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.
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 165 5 THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS INTRODUCTION This chapter uses several case studies to illustrate how epidemiologic strategies and study designs involving both long-term planning and flexible short-term responses can be applied to important current research questions. We have stressed that the development of protocols for epidemiologic research in air pollution requires substantial amounts of focused interdisciplinary planning and coordination. Therefore, the purpose of this chapter is not to write protocols for specific studies. Instead, for each selected problem dealt with here, we recommend approaches or guides to particular components of strategy and design development. The following components are addressed: â¢ Selection of a study method. â¢ Selection of study populations. â¢ Selection of exposure variables. â¢ Selection of effect variables. â¢ Selection of confounders to be considered. â¢ Special opportunities for studies to be considered. â¢ Limitations on the types of answers possible.
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 166 ACUTE RESPIRATORY INFECTION âDo any current patterns of exposure to air pollutants lead to increased frequency or severity of acute respiratory infections?â The study methods likely to be of use for this particular question are the prospective cohort method and the case-control method. Ecologic studies are not foreseeable, because widely reported data on the incidence of these conditions are lacking. The prospective cohort method requires the study of a relatively common end point, perhaps upper respiratory infections, as opposed to more serious lower respiratory infections (such as acute bronchitis and pneumonia). To achieve acceptable statistical power with the cohort method in studying less common events, large numbers of people must be followed for long periods, and that creates problems related to expense and the potential for loss of subjects to followup. However, small cohorts (panels) of children could provide a sensitive basis for studies. The high frequency of acute respiratory infection among preschool children (four to eight episodes per year) makes it possible to detect even small increases in the relative risk among exposed populations. Thus, an increment of 25% typically could be detected readily in a population of 200-300 children, whereas such small increments in chronic disease prevalence would require several thousand person-years of observation. Acute respiratory infection in children is readily detected and reported by most parents, but frequent contact with subjects or their parents is necessary, because the events are short-lived and easily forgotten. Methods to validate the presence of an acute respiratory infection might present themselves when our ability to measure immune responses in humans has advanced. Case-control methods might work nicely for the study of serious but infrequent infections. Acute bronchiolitis in infants, for example, can be caused by a number of viruses and is a frequent reason for hospital admission of children under the age of 2. Such infections have been linked with a later risk of developing chronic obstructive pulmonary disease (COPD).21 Cases of bronchiolitis can readily be accumulated by active pediatric hospital services, and the diagnoses can be
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 167 accurately verified, with viral antibody titers or cultures if necessary. Controls could be selected from infants admitted with diagnoses not conceivably related to air pollution. The youth of the subjects and the acute nature of the illness make it fairly easy to determine lifetime exposures, both inside and outside the home. Exposures are recent enough to permit monitoring in homes for specific pollutants or for particular emission sources, perhaps for a sample of the study group. The specific pollutants of interest in studies of respiratory infection might be nitrogen oxides, gas stove emission, passively received cigarette smoke, woodsmoke, acid aerosols, and recurrent summertime haze and ozone. If a small-cohort frequent-event approach is taken, it should be possible to use individual, rather than aggregate, monitoring data on pollutant exposure. With a large cohort, individual measurements could be made on a sample to âanchorâ or validate the aggregate data. Cigarette smoking would be a serious confounding variable in any studies in this field, and it is tempting to restrict studies to children or nonsmokers. But the exclusion of smokers from all studies would be a serious mistake, because cigarette smoke and air pollution might act synergistically to multiply the risk of developing a respiratory infection. Smokers would therefore constitute a useful hypersusceptible or sensitive population for study. Furthermore, nonsmokers might be subjected to some degree of risk through passive exposure. Besides the usual confounders, studies of acute respiratory infection will have to deal with unusual and difficult codeterminants of risk, including epidemics in the community, social class, season, number and age of siblings, and frequency of contact with other people. Some special opportunities for studies in this field can be identified. The utility of national survey samples--such as the National Health Information Survey--will be increased by improvements in routinely collected pollution surveillance data. Countries or regions with centralized health services can provide excellent opportunities for studies that require the recording of discrete events, such as hospital admissions due to respiratory infection.4 Finally, epidemiologic surveillance systems that track the response of respiratory infection rates in communities to fluctuations in
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 168 pollution could prove useful as a preliminary form of study. The use of small cohorts with individual exposure measurements could be a particularly sensitive means for studying the impact of small excursions above existing air quality standards. This is one of the circumstances in which epidemiologic studies might be able to give answers to threshold questions. The other study designs mentioned above are better suited to âany-associationâ or âdose-responseâ questions, depending largely on the extent of detail in the exposure data. Obviously, studies based on categorical data (e.g., frequency of use of gas stoves, woodstoves, or kerosene heaters) will be less suitable for the construction of dose-response curves than studies based on continuous numerical data. Many studies of acute respiratory infection will be able to determine which among numerous pollution sources is responsible for an effect. But few of them, even if conducted with individual measures of several pollutants, will easily be able to attribute an effect to a specific pollutant. CHRONIC OBSTRUCTIVE PULMONARY DISEASE âDo any current patterns of exposure to air pollutants lead to an excess risk of developing chronic obstructive pulmonary disease (COPD) or to exacerbation of existing COPD?â In many ways, this is the most difficult question for epidemiologic studies and therefore the one that requires the greatest care and ingenuity. The incidence of COPD is relatively low, so prospective studies that use clinical cases to determine the etiologic role of air pollution are unfeasible, for reasons of cost and time. That leaves two alternative strategies: studying the onset of COPD retrospectively and studying the development of early markers (physiologic or biochemical) that presage the clinical disease. All retrospective methods for studying this question suffer from the lack of reliable, precise data on individual lifetime exposure to ambient pollutants. A case-control study of COPD would be attractive if this difficulty could be overcome. Given lifetime residential histories obtained through interview of the subjects,
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 169 some index of cumulative exposure to pollutants could be constructed. It would probably require expert subjective assessment of pollution in various locales at various past times, inasmuch as measurements would not likely be available for the early years of the subjects' lives. This approach is not unlike the one common to retrospective studies of occupational cohorts, except that fewer variables need to be considered in reconstructing workplace exposures. The result would be a cumulative pollution exposure index that would be relatively coarse and insensitive to the role of individual pollutants. A case-control study of nonsmokers with COPD is worth considering because of the major etiologic role of smoking in most cases. The incidence of COPD in nonsmokers, however, is extremely low and would necessitate collaboration of several medical institutions for assembly of an adequate number of cases. (For example, in the Tucson cohort, the prevalence of COPD in nonsmoking adults was 2 per 1,000 persons.13) If air pollution causes COPD, it probably causes more cases among smokers than among nonsmokers; therefore, studies that yield information about the interaction of smoking and air pollution in the etiology of COPD would be extremely valuable. That brings us back to the prospective cohort method. It is the most effective means for addressing the question of COPD etiology, if costs can be kept down and the precision of exposure information and sample size can be maintained. It should soon be possible to develop predictive biochemical tests that capitalize on early injury to connective tissue in COPD. So far, the studies directed at detecting degradation products of elastin and collagen appear promising. The use of rapid decline of FEV1 or of other measures of pulmonary function as markers of COPD is desirable, in that rapid decline is far more common than clinically apparent disease; etiologic studies can therefore be shorter and cheaper and have satisfactory statistical power. Berry, Schlesselman, and others have shown that varying the number of subjects followed, the frequency of lung function measurement, and the length of followup makes it possible to maximize the chances of detecting an effect of a given magnitude without wasting resources on excess measurements.6 22 Table 1 shows the numbers of
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 170 TABLE 1 LONGITUDINAL OBSERVATIONS Number of Subjects Required in Followup Study of FEV1 in Each of Two Groups, to Detect Differences of 0.03 L/year at Significance Level of 0.05 and Power of 0.8a Length of Intervals Between Measurements Cross- Followup, 1 3 6 1 Beginning Sectional years Month Months Months Year and End Study on Only Subjects Previously Exposed for Length of Followup 1 226 429 530 530 530 4,356 2 56 95 128 153 153 1,089 3 36 50 64 78 84 484 4 32 38 45 53 59 272 5 30 33 37 42 48 174 6 29 31 33 37 42 121 7 29 30 31 34 38 89 8 28 29 30 32 36 68 9 28 29 30 31 34 54 10 28 29 29 30 33 44 a Reprinted with permission from Berry.4
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 171 subjects required for different followup periods. In fact, only small panels of subjects (fewer than 100 in the âexposedâ cohort) are required, if followup extends for 3 years and measurements are taken only once a year. Therefore, a large study can follow several cohorts simultaneously, each with a different pattern of pollution exposure. The inclusion of multiple cohorts mitigates the âsample size of twoâ problem encountered in some large cohort studies that use only two kinds of populations (see Appendix C). Perhaps more important, the use of small cohorts makes it easy to incorporate precise individual exposure measurements or to apply them to a substantial fraction of each cohort. Cross-sectional studies on the COPD question should not be ignored. The first slice in time looked at in a longitudinal study might be suitable for analysis as an independent cross-sectional study. This requires planning from the beginning, to ensure that estimators of past exposure are available at baseline. Furthermore, the last column in Table 1 shows that appreciable power can be achieved in a cross-sectional study, if the population consists solely of subjects who have had several years of exposure before testing. For example, in a two- population study (âexposedâ and ânonexposedâ), only 174 persons exposed for a period of 5 years would have to be studied to meet standard power requirements. As previous discussions have pointed out, however, this consideration accounts only for random errors; in the case of cross-sectional studies of chronic respiratory disease, the nonrandom error associated with retrospective exposure assessment would be a dominant concern and would limit the types of answers that could be expected. A research strategy on the COPD question would be enhanced by prior consideration of descriptive data on COPD incidence or mortality. Such data (although not perfectly reliable) could be used to locate âhot spots,â where the occurrence of pollution-related cases might be greatest. Figure 7 is a map of the United States showing the distribution of mortality from chronic bronchitis among white females.between 1965 and 1971.25 Mortality among women during this period would presumably be less influenced by smoking and occupational exposure than mortality among men. One potentially interesting area for the location of a study would be the Ohio Valley, where mortality rates were particularly high in a multistate area that included worrisome pollution sources. Given the limitations of mortality data for respiratory disease across geographic boundaries, a preliminary ecologic study would not be advisable. However, pilot studies in an area of high rates can explore the role of local artifacts in diagnosis or reporting or the impact of confounders, such as smoking. Case-control studies should follow, and then perhaps cohort studies.
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS FIGURE 7 Chronic bronchitis mortality among white females, 1965-1971. Reprinted from U.S. Department of Health and Human Services.25 172
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 173 Special opportunities for studying the role of air pollution in the etiology of COPD might be found. One important opportunity could be provided by the anticipated decreases in the emission of acid-precipitation precursors in some areas. The impact of sudden decreases on the respiratory health of a surrounding population--for instance, in a community near a midwestern power generating plant--would be a potentially fruitful subject for a cohort study. The growth of lung function in children offers another opportunity. Rates of increase in such indexes as FEV1 are sensitive indicators of toxic lung injury, according to recent studies on the effects of passive smoking and gas cooking.27 Children who fail to reach their highest potential plateau have less reserve lung function and are presumed to be at greater risk of dangerous decline in function. The opportunity afforded by studies of heavily exposed occupational groups or populations abroad should not be overlooked. Studies of the exacerbation of COPD call for a strategy entirely different from the strategy for COPD etiology, because exacerbation is considered acute, rather than chronic. Small cohorts (panels) of carefully screened subjects with chronic respiratory diseases can be followed longitudinally. In an approach very similar to that recommended for asthma or acute respiratory infection, the goal is to ascertain the determinants of the probability of exacerbation in a person over a given period. Personal monitoring can be incorporated into the design, to increase the precision of the study. The kinds of answers that epidemiologic studies might provide on the COPD question depend largely on the accuracy of the exposure information used. Further development of biologic markers that indicate the breakdown of connective tissue in the lung will increase
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 174 study sensitivity. Threshold questions on the etiology of COPD do not appear to be feasible within reasonable cost bounds, particularly for thresholds related to individual pollutants. In the future, the availability of better nationwide monitoring data on pollutants (collected since the 1960s) will provide a basis for accurate estimation of long-term cumulative individual exposure. Such data, when combined with data on individual effects, will permit a great increase in the overall power of almost any study method, except perhaps when outdoor exposure is a poor predictor of total exposure. ASTHMA âDo any current patterns of exposure to air pollutants lead to an increased risk of developing asthma or to exacerbation of existing asthma?â It is unlikely that air pollution plays a major role in the etiology of asthma, although some brief but intense exposures like the Yokkaichi episode (involving sulfuric acid mist) in Japan apparently can increase the prevalence of the disease in a community.12 Genetic factors, such as atopic status, are known to be important, so a good strategy might include twin studies to determine the relative contributions of genetic and environmental variables. Other studies related to asthma etiology will be constrained by the same kinds of factors as studies of COPD etiology. The relative rarity of new asthma cases makes case-control studies worthy of consideration, if problems with retrospective exposure assessment can be addressed. Most epidemiologic studies of air pollution and asthma have focused on acute exacerbation of the disease--definitely a subject in which recent developments can be applied to increase the sensitivity of studies. In the future, small-cohort studies of asthmatics will be able to take advantage of advances in statistical techniques (time series or individual regression), personal monitoring for estimation of short-term exposures, and portable devices, such as the peak flow meter, for frequent recording of pulmonary function by subjects themselves. Such individual monitoring will automatically include exposures to indoor pollutants and allergens
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 175 whose composition and concentrations have been shown to differ greatly from those prevailing outdoors. The preferred research strategy will follow several panels of asthma patients and controls over time. The panels might be of different socioeconomic, racial, or ethnic composition and might be subjected to different environmental conditions. As the number of subjects in a study increases, persons with similar dose-response patterns could be grouped, in recognition of the heterogeneity of the population labeled asthmatic. Some epidemiologic studies of air pollution and asthma will be complicated by recent observations that asthma morbidity, reflected in hospital admissions, appears to be increasing in the United States and other parts of the world.4 16 26 Air pollution is probably not responsible for these increases, which have occurred in spite of some notable therapeutic advances in the last 15 years. Nevertheless, until the reasons become clear, trend data on asthma morbidity in large populations and data covering long periods must be interpreted cautiously. There is every reason to believe that the types of studies described in this report will soon be able to answer questions about the relationship between the frequency of asthmatic attacks and relatively minor fluctuations in exposure to specific pollutants. LUNG CANCER âDo any current patterns of exposure to air pollutants lead to an excess risk of developing lung cancer, and does this exposure interact with cigarette smoking in producing excess risk?â Exploring the relationship between lung cancer and air pollution presents many of the same obstacles encountered in studying chronic respiratory diseases. The presence of multiple causes, dominated by smoking, and the existence of a long latency between initial exposure and the appearance of clinical disease head the list. In 1981, Doll and Peto suggested that a large case-control study of lung cancer be undertaken to evaluate the etiologic contribution of such factors as air pollution, passive smoking, and occupational exposures.9 This is an important suggestion and deserves consideration with high priority.
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 176 The proposed case-control study would require perhaps 10,000 cases (some 140,000 new cases occur in the United States each year1) and a similar number of controls. Enough smokers and nonsmokers should be included for both independent and interactive effects of air pollution to be discerned. The cases and controls should form a nationally representative sample, so that results can be used to estimate the impact of air pollution on cancer throughout the country, rather than in one area. If the sample is geographically broad enough, it should be possible to avoid bias due to matching by residential area. Control over bias and confounding must be very tight; for instance, failure to distinguish cigarette use in detail might mean ignoring variations far more powerful in determining risk than exposure to air pollution. The chief limitation of the study, however, would lie in the degree of detail possible in the exposure histories, which will limit the study's ability to find small relative risks due to specific pollutants and large relative risks that affect only a small proportion of cases. But it should not prevent the study from providing reasonable estimates of attributable risk, that is, the proportion of cases due to air pollution. The answer to this question is invaluable, because it provides perspective for preventive efforts. Other approaches to the study of air pollution and lung cancer are possible. Most prospective designs would fail, because of the relative infrequency of lung cancer in the general population. But the American Cancer Society is conducting its second massive prospective study of cancer in a cohort of 1.2 million persons; such an extraordinary cohort is large enough to detect small relative risks and yield actual incidence rates among the exposed groups while permitting adjustment for important confounders, such as smoking. The baseline interviews of the cohort have already been conducted. The initial questionnaire did not obtain information on the history of indoor or outdoor exposures outside the workplace. (Personal communication from S. Stellman, American Cancer Society) However, one question does identify those who have remained in the same residence for at least 10 years, and it could be used to define a nonmobile sample from whom histories of long-term residential exposures to air pollutants could be obtained. Even if only 5% of the entire cohort were so defined, that would amount to 60,000 long-term residents at risk of cancer for a
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 177 followup period of 6 years. Thus, with relatively little investment, the American Cancer Society study could be modified to explore hypotheses relating lung cancer and air pollution. The time available for seizing this opportunity is short, though, inasmuch as the next biennial followup interview of the cohort is approaching and the total planned followup period for the study ends in 1990. Many critics contend that traditional epidemiologic studies of cases of chronic disease are most limited by their after-the-fact, observational nature. It is true that these cases include some that might have been prevented; that is one reason why preclinical markers that point to an increased risk of developing lung cancer are of strategic interest. A variety of short-term laboratory tests have been developed to measure the interaction between environmental carcinogens and human genetic material in vivo. Most are performed on targets (such as peripheral lymphocytes and urinary nucleotides) that have only an indirect relationship to critical targets for lung cancer in bronchial cells. Others can measure the mutagenic potential of carcinogens in body fluids. Combinations of these tests could be applied to small cohorts that are well characterized with respect to patterns of exposure to air pollutants. Evaluation of some of the short-term tests as variables for epidemiologic studies is urgently needed, to confirm that they measure what they intend to and adequately control for confounding and to determine what positive results mean with regard to disease risk. The need for this information varies according to the specific test. Laboratory studies are the most practical means of evaluation. Until predictive value is ascertained, the finding that a particular group or community exposed to air pollution has a higher prevalence of markers of genetic damage than some controls must be interpreted as a warning of potential risk. EXPOSURE TO WOODSMOKE âDoes current exposure to woodsmoke cause or worsen acute or chronic respiratory disease, lead to acute or chronic changes in pulmonary function, or increase the risk of lung cancer?â
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 178 Descriptive studies that define the exposed populations and clarify the nature of woodsmoke exposures under various conditions should form an early part of the research strategy on this question. Ecologic studies comparing disease rates in communities with different patterns of wood use as fuel can be performed quickly and easily to form hypotheses and focus on populations at risk. Case-control studies like the one described in this chapter for acute bronchiolitis could easily incorporate information on woodstove use. The major thrust of the strategy, however, must be to establish cohorts exposed to woodsmoke for prospective study. Selecting the exposure variables to be measured presents several choices. Categorical descriptors, such as hours of woodstove use per day or concentration of woodstoves in a neighborhood, must be developed. Indoor and outdoor source contributions to individual woodsmoke exposure must be distingusihed. Woodsmoke is a complex mixture, so various surrogates or index of exposure must be chosen. These could include benzo[a]pyrene, carbon monoxide, nitrogen oxides, or respirable particles. Each would have its own advantages, which would depend on the end points under study. Modifying factors--such as humidity, stove tightness, temperature fluctuations, and house ventilation--might have to be considered. If carcinogenicity is of concern, biologic markers of exposure, such as urine mutagenicity in the Salmonella/microsome assay or detection of DNA- benzo[a]pyrene adducts in nonsmokers, might prove useful. We should also note the recent appearance of a small passive monitor for polycyclic aromatic compounds. (Personal communication from Tuan Vo Dinh, Oak Ridge National Laboratory) The effect measurements in woodsmoke studies could include symptom frequency, frequency of respiratory infection or asthmatic attacks, transient or persistent changes in lung function, changes in lung clearance, and markers of genetic damage in somatic cells. Reporting of acute respiratory symptoms in children might be particularly productive, as indicated by a recent study in Michigan.11 The increase in the use of wood as a fuel in the United states has been relatively recent (except in some distinct populations, such as the Navajo Indians); that makes it difficult to assemble groups that had long-term exposure for studies of chronic effects, such
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 179 as lung cancer. The greatest success could be achieved with small-panel, time- series studies. The existence of communities (e.g., in New England and the Rocky Mountains) that have experienced rapid increases in the amount of woodsmoke pollution suggests some possibilities for fruitful studies, particularly if local ordinances lead to precipitous declines in exposure. Many populations in developing countries that use wood or other biomass as a primary fuel in village stoves have lived under conditions involving extraordinary exposure. A study of women cooks in central India measured total daily exposures to benzo[a]pyrene equivalent to the smoking of 20 packs of cigarettes a day. For several years, epidemic cor pulmonale (terminal heart failure usually due to lung damage) has been noted among young women in Nepal who work at cookstoves.18 19 23 These situations are difficult to compare with the U.S. experience, and the primary usefulness of studying them would be in determining whether there could be any possible associations between exposure and a given health effect and in validating biologic markers. The introduction of relatively simple and acceptable devices to reduce pollution from village cookstoves, such as chimneys and vents, would provide an excellent basis for health studies and at the same time take action on what is already perceived as a serious and nearly worldwide public health problem.8 The types of answers to be expected from foreseeable epidemiologic studies on woodsmoke are limited. Such studies are not likely to identify the specific components of woodsmoke responsible for an effect. But dose-response information for an index or surrogate of woodsmoke could be obtained, particularly in relation to acute health effects. Epidemiologic studies on this question should make it possible to weigh the importance of woodstove emission against that of other prevalent sources of exposure. EXPOSURE TO NITROGEN DIOXIDE âDoes total exposure to nitrogen dioxide, from both indoor and outdoor sources, lead to or exacerbate acute or chronic respiratory diseases?â
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 180 It now appears that, except perhaps in a few selected areas, population exposures to nitrogen dioxide, NO2, both average and peak, are determined primarily by emissions from unvented indoor combustion sources. Therefore, future studies must consider the joint impact of indoor and outdoor sources of this pollutant. Characterization studies will refine the understanding of the categorical descriptors of NO2 exposure, but individual exposure monitoring of relatively small cohorts will still generally be necessary. Anchoring of aggregate exposure estimates in a larger cohort through intensive study of a selected sample is also possible. The appropriate monitoring equipment is available and awaits application in full-scale epidemiologic studies. The frequency of respiratory infection is an important end point in NO2 studies, to judge by the results of laboratory animal work. The role of NO2 exposure in serious respiratory infections among infants has already been addressed in connection with the case-control study discussed early in this chapter.27 Transient and persistent changes in lung function and symptom frequency are also of concern. In particular, new data on the effect of specific NO2 sources on rates of lung growth in children need confirmation and expansion. As mentioned in Chapter 2, results of Japanese studies indicate that urinary hydroxyproline excretion might provide a useful early indicator of pulmonary connective-tissue degradation due to NO2 exposure.15 Further evaluation of the hydroxyproline marker in population studies involving high exposure is needed, to clarify the sources of variability in these measurements. The development of categorical descriptors for NO2 exposure (e.g., questions concerning kerosene-heater or gas-stove use that delineate exposure magnitude) will advance the testing of hypotheses regarding NO2 and COPD in retrospective studies. The most precise data, however, will be gained through longitudinal studies of pulmonary function and measurement of individual exposures. Once again, studies on acute health effects should yield data that might eventually be used to identify a threshold exposure.
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 181 PERSISTENT EXPOSURE TO OZONE AND ACID AEROSOLS âDoes persistent exposure to ozone or acid aerosols (including sulfates) at high concentrations lead to an increased risk of developing chronic obstructive pulmonary disease?â Average ozone concentrations remain high in some urban areas of the country and are rising in others where the density of automobile traffic has been increasing dramatically. As mentioned in Chapter 3, persistently high concentrations of acid aerosols (largely sulfur species) are being reported in some areas. Populations exposed to sulfates at annual average concentrations of 1 Âµg/m3 to nearly 20 Âµg/m3 can be identified, and they offer ample contrasts for study. There is a concern over the potential influence of these patterns of exposure on the risk of developing COPD. Results of long-term animal exposure studies support this possibility and a recent longitudinal study of an ozone-exposed cohort in Los Angeles pointed to decrements in pulmonary function among several age groups (unpublished manuscript, R. Detels, University of California, Los Angeles). The most important knowledge can now be gained by prospective cohort studies, which alone will have the capacity to describe exposure with the needed accuracy. That statement presupposes that specially designed air monitoring procedures will be relied on in lieu of routinely collected data. Personal monitoring of a sample of the cohort must also be used to determine the extent of misclassification error involved in the use of central, aggregate exposure data. Cross-sectional studies, if large enough, will also be useful and should be derived automatically from baseline measures in longitudinal studies. Large cross-sectional studies of pulmonary function in long-term residents of polluted areas, compared with suitable controls, should be particularly useful. The availability of standard population values for pulmonary function that allow adjustment for smoking is a recent development of note (unpublished manuscript, A. Miller, Mt. Sinai School of Medicine). Exposures can be measured accurately only at or somewhat before the time of lung function measurement, so the impact of past and current exposures must be sorted out. Cohorts representing several different average exposures to ozone should be followed, including
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 182 some with widely contrasting exposures. Optimal sample size, measurement precision, and control of nonrandom error must all be considered, to ensure the feasibility of such a study. Cohort size and followup time must be kept to a minimum, so that loss to followup, a crucial source of error in previous studies, can be restricted. Frequent contact with cohort members is also essential for this purpose. EPISODIC EXPOSURE TO OZONE AND ACID AEROSOL HAZE âDoes episodic exposure to ozone or acid aerosols at high concentrations lead to excess morbidity from acute and chronic respiratory diseases?â In addition to the problem of persistent exposure to ozone and acid aerosols, there is an emerging regional pattern of summertime âhazeâ episodes in the Northeast. These episodes are characterized by high concentrations of ozone and acid aerosol that are formed in the atmosphere and transported over wide geographic areas. Rural areas are affected by this phenomenon, and the highest seasonal variations in these pollutants might be found there. That suggests that a successful strategy might involve study of acute responses to ozone and acid aerosol during winter and summer in small cohorts of rural subjects. The most sensitive population imaginable would be exercising asthmatics, but any group likely to be outdoors and active during the haze episodes would be a reasonable study group. In fact, Lippmann and co-workers used such considerations in designing their study of transient lung function changes in children at rural summer camps in Pennsylvania and New Jersey.14 Exacerbation of COPD in rural persons with this disease would also be an important topic for study, provided that likely exposures of persons remaining mostly indoors proved to be significant. The regional episodic haze problem is also ideal for the use of epidemiologic surveillance networks. Once in place, these networks could detect increases in adverse health effects on the basis of hospital admissions, physician visits, or medication sales. During the summer of 1984, one haze episode resulted in 7 consecutive days of ozone above the national standard in the area from Philadelphia to New York. Morbidity surveillance was
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 183 nonexistent, and mortality surveillance nearly so. The chief problem in such surveillance will be to separate out the effects of high heat and humidity, which invariably accompany these polluted air masses. The potential chronic effects of episodic ozone and acid aerosol will remain more difficult to study, in part because of the relatively recent appearance of the pollution pattern and because of the difficulty involved in separating persistent from episodic exposures. Again, rural areas that generally have low exposure to ozone and primary acid aerosol, except during the summer, might be preferable sites for studies. Persons without air conditioning who spend much time outdoors would be receiving the highest exposures. It might be feasible to couple the summertime acute studies on children with wintertime followup of the same children over a period of 2-3 years. The expense of establishing a new cohort could thus be reduced, exposures to the pollutants in the summer verified, and the acute-study measurements used as a baseline for observation of the rate of growth in pulmonary function. Given careful planning and adequate time and funding, currently available epidemiologic techniques could establish the acute impact of summertime excursions in ozone and acid aerosol on sensitive populations. EXPOSURE TO RADON âDoes exposure to radon and its progeny at current indoor concentrations increase the risk of developing lung cancer?â A survey of radon concentrations in soil, water, and indoor air in various parts of the United States would assist the development of an epidemiologic strategy on this question in several ways. First, although several areas with relatively high radon concentrations in buildings have already been identified, increasing the list would expand the options for selecting study populations and determine the potential size of the population at risk. For example, a national survey in Sweden has sampled radon concentrations in approximately 50,000 dwellings.24 Second, comparable data on radon concentrations in numerous areas might be used in a
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 184 preliminary ecologic analysis that combines radon exposure data with local data on lung cancer mortality. Third, if some descriptive studies were detailed enough, useful information on factors influencing the entry, exit, and transformation of radon in buildings could be generated. This would sharpen the focus in assigning groups or individuals to exposure levels and provide better categorical descriptors of exposure for questionnaires. For example, it has recently been demonstrated that radon progeny readily attach to particles found in cigarette smoke, thereby increasing and prolonging the effective dose experienced by both active and passive smokers.5 In addition to descriptive studies, case-control studies of lung cancer in areas with wide contrasts in radon exposure would be feasible. Relatively large numbers of cases would be required for adequate statistical power, because the risk is probably low, compared with that associated with smoking. Natural experiments, in which exposures within a community vary sharply according to the specific location of a house or the use of a different water supply or building materials, should be actively sought. One such study, comparing residence in wood versus stone houses, has already been conducted in Sweden.3 Aside from such geographically focused studies, information on radon and lung cancer might be derived from the large-scale case-control study or prospective studies mentioned earlier in connection with lung cancer. Such information will not be very substantial, because reliable, means for evaluating exposure of such large, heterogeneous populations are still lacking. Cohort studies of uranium miners and other types of miners have successfully demonstrated that radon and its progeny can cause lung cancer in humans. However, the least-exposed miner group that showed a significant excess of lung cancer still had cumulative lifetime exposures at least several times higher than that likely to be experienced by residentially exposed persons.17 Cohort studies (either prospective or historical-prospective) of residential exposure are therefore not likely to be large enough to detect the potential excess risk involved, unless short-term measures of risk that occur more frequently (such as markers of genotoxicity) are used. Sputum cytology analysis has been positive in studies of uranium miners2 and could be clinically
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 185 useful for screening these men, but morphologic abnormalities in exfoliated cells might occur so late as to have almost the same frequency as lung cancer itself. Chromosomal aberrations in peripheral blood lymphocytes, however, have also been observed in radon-exposed miners and in one study of home exposure and might prove to be a useful end point for residential cohort studies.7 20 Radon has decreased in mines in recent decades, and many miners' long-term exposures are now comparable with those of persons in some kinds of dwellings. Furthermore, some homes have been shown to exceed the permitted occupational radon concentration. The current followup of miners, if done carefully and with large enough cohorts, should be supported, to estimate risks in this dose range. Recent work in dose modeling has clarified some of the factors (including dust concentrations) involved in extrapolating radon dosage from the mine environment to the home.10 The most useful direct approach to the radon-lung cancer question is through the case-control method, so exposure measures will most often be retrospective and hence subject to considerable uncertainty. Nevertheless, radon sources are relatively stable emitters, and, assuming that changes over the years in exposure- modifying factors (such as the sealing of buildings) can be accounted for, retrospective estimates of exposure of long-term residents to radon are likely to be more accurate than those of exposure to most other air pollutants. Estimation of rough dose-response relationships should be possible, if a positive association can be found. The accuracy of such estimates will depend heavily on how well local exposure conditions have been characterized. Questions of attributable risk and interactive effects with smoking are important and should be addressed in these studies. REFERENCES 1. American Cancer Society. 1985 Cancer Facts and Figures, p. 8. Publication No. 5008-LE. Washington, D.C.: American Cancer Society, 1985.
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 186 2. Auerbach, O., G. Saccomanno, M. Kuschner, R.D. Brown, and L. Garfinkel. Histological findings in the tracheobronchial tree of uranium miners and non-miners with lung cancer. Cancer 42:483-489, 1978. 3. Axelson, O., C. Edling, and H. Kling. Lung cancer and residency: A case referent study on the possible impact of exposure to radon and its daughters in dwellings. Scand. J. Work Environ. Health 5:10-15, 1979. 4. Bates, D.V., and R. Sizto. Relationship between air pollutant levels and hospital admissions in Southern Ontario. Can. J. Pub. Health 74:117-122, 1983. 5. Bergman, H., C. Edling, and O. Axelson. Indoor radon daughter concentrations and passive smoking, pp. 79-84. In B. Berglund, T. Lindvall, and J. Sundell, Eds. Indoor Air. Stockholm, Sweden: Swedish Council for Building Research, 1984. 6. Berry, G. Longitudinal observations: Their usefulness and limitations with special reference to the forced expiratory volume. Bull. Physio-pathol. Respir. 10:643-655, 1974. 7. Brandom, W.F., G. Saccomanno, V.E. Archer, P.G. Archer, and A.D. Bloom. Chromosome aberrations as a biological dose-response indicator of radiation exposure in uranium miners. Radiat. Res. 76: 159-171. 8. DeKoning, H., K.R. Smith, and J.M. Last. Biomass Fuel Combustion and Health. Report No. EFP/84.64. Geneva: World Health Organization, 1984. 9. Doll, R., and R. Peto. The causes of cancer: Quantitative estimates of avoidable risks of cancer in the United States today. J. Nat. Cancer Inst. 66:1191-1308, 1981. 10. Harley, N.H., and B.S. Pasternak. A model for predicting lung cancer risks induced by environmental levels of radon daughters. Health Physics 40:307-316, 1981.
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 187 11. Honicky, R.E., S. Osborne, III, and C.A. Akpom. Symptoms of respiratory illness in young children and the use of wood-burning stoves for indoor heating. Pediatrics 75:587-592, 1985. 12. Kitagawa, T. Cause analysis of the Yokkaichi asthma episode in Japan. J. Air Pollut. Control Assoc. 34:743-746, 1984. 13. Lebowitz, M.D. Smoking habits and changes in smoking habits as they relate to chronic conditions and respiratory symptoms. Am. J. Epidemiol. 105:534-543, 1977. 14. Lippmann, M., P.J. Lioy, G. Liekauf, K.B. Green, D. Baxter, M. Morandi, B. Pasternak, D. Fife, and F.E. Speizer. Effects of ozone on the pulmonary function of children, pp. 423-446. In M.A. Melman, S.D. Lee, and M.G. Mustafa, Eds. Advances in Modern Environmental Toxicology. Vol. V. Princeton, N.J.: Princeton Scientific Publishers, 1983. 15. Matsuki, H., F. Osaka, H. Kasuga, and M. Sugita. Epidemiological study on the effects of smoking and air pollution using urinary hydroxyproline: creatinine ratio on the healthy schoolchildren and adults. Jap. J. Pub. Health 28:505-515, 1981. 16. Mitchell, E.A., and D.R. Cutler. Pediatric admissions to Auckland Oospital for asthma from 1970 to 1980. New Zeal. Med. J. 97:67-70, 1984. 17. National Research Council, Committee on Indoor Pollutants. Indoor Pollutants. Washington, D.C.: National Academy Press, 1981. 537 pp. 18. Pandey, M.R. Domestic smoke pollution and chronic bronchitis in a rural community of the Hill Region of Nepal. Thorax 39:337-339, 1984. 19. Pandey, M.R. Prevalence of chronic bronchitis in a rural community of the Hill Region of Nepal. Thorax 39:331-336, 1984.
THE APPLICATION OF EPIDEMIOLOGY TO SELECTED RESEARCH QUESTIONS 188 20. Pohl-Ruling, J., and P. Fischer. The dose-effect relationship of chromosome aberrations to Î± and Î³ irradiation in a population subjected to an increased burden of natural radioactivity. Radiat. Res. 80:61-81, 1979. 21. Samet, J.M., I.B. Tager, and F.E. Speizer. The relationship between respiratory illness in childhood and chronic air-flow obstruction in adulthood. Am. Rev. Respir. Dis. 127:508-523, 1983. 22. Schlesselman, J.J. Planning a longitudinal study: II. Frequency of measurement and study duration. J. Chron. Dis. 26:561-70, 1973. 23. Smith, K.R., A.L. Aggarwal, and R.M. Dave. Air pollution and rural biomass fuels in developing countries: A pilot village study in India and implications for research and policy. Atmos. Environ. 17:2343-2362, 1983. 24. Swedjemark, G.A. Indoor measurements of natural radioactivity in Sweden. SSI 1979-026. Stockholm: National Institute for Radiation Protection, 1979. 25. U.S. Department of Health and Human Services. Chronic bronchitis, white females, 1965-1971. In T.J. Mason, J.F. Fraumeni, Jr., R. Hoover, and W.J. Blot, Eds. An Atlas of Mortality from Selected Diseases. NIH Publication No. 81-2397. Washington, D.C.: U.S. Department of Health and Human Services, National Institutes of Health, 1981. 26. U.S. National Center for Health Statistics. Detailed Diagnosis and Surgical Procedures for Patient Discharge from Short Stay Hospitalization. Publication No. (PHS) 85-11743. Hyattsville, Md.: U.S. Department of Health and Human Services, National Center for Health Statistics, 1984. 278 pp. 27. Ware, J.H., D.W. Dockery, A. Spiro, III, F.E. Speizer, and B. G. Ferris, Jr. Passive smoking, gas cooking, and respiratory health of children living in six cities. Am. Rev. Respir. Dis. 129:366-374, 1984.
189 Chapter 6 Conclusions and Recommendations