Air Pollution and Respiratory Symptoms in Preschool Children 1- 3

CHARLOTTE BRAUN-FAHRLANDER, URSULA ACKERMANN-LIEBRICH, JOEL SCHWARTZ, HANS PETER GNEHM, MARKUS RUTISHAUSER, and HANS URS WANNER

Introduction

Despite considerable effort in the past 20 yr, the effects of both long-term and short-term exposure to air pollution are only partially understood. A number of studies have associated long-term exposure to particulates and sulfur dioxide (S02) with increase in chronic respiratory symptoms (1-5). Other studies have examined oxidant pollutants (6-8). Most of those studies have used annual questionnaires of chronic conditions, such as asthma or chronic bronchitis, as their outcome measure and areawide pollution monitors as their exposure index. To assess the impact of short-term exposure to air pollution, daily diaries of acute respiratory symptoms represent an attractive tool. Diary data may be a more sensitive indicator of respiratory symptoms and less subject to recall bias. These have recently been used to look at acute effects of air pollution with some success (9, 10). Individual monitors represent an improvement in the specificity of the exposure measure. More sensitive indicators of respiratory symptoms and more precise indices of exposure should aid in discovering the impact of air pollution, particularly since it is not the major source of respiratory symptoms. The principal hypothesis of this study was that moderate-term (6 wk) exposure to air pollution is associated with an increased incidence and duration of respiratory symptoms in preschool children, a population with airways particularly sensitive to the irritant effects of air pollution. In 1984, when the study was planned, the main interest in Switzerland focused on N0 2 as an important air pollutant because it clearly exceeds the Swiss air quality standards in urban and suburban areas, whereas S02 and total suspended particulates (TSP) are close to or within these standards (table 1). Elevated ozone concentrations have only recently been recognized as a major air pollution problem in Switzerland. The Swiss air quality standards correspond to World Health Organization recommendations but are generally lower than current U.S. standards. 42

SUMMARY A diary stUdy on a random sample of 625 Swiss children aged 0 to 5 yr was conducted In two cities In Switzerland to Investigate the association between air pollution and respiratory symptoms. Total suspended particulates (TSP), SO, and NO, were measured by city monitor. In addition, passive samplers Inside and outside the home measured NO, concentration during the 6 wk each child was on the diary. Diaries were filled out by parents, and 20% were validated with the attending pediatrician'S case notes. Incidence and duration of symptom episodes were examined separately. The study included any episode, episodes of coughing without runny nose, upper respiratory episodes, and episodes of breathing difficulty. In regressions using 6-wk average pollution that controlled for medical history, NO, measured outdoors but not indoors was associated with the duration of any symptom. Total suspended particulates were a more significant predictor of duration of any symptom than NO,. The 6-wk average TSP was significantly associated with Incidence of coughing episodes and marginally significant as a predictor of upper respiratory episodes. Previous day'S TSP wes a significant predictor of Incidence of upper respiratory symptoms. Annual average of NO, was associated with the duration of any episode and of upper respiratory episodes. We conclude that the Incidence and duration of respiratory symptom episodes are likely associated with particulate concentrations and duration may be associated with NO,. AM REV RESPIR DIS 1992; 145:42-47

Respiratory symptom diaries werefilled out over a 6-wk period. Exposure data included N0 2 concentrations measured by passive samplers (Palmes tubes) inside each child's home and immediately outside the home, as well as additional pollutants from existing monitors in the two cities, Basel and Zurich. In the suburban and rural study area no air pollution monitors existedand the N0 2passive sampler was the only exposure measure. This paper mainly focuses on the subsample of children living in Basel and Zurich for whom more complete exposure data were available. The averaging times appropriate for studying the health effects of air pollution have not yet been established. Some studies have used annual average pollution measures (1-3, 11, 12) and other short-term averages (9, 13, 14). A first cross-sectional analysis of the present study looked at the prevalence of respiratory symptoms and the 6-wk average exposure of N0 2and found it significantly related after adjustment for relevant covariates (15). We have now examined three different time scales-6-wk averages, short-term exposure (1 day or means of several days), and in addition annual averages - for both incidence and duration and included additional pollution and weather data.

Methods Diary Data A sample of 1,800children was chosen in four locations in Switzerland: the cities of Basel and Zurich, the suburban community of Wetzikon, and a rural area Rafzerfeld in the canton of Zurich. From a list of inhabitants in each region a random sample of children up to 5 yr of age was drawn. The subsample in Zurich and Basel originally consisted of 840 children of whom 625 were included in this analysis. Because of logistic considerations, 1/12 of the random sample was recruited each month between November 1985and November 1986. Participation was evenly spread over the year. Recruitment was initiated by a letter explain-

(Received in original form April 9, 1990 and in revised form July 16, 1991) 1 From the Division of Social and Preventive Medicineand the Department of Pediatrics, University of Basel, the Department of Pediatrics, University of Zurich, and the Institute of Hygiene and Occupational Physiology, Federal Institute of Technology, Zurich, Switzerland, and the U.S.Environmental Protection Agency, Washington, D.C. , Supported by Swiss National Research Foundation Grant No. 3.809.0.84 and by the Government of the Canton of Zurich. 3 Correspondence and requests for reprints should be addressed to Charlotte Braun-Fahrlander, Divisionof Social and Preventive Medicine,University of Basel, Steinengraben 49, CH-4051 Basel, Switzerland.

43

AIR POLLUTION AND RESPIRATORY SYMPTOMS IN PRESCHOOL CHILDREN

TABLE 1 TYPICAL ANNUAL MEAN CONCENTRATIONS OF POLLUTANTS IN SWITZERLAND

Rural area Suburbs Cities Swiss air quality standards

30

SO, (;tg/m 3 )

NO, (;tg/m 3 )

TSP (;tg/m 3 )

8-12 30-40 50-70

20-30 30-50 60-140

35-40 40-45 50-55

~g/m3

30

annual mean

80

U.S. air quality standards

~g/m3

annual mean

~g/m'

70

annual mean 100 ~g/m' annual mean

0" 600-800 350-450 80-150

~g/m3

annual mean 75 ~g/m' annual mean

120 ~g/m3 maximum 1-h average

235

~g/m'

maximum 1-h average

• Number of hours with ozone concentration> 120 ug/m3 (April through September).

ing the aims of the study. All participating families werevisited by a specially trained female physicianwho conducted a standardized interview with the parents using a questionnaire. The questionnaire included questions on their social situation, family size, and living condition, the health status of the child, the parents' respiratory health, and the smoking habits of the family. Parents were asked to record their child's respiratory symptoms (from a given list) daily on the diary form. This form was carefully explained to every family. Every child participated in the study for 6 wk. The symptoms recorded included cough during the day, cough at night, sore throat, runny nose, fever, earache, and breathing difficulty (symptoms easily recognized by lay people). Physician visits or administration of drugs to the child and respiratory symptoms of other members of the family were also recorded. Of the diaries 20070 were validated with the attending pediatrician's case notes, and an agreement of 87% was found. This analysis is limited to Swiss nationals because the number of immigrant familieswas small overalland the participation rate among immigrants below 50%. We further excluded from analysis families who did not return the passive NO, samplers properly or did not record the exposure time properly (102). This left a sample of 625 subjects living in the cities of Basel and Zurich. The numbers of subjects and participation rates, by age and sex, in each of the four locations is shown in table 2. The study population was still representative for the target population. There was a modest tendency of "unreliable" families to belong more often to the lower social class

and to smoke more often, but the differences were small and far from being statistically significant.

A ir Pollution Data Ambient air NO, was measured with passive samplers (Palmes tubes) (16)located outside the apartment and in the room in which the child stayed most frequently. The samplers were changed weekly for 6 wk. The parents received a new set of sampler every week and were thus reminded of the study. NO, absorbed by the samplers' coating was extracted and analyzed with a colorimetric reaction (Salzmann) at the Federal Institute of Technology (17). In addition, data from all outdoor air pollution monitors in Basel and Zurich were obtained, and average concentration of TSP, NO" SO" and ozone over the same 6-wk time periods for each child were computed. Weather data were obtained for all four locations and again averaged over 6 wk. Daily averages of TSP, NO" SO" ozone, and weather werealso computed. Children in Basel and Zurich were living within 4 to 6 km of the city monitors SO, was measured as half-hour means with an ultraviolet fluorescent analyzer (Horiba APSA 35DE; Horiba Ltd., Kyoto, Japan). Half-hour means of NO, were measured by chemiluminescence with a Tecan CLD 502 (Tecan AG, Hombrechtikon, Switzerland). The half-hour mean concentrations of ozone weremeasured by ultraviolet absorption using a Monitor Labs 8810(Monitor Labs Ltd., San Diego,CA). TSP wasmeasured as daily means using a digital high-volume sampler with an automatic filter exchange (Digital AG, Hegnav, Switzerland).

TABLE 2 PARTICIPANTS IN THE FOUR STUDY REGIONS ACCORDING TO AGE, SEX, AND PARTICIPATION RATES Age (Months)

Basel Zurich Total cities Wetzikon Rafzerfeld Total

0-12

13-24

25-36

37-48

49+

Boys

Girls

Total

Total contacted

95 45 140 22 35 197

94 45 139 49 45 233

84 38 122 47 57 226

88 38 126 39 60 225

65 33 98 36 48 182

201 98 299 89 122 510

225 101 326 104 123 553

426 199 625 193 245 1,063

575 265 840 277 314 1,431

Participation Rate, Reliables Only

(%) 74 75 74 70 78 74

Statistical Methodology The symptoms reported on the diary do not occur independently. Rather, certain complexes of symptoms are associated with different types of illness. To identify the appropriate groupings of symptoms into symptom complexes,weused factor analysis combined with medical judgment. The complexes identified were coughing without runny nose, breathing difficulty, and fever combined with earache and sore throat, which we refer to as an upper respiratory complex (URI). The factors that increase the risk of acquiring an illness are not necessarily the same as those that increase its duration. The best predictor of symptom prevalence today is symptom prevalence yesterday. This serial correlation can obscure important relationships with the incidence of illness. To reduce these problems we focused separately on the incidence and duration of the symptom complexes defined previously. This approach has proved successful in other diary data (9, 10). Incidence requires that the previous day be symptom free. If symptoms were present the previous day, incidence was defined as missing. The analysis focused on the 625 children of Basel and Zurich and proceeded in several stages. First, models for incidence and duration werechosen without using air pollution. Two approaches were used to choose the models. Our primary approach was to use a priori judgment to determine the covariates that belong in the regression model. As risk factors for symptom incidence wechose asthma history as a measure of bronchial hyperresponsiveness, bronchitis history and frequent colds as a measure of susceptibility to infections, social class as a possible modifier of reporting behavior, sex, and passive smoking exposure. A dummy variable for region was also included. Our priors on the impact of weather wereless strong, and dummy variables for season and a continuous temperature variable were included when significant. For symptom duration our priors were different. Two recent studies (9, 10)reported that risk factors for symptom incidencegenerally do not modify the length of symptom persistence for the symptoms of such diary studies. Here we controlled for social class as a modifier of reporting behavior and membership in a play group, because reinfection could act to prolong some symptoms. Again, weather terms wereincluded when significant. As an alternative to this procedure, weused stepwiseregression to choose models for each outcome from among the variables listed in table 3. The second step of this analysis tests the hypothesis that medium-term averagesof pollution, defined as 6-wk means for each subjet, could influence either symptom incidence or duration. The incidence of symptoms was relatively rare, and therefore the counts of the number of incidents of each symptoms complex that each child suffered during the child's 6-wk monitoring period were modeled using Pois-

44

BRAUN·FAHRLANDER, ACKERMANN-LIEBRICH, SCHWARTZ, GNEHM, RUTISHAUSER, AND WANNER

TABLE 3 DISTRIBUTION OF ALL COVARIATES ACCORDING TO STUDY SITE

Covariates Age Under 2 yr 2-5 yr Sex Male Female Parental education Level I (low) Level II Level III Level IV (high) Crowding More than 1 person/room 1 person/room Less than 1 person/room Play group Member of play group or kindergarten No Cooking range Gas Electric Smoker in family No Yes Total number of cigarettes/day Up to 10 cigarettes/day 10-19 cigarettes/day 20 and more cigarettes/day Number of cigarettes/day smoked by the mother Nonsmoker Up to 10 cigarettes/day 10-19 cigarettes/day 20 and more cigarettes/day History Asthma Bronchitis Tonsillitis Otitis media Pneumonia Croup Susceptibility to colds Number of participants Winter Spring Summer Autumn

n

%

n

%

n

%

189 237

44.4 55.6

90 109

45.2 54.8

279 346

44.6 55.4

201 225

47.2 52.8

98 101

49.2 50.8

299 326

47.8 52.2

28 121 174 90

6.8 29.3 42.1 21.8

19 42 62 66

10.1 22.2 32.8 34.9

47 163 236 156

7.8 27.1 39.2 25.9

132 173 121

31.0 40.6 28.4

56 104 38

28.3 52.5 19.2

188 277 159

30.1 44.4 25.5

103 323

24.2 75.8

66 133

33.2 66.8

169 456

27.0 73.0

167 259

39.2 60.8

55 144

27.6 72.4

222 403

35.5 64.5

221 205

51.9 48.1

125 74

62.8 37.2

346 279

55.4 44.6

63 57 83

14.9 13.4 19.6

22 17 28

11.5 8.9 14.6

85 74 111

13.8 12.0 18.0

80 39 52 34

39.0 19.0 25.0 16.6

32 14 16 12

43.2 18.9 21.6 16.2

112 53 68 46

40.1 19.0 24.4 16.5

6 120 70 100 18 47 145

1.4 28.2 16.4 23.5 4.2 11.0 34.1

4 50 38 57 1 25 73

2.0 25.1 19.1 28.6 0.5 12.6 37.1

10 170 108 157 19 72 218

1.6 27.2 17.3 25.1 3.0 11.5 35.0

95 146 104 81

22.3 34.3 24.4 19.0

65 52 44 38

32.7 26.1 22.1 19.1

160 198 148 119

25.6 31.7 23.7 19.0

Results

Zurich

Basel Annual Mean Meteorologic data Daily mean temperature, 0 C Daily mean relative humidity, % Daily sunshine duration, h/day Daily amount of rain, mm/day

8.5 76.5 3.8 2.3

Minimum

Maximum

Annual Mean

-12 52.3

27.1 95.8 14.3 37.0

7.5 82.6 3.9 3.2

son regression(18). In such a model weassume In E(Yj)

= Xi13 +

In

N

where Y, is the number of incidents for the ith child, Xj the risk factors for the ith child, and Nj the number of days the child was eligible to be incident (the number of nonmissing days). E denotes the expected value. Maximum likelihood estimation was used. Duration of illness was modeled using linear regression of the natural logarithm of

Total

Zurich

Basel

ates. Because some causes of respiratory illness are serially correlated over time, failure to include all possible factors in a model can result in correlation between the rates on different days. This lack of independence can, at a minimum, lead to misestimating the standard errors of regression coefficients. Methods for estimating logistic regression models that incorporate the possibility of serial correlations have been developed and used in similar diary studies (9, 10, 19,20). Weused these methods for our logistic regressions. In addition, the hypothesis was tested that annual average differences in air pollution are associated with either incidence or duration (this analysis included all four study areas and was therefore limited to NO z passive sampler as the only exposure measure). Dummy variables for location were used to determine regional differences in symptoms after controlling for individual risk factors. These were then regressed against regional differences in annual average air pollution in a second stage.

o

Minimum

Maximum

-12.8 45.0

24.8 97.9 14.2 49.5

o o

symptom duration for each child. To avoid giving undue influence to a few children with multiple incidents, the mean duration for each child was modeled. Third, the daily rate of symptom incidence was examined, controlling, by stratification, for those individual (time-independent) risk factors that were significant in the first analysis. Logistic regressions of these incidence rates were performed using season, weather, and air pollution as the time-varying covari-

Six-week Averages of Air Pollution Incidence models. For incidence of each of the symptom complexes over a 6-wk period, each pollutant in turn was inserted into the regression model. Neither measure of NO, by passive sampler was associated with incidence of cough in these models, nor was the 6-wk average of the city NO, monitors. SO, and ozone likewise showed no significant associations. TSP, however, was significantly associated with the number of coughing incidences (RR = 1.16,950/0 confidence interval, CI = 1.07, 1.26). The relative risk (RR) shown for TSP is for a 20 ug/m" change in exposure, which is approximately a 1 standard deviation (SD) change. For the incidence of upper respiratory symptoms both TSP (RR = 1.12,95% CI = 1.00, 1.24) and NO, (RR = 1.23, 95% CI = 1.03, 1.48) were significant risk factors. NO, by outdoor passive sampler was marginally significant (RR = 1.19, 95% CI = 0.99, 1.42), but NO, by indoor passive sampler was highly insignificant (RR = 1.03, 95% CI = 0.89, 1.18).All the relative risks are computed for a 20 ug/m" increase in pollution concentration. When both TSP and NO, were included simultaneously as predictors, the significance and magnitude of both were somewhat decreased (RR = 1.08,95% CI = 0.96, 1.22 for TSP and RR = 1.14, 95% CI = 0.93, 1.40 for NO,). No association was found between pollution and the incidence ofbreathing difficulty or any symptom. Duration models. For mean duration of each symptom complex over a 6-wk peri-

45

AIR POLLUTION AND RESPIRATORY SYMPTOMS IN PRESCHOOL CHILDREN

TABLE 4

The mean percentage of children with incident cases of upper respiratory symptoms by quartile of TSP, plotted against the mean TSP level in each quartile, is shown in figure 1. The results have been adjusted by logistic regression to control for season of the year, city of residence, temperature, and risk strata. Note that the interpretation of the seasonal dummies is complicated by the presence of a continuous temperature term.

LOGISTIC REGRESSION FOR UPPER RESPIRATORY SYMPTOM 625 INCIDENCE (DAILY INCIDENCE RATES), N

Intercept Temperature, Springt Winterl Fall§ Basel Risk strataII TSP'

0

C'

J3

SEM

-4.234 -0.0230 -0.2455 -0.703 -0.332 -0.1606 0.413 0.00454

0.2798 0.0131 0.1934 0.319 0.184 0.1266 0.121 0.00174

OR (95% CI)

0.78 0.50 0.72 0.85 1.51 1.10

(0.53-1.15) (0.26-0.94) (0.50-1.04) (0.66-1.10) (1.19-1.93) (1.02-1.19)

Annual Means of Air Pollution

• Previous day's temperature.

t March through May. l December through February. § September through November.

II Otitis history or frequent colds. ~ Previous day's TSP level; relative odds are tor a 21.6 ug/m3 change.

od, again each pollutant was inserted in turn into the regression models previously developed. TSP was a significant predictor of the duration ofany respiratory episode, with a 20 ug/m" increase in TSP associated with a relative duration of 1.12 (95070 CI = 1.004 to 1.24). N0 2 by outdoor individual sampler was also significantly associated with the duration of anyrespiratoryepisode(RD = 1.13,95% CI = 1.01, 1.27). N0 2 by indoor individual monitor was not significantly associated with the duration of any episode (RD = 1.05,95% CI = 0.95, 1.16) and had a substantially smaller effect than outdoor individual sampler. N0 2 by city monitor was highly insignificant. When TSP and N0 2 by outdoor individual sampler were both in the regression model, TSP remained a significant predictor of the duration of any respiratory episode (RD = 1.11,95% CI = 1.03, 1.14), with little change in effect size, and N0 2

1.8

~

'" ::l g ~

, 1.6

::l

>'" >~ 0

1.4

by outdoor individual sampler became insignificant. There was a marginally significant association between TSP and the duration of coughing episodes (RD = 1.06,95% CI = 0.99, 1.14) and a positive trend for the duration of upper respiratory episodes(RD = 1.04,95% CI = 0.97,1.11). None of the other pollutants were even marginally significant as risk factors for these outcomes. For duration of episodes ofbreathing difficulty, N0 2 outdoor individual sampler was a significant risk factor (RD = 1.50,95% CI = 1.04,2.16) and TSP was a marginally significant risk factor (RD = 1.37,95% CI = 0.96, 1.96). N0 2 by indoor individual sampler, in contrast, was quite insignificant, with a smaller effect (RD = 1.12,95% CI = 0.78, 1.59).

Daily Means of Pollution In logistic regressions of daily incidence rates, controlling for city, risk strata, and season, the previous day's temperature was the only significant weather factor. When daily air pollution was added to these models, the only significant relationship was between the previous day's TSP and upper respiratory symptom incidence (/3 = 0.00454 ± 0.0017). This model is shown in table 4.

This analysis added the symptom data and passive N0 2 samplers for the suburban and rural areas. For all four study areas, regional mean incidences rates, adjusted for individual covariates and weather data, were regressed against regional differences in annual mean N0 2 concentrations, measured by passive samplers indoors and outdoors. Only the N0 2 passive sampler could be examined in this additional analysis. The adjusted mean annual incidence rate of all symptom complexes and the mean values of N0 2 measured by passive samplers indoors and outdoors are shown in table 5. As is evident from table 5, there was no association between long-term difference in N0 2levels by region, as indexed by indoor or outdoor samplers, and mean annual rates of respiratory incidence. The adjusted annual mean symptom duration by region and the corresponding N02leveis (measured by passive samplers) are shown in table 6. The pattern here is different, and a second-stage regression of the adjusted natural logarithm of regional mean duration on N0 2 levels yields significant associations between outdoor N0 2levels (by individual sampler) and the average duration of any respiratory episode (RD = 1.11,95% CI = 1.07, 1.16) and upper respiratory episodes (RD = 1.14,95% CI = 1.03, 1.25). A positive trend for the duration of coughing episodes was also seen (RD = 1.09,95% CI = 0.97, 1.22). No association was seen with the duration of breath-

!( ~

li:

00l '"

~

"

"

TABLE 5

1.2

ADJUSTED ANNUAL SYMPTOM INCIDENCE RATES AND NO, LEVELS BY REGION (MEAN NUMBER OF SYMPTOM EPISODES/CHILDIYEAR), N = 1,063

~

00l

e,

~

::>

"

1.6 0

20

40

60

80

TOTAL SUSPENDED PARTICULATES (ug/m3)

Fig. 1. Mean percentage of children with incident cases of upper respiratory symptoms, by quartile of TSP, plotted against the mean TSP level in each quartile (n =

625).

URI

Cough Incidence

Breathing Difficulty Incidence

NO, In (jlg/m 3 )

NO, Out (jlg/m 3 )

4.38 4.77 5.22 3.72

6.85 4.77 5.27 3.72

0.47 0.47 1.16 0.51

31.31 22.31 19.42 11.10

51.25 46.88 32.65 25.12

100

Region Basel Zurich Wetzikon Rafzerfeld

Any Incidence

14.6 14.6 16.8 13.1

46

BRAUN-FAHRLANDER, ACKERMANN-L1EBRICH, SCHWARTZ, GNEHM, RUTISHAUSER, AND WANNER

TABLE 6 ADJUSTED ANNUAL SYMPTOM DURATION (DAYS) AND NO, LEVELS BY REGION, N ; 1,063

Region Basel Zurich Wetzikon Rafzerfeld

Any Symptom Duration

URI Duration

Cough Duration

Breathing Difficulty Duration

NO, In ()lg/m 3 )

NO, Out ()lglm 3 )

4.50 4.21 4.00 3.88

1.99 1.85 1.62 1.72

2.32 2.01 2.10 2.02

1.55 1.72 3.47 1.25

31.31 27.31 19.42 11.10

51.25 46.88 32.69 25.12

ing difficulties. NO z by indoor individual sampler was also predictive of the duration of any episode (RD = 1.16, 95070 CI = 1.12, 1.21), upper respiratory episodes (RD = 1.18,95% CI = 1.01,1.38), and coughing episode (RD = 1.15,95070 CI, 1.03 to 1.29). When stepwise regression was used to select the variables in table 3 to include in each model for the incidence and duration of symptom complexes, similar results were found. The 6-wk average TSP concentrations were associated with the incidence of coughing and upper respiratory episode with similar effect sizes. None of the NO z levels measured were associated with the incidence of upper respiratory episodes in those models. Similarly, TSP and NO z by outdoor individual sampler, but not indoor sampler, were associated with the duration of any episode, again with similar magnitude. However, the association of TSP with breathing difficulty was more significant (RD = 1.34, 95% CI = 1.01, 1.80) but similar in magnitude. Hence, the results changed little with model specification. Discussion

Incidence of Respiratory Symptoms The previous day's TSP was a significant predictor of the incidence of upper respiratory symptoms in Basel and Zurich. In addition, the 6-wk average ofTSP was significant as a predictor of the number of upper respiratory incidents each child suffered during the 6 wk the child was on the diary. The magnitudes of the associations were quite similar for the two analyses. NO z measured by city monitor was also predictive of 6-wk average incidence rate but was not associated with daily incidence rate. The weaker association with NO z by individual sampler and the high correlation between NO z and TSP suggest that this NO z association may reflect confounding with TSP. TSP was also a significant predictor of the number of coughing episodes a child had during the 6-wk diary period. It was not significant in regressions of daily rates. The finding of an association between

short- and medium-term particulate exposure and the incidence of respiratory conditions is consistent with a developing literature. The Harvard Six-City Study, for example, reported that the prior day's particulate concentration was associated with the incidence of coughing episodes and lower respiratory episodes in children (10). As in this study, the relationship was significant at particulate levels lower than the current U.S. or Swiss standard. A recent study of mildly asthmatic children has associated particulate pollution with an aerodynamic diameter equal to or less than a nominal 10 micrometers [PM 10] with peak flow decrements and respiratory symptoms in Utah (21). In the Tucson study TSP was linked to the prevalence of wheezing symptoms in adults (22). The 2-wk averageparticle concentrations were also associated with the number of days of restricted activity due to respiratory conditions in adults (23). Several other studies indirectly support this association. These include studies linking acute exposure to particulates with decrements in lung function (24-26), with hospital admissions for respiratory conditions (27, 28), with hospital emergency room visits for acute respiratory conditions (29-31), and with increased bronchial secretion in children during colds (32). Acute exposure to particulates has also been associated with increased daily mortality (33, 34). Longer term exposure to particulates have been associated with increased rates of acute bronchitis in children (35) and with the prevalence of chronic bronchitis in adults (36).

Duration of Respiratory Symptoms NO z measured by outdoor personal sampler was associated with the mean duration of any respiratory symptom. However, NOz by city monitor and NO z measured by indoor personal sampler were not associated with symptom duration. The difference between indoor and outdoor measurements is striking and suggests some confounding. The outdoor NO z measurement may be representing some other exposure.

Several pieces of internal and external evidence support this hypothesis. The most likely confounder is TSP. TSP is also associated with the duration of any symptom in this data, and the association is slightly stronger, despite the greater opportunity for misclassification of exposure from the use of a city monitor. The correlation between outdoor passive NO z sampler and TSP measurements was quite high (0.52) and indeed was as high as the correlation between the passive NO z measurement and the city NO z monitors. Moreover, when both pollutants were simultaneously included in the model for symptom duration, TSP remained significant with no change in effect size, and NO z became insignificant. This argues for TSP as the responsible agent. TSP also showed a stronger trend in analyses of the duration of coughing and upper respiratory symptoms, although outdoor NO z correlated more strongly with breathing difficulty. There too, however, the indoor correlation was much weaker. The cross-sectional comparison of the mean duration of respiratory symptoms in each of the four regions in the expanded sample, after adjusting for both individual risk factors and weather, showed a strong association with the NO z passive samplers. The magnitude of that association was quite similar to the effect size seen in the longitudinal study of 6-wk outdoor exposure to NO z in Basel and Zurich. The lack of TSP data for the other two regions precludes our eliminating TSP as a possible confounder in the cross-sectional analysis, but the consistency of the NO z findings cannot be ignored. A recent review of the literature on indoor NO z studies shows that the case for an effect of indoor NO z exposure is weak (37). A recent study with indoor samplers found no significant association with chronic respiratory symptoms (38). It should be noted, however, that essentially all studies have reported positive trends. The number of studies using outdoor NO z measurements is smaller, but a significant association has been reported between daily NO z concentrations and the incidence of respiratory symptoms in a diary study of student nurses (9) and in the prevalence of cough in a diary study of subjects with chronic respiratory conditions (22). In conclusion, the relationship between TSP and both symptom incidence and symptom duration reported in this paper occurs at concentrations that ranged from 30 to 117 ug/m". The median TSP level was 43 ug/rn", and the 75th percentile was only 66Ilg/m3. Hence this entire

AIR POLLUTION AND RESPIRATORY SYMPTOMS IN PRESCHOOL CHILDREN

relationship occurs at particulate concentrations that never exceeded half the U.S. 24-h ambient air quality standard for particulates, when it was measured as TSP. The current PM 10 standard is about equal in stringency. The TSP association was seen using both daily and 6-wk average exposure measures. Duration of symptom episode showed a consistent pattern with trends for each symptom individually as well as for duration of any episode. Many of the symptoms may be of viral etiology, but the role of an irritant in enhancing infectivity and exacerbating symptoms is quite plausible. Given the consistency of these results with other recent findings, we believe the association is likely causal. If the relationship between TSP and symptom incidence and duration in this study is causal, it is likely the particulates less than 10 urn, rather than total particulates, are responsible, since larger particles are not deposited in the respiratory tract. PM 10 and TSP are highly correlated in the same location, but rural areas can have substantially lower ratios of PM lO to TSP. This is because of the higher concentrations of windborne dust, whose size distribution is mostly out of the respirable range. Urban particulates can also differ in chemical composition from rural particles. Hence the associations reported here should be interpreted as relevant for urban particulates, and any rural relationship may be different. The situation with N02 is less clear. Although the association with symptom duration in Zurich and Basel may well be due to confounding with TSP, the cross-sectional association across the four regions supports a possible role. Thus the association between mediumand long-term N0 2 exposure and symptom duration deserves serious consideration as possibly causal. High doses of N02 are associated with impaired intrapulmonary killing of infectious organisms in mice (39). Acute exposures have been associated with increased incidence of respiratory symptoms (7, 9, 40). Whether ambient exposure can prolong the duration of acute respiratory episodes requires further examination. References 1. Ware JH, Ferris BG Jr, Dockery DW, Spengler JD, Stram DO, Speizer FE. Effects of ambient sulfur oxides and suspended particles on respiratory health of pre-adolescent children. Am Rev Respir Dis 1986; 133:834-42. 2. Lunn JE, Knowelden J, Handyside AJ. Patterns of respiratory illness in Sheffield infant school-

children. Br J Prevent Soc Med 1967; 21:7-16. 3. Dockery DW, Speizer FE, Stram DO. Effects of inhaled particles on respiratory health of children. Am Rev Respir Dis 1989; 39:587-94. 4. Douglas JWB, Waller RE. Air pollution and respiratory infection in children. Br J Prevent Soc 1966; 20:1-8. 5. Sawicki F, Lawrence PS, eds. Chronic nonspecific respiratory diseases in the city of Cracow, report of a 5-year follow-up study among adult inhabitants of the city of Cracow. Warsaw: National Institute of Hygiene, 1977. 6. Detels R, Sayre JW, Coulson AH, et al. The UCLA population studies of chronic obstructive respiratory disease. Respiratory effect of long term exposure to photochemical oxidants, nitrogen dioxide and sulfates on current and never smokers. Am Rev Respir Dis 1981; 124:673-80. 7. Lebowitz MD, Holbert CJ, Boyer B, Hayes C. Respiratory symptoms and peak flow associated with indoor and outdoor pollutants in the Southwest. J Air Pollut Control Assoc 1985; 35:1154-8. 8. Shy CM, Creason JP, Pearlman ME, McClain KE, Benson FB, Young MM. The Chattanooga schoolchildren study: effects of community exposure to nitron dioxide. 2. Incidence of acute respiratory illness. J Air Pollut Control Assoc 1970; 20:582-88. 9. Schwartz J, Wypig D, Dockery DW, et al. Daily diaries of respiratory symptoms and air.pollution: methodological issues and results. Environ Health Perspect 1991; 90:181-7. 10. Schwartz J, Dockery DW,Ware JH, et al. Acute effects of acid aerosols on respiratory symptom reporting in children. Preprint 89-92.1, Annual Meeting of Air and Waste Management Association, Anaheim, California, June 25-30, 1989. 11. Mostardi RA, Woebkenberg NR, Ely DL, Coulon M, Atwood G. The University of Akron study on air pollution and human health. Effects on adults respiratory illness. Arch Environ Health 1981; 36:250-5. 12. Imai M, Yoshida K, Kitabatake M. Mortality from asthma and chronic bronchitis associated with changes in sulfur oxides air pollution. Arch Environ Health 1986; 41(1):29-35. 13. Ostro BD. The effects of air pollution on work loss and morbidity. J Environ Econ Man 1983; 10:371-82. 14. Ayres J, Fleming D, Williams M, McInnes G. Measurement of respiratory morbidity in general practice in the United Kingdom during the acid transport event of January 1985. Environ Health Perspect 1989; 79:83-8. 15. Braun-Fahrlander C, Ackermann-Liebrich U, Wanner HU Auswirkungen von Luftschadstoffen auf die Atemwege von Kleinkindern. Schweiz Med Wochenschr 1989; 119:1424-33. 16. Palmes ED, Gunnison AF, Di Matto J, Tomczyk C. Personal samplers for nitrogen dioxide. Am Ind Hyg Assoc J 1976; 34:78-81. 17. Hangartner M, Burri P, Monn C. Passive sampling of nitrogen dioxide, sulfur dioxide and ozone in ambient air. Proceedings of the Eighth World Clean Air Congress, Vol. 3, Den Haag, September 11-15, 1989; 681-6. 18. McCullogh 0, Neider JA. Generalized linear models. London: Chapman and Hall, 1983. 19. Liang KL, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986; 73:13-22. 20. Zeger SL, Liang KY. Longitudinal data analysis of discrete and continuous outcomes. Biometrics 1986; 42:121-30. 21. Pope CA, Dockery DW,Spengler JD, Raizenne ME. Respiratory health and PM IO pollution: a daily time series analysis. Am Rev Respir Dis 1991;

47 144:668-74. 22. Robertson G, Lebowitz MD. Analysis of relationships between symptoms and environmental factors over time. Environ Res 1984; 33:130-43. 23. Ostro BD, Rothschild S. Air pollution and acute respiratory morbidity: an observational study of multiple pollutants. Environ Res 1989;50:238-47. 24. Dockery DW, Ware JH, Ferris BG Jr, Speizer FE, Cook NR, Hermann SM. Change in pulmonary function in children associated with air pollution episodes. J Air Pollut Control Assoc 1982; 32:937-42. 25. Dassen W, Brunekreef B, Hoek G, et al. Decline in children's pulmonary function during an air pollution episode. J Air Pollut Control Assoc 1986; 36:1223-7. 26. Spektor DM, Hofmeister VA, Artaxo P, et al. Effects of heavy industrial poll ution on respiratory function in the children of Cubatao, Brazil. A preliminary report. Environ Health Perspect (in press). 27. Pope CA. Respiratory disease associated with community air pollution and a steel mill, Utah Valley. Am J Public Health 1989; 79:623-28. 28. Bates DV, Sizto R. Hospital admissions and air pollution in Southern Ontario: the acid summer haze effect. Environ Res 1987; 43:317-31. 29. Gross J, Goldsmith JR, Zangwill L, Lerman S. Monitoring of hospital emergency room visits as a method for detecting health effects of environmental exposures. Sci Total Environ 1984; 32: 289-302. 30. Bates DV, Baker-Anderson M, Sizto R. Asthma attack periodicity: a study of hospital emergency visits in Vancouver. Environ Res 1990; 51:51-70. 31. Samet JM, Speizer FE, Bishop Y,Spengler JD, Ferris BG Jr. The relation between air pollution and emergency room visits in an industrial community. J Air Pollut Control Assoc 1981; 31:236-40. 32. Spinaci S, Arossa W, Bugiani M, Natale P, Bucca C, de Candussio G. The effects of air pollution on the respiratory health of children: a crosssectional study. Pediatr Pulmonol 1985; 1:262-6. 33. Schwartz J, Marcus A. Mortality and air pollution in London: a time series analysis. Am J Epidemiol 1990; 131:185-94. 34. Wichmann HE, Mueller W, Allhoff P, et al. Health effects during a smog-episode in West Germany in 1985. Environ Health Perspect 1989; 79:89-99. 35. Dockery DW, Speizer FE, Stram DO, Ware JH, Spengler JD, Ferris BG Jr. Effects of inhaled particles on respiratory health of children. Am Rev Respir Dis 1989; 139:587-94. 36. Euler GL, Abbey DE, Hodgkin JE, Magie AR. Chronic obstructive pulmonary disease symptom effects of long term cumulative exposure to ambient levels of total oxidants and nitrogen dioxide in California Seventh-Day Adventist residents. Arch Environ Health 1988; 43:279-85. 37. Samet JM, Marbury MC, Spengler JD. Health effects and sources of indoor air pollution. Part I. Am Rev Respir Dis 1987; 136:1486-508. 38. Dijkstra L, Houthuijs D, Brunekreef B, Akkerman I, Boleij JSM. Respiratory health effects of the indoor environment in a population of Dutch children. Am Rev Respir Dis 1990; 142:1172-8. 39. Parker RF, Davis JK, Cassell GH, etal. Shortterm exposure to nitrogen dioxide enhances susceptibility to murine respiratory mycoplasmosis and decreased intrapulmonary killing Mycoplasma putmonis. Am Rev Respir Dis 1989; 140:502-12. 40. Lebowith MD, Collins L, Holberg CJ. Time series analyses of respiratory responses to indoor and outdoor environmental phenomena. Environ Res 1987; 43:332-41.