Short-term effects of ambient air pollution and childhood lower respiratory diseases

The association between air pollution and childhood respiratory disease is inconsistent. In the present study, we investigated a short-term effect of ambient air pollutants and daily childhood lower respiratory diseases (CLRD). Daily air pollutants, weather data, and CLRD data were collected from January 2014 to April 2015 (452 days) in Nanjing, China. Time-series regression and generalized additive models were used to assess the effects of air pollutants (PM10, PM2.5, NO2, SO2, O3, and CO) on CLRD. We observed that an interquartile range (IQR) increase in concentrations of PM10, NO2, and SO2 significantly increased the daily CLRD with 6 days cumulative effects (difference of estimates: 2.8%, 95% CI: 0.6–5.0%; 4.1%, 1.2–7.0%; 5.6%, 2.6–8.6%, respectively). However, no significant association was found in IQR concentrations of PM2.5, O3, and CO. Specifically, elevated PM10, PM2.5, NO2, and SO2 significantly increased the numbers of CLRD in cool season (3.6%, 1.5–5.7%; 2.4%, 0.3–4.5%; 4.9%, 2.9–7.0%; 6.3%, 3.7–9.0%, respectively). Additionally, the effect estimates of PM10, NO2, and SO2 in female and age >27 months were more pronounced than in male and age ≤27 months. This study suggested that short-term exposure to ambient PM10, NO2, and SO2 were associated with the increased CLRD numbers.

the higher childhood asthma 10 . Liu et al. collected the daily concentrations of PM 10 , NO 2 , and SO 2 during the children life time (2006)(2007)(2008)(2009)(2010)(2011)(2012) where they lived, and found that exposure to NO 2 was associated with respiratory diseases in children 11 . Although the association studies of air pollutants and CLRD have been widely studied, the results remain inconsistent. Molter et al. found that outdoor air pollutants were not related with the multi-center childhood asthma in Europe 12 . Studies demonstrated that different NO 2 exposure years of life were not associated with childhood asthma and asthma-related symptoms in the United States 13 .
It was worth to note that these studies had provided the effects of air pollutants on CLRD in location for public health assessments, however, the results might be affected by the local air pollutants levels, components and population susceptibility. It is important to study the effects of city air pollutants on CLRD numbers for the local city policy development 14 . Nanjing is a city located in the east of China, which is the provincial capital of Jiangsu province. With the development of economics, the concentrations of Nanjing's air pollutants in central monitoring sites are above the nation's levels. In the present study, we aim to conduct a short-term effect of ambient air pollutants, including PM 10 , PM 2.5 , NO 2 , CO, and O 3 , and daily CLRD in Nanjing, China.

Material and Methods
Data collection. Nanjing is a large city of Jiangsu province with a moderate subtropical climate (6597 km 2 ). It had a population of 8.22 million in 2015. Daily CLRD data from 28 January 2014 to 24 April 2015 were obtained from Children's Hospital of Nanjing Medical University, The Second Affiliated Hospital of Nanjing Medical University, Jiangsu Women and Clinical Health Hospital, Nanjing Jiangning Hospital, Nanjing First Hospital, Nanjing Gaochun People's Hospital, Nanjing Maternity and Child Health Care Hospital. In this study, CLRD were comprised of doctor-diagnosed pneumonia (J18), bronchitis (J20), capillary bronchitis (J21), and acute asthma (J45), which were diagnosed based on the International Classification of Diseases 10th revision 15 .
Air pollution data including PM 10 , PM 2.5 , NO 2 , SO 2 , O 3 , and CO were obtained from China Environmental Monitoring Centre. These monitoring stations are not in direct vicinity to traffic or industrial sources, housing emissions such as oil, coal, or waste burning. Thus, the daily air pollution data of monitoring stations can represent the environmental air pollutants' levels. To control the confounding effects of meteorological conditions on CLRD, we also obtained the data of daily weather data (relative humidity and mean temperature) from the Meteorological Bureau.
Statistical analysis. Spearman's correlation coefficient was applied to investigate the association between air pollutants and weather variables. We used the time-series regression model to explore the effects of each air pollutant (PM 10 , PM 2.5 , NO 2 , SO 2 , O 3 , and CO) on CLRD. We also used generalized additive model (GAM) to calculate the data 16 . Herein, the data of the respiratory symptoms followed the over-dispersed Poisson distribution, thus, we applied the quasi-Poisson regression in the GAM 17 . In order to control the potential confounding effect, firstly, we included a natural cubic smooth function of calendar time with 7 degrees of freedom (df) per year to remove unmeasured long-term and seasonal trends 18 ; secondly, we included the natural smooth functions of the current-day relative humidity (3 df) and mean temperature (3 df) to exclude the weather confounding effects 19 ; thirdly, we also included the indicator variable for "day of the week (DOW)" in the model. In this study, we also calculated the data in the cool period (November through April, 4 df) and warm period (May to October, 4 df) to separately analyze the effect of air pollutants on CLRD.
Further, we used single-pollutant model to explore the air pollution's effect on CLRD with single lag days (lag0, 1, 2, 3, 4, 5, 6). To remove the potential misalignment of single day lag exposure, we considered the moving average exposure of multiple days (lag 0-1, 0-2, 0-3, 0-4, 0-5, 0-6) in additional analyses. We conducted the concentration-response relationship curves for the air pollutants concentrations with daily CLRD using a 3 df for the smoothing function in single-pollutant model. Herein, we performed the two-pollutants model to exam the stability of results.
The effect estimates were shown as the percentage change and 95% confidence intervals (CIs) in daily CLRD per interquartile range (IQR) increased of air pollutants' levels. In this study, all models were conducted in R software (version 3.2.1) with mgcv package. All tests were two-sided, and P < 0.05 were considered as statistically significant.  24 April 2015 (452 days), among which there were 5,551 bronchitis, 19,098 pneumonia, 1,560 capillary bronchitis, and 214 asthma cases. The daily average CLRD numbers were 58.5, and the maximum was 105.0. The daily average concentrations were 117.4 µg/m 3 for PM 10 , 69.9 µg/m 3 for PM 2.5 , 50.1 µg/m 3 for NO 2 , 21.5 µg/m 3 for SO 2 , 0.94 mg/m 3 for CO, and 95.8 µg/m 3 for O 3 . The daily average relative humidity and temperature were 72.5% and 15.1 °C, respectively.

Data description.
As shown in Table 2, there were positively correlated among 6 air pollutants, except CO and O 3 (r = −0.17, P = 0.001). In addition, air pollutants were also correlated with temperature and humidity.
Concentration-response relationships of CLRD numbers with air pollutants also were performed ( Figure 3). The curves for PM 10 , NO 2 and O 3 were similar and linearly positive and flat at higher concentrations. The effects of SO 2 tend to be linearly positive with its concentrations, and appeared to be stable when concentration was beyond 40 µg/m 3 . However, the curves for PM 2.5 and CO were linearly associated without any thresholds.

Discussion
The source of air pollution is mainly manufacturing industries, traffic, agriculture, forest fires, etc., which result in increasing air pollutants level. Exposure to outdoor air pollutants such as particulate matter, NO 2 , SO 2 had the adverse effects on the airway responsiveness and affected the immune competent cells 20,21 . Epidemiological association studies between the air pollutants and childhood respiratory health have been studied widely; however, results are conflicting. For example, in a prospective birth cohort, Molter et al. did not find any evidence of the effect on the association between PM 10 and NO 2 and childhood asthma or wheeze 22 . Similar to their study, a study in Norway suggested that exposure to NO 2 was not associated with childhood asthma at age 9-11 years old 23 . European BAMSE study also observed no significant association between PM 10 exposure and childhood asthma 24 . Nevertheless, Clarket et al. found the significant association between NO 2 , PM 10 , SO 2 exposure and Canadian childhood asthma 25 .
Children exposure to air pollutants is vulnerable to suffering from pulmonary diseases, owing to a larger surface area of lung space than adult 26 . Exposure to higher air pollutants levels may induce the pathologic changes in childhood airway mucous and make children prone to the prevalence of chronic inflammation in respiratory   Table 2. Spearman correlation between air pollutants and weather variables in Nanjing. **P < 0.001. system 27 . Many studies have investigated the association between air pollutants and CLRD, of which the majority suggested the effects of short-term exposure to air pollutants up to 4 or 5 days 28,29 . Similarly, we also found a cumulative effect of PM 10 , NO 2 , and SO 2 (lag0-5) on the short term effects with CLRD. Besides, we also investigated the effect estimates of PM 10 , NO 2 , and SO 2 on CLRD by stratification of diseases (bronchitis, pneumonia, capillary bronchitis, and asthma cases). Some differences observed between PM 10 , NO 2 , and SO 2 and diseases. Exposure to PM 10 can induce the lung function disorder of children. In this study, we observed that an IQR increase in PM 10 daily concentrations increased the daily numbers of CLRD with 5, 6, or 7 days cumulative effects (lag0-4, lag0-5, lag0-6, respectively). Recently, a retrospective study also showed a positive association between the cumulative effects of 7 days (lag0-6) for PM 10 and CLRD 15 . Further, we found the significant association between PM 10 and pneumonia. A Jinan study revealed that PM 10 were associated with increased number of Childhood pneumonia 30 . In 2010, Weinmayr and his colleagues conducted a systematic literature review to assess the effect of PM 10 on respiratory health in children, which had provided the clear evidence of PM 10 as the risk factor on childhood asthma symptom episodes 31 . However, we did not observe the similar association. It is possible that genetic background can contribute to the differences. NO 2 exposure can damage the lung function, and induce the change of allergic airway inflammation, and responses to allergens 32 . Similar to PM 10 , we found that exposure to NO 2 was significantly associated with CLRD. The estimated effect for NO 2 was consistent with many other studies 33,34 . In Italy, Bono et al. found the adverse health effect of NO 2 on the risk of emergency room admissions for childhood respiratory diseases 34 . A meta-analysis included seven countries also confirmed that exposure to NO 2 was significantly associated with the asthma and wheezing of children aged 0-18 35 . In the present study, the significant effect estimate of NO 2 on childhood pneumonia and capillary bronchitis was more pronounced than other diseases (bronchitis, and asthma).    SO 2 is a well-known gas with strong irritant, and can induce the responses of respiratory system. A study showed that children living near a petrochemical complex with high SO 2 exposure had significantly increased incidence of bronchitis, asthma and allergic rhinitis than low exposure group 36 . In addition, Samoli et al. also observed that exposure to SO 2 (10 µg/m 3 ) significantly increased the number of childhood asthma hospital admissions 37 . In this study, we also found that exposure to SO 2 had a positive association with the numbers of CLRD, especially with childhood bronchitis, pneumonia, and capillary bronchitis.
Season was an important factor in modifying the association between air pollutants and population health effects 38 . In this study, we found the effects of air pollutants on CLRD numbers varied by season. Specifically, the effect estimates of PM 10 , NO 2 , and SO 2 were positive associated with CLRD numbers in cool season. Chen's findings also suggested that exposure to PM 10 and NO 2 had the largest effects in childhood asthma in cool season 39 . A system review and meta-analysis also proposed that children exposure to PM 10 , NO 2 , and SO 2 have the stronger association of asthma-related hospital admission in warm season 40 . The exact seasonal reasons were still not clear, and might be related with the difference in air pollution levels, surrounding environment, or races.
Several studies have found that age and sex can influence the effect estimates of air pollutants on population health. In the present study, we explored the effect estimates of air pollutants on CLRD in the subgroup of sex and age. Results showed that the effect estimates of PM 10 , NO 2 , and SO 2 in female were more pronounced than in male, which were consistent with previous study 41 . Similar results were observed in children with age >27 months. It was possible that children with age >27 months spent more time outdoors than age ≤27 months.
Herein, we should mention several limitations. Firstly, we obtained the data only from one city. Exposure to air pollutants effects may be influenced by person's living and other factors. Results were difficult to spread to other cities. Secondly, we did not obtain the exactly individual exposure levels of air pollutants, thus, we merely calculated the average air pollution levels of local region to evaluate the estimate effects of air pollutants. Thirdly, we collected the useful data based on one and a half years data from the seven hospitals, which may not represent the entire populations. Besides, our study is the retrospective study, we cannot obtain the biochemical indexes of CLRD cases. Further larger studies should be conducted to comprehensively clarify the effects of air pollutants on CLRD. This study also had several strengths. Firstly, it was the first time to investigate the effect estimate of air pollutants on CLRD in Nanjing, China. Secondly, we used the time-series regression and GAM analysis to evaluate the effects of air pollution. Thirdly, although we obtained the data in a city, the results were consisted with many other studies 39,40,42 .
In summary, we found that the short-term exposure to ambient PM 10 , NO 2 , and SO 2 were associated with increased CLRD numbers. It was worth to note that our data provided the limited information; therefore, further studies should be conducted to validate our findings.