The study was undertaken in winter 2004 in Christchurch, New Zealand, a city of ~330,000 situated on the east coast of the South Island. Significant air pollution in Christchurch almost invariably occurs in winter, and is characterized by PM₁₀, ~90% of which is PM_2.5, that is largely generated by domestic wood-burning open fires and enclosed wood-burners 7. Mean daily PM₁₀ in central and suburban Christchurch exceeds 50 μg/m³ on ~30 days per year and can reach up to 250 μg/m³ 8. The health effects of Christchurch air pollution have been the subject of a number of local studies 91011.

The study was undertaken in the boarding houses at Christ's College, a single-sex private school situated in the central city, but inset into the grounds of Hagley Park, a five km² area of grassland and trees to the immediate west of the CBD but surrounded by the urban area on all sides.

Two pollution monitoring sites were established for this project. The indoor site was a second floor corridor leading to rooms occupied by boarding students. The building is naturally ventilated and is heated by a thermostatically controlled central heating system. The site was not near any obvious indoor sources of particulates such as kitchens which are located in a different building. The outdoor site was a courtyard within the school grounds. Indoor and outdoor particulate pollution levels (PM₁₀, PM_2.5, PM₁) were measured continuously using a TEOM Series 1400a Ambient Particulate Monitor, and expressed as 10 minute averages over a 24 hour period, and 24 hour averages, from midday to midday. It has been shown that this TEOM under-reads the true PM₁₀ mass because of the loss of semi-volatile compounds in the heated air stream. Consequently the figures were corrected based on an established factor. Details about this and the more information about the indoor and outdoor pollution monitoring at Christ's College are given in Kingham and co-workers 5. Weather data including wind speed, temperature, temperature difference and relative humidity were also continuously collected from the outdoor site and expressed as 10 minute averages.

The participants were male secondary school students, aged between 12 and 18 years, attending the boarding houses of Christ's College. All were healthy non-smokers with no significant medical conditions. Smokers were specifically requested not to volunteer for the study, and the non-smoking status of the participants was confirmed by school staff. A sub-group of students with a previous diagnosis of asthma (defined as "doctor diagnosed asthma ever") were also studied. They were being treated with as-required short-acting beta agonist reliever medication, with some taking regular inhaled corticosteroids (see Results section for details). The regular use of oral corticosteroids was an exclusion criterion. Investigators made no alterations to asthma management and gave no advice about reliever inhaler use during the study.

Prior to commencing the study the health status of the students was assessed by taking a brief medical history. Baseline spirometry and skin prick testing for atopic status were also undertaken.

The study protocol and conduct was approved by the Canterbury Regional Ethics committee. All participants provided written informed consent prior to commencement, and parental approval to participate was also sought.

All students were issued with personal downloadable electronic spirometers (Piko-1, Ferraris Respiratory Europe Ltd, Hertford, UK) and trained in their use. These spirometers measure Forced Expiratory Volume in 1 second (FEV₁) and peak expiratory flow rate (PEFR) to American Thoracic Society standards 12. Lung function measurement was undertaken twice daily, in the early morning, and around 6.00 pm, supervised by House Matrons, who were all trained nurses given training in spirometry. No specific instructions about withholding reliever medications prior to spirometry were given to the students. Spirometry data were downloaded at least fortnightly by investigators.

All students were issued with daily diary cards to detail respiratory, nasal and eye symptoms. In addition, students with asthma completed diaries detailing asthma symptoms and reliever medication use.

Lung function and diary card data collection occurred during school terms for the duration of the project. When a student was away from the school for more than one night (weekend and school holidays), diaries and lung function measurements were not undertaken.

Within 24 hours of two separate high pollution days all participating students underwent a more detailed assessment. The investigators were informed by Environment Canterbury the morning after a high pollution night. The school was contacted, and all participating boys were excused their evening study activities, to be assessed by the investigating team. This involved collection of urine samples for analysis of 1-hydroxypyrene and collection of exhaled breath condensate for measurement of pH and hydrogen peroxide. Baseline assessments were also undertaken at the beginning of the study during a period of low pollution (autumn control) and during the winter, after a 1 week period of low pollution levels (winter control).

Urine samples were stored at -20°C until analysis. Urinary 1-OHP was quantified using reverse phase high-performance liquid chromatography (HPLC) using a modification of a previously described method 13 as described in Cavanagh and co-workers 6. The lower limit of detection was 0.23 nmol/L. Hydroxypyrene concentrations were corrected for urinary creatinine to account for hydration status and expressed as μmol 1-OHP/mol creatinine.

Exhaled breath condensate (EBC) was collected using the RTube™ system (Respiratory Research Inc, Charlottesville, USA), using an aluminium cooling tube. The aluminium cooling tube was stored at -20°C for at least 24 hours prior to collection. EBC was collected for ten minutes on each sampling occasion, after rinsing the mouth to reduce saliva contamination. Approximately 2 ml of condensate was collected from each student at each collection.

EBC pH was immediately measured by applying approximately 30 μl of EBC to a pre-calibrated Shindengen ISFET pH meter (model KS701). EBC was not de-gassed prior to pH measurement.

H₂O₂ was measured in EBC samples using fluorometry with 4-hydroxyphenylacetic acid 14. Briefly, following the H₂O₂-dependent conversion of 4-HPAA to the dimer 2,2'-dihydroxybiphenyl-5,5'-diacetate, the fluorescence of each sample was quantified within four hours using a Hitachi F4500 Fluorescence Spectrophotometer (excitation 295 nm, emission 405 nm; slit width 5 nm ex/10 nm em; PMT voltage 700 V). Linear regression analysis allowed the estimation of the concentration of H₂O₂ in samples based upon a corresponding standard curve. More details of the H₂O₂ assay as used in our laboratory, including validation and optimisation can be found in Brooks and co-workers 15. The limit of detection of the assay used for these experiments was 3.4 nM H₂O₂.

For seasonal and event data comparisons of EBC pH, hydrogen peroxide, and 1-OHP concentrations were compared using the non-parametric Friedman's test and Wilcoxon-signed rank tests as appropriate to allow for repeated observations on individuals and due to the extreme non-normality of these measures. Readings below the limit of detection for any assay were arbitrarily assigned a value corresponding to 50% of the limit of detection to avoid division by zero in any subsequent calculation. Longitudinal data was cleaned to exclude spurious or physiologically impossible data points by an a priori defined protocol. Peak flow and FEV₁ data for each student was expressed as a Z-score relative to the student's mean score over the period of the study. Longitudinal data analysis exploring bivariate association between lung function (Z-scores of PEFR or FEV₁) or symptom data and pooled air pollution measures (both indoor and outdoor) and climate related measures were conducted using the non-parametric Spearman's correlation coefficient. The air pollution measures were lagged daily out to seven days and correlated with lung function and symptom scores. In this manner the potential for any lagged effect of air pollution on these measures was analysed. Variables showing some association (p < 0.15) with lung function or symptom data from these analyses were then entered into a mixed-model linear regression analysis which included the daily use of a reliever as a covariate to confirm the independent association with lung function or symptoms. These analyses were undertaken for the dataset as a whole and repeated separately for asthmatic and non-asthmatic students.

In addition, mean levels of each of the longitudinal outcome variables for each student were calculated within moderate/high (24 hour average> 20 μm/m³ outdoor) pollution days and low pollution days. These levels were compared using Mann-Whitney U tests. Longitudinal data analysis exploring bivariate association between lung function (Z-scores of PEFR or FEV₁) and pooled air pollution measures on moderate/high pollution (24 hour average> 20 μm/m³ outdoor) days alone were then conducted using the non-parametric Spearman's correlation coefficient.

The power of the study was of necessity dependant on the number and magnitude of high pollution days during the study period. In addition, limiting the sampling to one school (to increase exposure validity) placed limitations on potential numbers of students willing/able to participate. Retrospective power analysis was therefore undertaken for lung function scores (expressed as Z-scores) between moderate/high (24 hour average> 20 μm/m³ outdoor) pollution days and low pollution days. Assuming a 2:1 split in air pollution days, and given n = 80 students, the study had an 80% power to detect differences in Z-score values of 0.30 in FEV₁ and 0.27 in PEFR with α = 0.05 (two-tailed). In addition, with 100 time points a correlation between Z-scores and pollution of >0.28 would have been detected as statistically significant i.e. R² >10%.

A total of 93 students took part in the study, from a total potential boarding house population of 240 students. Eighty-nine students identified themselves as New Zealand European, whilst four (4.3%) came from Asia. There were no students who identified themselves as Maori. In 2001, 6.9% of the population of Christchurch identified themselves as Maori, while 5.5% identified themselves as Asian 16. Twenty-six (28%) of the students had a doctor diagnosis of asthma. The prevalence of asthma in this age group in New Zealand is 24.4% 17. Seventeen of these students used short-acting beta agonists on an as-required basis, while eight students were taking regular inhaler corticosteroids. Mean age of the students at study start was 14 years 5 months (range 12 years 9 months – 17 years 6 months). There were no significant age differences between students with asthma and those without asthma. There was no significant difference in % predicted FEV₁ between asthmatics and non-asthmatics (97.54% ± 2.481 (mean ± 1 s.e.) n = 26, vs 95.49% ± 1.635 n = 57), though FEV₁/FVC ratio was significantly lower in the asthmatic students (FEV₁/FVC ratio asthma group 0.803 ± 0.013 n = 26 vs normal group 0.834 ± 0.009 n = 57, p = 0.05).

Reliever use in the asthmatic students was generally low. Reliever use was only documented on 3.26% of the days in the diary cards filled in by the asthmatic students. There was no significant difference between the number of puffs of reliever used by the asthmatics on low pollution days when compared to moderate/high days (Mann-Whitney U, p = 0.488). There was however a relationship between days of ANY reliever use and the occurrence of a moderate/high pollution day. For this reason, the daily use of a reliever was used as a covariate in the mixed-model linear regression analysis.

The study was undertaken from the end of March to early September 2004, encompassing autumn and winter. Indoor and outdoor pollution levels are shown in Figure 1 (24 hour average) and Figure 2 (10-min peak). Pollution levels during the winter of 2004 were generally lower than previous years 18, with the majority of high pollution nights occurring during the school vacation period, when data was not being collected. There were however a number of high pollution events where data was collected. Detailed sampling took place after high pollution nights on June 24 and July 21. The 24-hour average outside PM₁₀ level recorded at the school on those dates was 43 μg/m³ and 72 μg/m³ respectively. Peak outdoor PM₁₀ levels for a 10-minute period reached 105 μg/m³ and 257 μg/m³ respectively. The majority of PM₁₀ pollution was in the PM_2.5 range, and indoor pollution levels were similar to outdoor. Occasional very high pollution peaks were identified indoors. These tended to be in the PM₁₀ range, rather than PM_2.5, and probably represented resuspension events due to activity in corridors. On average, pollution levels measured at the school were 22.5% lower than those measured at the central Environment Canterbury monitoring station.

Urine 1-OHP levels are shown in Figure 3. Urine 1-OHP levels were significantly higher on the high pollution days compared to both low pollution control days. Median (25–75% range, number of sample analysed) 1-OHP levels corrected for creatinine were: Autumn control 0.0195 (0.009–0.036, n = 88) μmol OHP/mol creatinine; Winter control 0.025 (0.013–0.038, n = 77) μmol OHP/mol creatinine; Pollution day 1 0.043 (0.030–0.073, n = 79) μmol OHP/mol creatinine; Pollution day 2 0.042 (0.022–0.064, n = 73); p < 0.0001 for all comparisons of high pollution days vs controls. Differences in 1-OHP levels between the two control days did not reach statistical significance. Similarly, the 1-OHP levels on the two high pollution days were not significantly different.

EBC pH readings are shown in Figure 4. There was a broad range of pH readings on all sampling days. pH differences between autumn and winter controls reached statistical significance, and pH at the first high pollution day was significantly higher than the second study day, and the winter control day, but not the autumn control day. There was no independent effect in students with asthma in comparison with healthy students.

EBC H₂O₂ readings are shown in Figure 5. Detectable hydrogen peroxide was present in most samples. There was no difference in measured H₂O₂ between any of the sampling days, either control or high pollution. There was also no difference in proportions of samples with detectable H₂O₂. There was no independent effect of asthma in comparison with healthy subjects. There was no significant correlation between EBC levels of hydrogen peroxide, and pH, and no correlation with either of these measurements and urinary 1-OHP.

FEV₁ data was available for a median of 129.5 time points per student (Range 4–189). PEFR data was available for a median of 128 time points (Range 2–186). Univariate analysis identified a number of biologically plausible correlations between lung function data and PM₁₀ levels which reached statistical significance. There was a significant correlation between morning FEV₁ and 24 hour average outdoor air pollution levels the previous day (Spearman's rho -0.201, p = 0.034). When analysing for the presence of asthma, asthmatic subjects showed correlations between FEV₁ and 24 hour average outdoor air pollution levels which just failed to reach statistical significance (Spearman's rho -0.187, p = 0.06), while non-asthmatic subjects showed no significant correlations. There was no significant correlation between peak PM₁₀ levels and lung function in either asthmatics or non-asthmatics. In the asthmatic subjects, there were a number of associations noted between pollution and temperature variables and the use of any reliever medication which reached statistical significance. These were daytime indoor pollution – one day lag; night-time indoor pollution – same night and one day lag; temperature; minimum and maximum inside temperature, and relative humidity. There was no significant relationship between the use of any reliever medication and any of the lung function variables in the asthmatic group.

When multivariate analysis was undertaken, using six dependent lung function variables (FEV₁ and PEF, morning and afternoon and daily variation in both measures), and lagging air pollution measures out to seven days, no pollution variable was retained in a regression model, with only minor temperature effects being retained. When asthmatics were studied separately, small but statistically significant associations between maximum outdoor air pollution levels and variability in FEV₁, and night time PEFR were noted. With higher maximum out door pollution, FEV1 variability between morning and afternoon dropped, and afternoon PEFR dropped. Effect size was small, in the case of afternoon PEFR having an R² of 3.6%. Reliever use was not retained in the regression models for any of the lung function data, except for daily variability in PEF.

When comparing median morning FEV₁ between moderate/high pollution days (>20 μg/m³) and low pollution days, students with asthma demonstrated lower readings on high pollution days (p = 0.043). No effect was seen in non-asthmatic students. However, studying moderate/high pollution days only, no significant correlation between lung function and pollution level could be detected, either in asthmatics or normal students.

Symptom score data was available for a median of 119 time points (Range 109–119). A number of biologically plausible associations were identified between symptoms and both indoor and outdoor air pollution levels. The proportion of students reporting cough and ear, nose and throat symptoms increased with increasing indoor air pollution the previous day. Maximum outdoor air pollution levels the same day were associated with increasing proportion of students reporting cough. The effect of maximum outdoor air pollution remained significant when multivariate analysis was undertaken, and remained in the model, while weather effects were excluded by the model. The effect size however was small, with a change in PM₁₀ of 50 μg/m³ being associated with a 1.5% increase in proportions of children reporting cough (equivalent to one extra report of cough on high pollution days in this cohort). The use of reliever medication was not retained in the model when asthmatics were studied separately, and in this sub-group, the effect of maximum outdoor air pollution remained statistically significant.