Traffic-Related Air Pollution and Childhood Respiratory Symptoms, Function and Allergies

Abstract

Background: 

Urban air pollution can trigger asthma symptoms in children, but there is conflicting evidence on effects of long-term exposure on lung function, onset of airway disease and allergic sensitization.

Methods: 

The spatial distribution of nitrogen oxides from traffic (traffic-NO_x) and inhalable particulate matter from traffic (traffic-PM₁₀) in the study area was assessed with emission databases and dispersion modeling. Estimated levels were used to assign first-year exposure levels for children in a prospective birth cohort (n = 4089), by linking to geocoded home addresses. Parents in 4 Swedish municipalities provided questionnaire data on symptoms and exposures when the children were 2 months and 1, 2, and 4-year-old. At 4 years, 73% of the children underwent clinical examination including peak expiratory flow and specific IgE measurements.

Results: 

Exposure to air pollution from traffic during the first year of life was associated with an excess risk of persistent wheezing (odds ratio [OR] for 44 μg/m³ [5th–95th percentile] difference in traffic-NO_x = 1.60; 95% confidence interval [CI] = 1.09–2.36). Similar results were found for sensitization (measured as specific IgE) to inhalant allergens, especially pollen (OR for traffic-NO_x = 1.67; 95% CI = 1.10–2.53), at the age of 4 years. Traffic-related air pollution exposure during the first year of life was also associated with lower lung function at 4 years of age. Results were similar using traffic-NO_x and traffic-PM₁₀ as indicators.

Conclusions: 

Exposure to moderate levels of locally emitted air pollution from traffic early in life appears to influence the development of airway disease and sensitization in preschool children.

Many epidemiologic studies have described adverse effects of urban air pollution on various aspects of respiratory health in children. Air pollutants from traffic are important triggers for asthma exacerbations,¹ but results regarding the role of air pollutants as causative agents in the development of disease are not consistent. Some cohort studies in children have indicated associations between traffic-related air pollution and doctor’s diagnosis of asthma,^2,3 while others find associations with prevalence of asthma symptoms.^4,5 On the other hand, some studies have failed to show a relation between air pollution and prevalence of asthmatic symptoms.^6,7

Several cross-sectional studies have observed negative correlations between air pollution and lung function in children at about 10 years of age.^8–11 Recent prospective studies also indicate that air pollution may impair lung function development at ages between 8 and 18 years, resulting in persistent damage.^12,13 However, no study on lung function and air pollution has been made in children below school age.

Some cross-sectional studies have shown positive associations between air pollution and elevated total IgE,¹⁴ sensitization to pollen,^15,16 and to outdoor allergens¹⁴ in children. These findings are supported by experimental data indicating effects of diesel exhaust on allergic sensitization in mice,^17,18 and the observation that pollen proteins may be chemically modified by air pollution.¹⁹ However, the evidence on children below school age or effects at comparatively low air pollution levels is limited.

The aim of this study was to assess the impact of exposure to source-specific air pollutants during the first year of life on wheezing, lung function and sensitization in children at the age of 4 years. We performed the study in a birth cohort exposed to moderate levels of air pollution, and with detailed modeling of exposure to air pollution from local traffic and house-heating.

METHODS

Study Subjects

Between 1994 and 1996, 4089 infants were recruited from Child Health Centers in a prospective study, described in detail elsewhere.²⁰ The study population comprised 75% of all eligible children born in predefined areas in 4 Swedish municipalities, representing urban and suburban environments. Data on parental allergic diseases, pet contact, detailed residential characteristics and socio-economic factors were collected with a postal questionnaire to the parents at recruitment (median child age = 2 months). When the children were approximately 1, 2, and 4-year-old, parents received similar questionnaires, with a main focus on the children’s symptoms related to wheezing and allergic diseases, and information on exposure factors. The response rates were 96%, 94%, and 91%, respectively. At approximately 4 years of age, 2965 children (73% of the full cohort) attended a clinical investigation in our department, including lung function test and blood sampling. The study was approved by the ethical committee of Karolinska Institutet.

Air Pollution Assessment

The assessment of exposure to locally emitted air pollution was based on a methodology developed to retrospectively estimate long-term source-specific exposure to air pollution in the study area, described in detail elsewhere.^21,22 It entails geocoding of an individual’s address information, and using an emission inventory together with dispersion models to map outdoor levels of selected pollutants from selected emission sources over time at the relevant geographical locations.

Residential address information was retrieved from the questionnaires at age 2 months and 1 year, and transformed into geographical coordinates (geocoded) using standard Geographical Information Systems computer software (MapInfo), in combination with an address database.²³ The geographical location was successfully assessed for 86% of the 4802 addresses, either automatically (77%) or interactively by an operator adjusting street spelling or street number within the software (9%). The database used in the computerized geocoding only covered densely populated areas within the studied county, and so addresses outside this area had to be geocoded manually (12%). Two percent of the addresses were outside the area for which air pollution exposure information was available and were thus excluded from the study.

Emission databases describing traffic-generated nitrogen oxides (traffic-NO_x) and particulate matter less than 10 μm (traffic-PM₁₀), as well as sulfur dioxide from house heating (heating-SO₂), within the county were available for the years 1990 (traffic-NO_x and heating-SO₂ only) and 2000 (traffic-NO_x, heating-SO₂, and traffic-PM₁₀). The geographical distribution of air pollution was assessed in 3 different resolutions, applied to regional/countryside (500 × 500 m), urban (100 × 100 m), and inner-city (25 × 25 m) areas. In addition, because the air pollution levels in the city also depend on very local traffic conditions, a street canyon contribution was added for addresses in the most polluted street segments in the city center (3% of the addresses).

The road network in the emission database for Stockholm County is described with 4600 individual links for which the length is defined by digitizing. To each link is assigned traffic flow data, signed speed, heavy traffic share, road type, and number of stops per kilometer. A total of 45 road types are defined depending on time variation of traffic over the day, the week, and the year. Vehicles are divided into passenger cars, trucks and buses, trucks with trailers and semi-trailers. Each of the 3 vehicle types are divided into 3 emission classes, depending on exhaust cleaning equipment on the vehicle. Emission factors for the 9 vehicle classes are given by the Swedish Road Authority.²⁴ Different emission values are defined for NO_x, CO, volatile organic compounds, CO₂ and particles representing link emissions, start and stop emissions, cold start emissions and evaporation emissions. The emission values are also differentiated in road types (highways, main roads, roads, residential streets, city center streets, etc.). The emission data for 1996 were validated by comparison with delivered amounts of gasoline and road transport diesel the same year. The corresponding consumption estimates from the emission database were 82% and 99% of these amounts, respectively. The lower proportion for gasoline is partly explained by unaccounted consumption of gasoline in pleasure boats, working tools, motorcycles and mopeds. A comparison of measured (average 1994–1997) and modeled (1996) total NO₂ levels in 16 locations showed good agreement: modeled level = 0.98 (standard error 0.07) × measured level + 1.07 (standard error 1.11) μg/m³ (r² = 0.93).²⁴ Domestic heating was modeled as an area source in areas without district heating. The source strength was modeled from estimated oil consumption and sulfur content. Since 1995 Swedish domestic heating oil by law contains less than 0.2% (2 g/kg) sulfur, and in the source calculations, 0.17% and 0.053% were used for 1990 and 2000, respectively.

The dispersion of the pollutants from the sources was estimated with a dilution model based on the average annual distribution of wind speed, direction and precipitation, implemented in the Airviro computer program (Swedish Meteorological and Hydrological Institute, Norrköping, Sweden). Residential outdoor levels of traffic-NO_x and heating-SO₂ for the children’s first year of life (1994–1997) were calculated by interpolation between 1990 and 2000, assuming a linear change in air pollution levels between these years in each point. For traffic-PM₁₀, the levels from the year 2000 were used for the whole study period. These trends were verified from graphs of annual means for the years concerned.

To validate the exposure assessments we compared the assessed levels of traffic-generated nitrogen dioxide (calculated with the empirical formula

²⁵; with measurements of ambient nitrogen dioxide made during 1995–1999 outside the homes of 487 of the children in the cohort.²⁶ For the same purpose, PM₁₀ and NO_x levels were estimated for 42 locations in which PM_2.5 measurements had been performed in a previous study.²⁷

Health Effect Assessment

The definition of wheezing was based on questionnaire information and is described in detail elsewhere.²⁸ Wheezing was subdivided on “transient,” “persistent,” and “late onset,” according to reported episodes of wheezing during the early (3 months to 2 years) and recent (last 12 months at 4 years) age periods. Transient wheezing was defined as ≥3 episodes of wheezing in the early period, but no episode in the recent period; persistent wheezing was defined as ≥1 episode early and ≥1 episode recently; and late-onset wheezing, as no episode of wheezing early, but ≥1 episode recently. “Wheezing up to age four” was defined as the sum of transient, late onset and persistent wheezing. Allergic rhinitis at 4 years was defined as sneezing, runny or blocked nose, or red itchy eyes after exposure to pollen or pets; a physician’s diagnosis of allergic rhinitis during the last 24 months; or both.

Peak expiratory flow measurements were performed using the normal range Ferraris Peak Flow Meters (Ferraris Medical Limited, Hertford, UK). Each child performed several peak expiratory flow assessments and the highest reading was used for analysis. A peak expiratory flow measurement was accepted if the operator had judged the child’s performance as adequate and if the best and second-best peak expiratory flow values were within 15%.²⁹ Of the 2965 who participated in the clinical testing, 17 children refused to do peak expiratory flow measurements and 349 children were not able to perform acceptable peak expiratory flow tests. Eighty-eight percent (n = 2599) of the children were able to perform acceptable lung function measurements that could be used for analyses.

Blood samples were drawn from 2614 children (88% of those who attended the clinical investigation). Serum IgE-antibodies were analyzed in 2543 (97%) of the blood samples, employing Pharmacia CAP System in a certified laboratory. Phadiatop containing inhalant allergens of cat, dog, horse, birch, timothy, mugwort, Dermatophagoides pteronyssinus and Cladosporium and fx5 containing food allergens of cow’s milk, egg white, soy bean, peanut, fish, and wheat were used. Samples that exceeded 0.35 kU_A/L were considered positive in Phadiatop or fx5 and were analyzed for specific IgE antibodies to the airborne and food allergens mentioned above.

Heredity for allergic disease was defined as parental (any parent) reports of allergy to furred pets or pollen, in combination with doctor’s diagnosis of asthma (medication required) or hay fever (no medication required).

Statistical Analyses

The associations between air pollution and wheezing or sensitization were analyzed using logistic and multinomial logistic regression, and the results are presented as odds ratios (ORs) and 95% confidence intervals (CIs). The association between air pollution and peak expiratory flow was analyzed with linear regression. Air pollution exposure was entered as a continuous variable and the results are presented for the 5th to 95th percentile difference within the cohort. All statistical analyses were performed with Stata Release 8.2 (StataCorp, College Station, TX).

Parental socioeconomic status was classified according to the Nordic standard occupational Swedish socioeconomic classifications.³⁰ The children were categorized on the basis of their parents’ highest socioeconomic status in 6 categories. Complete confounder sets were available for 3515 children (86% of cohort; Table 1) for the analysis of wheezing, for 2565 (63%) for the analysis of peak expiratory flow, and for 2543 (62%) for analysis of sensitization. The model included known risk factors for respiratory disease and variables that had influence on the effect of air pollution.

TABLE 1: Descriptive Data for the Air Pollution Study in the Birth Cohort (n = 3515)

RESULTS

Outdoor air pollution from traffic differed among the 4 study areas, with the highest levels in the city center and lowest in the municipality farthest away. The average contribution above regional background to the children’s residential outdoor air pollution levels was 4 μg/m³ PM₁₀ and 23 μg/m³ NO_x from traffic, and 3 μg/m³ SO₂ from home heating (Table 1). Traffic-PM₁₀ and traffic-NO_x were highly correlated (r = 0.94). Children from families of high socioeconomic status had the highest exposure levels, largely because they tended to live in the inner city where air pollution levels are higher (data not shown). Our 1-year outdoor residential estimates of NO_x from traffic were recalculated to NO₂ from traffic and compared with 1-month outdoor measurements of NO₂ for 487 of the children, showing a correlation of 0.74 (Fig. 1). Estimated NO_x for 2000 showed a similar correlation with seasonally adjusted PM_2.5 observations in 1999–2000 (r = 0.72), while the correlation with PM₁₀ was 0.61.

FIGURE 1.: The relation between measured ambient NO₂ levels and traffic-related NO₂ calculated from dispersion-modeled NO_x from traffic (annual mean) for 487 addresses in the study area. Fitted regression: calculated NO₂ from traffic = 0.64 × measured NO₂ − 0.77 (r = 0.74).

Wheezing up to the age of 4 (cumulative incidence) was reported in 22% of the children. For a 5th to 95th percentile difference in residential level, persistent wheezing was associated similarly with exposure to traffic-PM₁₀ and traffic-NO_x but not to heating-SO₂ (Table 2). Categorical analyses in quartiles did not reveal any clear dose-response pattern (details not shown). Girls tended to show somewhat stronger effects than boys. Transient and late-onset wheezing did not seem to be associated with exposure to any of the air pollutants. There was no relation between reported doctor’s diagnosis of asthma in combination with current symptoms and any of the air pollutants (data not shown). Furthermore, no association was seen to symptoms of allergic rhinoconjunctivitis at the age of 4 years.

TABLE 2: Association of Residential Levels of Air Pollution During the First Year of Life With Various Wheezing Outcomes up to Age 4 yr (n = 3515)

Lower peak expiratory flow at age 4 was significantly associated with exposure to traffic-PM₁₀ during the first year of life and a similar tendency was seen for traffic-NO_x (Table 3). For heating-SO₂ there was also a tendency of an effect, however weaker and less precise. No consistent difference was seen between girls and boys. The effect of air pollution on lung function did not differ between children with and without wheezing (data not shown).

TABLE 3: Associations of Residential Levels of Air Pollution During the First Year of Life With Peak Expiratory Flow (L/min) at age 4 yr (n = 2565)

A total of 614 children (23% of 2614) were sensitized to inhalant or food allergens at age 4, as indicated by increased serum IgE-levels. Eleven percent of the children were sensitized to any pollen (9% to birch pollen) and 8% were sensitized to furred pets. Sensitization to cow’s milk dominated among the food allergens and occurred in 8%. Exposure to air pollution from traffic or house heating during the first year of life showed a tendency to association with sensitization overall (inhalant or food allergens; Table 4). For air pollution from traffic, but not from home heating, this was driven by a strong effect for sensitization to pollen.

TABLE 4: Associations of Residential Levels of Air Pollution During the First Year of Life With Allergic Sensitization (IgE-Antibodies ≥0.35 kU_A/L) at age 4 yr (n = 2543)

Children with wheezing at 4 years (early persistent wheezing and late onset wheezing) were further subgrouped regarding sensitization to pollen at the age of 4, thus representing nonatopic and atopic wheeze. Exposure to traffic pollutants tended to be primarily associated with nonatopic wheeze (OR for traffic-NO_x = 1.46 [95% CI = 1.00–2.13]), rather than with atopic wheeze (OR for traffic-NO_x = 1.11 [0.55–2.22]) (P for interaction = 0.43).

Children with parents reporting an allergic disease tended to have a stronger association between air pollution and sensitization to pollen (OR = 2.02 [CI = 1.20–3.42]) for NO_x than if the parents did not report any allergic disease (1.36 [0.82–2.24]) (P for interaction = 0.17). No such tendencies were seen for the asthma/wheeze or peak expiratory flow outcomes.

Including paternal smoking did not change any of the estimates, nor did it increase the explanatory power of the models. Some of the children lived in homes with gas stoves (11%), but these were too few to allow for a separate analysis. Although many of the children’s homes were equipped with a fireplace or similar heating device (18%), these were mostly used intermittently—typically less than 4 times a month. Very few families used wood stoves as the primary heating source (n = 9).

DISCUSSION

In a Swedish birth cohort we found associations between residential outdoor levels of locally emitted air pollution from traffic during the first year of life, and 3 different indicators of airway disease in 4-year-old children: persistent wheezing, lower peak expiratory flow and sensitization to pollen. Previous cohort studies in children give some support for an association between traffic-related air pollution and airway disease.^2,3,5 These studies, however, either included children only up to the age of 2 years,^2,5 with unspecific wheezing often triggered by respiratory infections, or they used less spatially discriminating exposure assessment,³ making it difficult to draw conclusions about long-term effects. It has been proposed that the term “asthma” in children should be replaced by a wheezing definition that may also include a temporal dimension: transient, persistent, and late-onset wheezing.³¹ There was no association between self-reported doctor’s diagnosis of asthma, combined with current symptoms, and air pollution in this study. We found an association between traffic-related air pollution and persistent wheezing, which was not studied specifically in the other cohort studies. A stronger effect was suggested in girls than in boys, which lends some support to earlier evidence of girls being more sensitive to air pollution.^32–34

Negative correlations between air pollution and lung function in children^8,35 have been shown in some cross-sectional studies, although no effects on pulmonary function were found in the Six Cities study on 5000 12-year-old children.³⁶ A few prospective studies on school children have found that high levels of air pollution may impair the lung function development.^11–13 Our study shows that such negative effects may also be present in preschool children, and at lower levels of air pollution.

The single lung function parameter peak expiratory flow may not correlate well to forced expiratory flow in 1 second³⁷ or other measures used in diagnostic procedures or in staging of severity of asthma. However, in the present study peak expiratory flow is used not as a tool for diagnosis, but rather to compare the effect on airflow of the central airways between groups of children. Full-flow volume curves would probably be difficult to obtain with similar success rates in 4-year-old children.

It has been proposed that air pollution may increase the sensitivity of the airway epithelium to airborne allergens. Diesel exhaust particles may induce IgE responses directly by acting on B-cells and enhancing the production of cytokines that favor the development of an allergy-prone immune response.³⁸ Interactions between pollen and particles in air have also been suggested.³⁹ Some epidemiologic studies have shown an association between exposure to air pollution and sensitization in school children.^14–16 Our results indicate that the air pollution effect on sensitization to inhalant allergens, especially to pollen, is present by the age of 4 years, and that traffic-related air pollution plays an important role. In contrast to some other studies,^14–16 we found no association between air pollution and sensitization to pets.

Residential outdoor levels of air pollution from traffic were assessed using residential address histories and dispersion modeling from an emission inventory with high spatial detail. Despite the lack of full comparability, there was a good agreement when comparing calculated traffic-related NO₂ levels, based on our assessed annual traffic-NO_x levels, with measured 1-month outdoor levels of ambient NO₂. A smaller dataset of spatially distributed measurements is available from the same study area, including PM_2.5 measurements.^27,40 The results confirm a close relation between different pollutants originating from traffic. Analyses using traffic-PM₁₀ and traffic-NO_x showed similar results, and modeled values correlated reasonably well with measured PM_2.5. The model provides outdoor levels and most children spend a large part of their time indoors, where traffic-related air pollution levels are lower. It is, however, highly unlikely that this overestimation of residential exposure would have resulted in differential bias. According to measurements from roof-top monitors, the annual average urban background level of NO_x and NO₂ during the year 2000 was 25 and 20 μg/m³, respectively, in the city center.⁴¹ Using total pollutant levels by adding the long-range transported fraction and contributions from other sources in the region would not notably change the spatial contrast in the study.

Home heating with oil is represented as area sources distributed according to population density in areas that lack district heating. Thus, the spatial resolution of estimated residential levels of air pollution from home heating is necessarily lower than for pollutants from traffic, which may have contributed to the weaker effects seen for heating-SO₂.

By using exposure only during the first year of life, we minimize possible reverse causality induced by avoidance behavior due to the child’s disease. At this age, the air pollution levels at the home address are highly relevant as exposure estimates, since only 1% of the children in the study started day care before 12 months of age. Although about half of the children moved between birth and the age of 4 years, individual exposure estimates for the different life years were highly correlated, limiting the possibility of estimating effects related to specific time windows of exposure.

Although the observed effects in this study were related to PM₁₀ and NO_x from traffic, other components from vehicle exhaust may be responsible for the effects. For example, the geographical distribution of NO_x from traffic is very similar to that from other components of vehicle exhaust, eg, ultrafine particles.⁴² For PM we had no emission inventory earlier than year 2000. Traffic-related PM₁₀ in the study area is dominated by coarse particles, mainly from the road surface. Traffic volume was nearly constant in the inner city during the 1990s, while it increased outside the center. We thus believe our PM estimates to be unbiased for the inner city children but slightly overestimated for children who resided outside the center, but this should have very little effect on our results. Children from families of high socioeconomic status had the highest air pollution exposure levels, largely because they tend to live in the inner city where air pollution levels are higher. Higher exposure levels were also seen for children who lived in old houses, which are more common in the inner city than the suburban areas, and for children whose mothers smoked during pregnancy. No other important correlations between air pollution exposure levels and other exposures were observed.

Our results are based on a cohort study, and the information about exposure factors and confounding variables was collected before the children had any symptoms. This minimizes the risk for disease-related misclassification of exposure. The response rate was high for all 4 questionnaires; of the 4089 children initially recruited at birth, 3515 (86%) had complete data for multivariate analyses. The children lacking information on outcome variables or variables used in the multivariate analyses showed similar exposure levels to those analyzed. For example, the mean traffic-NO_x was 23.1 μg/m³ for children lacking peak expiratory flow measurements and 23.2 μg/m³ for those with measurements. Questionnaire answers are inherently subjective and could be related to such factors as socioeconomic status, but measurements of lung function and serum antibodies provide additional objective evidence. Further, the odds ratios are adjusted for a range of potential confounders.

In summary, this study provides evidence that exposure to moderate levels of air pollution from traffic during the first year of life may increase the risk of airway disease in preschool children. Further follow-up of our cohort will shed light on the persistence of these effects.

ACKNOWLEDGMENTS

We thank all children and parents in the BAMSE cohort, and the nurses and other staff working in the BAMSE project. We thank Marianne van Hage for analyses of IgE and Stina Gustavsson, André Lauber, Anna Boberg, Eva Hallner, and Anne-Charlotte Öhman Johansson for help with the geocoding. Dispersion modeling was performed by SLB-analys, Stockholm.