assessment for the most likely pathogens. Cohort Methods. After deciding on the organisms) to evaluate, investigators could study cohorts of high-risk populations (for example, chronic bronchitics) residing in high- versus low-level oxidant areas. Serial measurements of antibody titers could be compared to assess whether the high-oxi- dant group had higher conversion rates (fourfold rise) than the low-oxidant group. Such a study would presuppose that the right organists) has been identified a pri- ori, that the populations have had the same likelihood of exposure to these organisms, that they are susceptible to infection with these organisms, and that the prevalence of exposure is high enough to ensure a suffi- cient incidence of clinical disease with a fourfold rise in antibody titers in both groups. Culture of organisms would not necessarily increase the ability to detect differences in infection rate between the two groups (Tager and Speizer 1975~. It is easy to calculate the sample size of previously uninfected subjects required to detect a significant clinical difference be- tween the high- and low-oxidant groups. Table 3 shows how the required sample

408 Epidemiologic Studies of Oxidant Exposure size varies depending on the frequency of infection in the low-oxidant group and the relative risk that the investigator would like to detect with a reasonable degree of cer- tainty (,8 = 0. 20) (see also Schlesselman 1974). For example, 267 subjects would be re- quired in each area (534 total) to detect a 50 percent Increase In Injection rate In the high-oxidant group, assuming that 20 per- cent of the low-oxidant group developed disease on exposure to the organism of interest. Note that the sample size require- ments are minimum estimates and do not account for loss to follow-up, failure to develop clinical disease on exposure to or- ganism, and the likelihood of being ex- posed to the organism. All these factors would increase the sample size needed to do the study. However, if infection with any one of several identified organisms was the outcome measure of interest, the sam- ple size required to detect a difference in the incidence of infection between the high- and low-oxidant group would de- crease. Retrospective Methods. A case-control study of the relationship between respira- tory infection and oxidant exposure would also be difficult to do. However, if one could specify the organisms of interest in advance and, on a prospective basis, iden- tify cases as they are presented to the hos- pital (or to a physician), one could then compare them to controls selected on the basis of diagnoses not conceivably related to oxidant exposure. Clearly, the investi- gator would require a reliable and valid Instrument tor assessing past exposure to oxidants and the presence of potential con- founders. The Medicaid data base described previously may be an excellent resource for doing this type of study. ~ Recommendation 13. Seroepidemio- logic surveys should be done in cohorts that have different exposures to oxidants. More than one organism should be se- lected as a suspected respiratory pathogen to maximize the likelihood of detecting differences in infection rate among the co- horts. Attributable Risk Air pollution research efforts have focused primarily on acute effects in attempts to address issues of threshold effects, dose/re- sponse relationships, and disease incidence and prevalence in sensitive groups. Investi- gators have paid little attention to assessing the potential impact of improved air quality standards on reductions in health risks. Since most of the diseases associated with exposure to oxidants (and other pollutants) have multifactorial origins, it would be use- ful to know how reducing average pollutant levels might proportionally reduce morbid- ity. The attributable risk of exposure to oxi- dants is the proportion of disease in the population that can be theoretically pre- vented by eliminating exposure to oxidants. Morganstern and Bursic (1982) outlined a mathematical method for estimating the potential impact of a public health policy on the health status of a target population. Investigators applying this method to am- bient oxidants would have to assume that (1) the distribution of oxidant levels in the population is known; (2) oxidants are true determinants of the diseased) in question (as measured by the relative risks); (3) a reduction in ambient oxidant levels would reduce the risk of diseases in previously exposed persons compared to the risk of diseases in previously unexposed persons; (4) there are no significant secular trends in the risk of diseases) because of changes unrelated to the reduction in ambient oxi- dant levels; and (5) there are no significant changes in other risk factors that would interact with oxidants to synergistically in- crease the risk of diseases. Because of the latter assumption, such models must be applied cautiously to air pollution research. When there are several, noninteracting causes of adverse health effects, the simple addition of the attributable risks associated with each causal factor will yield the total benefit in risk reduction that is theoretically achievable. However, in the case of inter- acting or synergistic pollutants, attributable risks must be viewed more carefully. When synergism exists, no single etiologic factor can be said to "cause" a percentage of the disease that is numerically equivalent to the

Eddy A. Bresnitz and Kathleen M. Rest 409 attributable risk. For example, it is true that smoking cessation could prevent x percent of disease in a target population, but it may also be true that the elimination of other occupational or environmental toxins could prevent y percent of the disease in the same population. The addition of x percent and y percent can yield a number much greater than 100 percent in the case of synergistic , , _ risk factors (Ashford 1985; Sterling 1985~. With these caveats, studies that assess attributable risk of oxidant exposure would be highly useful to policy makers and reg- ulators, as well as of great interest to the general public, the automotive industry, and public health professionals. Mathemat- ical modeling using data collected from previous studies could provide some esti- mate of the amount of morbidity that could be detected in populations with different diseases and potential exposures. In addi- tion, mathematical modeling could help esti- mate the expected decree of interaction be- tween air pollutants and other risk factors. Recommendation 14. The potential attributable risk for disease caused by ex- posure to oxidants and other air pollutants should be determined and expressed in a meaningful way. measures of acute respiratory morbidity. Review of selected air pollution studies shows that higher oxidant exposures are associated with a higher prevalence of res piratory symptoms, more frequent asthma attacks, greater decreases from expected pulmonary function, and more frequent utilization of health care services. The long term and chronic effects of low-level expo sure to oxidants is essentially unknown. The major gaps in oxidant air pollution research include information on threshold levels of effects and dose/response relation ships, long-term effects on sensitive groups, and the quantification of disease burden on public health. Investigators must address several meth odological issues when conducting epide miologic research in this area. They involve developing more specific tools for assessing personal exposure and adverse health ef fects; improving assessment and control of confounders and effect modifiers; identify ing sensitive populations for more detailed study; and assessing potential biases in pop ulation selection. Several approaches for addressing these methodological issues and for answering important research questions have been ad dressed in this paper. Cross-sectional, case control, and cohort studies can be used to examine the relationship between oxidant exposure and increased risk for or exacer Summary bation of existing COPD, asthma, or pul Epidemiologic studies in air pollution re- search are necessary to assess the effects of long-term oxidant exposure on human health. Most studies done to date have utilized the cross-sectional design, which is best suited for correlating current levels of exposure with the prevalence of various monary injections. leach approach has its strengths and weaknesses and will be ap- propriate in different situations. The con- sistency and reproducibility of findings from different epidemiologic studies are necessary to allow causal inference between oxidant exposure and adverse health ef- fects. Summary of Research Recommendations Specific Research Issues HIGH PRIORITY Studies that will provide data on the long-term, chronic health effects of exposure to low levels of photochemical oxidants are of highest priority.

410 Epidemiologic Studies of Oxidant Exposure Recommendation 1 A cohort study of the effects of long-term exposure to oxidants on respiratory morbidity in children, normal adults, and sensitive populations should be done if the Six-Cities study and the UCLA population studies cannot assess these issues. Recommendation 6 Longitudinal cohort studies should assess small airway function in nonsmokers. If residents of high-oxidant environments have decreases in small airway function compared to residents of low oxidant environments, the cohorts should be followed for a sufficient amount of time to assess the development of COPD. Recommendation7 Small panels of individuals with sensitive airways should be studied carefully to assess whether oxidants are risk factors for various outcome measures of respiratory morbidity. Recommendation 11 A variety of design methods should be used to study nonsmok ing and smoking individuals with bronchial hyperresponsiveness to assess whether long-term residence in high-oxidant environments causes accelerated rates of decline in pulmonary function, more frequent encounters with the health care system, and increased severity of pulmonary symptoms. MODERATE PRIORITY Existing data bases may be an untapped resource of information on health effects. New environments may also provide an oppor tunity to explore exposure/effect relationships. Recommendation 8 Extant data bases should be explored for their use in studying the effects of air pollution on health. Recommendation 10 New point-source emissions should be identified and their effects on air pollution levels and respiratory morbidity should be studied longitudinally. LOW PRIORITY Causal hypotheses are strengthened if a variety of study designs, methods, and data sources are used and yield similar results. Recommendlation2 Future epidemiologic studies of air pollution should include the use of retrospective methods to investigate the relationship between oxidant exposure and respiratory morbidity if accu ~ 1 rate methods of assessing past cumulative exposure can be devel- oped. Recommendation 9 Investigators should do case-control studies using existing health service data bases to assess the relationship between specific diseases and long-term residence in high-pollution areas. Recommendation 13 Seroepidemiologic surveys should be done in cohorts that have different exposures to oxidants. More than one organism should be

Eddy A. Bresnitz and Kathleen M. Rest 411 selected as a suspected respiratory pathogen to maximize the likelihood of detecting differences in infection rate among the cohorts. Recommendation 14 The potential attributable risk for disease caused by exposure to oxidants and other air pollutants should be determined and ex pressed in a meaningful way. Generic Issues Methodological innovation will greatly enhance the ability of epidemiologic studies to investigate dose/response relationships, long-term effects of low-level exposure, and attributable risk. Such methodological research deserves the highest priority. Recommendation 3 Future epidemiologic studies should use personal monitors and a greater number of better-placed area monitors to improve the assessment of true exposure in at least a sample of the individuals under study. Recommendation 4 Valid and reliable questionnaires that help assess personal expo sure to air pollutants should be developed and used in future epidemiologic studies. Recommendations Valid and reliable questionnaires or diaries that assess health effects should be developed further to facilitate epidemiologic studies in air pollution. Ongoing cohort studies (Six-Cities, UCLA) could be used to develop and test such data collection instruments. . Recommendation 12 All epidemiologic studies should analyze the outcome measure of interest by stratifying groups on their length of residence in the study areas. Acknowledgment The authors wish to thank the following individuals for their careful reviews and thoughtful comments: Nicholas Ashford, Ph.D., I. D.; Peter Gann, M.D.; Murray Oilman, M.D.; Tee Guidotti, M.D.; Sandy Norman, Ph.D.; and Brian Strom, M.D. The authors also appreciate the unflagging clerical support of Karen Menna and Frances Seeds. Correspondence should be addressed to Eddy A. Bresnitz, Division of Occupational and Environmen- tal Health, Department of Community and Preven- tive Medicine, The Medical College of Pennsylvania, 3300 Henry Avenue, Philadelphia, PA 19129. References Ashford, N. A. 1985. Unpublished comments on the OTA Report on Smoking-Related Death and Fi- nancial Costs. Barter, C. E., and Campbell, A. H. 1976. Relation- ship of constitutional factors and cigarette smoking to decrease in 1-second forced expiratory volume, Am. Rev. Respir. Dis. 113:30~314. Bates, D. V. 1985. The strength of the evidence relating air pollutants to adverse health effects, Carolina Environmental Essay Series VI, Univer- sity of North Carolina Institute for Environmental Studies, Chapel Hill, N.C. Bates, D. V., and Sizto, R. 1983. Relationship be- tween air pollution levels and hospital admissions in Southern Ontario, Can. J. Public Health 74:117-133. Becklake, M. R., and Permutt, S. 1979. Evaluation of tests of lung function for "screening" for early detection of chronic obstructive lung disease, In: The Lung in the Transition Between Health and Disease

412 (P. T. Macklem and S. Permutt, eds.), pp. 345-387, Marcel Dekker, New York. Bennett, A. E. 1981. Limitations of the use of hospital statistics as an index of morbidity in environmental studies, J. Air Pollut. Control Assoc. 31: 1276-1278. Chan-Yeung, S., Vedal, S., and Enarson, D. 1986. Asthma, asthma-like symptoms, chronic bronchitis and the degree of bronchial hyperresponsiveness in epidemiologic surveys (abs.), Am. Rev. Respir. Dis. 133:A155. Cohen, C. A., Hudson, A. R., Clausen, J. L., and Knelson,J. H.1972. Respiratory symptoms, spirom- etry, and oxidant air pollution in nonsmoking adults, Am. Rev. Respir. Dis. 105:251-261. Comstock, G. W., Meyer, M. B., Helsing, K. J., and Tockman, M. S. 1981. Respiratory effects of house- hold exposures to tobacco smoking and gas cook- ing, Am. Rev. Respir. Dis. 12:143-148. Detels, R., Rokaw, S. N., Coulson, A. H., Tashkin, D. P., Sayre, J. W., and Massey, R. J. 1979. The UCLA population studies of chronic obstructive respiratory disease. I. Methodology and compari- son of lung function in areas of high and low pollution, Am. J. Epidemiol. 109:33-58. Detels, R., Sayre, J. W., Coulson, A. H., Rokaw, S. N., Massey, F. J., Tashkin, D. P., and Wu, M. M. 1981. The UCLA population studies of chronic obstructive respiratory disease. IV. Respi- ratory effect of long-term exposure to photochem- ical oxidants, nitrogen dioxide, and sulfates on current and never smokers, Am. Rev. Respir. Dis. 124:673-680. Detels, R., Tashkin, D. P., Simmons, M. S., Carmi- chael, H. E., Sayre, J. W., Rokaw, S. N., and Coulson, A. H. 1982. The UCLA population stud- ies of chronic obstructive respiratory disease, Chest 82:630-638. Ferris, B. G. 1978. Epidemiology standardization project, Am. Rev. Respir. Dis. 118(Suppl.):7-53. Ferris, B. G., Speizer, F. E., Spengler, J. D., Dock- ery, D., Bishop, Y. M. M., Wolfson, M., and Humble, C. 1979. Effects of sulfur oxides and respirable particles on human health. Methodology and demography of populations in study, Am. Rev. Respir. Dis. 120:767-779. Ferris, B. G., Dockery, D. W., Ware, J. H., Speizer, F. E., and Spiro, R. 1983. The six-city study: examples of problems in analysis of the data, Envi- ron. Health Perspect. 52:115-123. Gunby, R. 1984. Intervention trial begins with young smokers, J. Am. Med. Soc. 252:2802-2803. Hill, A. B. 1965. The environment and disease: asso- ciation or causation, Proc. Roy. Soc. Med. 58: 295-300. Holguin, A. H., Contant, C. F., Noel, L. M., Mei, M., Buffler, P. A., and Hsi, B. P. 1983. Defining an asthmatic attack in epidemiologic studies, In: Inter- national Symposium on the Biomedical E.~ects of Ozone and Related Photochemical Oxidants (S. D. Lee, M. G. Mustafa, and M. A. Mehlman, eds.), pp. 527-538, Princeton Scientific Publishers, Inc., Princeton, N.J. Holguin, A. H., Buffler, P. A., Contant, C. F., Stock, T. H., Kotchmar, D. J., Hsi, B. R., Jenkins, D. E., Epidemiologic Studies of Oxidant Exposure Gehan, B. M., Noel, L. M., and Mei, M. 1985. The effects of ozone on asthmatics in the Houston area, In: Transactions of the APCA International Specialty Conference: Evaluation of the Scienti.tc Basis for Ozone/ Oxidant Standards (S. D. Lee, ed.), pp. 262-280, Air Pollution Control Association, Pittsburgh, Pa. Johnson, D., and Travis, J. 1979. The oxidative inactivation of human alpha-1-proteinase inhibitor, J. Biol. Chem. 254:4022. Kagawa, J., and Toyama, T. 1975. Photochemical air pollution: its effects on respiratory function of ele- mentary school children, Arch. Environ. Health 30:117-122. Keller, M. D., Lanese, R. R., Mitchell, R. I., and Cote, R. W. 1979a. Respiratory illness in house- holds using gas and electricity for cooking. I. Sur- vey of incidence, Environ. Res. 19:495-503. Keller, M. D., Lanese, R. R., Mitchell, R. I., and Cote, R. W. 1979b. Respiratory illness in house- holds using gas and electricity for cooking. II. Symptoms and objective findings, Environ. Res. 19:504-515. Khan, A. U. 1977. The role of air pollution and weather changes in childhood asthma, Ann. Allergy 39:397-400. Kucich, U., Abrams, W. R., Christner, P., Rosen- bloom, J., Kimbel, P., and Weinbaum, G. 1984. Molecular weight distribution of elastin peptides in plasmas from human non-smokers, smokers and emphysema patients, Am. Rev. Respir. Dis. 129:307. Kurata,J. H., Glovsky, M. M., Newcomb, R. L., and Easton, J. G. 1976. Multifactorial study of patients with asthma. 2: Air pollution, animal dander, and asthma symptoms, Ann. Allergy 37:398~09. Lippmann, M., Lioy, P. J., Leikauf, G., Green, K. B., Baxter, D., Morandi, M., and Pasternack, B. S. 1983. Effects of ozone on the pulmonary function of children. In: International Symposium on the Biomedi- cal Effects of Ozone and Related Photochemical Oxidants (S. D. Lee, M. G. Mustafa, and M. A. Mehlman, eds.), pp. 423 446, Princeton Scientific Publishers, Inc., Princeton, N.J. Medical Research Council, Committee on the Aetiol- ogy of Chronic Bronchitis. 1960. Standardized questionnaires on respiratory symptoms, Brit. Med. J. 2:1665. Michaeli, D., and Fudenberg, H. H. 1974. Antibodies to collagen in patients with emphysema, Clin. Im- munol. Immunopathol. 3:187-192. Monson, R. R. 1980. Occupational Epidemiology, CRC Press, Inc., Boca Raton, Fla. Morganstern, H., and Bursic, E. S. 1982. A method for using epidemiologic data to estimate the poten- tial impact of an intervention on the health status of a target population, J. Commun. Health 7:292-307. National Center for Health Services Research. 1980. Health Care Surveys Using Diaries, p. 286, Publi- cation No. PHS 80-3279, U. S. Department of Health and Human Services, National Center for Health Services Research, Hyattsville, Md. National Research Council. 1985. Epidemiology and Air Pollution, National Academy Press, Washington, D.C. Pearlman, M. E., Finklea, J. F., Creason, J. P., Shy,

Eddy A. Bresnitz and Kathleen M. Rest 413 C. M., Young, M. M., and Horton, H. S. M. 1971. Nitrogen dioxide and lower respiratory illness, Pe- diatrics 47:391-398. Richards, W., Azen, S. P., Weiss, J., Stocking, S., and Church, J. 1981. Los Angeles air pollution and asthma in children, Ann. Allergy 47:348-354. Rokaw, S. N., Detels, R., Coulson, A. H., Sayre, J. W., Tashkin, D. P., Allwright, S. S., and Massy, F. J. 1980. The UCLA population studies of chronic obstructive lung disease. III. Comparison of pulmo- nary function in three communities exposed to photochemical oxidants, multiple primary pollu- tants, or minimal pollutants, Chest 78:252-262. Samet, J. M., Tager, I. B., and Speizer, F. E. 1983. The relationship between respiratory illness in childhood and chronic air flow obstruction in adult- hood, Am. Rev. Respir. Dis. 127:508-523. Schlesselman, J. J. 1974. Sample-size requirements in cohort and case-control studies of disease, Am. J. Epidemiol. 99:381-384. Schoettlin, C. E., and Landau, E. 1961. Air pollution and asthmatic attacks in the Los Angeles area, Public Health Rep. 76:545-548. Selwyn, B. J., Stock, T. H., Hardy, R. J., Chan, F. A., Jenkins, D. E., Kotchmar, D. J., and Chap- man, R. S. 1985. Health effects of ambient ozone exposure in vigorously exercising adults, In: Trans- actions of the APCA International Specialty Conference: Evaluation of the Scientific Basis for Ozone/Oxidant Standards. (S. D. Lee, ed.), pp. 281-296, Air Pollu- tion Control Association, Pittsburgh, Pa. Shy, C. M., Creason, J. P., Pearlman, M. E., Mc- Clain, K. E., Benson, F. B., and Young, M. M. 1970a. The Chattanooga school children study: effects of community exposure of nitrogen dioxide. I. Methods, description of pollutant exposure and results of ventilatory function testing, J. Air Pollut. Control Assoc. 20(8):539-545. Shy, C. M., Creason, J. P., Pearlman, M. E. Mc- Clain, K. E., Benson, F. B., and Young, M. M. 1970b. The Chattanooga school study: effects of community exposure to nitrogen dioxide. II. Inci- dence of acute respiratory illness, J. Air Pollut. Control Assoc. 20(9):582-588. Speizer, F. E., Ferris, B., Jr., Bishop, Y. M. M., and Spengler, J. 1980. Respiratory disease rates and pulmonary function in children associated with NO2 exposure, Am. Rev. Respir. Dis. 121:3-10. Spengler, J. D., Ferris, B. G., Jr., and Dockery, D. W. 1979. Sulfur dioxide and nitrogen dioxide levels inside and outside homes and the implication on health effects research, Environ. Sci. Technol. 13:1276-80. Sterling, T. D., 1985. Letter to the Honorable Don Sundquist, U.S. House of Representatives. Com- ments on OTA Report on Smoking Related Death and Financial Costs (September 20). Sterling, T. D., and Weinkam, J. J. 1985. The "healthy worker effect" on morbidity rates, J. Oc- cup. Med. 27:477082. Sterling, T. D., and Weinkam, J. J. 1986. Extent, persistence, and constancy of the healthy worker or healthy person effect by all and selected causes of death, J. Occup. Med. 28:348-353. Sterling, T. D., Phair,J. J., Pollack, S. V., Schurnsky, D. A., and Degroot, I. 1966. Urban hospital mor- bidity and air pollution. A first report, Arch. Envi- ron. Health 15:362-374. Sterling, T. D., Pollack, S. V., and Phair, J. H. 1967. Urban hospital morbidity and air pollution. A second report. Arch. Environ. Health 15:362-374. Tager, I. B., and Speizer, F. E. 1975. Role of infection in chronic bronchitis, N. Engl.J. Med. 292:563-571. Tager, I. B., Weiss, S. T., Rosner, B., and Speizer, F. E. 1979. Effect of parental cigarette smoking on the pulmonary function of children, Am. J. Epide- miol. 110:15-26. U. S. Environmental Protection Agency. 1986. Air Quality Criteria for Ozone and Other Photochem- ical Oxidants, Report EPA/600/8-84/020EF, vol- ume V, Research Triangle Park, N.C. Ware, J. H., Dockery, D. W., Spiro, A., Speizer, F. E., and Ferris, B. G. 1984. Passive smoking, gas cooking, and respiratory health of children living in 6 cities, Am. Rev. Respir. Dis. 129:366-374. Whittemore, A. S., and Korn, E. L. 1980. Asthma and air pollution in the Los Angeles area, Am. J. Public Health 70:687-696. Woolcock, A. J., Peat, J. K., and Cullen, K. J. 1986. The effect of asthma on rate of decline of lung function (abs.), Am. Rev. Respir. Dis. 133:A156. Yanagisawa, Y., Nishmura, H., Osaka, F., and Ka- suga, H. 1986. Personal exposure and health effect relation for NO2 with urinary hydroxyproline to creatinine ratio as indicator, Arch. Environ. Health 41 :41-48. Zagraniski, R. T., Leaderer, B. P., and Stolwijk, J. A. J. 1979. Ambient sulfates, photochemical oxidants, and acute health effects: an epidemiological study, Environ. Res. 19:30~320.

Next: Biochemical and Cellular Interrelationships in the Development of Ozone-Induced Pulmonary Fibrosis »
Air Pollution, the Automobile, and Public Health Get This Book
×
Buy Paperback | $170.00
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

"The combination of scientific and institutional integrity represented by this book is unusual. It should be a model for future endeavors to help quantify environmental risk as a basis for good decisionmaking." —William D. Ruckelshaus, from the foreword. This volume, prepared under the auspices of the Health Effects Institute, an independent research organization created and funded jointly by the Environmental Protection Agency and the automobile industry, brings together experts on atmospheric exposure and on the biological effects of toxic substances to examine what is known—and not known—about the human health risks of automotive emissions.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!