Document 4v7DOGZ1E1QgggbqoXmGyGp9x

90.00 (DO- and Public additional cy, 2007 E. SOPPOrted : 410-955- KanawhaCounty, West Virginia, is one of the largest chemical manufacturingcenters in the United States. In 1988, a survey of respiratory and irritant symptoms was administered to all third grade to fifth grade children attending 74 elementary schools in Kanawha County, and concentrations of 15 volatile organic compounds were measured at each school. Exposures were characterized by school location, by the sum of the concentrations of five petroleum-related compounds, and by the sum of the concentrationsof 10 cornpounds more specific to industrial processes. Childrenenrolled in schools within the valley had higher rates of doctor-diagnosed asthma (odds ratio (OR)= 1.27,95%confidenceinterval(CI) 1.09-1.48)and a higher score on a composite indicator of five chronic lower respiratory symptoms (OR = 1.13, 95% CI 1.02-1.26) than children who were enrolled in schools outside of the valley. The incidence of chronic respiratory symptoms was also positively associated with the concentrations of volatile organic compounds. The estimated change in the odds ratio for chronic lower respiratory symptoms associated with a 2-pg/m3change in process-relatedcompounds was 1.08 (95% CI 1.02-1.14). No consistent pattern was found between acute irritant symptoms in the 2 weeks preceding questionnaire administration and either proximity to industry or exposure to volatile organic compounds. The authors conclude that exposure to volatile organic compounds, including emissions from chemical rnanufacturing plants, is associated with increased rates of chronic respiratory symptoms characteristic of reactive airways. Am J Epidemiol 1993337:1287-1301. air pollution; child; environmental exposure; respiratory tract diseases The Kanawha Valley region of West Vir- rounding communities, is one of the largest ginia, which includes Charleston and its sur- centers of chemical manufacturing in the Received for publication June 18, 1992, and in final form March 22. 1993 Abbreviations CI. confidence interval, OR.odds ratio, OVM, organic vapor monitor ' Department of Biostatistics.Harvard School of Public Health. Boston, MA Department of Environmental Health Harvard School Of Public Health, Boston, MA Departmentof Epidemiology, HarvardSchool of Public Health, Boston, MA 'Channing Laboratory, Brigham and Women's Hospi- tal. Harvard Medical School,Boston, MA. Pulmonary Division. University of New Mexico School of Medicine, Albuquerque, NM Departmentof Medicine,Marshall University, Huntington, WV 1287 1288 Ware et al United States. The industrial facilities are concentrated along the floor of a narrow valley, in many cases close to residential neighborhoods. Public concern about health risks increased sharply after the widely publicized disaster in Bhopal, India, in 1984. Methyl isothiocyanate, the compound released in the disaster, is also produced in Institute, a small community located within the Kanawha Valley. An accidental release of aldicarb oxime from the Institute plant in 1985 stimulated additional concern. In December 1986, Harvard University and the National Institute for Chemical Studies, a private nonprofit institution, began an Environmental Protection Agencysponsored study of air pollution and health in the Kanawha Valley. A preliminary phase combined measurement of air pollutants with an evaluation of the feasibility of conducting an epidemiologic study in the Kanawha Valley area (1). A panel of scientific experts and community representatives was convened to discuss community concerns, health endpoints, and epidemiologic strategies for assessing the health effects of community exposure to ambient air pollutants. The panel identified cancer and respiratory/ irritant effects as areas of potential health impact and community concern, and recommended that an epidemiologic study be undertaken combining ambient monitoring of chemical air pollutants with a survey of respiratory, allergic, and other symptoms and illnesses in preadolescent schoolchildren (2). The Kanawha County Health Study was conducted to test the hypothesis that residence close to chemical manufacturing facilities and/or exposure to volatile organic compounds would be associated with increased rates of both acute and chronic re- spiratory symptoms. This report characterizes the incidence of respiratory and other symptoms in Kanawha County schoolchildren and investigates the associations between symptom rates and these measures of chemical air pollution exposure. An assessment of acid aerosol pollutant concentrations and their association with rates of respiratory symptoms in children living in Kanawha County and in six other US communities is reported elsewhere (3). MATERIALS AND METHODS Site characteristics Kanawha County is transected by the Kanawha River for about 82 km from the southeast to the northwest, with a narrow valley between 0.8 and 2.5 km wide and about 120 m deep. The urban areas are arrayed as a series of small towns on alternating sides of the river. The major chemical manufacturing industries are located on the valley floor in close proximity to the residential areas, which include Charleston and the surrounding communities. Meteorologically,the Kanawha Valley urban area is dominated by temperature inversion conditions during about two thirds of the total hours throughout the year, with at least an 18-m inversion on over 90 percent of the sunrises (4).Wind speeds within the valley are typically half of those above the valley walls, with an average valley wind speed of 9 km per hour in the winter and 6 km per hour in the summer. The public school system is county-wide, and nearly 95 percent of all children between the ages of 7 and 13 years are enrolled in a public school. The 75 elementary schools are typically small, and most children attend a neighborhood elementary school. Reprint requests to Dr James H Ware, Department of Biostatistics.Harvard School of Public Health. 665 Huntington Ave , Boston, MA 02115 This study was supported in part by the National Institute for Chemical Studies, Charleston, West Virginia Or Cucas M Neas was supported in part by NationalInstitutes of Health NationalResearchService Awards ES-07069and HL-07427 The authors thank Andrew I Damokosh for his assistance with data processingand model development Geographic indicators of exposure A prior hypothesis was that exposure to chemical industry emissions would be inversely related to the distance of schools from the plants. Schools were chosen as a surrogate for residential location because the elementary schools in Kanawha County are community-based and serve very small areas (usual1, closest 100-2( industr the co: was de: each sc nine m We c ity-to-i r gion I ) : near (rt 4). The sch 001s valley 1 '1-64\ & 6 t 0 - FIGURE of the scr haracternd other hoolchilions bem r e s of n assessbncentraes of reiving in JS com- by the ?-om the narrow de and eas are )n alterlernical :on the -esidenind the iley urore inthirds r, with xrcent un the me the wind and 6 -wide, tween d in a :hook lttend ire to x! in- hook asa ;e the Y are areas Health Effects of Ambient Volatile Organic Compounds 1289 (usually of less than a 2-km radius). The closest residential neighborhoods are within 100-200 m of the fences of many chemical industry complexes or within 300-400 m of the complex center. Proximity to industry was defined as the smallest distance between each school and the geographic centers of nine major industrial complexes. We divided the schools into four proximity-to-industry regions: in valley, near (region t); in valley, far (region 2);out of valley, near (region 3); and out of valley, far (region 4). The "in valley, near" region included schools within 2 km up-valley or across the valley from a major industrial facility or within 3 km down-valley of a major facility. The second region included all other schools located on the valley floor. The third group included schools located outside of the valley but within 4 km of the valley floor. The final region included schools more than 4 km from the valley floor. Figure 1 shows the locations of the schools by their proxirnityto-industry group. Selection of volatile organic compounds Volatile organic compounds were chosen as the primary focus of monitoring because 1) previous studies in the Kanawha Valley 0 4N a lndustrlal Sltc Schools by Region D Out-of-valleyFar & Out-af-valleyNear & In-valley FW b In-valleyNear Marmct Grove a i miles A 'f at FIGURE 1. Locationsof 75 public elementary Schools and 11chemical manufacturingplants, identified by proximity Ofthe schools to the industrial sites, Kanawha County, West Virginia, 1988. t -v t: showed that these compounds were present in the outdoor air (1, 5 , 6); 2) these compounds are widely used in chemical manufacturing and are found in other US communities (7); and 3) some of the compounds can induce toxic effects at low concentrations (8, 9). Fifteen compounds were selected through a multistage process. First, the investigators identified 5 1 pollutants emitted in quantities in excess of 50,000 pounds (22,700 kg) per year (10).Several compounds that had been monitored previously but did not meet the 50,000 pound selection criterion were added to this list. The list was then restricted to chemicals for which toxicity ratings were available (1 1-1 3). The last step was to use vapor pressure data to select compounds that could be measured accurately using available technology. Table 1 shows the final list of compounds selected for monitoring. These compounds can serve as markers for other chemical air pollutants released at the same location. Environmental measurementsof volatile organic compounds The study plan called for one period of sampling at each elementary school. An initial round of sampling during the summer of 1987 was affected by the theft of samplers and by sample contamination by a laboratory reagent. The monitoring was repeated between October 20 and December 28, 1988, at the same time that the questionnaires were administered. Monitors were placed outside of each of the 75 public elementary schools for an 8-week period of continuous sampling. In this second round, no samplers were lost and no problems were encountered in the laboratory. Only the measurements from the second round of sampling were used in the health effects analysis. A second monitor was placed at seven of the sitesto collect data on sampler reliability. The data from these seven pairs and from five paired samples collected during the third TABLE 1. Descriptive statistics for 15 volatile organic compounds measured at 74 elementary schools, Kanawha Valley, West Virginia, November and December 1988 Volatile organic compound Maximum Median Mean Standard deviation Petroleum-related compounds Toluene m, p-Xylene Benzene o-Xylene Decane 117.4 25.10 6.89 7.45 1.85 4.8 3.06 3.02 1.13 0.66 9.7 4.13 3.18 1.45 0.77 17.2 3.78 1.22 1.09 0.41 Total 154.40 13.12 19.19 22.27 Process-related cornpounds 1 , l ,1-Trichlorethane Carbon tetrachloride 1-Butanol Chloroform Perchloroethylene Methyl isobutyl ketone 1,2-Dichloroethane Styrene Mesityl oxide 2-Ethoxyethyl acetate Total 2.03 1.45 2.55 7.92 2.24 0.90 0.60 0.23 0.46 0.55 13.02 0.96 0.62 0.58 0.56 0.46 0.42 0.26 0.12 0.02 0.02 4.20 0.98 0.64 0.62 0.88 0.57 0.42 0.26 0.12 0.08 0.04 4.61 0.22 0.19 0.38 1.11 0.37 0.18 0.13 0.02 0.11 0.07 1.72 phase of the k cussed below samples were oratories for variation. Th this report. A os01 acidit!. ozone, and 9 sured betwee several sampi Valley (6, 14 Although I for monitorir integrated sar por monitor effective me concentratior period (1 5). V developed foexposures, th sure ambien them for lor. plers work pounds thro: carbon. A p vironmental analyzed thc selected-ionphy with m2 the ambient vidual volat: The anal? a division o!' composite r general expc pounds (tab1 tor was the compounds stations and petroleum fi industry ( 19 goup inch lated to che are minima; eration. Health outc The ques: ments used ents of volatile ir one period of v. school. An iniing the summer heft of samplers on by a labora7g was repeated December 28, it the questionMonitors were e 75 public ele.week period of second round, 5 problems were :ory. Only the -ond round of health effects ced at seven of :pler reliability. 3airs and from juring the third Standard deviation 17.2 3.78 1.22 1.09 0.41 22.27 0.22 0.19 0.38 1.11 0.37 0.18 0.13 0.02 0.11 0.07 1.72 Health Effectsof Ambient Volatile Organic Compounds 1291 phase of the Kanawha Valley Study are discussed below. An additional seven pairs of samples were analyzed at two different laboratories for assessment of interlaboratory variation. These data are also discussed in this report. Ambient concentrations of aerosol acidity, particles, nitrogen dioxide, ozone, and particulate sulfate were measured between May and August of 1986 at several sampling locations in the Kanawha Valley (6, 14). Although it is not the standard method for monitoring volatile organic compounds, integrated sampling with passive organic vapor monitor badges provided the only costeffective method for measuring ambient concentrations at 75 sites over an extended period( 15). While this type of dosimeter was developed for measurement of occupational exposures, the samplers can be used to measure ambient concentrations by exposing them for longer periods (16-18). The samplers work by diffusion of volatile compounds through air and their sorption onto carbon. A private laboratory (Clayton Environmental Consultants, Novi, Michigan) analyzed the data from the badges using selected-ion-monitoring gas chromatography with mass spectrometry and quantified the ambient concentrations of the 15 individual volatile organic compounds. The analyses reported here were based on a division of these 15 compounds into twc composite measures of the community's general exposure to volatile organic compounds (table 1). The first exposure indicator was the sum of five petroleum-related compounds which are released from service stations and motor vehicles as well as the petroleum feedstocks of the petrochemical industry (1 9-2 l), while the second indicator group included 10 compounds that are related to chemical production processes but are minimally related to motor vehicle operation. Health outcome measures The questionnaire was based on instruments used in the Six Cities Study of Air Pollution and Health (22) and included the standard respiratory health questions from the American Thoracic Society questionnaire (23). For this study, questions were added about acute symptoms that had potentially been caused by imtant air pollutants. Questionnaires were distributed in three rounds, and 97 percent of the returned questionnaires were completed between October 15 and December 30, 1988. The schools represented in each round of distribution were geographically representative of the Kanawha Valley. A field surveyor visited each classroom to engage the children's interest and explain the survey. The actual distribution and collection of the questionnaires was done by teachers. The chronic respiratory symptom questions obtained information on chronic cough, chronic phlegm, bronchitis, persistent wheezing, and attacks of shortness of breath with wheezing that had taken place during the previous year. Two composite indicators of chronic respiratory conditions were created from these individual symptoms. Lower respiratory symptoms were considered present if the child was reported to have any of the five chronic symptoms. Chronic lower respiratory response was considered present if the child was reported to have either chronic cough, persistent wheezing, or a physician's diagnosis of asthma. A second set of questions focused on acute eye, nose, throat, and chest symptoms potentially effected by exposure to imtants. These questions asked about the occurrence of these symptoms during the 2-week period prior to the date on which the questionnaire was completed. Analytic and statistical methods The relative reliability of the measurements of volatile organic compounds was estimated by the relative standard deviation, expressed as a percentage, defined as +100 x J;i x lx, - .uzl/(x, XZ), where xI and x z are the first and second values of a paired sample. We report the 1292 Ware et al. mean and standard deviation of the relative standard deviation for sets ofpaired samples. The sample included all children enrolled in the third, fourth, and fifth grades of all general purpose public elementary schools in Kanawha County. Children placed in special instructional units were excluded unless they resided in the community in which their school was located. The analysis examined the association between the health outcomes and either the geographic variables or the environmental exposure measurements, while controlling for parental smoking, familial socioeconomic status, and other potentially confounding variables such as the child's age and characteristics of the child's home that were predictive of indoor air quality. The socioeconomic status indicator was a composite variable based on parental education, occupation, and head-of-household status (details available from the authors upon request). The data were analyzed by standard multiple logistic regression methods (2426). Results of individual logistic regression analyses were summarized by the estimated adjusted odds ratios and their 95 percent confidence intervals. RESULTS Concentrations of volatile organic compounds Because data from one school were lost during laboratory analysis, analyses were based on a total of 74 observations. Values below the limit of detection for each volatile organic compound were assigned a value equal to the limit of detection. Table 1 gives the maximum, median, and mean values and standard deviations for each of the 15 volatile organic compounds and for the summary measures of petroleum-related and process-related volatile organic compounds. Correlation between the petroleumrelated compounds and the process-related compounds was 0.73. Note that for toluene there was a large standard deviation and a substantialdifferencebetween the mean and median values for the 74 sites. The mean values for both petroleum-related and process-related compounds tended to be higher within the valley, particularly at the in-valley-near schools, than outside of the valley (table 2). The trend did not achieve statistical significance for the petroleumrelated compounds because of the large var- TABLE 2. Composite levels of volatile organic compounds by the proximity to industry of 74 elementary schools, Kanawha Valley, West Virginia, November and December 1988 Proximity-to-industryregion. No. of schools Median (dm? Mean (dm3) Standard error Pt Petroleum-related compounds$ In valley, near In valley, far Out of valley, near Out of valley, far 10.375 16 19.71 27.53 8.53 26 12.05 18.98 3.76 14 10.55 15.50 3.85 18 11.01 14.94 4.13 Above category excluding toluene In valley, near In valley, far Out of valley, near 16 11.62 13.11 1.71 26 7.34 9.28 0.90 14 6.51 7.87 1.25 Out of valley, far 18 6.86 8.00 1.28 0.02 Process-related compounds In valley, near In valley, far Out of valley, near Out of valley, far 16 5.00 26 4.76 14 3.89 18 3.77 See text for detailed description of regions. +t Probabilityassociatedwith the F statistic in one-way analysisof VarianCe. See table 1. 5.68 4.85 3.83 3.91 0.57 0.33 0.26 0.24 <0.005 :ile organic e school were lost sis, analyses were xervations. Values 3n for each volatile assigned a value xion. Table 1 gives and mean values for each of the 15 inds and for the petroleum-related tile organic com:en the petroleumhe process-related te that for toluene d deviation and a feenthe mean and ! sites. The mean cum-related and b tended to be 7articularly at the in outside of the i did not achieve - the petroleum- e of the large var- of 74 elementary j Pt x.375 0.02 <0.005 - Health Effects of Ambient Volatile Organic Compounds 1293 iability induced by a few very large toluene measurements. When the toluene data were excluded from the petroleum-related group, the trend in concentrations remained and became statistically significant. An analysis of 12 paired samples gave intralaboratory mean values for the relative standard deviation of 18.8 k 5.3 percent (standard error) and 15.1 k 3.6 percent for the process-related and petroleum-related compounds, respectively. An interlaboratory comparison of seven paired samples gave mean values of 23.1 k 4.5 percent and 16.1 f 2.3 percent for the process-related and petroleum-related compounds, respectively. Health characteristics Completed health questionnaires were obtained for 8,549 children, a return rate of 97 percent. Because of the restrictions noted above, 753 children (9 percent) were excluded from the analysis data set. The 7,796 children included in the analysis data set were approximately equally distributed across the third, fourth, and fifth grades. The percentages of children belonging to the four proximity-to-industry regions ranged from 15 percent to 37 percent. Information on race, familial socioeconomic status, parental smoking status, and other characteristics of the children's home environment is shown in table 3. Symptom rates for the entire sample and for the four proximity groups are displayed in table 4. The latter rates were adjusted for familial socioeconomic status and parental smoking. The cumulative incidence of chronic respiratory symptoms ranged from 8 percent for bronchitis or chronic phlegm to 14 percent for persistent wheezing. The cumulative incidence of chronic cough was 9 percent, and that of attacks of shortness of breath with wheezing was 1 1 percent. For the 2-week period preceding administration of the questionnaire, a small proportion of children were reported to have had eye symptoms (7 percent) or lower respiratory symptoms (10 percent), while a much larger proportion were reported to have had nasal symptoms (25 percent) or throat symptoms (21 percent). Potential confounders Preliminary analyses explored associations with potentially confounding factors. Because of the narrow age range of the children, child's age was not associated with any of the health outcomes. The dampness variable showed a strong relation to health out- TABLE 3. Characteristicsof children by the proximity to industry of 74 elementary schools, Kanawha County, West Virginia, 1988* Sample characteristic In valley Near Far No. of schools 16 No. of children 1,566 YOof children 20.6 Male sex 49.4 Race White Black 90.2 8.0 Other 1.8 Passive smoke exposure 50.7 Socioeconomic status High 12.1 Medium high 21.7 Medium low 30.5 Low 35.8 Gas cooking stove in the home 40.0 Exposure to mold, mildew. or water damage 0 24.7 Ay data shown are percentages unless otherwise indicated. 26 2.768 36.5 51.1 84.6 13.2 2.3 52.6 14.0 19.4 30.0 36.6 38.3 22.4 Out of valley Near FW 14 1.122 14.8 52.3 18 2,136 28.1 50.3 91.9 4.9 3.2 47.2 97.5 1.1 1.4 48.4 23.3 19.8 24.9 32.1 35.0 13.9 17.5 25.7 42.9 30.8 25.5 21.3 1294 Ware et aJ. TABLE 4. 74 ~ m Adjustdo curnulahe incidence (%) of selected health OUtCOmeS bY the proximity to industry of O X hnd S ~, Kan~awha County, west Virginia, 1988 W t h outarne Total In valley Near Far Out of valky Near Far Trend test p Chronc lower respiratory symptoms Chronic coughtd Chronic phlegm? Bronchitist persistent wheezingt I $ Attacks of shortness of breath with wheezingt Physician'sdiagnosis of asthma$ composite indicators 8.7 7.7 8.2 14.0 11.1 10.1 9.7 8.2 8.0 14.1 11.6 10.4 8.6 8.5 8.9 15.1 11.9 11.2 Lower respiratory symptoms Chronic lower respiratory 26.1 26.6 27.3 response Acute symptoms 20.9 21.6 22.1 Eyes Nose Throat Lower respiratory symptoms 7.3 25.4 21.2 9.9 7.8 24.8 21.1 11.2 8.1 24.9 20.5 9.4 Adjusted for parentalsmoking and familial socioecommic status. t HealthOutcome included in lower respiratory symptoms compositeindicator, $ HealthOutcome includedin chronic lower respiratory responseampositeindimtor, 7.4 6.4 8.4 12.2 10.6 9.5 23.8 18.6 7.3 24.2 19.9 8.7 8.7 7.1 7.5 12.7 9.8 8.5 25.1 19.7 5.7 26.6 22.6 10.1 0.25 0.06 0.37 0.04 0.03 <0.01 0.09 0.04 <0.01 0.23 0.22 0.43 comes but no association with proximity to industry. Race and type of home heating fuel showed no association with health outcomes in these data. Thus, the final model did not adjust for these variables. Although both parental smoking and familial socioeconomic status had only weak associations with health outcomes and proximity to industry, subsequent analyses were adjusted for these variables. Incidence of respiratory symptoms and geographic location After adjustment for parental smoking and familial socioeconomic status, the reporting of health outcomes differed between the four proximity-to-industry regons (table 4). The trend test is a I df test for trend in symptom rates across the four proximity-toindustry groups, with the groups being assigned values of 1 to 4. Significant trends across the four proximity-to-industry regions were observed for acute eye symptoms and for asthma-related responses such as a physician's diagnosis of asthma, persistent wheezing, and attacks of shortness of breath with wheezing. Chronic cough and chronic phlegm did not show statistically significant trends. For all of the health outcomes, the greatest contrasts appeared to be between the two invalley regions and the two out-of-valley regions. A physician's diagnosis of asthma was more prevalent among children attending schools within the valley than among children attending schools outside of the valley (odds ratio (OR) = 1.27, 95 percent confidence interval (CI) 1.09-1.48). A significant in-valley effect was also noted for the composite indicators of lower respiratory symptoms (OR = I . 13, 95 percent CI 1.02-1.26) and chronic lower respiratory response (OR = 1.17, 95 percent CI 1.04-1.31), as well as for chronic phlegm, persistent wheezing, and attacks of shortness of breath with wheezing. After this in-valley effect was controlled, the near/far dichotomy was not associated with any health outcome. Among the acute symptoms, only acute eye imtation was associated with attending a school within the valley (OR = 1.30, 95 percent CI I .09- 1.56). Symptom rates and environmental measurements Most of the chronic respiratory symptom variables were positively and significantly associated with pollutant concentrations, TABLE5. Adjusted. odd change in volatile organic c and health out Health ou: Chronic lower resp symptoms Chronic cough7 Chronic phlegmBronchitist Persistent whee Attacks of short, wlth wheezinc Physaan's diagno Composite indlcatc Lower resprator Chronc lower re response Acute symptoms Eyes Nose Throat Lower respirato Adjusted for parental smc +t Healthoutcome included Health outme included and the results were summary measures (t. the summary exposur eled as the linear effec exposure (in pg/m3). been rescaled separatl reflect a change in e. equal to the interqi summary exposure factor was 10 pg/n related compounds process-related c o w An increased CUI lower respiratory sq i` with both a lO-pg/m related compounds ( t m3 change in proc: (OR = 1.08). A simi served for the chron sponse composite 3 chronic respiratory s` all of the confidenct vidual symptoms i n L icance at the 95 pe The prevalence of . of asthma was aSSOc *proximity to industry of valley ~ Far Trend test p 8.7 7.1 7.5 12.7 9.8 8.5 25.1 19.7 5.7 26.6 22.6 10.1 0.25 0.06 0.37 0.04 0.03 co.01 0.09 0.04 co.01 0.23 0.22 0.43 itcomes, the greatest between the two inwo out-of-valley renosis of asthma was children attending y than among chil)utside of the valley *. 95 percent confi- 1.48).A significant noted for the com'r respiratory symprcent CI 1.02-1.26) spiratory response 1 CI 1.04-1.31), a~ n, persistent wheezness of breath with illey effect was conlotomy was not asi outcome. Among f y acute eye imtaattending a school I .30,95 percent CI {ironmental spiratory symptom and significantly I t concentrations, Health Effects of Ambient Volatile Organic Compounds 1295 TABLE 5. Adjusted* odds ratios (ORs) and 95 percent confidence intervals (CIS)for an interquartile change in volatile organic compound exposure as measured at 74 elementary schools, by type of compound and health outcome, Kanawha County, West Virginia, 1988 Healthoutcome Petroleum-related compounds (A 10 r9/m? OR 95% CI Process-related compounds (a 2 rgIm3) OR 95% CI Chronic lower respiratory symptoms Chronic cought ,$ Chronic phlegmt Bronchitist Persistent wheezingt,$ Attacks of shortness of breath with wheezing? Physician's diagnosis of asthma$ Composite indicators Lower respiratory symptoms Chronic lower respiratory response Acute symptoms Eyes Nose Throat Lower respiratow symptoms 1.03 1.00-1.06 1.03 0.99-1.06 1.05 1.02-1.08 1.01 0.99-1.04 1.04 1.01-1.07 1.05 1.02-1.08 1.05 1.02-1.07 1.03 1.01-1.05 0.99 0.95-1.03 1 .oo 0.98-1.02 1.01 0.99-1.04 1.03 1.00-1.06 Adjusted for parentalsmoking and familial sodoeconomicstatus. t Health outcome includedin lower respiratwysymptoms compositeindiitw. $Healthoutcome includedin chronic lower respiratoryresponsecomposite indicator. 1.08 1.06 1.03 1.04 1.05 0.99 1.08 1.07 1.08 0.99 1.02 1.02 1 .OO-1.18 0.97-1.17 0.94-1.13 0.97-1.12 0.97-1.13 0.91-1.08 1.02-1.14 1.01-1.14 0.99-1.19 0.93-1.05 0.96-1.09 0.93-1.11 and the results were consistent for the two summary measures (table 5). While both of the summary exposure measures were modeled as the linear effect of a 1-unit change in exposure (in crg/m'), the odds ratios have been rescaled separately for each measure to reflect a change in exposure approximately equal to the interquartile range for each summary exposure measure. This scaling factor was 10 pg/m3 for the petroleumrelated compounds and 2 pg/m3 for the Process-related compounds. An increased cumulative incidence of lower respiratory symptoms was associated with both a 10-pg/m3 change in petroleumd a t e d compounds (OR = 1.OS) and a 2-pg/ m3 change in process-related compounds (OR= 1.08). A similar association was ob- served for the chronic lower respiratory response composite and for the individual chronic respiratory symptoms, although not 41 of the confidence intervals for the individual symptoms indicated statistical significance at the 95 percent confidence level. The prevalence of a physician's diagnosis of asthma was associated with exposure to petroleum-related compounds (OR = 1.OS, 95 percent CI 1.02-1.08) but not with exposure to process-related compounds (OR = 0.99, 95 percent CI 0.9 1-1.08). Acute lower respiratory symptoms were marginally associated with the ambient concentrations of both petroleum-related compounds (OR = 1.03) and process-related compounds (OR = 1.02). Rates of acute symptoms in the eyes, nose, and throat did not show an association with the measured concentrations of these compounds, except for acute eye symptoms, which were possibly associated with process-related compounds (OR = 1.08, 95 percent CI 0.99-1.19). As figure 2 shows, one school had a very extreme measured level of petroleum-related compounds and a relatively high cumulative incidence of lower respiratory symptoms. The level at the school with the highest exposure to petroleum-related compounds ( 1 54.4 &m3) was an order of magnitude higher than the median and was much higher than the level at the school with the second highest exposure b e l (85.1 pg/m3). The exposure-response association for r- 1296 Ware et al. 0.40 W U C W .--0 0.30 U -c i.W>- 4 -0 3 0.20 E 2 0 *" I0.10 I 0 IIIIII I 25 50 75 100 125 150 175 M e a s u r e d Concentration (,ug/rn 3) FIGURE 2. Cumulative incidence of lower respiratory symptoms according to measured concentrations of petroleum-related compounds at 74 elementary schools. Kanawha County, West Virginia, 1988. (Plotted on the bgit scale). 0.50 W U C .5m- 0.30 U -C .a>-l 4 -0 3 0.20 E 3 V c* * * *e -0 ****A$x* *** ** :** : 0.10 I I I I II I I I 1 0 2 4 6 8 10 12 14 1 6 18 M e a s u r e d C o n c e n t r a t i o n (pg/m 3) FIGURE 3. Cumulative incidenceof lower respiratorysymptoms according to measured concentrations of process- related compounds at 74 elementary schools, Kanawha County, West Virginia, 1988. (Plottedon the logit scale). petroleum-related com; influenced by this sing1 ever. the plot for proces. (figure 3) did not indic distribution. For the la the school with the hi; pg/m3) was only a facthe median and was on I the level at the school v est exposure (8.6 pg/m- Because of concerns of the toluene measure: sites, due in large par values, we repeated tianalyses using the pet pound concentrations uene. The adjusted odc to the smaller range ot leum-related summa? only slightly, and no c?in the pattern of sigr This is not surprising. between the sum of L compounds and the su was 0.95. DISCUSSION The volatile organic ments collected durir showed that the air of contains organic comr tions not typically foui example, unusually h: chloroform, carbon te` ityl oxide were measur spatial distribution of compounds across the sistent with the West Control Commission' ventory (9),which sug plants are an importar pounds. The community w= erative with the s u m data base was establisi promised by either a p large number of care! tionnaires. Symptom range reported from munities that have 1: naires (22). _I_L___J 1 150 175 m3) %wed mmmtrations of mia, 1988. (plotted on the u 4 16 18 ) lentrationsof processed or~the bgit scale). HealthEIffectsof Ambient Volatile Organic Compounds 1297 petroleum-related compounds was strongly influenced by this single observation. However, the plot for process-related compounds (figure 3) did not indicate such an extreme distribution. For the latter plot. the level at the school with the highest exposure (13.0 pg/m3) was only a factor of 3 higher than the median and was only slightly higher than the level at the school with the second highest exposure (8.6 pg/m3). Because of concerns about the variability of the toluene measurements across the four sites, due in large part to a few extreme values, we repeated the logistic regression analyses using the petroleum-related cornpound concentrations after excluding toluene. The adjusted odds ratios, when scaled to the smaller range of the modified petroleum-related summary variable, changed only slightly, and no changes were observed in the pattern of significant associations. This is not surprising, since the correlation between the sum of all petroleum-related compounds and the sum excluding toluene was 0.95. DISCUSSION The volatile organic compound measurements collected dunng this investigation showed that the air of the Kanawha Valley contains organic compounds at concentrations not typically found in outdoor air. For example, unusually high concentrations of chloroform, carbon tetrachloride. and mesityl oxide were measured at many sites. The Spatial distribution of the 15 volatile organic Compounds across the valley (14) was consistent with the West Virginia Air Pollution Control Commission's 1984 emissions inventory (9), which suggests that the chemical plants are an important source of these compounds. The community was exceptionally c o o p erative with the survey. As a result, a large data base was established that was not comPromised by either a poor response rate or a large number of carelessly completed questionnaires. Symptom rates were within the range reported from studies in other communities that have used similar questionnaires (22). Three measures of exposure were used in this study: two composite measures based on the volatile organic compound measurements and a measure of proximity to industry. Incidence rates for respiratory s y m p toms were higher for schools located within the valley. Trend analysis showed that these differences were statistically significant for physician's diagnosis of asthma, wheezerelated symptoms, and the composite variables that included the symptoms characteristic of reactive airway disease (table 4). The regression analysis based on the two composite measures of volatile organic compound concentrations suggested that rates of chronic symptoms increased as the levels of the two exposure variables increased (table 5). Taken as a whole, these data show that children attending schools closer to chemical plants and children attending schools with higher concentrations of volatile organic compounds tended to report more frequent respiratory symptoms, and also were more likely to report acute eye symptoms. Limitations Epidemiologic studies of air pollutants inevitably have limitations arising from measurement error in both exposure data and health data and the possibility of incomplete control for other factors that affect the health outcomes. Since location is not the only determinant of the dispersal of emissions, the proximity-to-industry groups were subject to substantial misclassification as indicators of exposure. Moreover, outdoor measurements are an imperfect predictor of total personal exposure, which also depends on indoor concentrations (27). The 15 pollutants selected do not fully characterize the exposure of Kanawha Valley residents to all volatile organic compounds. Emissions of some compounds occur episodically, and their peak concentrations may be much higher than was suggested by the timeaveraged values over the 8-week monitoring period. Finally, the air quality during the monitored period may not be representative of the children's prior exposure history. The consequent random misclassification of exposure, which should have affected persons with and without the health outcomes equally, would have tended to reduce the strength of the apparent relation between exposure to volatile organic compounds and chronic respiratory symptoms. The measurements made during this investigation and during an earlier environmental monitoring study (5) show that ambient concentrations of some volatile organic compounds are unusually high in the Kanawha Valley. The organic vapor monitor (OVM) badges used in this study had not previously been used for prolonged passive sampling of outdoor air at concentrations measured in the Kanawha Valley. A 2 1-day chamber study was performed to compare measurements obtained by OVM badges to predicted concentrations and measurements obtained using active sampling with activated charcoal. For most of the organic compounds investigated, measurements obtained with the passive sampler were within 25 percent of predicted concentrations and were consistent with values obtained using active sampling, suggesting that these samplers can provide an inexpensive alternative to active sampling ( 18).Paired badges placed at 12 locations provided data on the reliability of paired measurements. Finally, concentrations of volatile organic compounds were compared with measurements obtained by active sampling at four sites over a 12-month period (April 1986March 1987) during an earlier phase of this research. The mean concentration of petroleum-related compounds was 18.5 pg/m3 in the earlier study, which is quite consistent with the mean of 19.2 pg/m3 observed in this study. Mean concentrations of chloroform and carbon tetrachloride were 5.43 pg/m3 and 1.87 pg/m3, respectively, raising the possibility that concentrations of these compounds might have been underestimated by the OVM badge methodology. However, these comparisons were qualified by the fact that the active sampling was done at a different time than the OVM badge sampling and at different sites. Further stud- ies are needed to improve our understanding of the performance of the OVM badge. In particular, other investigators considering studies should consider the feasibility of placing pairs of badges at each sampling site. This would provide more extensive reliability data, and the mean of the two sample values would have smaller sampling variability. Regional air pollutants such as particulate matter and acidic air pollution were not potential confounders in this study. Cohen et al. (6, 14) found no significant differences at four Kanawha County sites in measurements of light scattering or in concentrations of nitrogen dioxide, sulfuric acid, and fine particulate matter. Despite the absence of confounding by acidity and particulate matter, the possibility remains that volatile organic compounds may be serving as markers for other pollutants emitted from industrial and mobile sources, including other emissions from the same facilities and emissions from vehicles and other activities around the plants. Many factors might influence the reporting of respiratory symptoms in children, including exposure to environmental tobacco smoke, a history of asthma or allergies, a past history of severe lower respiratory infections, and socioeconomic status (28-30). The symptoms present in the reporting parent also influence the reporting of symptoms for the child (31-33). An attempt was made in this study to measure and control for factors that might have affected the endpoints selected for study, and parental smoking and familial socioeconomic status were included in the regression models. However, because of misclassification, control for such confounders may not have been complete. Symptom reporting may also have been influenced by parental attitudes about the petrochemical industry and by their general knowledge of health: these factors, however. could not be readily measured or controlled in the analysis. Parents living near the chemical plants could have introduced a bias in two ways. Parents who were concerned about exposure could have overreported respiratory symptoms, while parents who were concerned about an adverse effect on employment could have underreported respi- r3to1-ysymptoms. t far-exposure dicho 13nt as either the \alley dichotomy e 10 volatile organic The choice of re facton related tc higher incomes nchemical plant, ar fected may move ;i study, the bias 1 choices as to plact fully detected or cc Conclusions The monitoring complement the re of chemical air pc Valley. Many of t pounds measured c ted by Kanawha C in quantities of at nually. These corn outdoor air at schc neighborhoods. P,, outdoor activities . of indoor air are ce of these compour, industrial sources. in automobile em! petroleum-related sured the combine trial and vehicular The indices of ex compounds were risk of several chrc and of a physicin These results inc within the commc reflect the childre sure to sources of compounds locate Previous studies 1exposures to pan: have been assock tory symptoms ( Z nary function (340). In this stud) using either the e dges in epidemiologic der the feasibility of j at each sampling site. lore extensive reliabili n of the two sample naller sampling varia- nts such as particulate r pollution were not in this study. Cohen significant differences .nty sites in measureg or in concentrations dfuric acid, and fine spite the absence of . and particulate matiains that volatile orbe servingas markers nitted from industrial ncluding other emisxilities and emissions r activities around the influence the report;toms in children, inL ironmental tobacco dhma or allergies, a ower respiratory inomic status (28-30). In the reporting parporting of symptoms n attempt was made ire and control for e affected the end- . and parental smok- m o m i c status were In models. However, ion, control for such ive been complete. nay also have been attitudes about the m d by their general *sefactors, however. m r e d or controlled ving near the chemntroduced a bias in 0 were concerned !ve overreported rele Parents who were effect on emnderreported respi- Health Effectsof Ambient Volatile Organic Compounds 1299 ratory symptoms. However, the near- versus far-exposure dichotomy was not as important as either the in-valley versus out-ofvalley dichotomy or the measured exposures to volatile organic compounds. The choice of residential location reflects factors related to health. Families with higher incomes may avoid living near a chemical plant, and persons adversely affected may move away. In a cross-sectional study, the bias introduced by personal choices as to place of residence cannot be fully detected or controlled. Conclusions The monitoring results from this study complement the results of previous studies of chemical air pollution in the Kanawha Valley. Many of the volatile organic compounds measured during this study are emitted by Kanawha Valley chemical industries in quantities of at least 50,000 pounds annually. These compounds were detected in outdoor air at schools situated in residential neighborhoods. Population exposures from outdoor activities and from contamination of indoor air are certain to take place. Some of these compounds are emitted only by industrial sources, while others are found in automobile emissions as well. Thus, the petroleum-related composite variable measured the combined exposures from industrial and vehicular sources. The indices of exposure to volatile organic compounds were associated with increased risk of several chronic respiratory symptoms and of a physician's diagnosis of asthma. These results indicate a gradient of risk within the community which is presumed to reflect the children's general level of exposure to sources of ambient volatile organic compounds located in the Kanawha Valley. Revious studies have shown that ambient exposures to particulates and acid aerosols have been associated with chronic respirafOrY symptoms (34-36), deficits in pulmonarY hnction (37, 38), and mortality (39: 40). In this study, the geographic analyses using either the exposure measurements at each school or the proximity-to-industry regions found associations with exposure to volatile organic compounds that were distinct from the effects of regional air pollutants such as particulates and acid aerosols. While the mechanisms are not well understood (4I), these compounds may cause disease through direct irritation and neurotoxic effects (42-44). Experimental human exposures to mixtures of volatile organic compounds also produce symptoms of eye, mucosal, and airway imtation (45). Thus, the finding that ambient exposures were associated with respiratory symptoms is biologically plausible and is consistent with the limited available evidence on the respiratory effects of volatile organic compounds. Whether the association with asthma represents a role for exposure in the onset or exacerbation of asthma cannot be established from this study. However, in both occupational and experimental settings, volatile organic compounds have been linked to asthma and bronchial hyperreactivity (46-48). This survey of respiratory symptoms in schoolchildren provides evidence that exposure to a mixture of volatile organic compounds, some emitted from chemical manufacturing plants, has adversely affected the health of some of the children living in the Kanawha Valley. REFERENCES I . Sullivan N, Ozkaynak H, Ware JH, et al. Report on ambient exposures to volatile organic compounds in the Kanawha Valley. Charleston, WV: National Institute for Chemical Studies, 1989. 2. Burbank B, Ozkaynak H, Ryan PB. et al. Monitoring of exposures and the design of a health effects study in the Kanawha Valley of West Virginia. Charleston. WV: National Institute for Chemical Studies. 1987. 3. Damakosh A. S.oenIder JD. Docken DW. et al. Effects of acidic particles on respiratory symptoms in 7 US communities. (Presented at the American Lung Association/American Thoracic Society International Conference. San Francisco. California, May 18, 1993). Am Rev Respir Dis 1993: 147 (in press). 4. National Air Pollution Control Administration. US Public Health Service. Kanawha Valley Air Pollution Study. Raleigh. NC: National Air Pollution Control Administration. 1970. (Publication no. APTD 70-1). 5 . Environmental Protection Agency. Kanawha Valley toxics screeningstudy: final report. Washington, DC: Region 111-Environmental Services Division and Office of Policy Analysis, Environmental Protection Agency, 1987. 6. Cohen MA. Ryan PB, Spengler JD. et al. Sourcereceptor study of volatile organic compounds and particulate matter in the Kanawha Valley, WV. 1. Methods and descriptive statistics. Atmos Environ (Urban Atmos) 1991;258:79-93. 7. Environmental Protection Agency. The Total Ex- posure Assessment Methodology (TEAM) Study: summary and analysis. Vol I. Washington, DC: Office of Research and Development, 1987. (EPA publication no.600-6-87/002a). 8. SterlingDA. Volatile organic compounds in indoor air: an overview of sources. concentrations, and 993-7. 19. Chan CC. Spender JD, Ozkaynak H. et al. Com- muter exposures to V W s in Boston. Massachu. setts. J Air Waste Manage Assoc 1991:41: 1594-1600. 20. Wixtrom RN. Brown SL. Individual and popula- tion exposures to gasoline. J Expo Anal Environ Epidemiol 1992223-78. 2 I . Edgerton SA. Shah JJ. Assessing total exposures to gasoline vapor using the source exposure model. J Expo Anal Environ Epidemiol 1992:2:109-15. 22. Ware JH, Spiro A, Dockery DW, et al. Respiratory illnessin children exposed to ambient air pollution, cigarettesmoke and gas stoves in six cities. Am Rev Respir Dis 1984:129:366-74. 23. Fems BG Jr. Epidemiology standardization project. Am Rev Respir Dis 1978:118(suppl):7-52. 24. Rothman KJ. Modem epidemiology. Boston: Little, Brown and Company, 1986. air pollution: a cross-sectit 11. Environ Res 198950:: 39. Schwaru J, Marcus A. hl in London: a time series 1990;I 31:I 85-94. 40. S c h w m J, Dockery D\h and daily mortality in S Epidemiol 1992;135:12-; 41. Wallace LA. Comparisoand indoor exposure to Health Perspea 1991:95 42. Andentn I, Lundquist G man exposure to controllr hour exposures. S a n d ; 1983;9:405-18. 43. Nielsen GD. Alaire Y. Ser ratory stimulation by airt benzenes. Toxic01 . ~ p r health effects. In: Gammage RB, Kaye SV, eds. 25. Hennekens CH. Buring JE. Epidemiology in med- Indoor air and human health. Chelsea, MI: Lewis icine. Boston: Little, Brown and Company, 1987. Publishers. Inc, 1985:387-402. 26. Hosmer DW, Lemeshow S. Applied logistic regres- 9. Burner PA, Crane M, Key MM. Possibilities of sion. New York: John Wiley & Sons, Inc, Publish- i i detecting health effects by studies of populations exposed to chemicals from waste disposal sites. Environ Health Perspect 1985:62:423-56. ers, 1989. 27. Wallace LA. Volatile organic compounds. In: Samet JM, Spengler JD. Indoor air pollution: a IO. West Virginia Air Pollution Control Commission. health perspective. Baltimore, M D The Johns 1984 emissions inventory for Kanawha Valley Hopkins University Press, 1991:252-72. chemical industry. Charleston, WV: West Virginia 28. Tisi GM, Braunwald E. Cough and hemoptysis. In: Air Pollution Control Commission, 1988. Braunwald E, Isselbacher KJ, Petersdorf RG, et al, 11. National Institute for Occupational Safety and eds. Hamson's principles of internal medicine. Health and Occupational Safety and Health Ad- 11th ed. New York: McGraw-Hill Book Company, ministration. Occupational health guidelines for 1987:138-41. chemical hazards. Rockville, MD: National Insti- 29. Speizer FE. Environmental lung diseases. In: tute for Occupational Safety and Health, 1981. Braunwald E, Isselbacher KJ, Petersdorf RG, et al, (DHHS publication no. 81-123). eds. Harrison's principles of internal medicine. 12. National Library of Medicine. IRIS: Integrated 11th ed. New York: Mkraw-Hill Book Company, Risk Information System. (Online electronic data 1987:1068-75. base). Ekthesda, M D National Institutes of Health, 30. Austen KF. Diseases of immediate type hypersen- 1988. sitivity. In: Braunwald E, IsselbacherKJ, Petersdorf 13. Environmental Protection Agency. Superfund pub- RG, et al, eds. Harrison's principles of internal lic health evaluation manual. Washington, DC: medicine. 1Ith ed. New York: McGraw-Hill Book Office of Emergency and Remedial Response, En- Company, 1987:1407-14. vironmental Protection Agency, 1986. (EPA pub- 31. Colley JRT.Respiratory symptoms in children and lication no. 540/ I -86/060). parental smoking and phlegm production. Br Med 14. Cohen MA, Ryan PB, Spender JD, et al. Source- J [Clin Res Ed] 1974;2:201-4. receptor study of volatile organic compounds and 32. Colley JRT, Holland WW, Corkhill RT. Influence paniculate matter in the Kanawha Valley, WV. 11. of passive smoking and parental phlegm on pneu- Analysis of factors contributing to VOC and parti- monia and bronchitis in early childhood. Lancet cle exposures. Atmos Environ (Urban Atmos) 1974;2:1031-4. 1991;25B:95-107. 33. Leeder SR, Corkhill RT, lnvig LM, et al. Influence 15. Minnesota Mining and Manufacturing. Organic of family factors on asthma and wheezing during vapor monitor #3S00 analysis guide. St. Paul, MN: the first five years of life. Br J Rev Soc Med 1976; Occupational Health and Safety Products Division, 302 13-18. Minnesota Mining and Manufacturing, 1982. 34. Ware JH, Fems BG Jr, Dockery DW, et al. Effects 16. Epstein PS,Morey PR, Ozkaynak H, et al. Expe- of ambient sulfur oxides and suspended particles riences using passive monitors to measure volatile on respiratory health of preadolescent children. Am organics during indoor and ambient air quality Rev Respir Dis 1986:133S34-42. surveys. Presented at Indoor Air '90 The Fifth 35. Samet JM, Speizer FE, Bishop Y, et al. The rela- International Conference on Indoor Air Quality tionship between air pollution and emergency and Climate, Toronto, Ontario, Canada, July 29- room visits in an industrial community. JAPCA August 3, 1990. I98 1:31:236-40. 17. Shields HC, Weschler CJ. Analysisof ambient con- 36. Ostro BD, Rothschild S. Air pollution and acute centrations or organic vapors with a passive sam- respiratory morbidity: an observational study of pler. JAPCA 1987;37:1039-45. multiple pollutants. Environ Res 1989;50:238-47. 18. Cohen MA, Ryan PB, Yanagisawa Y, et al. The 37. Dockery DW, Ware JH, Fems BG Jr. Change in validation of a passive sampler for indoor and pulmonary function in children associated with air outdoor concentrations of volatile organic com- pollution episodes. JAPCA 1982;32:937-42. pounds. J Air Waste Manage Assoc 199Q40: 38. Schwartz J. Lung function and chronic exposure to JD. Ozkaynak H, et al. Corn- ~ .) V ~ in SBoston, Massachu,re Manage ASOC 1991;41: . ~ SnL. Individual and popula.ssoline. J Expo Anal Environ ..L 7 8 . JJ. Assessing total exposures to 4 the source exposure model. J Epidemiol 1992;2:109-15. hckery DW, et al. Respiratory posed to ambient air pollution, gas stoves in six cities. Am Rev '1 366-74. rniology standardization projDis 1978;118(suppl):7-52. .rn epidemiology. Boston: Lit.pany, 1986. m g JE. Epidemiology in med- Brown and Company, 1987. .howS. Applied logistic regresIn Wiley & Sons, inc, Publish- :ile organic compounds. In: :JD. Indoor air pollution: a Baltimore, M D The Johns Press, 1991:252-72. . E. Cough and hemoptysis. In: icher KJ, PetersdorfRG, et al, :ciples of internal medicine. kGraw-Hill Book Company, .nrnental lung diseases. In: sher KJ, Petendorf RG, et al, ciples of internal medicine. .IcGraw-HiU Book Company, *>fimmediate type hypersen: E, Isselbacher KJ, Petersdorf ison's principles of internal :w York: McGraw-Hill Book -14. )qsymptoms in children and : phlegm production. Br Med xo1-4. w,Corkhill RT. Influence Id, Parental phlegm on pneu1 In early childhood. Lancet T,lnvig LM, et al. Influence sthma and wheezing during fe. Br J Prev Soc Med 1976; .. Dockery DW, et al. Effects .*.and suspended particles i preadolescentchildren. Am 3k834-42. . Bishop Y, et al. The rela- Pollution and emergency Ustrial community. JAPCA s. Air pollution and acute an observational study of .<ironRes 1989;50:238-47. 3. Ferris BG Jr. Change in children associated with air r.41982;32:93742. on and chronic exposure to Health Effects of Ambient Volatile Organic Compounds 1301 air pollution: a cross-sectionalanalysisof NHANES 11. Environ Res 1989;50:309-2 I. 39. Schwartz J, Marcus A'. Mortality and air pollution in London: a time series analysis. Am J Epidemiol 1990;13 1: 185-94. 40.SchwartzJ, Dockery DW. Particulate air pollution and daily mortality in Steubenville, Ohio. Am J Epidemiol 1992;135:12-19. 41. Wallace LA. Comparison of risks from outdoor and indoor exposure to toxic chemicals. Environ Health Perspect 1991;95:7-13. 42. Andersen I, Lundquist GR, Molhave L, et al. Human exposure to controlled levels of toluene in sixhour exposures. Scand J Work Environ Health 1983;9:405-18. 43. Niclsen GD, Alaire Y. Sensory irritation and respi- ratory stimulation by airborne benzene and alkyl- benzenes. Toxicol Appl Phannacol 1982;65: 459-77. 44. Harving H, Dahl R, Molhave L. Lung function and bronchial reactivity in asthmatics during ex- posure to volatile organic compounds. Am Rev Respir Dis 1991;143:7514. 45. Molhave L, Bach B, Pedersen OF. Human reac- tions to low concentrations of volatileorganic compounds. Environ Int 1986;12:167-75. 46. Lee HS, Wang YT, Cheong TH, et al. Occupational asthma due to maleic anhydride. Br J Ind Med 199 1;48:283-5. 47. Harving H, Korsgaard J, Dahl R, et al. Low concentrations of formaldehyde in bronchial asthma: a study of exposure under controlled conditions. Br Med J [Clin Res Ed] 1986;293:3IO. 48. Venables KM. Low molecular weight chemicals, hypersensitivity, and direct toxicity: the acid an- hydrides. Br J Ind Med 1989;46:222-32.