Journal of Exposure Analysis and Environmental Epidemiology (1999) 9,293±299 # 1999 Stockton Press All rights reserved 1053-4245/99/$12.00

http://www.stockton-press.co.uk

Respiratory function as measured by rate and PM10: six communities study

JONG-TAE LEE a,b AND CARL M. SHYa a Department of Epidemiology, School of Public Health, University of North Carolina, Chapel Hill, NC b Department of Preventive Medicine, College of Medicine, Yonsei University, Seoul, Korea

This Six Communities Studyconducted at six communities in southwestern North Carolina investigates the respiratoryhealth status of residents who se households are located near an incinerator. This diarystudymakes it possible to estimate the dailyvariation of pulmonaryfunction measured as peak expiratory flow rate (PEFR) related to 24-h mean PM10 levels, which were observed at each monitoring station placed in the six studycommunities, as a surrogate exposure measure of outdoor air pollution. Observations of PEFR among participants in each communitywere analyzedto determine how theyvaried according to the degree of exposure to ambient pollutants as well as to other cofactors including, sex, age, respiratoryhypersensitivity,hours spe nt outdoors within the area of the selected community, and surrogate measures for indoor air pollution exposure (vacuum use and experience of air irritants at work). The findings revealed that respiratoryhypersensitivitystatus is a predictor of declining PEFR. PM 10 concentrations measured in each studyarea did not seem to be related to the variations of respiratoryhealth as measured byPEFR. This studydid not show anydifference in respiratoryhealth between subjects of a n incinerator communityand those of its comparison community.Even though this community-based studywith free-living subjects shows negative find ings on the relationship between respiratoryhealth and PM 10, it is worth noting that these findings must be interpreted cautiouslybecause exposure estimation based on monitoring of ambient air likelyresults in misclassification of true exposure levels.

Keywords: air pollution, environmental epidemiology, epidemiologic methods, incinerators, longitudinal data, particulate matters, PEFR, respiratory function.

Introduction nominal 10 m (PM10). The study, however, failed to see an association between the prevalence of acute respiratory symptoms and these pollutants. Because the volume of inhaled air is so large, a biologically Incineration, disposing about 10% of solid waste in the significant dose of a potentiallyhazardous contaminant may United States, is an important and useful method for be delivered to the respiratorytract even though the destroying any type of hazardous waste (Environmental contaminant is present in the air at a low concentration. Protection Agency, 1986). A total of 260 hazardous waste Studies of trends over time in air pollution and disease incinerators were operating throughout the countryin 1986 patterns have produced a growing bodyof literature that has (Environmental Protection Agency, 1986). Pollutants associated day-to-day fluctuations in air pollution with daily emitted from an incinerator stack are particulate matter, fluctuations in acute morbidityor mortality(Whittemore carbon monoxide, hydrogen chloride, sulfur dioxide, and Korn, 1980; Dockeryand Schwartz, 1992; Schwartz nitrogen oxides, various metals, dioxins/furans, and volatile and Dockery, 1992 ; Schwartz et al., 1993). The acute organic compounds (VOCs). Much of the data that we have respiratoryeffects of ambient winter air pollution episodes on adverse health effects from exposure to incinerated waste were studied in a sample of children aged 7±12 years old in comes from cases of accidental or exceptionallyhigh levels nonindustrial communities in the Netherlands (Hoek and of exposure to pollutants. There is good evidence that Brunekreef, 1994). The studyreported significant negative particulate matter and other combustion products released associations between pulmonaryfunction and the concen- byindustrial sources can produce chronic respiratory tration of sulfur dioxide (SO ), black smoke, and particulate 2 disease (Environmental Protection Agency, 1982). Building matter with an aerodynamic diameter equal to or less than a a new incinerator, therefore, is often controversial because of concerns related to the possibilityof adverse health 1. Address all correspondence to: Dr. Jong-Tae Lee, Department of effects and environmental contamination from long-term Preventive Medicine and Public Health, College of Medicine, Yonsei exposure to these pollutants. University, 134 Shinchon-Dong, Seodaemun-Gu, Seoul, South This studyis a community-based diarystudyand uses a Korea. Tel.: +82-2-361-5357. Fax: +82-2-392-0239. E-mail: [email protected] classic design for the respiratoryeffects of air pollution. The Accepted 20 November 1998. studydesign consisted of a cross-sectional survey,yearly Lee and Shy Respiratoryfunction as measured bypeak expiratoryflow rate and PM 10: six communities study repeated cross-sectional observations (baseline sities. Three comparison communities were rep resented by information), and longitudinal information for 5 weeks in the labels MCo, BCo, and HCo. There is no other major air each year of follow-up. Over cross-sectional studies, the pollution source in these selected communities, except those longitudinal methods have a unique benefit of which the incinerators. possibilityof confounding byother time-varyingfactors At each studysite, we contacted everyselected house- that are difficult to control maybe reduced. The analyses hold byan introductoryletter, followed bypersonal home presented here address the following research questions. (1) and telephone interviews. The participants were informed

Is the outdoor PM10 level associated with respiratory that theywere part of an environmental study,but no function, as measured bypeak expiratoryflow rate (PEFR)? reference was made to the area incinerator as a pollution (2) Are some indoor sources of air pollution in the source of interest. Through the telephone interview, we workplace or home associated with PEFR? (3) Does the obtained information on sociodemographic factors, smok- function of participants in an incinerator community ing habits, respiratorysymptoms compatible with chronic differ from that in its comparison community? (4) Are the or with respiratoryhypersensitivity, results more conclusive or different when anyquestion frequencyof lower respiratorytract disease over the past described above is applied onlyto a subgroup of participants 2 years, other medical conditions, occupation and place with respiratory hypersensitivity? This study investigated of work as well as potential for exposure to irritating each of these research questions byanalyzing the respiratory substances at work, and perception of neighborhood function measured as PEFR of the original longitudinal data environmental pollution associated with respiratorysymp- of the Health and Clean Air (HCA) studyduring 1992±1993 toms. We identified active smokers and persons usually in all six communities (Shyet al., 1995). exposed to tobacco smoke at home, and excluded these subjects from the follow-up study. Based on information obtained from the community- wide cross-sectional survey, about 80 subjects aged 8±80 in Methods and materials each communitywere selected and asked to participate in the 1-month longitudinal study. The 80 subjects were This studyused a longitudinal component from the HCA selected such that half of them gave a positive response on studyconducted in southwestern North Carolina during the question regarding wheezing or -like symptoms 1992 and 1993. All information were obtained from during the past 12 months, and the other half included nonsmoking participants. Participants were instructed to persons who did not have anyactive respiratorydisorders. perform the peak flow test three times in the standing The former subjects were termed `respiratory-hypersensi- position, twice daily(in the morning and in the evening), tives' and the latter `normals'. All selected subjects were and to report simple time±activitypatterns of the day.Daily neither active nor passive smokers. The baseline assessment measurements of 24-h average PM concentrations were 10 of lung function was done with routine spirometryusing a obtained from an air sampling tower in each study computerized dryrolling seal . Each subject was communityduring the month of the longitudinal study. provided with a mini-Wright peak flow meter (Clement Clark, Inc., Columbus, OH) and a preformatted health Subject Selection and Health Outcomes symptom diary for 5 weeks, and was trained on their proper Based on aerial photographs and site visits, we located three use. Participants were instructed to perform the peak flow paired communities. There were about 400±500 homes in test three times in the standing position, twice daily(in the each selected community. Each pair of communities morning and in the evening), and to record all the readings consisted of a target (exposed, or incinerator community) along with the symptoms experienced that day, any and a comparison (nonexposed) area. This studywas medication use, and simple time±activitypatterns of the directed at the population surrounding three distinct waste day. Respiratory and nonrespiratory symptoms were listed combustion facilities; two located in Mecklenburg County, on the diaryform, with following blank columns in which North Carolina, and one in StanleyCounty,North Carolina. subjects could check the presence of each symptom for each The facilities varywith respect to waste type(municipal and dayof the month. Subjects also reported hours spent biomedical waste incinerators, and a hazardous industrial outdoors, use of a vacuum cleaner each day, hours spent waste furnace). Three communities, each located within 2 outside the neighborhood, and use of anyasthma medica- miles of and surrounding one of the waste incinerators, were tion. identified as target (exposed) communities and were represented bythe labels MWI, BWI, and HWI, respec- Measurement of Exposure tively. Comparison (nonexposed) communities were located An air monitoring station was located in each of the six >2 miles from these sources and selected on the basis of communities. Dailymeasurements of air qualitywere similar socioeconomic characteristics and population den- obtained in each studyarea during the month of the

294 Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(4) Respiratoryfunction as measured bypeak expiratoryflow rate and PM 10: six communities study Lee and Shy

Table 1. Compliance rate of peak flow diaries in each community, 1992 and 1993.

Year Contents Community BWIa Bcoa MWIa Mcoa HWIa Hcoa 1992 Recruited and trained for PEFRa 86 82 109 83 95 93 Completed 4 weeks peak flow diaries 73 (85%) 67 (82%) 94 (86%) 77 (93%) 89 (94%) 83 (89%) 1993 Recruited and trained for PEFR 111 84 112 105 110 76 Diaries mailed 15 16 18 6 70 37 Completed 4 weeks peak flow diaries 110 (87%) 86 (86%) 110 (85%) 94 (85%) 107 (91%) 74 (65%) 1992+1993 Diaries completed both in 1992 and 1993 36 46 70 59 79 56 a BWI, biomedical incinerator community; Bco, comparison community of BWI; MWI, municipal waste incinerator community; Mco, comparison communityof MWI; HWI, hazardous waste furnace community;Hco, comparison communityof HWI; PEFR, peak expiratoryflow rate.

longitudinal study. The tower included a Versatile Air search Triangle Park, NC. In this study, the PM10 level was Pollution Sampler (VAPS) system and equipment to used as an outdoor air qualityparameter and a surrogate measure wind speed and direction (Stevens et al., 1993). measure specific for the incinerator. Dailyestimated 24-h 3 The ambient pollutants were then analyzed at the U.S. means of PM10 (g/m ) were considered as an index of Environmental Protection Agency's Atmospheric Research ambient air quality. Cigarette smokers and persons regularly and Exposure Assessment Laboratory(AREAL) in Re- exposed to a household smoker were excluded from the study, but data were obtained daily on any passive smoke exposure. We considered use of a vacuum cleaner at home Table 2. Distributions of selected demographic variables in each as a source (or an indicator) of indoor air pollution on the communitybycalendar year. dayof use. In addition, information on hours spent outdoors was available from dailydiaries. Variable Community BWIa Bcoa MWIa Mcoa HWIa Hcoa Participants (numbers) Analysis 1992 78 75 105 80 92 90 This analysis relies on the combined data collected from 1993 116 91 112 96 113 75 dailydiaries and air qualitylogs from a sampling tower in 1992 and 39 (25.2) 53 (46.9) 73 (51.0) 59 (51.3) 79 (62.7) 59 (55.7) each of six communities for 35 consecutive days in 1992 1993 (%) and 1993. Our analysis of this project focuses on the daily respiratoryfunction (PEFR) measured bynonsmoking Age (year) participants living both in a studyarea with an incinerator 1992 38.9 39.8 34.0 45.6 40.2 40.9 and in a comparison communitywhose households were 1993 36.8 36.2 34.3 45.8 41.7 43.4 located more than 2 miles from the incinerator. The frequencydistribution of each variable was checked Female (%) to find the missing and unreasonable data points (e.g., if a 1992 41 (53) 39 (52) 57 (54) 47 (59) 56 (61) 47 (52) dailyPEFR was not between 100 and 800 l/min, the value 1993 59 (51) 54 (59) 66 (58) 59 (61) 69 (61) 42 (56) was checked for clerical error, and corrected or treated as an

Respiratory-hypersensitives (%) Table 3. Geometric mean and 95% CI of PEFR (l/min) byage or 1992 25 (32.1) 31 (41.3) 48 (45.7) 44 (55.0) 43 (46.7) 36 (40.0) respiratory hypersensitivity in each year. 1993 52 (44.8) 39 (42.9) 47 (42.0) 43 (44.8) 43 (38.1) 31 (41.3) Variable PEFR (95% CI) Hay fever (%) 1992 1993 1992 23 (29.5) 17 (22.7) 22 (21.0) 15 (18.8) 14 (15.2) 21 (23.3) Age (year) 1993 28 (24.1) 16 (17.6) 30 (26.8) 26 (27.1) 26 (23.0) 20 (26.7) <18 368.2 (231.0±587.3) 389.2 (249.4±607.2) 18±35 503.7 (349.1±726.7) 486.3 (327.3±722.6) Education (year) 35< 475.3 (302.8±746.1) 471.9 (312.7±712.2) 11.1 11.0 11.6 12.2 11.3 11.2 a BWI, biomedical incinerator community; Bco, comparison community of Respiratory sensitivity BWI; MWI, municipal waste incinerator community; Mco, comparison Normal 478.1 (302.5±755.7) 460.8 (286.2±741.9) communityof MWI; HWI, hazardous waste furnace community;Hco, Respiratory-hypersensitive 425.0 (253.8±711.6) 436.2 (276.5±687.9) comparison communityof HWI.

Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(4) 295 Lee and Shy Respiratoryfunction as measured bypeak expiratoryflow rate and PM 10: six communities study

Table 4. Distribution of average PEFRa variabilityindices in respiratory variable of respiratoryfunction as measured byPEFR hypersensitive and normal participants. reported on successive days will usually be autocorrelated. Thus, the response vector for each individual is a vector of b Type of variability Type of participants Percentage (%) correlated outcomes in this paper. This longitudinal study Respiratory- NormalŠ- difference implies an underlying ordering of the times of repeated hypersensitive measures (Liang and Zeger, 1986; McLean et al., 1991). Abs Ampa 19.64 16.58 18 All models considered here assumed that the expected Amp percentage meana 0.05 0.04 25 response varied linearly, for each participant, with a slope Morning dip À7.14 À5.28 35 and intercept that maydepend on fixed (SEX: female) or a PEFR, peak expiratoryflow rate; Abs Amp, absolute amplitude; Amp time-varying covariates (OUTDOOR: time spent outdoor; percentage mean, amplitude percentage mean. VACUUM: use of vacuum cleaner; PM :PM level; b 10 10 Percentage difference=(variabilitydifference between sensitive and AGE: age; HEIGHT: height; and AIRIRR: experience of air normal)/(normal variability). irritants at work). Separate analyses were done for each dataset stratified bythe presence of respiratoryhypersensi- tivity (hypersensitive and normal). All statistical analyses error). We also excluded the records on PEFR trials of the were performed using the MIXED procedure in Windows/ first day, because it was assumed that PEFR practice on the SAS Software version 6.09 (SAS, 1992). first dayof follow-up might be poorlydone because of insufficient exercise or training period. The information we collected was divided into three categories. Firstly, the cross-sectional baseline information Results was collected at the time of subject recruitment. These baseline data consisted of demographic factors, respiratory Descriptive Analysis symptoms/diseases (asthma attack, wheezing, coughing, Response rates to an initial cross-sectional questionnaire for and hayfever) over the past 12 months, information on the six communities in 1992 ranged between 66% and 80%. occupational chemical exposure, and smoking history. Table 1 presents the numbers and proportion of subjects Secondly, diary data included daily respiratory symptoms, recruited and trained for PEFR in each communitybyyear. twice dailyPEFR measures, medication information, and We considered that a participant completed the 35 days of surrogate measures on dailytime±activitypatterns and diarystudyif he/she recorded the peak flow diaryfor at least indoor air pollution. The third element of data involved four weeks. The present studyresults also indicated that the information on ambient air quality. Measurements of air PEFR measurements of the first trial (day) varied incon- qualitywere given as 24-h average concentrations of PM 10. sistentlyfrom those of the rest of the trials (days).Looking The relationship between community-averaged daily at crude graphical plots of individual PEFR values over the PEFR and ambient air quality(PM 10), time±activity 35-dayperiod of measurement, we decided to exclude patterns, and work exposure to irritants was studied in each PEFR values of the first days. Therefore, there are, at most, of the six communities. Because repeated observations are 34 consecutive observations for each studyparticipant in a made on the same participant, the continuous response year.

a 3 Table 5. Distributions of 12-hour mean PM10 (g/m ) by communities and surveyed year.

Community12-h Mean concentration (minimum±maximum) 1992ŠŠŠ- 1993 ŠŠ- Daytime Nighttime Daytime Nighttime BWIb 35.3 (8.4±76.6) 33.4 (12.0±72.7) 32.4 (5.6±68.0) 32.4 (7.8±59.3) BCob 35.1 (7.9±68.2) 37.8 (12.3±85.5) 29.9 (12.3±55.5) 30.5 (15.6±52.5) MWIb 20.9 (6.9±46.6) 22.0 (7.1±51.8) 20.8 (5.3±45.8) 24.0 (5.5±53.9) MCob 21.5 (9.2±44.1) 24.6 (8.0±55.6) 22.0 (10.2±51.4) 25.7 (6.4±49.3) HWIb 14.8 (3.5±28.6) 21.3 (10.2±39.5) 18.9 (4.6±34.3) 19.3 (6.2±33.3) HCob 18.0 (5.7±33.0) 24.2 (8.8±65.4) 18.9 (4.5±38.0) 24.2 (5.7±41.8) a 3 3 PM10, particulate matters less than 10 m in diameter; 12-h mean concentration over 2 years was 26.4 g/m (SE=13.0) for nighttime and 23.9 g/m (SE=13.1) for daytime. The 24-h concentration over 2 years was 25.1 g/m3 (SE=12.1). b BWI, biomedical incinerator community; Bco, comparison community of BWI; MWI, municipal waste incinerator community; Mco, comparison communityof MWI; HWI, hazardous waste furnace community;Hco, comparison communityof HWI.

296 Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(4) Respiratoryfunction as measured bypeak expiratoryflow rate and PM 10: six communities study Lee and Shy

Table 6. Geometric mean and 95% CI of PEFR (l/min) in each communitybycalendar years.

Year Community BWIa Bcoa MWIa Mcoa HWIa Hcoa 1992 457.6 (279.9±748.4) 464.1 (296.2±726.9) 450.3 (267.4±758.5) 437.0 (251.5±759.5) 454.9 (292.1±708.4) 464.1 (287.7±748.6) 1993 456.2 (288.4±721.7) 449.3 (278.0±726.3) 442.3 (272.0±719.2) 445.9 (294.8± 674.2) 452.8 (297.7± 688.7) 460.4 (291.0±728.2) a BWI, biomedical incinerator community; Bco, comparison community of BWI; MWI, municipal waste incinerator community; Mco, comparison communityof MWI; HWI, hazardous waste furnace community;Hco, comparison communityof HWI.

3 About 48% (n=362/756) of subjects participated in both standard of 150 g/m PM10 was never exceeded in all six of the first two years of follow-up study (Table 2). In all communities during the studyperiod (24-h average PM 10 communities, the number of female participants was higher concentration in 1992 and 1993 was 25.9 g/m3). Only15 than for male participants (Table 2). The age range of days (12 days in 1992 and 3 days in 1993) were above 50 3 participants in the diarystudywas 8±81 yearsand the g/m . This studydid not reveal anydifference in PM 10 mean age of participants was 39.1 years. The range of levels between studyand comparison communities (Table dailymeasured PEFR in this project was 150±765 l/min. 5). The variations of PM10 levels wiithin each pair of The geometric mean of PEFR is 452.9 l/min. There was communities show verysimilar trends in 1992 and 1993. not anyparticular trend or variation of mean PEFR over The means of PM10 byyearsof data collection within a the six communities between the years of 1992 and 1993 communityindicate that there is little difference between (Table 3). years. Table 6 presents the analysis of the effect of The results (Table 3) from the cross-sectional study communityresidence on respiratoryfunction. This study indicate that persons classified as respiratoryhypersensi- compared three distinct pairs of communities bythe tives showed a slightlylower PEFR on average than those as defining characteristic of a nearbyincenerator. The results normal subjects. The finding also shows a greater variability show that there is little evidence of differences in PEFR among the hypersensitive group in any type of variability response comparing the studies of comparison commu- index, such as absolute amplitude % mean, and morning nities, after adjusting for sex, age, height, exposure to air dip. These values of respiratoryhypersensitives show 18%± irritants at work, use of vacuum cleaner, time spent outdoors

35% more variabilitythan those of normal participants and PM10. (Table 4).

The 24-h mean PM10 concentrationover two years was 25.1 g/m3 (SE=12.1). In our study, the national 24-h Longitudinal Analysis (PM10 and PEFR) The results of some of the regression models are presented in Table 7. Each model was separatelyapplied to data Table 7. Estimates of the regression coefficients and its SEs byyearusing stratified byyear. DailyPEFR measured in the afternoon a mixed model. was regressed against the 24-h average PM10 level lagged by1 day,time spent outdoors, and the use of vacuum 1992 1993 cleaner at home, gender difference, reporting of air irritants Model 1 at work, height, and respiratoryhypersensitivitystatus. We Intercept À506.1 (46.59)** À422.7 (47.20)** a 3 PM10 (g/m ) À0.007 (0.015) À0.011 (0.008) Table 8. Estimates of the regression coefficients of predicting amplitude Outdoorb (h) À0.072 (0.075) 0.072 (0.076) of PEFRa using a mixed model. Female À27.4 (6.89)** À31.85 (6.65)** c Vacuum À0.257 (0.43) À1.156 (0.44)** Variable 1992 1993 AIRIRRc À1.496 (0.72)* À0.096 (0.79) Intercept 19.3 (6.6)** 24.88 (7.19)** Age (year) À1.022 (0.19)** À0.635 (0.17)** 3 PM10 (g/m ) 0.008 (0.012) À0.012 (0.013) Sensitivec À27.23 (6.20)** À17.25 (5.90)** Outdoor (h) À0.014 (0.058) 0.159 (0.060)** Height (cm) 15.41 (0.73)** 13.99 (0.73)** Female À1.931 (0.968)* À2.718 (1.012)** a Dailyaveraged concentration considered 1-daylag factor. Age À0.085 (0.026)** À0.033 (0.025) b Time spent outdoor during the previous day. Height (cm) 0.066 (0.103) À0.026 (0.111) c Vacuum: use of vacuum cleaner at home during the previous day; AIRIRR: experience of air irritants at work; Sensitive: an indicator variable Sensitive 3.380 (0.868)** 2.538 (0.897)** for respiratory-hypersensitives. a Amplitude of PEFR=(highest readingÀlowest reading). * 0.01 p-value < 0.05. * 0.01 p-value < 0.05. ** p-value < 0.01. ** p-value < 0.01.

Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(4) 297 Lee and Shy Respiratoryfunction as measured bypeak expiratoryflow rate and PM 10: six communities study

Table 9. Estimates of the selected regressiona coefficients and its standard air pollution episode varies significantlyamong study errors byrespiratoryhypersensitivityusingdata of participants who are subjects. 16 years-old and exposed to chemicals at work. There was no difference in 24-h average PM10 levels between the studyand comparison communities. This Variable Hypersensitive Normal indicates that regional air pollution sources rather than air and hayfever pollution from an incinerator are likelyto be the major Vacuum À2.360‹1.19* 0.860‹1.01 factors in local air quality, and the distribution of total air AIRIRR À2.998‹1.18* 0.315‹1.02 pollution sources is similar within each pair of communities. a Regression models also included sex, age, height, time spent outdoor, We need to applymore specific exposure measures using PM10 and an interaction term of PM10 and time spent outdoor. * p-value < 0.05. wind direction and speed, meteorological information, microenvironmental air pollution measurement, and de- tailed activitypatterns of each participant, etc. Indoor pollutants including environmental tobacco had separatelyrun the regression models with PM 10 levels smoke (ETS), indoor allergens, nitrogen dioxide, and VOCs lagged by0 h, a half day,and a full day(24 h). Table 8 have been emphasized as a major factor to explain the presents the estimates of the regression coefficient for PM10 increasing mortalityand morbidityof the subpopulation byconsidering these lag factors. Two different measures of who have hyperresponsiveness or hypersensitivity (i.e., PEFR (one measured in the morning; the other measured in asthma) (Gergen et al., 1992, Ostro et al., 1994). This study the afternoon) were also considered as a response variable. considered the use of a vacuum cleaner as an indoor None of the estimates was found to be significant. This pollutants source after excluding anyparticipant who was practice shows that lagging by24 h yields consistently regularlyexposed to ETS. We did not use information on the negative coefficients in both PEFRs measured in the types of combustion source at home because the informa- afternoon and in the morning. tion on dailyuse of the combustion was not available. This The results suggest that a subject defined to have a diarystudydid not include questions asking whether or not respiratoryhypersensitivitymayshow increased suscept- the stove was used during each reporting day. A gas or wood ibilityor variabilityto air pollutants. Overall, this studydid stove is believed to be an important source of a complex not reveal a statisticallysignificant negative relationship mixture of pollutants, manyof which are respiratory between PM10 level and PEFR, though the association was irritants. always in the hypothesized negative direction. Our findings suggest that there is an adverse effect on Table 9 presents an analysis of the effect of PM10 on the respiratoryfunction from using a vacuum cleaner at home. intra-individual variabilityof PEFR using a mixed model. Use of a vacuum cleaner indicates either a good or bad sign PM10 was not a significant predictor of variabilityas of indoor air quality. People are accustomed to cleaning measured bythe amplitude of PEFR. The result also their house since the house is dirty. Therefore, it could be an indicates that male and respiratoryhypersensitives show a indication of indoor air pollution, which is a risk factor for higher variabilityof PEFR. respiratoryhealth. Using them in a confined area may contribute to airborne dust byboth leakage through the cleaner and disturbance of floor dust. Some models of Discussion vacuum cleaners mayincrease total airborne allergens and selectivelyincrease certain particle size. A studyby The idea of conducting this studywas partlyinduced bythe Haddock and Nocon (1994) showed that some cases of results from manystudies showing an association between infant salmonellosis resulted from the use of contaminated

PM10 and pulmonaryfunction and symptoms (Dockeryet vacuum cleaner bags. al., 1982; Environmental Protection Agency, 1987; Pope et Our studyhad another strength in that it allowed us to al., 1991). The findings of this studyconsistentlyshowed compare and validate PEFRs using a mini-Wright PEF that the presence of respiratoryhypersensitivitywas a risk meter with PEFR from a spirometer exercise. We found that factor for decreased PEFR. It can also be seen that the these two measures were highlycorrelated (correlation within-individual variations of PEFR were larger in coefficient=0.90, p-value<0.0001). Therefore, dailymea- hypersensitive participants than in normal participants. This sured PEFRs using a mini-Wright meter provided a reliable suggests that these two groups have a different degree of index as an end-point for a participant's respiratoryhealth biological susceptibilityto respond to air pollutants. The status. A possible disadvantage of using PEFR variabilityin Clean Air Act requires that sensitive subgroups of exposed community-based studies is that the values obtained might populations be protected from adverse health effects of air be altered byuse of inhalers, which could, in pollution exposure. The existence of a respiratory-hyper- theory, either increase or decrease PEFR variability sensitive subgroup implies that the expected response to an (Higgins et al., 1989). This studydid not consider

298 Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(4) Respiratoryfunction as measured bypeak expiratoryflow rate and PM 10: six communities study Lee and Shy information on the use of inhalers as a predictor in the Environmental Protection Agency. Air quality criteria for particulate statistical model because the diarydata showed that the matter and sulfur oxide. Epidemiological Studies of Health Effects, Chap. 14. EPA, Research Triangle Park, NC, 1982. participants rarelyused the inhaler during the studyperiod. Environmental Protection Agency. Incineration and treatment of hazardous waste. Proceedings of the 11th Annual Research Symposium, Conclusion Cincinnati, OH, 1986, EPA/600/9-85-028. Environmental Protection Agency. Revisions to the national ambient air We examined age, sex, occupational exposure to chemicals, qualitystandards of particulate matter. Federal Register July1, 1987: use of vacuum cleaners, hypersensitivity status, and height 52 (126): 24634±24669. Gergen P.J., and Weiss K.B. The increasing problem of asthma in the of participants as potential predictors of the variation of United States. Am. Rev. Respir. Dis. 1992: 146: 823±824. PEFR. PEFRs for hypersensitives were lower and more Haddeck R.L., and Nocon F.A. Infant Salmonellosis and vacuum cleaners. variable than those of normals. For hypersensitive subjects, J. Trep. Pediatr. 1994: 40: 53±54. we found that the use of vacuum cleaners at home and Higgins B.G., Britton J.R., and Chinn S. et al. The distribution of peak exposure to air irritants at work were significantly expiratoryflow variabilityin a population sample. Am. Rev. Respir. Dis. 1989: 140: 1368±1372. associated with the variation of PEFR. There was no Hoek G., and Brunekreef B. Effects of low-level winter air pollution difference in lung function as measured byPEFR between concentrations on respiratoryhealth of Dutch children. Environ. Res. participants in an incinerator communityand those in its 1994: 64 (2): 136±150. comparison community. There was no indication that Liang K.Y., and Zeger S.L. Longitudinal data analysis using generalized variabilityof pulmonaryfunction was associated with air linear models. Biometrika 1986: 73: 13±22. McLean R.A., Sanders W.L., and Sroup W.W. A unified approach to pollution exposure within the ranges of PM10 levels we mixed linear models. Am. Stat. 1991: 45: 54±64. observed. Overall, this studycould not reject the null Pope C.A., DockeryD.W., and Spengler J.D. et al. Respiratoryhealth and hypothesis of no relationship between PM10 level and PM10 pollution. Am. Rev. Respir. Dis. 1991: 144: 668±674. PEFR. The measure used to assess a personal exposure may Ostro B.D., Lipsett M.J., and Mann J.K., et al. Indoor air pollution and not actuallybe representative. This studyis limited bythe asthma: results from a panel study. Am. J. Respir. Crit. Care Med. 1994: 149: 1400±1406. absence of information, which makes it possible to estimate SAS Institute Inc. SAS/STAT software: changes and enhancements. SAS total individual exposure. We need further data which Technical Report P-229, Release 6.07, Chap. 16. SAS Institute Inc., attempt to address the full arrayof environmental factors Cary, NC, 1992. determining the variation of respiratoryfunction. This Schwartz J., and DockeryD.W. Particulate air pollution and dailymortality problem makes it harder to measure the health effects in Steubenville, Ohio. Am. J. Epidemiol. 1992: 135: 12±19. Schwartz J., Slater D., Larson T.V., Pierson W.E., and Koenig J.Q. associated with polluted air specific to a pollution source. Particulate air pollution and hospital emergencyroom visits for asthma For subsequent analyses, it would be desirable to collect in Seattle. Am. Rev. Respir. Dis. 1993: 147: 826±831. information to estimate total personal exposure. ShyC.M., Degnan D., Fox D.L., Mukerjee S., Hazucha M.J., Boehlecke B.A., Rothenbacher D., Briggs P.M., Devlin R.B., Wallace D.D., Stevens R.K., and Bromberg P.A. Do waste incinerators induce References adverse respiratoryeffect? An air qualityand epidemiologic studyof six communities. Environ. Health Perspect. 1995: 103: 714±724. DockeryD.W., and Schwartz J. Authors' response to Waller and Swan Stevens R., Pinto J., Mamane Y., Ondov J., Abdulraheem M., Al-Majed commentary(pp. 20±23) on their article published on pp. 12±19. Am. N., Sadek M., Cofer W., Ellenson W., and Kellogg R. Chemical and J. Epidemiol. 1992: 135: 23±25. physical properties of emissions from Kuwaiti oil fires. Water Sci. DockeryD.W., Ware J.H., and Ferris B.G. et al. Change in pulmonary Technol. 1993: 27: 223±233. function in children associated with air pollution episodes. J. Air Whittemore A.S., and Korn E.L. Asthma and air pollution in the Los Pollut. Control 1982: 32: 937±942. Angeles area. Am. J. Public Health 1980: 70: 687±696.

Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(4) 299