Association of short-term air pollution with systemic inflammatory biomarkers in routine blood test: a longitudinal study

The biological mechanisms behind health effects of air pollution have not been well known. Inflammation plays an important role in occurrence and development of a wide range of diseases. In this study, we assessed the effects of short-term exposure to ambient air pollution on systemic inflammatory biomarkers among 12 508 participants who underwent routine physical examination annually at the Hebei General Hospital in Shijiazhuang, China. For each participant, white blood cell count (WBC), lymphocytes, neutrophils and eosinophils were measured for two or three times during September 2016 to December 2018. Daily concentrations of nitrogen dioxide (NO2), sulfur dioxide (SO2), ozone (O3) and particulate matter less than 2.5 µm in aerodynamic diameter (PM2.5) were interpolated to each district, where the participants worked. The linear mixed-effects regression with a constrained distributed lag model was applied to examine the associations between air pollution and inflammatory biomarkers during lag 0–14 d. It was observed that WBC, neutrophils and eosinophils [percent change (%Δ) and 95% confidence interval (95%CI)] significantly decreased by −0.07 (−0.11, −0.04), −0.08 (−0.12, −0.03) and −0.15 (−0.25, −0.05) at lag 14 d, associated with per 10 µg m−3 increase in O3. WBC, lymphocytes and eosinophils (%Δ and 95%CI) significantly elevated by 0.08 (0.04, 0.12), 0.16 (0.11, 0.21) and 0.22 (0.10, 0.35) at lag 0 d, associated with per 10 µg m−3 increase in PM2.5. This study reveals short-term effects of air pollution on systemic inflammatory biomarkers in routine blood test, which is helpful for further study to explore the biological mechanisms.


Introduction
A growing body of studies have revealed that ambient air pollution was associated with a wide range of adverse health outcomes, such as cardiopulmonary and autoimmune diseases, thrombosis and cognitive impairment [1][2][3][4][5]. Air pollution is one of leading causes of global burden of disease [6,7]. Despite the growing epidemiological evidence, the biological mechanisms by which air pollution causes specific diseases remain unclear.
Inflammation plays an important role in occurrence and development of a wide range of diseases [8,9]. Investigating the inflammatory response can help to reveal the pathogenesis of diseases induced by air pollution [9][10][11][12]. According to a cross-sectional study in the U.S., exposures to higher levels of particulate matter ⩽10 µm in aerodynamic diameter (PM 10 ), nitrogen dioxide (NO 2 ) and sulfur dioxide (SO 2 ) pollution were associated with significant changes in levels of white blood cell count (WBC) and fibrinogen. The relative odds (95%CI) of WBC and fibrinogen were associated with an interquartile range (IQR) increase in PM 10 [1.77 (2.26, 2.49) and 1.64 (1.17, 2.30), respectively]. These changes were relevant to cardiovascular effects of air pollution [10]. A case crossover study in the U.K. illustrated that exposure to ozone (O 3 ) caused a 39% increase in sputum neutrophils in normal subjects, and neutrophils remained at high levels over a longer time period. This change further led to a neutrophil inflammatory response which induced reduced lung function [11]. Similar findings were reported by a study in northern France that levels of O 3 , NO 2 and PM 10 were associated with pulmonary inflammation markers like eosinophils and high-sensitivity C-reactive protein (hs-CRP). An increment in NO 2 level was significantly associated with a higher level of hs-CRP [5.03 (0. 36, 9.91)] on the day of the examination. These inflammatory reactions were one of the pathways leading to subclinical decrease in distal lung function [9]. In summary, exposure to particulate matter and gaseous pollutants may cause changed levels of inflammatory biomarkers, which further triggers inflammatory responses.
According to previous studies, alterations in inflammatory pathways have been proposed to be important mechanisms by which air pollution induces cardiovascular disease, reduced lung function and diabetes mellitus [13]. Meanwhile, the expression of pro-inflammatory cytokines by air pollution may lead to the anomalous change of inflammatory biomarkers, such as WBC, lymphocytes, neutrophils and eosinophils [9][10][11]. These four biomarkers are different types of peripheral white blood cells which can reflect the level of inflammation in human body at the early stage. Furthermore, abnormal changes in these four biomarkers in lung and other parts of the body are associated with increased risk of relevant diseases.
Although existing evidence have showed the associations of air pollution with inflammatory markers, however, the majority of the evidence is from crosssectional and case-control studies [14][15][16]. Evidence from large-scale prospective studies is very limited. In addition, most of previous studies have been conducted in high-income countries, but it is uncertain whether the conclusions also apply to low-and middle-income countries, where levels of air pollution, types of pollutants and characteristics of population are different from those of high-income countries [17][18][19].
Shijiazhuang is almost the most polluted city in China located on the North China Plain, due to rapid industrial development and high population density [20]. This study aims to examine the shortterm effects of air pollution on four inflammatory

Data collection
The participants included in this study were urban residents or office workers, who visited the Hebei General Hospital for annual physical examination during September 2016 to December 2018. In total, 12 508 participants were recruited who provided complete blood samples and finished all items of physical examination, among which, 11 885 underwent physical examination twice and 623 for three times during the study period. We excluded the participants who only underwent physical examination once. The locations of Shijiazhuang and the hospital are shown in figure S1 (available online at stacks.iop.org/ERL/16/035007/mmedia) in the supplementary material. The study has been approved by the scientific review and ethics committee of the Hebei General Hospital.

Inflammatory biomarkers
Four inflammatory biomarkers in peripheral blood were measured for all participants at each hospital visit, including WBC, lymphocytes, neutrophils and eosinophils. Specifically, 2 ml overnight fasting blood samples were extracted from the antecubital vein and were put in EDTA tubes, and then they were stored in refrigerator (2 • C-8 • C) until sent to analytical laboratories for testing. Blood routine examination was performed using a flow cytometry in automated These biomarkers have the same sample size (n = 25 639). Abbreviation: WBC, white blood cell count. complete blood count analyzer (Sysmex XN-B4), through which the total WBC (×10 9 l −1 ), lymphocytes (×10 9 l −1 ), neutrophils (×10 9 l −1 ) and eosinophils (×10 9 l −1 ) were determined.

Data on air pollutants and weather conditions
Daily mean levels of NO 2 , SO 2 , O 3 and particulate matter ⩽2.5 µm in aerodynamic diameter (PM 2.5 ) were obtained from 15 monitoring stations of the China National Environmental Monitoring Center in Shijiazhuang and surrounding cities during the study period. The measurements of air pollutants have been reported in details previously [21,22]. The level of pollutant was interpolated to each district where the participants worked using the inverse distance weighted (IDW) method [23,24]. The interpolation of air pollutants was validated using a leave-one-out cross-validation method [25]. The results of crossvalidation are shown in table S1. Meteorological data in Shijiazhuang and surrounding cities were obtained from six weather stations of the China Meteorological Data Sharing Service System. Daily mean temperature ( • C) and relative humidity (%) were estimated using the same IDW method as the air pollution data during the same exposure period.

Statistical analysis
To examine the associations between air pollution and inflammatory biomarkers, a linear mixedeffects regression model was used in this study. A range of potential confounders were controlled as fixed-effect terms in the model, including age ('⩽60 years' and '>60 years'), gender ('Male' and 'Female'), body mass index (BMI, '⩽24 kg m −2 ' and '>24 kg m −2 '), smoking and drinking status ('Current smoker/drinker' , 'Ex-smoker/drinker' and 'Nonsmoker/drinker') [26][27][28]. Each participant (personal ID) was also included as the random-effect term in the model. In order to test both the current-day and lag-effect of air pollutants, concentrations of each pollutant on the current day and previous 14 d (lag 0-14 d) were fitted using the constrained distributed lag model (CDLM) (natural cubic spline given three degrees of freedom for lag time) [29], according to our previous works [30,31]. Moreover, meteorological variables (including daily mean temperature and relative humidity) during lag 0-14 d were also controlled in the model using natural cubic splines given three degrees of freedom [32]. Both the single-day and cumulative estimated effects of air pollution were examined during lag 0-14 d [14,18,33]. Based on the overall estimated effects of air pollutants during lag 0-14 d, multi-pollutant models were developed to examine the joint effects of different air pollutants on biomarkers. In addition, a series of analyses were also performed stratified by age ('⩽60 years' and '>60 years'), gender ('Male' and 'Female'), BMI ('⩽24 kg m −2 ' and '>24 kg m −2 '). A two-sample test was performed to examine the statistically difference between different subgroups [30,34]. As data on WBC, lymphocytes, neutrophils and eosinophils were not normally distributed, levels of these three biomarkers were log transformed in the model [19,35,36], and their results were showed as percent change (%∆) and corresponding 95% confidence interval (95%CI), in relation to per 10 µg m −3 increment in each pollutant. Percent change (%∆) was back-transformed using the formula [100 × (exp β −1)] (β was coefficient from the linear regression model) [36,37]. As some previous studies considered the maximum lag time for shortterm effects of air pollution as 10 d [38,39], we have performed sensitivity analyses by shortening the lag period from lag 0-14 d to 0-10 d. As the current day exposure of each participant is the 24 h average level of pollutant, which may overlap with the time when the participant underwent examination, we performed sensitivity analyses by only considering the exposure during lag 1-14 d (excluding the current day exposure). The random-effect term was ID. Models adjusted for age, gender, BMI, smoking status, drinking status, mean daily temperature and relative humidity.
All statistical analyses were performed using R software (version 3.6.2). The 'dlmn' and 'nlme' packages were used for the CDLM and linear-mixed model.

Results
A summary of participants' basic information is shown in table 1. The mean (SD) age of all participants was 53.08 (14.24) years, and the majority of them were men (55.9%). Most of them never smoke or drink (75.0% and 60.3%). The measurements of four inflammatory biomarkers are summarized in table 2. The mean (SD) of WBC, lymphocytes, neutrophils and eosinophils were 6.33 ×10 9 l −1 (1.52), 2.14 ×10 9 l −1 (0.61), 3.66 ×10 9 l −1 (1.14) and 0.15 ×10 9 l −1 (0.14), respectively. The mean levels of daily air pollution in Shijiazhuang City during the study period are summarized in table 3. The mean  The results of stratified analyses for cumulative estimated effects of air pollutants are shown in figure 2. For the association between Lymphocytes and NO 2 or PM 2.5 , it was shown that older adults were more affected by pollutants. Compared with younger participants, more remarkable changes in lymphocytes were observed among older adults associated with 10 µg m −3 increase in NO 2  The results for overall estimated effects of air pollutants during lag 0-10 d were also similar to those during lag 0-14 d (table S2 in the supplementary material). Shortening maximum lag day from 14 d to 10 d did not substantially change the results. The results did not substantially change by only considering the exposure during lag 1-14 d (tables S3, S4 and figure S2). The Pearson correlation coefficients of air pollutants is shown in table S5. It showed O 3 was negatively associated with other pollutants.

Discussion
Our study has revealed the short-term effects of air pollution on inflammatory biomarkers. Higher levels of inflammatory biomarkers were significantly associated with increase in NO 2 , SO 2 and PM 2.5 , while their lower levels were associated with increase in O 3 .
Consistent with this study, previous studies also reported significant changes in inflammatory biomarkers induced by air pollution [18,[40][41][42]. A study conducted in Seoul, South Korea, reported that WBC in non-smokers was negatively associated with exposure to O 3 at lag 1-4 d, lag 6 d and lag 8 d [40]. A panel study in Germany stated that increased level of 24 hlymphocytes was associated with nitrogen monoxide (NO) and carbon monoxide (CO) pollution, but such association was not observed for eosinophils [41]. A Table 5. WBC, lymphocytes, neutrophils and eosinophils were expressed as percent change (%) and 95%CI. We used a linear mixed-effects regression model in this study. The lag time for daily mean temperature and relative humidity was 0-14 d. The random-effect term was ID. Models adjusted for age, gender, BMI, smoking status, drinking status, mean daily temperature and relative humidity. The bold values means significant changes, and p-value < 0.05.

Figure 2.
Changes in four inflammatory biomarkers associated with cumulative effects of air pollutants (per 10 µg m −3 increase) during lag 0-14 d modified by age, gender and BMI. WBC, lymphocytes, neutrophils and eosinophils were expressed as percent change (%) and 95%CI. We used a linear mixed-effects regression model in this study. The lag time for daily mean temperature and relative humidity was 0-14 d. The random-effect term was ID. Models adjusted for age, gender, BMI, smoking status, drinking status, mean daily temperature and relative humidity. * refers to p < 0.05. * * refers to p < 0.01. * * * refers to p < 0.001.
nested case-control study in Beijing among prediabetic and healthy individuals found that WBC and neutrophils were associated with PM 2.5 at lag 1-3 d and lag 1-4 d, respectively, and the maximum elevations of WBC and neutrophils were both observed on lag 2 d [%∆ (95%CI): 3.0 (1.5, 4.6) and 3.9 (1.6, 6.2) for per IQR increase] [18]. A longitudinal cohort study in Taiwan [42]. Changes in inflammatory biomarkers induced by air pollution may play an important part in the occurrence and development of various diseases, such as chronic obstructive pulmonary disease [43] and asthma [44]. However, their biological mechanisms are complex and have not been well understood. One potential pathway is that exposure to air pollutants triggers the release of pro-inflammatory mediators which further leads to the migration, infiltration, or extravasation of relevant inflammatory cells [45][46][47][48][49][50][51][52][53]. Evidence also indicated alveolar macrophages involved in changes of inflammatory markers that related to particulate matter exposure. An animal study revealed that phagocytosis of fine particles by rabbit alveolar macrophages could elevate white blood cells by releasing precursors from the bone marrow, and then white blood cells quickly entered into bloodstream [46]. Gaseous pollutants can also have an impact on inflammatory markers by facilitating the release of chemoattractant and stress hormones in vivo. After 5 h exposure to SO 2 , the rats showed increased neutrophil chemoattractant GRO/KC coinciding with early signs of acute airway inflammation with neutrophilic and macrophage infiltrates, followed by a shift from neutrophils to eosinophils [47]. The increase and extravasation of neutrophils in lungs were associated with the release of O 3induced stress hormones [49]. Moreover, increased circulating epinephrine and corticosterone induced by O 3 could trigger innate neutrophilic immune response [50]. Blocking the receptors of the stress hormones could minimize O 3 -induced neutrophilic inflammatory, lymphopenia, and pulmonary inflammatory cytokine expression in rats [51].
Some results for single pollutant changed after adjusting for other pollutants, which suggested that these four pollutants were not associated with levels of biomarkers independently. According to the correlation table of pollutants (table S5), the correlation coefficients between O 3 and NO 2 , SO 2 and PM 2.5 were −0.53, −0.36 and −0.33, respectively, which showed O 3 was negatively associated with other air pollutants. This may be one of the reasons for the opposite association between O 3 and inflammatory biomarkers compared with other pollutants [54]. Shijiazhuang is almost the most polluted city in China. Some studies indicated that the level of O 3 was associated with different changes of biomarkers. Changes in biomarkers under high-level O 3 exposure may be opposite to those under low-level O 3 exposure. A multicenter crossover study found that with the increase in level of O 3 , nitrotyrosine increased when the level of O 3 was low (0-70 ppb), but decreased when the level of O 3 was high (70-120 ppb) [55]. Similar associations between O 3 and hs-CRP were also reported by a cross-sectional study in Germany [14]. In addition, it has been reported that the toll-like receptor 4 and the nuclear factor-kB pathway could be suppressed by O 3 , followed by a reduction of inflammatory cytokines levels [56,57].
Although our study also found the significant associations between air pollution and inflammatory biomarkers, inconsistency in effect estimate (e.g. percent change) and lag time remains comparing with previous studies. The inconsistency may be due to different study populations, designs and statistical approaches. For study population, most of the participants in our study were urban residents or office workers who visited hospital for routine health examination. They tended to have comparable occupation, living environment and working schedule, which helps to reduce the impact of some socialeconomic confounders on the results. Our study is a large-scale prospective study based on routine heath examination. It takes into account the variabilities of measurements both between-and withinindividual over time. Thus, the findings of our study can provide valuable information for the causal relationship of air pollution with health outcomes. For statistical approach, the CDLM we used facilitates users to model the complex exposure-responses association (linear or nonlinear) and its lag structure, and it provides the additional temporal dimension that helps to express the association [29].
In our stratified analyses, we found older adults were more vulnerable to air pollution, showing greater changes in inflammatory markers. Similar findings were also reported by several previous studies [58][59][60]. Inconsistent results were reported by previous studies regarding the interaction of gender and BMI in the association between air pollution and inflammatory biomarkers [61][62][63]. Therefore, further investigation is needed to identify the vulnerable population. Few parameters were statistically significant, indicating that no statistical differences in associations between air pollution and inflammatory biomarkers among different subgroups. Moreover, the smaller sample size in each stratum compared with main analyses resulted reduced statistical power.
Our study has several limitations. Due to the lack of individual's home address, we interpolated level of air pollution to each district where the participants worked rather than individual's home address [64]. This may bring some measurement errors for the assessment of exposure. Due to lack of data, we cannot consider participants' hourly level of exposure, but it does not have substantial impacts on the mean level of exposure during lag period. Levels of air pollutants were calculated on an average of 0-24 h on any given day. The 24 h mean O 3 concentration was lower than the daily maximum 8 h O 3 concentration, but it had already reached the threshold value that induced abnormal changes of risk indicators of certain diseases, such as cardiovascular diseases [65]. In addition, some potential confounders were not controlled in our study such as household income and education, as the information was not collected at the hospital visit.

Conclusion
In this large-scale longitudinal study, short-term exposure to ambient air pollution showed significantly estimated effects on levels of WBC, lymphocytes, neutrophils and eosinophils. Considering the lag effects of air pollution and joint effects of different air pollutants on health, the government should take effective measures to curb air pollution and protect vulnerable population. Regarding the complex relationships between air pollution and various diseases or biomarkers, more researches should be conducted in future to explore the biological mechanisms for health effects of air pollution.

Data availability statement
All data that support the findings of this study are included within the article (and any supplementary files).