Paper 210 Disc

Paediatric and Perinatal Epidemiology 1999, 13, 475488

Air pollution and respiratory illness of children in Sao Paulo, Brazil

Ahed Bhed Ched Dhed Ref marker Fig marker Table marker Ref end
Ref start

Laboratory of Experimental Air Pollution, School of Medicine, o University of Sa Paulo, Brazil, and bDepartment of Economics, Brigham Young University, Utah, USA

Chin A. Lina, Milton A. Martinsa, Sylvia C. L. Farhata, C. Arden Pope IIIb, Gleice M. S. Conceicaoa, Veruska M. Anastacioa, Marcelo Hatanakaa, Wagner C. Andradea, William R. Hamauea, Gyorgy M. Bohma and Paulo H. N. Saldivaa
a

Summary. This investigation reports the association between air pollution and paediatric respiratory emergency visits in Sao Paulo, Brazil, the largest city in South America. Daily records of emergency visits were obtained from the Children's Institute of the University of Sao Paulo for the period from May 1991 to April 1993. Visits were classified as respiratory and non-respiratory causes. Respiratory visits were further divided into three categories: upper respiratory illness, lower respiratory illness and wheezing. Daily records of SO2, CO, particulate matter (PM10), O3 and NO2 concentrations were obtained from the State Air Pollution Controlling Agency of Sao Paulo. Associations between respiratory emergency visits and air pollution were assessed by simple comparative statistics, simple correlation analysis and by estimating a variety of regression models. Significant associations between the increase of respiratory emergency visits and air pollution were observed. The most robust associations were observed with PM10, and to a lesser extent with O3. These associations were stable across different model specifications and several controlling variables. A significant increase in the counts of respiratory emergency visits more than 20% was observed on the most polluted days, indicating that air pollution is a substantial paediatric health concern in Sao Paulo.

Address for correspondence: Dr Paulo Hilario Nascimento Saldiva, Laboratorio de Poluicao Atmosferica Experimental, Departamento de Patologia, Faculdade de Medicina da Universidade de Sao Paulo, Av. Dr Arnaldo 455 cep: 01246-903, Sao Paulo, SP, Brazil. E-mail: pepino@usp.br

475

# 1999 Blackwell Science Ltd.

Paper 210 Disc

476 C. A. Lin et al.

Introduction
Events of urban air pollution have been associated with significant health effects on the exposed population in many studies, in terms of both morbidity111 and mortality.1224 These effects have been observed at pollution levels below current national and international ambient air quality health standards. However, there are still some points that need further clarification, especially the identification of the segments of population at higher risk of being affected during the pollution episodes, and the characterisation of which specific pollutants are responsible for the adverse health effect. In this context, it is useful to conduct studies in a region where daily measurements of several pollutants are available in combination with a large exposed population, to allow the division of the health outcomes into smaller groups of interest, such as children, elderly people and individuals with chronic diseases (mainly respiratory and cardiovascular). Sao Paulo meets most of the criteria and is an excellent place to assess the effects of air pollution on health. It is the most industrialised centre of Latin America and has * 16 000 000 inhabitants. Socio-economic deprivation is present in a significant fraction of this population. There are about 4 500 000 vehicles in the area using three different types of fuel: gasoline, diesel and pure ethanol. o Because of its geographical characteristics, Sa Paulo presents frequent thermal inversions, resulting in substantial increases in air pollution. Furthermore, the city has a network of air pollution-monitoring stations providing daily concentration levels thus facilitating time-series studies relating air pollution with adverse health effects. Elevated levels of air pollution in Sao Paulo have been associated with increased mortality among children and elderly people.23,24 In addition, inflammatory changes in the airways and pulmonary parenchyma have been o observed in experimental animals reared in downtown Sa Paulo.2527 The present investigation reports the association between air pollution and paediatric emergency visits in the city associated with respiratory diseases.

Materials and methods

Daily records of emergency visits were obtained from the Children's Institute of o the University of Sa Paulo (CIUSP) for the period from May 1991 to April 1993. The CIUSP is the largest paediatric hospital in Sao Paulo and receives children under 13 years of age from all parts of the city. This institute is located in the o centre of downtown Sa Paulo, but it serves as a reference centre for the entire city. During the study period, the emergency service of the CIUSP received on average about 180 cases per day. A large proportion of these (30%) represented respiratory illness. In addition, the CIUSP is a teaching hospital with a solid programme of residence in Paediatrics, thus maintaining qualified staff (24 h per
# 1999 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 13, 475488

Paper 210 Disc

Air pollution and respiratory illness in Brazil 477

day, 7 days per week) in its emergency service, providing reliable clinical information on the cause of admission. All medical diagnoses were reviewed by two of us (C.A.L. and S.C.L.F.). On the basis of the diagnosis after hospital discharge, the visits were classified as respiratory (RESP) and non-respiratory causes. Respiratory causes were further divided into three categories: upper respiratory illness (URI), lower respiratory illness (LRI) and wheezing (WHE). URI comprises primarily viral infections that affect the upper airways; LRI comprises children who presented evidence (clinical and/or radiological) of infection of the lower respiratory tract or lung parenchyma; WHE indicates children whose primary reason for assistance was persistent wheezing without evidence of pulmonary infection. Daily records of SO2, CO, particulates (inhaled fraction, PM10), O3, NO2 concentrations, temperature (daily mean of the lowest temperature of the day) and humidity (daily mean of values measured at noon) were obtained from the State Air Pollution Controlling Agency in Sao Paulo (CETESB). Although there are 12 monitoring stations around the city, not all of them provide measurements of all the pollutants listed above. In addition, there are large areas of the town not covered by any monitoring station. As the CETESB monitoring network was designed mainly to assess the air pollution profile of the central parts of the city, we considered the average of the 10 stations more centrally located as indicative of the citywide pollution levels. The number and characteristics of the monitoring stations are presented in Table 1.

Table 1. Pollutant

Characteristics of the air pollution monitoring system Number of stations 10 3 9 3 4 Method Coulometry Chemiluminescence b-Monitor Infrared Chemiluminescence Type of measurement Moving average of 24 h starting at 16.00 h of the preceding day Highest 8-h moving average starting at 16.00 h of the preceding day Highest hourly average starting at 16.00 h of the preceding day

SO2 (mg/m3) NO2 (mg/m3) PM10 (mg/m3) CO (ppm) O3 (mg/m3)

Information on daily hospital emergency visits, pollution and weather variables with the number of days of observation, mean, maximal value, minimal value and standard deviation is presented in Table 2.
# 1999 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 13, 475488

Paper 210 Disc

478 C. A. Lin et al.

Table 2. Summary information on daily hospital emergency visits, pollution and weather variables No. of days of observation Morbidity counts Total Respiratory Lower respiratory Upper respiratory Wheezing Non-respiratory causes Pollutants PM10 (mg/m3) NO2 (mg/m3) SO2 (mg/m3) CO (ppm) O3 (mg/m3) Weather variables Low temperature (8C) Relative humidity (%) 712 617 623 637 692 617 713 621 713 710 705 713 712 Mean 179 56 8 39 9 126 65 163 20 5 67 16 67 s.d. 62 20 5 16 4 54 27 85 8 2 44 3 14 Min. 59 11 1 1 0 17 15 2 4 1 8 5 33 Max. 355 158 33 109 25 307 193 688 60 12 272 21 100

Statistical analysis
Associations between respiratory emergency visits and air pollution were assessed by performing simple comparative statistics, simple correlation analysis and by estimating a variety of regression models. Daily counts of emergency visits can be modelled as potentially following a Poisson process, which presents independent and random occurrences across time or space. If time were divided into discrete intervals such as 24-h periods, or days, daily counts of visits theoretically would be distributed as a Poisson distribution. However, because there may be various factors that influence risk of illness, the Poisson process need not be stationary, i.e. the underlying mean of the process may change because of changes in risk factors. Poisson regression techniques are often used to estimate effects of various factors on illness. For large counts (4 10), the Gaussian distribution is a close approximation of the Poisson distribution and Gaussian (or Normal) regression models may be used. In either case, Poissonian variation would probably account for most of the day-to-day variation in the counts, but the underlying mean on the process may be determined by pollution levels, season, weather or other factors.28 Preliminary analyses were conducted using Gaussian models that allowed for first-order autocorrelation. Because the effects of pollution on respiratory health may be caused by the immediate exposure and previous exposure history, it is necessary to determine an adequate time lag when performing the statistical modelling. The lag structure of these data was evaluated by looking at concurrent# 1999 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 13, 475488

Paper 210 Disc

Air pollution and respiratory illness in Brazil 479

day exposure, prior-day exposure, and 2- to 6-day lagged moving average of exposure. In addition, polynomial distributed lag models were estimated that allowed for lag structures for 1021 days. Different lag structures occurred in the pollutants, for example CO and SO2 did not present consistent results for any lag structure ranging from 1 to 21 days. The effects of O3 and PM10 were consistent for the concurrent day up to a period 4 5 days. The results of this preliminary analysis suggested that a 5-day moving average model accounted for most of the observable association between air pollution and respiratory emergency visits. After these exploratory analyses, Poisson regression was then carried out, and 5-day lagged moving averages of the pollution exposure were used. Daily counts of visits for respiratory causes (RESP, URI, LRI and WHE) were considered as dependent variables in the models. To control for potential long-term time trends and/or seasonality, indicator variables for each month of the study period (a total of 24 indicator variables) were included as independent variables in the models. Potential day-of-week effects were controlled by using indicator variables for each weekday. Weather effects were included in the models by including variables for various ranges of temperature (5 108C, 11148C, 15188C and 4 188C) and two ranges of relative humidity (5 55% and 4 75%). To evaluate how sensitive the estimate effect of pollution was to the variables controlled for in the regression models, a series of models that included various combinations of the independent variables were estimated. Furthermore, to evaluate the exposureresponse relationship, pollution was divided into quintiles and the average numbers of visits for the different classifications of respiratory illness were compared across the quintiles. Regression models that used indicator variables for each quintile of pollution, instead of continuous pollution variables, were also estimated.

Results
Pearson correlation coefficients for selected variables used in the analysis are presented in Table 3. A high degree of collinearity between pollutants and among the different causes of respiratory emergency visits is evident. Based on these correlation coefficients, it is also evident that some measures of air pollution (PM10, SO2 and CO) were significantly associated with respiratory illness. The Poisson regression coefficients and standard errors from the model relating PM10 and all respiratory causes are presented in Table 4. Based on the results of this regression model, there is a statistically significant association between PM10 and paediatric emergency admission as a result of respiratory diseases even while controlling for season, weather and day of week. Significant associations with season and day of the week are also observed. The same basic model presented in Table 4 was also estimated while including all of the other measured pollutants individually and together. These models were
# 1999 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 13, 475488

480 C. A. Lin et al.

# 1999 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 13, 475488

Table 3. RESP LRI URI WHE Other PM10 NO2 SO2 CO O3 Temp. RH

Selected Pearson correlation coefficients for the variables in analysis Total 0.53*** 0.36*** 0.45*** 0.40*** 0.53*** 0.17*** 0.21*** 0.28*** 0.25*** 0.05*** 0.10** 0.17*** RESP 0.70*** 0.95*** 0.44*** 0.23*** 0.27*** 0.06 0.23*** 0.26*** 0.06 0.04 0.15*** LRI URI WHE Other PM10 NO2 SO2 CO O3 Temp

0.53*** 0.23*** 0.53*** 0.15** 0.17*** 0.25*** 0.24*** 0.02 0.06 0.17*** 0.22*** 0.21*** 0.27*** 0.05 0.04 0.14** 0.03 0.14** 0.14**

0.41*** 0.11** 0.02 0.07 0.04 0.01 0.02 0.03

0.09* 0.24*** 0.25*** 0.18*** 0.05 0.18*** 0.16***

0.40*** 0.73*** 0.50*** 0.27*** 0.15*** 0.29***

0.38*** 0.35*** 0.15** 0.12** 0.22***

0.56*** 0.21*** 0.05 0.27***

0.04 0.03 0.12**

0.20*** 0.33***

0.16***

*P 5 0.05; **P 5 0.001; ***P 5 0.0001.

Paper 210 Disc

Paper 210 Disc

Air pollution and respiratory illness in Brazil 481

Table 4. Poisson regression estimates of excess emergency admissions due to respiratory diseases of children in Sao Paulo, Brazil: coefficient (standard error) Variable Pollutant PM10610 (average of 5 days) Month of year indicators June 91 July 91 August 91 September 91 October 91 November 91 December 91 January 92 February 92 March 92 April 92 May 92 Weather variables Low temperature 11148C Low temperature 15188C Low temperature 4 188C Relative humidity (noon) 5 55% Relative humidity (noon) 4 75% *P 5 0.01; **P 5 0.0001. Coefficient (standard error) 0.039 (0.003)** Coefficient (standard error)

0.256 0.359 0.365 0.111 0.339 0.190 0.361 0.533 0.517 0.366 0.218 0.146

(0.039)** (0.040)** (0.039)** (0.039)* (0.039)** (0.039)** (0.043)** (0.044)** (0.044)** (0.042)** (0.040)** (0.140)

June 92 July 92 August 92 September 92 October 92 November 92 December 92 January 93 February 93 March 93 April 93

0.065 0.117 0.966 0.348 0.239 0.238 0.228 0.426 0.450 0.030 0.314

(0.037) (0.039)* (0.181)** (0.042)** (0.039)** (0.040)** (0.042)** (0.045)** (0.044)** (0.039) (0.043)**

0.115 0.034 0.005 0.006 0.103

(0.036)* (0.022) (0.016) (0.014) (0.015)**

Day of week indicators Monday 0.315 Tuesday 0.501 Wednesday 0.340 Thursday 0.351 Friday 0.298 Saturday 0.183

(0.022)** (0.021)** (0.022)** (0.022)** (0.022)** (0.223)**

estimated for all respiratory illness and for the three subcategories of respiratory emergency visits. The Poisson regression coefficients on the pollutants from these models are presented in Table 5. Positive, statistically significant effects of PM10 were observed for all categories of respiratory emergency visits. The inclusion of other pollutants in the models did not substantially modify the estimated coefficients of PM10, suggesting that this pollutant has an independent effect on RESP. O3 also presents an independent and significant role on respiratory visits. The remaining pollutants presented significant associations with RESP when considered individually in the models. However, for these other pollutants, significant positive associations were not observed when controlling for copollutants. In fact, after the inclusion of other pollutants, some of them presented negative estimated coefficients. Table 6 presents the estimated coefficients of the association between PM10 and respiratory admissions across regression models that progressively add additional variables to control for season, day of week, weather, non-respiratory visits, ozone
# 1999 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 13, 475488

Paper 210 Disc

482 C. A. Lin et al.

Table 5. Relative risks (RR) of all respiratory illness, lower and upper respiratory illness, wheezing and corresponding 95% confidence interval [95% CI] for each pollutant included individually or together with other pollutants (5-day moving average) Pollutant Respiratory illness PM10 (mg/m3) O3 (mg/m3) SO2 (mg/m3) NO2 (mg/m3) CO (ppm) Lower respiratory illness PM10 (mg/m3) O3 (mg/m3) SO2 (mg/m3) NO2 (mg/m3) CO (ppm) Upper respiratory illness PM10 (mg/m3) O3 (mg/m3) SO2 (mg/m3) NO2 (mg/m3) CO (ppm) Wheezing PM10 (mg/m3) O3 (mg/m3) SO2 (mg/m3) NO2 (mg/m3) CO (ppm) *P 5 0.03. Individually RR [95% 1.040 [1.034, 1.022 [1.016, 1.079 [1.052, 1.003 [1.001, 1.206 [1.066, CI] 1.046]* 1.028]* 1.107]* 1.005]* 1.364]* Together RR [95% CI] 1.052 [1.040, 1.065]* 1.015 [1.009, 1.021]* 0.938 [0.900, 0.977] 0.996 [0.994, 0.998] 0.945 [0.808, 1.105] 1.080 [1.049, 1.112]* 1.010 [0.994, 1.026] 0.872 [0.783, 0.971] 0.990 [0.982, 0.998] 0.971 [0.641, 1.472] 1.048 [1.036, 1.061]* 1.014 [1.006, 1.022]* 0.951 [0.906, 0.999] 0.996 [0.992, 1.000]* 0.944 [0.785, 1.135]* 1.047 [1.019, 1.076]* 1.018 [1.002, 1.034] 0.908 [0.824, 1.002] 0.991 [0.983, 0.999] 0.740 [0.505, 1.085]

1.042 [1.024, 1.060]* 1.002 [0.988, 1.016]* 1.052 [0.984, 1.125) 0.999 [0.991, 1.007] 1.203 [0.867, 1.669] 1.039 [0.960, 1.123]* 1.014 [1.008, 1.020]* 1.075 [1.044, 1.107]* 1.003 [0.999, 1.007] 1.237 [1.072, 1.428]* 1.029 [1.013, 1.046]* 1.025 [1.013, 1.037]* 1.034 [0.975, 1.096] 0.996 [0.990, 1.002] 0.813 [0.606, 1.091]

and the remaining pollutants. The association with PM10 is highly robust and stable across the different combinations of independent variables considered in the models. Table 7 presents the results of similar analysis for O3. The effect of O3 on respiratory illness was more sensitive to model specifications, especially when PM10 was also considered as an independent variable. The study period is divided up by quintiles of PM10. The mean PM10 levels and the unadjusted counts of the different categories of respiratory admissions observed for each of the quintiles of PM10 concentration are presented in Table 8. In these unadjusted data, there appears to be a general concentrationresponse relationship. Regression models that used indicator variables for each quintile of pollution, instead of continuous pollution variables and also included the monthly indicator, weather and day-of-week variables were also estimated. The odds ratios estimated from these models for each of the quintiles are presented in Fig. 1. This
# 1999 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 13, 475488

# 1999 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 13, 475488

Table 6. Comparison of relative risk (RR) and corresponding 95% confidence interval [95% CI] (for an increase of 10 mg/m3 of PM10 and 5day moving average) for the different causes of respiratory emergency visits across different model specifications Model PM10 (mg/m3) only +Month indicators +Day of week indicators +Weather indicators +Non-respiratory visits +O3 (mg/m3] +The remaining pollutants All respiratory RR [95% CI] 1.045 1.039 1.046 1.040 1.044 1.036 1.052 [1.041, [1.033, [1.040, [1.034, [1.038, [1.028, [1.042, 1.049] 1.045] 1.052] 1.046] 1.050] 1.044] 1.063] Lower respiratory RR [95% CI] 1.073 1.043 1.050 1.042 1.047 1.044 1.080 [1.060, [1.027, [1.034, [1.024, [1.029, [1.024, [1.049, 1.085] 1.059] 1.067] 1.060] 1.066] 1.065] 1.112] Upper respiratory RR [95% CI] 1.046 1.040 1.046 1.039 1.046 1.039 1.048 [1.040, [1.032, [1.038, [1.031, [1.038, [1.031, [1.034, 1.052] 1.048] 1.054] 1.047] 1.054] 1.047] 1.063] Wheezing RR [95% CI] 1.027 1.018 1.026 1.029 1.029 1.016 1.047 [1.015, [0.998, [1.010, [1.013, [1.013, [0.998, [1.019, 1.040] 1.038] 1.043] 1.046] 1.046] 1.034] 1.076]

Air pollution and respiratory illness in Brazil

Table 7. Comparison of relative risk (RR) and corresponding 95% confidence interval [95% CI] (for an increase of 10 mg/m3 of O3, 5-day moving average) for the different causes of respiratory emergency visits across different model specifications Model O3 (mg/m3) only +Month indicators +Day of week indicators +Weather variables +Non-respiratory visits +PM10 (mg/m3) +The remaining pollutants All respiratory RR [95% CI] 1.015 1.028 1.026 1.022 1.025 1.013 1.015 [1.011, [1.024, [1.022, [1.016, [1.019, [1.007, [1.009, 1.019] 1.032] 1.030] 1.028] 1.031] 1.019] 1.021] Lower respiratory RR [95% CI] 1.013 1.027 1.026 1.020 1.020 1.005 1.010 [1.005, [1.015, [1.014, [1.006, [1.006, [0.991, [0.994, 1.021] 1.040] 1.038] 1.034] 1.034] 1.019] 1.026] Upper respiratory RR [95% CI] 1.015 1.021 1.020 1.014 1.025 1.012 1.014 [1.009, [1.015, [1.014, [1.008, [1.019, [1.006, [1.006, 1.021] 1.027] 1.026] 1.020] 1.031] 1.018] 1.022] Wheezing RR [95% CI] 1.004 1.026 1.025 1.025 1.026 1.021 1.018 [0.996, [1.014, [1.013, [1.013, [1.014, [1.007, [1.002, 1.012] 1.038] 1.037] 1.037] 1.038] 1.035] 1.034]

Paper 210 Disc

483

Paper 210 Disc

484 C. A. Lin et al.

Table 8. Data on the quintiles of PM10 concentration (expressed in mg/m3) with the corresponding relative risk of all categories of respiratory illness and 95% confidence interval PM10 quintile mean All respiratory 44 52 60 69 98 44 52 60 69 98 44 52 60 69 98 44 52 60 69 98 No. of episodes 46 54 56 59 63 6 7 8 8 10 32 38 40 42 43 8 9 8 9 10 Relative risk 1.00 1.08 1.06 1.21 1.23 1.00 1.12 1.14 1.23 1.34 1.00 1.07 1.08 1.20 1.18 1.00 1.04 1.00 1.12 1.19 95% confidence interval Reference [1.04, 1.12] [1.02, 1.11] [1.16, 1.26] [1.18, 1.28] Reference [1.01, 1.24] [1.02, 1.27] [1.10, 1.38] [1.19, 1.50] Reference [1.02, 1.12] [1.03, 1.13] [1.14, 1.26] [1.12, 1.24] Reference [0.96, 1.14] [0.91, 1.09] [1.03, 1.23] [1.08, 1.31]

Lower respiratory

Upper respiratory

Wheezing

illustrates concentrationresponse relationships that suggest that the effects of PM10 on respiratory admissions are dose dependent with no clear evidence of a safe threshold.

Discussion
The results of the present study demonstrate a significant association between the increase in emergency visits due to respiratory diseases and PM10 and O3 levels in Sao Paulo. These associations were quite stable across different model specifications and across different combinations of several controlling parameters. Although association does not necessarily mean causation, the results of the present study suggest that air pollution induces adverse health effects in the exposed population. First of all, the models used in this study were quite restrictive in terms of the effects of possible confounding variables, as several controlling parameters for weather were used. Secondly, there is coherence between the results of this study with those provided by animal field studies2527
# 1999 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 13, 475488

Paper 210 Disc

Air pollution and respiratory illness in Brazil 485

Figure 1. Log scale of relative risk of all categories of respiratory illness as function of PM 10 quintiles, estimated from a Poisson regression model, controlled for season, weather and day of week.

and epidemiological mortality studies,23,24 which also reported adverse health effects as a consequence of the air pollution in Sao Paulo. One significant problem of ecological studies like this one is to ascribe to individual pollutants the responsibility for the observed effects, because the air pollution scenario in Sao Paulo is quite complex. The measured pollutants exhibit a synchronous variation and are highly correlated (Table 3). Other pollutants relevant to health, such as aldehydes, sulphates and nitrates, are not routinely measured in Sao Paulo and, consequently, were not taken into account in this study. Nevertheless, in this study, the association between air pollution and respiratory illness was dependent primarily on PM10 and O3. Other studies focusing on both morbidity4,7,8,10 and mortality14,1719,21 have also observed adverse health effects associated with PM10 levels in urban areas. PM10 levels outdoors and indoors correlate over time, and this correlation is highest in non-smoking homes.29 It is reasonable to suppose that PM10 levels would be a good proxy of indoor penetration of outdoor pollution, especially in Sao Paulo, where home heating and other isolating procedures are negligible. In fact, PM10 was the pollutant that exhibited the most robust and stable association with all categories of respiratory diseases. The adverse health effects of air pollution were quite acute, being observed within 5 days. The same lag structure was observed for both high30 and low level events11,24 of air pollution and indicates that some respiratory diseases could be
# 1999 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 13, 475488

Paper 210 Disc

486 C. A. Lin et al.

induced by polluted air in a short period of time. Alternatively, it is possible that air pollution exacerbates previously existing respiratory diseases. Animal studies previously performed in Sao Paulo support this hypothesis.26,27 Considering the amount of evidence already available,310,1324 it is likely that the observed association between air pollution and adverse health effects is causal and represents a serious public health problem for children in Sao Paulo. Based on the coefficients depicted in Fig. 1, a large increase in the counts of respiratory emergency visits more than 20% can be observed on the most polluted days. These results are important because they provide clues in estimating the magnitude of the adverse effects of urban air pollution on children's health profile. In addition, these results indicate that the deterioration of air quality may consume a proportion of the economic resources of the health system, which is evidently of social significance in developing countries such as Brazil. As the first o study linking acute respiratory diseases and pollutant levels performed in Sa Paulo, the largest city in South America, the results again confirmed those observed in studies carried out in other cities worldwide.310,1324 The present study was designed to investigate the acute effects of air pollution on children's health. However, the chronic effects of these levels of air pollution should be seriously considered and may explain the high rate (30%) of respiratory diseases as the cause of emergency medical care in our sample. In this context, chronic exposure to Sao Paulo's air pollution has been shown to induce pathological changes in the airways, promoting significant impairment of lung defence mechanisms.3133 Therefore, it is possible that children living in Sao Paulo are more likely to need acute medical support due to respiratory events because of the maintenance of relatively high pollution levels over long periods of time. As an ecological study, we could not answer certain important questions, such as the relationship between socio-economic status and the incidence of acute respiratory diseases in children, and the influence of the children's nutrition status in the incidence of acute respiratory diseases, both in association with air pollution. The population served by the Children's Institute of the University of Sao Paulo Medical School is classically represented by those that need medical care from the public health system and, thus, probably represent the less fortunate or favoured portion of the population. Therefore, there may be some important relationship between the high levels of pollutants and socio-economic conditions of the population studied. These questions are important, but further studies should be performed to provide a better answer. Air pollution in Sao Paulo is caused mainly by automotive emissions,34 with industrial sources playing a minor role. In this context, the city provides a representative picture of the health consequences of a chaotic proliferation of individual transportation without adequate emission control, which may be used as an alert for other large conurbations. Our results suggest substantial potential health benefits of controlling air pollution emission from motor vehicles.
# 1999 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 13, 475488

Paper 210 Disc

Air pollution and respiratory illness in Brazil References
1 Lawther PJ, Waller RE, Henderson M. Air pollution exacerbation of bronchitis. Thorax 1970; 25:525529. 2 Gross J, Goldsmith JR, Zangwill L, Leman S. Monitoring of hospital emergency room visits as a method for detecting health effects of environmental exposures. Science of the Total Environment 1984; 32:289302. 3 Ostro BD, Rothschild S. Air pollution and acute respiratory morbidity: an observational study of multiple pollutants. Environmental Research 1989; 50:238247. 4 Dockery DW, Speizer FE, Stram DO, Ware JH, Spengler JD, Ferris BG Jr. Effects of inhaled particles on respiratory health of children. American Review of Respiratory Disease 1989; 39:587594. 5 Bates DV, Baker-Anderson M, Sizto R. Asthma attack periodicity: a study of hospital emergency visits in Vancouver. Environmental Research 1990; 51:5170. 6 Krupnick AJ, Harrington W, Ostro BD. Ambient ozone and acute health effects: evidence from daily data. Journal of Environmental Economics and Management 1990; 18:118. 7 Pope CA III. Respiratory hospital admissions associated with PM10 pollution in Utah, Salt Lake, and Cache valleys. Archives of Environmental Health 1991; 46:9097. 8 Sunyer J, Anto JM, Murillo C, Saez M. Effects of urban air pollution on emergency room admissions for chronic obstructive pulmonary disease. American Journal of Epidemiology 1991; 134:277286. 9 Thurston GD, Ito K, Kinney PL, Lippman M. A multiyear study of air pollution and respiratory hospital admissions in three New York State metropolitan areas: results for 1988 and 1989 summers. Journal of Exposure Analysis and Environmental Epidemiology 1992; 2:429450 10 Schwartz J, Koenig J, Slater D, Larson T. Particulate air pollution and hospital emergency visits for asthma in Seattle. American Review of Respiratory Disease 1993; 147:826831. 11 Schwartz J. Air pollution and hospital admissions for the elderly in Detroit,Michigan. American Journal of Respiratory and Critical Care Medicine 1994; 150:648655. 12 Glaser M, Greenburg L. Air pollution, mortality, and weather: New York City, 1960 1964. Archives of Environmental Health 1971; 22:334343. 13 Fairley D. The relationship of daily mortality to suspended particulates in Santa Clara county, 19801986. Environmental Health Perspectives 1990; 89:159168. 14 Katsouyanni K, Hatzakis A, Kalandidi A, Trichopoulos D. Short-term effects of atmospheric pollution on mortality in Athens. Archives of Hellenic Medicine 1990; 7:126132. 15 Kinney PI, Ozkaynak H. Associations of daily mortality and air pollution in Los Angeles County. Environmental Research 1991; 54:99120. 16 Schwartz J. Particulate air pollution and daily mortality in Detroit. Environmental Research 1991; 56:204213. 17 Pope CA, Schwartz J, Ransom M. Daily mortality and PM10 pollution in Utah Valley. Archives of Environmental Health 1992; 42:211217. 18 Schwartz J. Particulate air pollution and daily mortality: a synthesis. Public Health Reviews 1992; 19:3960. 19 Schwartz J, Dockery DW. Particulate air pollution and daily mortality in Steubenville, Ohio. American Journal of Epidemiology 1992; 135:1220. 20 Schwartz J, Dockery DW. Increased mortality in Philadelphia associated with daily air pollution concentrations. American Review of Respiratory Disease 1992; 145:600604. 21 Dockery DW, Pope CA III, Xu X, Spengler JD, Ware JH, Fay ME, et al. The association between air pollution and mortality in six US cities. New England Journal of Medicine 1993; 329:17531754.
# 1999 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 13, 475488

487

Paper 210 Disc

488 C. A. Lin et al.

22 Schwartz J. Air pollution and daily mortality in Birminghan, AL. American Journal of Epidemiology 1993; 137:11361147. 23 Saldiva PHN, Lichtenfels AJFC, Paiva PSO, Barone IA, Martins MA, Massad E, et al. Association between air pollution and mortality due to respiratory diseases in children in Sao Paulo, Brazil: a preliminary report. Environmental Research 1994; 65:218225. 24 Saldiva PHN, Pope CA, Schwartz J, Dockery DW, Lichtenfels AJ, Salge JM, et al. Air pollution and mortality in elderly people: a time-series study in Sao Paulo,.Brazil. Archives of Environmental Health 1995; 50:159163. 25 Bohm GM, Saldiva PHN, Pasqualucci CA, Massad E, Martins MA, Zin WA, et al. Biological effects of air pollution in Sao Paulo and Cubatao. Environmental Research 1989; 49:208216. 26 Saldiva PHN, King M, Delmonte VLC, Macchione M, Parada MAC, Daliberto ML, et al. Respiratory alterations due to urban air pollution: an experimental study in rats. Environmental Research 1992; 57:1933. 27 Lemos M, Lichtenfels AJFC, Amaro E Jr, Macchione M, Martins MA, King M, et al. Quantitative pathology of nasal passages in rats exposed to urban levels of air pollution. Environmental Research 1994; 66:8795. 28 Pope CA, Schwartz J. Time series for the analysis of pulmonary health data. American Journal of Respiratory and Critical Care Medicine 1996; 156(6Pt.2):S229S233. 29 Quackenboss JJ, Leibowitz MD, Crutchfield CD. Indoor-outdoor relationships for particulate matter: exposure classifications and health effects. Environment International 1989; 15:353360. 30 Her Majesty's Stationery Office. Mortality and Morbidity During the London Fog of December 1952. Report No 95 on Public Health and Medical Subjects. London: Her Majesty's Public Health Service, 1954. 31 Kehrl HR, Vincent LM, Kowalsky RJ, Horstman DH, O'Neil JJ, McCartney WH. Ozone exposure increases respiratory epithelial permeability in humans. American Review of Respiratory Disease 1987; 135:11241128. 32 Devlin RB, McDonell WF, Mann R, Becker S, House DE, Schreinemachers D, et al. Exposure of humans to ambient levels of ozone for 6.6h causes cellular and biochemical changes in the lung. American Journal of Respiratory, Cell and Molecular Biology 1991; 4:406 411. 33 Koren HS, Devlin RB, Becker S, Perez R, McDonnell WF. Time-dependent changes of markers associated with inflammation in the lungs of humans exposed to ambient levels of ozone. Toxicologic Pathology 1991; 4:406411. 34 Massad E, Saldiva PHN, Saldiva CD, Caldeira MPR, Cardoso LMN, Morais AM, et al. Toxicity of prolonged exposure to ethanol and gasoline auto engine exhaust gases. Environmental Research 1986; 40:479486.

# 1999 Blackwell Science Ltd. Paediatric and Perinatal Epidemiology, 13, 475488