Urban air pollution has been associated with morbidity but little information exists on how it affects diurnal variation of lung function in children with asthma. The purpose of this study was to investigate the acute effects of traffic-related pollution on lung function among children with asthma.
We recorded morning and evening forced expiratory volume in 1 s (FEV1) for 28 consecutive days in 182 elementary schoolchildren with physician-diagnosed asthma, and monitored ambient hourly air pollution concentrations.
An interquartile range (IQR) increase (6.0 μg·m−3) in the previous 24-h (20:00 h to 20:00 h) mean concentration of fine particulate matter 2.5 μm in diameter (PM2.5) was associated with a 0.54% (95% confidence interval (CI) 0.06–1.02) decrease in bedtime FEV1 (p = 0.027). This association persisted in two-pollutant models with ozone, nitrogen dioxide and sulphur dioxide. An IQR increase in mean daytime (08:00 h to 20:00 h) PM2.5 of 6.5 μg·m−3 was associated with a 0.73% (95% CI 0.10–1.37) decrease in FEV1 over the course of the day expressed as 100×(FEV1 bedtime − FEV1 morning)/FEV1 morning (p = 0.024).
This study suggests that, in children with asthma, relatively low concentrations of urban air pollution worsen lung function over a short period of time, even within a day. Of the pollutants measured, PM2.5 appears to be the most important.
Acute and chronic exposure to urban air pollution in North America and Europe has been associated with increased respiratory symptoms, reduced lung function, increased number of hospitalisations and increased number of deaths from respiratory diseases 1–5. Although many different study designs have been used, the majority of the evidence comes from comparisons of daily concentrations of air pollutants with daily mortality or hospital admission counts. There have also been several panel studies which have reported associations between daily symptoms and/or lung function and daily measures of various air pollutants 6–14. The present study focuses on acute changes in lung function. We studied the effects of between-day changes in air pollution and tested a relatively unique hypothesis that acute change in lung function within a day, between morning and evening (diurnal change), was associated with fine particulate concentrations on the same day.
The study was carried out in Windsor, ON, Canada, with an estimated population of 216,473 in 2006 15. A unique aspect of the city is the Ambassador Bridge, which carries several thousand trucks daily between Windsor and Detroit, MI, USA. We identified children with asthma from a questionnaire survey carried out in the previous year of ∼16,000 elementary schoolchildren in Windsor. For the present study, we selected children whose parent or guardian had given a positive response to the question “Has a physician ever told you this child had asthma?”, and who had agreed to be contacted about future research. We selected children who were between 9 and 14 yrs of age when the initial questionnaires were completed, who spoke either of Canada’s official languages (English or French) and who lived in homes without cigarette smoke. A list of all eligible children satisfying these criteria was compiled. We telephoned their homes in random order and recruited the first 182 subjects who agreed to participate. The number of respiratory therapists available in Windsor to work on this study limited the number of subjects who could participate. The parent or guardian of each child gave written informed consent before the child participated in the study. The study protocol was approved by the Research Ethics Board of Health Canada.
A short-term longitudinal study design with repeated measures was used. Each child completed a daily symptom diary, and performed forced expiratory flows in the morning and at bedtime for 28 consecutive days. Owing to resource constraints, children were studied in one of two test periods: October 11 to November 7, 2005 or November 14 to December 11, 2005.
Air pollution measurements
City-wide air pollution hourly levels were estimated by averaging measures from two fixed-site monitors located in the western area of the city, upwind from the prevailing winds. These monitors, which are part of Environment Canada’s National Air Pollution Monitoring System, provided ambient levels of hourly fine particulate matter with a mass median aerodynamic diameter <2.5 μm in diameter (PM2.5), nitrogen dioxide (NO2), ozone (O3) and sulphur dioxide (SO2). The majority of the population lived within 10 km downwind.
Forced expiratory volume in 1 s
Respiratory therapists instructed parents and children on the use of hand-held flow meters and daily symptom diaries. Forced expiratory volume in 1 s (FEV1) was estimated using a PiKo-1 electronic peak expiratory flow/FEV1 meter (Ferraris Medical, Louiseville, CO, USA). Three flows were to be recorded twice each day, first thing on arising in the morning and again at bedtime, before taking breathing medications. For analyses, we used the maximum evening and morning FEV1 values expressed as a percentage of the predicted value based on the equations derived by Hankinson et al. 16. We also analysed the diurnal (within-day) variation in lung function, which was expressed as 100×(FEV1 bedtime − FEV1 morning)/FEV1 morning.
All statistical analyses were performed using the SAS software package version 9.1 (SAS Institute, Cary, NC, USA). Differences in the distribution of children’s characteristics between the two study periods were evaluated using the Chi-squared test. Daily mean concentrations of the ambient pollutants as well as temperature and relative humidity were calculated, and Pearson correlation coefficients were estimated to better understand their inter-relationships.
To estimate the mean pollution values 12 and 24 h before the FEV1 measurement, we assumed that, on average, morning FEV1 was performed at approximately 08:00 h and bedtime testing at 20:00 h. To evaluate temporal associations between ambient levels of air pollution and FEV1, we calculated the mean of the hourly air pollution measures for the following periods preceding the FEV1 measure: <12 h, 12 to <24 h and 0 to 24 h. A similar approach was used to create temporal metrics for temperature and relative humidity.
The results obtained from linear mixed models were expressed as the percentage change in median FEV1 corresponding to an increase in the interquartile range (IQR) in air pollutant levels. The use of the IQR serves to standardise each pollutant, facilitating comparisons between the magnitude of effect of different pollutants 17.
We used an autoregressive covariance structure (AR1), which allows for a greater within-subject autocorrelation for FEV1 measures taken more closely in time. We assumed a random intercept and a fixed slope. There were no statistically significant differences in the fit of the models using fixed slopes compared with random slopes.
The same approach was used to control for confounding in the logistic and linear mixed model. We adjusted symptoms and FEV1 for all variables that were associated at a level of significance of p≤0.15 with both ambient measures of air pollution and FEV1 (table 1⇓). This list of potential confounders included the daily mean temperature, relative humidity, day of the week, number of hours spent on outdoor activities (<2 or ≥2 h), sex and study period. Two pollutant models were used to characterise which pollutant was more strongly associated with FEV1 after adjusting for the effects of other correlated ambient measures.
Characteristics of the participants are shown in table 1⇑. During the first period of enrolment, a smaller percentage of participants were female (30.5%) compared with the second period (44.8%; p<0.05). Overall, 58.8% of children’s households were reported as having a household pet. To be included in the study, all subjects had a history of “ever had asthma”, and most (95.1%) reported that they still had asthma. Of the participants, 96.7% reported at least one respiratory symptom during the 28-day study period, but 42.3% did not report taking any asthma medication during this same time. Children in the first study group reported a mean of 2.2 h outside daily compared with 1.6 hours in the second group (p<0.0001).
For each child, we calculated the morning and bedtime FEV1 % predicted averaged over their 28 days of study (table 2⇓). The median values for both these measures were similar to the mean values (<2%), which were consistent with a near normal distribution. The diurnal change in FEV1 from morning to evening, expressed as 100×(FEV1 bedtime − FEV1 morning)/FEV1 morning was 2.3%, whereas the median (IQR) change was only 0.4 (0.4–7.6)%. We assessed the effect of time on lung function to determine whether there was evidence of a learning effect. The variables time (number of days each subject was studied), study period (October–November, November–December) and time×study period were not significantly associated with FEV1 (all p-values greater than 0.2).
Concentrations of air pollutants and IQRs were well below the US Environment Protection Agency ambient air quality guidelines for PM2.5 of 35 μg·m−3 for a 24-h average and 15 μg·m−3 for an annual average 18. In the present study, the 24-h mean was 7.8 μg·m−3 and the IQR was 6.0 μg·m−3 (table 3⇓). O3 decreased, SO2 increased and PM2.5 remained stable over the study period (fig. 1⇓).
SO2, NO2 and PM2.5 were positively correlated with each other, whereas ozone was negatively correlated with these three pollutants (table 4⇓). The strongest correlation between two pollutants was 0.68 (p<0.0001), which was observed with NO2, and PM2.5.
Adjusted for daily mean temperature and relative humidity, day of the week, duration of outdoor activity, sex and study period, an IQR increase (6.0 μg·m−3) in the previous 24-h (20:00 h to 20:00 h) mean concentration of PM2.5 was associated with a 0.54% (95% confidence interval (CI) 0.06–1.02) decrease in bedtime FEV1 % predicted (p = 0.03) (table 5⇓). No statistically significant associations were found between bedtime FEV1 and prior PM2.5 concentrations averaged over 24–48 h, 0–48 h or 0–72 h before lung function was measured. We found no statistically significant associations between evening FEV1 and NO2, SO2 and O3. In two-pollutant models with SO2, NO2 or O3, the PM2.5 effect remained significant at p<0.05 (fig. 2⇓).
We found no associations between morning FEV1 % predicted and pollutant concentrations with averaging times of 08:00 h on the day before to 08:00 h on the test day, 20:00 h to 08:00 h and 24:00 h to 08:00 h (overnight period). For the latter, adjusted mean interquartile increases in air pollution were associated with changes of 0.41 (95% CI -0.18–0.99) for SO2, 0.09 (95% CI -0.25–0.43) for NO2, -0.17 (95% CI -0.69–0.35) for O3 and 0.28 (95% CI -0.12–0.68) for PM2.5.
Diurnal change in FEV1
An interquartile increase in daytime PM2.5 (6.5 μg·m−3) averaged between 08:00 h and 20:00 h was associated with a 0.73% (95% CI 0.10–1.37) decrease in diurnal FEV1 change, expressed as 100×(FEV1 bedtime − FEV1 morning)/FEV1 morning (p = 0.024) (table 6⇓). No statistically significant association was found between diurnal FEV1 and PM2.5 concentrations averaged 24–48 h, 0–48 h or 0–72 h before lung function was measured. Diurnal declines in FEV1 during the daytime were also associated with increases in daytime NO2 (p = 0.024) and SO2 (0.036) (table 6⇓). In two-pollutant models with SO2, NO2 or O3, the PM2.5 effect remained statistically significant (p = 0.016) only with O3 (fig. 3⇓).
Effect modification of the FEV1 response to air pollutants
For the association between FEV1 and PM2.5, the first-order interaction term for corticosteroid use (on at least 50% of the study days versus less) was not significant, but power was poor with only 35 subjects reporting corticosteroids on at least half of the days. No significant differences in the FEV1–pollutant association were found for any of the pollutants and the following groups: male versus female subjects, those who spent at least 2 h outdoors daily versus those who spent less, and those whose baseline FEV1 was <85% versus ≥85%. For the sex comparison, the change in FEV1 % predicted for an interquartile change in 24-h mean PM2.5 was -0.54 (95% CI -1.13–0.04) for males and -0.54 (95% CI -1.37–0.28) for females. For baseline FEV1 level, the change in FEV1 % predicted was -0.88 (95% CI -1.53–-0.22) in the group with FEV1 ≥85% and -0.19 (95% CI -0.89–0.52) in those with lower FEV1.
Other response variables
The presence or absence of difficulty breathing, cough, wheeze and chest tightness was indicated on the daily diary. The odds ratio for reporting chest tightness was 1.30 (1.06–1.58) for days with a mean SO2 in the greatest quartile (≥8.8 ppb) versus the lowest quartile (<2.3 ppb). No other similar contrasts between air pollutants and symptoms were significant at p<0.05. We did not find any significant effects of air pollutants on the peak expiratory flow measures but all point estimates were negative. For an interquartile increase in 24-h mean PM2.5, percentage predicted peak flow decreased by -0.17 (95% CI -0.70–0.35).
Among children with a history of asthma, bedtime lung function decreased with increased ambient concentrations of PM2.5 even after adjustment for other pollutants. There was an approximate 0.5 percentage point decline in FEV1 % predicted for an interquartile increase in pollutant. FEV1 declines were associated with recent air pollution exposure, both the 24-h average before bedtime flows and the 12-h average during the day. Others have detected minimal or no effects within the past 24 h, although significant effects were detected from 5-day averaged particulate concentrations 9, 13. Although both PM2.5 and NO2 may represent mobile combustion-related air pollution, two-pollutant models indicated that PM2.5 had the most robust effect on lung function. The air pollution concentrations in Windsor were much lower than in many large US cities where particulate pollution has been associated with mortality 4. We observed decreases in lung function at 24-h PM2.5 concentrations approximately five times less than the US National Air Quality standard of 35 μg·m−3.
Significance of findings
We were able to detect an adverse biological effect on the lungs from air pollution at levels within currently accepted guidelines for public safety, suggesting that air pollution standards should be revisited. Although the FEV1 change during a clinically diagnosed asthma exacerbation is much larger than the 0.54% change observed in this study, the latter is of public health importance. All children in the community are exposed to the ambient air pollution, resulting in a large number of children affected. Assuming that there is a population distribution of asthma severity and of sensitivity to air pollution, children with severe disease and increased susceptibility to air pollution would be more likely to have a clinically important exacerbation of asthma on high-air pollution days. This argument is supported by the empiric evidence. Norris et al. 19 reported that a change in daily PM2.5 of 11 μg·m−3 was associated with a relative risk estimate of 1.15 (95% CI 1.08–1.23) for a visit to the emergency department for asthma.
Strengths and limitations of the present study
We did not have objective measurements of asthma. The use of medication would be more likely among those with more severe asthma, but reported use was not an effect modifier, suggesting that asthma severity is not a determinant of the response to air pollution. A meta-analysis of air pollution studies in children did not find children with asthma to be more susceptible than children without diagnosed asthma, which also suggests that asthma severity is not a consistent effect modifier 20. We do not know the reason for the stronger association between air pollution and bedtime flows compared with morning flows. Perhaps the children’s air pollution “dose” would be less in the 12 h before the morning FEV1 measurement because they would be indoors and exposed to lower concentrations of ambient pollution. Also, minute ventilation would be less while sleeping than during wakefulness. We did not measure indoor air quality, but only children living in smoke-free homes were included. Associations between FEV1 and day-to-day changes in ambient air pollution would not be confounded by indoor air allergens or irritants unless they changed on a day-to-day basis in concert with outdoor air pollution. Janssen et al. 21 studied the association between personal and outdoor concentrations of PM10 over time among 37 nonsmokers. The median Pearson correlation coefficient was 0.71 when not exposed to passive smoke, indicating that changes in outdoor levels reflect changes indoors. The forced vital capacity manoeuvre required for the FEV1 measurement was observed only during the training period. Increased variability of the test could have reduced the power to detect a true association with air pollution. However, we did detect an adverse relation with particles. The Pearson correlation coefficient between morning and evening values was r = 0.82 (p<0.0001), indicating a moderate degree of reproducibility.
Comparison of the present results with previous panel studies of children with asthma
Peters et al. 13 studied a panel of 82 Czech children with asthma between November 1991 and February 1992. An increase of 6.5 mg·m−3 in 5-day mean sulphate was associated with a 5.62 L·min−1 (95% CI -9.93– -1.30) decrease in peak flow. Another report from the Czech panel by Peters et al. 13 showed significant associations between air pollution and peak flow during September 1991 and March 1992. Mortimer et al. 10 also detected a negative association between ambient O3 and morning peak flow among 846 children with a history of asthma recruited from eight US urban centres. Peak flow decreased by 0.59% (95% CI 0.13–1.05) for each 15 ppb increase in the 5-day moving average for O3. In one panel study of 22 Hispanic children with asthma by Delfino et al. 8, peak flows were not found to be associated with ambient ozone, NO2 or SO2. In another panel of 19 children with asthma by the same lead investigator, a 0.7% (95% CI -1.9–0.4) decrease in FEV1 was associated with a 7.5 μg·m−3 increase in PM2.5 averaged over the preceding 24 h at a central site 9. Using a passive nephelometer for personal monitoring, the observed effect size was statistically significant with a -5.9% (95% CI -10.8–-1.0) change in FEV1 for a 30 μg·m−3 change. Multiday moving averages for PM2.5 resulted in larger effect sizes but these were not significantly different from those for the previous 24-h average. In our study, earlier changes were detected with significant effects seen between morning and evening lung function associated with same-day PM2.5. The effect size per μg·m−3 of PM2.5 in our study was 0.08%, in close agreement with the 0.09% estimated in the study by Delfino et al. 9 despite the mean PM2.5 during the studies being 7.8 μg·m−3 and 10.3 μg·m−3, respectively. Whether exposed to higher or lower daily average concentrations of PM2.5, it appears that equivalent changes in PM2.5 have a similar magnitude of effect on lung function. Trenga et al. 14 reported an adverse effect of centrally monitored PM2.5 on the FEV1 of children with asthma, only if they were not receiving anti-inflammatory medications. We did not detect a significant difference between those with and those without asthma medications. A study similar to ours took place in Detroit (MI, USA), which is connected to Windsor by the international Ambassador Bridge 22. During six periods, each of 14 days, 86 children performed FEV1 measurements using a portable device and recorded daily PM2.5, PM10 and O3. For the entire study group, no significant associations were reported between FEV1 and pollutants. Subgroup analyses restricted to those reporting maintenance inhaled corticosteroids, and using three different lags, revealed that four out of 24 associations tested were statistically significant between single pollutants and diurnal FEV1 variation and/or lowest daily FEV1. For the subgroup reporting upper respiratory symptoms, eight out of 24 associations with single pollutants were positive at p≤0.05. The air pollutant association with diurnal variation was positive, whereas it was inverse in our study. Whether or not an increase or decrease in diurnal variation is considered an adverse response depends on the reason. In our study, the morning flows did not change significantly with air pollution. Thus, the decreased diurnal variation was due to a smaller than expected FEV1 improvement that normally occurs during the day and is unrelated to air pollution. A more recent study also compared air pollution levels with self-administered spirometry in a panel of children 23, in which 53 subjects with asthma were studied for 10 days. Peak hourly personal PM2.5 (averaging 90 μg·m−3), but not ambient levels, were associated with decreased FEV1. Maximum morning FEV1 decreases were seen with approximate 8-h lags, and maximum afternoon and evening decreases were associated with 24-h lags. No air pollution effects were seen in a group of 16 children taking a β-agonist, theophylline, and/or an anti-cholinergic medication. In our study, with much lower concentrations of ambient PM2.5 but 10 times the number of person-days of observation, we found maximum FEV1 effects within a lag time of 24 h after exposure and also effects on diurnal variation within a 12-h period.
In summary, this study provides evidence that acute changes in urban air pollution, even within a day, at concentrations less than in many North American urban centres, worsens lung function in children with asthma.
Statement of interest
The authors gratefully acknowledge the Centre for Environmental Health of Ontario, the City of Windsor school boards, the Windsor community parents and children, the Windsor Medical Officer of Health, the Essex County Health Unit, the University of Windsor, the Ministry of Environment (all Windsor, ON, Canada), Environment Canada (Toronto, ON, Canada), the United States Environmental Protection Agency (Washington, DC, USA) and the International Joint Commission (Canada and USA).
- Received September 9, 2008.
- Accepted February 5, 2009.
- © ERS Journals Ltd