The design of this study has already been described 11. In brief, one hundred and fity eight asthmatic children, attending the Hospital Infantil de Mexico Federico Gómez, one of the largest pediatric hospitals in the city (Mexico City) were invited to participate in the study. The diagnosis and severity of their asthma was assessed in terms of clinical symptoms and response to treatment and rated by a pediatric allergist as either mild (intermittent or persistent), moderate, or severe, following the guidelines described in the Global Initiative for Asthma (GINA) (Global Initiative for Asthma 2006) 12. Fifty nonasthmatic children were recruited on a voluntary basis, by requesting that each asthmatic child should invite a schoolmate or a friend from their neighborhood. The children in each group ranged between 6 and 14 years of age. They lived in the study area, attended public schools located close to their homes, their attendance was voluntary and they were not selected using probability-based sampling.

Asthmatic and healthy children were recruited between July 2003 and March 2005 and followed up for an average of 22 weeks and evaluated at the same hospital every 2 weeks. At the beginning of the study, parents completed a general-purpose questionnaire, (adapted from existing survey material), outlining sociodemographic variables, past health history, and potential indoor environmental exposures (tobacco smoke and pets in the home). Information concerning allergy test results, medication, and medical visits during the 2 preceding years was obtained from the medical record. At the first visit and every 15 days thereafter, children were given a symptoms diary to be filled out by the mother. This diary was reviewed by the health staff at each visit. All procedures were explained to the parents, who signed an informed consent form. The children also gave their informed assent. Complete data from the daily dairies was available for the147 asthmatic children and 50 healthy children and this was included in the analysis.

Daily records from the health diary provided information on the presence or absence of coughing, wheezing (defined as wheezing and/or difficulty breathing) and on the child's medication use (corticosteroids, such as beclometasone and flixotide among others, and bronchodilators for example salbutamol, as well as others).

Exposure was estimated from outdoor PM_2.5 (particulate matter < 2.5 μm in aerodynamic diameter), NO₂ and O₃ concentrations, as recorded by the Mexico City government at four fixed-sites for central monitoring [Red Automática de Monitoreo Afmosférico (RAMA)] at locations within the study area (Cerro de la Estrella and Hangares, Merced, Universidad Autonoma Metropolitana, Iztapalapa and La Perla). Daily average, maximum moving average and 8-hr maximum ozone, nitrogen dioxide, and PM_2.5 concentrations along with meteorologic data (temperature and humidity) were obtained for all (505) days, during the study period. Sulfur dioxide and carbon monoxide were not taken into consideration because of the low levels of these pollutants during the study period and their high correlation with other pollutants (correlation CO with NO₂ r = 0.65 p < 0.00 and correlation of SO₂ and PM_2.5 r = 0.47 p < 0.00).

The home of each participating child was georeferenced using a geographic information system (GIS), and each child was assigned to the closest monitoring station. All children attended public schools, located close to their homes, and no fixed-site monitoring station was located > 5 km from a child's residence or school.

We also conducted monitoring at each school for three 15-day periods during the follow-up, in order to validate data obtained from the fixed-site monitoring stations (RAMA). Local, daily 24-hr average PM_2.5 was determined using Mini-Vol portable air samplers (version 4.2; Airmetrics, Eugene, Oregon, USA) with 47-mm Teflon filters (R2PJ047; Pall Gelman, Ann Arbor, MI, USA) and flows set at 5 L/min, and 7-day integrated data for NO₂ and O₃ concentrations were obtained using Ogawa passive samplers (Ogawa USA, Pompano Beach, FL, USA) 13. Samplers were located outside the 37 schools, usually on the roof, at a height of up to 4 m and far from any objects (e.g., trees, buildings) that might prevent air flow. Gravimetric analysis of the 47-mm Teflon filtres was performed at the air laboratory of the National Center for Environmental Research and Training (CENICA), in Mexico City. The NO₂ and O₃ filters were assembled at the Mexico City laboratory. Following exposure, all badges were placed in sealed bags and sent to the Harvard School of Public Health for chemical analysis 14. In addition, in a subsample of filters (n = 207, 11.5%), PM_2.5 absorbance (as a marker of diesel "soot") was measured according to the procedures previously described by Watson,13. Black or elemental carbon (EC) mostly originates from incomplete combustion of fossil fuels and is the main factor for light particle absorption. The absorbtion coefficient is a good marker for EC when compared with the variability of PM_2.5 levels and absorbance of PM_2.5 filters provide reliable information on the variability of EC levels related to traffic exhaust 15. By referring to a digital map of the Iztapalapa area (ArcView software), we selected the intersections between avenues, close to the children's residences and carried out traffic counts, applying categories to types of motor vehicle (Table 1). Traffic density was measured with pneumatic sensors every day for one week, at 7 geographical points from May to July, 2004. The week selected was considered representative concerning traffic in that area. The average density of motor vehicles according to time of day, type of vehicle, and day of the week was calculated for the study period. Vehicle type was defined, depending on the type of fuel used, as follows: A: Private cars (Gasoline); B: Small buses for public transportation (SBPT) (Gasoline or natural gas); C: School buses, other buses, pick up trucks and heavy trucks (Diesel) because our intention was to evaluate whether diesel-fueled vehicles had a greater impact on respiratory symptoms, than vehicles using other types of fuel.

The protocol for this study was approved by the biosecurity, ethics and research committees of the participating institutions (Instituto Nacional de Salud Publica de Mexico and Hospital Infantil de Mexico "Federico Gomez", Hospital Pediatrico Iztapalapa).

A bivariate analysis was carried out, where the basal characteristics of asthmatic and healthy children were compared, using the t-test (under the normality assumption), the Fisher exact test, or the χ² test where appropriate. The incidence density of symptoms and bronchodilator use episodes were calculated with reference to the health diary. An epidsode was defined as a manifestation of respiratory symptoms or the need for bronchidilator use, during 2 or more days and a child had to be free of symptoms or bronchodilator use for at least 3 days in order for the initiation of a new episode to be defined.

In order to analyze the occurrence of daily symptoms or bronchodilator use, we employed mixed models with random intercept, with a binary response for longitudinal data, in order to determine the effects of a pollutant (O₃, NO₂, PM_2.5) 16. The models were run, evaluating pollution effects on the same day and as a result of days prior to the occurrence of the event, assessing accumulation lags for pollutants (from 2 to 5 days). Models were adjusted for sex, severity of asthma, atopy, minimum temperature from the previous day and chronologic time. Other variables such as age, body mass index (weight/height²), socioeconomic status, outdoor activities, exposure to environmental tobacco smoke, pets, carpet in the home, and season did not modify the regression coefficients by more than 1%. The goodness of fit for longitudinal models was evaluated and χ² values relative to deviance in each of these models did not indicate any lack of adjustment 16. We also determined the association between respiratory symptoms and the amount of PM_2.5 absorbed in a subsample of filters collected in 20 schools. The analysis included 66 children for a total of 376 days of follow up (mean of 6 days per child), using mixed models.

Poisson models were used in order to analyze the association of road traffic with the incidence of symptoms and bronchodilatador use 17. The dependent variable included the number of new events of respiratory symptoms or bronchodilator use. The exposure variable included the assigned vehicle count, considering the nearest avenue to the child's residence (that is because traffic density was measured on a specific day during the study period, for each intersection selected). The time that each child participated in the study was considerated as offset. These models were also adjusted by sex, severity of asthma, atopy and the distance between the traffic road and the child's residence. We explored the absence of overdispersion and the residuals. We used STATA 9.0 (StataCorp, Texas USA 2005) for statistical analysis.

The 24-hr average PM_2.5 level was 27.8 μg/m³ (standard deviation (SD) = 14.9). The 1-hr maximum average O₃ level was 86.5 ppb (SD = 34.4), and the 1-hr maximum average NO₂ level was 68.6 ppb (SD = 25.8). The mean daily PM_2.5 absorbance was 10.3 (SD = 4.9) 10^-5 m^-1 (Table 1). PM_2.5 were significantly correlated with O₃ and NO₂ (r = 0.54, and 0.62, respectively). The correlation between O₃ and NO₂ was 0.48. Traffic density was regrouped according to vehicle type and expressed as a 24-hour hourly average. The largest fraction of vehicles consisted of private cars (76.1%); SBPT represented 15.3%, school buses and other buses, pick up trucks and heavy trucks represented 8.7% of the hourly 24-h average(data not shown).

The incidence of symptoms episode among asthmatic children during follow-up was remarkably greater than that among healthy children. The incidence of coughing episode was 3 times greater (IRR = 3.1; 95% CI: 2.8 to 3.4) among asthmatic children than healthy children, and the incidence of wheezing was 8 times greater (IRR = 8.5; 95% CI: 6.4 to 11.4) (Table 3).

Among asthmatic children, coughing and wheezing at any time of the day were associated with O₃ and NO₂ and the effect appeared to increase with cumulative exposure over several days. For PM_2.5 the greatest effect was observed on the same day as the exposure. The risk of coughing among asthmatics increased by 8.9% (95% CI: 3.0% to 15.1%) and the risk of wheezing increased by 10.0% (95% CI: 3.2% to 17.3%), where there was an increase of 48 ppb interquartile range (IQR) in O₃ levels (1 hr maximum) on the previous day. These effects were stronger when cumulative exposure during 2 to 5 days was considered. The same pattern was observed for NO₂. The risk of coughing increased by 7.5% (95% CI: 1.8% to 13.6%) and the risk of wheezing increased by 9.1% (95% CI: 2.3% to 16.4%) when NO₂ levels (1-hr maximum) increased by 34 ppb (IQR) on the previous day, and these effects were stronger when the cumulative exposure from 2 to 5 previous days was considered (Figure 1). PM_2.5 24-hr average levels were only significantly related to respiratory symptoms at night, when recorded that same day. An increase in the IQR (17.4 μg/m³) of PM_2.5 was associated with an 11.1% increase in coughing (95% CI: 5.6% to 16.9%) and with an 8.8% increase in wheezing (95% CI: 2.4% to 15.5%). The effects of all pollutants on bronchodilator use were similar to those observed concerning symptoms (Figure 1). No significant interaction between PM_2.5, NO₂ or ozone and season (warm) (May to September) or cold (October to April)) was observed to effect the risk of respiratory symptoms. However the effect of ozone and PM_2.5 on the risk of wheezing was slightly greater during the warm season.

PM_2.5 absorbance was positively related to respiratory symptoms. An increase of an IQR (8.5 absorbance 10^-5m^-1) was associated with a 55% increase in coughing risk (OR = 1.55; 95% CI: 0.89 to 2.69) and with a 50% increase in wheezing risk (OR = 1.50; 95% CI: 0.83 to 2.72). Nevertheless, these associations were not significant, probably because of the small sample size employed for this subanalysis.

In multipollutant models, including O₃, NO₂ and PM_2.5, a similar pattern was observed. PM_2.5 levels were significantly related to coughing on the same day as exposure. Ozone levels were significantly associated with coughing and wheezing, following a 1-day lag and these effects increased according to the number of days of cumulative exposure (2 to 5 days). NO₂ levels were mostly related to coughing, with an increase of 11.3% (95% CI: 3.3% to 19.8%) where there was an increase of 24.5 ppb in 1-h maximum NO₂ level, over 4 days (Figure 2).

Among healthy children, we only observed a significant association between NO₂ levels and coughing (OR = 1. 22; 95% CI: 1.03 to 1.45 for an increase of 28.5 ppb in 1-hr maximum NO₂ levels over 2 days). No significant association was observed between PM_2.5 or O₃ levels and respiratory symptoms (data not shown).

Traffic density was significantly related to respiratory symptoms (coughing and/or wheezing) in asthmatic children (Incidence rate ratio (IRR) = 1.88; 95% CI: 1.14 to 3.09 for an IQR of 499 vehicles of SBPT) (data not shown). When considering the type of vehicle, wheezing, and bronchodilator use were positively associated with the 24-hr hour average traffic count. Table 4 presents changes in the risk of each respiratory symptom, related to the type of vehicle for an interquartile (IQR) increase. Both SBPT (type B) and heavier vehicles (type C) were associated with an increase in wheezing and bronchilator use. SBPT had the greatest effect on bronchodilator use (IRR = 1.99; 95% CI: 1.10 to 3.61 for an increase of 499 SBPT (IQR)).

When we considered distance to the main road as representing an index for exposure to traffic, the risk of wheezing decreased significantly with greater distance (IRR = 0.69; 95% CI: 0.49 to 0.98, with an increase in distance of 212 m (IQR)).