Recurrent wheeze represents a serious health burden in young children and is the cause of a considerable number of hospital admissions and other health care expenditure.1 Evidence for the role of air pollution in causing such symptoms remains equivocal.2 3 Particulate air pollution in terms of particulate matter <10 and 2.5 μm in diameter (PM₁₀ and PM_2.5) has been linked to asthma exacerbations in school children,4^–12 suggesting a role as a trigger of symptoms, although two large multicentre studies failed to confirm this.13 14 Recently, ultrafine particles (UFPs; particles <0.1 μm in diameter) generated by traffic emissions were suggested to have particularly strong effects in the airways due to a high level of pulmonary deposition and their ability to induce inflammation and oxidative stress.15 However, the only epidemiological study of the effects of UFPs in children to date found only a weak association with asthma symptoms in school children.8 Traffic-related air pollution, assessed by nitrogen dioxide (NO₂) exposure or proximity to traffic, has been more consistently associated with asthma exacerbations in school children, which suggests that air pollution is an important trigger of symptoms.16^–20

Only a few studies of the effects of air pollution are available in young children and provide mixed evidence.21^–27 A greater susceptibility of younger children compared with school children may be suspected because of the development of the lung and the higher ventilation rates in early life.2 A case-control study has found significant associations between traffic pollution and asthma before the age of 3 years which attenuated between the ages of 3 and 14 years.20 Understanding the effects of early exposure to air pollution on the development and triggering of wheeze may give important clues to its role in the development of asthma later in life.3 To date, most studies in young children have used geographical variations in air pollutant levels and have focused on traffic.20 21 23^–26 Only one studied PM₁₀22 and one studied ozone (O₃),27 with no evidence to date on the effect of UFPs. One study in preschool children found an association between soot and incident wheeze, asthma and ear/nose/throat infections in a birth cohort of children up to 4 years of age,21 and a second study in preschool children found an association between PM₁₀ and the prevalence and incidence of respiratory symptoms in a survey of children aged 1–5 years.22 While proximity to traffic was recently reported to be associated with wheeze in infants,23 only two birth cohort studies have examined susceptibility in small children by age, reporting associations between doctor-diagnosed asthma24 and respiratory symptoms (cough)25 and traffic to be stronger in the first year of life than in the second year of life. Finally, the potential importance of genetic susceptibility has been suggested by a particularly strong association between triggering of wheezing bronchitis or wheeze and PM_2.5 and O₃, respectively, among infants with a family history of asthma.26 27

In this study we used the unique data on daily symptom recording in the birth cohort of children of mothers with asthma from the Copenhagen Prospective Study on Asthma in Childhood (COPSAC) to determine the short-term impact of particulate (PM₁₀ and UFPs) and gaseous (nitrogen oxide (NO_x), NO₂ and carbon monoxide (CO)) air pollution on the triggering of wheezing symptoms during the first 3 years of life in this high-risk population. We examined the susceptibility by age (infants (0–1 year), 1–2 years and 2–3 years) to study the hypothesis of increased susceptibility of infants to air pollution, and tested for effect modification by medication use, gender and paternal asthma history.

METHODS

Study population

The COPSAC prospective longitudinal birth cohort study includes 411 Danish children born to mothers with asthma and was designed to study the gene-environment interaction in the development of asthma and other atopic diseases.28^–30 In brief, pregnant women identified through the Danish National Birth Cohort Study and prenatal clinics, living in Copenhagen, fluent in Danish, with a history of physician-diagnosed asthma after the age of 7 years and a history of daily treatment with inhaled β₂ agonists or glucocorticoids (minimum of 2 weeks during two seasons or continuously for 1 year) were invited to participate in the study. Infants with a severe congenital abnormally, a gestational age <36 weeks, a need for mechanical ventilation or a lower respiratory tract infection were excluded. The eligible 411 children of 394 mothers (nine pair of twins and eight siblings) were born between 2 August 1998 and 12 December 2001 and enrolled in the cohort at the Clinical Research Unit (CRU) visit at 1 month of age, with follow-up clinical investigations every 6 months until the age of 7 years and at any episode of symptoms.

At the first CRU visit parents were given diary cards (http://ipaper.dk/copsac/Asthma_in_young) and instructed to record their child’s wheezing symptoms daily as dichotomised scores (yes/no), as previously described.29 The CRU doctor reviewed the symptom definitions and the diary entries with parents at the 6-monthly clinical sessions and at acute episodes of wheeze. Diary recordings for the first 3 years of life were used in the current analyses. Information on exposure to smoking at home and paternal history of asthma was obtained by interviews. Medication use followed the algorithm described in detail elsewhere.29

Information on addresses, including relocations, was available from the COPSAC database and the Danish National Person Registry, allowing for geocoding and calculation of the exact distance from the residence to the air pollution monitor in the centre of Copenhagen on a daily basis. Thus, 205 children living within a 15 km radius of the central monitor during the first 3 years of life were selected for the study.

Assessment of air pollution exposure

The pollutant and meteorological data were measured by the Danish National Environmental Research Institute at a fixed urban background monitor (20 m height) in the centre of Copenhagen with minimal contribution from local sources, in accordance with WHO guidelines. Main streets are located about 300 m west and 50 m east of the monitor, with 26 000 and 56 000 vehicles passing per working day, respectively. For the study period (12 December 1998 to 19 December 2004), daily (24 h, midnight-to-midnight) mean concentrations were available for PM₁₀ (SM200 monitor; Opsis, Sweden), CO (M300; API, San Diego, USA), NO₂ and NO_x (M200A; API), O₃ (M400; API) and temperature, with missing data on days with equipment malfunctions. Although the total number of concentration measurements included particles 10–700 nm in diameter, we define them as UFP_NC in this study because particles <100 nm (by definition, UFP) comprised more than 95% of the total concentration measurements. Monitoring of UFP_NC (Differential Mobility Particle Sizer, Roskilde, Denmark)31 and PM_2.5 (TEOM Ambient Particulate Monitor; Rupprecht and Patashnick, USA) began in Denmark on 15 May 2001 and 3 October 2003, respectively.

Statistical analysis

The incidence of wheezing symptoms was defined as the first day of a registered symptom. Logistic regression analyses were performed using generalised estimating equations (GEE)32 by the GENMOD procedure with exchangeable correlation structure using SAS Version 9.1. Pollutant concentrations were log-transformed. Analyses were conducted for all 3 years and separately for infants and age groups 1–2 and 2–3 years. We first fitted single-pollutant models, adjusted for age (dummy for each year), gender, exposure to passive smoking, paternal history of asthma, 24 h mean temperature (linearly) and calendar season (dummy). Pollutant concentrations on the same day (lag 0), previous day (lag 1), up to 4 days before a new symptom (lag 4) and the 3-day mean (2–4 days) were considered. In cases where monitor data were missing from single days, 3-day means of available measurements were used. Two-pollutant models were fitted for 3-day mean pollutant concentrations to examine the robustness of one-pollutant associations. Separate analyses were conducted for a subset of data where UFP_NC data were available (15 May 2001 to 19 December 2004) to ensure comparability between the effect of UFP_NC and other pollutants. Finally, we tested for effect modification by gender, medication use and paternal history of asthma. Effects were reported as odds ratios (ORs) per interquartile range (IQR) increase in exposure.

A number of sensitivity analyses were performed. The results remained unchanged when fitting a GEE model with alternative correlation structures (independent, autoregressive). Various adjustments for meteorology and seasonal/time trend effects were tested. A model with temperature defined by a dummy per 25th percentile showed similar results and indicated a strong negative trend on symptom development by increasing quartiles of temperature, validating linear modelling of temperature. Season was modelled by dummy variables for calendar seasons, with a significant trend seen for an increase of symptoms in colder seasons. After adjusting for season and temperature, additional time trends including calendar time (linearly and by smoothing spline) and dummy for each month were not significant and not included in the model. Finally, we performed analyses for an alternative choice of study population with 110 COPSAC children living within a 5 km radius from the central monitor.

RESULTS

The 205 children (99 boys) in the COPSAC cohort lived within a 15 km radius of the central monitor, with a mean (SD) distance of 6.1 (4.0) km and a minimum distance of 0.3 km. Of the 205 children, 35 had a paternal history of asthma and 94, 64 and 47 were exposed to passive smoking for <10, 10–100 and >100 days/year, respectively. The 205 children seemed representative of the whole COPSAC cohort,29 with a dropout rate of 3 (1.5%), 6 (3%) and 2 (1%) children in the first, second and third year of life, respectively, and 194 (95%) active children at their third birthday. The average observation period per child was 850 days (minimum 23, maximum 1097), with a total of 174 259 person-days and 15.4% diary entries missing. Thirty-three children (16%) experienced no wheezing symptoms in the first 3 years of life. A prevalence of 6.2 per 100 person-days (total of 10 779 symptom-days) and an incidence of 1 per 100 person-days (1591 new symptom-days) was observed.

Tables 1 and 2 show the pollution and meteorological conditions in Copenhagen during the study period. Frequent breakdown or other utilisation of the UFP_NC measuring equipment caused missing data gaps (54%). PM_2.5 and O₃ data were not sufficient for analyses of all three age groups due to the late start of PM_2.5 monitoring (3 October 2003) and a monitoring gap for O₃ in the middle of the study period (13 July 1999 to 1 July 2003). Strong correlations were observed between PM₁₀ and PM_2.5, CO and NO₂/NO_x, and UFP_NC and NO₂/NO_x (table 2) and weak correlations were observed between PM₁₀ and UFP_NC and PM₁₀ and NO₂/NO_x.

Air pollution showed delayed adverse effects, which were strongest and most significant in infants with a 4-day delay for PM₁₀ and a 3-day delay for NO₂, NO_x and CO. In this group an increase of one IQR in exposure to PM₁₀, NO₂, NO_x and CO resulted in an increase in wheezing symptoms of 23% (95% confidence interval (CI) 2% to 48%; lag 4; IQR 14 μg/m³), 42% (95% CI 15% to 77%; lag 3; IQR 6.5 ppb), 30% (95% CI 9% to 53%; lag 3; IQR 8.9 ppb) and 47% (95% CI 10% to 96%; lag 3; 0.12 ppm), respectively (table 3). Associations for NO₂ and NO_x were also significant with a 4-day lag and 3-day means. Similar lag structures were observed for the pollutants in all 3 years, but the effects were attenuated after 1 year of age. The associations remained positive and significant in all 3 years only for NO₂ and NO_x, for which an increase of one IQR in 3-day mean exposure was associated with an increase in wheezing symptoms of 19% (95% CI 1% to 30%; 6.5 ppb) and 14% (95% CI 0% to 30%; 8.9 ppb), respectively. The effects of NO₂ and NO_x were stronger than those of PM₁₀ throughout the study period (table 3), as confirmed in two-pollutant models for infants (table 4).

UFP_NC showed a relatively strong adverse effect with delays of 2–4 days in infants, but without reaching statistical significance, and this changed after the age of 1 year to an apparently protective effect. Analyses for 3-day mean concentrations of other pollutants (PM₁₀, NO₂, NO_x and CO) were repeated for the subset of data where UFP_NC measurements were available (15 May 2001 to 19 December 2004) for infants to enable direct comparisons (table 5). Associations for all pollutants were enhanced, with estimates for UFP_NC comparable to those of PM₁₀ and gases in one-pollutant models and higher than the others in two-pollutant models. For COPSAC children living within a 5 km radius of the monitor, the effects of UFP_NC were enhanced reaching statistical significance for 3-day means (table 6) and for lags 2–4 (data not shown).

No effect modification was detected between air pollution and gender, paternal asthma history or medication use.

DISCUSSION

We found strong adverse effects of air pollution with PM₁₀, NO₂, NO_x and CO on the triggering of wheezing symptoms in infants which attenuated after the age of 1 year. Independent effects of PM₁₀ and gases were observed, with consistently stronger estimates for gases (NO₂ and NO_x), proxies of traffic-related pollution. Furthermore, adverse effects of UFP_NC in infants, which were enhanced in infants living within a 5 km radius of the monitor, support the relevance of traffic, although an apparently protective effect of UFP_NC was observed at the age of 3 years.

The finding that traffic triggers wheezing in infants is in agreement with recent evidence that infant wheezing is associated with proximity to traffic23 and infant wheezing bronchitis is associated with PM_2.5.26 Furthermore, two studies found an association between doctor-diagnosed asthma and respiratory symptoms (cough) in children aged <2 years and modelled levels of PM_2.5 and NO₂ at the children’s residence.24 25 These associations were strongest in the first years of life, as observed in our study. Lung anatomy and physiology, ventilation rates and organ maturity change with age, which may explain the greater vulnerability of infants to air pollution and suggest biological plausibility for these findings.2 Consistent with our findings, a study of asthma symptoms in preschool children21 and several studies of asthma exacerbations in school children have also shown adverse effects of various proxies of traffic.16^–19

Our study is the first to study the effects of UFPs on triggering of wheezing symptoms in young children. A positive association was found in the first year of life, which was significant in children living within 5 km of the monitor. The only other study of UFPs in children found a significant effect of PM₁₀ and a comparable but not significant effect of UFPs on peak expiratory flow in 39 asthmatic school children over 57 days.8 Three studies of the effects of pollution on asthma exacerbations in adults found that UFPs and PM_2.5 had a greater effect than larger particles33^–35 in a radius of 2 km33 and 5 km34 from the pollution monitor. However, our results should be taken with caution and need to be confirmed because of the unexpected protective effect of UFPs found in children over 1 year of age.

To date, the effects of PM₁₀ have been studied in only one survey of preschool children22 and there have been no studies in infants. In Copenhagen the main source of PM₁₀ is long-range transport from secondary biomass and oil combustion sources with a smaller contribution from traffic,36 which suggests that PM₁₀ is a poor proxy for traffic and enables its effects to be separated from and compared with the effects of traffic proxies (NO₂, NO_x, and UFP_NC). We found that the adverse effects of PM₁₀ were significant only in infants (lag 3) and were generally smaller than those of gases (table 4) and UFP_NC (table 5). This suggests an independent effect of particle mass and gases or particle numbers mainly related to traffic.

The increased susceptibility to air pollution in genetically susceptible children was previously suggested by the finding of an increase in the association of incident wheezing bronchitis with PM_2.5 in children with a family history of asthma26 and the stronger effects of O₃ in triggering wheeze in children of mothers with asthma.27 In our cohort of high-risk infants, associations with traffic-related air pollution were higher than those reported elsewhere in healthy infants,22 24 25 but direct comparisons are difficult because of differences in outcome definitions and the use of geographical variation in exposure assessments in those studies. Furthermore, the association of asthma prevalence with traffic was surprisingly stronger for school children without a family history of asthma.17 This may indicate that early symptoms in infants and later diagnosed asthma—although strongly related—represent children with different susceptibilities.

The incidence of wheezing symptoms was chosen as the health outcome in this study rather than diagnoses of asthma because of the inaccurate diagnostic criteria for asthma in early life.29 30 As the term “wheeze” carries little specific meaning in lay terms and even between specialists,37 we explained the term to the parents in the generic sense as lung symptoms severely affecting the child’s well-being including wheeze or whistling sounds, breathlessness, shortness of breath and persistent troublesome cough. This approach was supported by diary reviews at the 6-monthly visits together with a dedicated book describing the relevant lung symptoms. The diagnosis and day-to-day management of respiratory conditions was conducted solely by the doctors employed for this purpose at the CRU in accordance with predefined algorithms, minimising the risk of symptom misclassification from influence of the prevailing and variable diagnostic criteria and treatment traditions in the medical community. The inaccuracy of symptom recognition and reporting bias was reduced by the fact that all mothers had a personal asthma history and familiarity with the disease. Furthermore, the state-of-the art information on the incidence of wheezing symptoms in the COPSAC cohort provided an opportunity to study the short-term effects of air pollution on triggering symptoms which has previously been reported in only two studies in infants.26 27 Analyses of the cohort at later ages when the diagnoses of health outcomes are more specific will help determine the long-term impact of early exposure to air pollution on the development of specific respiratory diseases, including asthma.

The strengths of our study include the prospective daily recording of symptoms which allowed the study of symptom incidence, and the long study period of 5 years which gave sufficient contrast in exposure over time and power to detect adverse effects. This is an advantage over earlier studies with a short follow-up of a few months,4 6^–10 27 which may limit the power to detect pollutant effects.38 An additional strength is the well-defined birth cohort of children genetically predisposed to asthma, giving a unique opportunity to study the effect of early exposure to air pollution in a high-risk group. Furthermore, continuously updating the residential information allowed inclusion criteria to be based on address (vicinity to the monitor) at the time of symptoms and not at the time of birth, as has previously been the case.21 24 Finally, the availability of both gaseous and particulate pollutants of different sizes allowed comparisons of the effects of traffic-related (gases and UFPs) and non-traffic-related (PM₁₀) air pollution in the development of respiratory disease in young children.

The limitations of the study include assessment of exposure from a single monitor which might result in exposure misclassification.39 A 15 km radius from the central monitor was chosen as it represents the municipality limits of Copenhagen City with similar population and traffic density, and was assumed to be representative of air pollution levels measured at the central monitor. We have previously shown that the UFP_NC levels at this site correlated well with levels at a monitoring site at the kerbside (2 m height) of a busy street 3000 m away (Spearman correlation coefficient R_S = 0.62) and a rural monitoring site located in a residential area 30 km southwest (R_S = 0.80).40 This was despite the fact that UFP_NC levels were on average 4.6 times higher at the kerbside and 0.7 times lower at the rural site than at the urban background site. This indicates that the daily oscillations in traffic-related air pollutants on busy streets due to variation in traffic intensity, weather conditions and other factors are also reflected by urban background monitoring, and that a radius of 15 km from the background monitor is reasonable for the assessment of daily variations in population exposure. In children living closer to the central monitor (5 km) there was a slightly stronger association between UFP_NC and triggering of wheeze which was statistically significant and exceeded that of other pollutants (table 6). However, if the apparent difference between estimates for a radius of 5 km and 15 km is real, we do not know whether this is due to a greater misclassification of exposure with the larger radius or whether the effects of traffic closer to the monitor—where traffic density and emissions and traffic pollutant levels are higher—really are greater. A further limitation of the study is the counterintuitive and biologically implausible change in the effect of UFP_NC from an adverse effect in infants to a protective effect in children over the age of 2 years which was statistically significant in children living within 5 km of the air pollution monitor (table 6). However, the effect of all other pollutants (PM₁₀, NO₂, NO_x and CO) also attenuated after the age of 2 years to become apparently protective, although this did not reach statistical significance (table 6). This indicates that a decrease in susceptibility to air pollution after the age of 2 years is not confined to UFPs and is therefore probably not the result of measurement error or change in exposure but, rather, is due to the underlying disease process and/or the treatment of wheezing symptoms in this group of children. As the medication protocol in the COPSAC cohort presumes a history of recurring wheezing episodes, children typically do not receive medication until the second year of life or later. The attenuation of the adverse effects of air pollution after the age of 1 year and further after the age of 2 years is therefore most likely explained by the decreased susceptibility of the children to air pollution as a result of medication for their wheezing symptoms. Likewise, the statistical significance with respect to UFP_NC is most likely explained by a type I error. Nevertheless, it may question the findings for UFP_NC in the first year of life as no other supportive data are available. Further limitations of the study include the large number (54%) of missing data for UFP_NC. However, missing data gaps were examined and were found to be random and not associated with pollution levels or weather conditions. Nevertheless, our UFP_NC measurements for 602 days are the longest available to date for the study of the effects of UFP on respiratory symptoms.8 33^–35

In summary, we found independent effects of PM₁₀ and traffic-related pollution measured by exposure to NO₂/NO_x, CO and UFP_NC (5 km radius) in triggering wheezing symptoms in young children genetically predisposed to asthma. Infants were found to be particularly vulnerable. Follow-up of the cohort at later ages will determine the effect of these associations on the development of asthma.