- Email: [email protected]
Ambient concentrations of particulate matter and hospitalization for depression in 26 Chinese cities: A case-crossover study
Environment International 114 (2018) 115–122
Contents lists available at ScienceDirect
Environment International journal homepage: www.elsevier.com/locate/envint
Ambient concentrations of particulate matter and hospitalization for depression in 26 Chinese cities: A case-crossover study
T
Feng Wanga,b,c,1, Hui Liud,1, Hui Lia,b,c,1, Jiajia Liua,b,c, Xiaojie Guoa,b,c, Jie Yuane, Yonghua Hud, ⁎ ⁎⁎ Jing Wangd, , Lin Lua,b,c, a
Peking University Sixth Hospital/Institute of Mental Health, 100191 Beijing, China National Clinical Research Center for Mental Disorders & Key Laboratory of Mental Health, Ministry of Health (Peking University), 100191 Beijing, China c Key Laboratory of Mental Health, Ministry of Health, Peking University, 100191 Beijing, China d Peking University Medical Informatics Center, Peking University, 100191 Beijing, China e North China University of Science and Technology, 063000, Hebei Province, China b
A R T I C L E I N F O
A B S T R A C T
Handling Editor: Yong Guan Zhu
Objective: Air pollution with high ambient concentrations of particulate matter (PM) has been frequently reported in China. However, no Chinese study has looked into the short-term effect of PM on hospitalization for depression. We used a time-stratified case-crossover design to identify possible links between ambient PM levels and hospital admissions for depression in 26 Chinese cities. Methods: Electronic hospitalization summary reports (January 1, 2014–December 31, 2015) were used to identify hospital admissions related to depression. Conditional logistic regression was applied to determine the association between PM levels and hospitalizations for depression, with stratification by sex, age, and comorbidities. Results: Both PM2.5 and PM10 levels were positively associated with the number of hospital admissions for depression. The strongest effect was observed on the day of exposure (lag day 0) for PM10, with an interquartile range increase in PM10 associated with a 3.55% (95% confidence interval: 1.69–5.45) increase in admissions for depression. For PM2.5, the risks of hospitalization peaked on lag day 0 (2.92; 1.37–4.50) and lag day 5 (3.65; 2.09–5.24). The elderly (> 65) were more sensitive to PM2.5 exposure (9.23; 5.09–13.53) and PM10 exposure (6.35; 3.31–9.49) on lag day 0, and patients with cardiovascular disease were likely to be hospitalized for depression following exposure to high levels of PM10 (4.47; 2.13–6.85). Conclusions: Short-term elevations in PM may increase the risk of hospitalization for depression, particularly in the elderly and in patients with cardiovascular disease.
Keywords: Depression Particulate matter Hospitalization China
1. Introduction Depression is a serious mental disorder that profoundly affects individual quality of life and can impair daily functioning, even leading to suicide (Moscicki, 2001). The prevalence of depression the US has been estimated at 9.0% (CDC, 2010). The estimated prevalence of major depressive disorder in China is 1.6% (2.1% for females and 1.3% for males) (Gu et al., 2013). After the first depressive episode, more than half of patients develop a recurrent or chronic disorder, causing personal suffering and economic problems that are a substantial burden to society (Kleine-Budde et al., 2013). In China, the total cost of
depression in 2002 has been estimated at 51,370 million RMB ($6264 million USD) (Hu et al., 2007). Some researchers suggest that prevention is critical in reducing this burden (Cuijpers et al., 2012). Understanding the factors involved in depression is therefore of practical importance. The onset of depression can be triggered by various factors. Apart from genetic susceptibility, environmental risk factors may contribute to the onset or aggravation of depression. Contributing factors identified in earlier studies include stressful events, such as losing jobs and loved ones (Rashid and Heider, 2008), and physical conditions, such as diabetes (Rustad et al., 2011) and stroke (Saravane et al., 2009). Air
⁎
Corresponding author. Correspondence to: L. Lu, Peking University Sixth Hospital/Institute of Mental Health, 100191 Beijing, China. E-mail addresses: [email protected] (F. Wang), [email protected] (H. Liu), [email protected] (H. Li), [email protected] (J. Liu), [email protected] (X. Guo), [email protected] (J. Yuan), [email protected] (Y. Hu), [email protected] (J. Wang), [email protected] (L. Lu). 1 These authors contributed equally to this work. ⁎⁎
https://doi.org/10.1016/j.envint.2018.02.012 Received 28 September 2017; Received in revised form 7 February 2018; Accepted 8 February 2018 0160-4120/ © 2018 Published by Elsevier Ltd.
Environment International 114 (2018) 115–122
F. Wang et al.
matter < 10 μm in aerodynamic diameter) between January 1, 2014 and December 31, 2015. The daily (24-hour) mean concentrations of pollutants averaged across all the stations in a given city were used as the reading for that city on that day. In addition, the daily (24-hour) average temperature and relative humidity were obtained from the Chinese Meteorological Bureau, allowing for adjustment for weather conditions.
pollution has also been shown to be involved, which is a heterogeneous mixture of both gaseous pollutants and particulate matter (PM), and especially fine PM (PM2.5, particles with an aerodynamic diameter ≤ 2.5 μm), Animal models have shown evidence of the neuropathological effects of exposure to PM (Campbell et al., 2005; Veronesi et al., 2005), and it has been suggested that high concentrations of inhaled air pollutants may reach the brain and cause neuroinflammation (Calderon-Garciduenas et al., 2008). Chronic inflammation has been linked to depression (Brites and Fernandes, 2015). Air pollution may contribute to a depressive mood by inducing dopaminergic neurotoxicity, possibly owing to oxidative stress (Block et al., 2004), as a decrease in dopamine in the central nervous system is one of the underlying pathophysiological mechanisms of depression (Hasler, 2010). Several studies found a relationship between depression and exposure to PM. Studies conducted in Canada indicated that emergency department visits for depression were associated with exposure to ambient air pollution (Szyszkowicz, 2007; Szyszkowicz et al., 2009, 2016). Lim et al. (2012) reported that increases in PM10, nitrogen dioxide (NO2), and ozone levels may increase depressive symptoms among the elderly. One prospective cohort study reported a possible association between both long-term PM2.5 exposure and depression onset and incident antidepressant use (Kioumourtzoglou et al., 2017). Kim et al. (2016) reported that long-term PM2.5 exposure may increase the risk of major depressive disorder. However, to our knowledge, no published study has investigated the association between PM exposure and depression in China, one of the most polluted countries in the world (Kan et al., 2012). With rapid industrialization and increased energy consumption over the past several decades, air pollution, and especially PM pollution, has become a severe environmental problem in China. Given the public health and economic burden of depression, it is important to determine whether specific air pollutants are linked to an increased risk of depression hospitalization; a better understanding of the underlying mechanisms may lead to evidence-based policymaking for primary prevention and specific interventions. This study investigated the association between ambient concentrations of PM and the number of depression-related hospital admissions in 26 Chinese cities between 2014 and 2015, stratifying by sex, age, and comorbidities.
2.3. Study design To investigate a possible association between ambient PM concentrations and hospitalization for depression, we adopted a timestratified case-crossover design where cases were used as their own controls (Carracedo-Martinez et al., 2010). With this approach, we were able to adjust for time-invariant characteristics such as age and sex, to ensure unbiased estimates from conditional logistic regression and to avoid time-trend bias (Janes et al., 2005). For each case, the patient's exposure to ambient PM on the day of hospital admission was compared with the exposure on three or four reference days. The reference days were the same days of the week within the same month and year of admission. Using this method, we controlled for the influence of day of week, seasonal and long-term trends, and potential individual-level risk factors (such as sex and genetics). 2.4. Statistical analysis Pooled analyses were applied in this study; observations from all included cities were combined, and each city was given a special indicator in the dataset. General and clinical characteristics of all 19,646 cases were described, and Spearman's correlation analysis was applied to assess the associations between exposure variables. We then used conditional logistic regression to examine associations between PM concentrations and hospitalizations for depression for each special lag day. Distributed lag non-linear models (DLNM) with three degrees of freedom in the natural cubic splines and a maximum lag of 3 days were used to adjust for the delayed and non-linear effects of temperature and humidity (Goldberg et al., 2011). To control for the meteorological effects on health in different areas, we added interactions between meteorology and the cities to the models. Public holidays were included in the models. The results are described as the percentage of PM concentration increase and the 95% confidence intervals (CIs) in daily admissions for depression per interquartile range (IQR = 75th percentile–25th percentile of air pollutants). We also analyzed the exposureresponse association between concentrations of PM2.5 and hospitalizations for depression. The DLNM was used to determine the non-linear delayed relationships between exposure and hospitalization for depression. The models included various lag structures—from the day of hospitalization (lag day 0) up to seven lag days (lag day 7)—to examine the temporal associations of PM concentrations and depression. In addition, we examined the associations with 3-day (lag days 0–2) and 6-day (lag days 0–5) moving mean PM concentrations, to avoid underestimating the effect of pollutants measured by single-day lag models (Bell et al., 2004). In order to explore the impact of PM on depression hospitalization varies across levels of other pollutants, including sulfur dioxide (SO2), NO2, and carbon monoxide (CO), two-pollutant models with first order interaction were added in the sensitivity analysis (CarbajalArroyo et al., 2011; Winquist et al., 2014). To determine whether the associations differed by sex and age (< 18, 18–65 and ≥65 years), stratified analyses using a Z-test were applied (Altman and Bland, 2003). To examine whether PM exerted different effects on patients with comorbid chronic diseases, we singled out patients who, in addition to a primary diagnosis of depression, had cardiovascular diseases (CVDs, ICD-10 codes I10.x–I15.x, I20.x–I25.x, or I60.x–I69.x), diabetes (E10.x–E14.x), and chronic obstructive pulmonary disease (COPD, J40–J44.x). All analyses were conducted using
2. Materials and methods 2.1. Study population Data on hospital admissions were collected from the electronic hospitalization summary reports of tertiary A hospitals in 26 Chinese cities (Supplementary Fig. 1), which recorded basic demographics (sex and age), dates of admission and discharge, hospitalization and discharge diagnoses in Chinese and their corresponding codes in the 10th revision of the International Classification of Diseases (ICD-10), treatments (mainly surgical information), discharge status (survival, drug allergy, and infection), and hospitalization expenses. We used ICD-10 codes to identify depression-related admissions between January 1, 2014 and December 31, 2015. These codes included F32 (mild depressive episode), F33 (recurrent depressive disorder), F34.1 (dysthymia), and F41.2 (mixed anxiety and depressive disorder). In total, 19,646 hospital admissions were depression-related. 2.2. Air pollution and meteorological data Data on air pollution were obtained from the National Air Pollution Monitoring System, which is run by the Chinese Ministry of Environmental Protection. The system fulfills the quality assurance and quality control mandates of the Chinese government through its ambient air-monitoring stations. These stations, ranging in number from 4 to 15 per city, provide hourly air pollution data to the system. We collected records on the levels of PM2.5 and PM10 (particulate 116
Environment International 114 (2018) 115–122
F. Wang et al.
Table 1 Demographic characteristics of depression admissions. Variable
Depression individuals
Total Gender Male (%) Female (%) Age (year) (mean ± SD) < 18 (%) 18–64 (%) ≥65 (%) Comorbidities Cardiovascular disease (%) Diabetes (%) COPD (%)
19,646 6493 (33.1) 13,153 (66.9) 46 ± 16.8 1089 (5.5) 15,716 (80.0) 2841 (14.5) 4687 (23.9) 1287 (6.6) 321 (1.6)
R (V.3.2.2, R Development Core Team). All statistical tests were twosided, with a significance level of 5%. 3. Results A total of 19,646 hospital admissions were identified to be depression-related (patients' mean age: 46.0 ± 16.8 years; 33.1% males and 66.9% females). Among the cases, 5.5% were adolescents (mean age: 15.2 ± 1.7 years) and 14.5% were elderly individuals (mean age: 71.4 ± 5.3 years). Comorbid CVDs were associated with 4687 cases; 1287 cases had diabetes; 321 had COPD (Table 1). 3.1. Air pollution and meteorological variables Table 2 displays descriptive statistics for air pollutants and meteorological measurements in all cities combined during the study period. The mean temperature (SD) was 14.5 °C (10.9 °C), and the relative humidity (SD) was 69.2% (33.2%). Concentrations of PM2.5 were positively associated with concentrations of PM10 (r = 0.87, p < 0.001), and both values were positively associated with the concentrations of CO, NO2, and SO2 (r = 0.54–0.68, p < 0.001). However, concentrations of air pollutants in general were negatively correlated with temperature and humidity. 3.2. Associations between air pollution and hospitalizations for depression Positive associations between PM concentrations and hospitalizations for depression were observed (Fig. 1). Table 3 shows increases in depression hospitalizations associated with an IQR increase in PM2.5 and PM10 concentrations for different lag structures. The associations between PM and hospitalization for depressive disorders were observed on lag day 0, and remained significant for 5 days after exposure. PM2.5 exerted the highest effects on lag day 0 (2.92% increase; 95% CI, 1.37–4.50) and lag day 5 (3.65; 2.09–5.24), whereas PM10 exerted the highest impact on lag day 0 (3.55; 1.69–5.45). DLNM analysis showed similar trends (Supplementary Fig. 1).
Fig. 1. The exposure-response curve of 6-day (lag0–5) moving average fine particulate matter (PM2.5) and PM10 concentrations (degree of freedom = 3) and ischemic stroke hospitalizations in 26 cities. Note: The X-axis is the 6-day (lag0–5) moving average PM concentrations (μg/m3). Y-axis is the predicted log (relative risk (RR)), after adjusting for temperature and relative humidity, is shown by the solid line, and the dotted lines represent the 95% CI.
Table 2 Summary statistics for air pollutants concentrations and meteorological variables. Variable
Mean ± SD
3
PM2.5 (μg/m ) PM10 (μg/m3) SO2 (μg/m3) NO2 (μg/m3) CO (mg/m3) Temperature (°C) Relative humidity (%)
63.5 ± 50.6 106.8 ± 71.9 29.6 ± 32.6 44.1 ± 19.4 1.15 ± 0.63 14.5 ± 10.9 69.2 ± 33.2
Minimum
5.1 7.4 1.9 4.5 0.14 −25.7 8
Percentile 25th
50th
75th
31.5 58.3 11.4 30.0 0.76 7.0 53
49.4 89.4 18.8 40.2 0.99 16.4 69
79.0 135.2 33.6 54.1 1.32 23.3 80
117
Maximum
IQR
897.5 977.3 316.9 175.8 8.41 35.5 97
47.5 76.9 22.2 24.1 0.56 16.3 27
Environment International 114 (2018) 115–122
F. Wang et al.
concentrations on lag day 0 corresponded with 5.64% (−0.59–12.25) and 3.1% (−1.55–7.96) increases in admissions, respectively. However, associations in adolescents did not reach statistical significance (p > 0.05), perhaps because of a smaller sample (1089) (Supplementary Table 2).
Table 3 Percentage change with 95% CI in depression admissions associated with an interquartile range increases in PM2.5 and PM10 concentration for different lag structures. Lag days
Percentage change
95% CI
p
PM2.5 Lag 0 days Lag 1 days Lag 2 days Lag 3 days Lag 4 days Lag 5 days Lag 6 days Lag 7 days Lag 0–2 days Lag 0–5 days
2.92 2.55 2.07 2.6 3.29 3.65 2.47 1.23 3.77 6.21
1.37–4.5 0.97–4.16 0.51–3.66 1.05–4.18 1.71–4.9 2.09–5.24 0.95–4.02 −0.3 to 2.78 1.84–5.74 3.85–8.63
< 0.001 0.002 0.009 < 0.001 < 0.001 0.028 0.001 0.115 < 0.001 < 0.001
PM10 Lag 0 days Lag 1 days Lag 2 days Lag 3 days Lag 4 days Lag 5 days Lag 6 days Lag 7 days Lag 0–2 days Lag 0–5 days
3.55 3.13 1.88 2.38 2.37 2.69 1.75 0.53 4.36 5.86
1.69–5.45 1.27–5.03 0.04–3.76 0.54–4.25 0.55–4.21 0.89–4.53 −0.05 to 3.6 −1.27 to 2.38 2.05–6.73 3.09–8.71
< 0.001 < 0.001 0.045 0.001 0.001 0.003 0.057 0.565 < 0.001 < 0.001
3.3. Effect modification by comorbidities Analyses revealed a significant association between exposure to air pollutants and hospitalizations for depression in individuals with CVDs. An IQR increase in PM10 concentrations on lag day 0 resulted in a 4.47% (2.13–6.85) increase in hospitalization for depression in patients with CVDs, but only in a 1.47% increase (0.17–2.79) in patients without comorbid conditions (Fig. 4, Supplementary Table 3). We did not find evidence of effect modification by diabetes or COPD in any lag structures (p > 0.05 in all cases). When the effect of PM2.5 and PM10 was evaluated within SO2, NO2 and CO level quartiles, we observed largest percentage changes of PM2.5 and PM10 for the first quantile of SO2, NO2 and CO. Within SO2, the positive association between the concentration of PM2.5 and the number of hospital admissions for depression in the two pollutants model consistent with the single pollutant model. The effect estimates for an IQR change in PM2.5 from the two pollutants model with interaction are higher than the single pollutant model and the two pollutants model without interaction. The strongest effect was observed on the lag 5 day in the two pollutants model without interaction, which is the same as the result in the single pollutant model. Although the strongest effect was on the lag 0 day in the two pollutants model with interaction, it is still consistent the double peak phenomenon in the results of single pollutant model. Similarly, within NO2/CO, the concentration of PM2.5 positively associated with the hospitalizations for depression in the two pollutants model. Within NO2, the strongest effect was observed on the lag 4 day (3.40; 1.16–5.69) in the two pollutants model without interaction, while in the two pollutants model with interaction the strongest effect was on the lag 1 day (5.83; 1.57–10.27). Within CO, the strongest effect was on the lag 5 day (5.14; 2.67–7.68) in the two pollutants model without interaction, while on the lag 2 day (6.73; 3.04–10.55) in the two pollutants model with interaction (Table 4). Within SO2, NO2 or CO, the PM10 also positively associated with the hospitalizations for depression in the two pollutants model. In the two pollutants model without interaction, the strongest effect was on the lag 2 day within SO2 (3.51; 0.84–6.26), on the lag 2 day within NO2, (3.81; 0.74–6.98) and on the lag 4 day within CO (4.65; 1.82–7.55). In the two pollutants model with interaction, the strongest effect was on the lag
The sex differences in associations of PM2.5 with hospitalization for depression on lag day 0 were on the threshold of significance (p = 0.057). An IQR increase in PM2.5 concentrations on lag day 0 led to a 3.97% (2.06–5.91) increase in admissions in females, compared with a 0.74% (−1.92–3.47) increase in males. There were no significant differences between females and males on other lag days for PM2.5 exposure, and no differences by sex were found for the effects of PM10 exposure on hospitalization for depression (Fig. 2, Supplementary Table 1). Fig. 3 shows the associations between PM concentrations and hospitalizations for depression, stratified by age. The associations for both PM2.5 and PM10 were strongest in the elderly. An IQR increase in PM2.5 and PM10 concentrations on lag day 0 resulted in 9.23% (5.09–13.53) and 6.35% (3.31–9.49) increases, respectively, in admissions among the elderly, compared with 1.49% (−0.25–3.27) and 1.32% (0.04–2.61) increases in adults aged 18–65. The differences between the two groups were statistically significant (PM2.5: p = 0.001; PM10: p = 0.003). Associations between PM concentrations and admissions were also observed in adolescents, where an IQR increase in PM2.5 and PM10
Fig. 2. Percentage change with 95% CI in depression admissions associated with an interquartile range increase in PM2.5 and PM10 concentrations stratified by gender. ⁎p < 0.1.
118
Environment International 114 (2018) 115–122
F. Wang et al.
Fig. 3. Percentage change with 95% CI in depression admissions associated with an interquartile range increase in PM2.5 and PM10 concentrations stratified by age. ⁎p < 0.1 ⁎⁎p < 0.05 ⁎⁎⁎ p < 0.01.
Fig. 4. Percentage change with 95% CI in depression admissions associated with an interquartile range increase in PM2.5 and PM10 concentrations stratified by comorbidities. ⁎p < 0.05.
symptoms among the elderly (Lim et al., 2012). These findings support the hypothesis that short-term PM exposure is linked to depressive symptoms, but the mechanism has yet to be determined. One possible explanation could be neuroinflammation, caused by air pollutants reaching the brain in people who are highly exposed (CalderonGarciduenas et al., 2008); depressive symptoms have been observed to be associated with inflammatory processes (Raison et al., 2006). PM10 has been identified as a strong inflammatory agent (CalderonGarciduenas et al., 2008), and PM2.5 is thought to exert greater toxicity than PM10 because smaller particles can deposit in deeper areas of the lungs, leading to higher concentrations of adsorbed or condensed toxicants in the body (Sarnat et al., 2016). Exposure to PM2.5 and PM10 could therefore increase the risk of depressive symptoms via neuroinflammatory processes. Notably, in our study, the effect of PM2.5 increased on lag day 0 and lag day 5, while PM10 effects were highest on lag day 0. The continued effects of PM2.5 on lag day 5 may reflect impaired function related to internal PM accumulation. The differences observed for the effects of PM2.5 and PM10 may therefore have resulted
0 day within SO2 (3.73; 0.97–6.56), on the lag 2 day within NO2, (5.59; −0.09-11.59) and on the lag 2 day within CO (6.75; 2.32–11.38) (Table 5).
4. Discussion This study investigated the effect of short-term exposure to ambient PM (especially PM2.5) on hospitalizations for depression in 26 Chinese cities and found that short-term exposure to increased levels of PM was significantly associated with increased risk of hospital admission for depression. The maximum effects were observed on lag day 0 and lag day 5 for PM2.5 and on lag day 0 for PM10. These findings are consistent with those reported by previous studies on PM and depression. Szyszkowicz (2007) and Szyszkowicz et al. (2009) reported that short-term increases in ambient air pollution were associated with increased emergency department visits for depression in Canada. Similarly, a Korean study found that the 3-day moving average (lag 0–2) concentration of PM10 was associated with depressive 119
Environment International 114 (2018) 115–122
F. Wang et al.
Table 4 Percentage changes with per IQR change in PM2.5 (95% Confidence Interval) for two pollutants models without and with interaction (within the first, second and third quantiles of the two-pollutants). PM2.5 Co-pollutant
Without interaction
With interaction First quartile
Second quantile
Third quantile
3.38 3.31 3.33 3.07 3.24 2.95 1.34 0.55
3.32 3.25 3.30 2.99 3.21 2.97 1.40 0.56
SO2 Lag 0 Lag 1 Lag 2 Lag 3 Lag 4 Lag 5 Lag 6 Lag 7
2.70 2.55 2.98 2.21 2.93 3.23 2.18 0.65
(0.84,4.59) (0.63,4.51) (0.74,5.27) (0.37,4.08) (1.08,4.81) (1.40,5.10) (0.42,3.97) (−1.13,2.45)
3.41 3.35 3.35 3.10 3.25 2.94 1.31 0.55
NO2 Lag 0 Lag 1 Lag 2 Lag 3 Lag 4 Lag 5 Lag 6 Lag 7
1.27 1.57 3.24 1.65 3.40 2.59 0.85 0.24
(−0.93,3.52) (−0.71,3.90) (0.70,5.84) (−0.56,3.92) (1.16,5.69) (0.38,4.85) (−1.27,3.00) (−1.9,2.43)
2.90 (−1.10,7.06) 5.83 (1.57,10.27) 5.72 (0.84,10.84) 4.46 (0.33,8.75) 3.33 (−0.78,7.61) 1.41 (−2.56,5.55) −1.47 (−5.22,2.42) −0.78 (−4.61,3.21)
2.85 (−1.07,6.92) 5.7 (1.53,10.04) 5.64 (0.87,10.63) 4.37 (0.33,8.57) 3.33 (−0.69,7.51) 1.45 (−2.44,5.49) −1.40 (−5.08,2.41) −0.75 (−4.5,3.15)
2.78 (−1.02,6.73) 5.52 (1.48,9.72) 5.53 (0.92,10.35) 4.25 (0.34,8.32) 3.34 (−0.57,7.39) 1.5 (−2.28,5.42) −1.31 (−4.89,2.4) −0.71 (−4.36,3.08)
CO Lag Lag Lag Lag Lag Lag Lag Lag
2.39 (−0.01,4.85) 3.45 (0.98,5.98) 4.14 (1.70,6.63) 3.02 (0.62,5.48) 4.64 (2.16,7.19) 5.14 (2.67,7.68) 1.81 (−0.52,4.19) −0.16 (−2.49,2.23)
3.23 (0.20,6.35) 6.25 (2.94,9.66) 6.73 (3.04,10.55) 5.66 (2.45,8.96) 5.61 (2.37,8.96) 6.34 (3.07,9.72) 1.93 (−0.96,4.92) −0.04 (−2.93,2.94)
3.20 (0.20,6.29) 6.16 (2.90,9.53) 6.64 (3.02,10.39) 5.58 (2.42,8.84) 5.58 (2.38,8.89) 6.31 (3.08,9.64) 1.93 (−0.94,4.88) −0.04 (−2.9,2.9)
3.17 (0.21,6.21) 6.04 (2.83,9.34) 6.52 (3.00,10.17) 5.47 (2.36,8.67) 5.54 (2.39,8.79) 6.26 (3.08,9.52) 1.93 (−0.9,4.83) −0.05 (−2.86,2.85)
0 1 2 3 4 5 6 7
(0.99,5.88) (0.9,5.86) (0.49,6.29) (0.72,5.53) (0.88,5.69) (0.60,5.34) (−0.96,3.63) (−1.78,2.93)
(1.00,5.80) (0.90,5.78) (0.52,6.23) (0.73,5.46) (0.90,5.64) (0.64,5.32) (−0.90,3.62) (−1.74,2.89)
(1.02,5.66) (0.91,5.64) (0.58,6.10) (0.73,5.31) (0.95,5.53) (0.73,5.27) (−0.77,3.62) (−1.66,2.83)
Table 5 Percentage changes with per IQR change in PM10 (95% Confidence Interval) for two pollutants models without and with interaction (within the first, second and third quantiles of the two-pollutants). PM10 Co-pollutant
Without interaction
With interaction First quartile
Second quantile
Third quantile
SO2 Lag 0 Lag 1 Lag 2 Lag 3 Lag 4 Lag 5 Lag 6 Lag 7
3.25 (1.05,5.5) 3.23 (0.99,5.51) 3.51 (0.84,6.26) 2.15 (−0.05,4.39) 1.91 (−0.22,4.1) 1.94 (−0.19,4.12) 0.86 (−1.27,3.03) −0.43 (−2.55,1.73)
3.73 (0.97,6.56) 3.57 (0.8,6.41) 3.31 (−0.01,6.74) 2.74 (0.03,5.53) 1.92 (−0.7,4.61) 1.07 (−1.49,3.71) −0.49 (−3.02,2.12) −0.74 (−3.35,1.93)
3.71 (0.99,6.51) 3.56 (0.82,6.37) 3.31 (0.03,6.71) 2.72 (0.04,5.48) 1.92 (−0.67,4.58) 1.1 (−1.45,3.71) −0.45 (−2.96,2.13) −0.73 (−3.32,1.92)
3.68 (1.01,6.43) 3.54 (0.85,6.3) 3.33 (0.1,6.65) 2.69 (0.05,5.39) 1.92 (−0.62,4.53) 1.15 (−1.36,3.71) −0.37 (−2.85,2.16) −0.72 (−3.25,1.88)
NO2 Lag 0 Lag 1 Lag 2 Lag 3 Lag 4 Lag 5 Lag 6 Lag 7
1.59 (−1.04,4.29) 2.23 (−0.47,5) 3.81 (0.74,6.98) 1.24 (−1.42,3.97) 1.79 (−0.76,4.41) 0.42 (−2.11,3.02) −1.53 (−4.05,1.07) −1.37 (−3.9,1.24)
2.57 (−1.82,7.16) 5.07 (0.48,9.87) 5.59 (−0.09,11.59) 2.95 (−1.44,7.54) 0.69 (−3.47,5.03) −2.71 (−6.73,1.48) −6.16 (−10.14,-2) −4.11 (−8.16,0.12)
2.55 (−1.79,7.08) 5.01 (0.48,9.74) 5.55 (−0.05,11.46) 2.91 (−1.42,7.44) 0.71 (−3.39,4.99) −2.64 (−6.61,1.49) −6.06 (−9.99,−1.95) −4.05 (−8.05,0.12)
2.52 (−1.73,6.96) 4.93 (0.49,9.56) 5.5 (0.02,11.28) 2.86 (−1.39,7.3) 0.75 (−3.28,4.94) −2.55 (−6.45,1.51) −5.93 (−9.8,−1.89) −3.97 (−7.9,0.13)
CO Lag Lag Lag Lag Lag Lag Lag Lag
2.76 (0.15,5.44) 3.93 (1.27,6.65) 4.65 (1.82,7.55) 2.42 (−0.09,4.99) 2.43 (0,4.93) 2.46 (0.03,4.95) −0.09 (−2.5,2.38) −1.22 (−3.63,1.25)
3.51 (0.07,7.07) 6.16 (2.53,9.92) 6.75 (2.32,11.38) 4.38 (1,7.87) 2.62 (−0.6,5.94) 2.5 (−0.72,5.83) −1.34 (−4.43,1.84) −1.93 (−5.03,1.27)
3.5 (0.09,7.03) 6.11 (2.51,9.83) 6.71 (2.33,11.28) 4.34 (0.99,7.79) 2.62 (−0.57,5.9) 2.5 (−0.69,5.79) −1.31 (−4.37,1.84) −1.91 (−4.98,1.26)
3.48 (0.11,6.96) 6.04 (2.5,9.71) 6.65 (2.35,11.14) 4.28 (0.98,7.69) 2.61 (−0.54,5.85) 2.5 (−0.65,5.75) −1.28 (−4.3,1.84) −1.89 (−4.93,1.24)
0 1 2 3 4 5 6 7
120
Environment International 114 (2018) 115–122
F. Wang et al.
observed effect modification by age and comorbidities, with significantly stronger associations in the elderly and in CVD patients. Future work should focus on elucidating the underlying biological mechanisms.
from deeper respiratory deposition of PM2.5, leading to a longer period of impairment. In this study, the effect of PM2.5 on lag day 0 showed a trend to be more pronounced in females, while other estimates did not show significant sex difference. The number of females in our sample was larger, which is in line with the sex-specific prevalence of major depression in China (Gu et al., 2013). Higher effects of air pollution on respiratory health have been observed among females than among males, and the possible sources of difference in air pollution response include life stage, work-related coexposures, hormonal status, and other factors (Clougherty, 2010). However, few studies have explored the effects of PM on depression in the context of sex-specific differences. Szyszkowicz et al. (2016) reported that exposure to PM2.5 was associated with increased risk in males one day after the exposure, while no such association was observed in females. This was not with the case in our study, and the reasons could be differences in the data sources, composition and relative concentrations of pollutants, or variations in patient characteristics. Further study is required to determine the effects of sex on the association between PM levels and hospitalizations for depression. Depressive symptoms are highly prevalent in the elderly (Thielke et al., 2010), yet few studies to date have looked at the association between PM2.5 exposure and depression in the elderly. Kioumourtzoglou et al. (2017) reported a possible association between long-term PM2.5 exposure and the onset of depression among older women. Another cohort study found that increases in PM10 were significantly associated with increased depressive symptoms in the elderly (Lim et al., 2012). In our study, the effects of both PM2.5 and PM10 appeared more pronounced in the elderly. The findings suggest that the elderly should get more protection from the air pollution as they might be more susceptible to ambient PM-related depression. Depressive disorder is frequently comorbid with other conditions, such as CVD (Musselman et al., 1998) and diabetes (Stuart and Baune, 2012). We found that CVD patients, when exposed to higher levels of PM, had a significantly higher risk of hospitalization for depression than non-CVD patients. This finding is consistent with that of a study on ambient PM that reported an increased risk of suicide among subjects with pre-existing CVD (Kim et al., 2010). Exposure to ambient PM has been shown to adversely affect vascular endothelial function, leading to increased plasma viscosity, a risk factor of blood clotting, thrombosis, and ischemic stroke (Gurgueira et al., 2002; Lucking et al., 2011; Liu et al., 2017). CVD has been linked to depression via systemic inflammation (Ridker et al., 1997). Air pollutants could therefore trigger depressive symptoms, especially among CVD patients. CVD can be both a mediator and a modifier and future studies are needed to figure it out. A previous study found an association between PM and depression in patients with diabetes or COPD (Kim et al., 2016), but we did not, perhaps owing to the smaller sample in our study. This study has some limitations. First, our data could not discriminate between emergent hospitalization and scheduled admission, making it difficult to determine an association between PM exposure and symptoms of depression. Second, using citywide air-pollutant mass concentrations as proxies for personal exposure may result in measurement error, underestimating the effects of the pollutants (Goldman et al., 2011). In addition, potential exposure and outcome misclassification should be considered, as the ICD-10 codes were the only data we could use to define patients with depression. In China, air pollution levels are much higher than in the US and Canada, the sources of most studies on air pollution and depression, resulting in limited comparability and generalizability of the findings. Moreover, we mainly discussed the findings of a linear model, although the non-linear findings have been added to the supplementary material; more research on non-linear associations is needed. In summary, our findings indicate that short-term elevations in the levels of PM2.5 and PM10 are significantly associated with an increased risk of hospitalization for depression in 26 Chinese cities. We also
Appendix A. Supplementary data Supplementary data to this article can be found online at https:// doi.org/10.1016/j.envint.2018.02.012. References Altman, D.G., Bland, J.M., 2003. Interaction revisited: the difference between two estimates. BMJ 326, 219. Bell, M.L., Samet, J.M., Dominici, F., 2004. Time-series studies of particulate matter. Annu. Rev. Public Health 25, 247–280. Block, M.L., Wu, X., Pei, Z., et al., 2004. Nanometer size diesel exhaust particles are selectively toxic to dopaminergic neurons: the role of microglia, phagocytosis, and NADPH oxidase. FASEB J. 18, 1618–1620. Brites, D., Fernandes, A., 2015. Neuroinflammation and depression: microglia activation, extracellular microvesicles and microRNA dysregulation. Front. Cell. Neurosci. 9, 476. Calderon-Garciduenas, L., Solt, A.C., Henriquez-Roldan, C., et al., 2008. Long-term air pollution exposure is associated with neuroinflammation, an altered innate immune response, disruption of the blood–brain barrier, ultrafine particulate deposition, and accumulation of amyloid beta-42 and alpha-synuclein in children and young adults. Toxicol. Pathol. 36, 289–310. Campbell, A., Oldham, M., Becaria, A., et al., 2005. Particulate matter in polluted air may increase biomarkers of inflammation in mouse brain. Neurotoxicology 26, 133–140. Carbajal-Arroyo, L., Miranda-Soberanis, V., Medina-Ramón, M., Rojas-Bracho, L., Tzintzun, G., Solís-Gutiérrez, P., Méndez-Ramírez, I., Hurtado-Díaz, M., Schwartz, J., Romieu, I., 2011. Effect of PM(10) and O(3) on infant mortality among residents in the Mexico city metropolitan area: a case-crossover analysis, 1997–2005. J. Epidemiol. Community Health 65, 715–721. Carracedo-Martinez, E., Taracido, M., Tobias, A., Saez, M., Figueiras, A., 2010. Casecrossover analysis of air pollution health effects: a systematic review of methodology and application. Environ. Health Perspect. 118, 1173–1182. Center for Disease Control and Prevention, 2010. Current depression among adults – United States, 2006 and 2008. Morb. Mortal. Wkly Rep. 59, 1229–1235. Clougherty, J.E., 2010. A growing role for gender analysis in air pollution epidemiology. Environ. Health Perspect. 118 (2), 167–176. Cuijpers, P., Beekman, A.T., Reynolds, C.F. rd, 2012. Preventing depression: a global priority. JAMA 307 (10), 1033–1034. Goldberg, M.S., Gasparrini, A., Armstrong, B., Valois, M.F., 2011. The short-term influence of temperature on daily mortality in the temperate climate of Montreal, Canada. Environ. Res. 111, 853–860. Goldman, G.T., Mulholland, J.A., Russell, A.G., Strickland, M.J., Klein, M., Waller, L.A., et al., 2011. Impact of exposure measurement error in air pollution epidemiology: effect of error type in time-series studies. Environ. Health 10, 10–61. Gu, L., Xie, J., Long, J., Chen, Q., Chen, Q., et al., 2013. Epidemiology of major depressive disorder in mainland China: a systematic review. PLoS One 8 (6), e65356. Gurgueira, S.A., Lawrence, J., Coull, B., Murthy, G.G., Gonzalez-Flecha, B., 2002. Rapid increases in the steady-state concentration of reactive oxygen species in the lungs and heart after particulate air pollution inhalation. Environ. Health Perspect. 110, 749–755. Hasler, G., 2010. Pathophysiology of depression: do we have any solid evidence of interest to clinicians? World Psychiatry Official Journal of the World Psychiatric Association 9 (3), 155–161. Hu, T.W., He, Y., Zhang, M., Chen, N., 2007. Economic costs of depression in China. Soc. Psychiatry Psychiatr. Epidemiol. 42 (2), 110–116. Janes, H., Sheppard, L., Lumley, T., 2005. Case-crossover analyses of air pollution exposure data: referent selection strategies and their implications for bias. Epidemiology 16 (6), 717–726. Kan, H., Chen, R., Tong, S., 2012. Ambient air pollution, climate change, and population health in China. Environ. Int. 42 (1), 10. Kim, C., Jung, S.H., Kang, D.R., Kim, H.C., Moon, K.T., Hur, N.W., Shin, D.C., Suh, I., 2010. Ambient particulate matter as a risk factor for suicide. Am. J. Psychiatry 167 (9), 1100–1107. Kim, K.N., Lim, Y.H., Bae, H.J., Kim, M., Jung, K., Hong, Y.C., 2016. Long-term fine particulate matter exposure and major depressive disorder in a community-based urban cohort. Environ. Health Perspect. 124 (10), 1547–1553. Kioumourtzoglou, M.A., Power, M.C., Hart, J.E., et al., 2017. The association between air pollution and onset of depression among middle-aged and older women. Am. J. Epidemiol. 27 (10), 1. Kleine-Budde, K., Müller, R., Kawohl, W., Bramesfeld, A., Moock, J., Rössler, W., 2013. The cost of depression — a cost analysis from a large database. J. Affect. Disord. 147 (1–3), 137–143. Lim, Y.H., Kim, H., Kim, J.H., Bae, S., Park, H.Y., Hong, Y.C., 2012. Air pollution and symptoms of depression in elderly adults. Environ. Health Perspect. 120 (7), 1023–1028. Liu, H., Tian, Y., Xu, Y., Zhang, J., 2017. Ambient particulate matter concentrations and hospitalization for stroke in 26 Chinese cities: a case-crossover study. Stroke 48 (8),
121
Environment International 114 (2018) 115–122
F. Wang et al.
Sarnat, S.E., Chang, H.H., Weber, R.J., 2016. Ambient PM2.5 and health: does PM2.5 oxidative potential play a role? Am. J. Respir. Crit. Care Med. 194, 530–531. Stuart, M.J., Baune, B.T., 2012. Depression and type 2 diabetes: inflammatory mechanisms of a psychoneuroendocrine co-morbidity. Neurosci. Biobehav. Rev. 36 (1), 658–676. Szyszkowicz, M., 2007. Air pollution and emergency department visits for depression in Edmonton, Canada. Int. J. Occup. Med. Environ. Health 20 (3), 241–245. Szyszkowicz, M., Rowe, B.H., Colman, I., 2009. Air pollution and daily emergency department visits for depression. Int. J. Occup. Med. Environ. Health 22 (4), 355–362. Szyszkowicz, et al., 2016. Air pollution and emergency department visits for depression: a multicity case-crossover study. Environmental Health Insights 10, 155–161. Thielke, S.M., Diehr, P., Unutzer, J., 2010. Prevalence, incidence, and persistence of major depressive symptoms in the cardiovascular health study. Aging Ment. Health 14 (2), 168–176. Veronesi, B., Makwana, O., Pooler, M., Chen, L.C., 2005. Effects of subchronic exposures to concentrated ambient particles. VII. Degeneration of dopaminergic neurons in Apo E−/− mice. Inhal. Toxicol. 17, 235–241. Winquist, A., Kirrane, E., Klein, M., Strickland, M., Darrow, L.A., Sarnat, S.E., Gass, K., Mulholland, J., Russell, A., Tolbert, P., 2014. Joint effects of ambient air pollutants on pediatric asthma emergency department visits in Atlanta, 1998–2004. Epidemiology 25 (5), 666–673.
2052–2059. Lucking, A.J., Lundback, M., Barath, S.L., Mills, N.L., Sidhu, M.K., Langrish, J.P., et al., 2011. Particle traps prevent adverse vascular and prothrombotic effects of diesel engine exhaust inhalation in men. Circulation 123, 1721–1728. Moscicki, E.K., 2001. Epidemiology of completed and attempted suicide: toward a framework for prevention. Clin. Neurosci. Res. 1 (5), 310–323. Musselman, D.L., Evans, D.L., Nemeroff, C.B., 1998. The relationship of depression to cardiovascular disease: epidemiology, biology, and treatment. Arch. Gen. Psychiatry 55 (7), 580–592. Raison, C.L., Capuron, L., Miller, A.H., 2006. Cytokines sing the blues: inflammation and the pathogenesis of depression. Trends Immunol. 27 (1), 24–31. Rashid, T., Heider, I., 2008. Life events and depression. Ann. Punjab Med. Coll. 2 (1), 128–149. Ridker, P.M., Cushman, M., Stampfer, M.J., Tracy, R.P., Hennekens, C.H., 1997. Inflammation, aspirin, and the risk of cardiovascular disease in apparently healthy men. N. Engl. J. Med. 336 (14), 973–979. Rustad, J.K., Musselman, D.L., Nemeroff, C.B., 2011. The relationship of depression and diabetes: pathophysiological and treatment implications. Psychoneuroendocrinology 36 (9), 1276–1286. Saravane, D., Feve, B., Frances, Y., et al., 2009. Drawing up guidelines for the attendance of physical health of patients with severe mental illness. L'Encéphale 35 (4), 330–339.
122
Contents lists available at ScienceDirect
Environment International journal homepage: www.elsevier.com/locate/envint
Ambient concentrations of particulate matter and hospitalization for depression in 26 Chinese cities: A case-crossover study
T
Feng Wanga,b,c,1, Hui Liud,1, Hui Lia,b,c,1, Jiajia Liua,b,c, Xiaojie Guoa,b,c, Jie Yuane, Yonghua Hud, ⁎ ⁎⁎ Jing Wangd, , Lin Lua,b,c, a
Peking University Sixth Hospital/Institute of Mental Health, 100191 Beijing, China National Clinical Research Center for Mental Disorders & Key Laboratory of Mental Health, Ministry of Health (Peking University), 100191 Beijing, China c Key Laboratory of Mental Health, Ministry of Health, Peking University, 100191 Beijing, China d Peking University Medical Informatics Center, Peking University, 100191 Beijing, China e North China University of Science and Technology, 063000, Hebei Province, China b
A R T I C L E I N F O
A B S T R A C T
Handling Editor: Yong Guan Zhu
Objective: Air pollution with high ambient concentrations of particulate matter (PM) has been frequently reported in China. However, no Chinese study has looked into the short-term effect of PM on hospitalization for depression. We used a time-stratified case-crossover design to identify possible links between ambient PM levels and hospital admissions for depression in 26 Chinese cities. Methods: Electronic hospitalization summary reports (January 1, 2014–December 31, 2015) were used to identify hospital admissions related to depression. Conditional logistic regression was applied to determine the association between PM levels and hospitalizations for depression, with stratification by sex, age, and comorbidities. Results: Both PM2.5 and PM10 levels were positively associated with the number of hospital admissions for depression. The strongest effect was observed on the day of exposure (lag day 0) for PM10, with an interquartile range increase in PM10 associated with a 3.55% (95% confidence interval: 1.69–5.45) increase in admissions for depression. For PM2.5, the risks of hospitalization peaked on lag day 0 (2.92; 1.37–4.50) and lag day 5 (3.65; 2.09–5.24). The elderly (> 65) were more sensitive to PM2.5 exposure (9.23; 5.09–13.53) and PM10 exposure (6.35; 3.31–9.49) on lag day 0, and patients with cardiovascular disease were likely to be hospitalized for depression following exposure to high levels of PM10 (4.47; 2.13–6.85). Conclusions: Short-term elevations in PM may increase the risk of hospitalization for depression, particularly in the elderly and in patients with cardiovascular disease.
Keywords: Depression Particulate matter Hospitalization China
1. Introduction Depression is a serious mental disorder that profoundly affects individual quality of life and can impair daily functioning, even leading to suicide (Moscicki, 2001). The prevalence of depression the US has been estimated at 9.0% (CDC, 2010). The estimated prevalence of major depressive disorder in China is 1.6% (2.1% for females and 1.3% for males) (Gu et al., 2013). After the first depressive episode, more than half of patients develop a recurrent or chronic disorder, causing personal suffering and economic problems that are a substantial burden to society (Kleine-Budde et al., 2013). In China, the total cost of
depression in 2002 has been estimated at 51,370 million RMB ($6264 million USD) (Hu et al., 2007). Some researchers suggest that prevention is critical in reducing this burden (Cuijpers et al., 2012). Understanding the factors involved in depression is therefore of practical importance. The onset of depression can be triggered by various factors. Apart from genetic susceptibility, environmental risk factors may contribute to the onset or aggravation of depression. Contributing factors identified in earlier studies include stressful events, such as losing jobs and loved ones (Rashid and Heider, 2008), and physical conditions, such as diabetes (Rustad et al., 2011) and stroke (Saravane et al., 2009). Air
⁎
Corresponding author. Correspondence to: L. Lu, Peking University Sixth Hospital/Institute of Mental Health, 100191 Beijing, China. E-mail addresses: [email protected] (F. Wang), [email protected] (H. Liu), [email protected] (H. Li), [email protected] (J. Liu), [email protected] (X. Guo), [email protected] (J. Yuan), [email protected] (Y. Hu), [email protected] (J. Wang), [email protected] (L. Lu). 1 These authors contributed equally to this work. ⁎⁎
https://doi.org/10.1016/j.envint.2018.02.012 Received 28 September 2017; Received in revised form 7 February 2018; Accepted 8 February 2018 0160-4120/ © 2018 Published by Elsevier Ltd.
Environment International 114 (2018) 115–122
F. Wang et al.
matter < 10 μm in aerodynamic diameter) between January 1, 2014 and December 31, 2015. The daily (24-hour) mean concentrations of pollutants averaged across all the stations in a given city were used as the reading for that city on that day. In addition, the daily (24-hour) average temperature and relative humidity were obtained from the Chinese Meteorological Bureau, allowing for adjustment for weather conditions.
pollution has also been shown to be involved, which is a heterogeneous mixture of both gaseous pollutants and particulate matter (PM), and especially fine PM (PM2.5, particles with an aerodynamic diameter ≤ 2.5 μm), Animal models have shown evidence of the neuropathological effects of exposure to PM (Campbell et al., 2005; Veronesi et al., 2005), and it has been suggested that high concentrations of inhaled air pollutants may reach the brain and cause neuroinflammation (Calderon-Garciduenas et al., 2008). Chronic inflammation has been linked to depression (Brites and Fernandes, 2015). Air pollution may contribute to a depressive mood by inducing dopaminergic neurotoxicity, possibly owing to oxidative stress (Block et al., 2004), as a decrease in dopamine in the central nervous system is one of the underlying pathophysiological mechanisms of depression (Hasler, 2010). Several studies found a relationship between depression and exposure to PM. Studies conducted in Canada indicated that emergency department visits for depression were associated with exposure to ambient air pollution (Szyszkowicz, 2007; Szyszkowicz et al., 2009, 2016). Lim et al. (2012) reported that increases in PM10, nitrogen dioxide (NO2), and ozone levels may increase depressive symptoms among the elderly. One prospective cohort study reported a possible association between both long-term PM2.5 exposure and depression onset and incident antidepressant use (Kioumourtzoglou et al., 2017). Kim et al. (2016) reported that long-term PM2.5 exposure may increase the risk of major depressive disorder. However, to our knowledge, no published study has investigated the association between PM exposure and depression in China, one of the most polluted countries in the world (Kan et al., 2012). With rapid industrialization and increased energy consumption over the past several decades, air pollution, and especially PM pollution, has become a severe environmental problem in China. Given the public health and economic burden of depression, it is important to determine whether specific air pollutants are linked to an increased risk of depression hospitalization; a better understanding of the underlying mechanisms may lead to evidence-based policymaking for primary prevention and specific interventions. This study investigated the association between ambient concentrations of PM and the number of depression-related hospital admissions in 26 Chinese cities between 2014 and 2015, stratifying by sex, age, and comorbidities.
2.3. Study design To investigate a possible association between ambient PM concentrations and hospitalization for depression, we adopted a timestratified case-crossover design where cases were used as their own controls (Carracedo-Martinez et al., 2010). With this approach, we were able to adjust for time-invariant characteristics such as age and sex, to ensure unbiased estimates from conditional logistic regression and to avoid time-trend bias (Janes et al., 2005). For each case, the patient's exposure to ambient PM on the day of hospital admission was compared with the exposure on three or four reference days. The reference days were the same days of the week within the same month and year of admission. Using this method, we controlled for the influence of day of week, seasonal and long-term trends, and potential individual-level risk factors (such as sex and genetics). 2.4. Statistical analysis Pooled analyses were applied in this study; observations from all included cities were combined, and each city was given a special indicator in the dataset. General and clinical characteristics of all 19,646 cases were described, and Spearman's correlation analysis was applied to assess the associations between exposure variables. We then used conditional logistic regression to examine associations between PM concentrations and hospitalizations for depression for each special lag day. Distributed lag non-linear models (DLNM) with three degrees of freedom in the natural cubic splines and a maximum lag of 3 days were used to adjust for the delayed and non-linear effects of temperature and humidity (Goldberg et al., 2011). To control for the meteorological effects on health in different areas, we added interactions between meteorology and the cities to the models. Public holidays were included in the models. The results are described as the percentage of PM concentration increase and the 95% confidence intervals (CIs) in daily admissions for depression per interquartile range (IQR = 75th percentile–25th percentile of air pollutants). We also analyzed the exposureresponse association between concentrations of PM2.5 and hospitalizations for depression. The DLNM was used to determine the non-linear delayed relationships between exposure and hospitalization for depression. The models included various lag structures—from the day of hospitalization (lag day 0) up to seven lag days (lag day 7)—to examine the temporal associations of PM concentrations and depression. In addition, we examined the associations with 3-day (lag days 0–2) and 6-day (lag days 0–5) moving mean PM concentrations, to avoid underestimating the effect of pollutants measured by single-day lag models (Bell et al., 2004). In order to explore the impact of PM on depression hospitalization varies across levels of other pollutants, including sulfur dioxide (SO2), NO2, and carbon monoxide (CO), two-pollutant models with first order interaction were added in the sensitivity analysis (CarbajalArroyo et al., 2011; Winquist et al., 2014). To determine whether the associations differed by sex and age (< 18, 18–65 and ≥65 years), stratified analyses using a Z-test were applied (Altman and Bland, 2003). To examine whether PM exerted different effects on patients with comorbid chronic diseases, we singled out patients who, in addition to a primary diagnosis of depression, had cardiovascular diseases (CVDs, ICD-10 codes I10.x–I15.x, I20.x–I25.x, or I60.x–I69.x), diabetes (E10.x–E14.x), and chronic obstructive pulmonary disease (COPD, J40–J44.x). All analyses were conducted using
2. Materials and methods 2.1. Study population Data on hospital admissions were collected from the electronic hospitalization summary reports of tertiary A hospitals in 26 Chinese cities (Supplementary Fig. 1), which recorded basic demographics (sex and age), dates of admission and discharge, hospitalization and discharge diagnoses in Chinese and their corresponding codes in the 10th revision of the International Classification of Diseases (ICD-10), treatments (mainly surgical information), discharge status (survival, drug allergy, and infection), and hospitalization expenses. We used ICD-10 codes to identify depression-related admissions between January 1, 2014 and December 31, 2015. These codes included F32 (mild depressive episode), F33 (recurrent depressive disorder), F34.1 (dysthymia), and F41.2 (mixed anxiety and depressive disorder). In total, 19,646 hospital admissions were depression-related. 2.2. Air pollution and meteorological data Data on air pollution were obtained from the National Air Pollution Monitoring System, which is run by the Chinese Ministry of Environmental Protection. The system fulfills the quality assurance and quality control mandates of the Chinese government through its ambient air-monitoring stations. These stations, ranging in number from 4 to 15 per city, provide hourly air pollution data to the system. We collected records on the levels of PM2.5 and PM10 (particulate 116
Environment International 114 (2018) 115–122
F. Wang et al.
Table 1 Demographic characteristics of depression admissions. Variable
Depression individuals
Total Gender Male (%) Female (%) Age (year) (mean ± SD) < 18 (%) 18–64 (%) ≥65 (%) Comorbidities Cardiovascular disease (%) Diabetes (%) COPD (%)
19,646 6493 (33.1) 13,153 (66.9) 46 ± 16.8 1089 (5.5) 15,716 (80.0) 2841 (14.5) 4687 (23.9) 1287 (6.6) 321 (1.6)
R (V.3.2.2, R Development Core Team). All statistical tests were twosided, with a significance level of 5%. 3. Results A total of 19,646 hospital admissions were identified to be depression-related (patients' mean age: 46.0 ± 16.8 years; 33.1% males and 66.9% females). Among the cases, 5.5% were adolescents (mean age: 15.2 ± 1.7 years) and 14.5% were elderly individuals (mean age: 71.4 ± 5.3 years). Comorbid CVDs were associated with 4687 cases; 1287 cases had diabetes; 321 had COPD (Table 1). 3.1. Air pollution and meteorological variables Table 2 displays descriptive statistics for air pollutants and meteorological measurements in all cities combined during the study period. The mean temperature (SD) was 14.5 °C (10.9 °C), and the relative humidity (SD) was 69.2% (33.2%). Concentrations of PM2.5 were positively associated with concentrations of PM10 (r = 0.87, p < 0.001), and both values were positively associated with the concentrations of CO, NO2, and SO2 (r = 0.54–0.68, p < 0.001). However, concentrations of air pollutants in general were negatively correlated with temperature and humidity. 3.2. Associations between air pollution and hospitalizations for depression Positive associations between PM concentrations and hospitalizations for depression were observed (Fig. 1). Table 3 shows increases in depression hospitalizations associated with an IQR increase in PM2.5 and PM10 concentrations for different lag structures. The associations between PM and hospitalization for depressive disorders were observed on lag day 0, and remained significant for 5 days after exposure. PM2.5 exerted the highest effects on lag day 0 (2.92% increase; 95% CI, 1.37–4.50) and lag day 5 (3.65; 2.09–5.24), whereas PM10 exerted the highest impact on lag day 0 (3.55; 1.69–5.45). DLNM analysis showed similar trends (Supplementary Fig. 1).
Fig. 1. The exposure-response curve of 6-day (lag0–5) moving average fine particulate matter (PM2.5) and PM10 concentrations (degree of freedom = 3) and ischemic stroke hospitalizations in 26 cities. Note: The X-axis is the 6-day (lag0–5) moving average PM concentrations (μg/m3). Y-axis is the predicted log (relative risk (RR)), after adjusting for temperature and relative humidity, is shown by the solid line, and the dotted lines represent the 95% CI.
Table 2 Summary statistics for air pollutants concentrations and meteorological variables. Variable
Mean ± SD
3
PM2.5 (μg/m ) PM10 (μg/m3) SO2 (μg/m3) NO2 (μg/m3) CO (mg/m3) Temperature (°C) Relative humidity (%)
63.5 ± 50.6 106.8 ± 71.9 29.6 ± 32.6 44.1 ± 19.4 1.15 ± 0.63 14.5 ± 10.9 69.2 ± 33.2
Minimum
5.1 7.4 1.9 4.5 0.14 −25.7 8
Percentile 25th
50th
75th
31.5 58.3 11.4 30.0 0.76 7.0 53
49.4 89.4 18.8 40.2 0.99 16.4 69
79.0 135.2 33.6 54.1 1.32 23.3 80
117
Maximum
IQR
897.5 977.3 316.9 175.8 8.41 35.5 97
47.5 76.9 22.2 24.1 0.56 16.3 27
Environment International 114 (2018) 115–122
F. Wang et al.
concentrations on lag day 0 corresponded with 5.64% (−0.59–12.25) and 3.1% (−1.55–7.96) increases in admissions, respectively. However, associations in adolescents did not reach statistical significance (p > 0.05), perhaps because of a smaller sample (1089) (Supplementary Table 2).
Table 3 Percentage change with 95% CI in depression admissions associated with an interquartile range increases in PM2.5 and PM10 concentration for different lag structures. Lag days
Percentage change
95% CI
p
PM2.5 Lag 0 days Lag 1 days Lag 2 days Lag 3 days Lag 4 days Lag 5 days Lag 6 days Lag 7 days Lag 0–2 days Lag 0–5 days
2.92 2.55 2.07 2.6 3.29 3.65 2.47 1.23 3.77 6.21
1.37–4.5 0.97–4.16 0.51–3.66 1.05–4.18 1.71–4.9 2.09–5.24 0.95–4.02 −0.3 to 2.78 1.84–5.74 3.85–8.63
< 0.001 0.002 0.009 < 0.001 < 0.001 0.028 0.001 0.115 < 0.001 < 0.001
PM10 Lag 0 days Lag 1 days Lag 2 days Lag 3 days Lag 4 days Lag 5 days Lag 6 days Lag 7 days Lag 0–2 days Lag 0–5 days
3.55 3.13 1.88 2.38 2.37 2.69 1.75 0.53 4.36 5.86
1.69–5.45 1.27–5.03 0.04–3.76 0.54–4.25 0.55–4.21 0.89–4.53 −0.05 to 3.6 −1.27 to 2.38 2.05–6.73 3.09–8.71
< 0.001 < 0.001 0.045 0.001 0.001 0.003 0.057 0.565 < 0.001 < 0.001
3.3. Effect modification by comorbidities Analyses revealed a significant association between exposure to air pollutants and hospitalizations for depression in individuals with CVDs. An IQR increase in PM10 concentrations on lag day 0 resulted in a 4.47% (2.13–6.85) increase in hospitalization for depression in patients with CVDs, but only in a 1.47% increase (0.17–2.79) in patients without comorbid conditions (Fig. 4, Supplementary Table 3). We did not find evidence of effect modification by diabetes or COPD in any lag structures (p > 0.05 in all cases). When the effect of PM2.5 and PM10 was evaluated within SO2, NO2 and CO level quartiles, we observed largest percentage changes of PM2.5 and PM10 for the first quantile of SO2, NO2 and CO. Within SO2, the positive association between the concentration of PM2.5 and the number of hospital admissions for depression in the two pollutants model consistent with the single pollutant model. The effect estimates for an IQR change in PM2.5 from the two pollutants model with interaction are higher than the single pollutant model and the two pollutants model without interaction. The strongest effect was observed on the lag 5 day in the two pollutants model without interaction, which is the same as the result in the single pollutant model. Although the strongest effect was on the lag 0 day in the two pollutants model with interaction, it is still consistent the double peak phenomenon in the results of single pollutant model. Similarly, within NO2/CO, the concentration of PM2.5 positively associated with the hospitalizations for depression in the two pollutants model. Within NO2, the strongest effect was observed on the lag 4 day (3.40; 1.16–5.69) in the two pollutants model without interaction, while in the two pollutants model with interaction the strongest effect was on the lag 1 day (5.83; 1.57–10.27). Within CO, the strongest effect was on the lag 5 day (5.14; 2.67–7.68) in the two pollutants model without interaction, while on the lag 2 day (6.73; 3.04–10.55) in the two pollutants model with interaction (Table 4). Within SO2, NO2 or CO, the PM10 also positively associated with the hospitalizations for depression in the two pollutants model. In the two pollutants model without interaction, the strongest effect was on the lag 2 day within SO2 (3.51; 0.84–6.26), on the lag 2 day within NO2, (3.81; 0.74–6.98) and on the lag 4 day within CO (4.65; 1.82–7.55). In the two pollutants model with interaction, the strongest effect was on the lag
The sex differences in associations of PM2.5 with hospitalization for depression on lag day 0 were on the threshold of significance (p = 0.057). An IQR increase in PM2.5 concentrations on lag day 0 led to a 3.97% (2.06–5.91) increase in admissions in females, compared with a 0.74% (−1.92–3.47) increase in males. There were no significant differences between females and males on other lag days for PM2.5 exposure, and no differences by sex were found for the effects of PM10 exposure on hospitalization for depression (Fig. 2, Supplementary Table 1). Fig. 3 shows the associations between PM concentrations and hospitalizations for depression, stratified by age. The associations for both PM2.5 and PM10 were strongest in the elderly. An IQR increase in PM2.5 and PM10 concentrations on lag day 0 resulted in 9.23% (5.09–13.53) and 6.35% (3.31–9.49) increases, respectively, in admissions among the elderly, compared with 1.49% (−0.25–3.27) and 1.32% (0.04–2.61) increases in adults aged 18–65. The differences between the two groups were statistically significant (PM2.5: p = 0.001; PM10: p = 0.003). Associations between PM concentrations and admissions were also observed in adolescents, where an IQR increase in PM2.5 and PM10
Fig. 2. Percentage change with 95% CI in depression admissions associated with an interquartile range increase in PM2.5 and PM10 concentrations stratified by gender. ⁎p < 0.1.
118
Environment International 114 (2018) 115–122
F. Wang et al.
Fig. 3. Percentage change with 95% CI in depression admissions associated with an interquartile range increase in PM2.5 and PM10 concentrations stratified by age. ⁎p < 0.1 ⁎⁎p < 0.05 ⁎⁎⁎ p < 0.01.
Fig. 4. Percentage change with 95% CI in depression admissions associated with an interquartile range increase in PM2.5 and PM10 concentrations stratified by comorbidities. ⁎p < 0.05.
symptoms among the elderly (Lim et al., 2012). These findings support the hypothesis that short-term PM exposure is linked to depressive symptoms, but the mechanism has yet to be determined. One possible explanation could be neuroinflammation, caused by air pollutants reaching the brain in people who are highly exposed (CalderonGarciduenas et al., 2008); depressive symptoms have been observed to be associated with inflammatory processes (Raison et al., 2006). PM10 has been identified as a strong inflammatory agent (CalderonGarciduenas et al., 2008), and PM2.5 is thought to exert greater toxicity than PM10 because smaller particles can deposit in deeper areas of the lungs, leading to higher concentrations of adsorbed or condensed toxicants in the body (Sarnat et al., 2016). Exposure to PM2.5 and PM10 could therefore increase the risk of depressive symptoms via neuroinflammatory processes. Notably, in our study, the effect of PM2.5 increased on lag day 0 and lag day 5, while PM10 effects were highest on lag day 0. The continued effects of PM2.5 on lag day 5 may reflect impaired function related to internal PM accumulation. The differences observed for the effects of PM2.5 and PM10 may therefore have resulted
0 day within SO2 (3.73; 0.97–6.56), on the lag 2 day within NO2, (5.59; −0.09-11.59) and on the lag 2 day within CO (6.75; 2.32–11.38) (Table 5).
4. Discussion This study investigated the effect of short-term exposure to ambient PM (especially PM2.5) on hospitalizations for depression in 26 Chinese cities and found that short-term exposure to increased levels of PM was significantly associated with increased risk of hospital admission for depression. The maximum effects were observed on lag day 0 and lag day 5 for PM2.5 and on lag day 0 for PM10. These findings are consistent with those reported by previous studies on PM and depression. Szyszkowicz (2007) and Szyszkowicz et al. (2009) reported that short-term increases in ambient air pollution were associated with increased emergency department visits for depression in Canada. Similarly, a Korean study found that the 3-day moving average (lag 0–2) concentration of PM10 was associated with depressive 119
Environment International 114 (2018) 115–122
F. Wang et al.
Table 4 Percentage changes with per IQR change in PM2.5 (95% Confidence Interval) for two pollutants models without and with interaction (within the first, second and third quantiles of the two-pollutants). PM2.5 Co-pollutant
Without interaction
With interaction First quartile
Second quantile
Third quantile
3.38 3.31 3.33 3.07 3.24 2.95 1.34 0.55
3.32 3.25 3.30 2.99 3.21 2.97 1.40 0.56
SO2 Lag 0 Lag 1 Lag 2 Lag 3 Lag 4 Lag 5 Lag 6 Lag 7
2.70 2.55 2.98 2.21 2.93 3.23 2.18 0.65
(0.84,4.59) (0.63,4.51) (0.74,5.27) (0.37,4.08) (1.08,4.81) (1.40,5.10) (0.42,3.97) (−1.13,2.45)
3.41 3.35 3.35 3.10 3.25 2.94 1.31 0.55
NO2 Lag 0 Lag 1 Lag 2 Lag 3 Lag 4 Lag 5 Lag 6 Lag 7
1.27 1.57 3.24 1.65 3.40 2.59 0.85 0.24
(−0.93,3.52) (−0.71,3.90) (0.70,5.84) (−0.56,3.92) (1.16,5.69) (0.38,4.85) (−1.27,3.00) (−1.9,2.43)
2.90 (−1.10,7.06) 5.83 (1.57,10.27) 5.72 (0.84,10.84) 4.46 (0.33,8.75) 3.33 (−0.78,7.61) 1.41 (−2.56,5.55) −1.47 (−5.22,2.42) −0.78 (−4.61,3.21)
2.85 (−1.07,6.92) 5.7 (1.53,10.04) 5.64 (0.87,10.63) 4.37 (0.33,8.57) 3.33 (−0.69,7.51) 1.45 (−2.44,5.49) −1.40 (−5.08,2.41) −0.75 (−4.5,3.15)
2.78 (−1.02,6.73) 5.52 (1.48,9.72) 5.53 (0.92,10.35) 4.25 (0.34,8.32) 3.34 (−0.57,7.39) 1.5 (−2.28,5.42) −1.31 (−4.89,2.4) −0.71 (−4.36,3.08)
CO Lag Lag Lag Lag Lag Lag Lag Lag
2.39 (−0.01,4.85) 3.45 (0.98,5.98) 4.14 (1.70,6.63) 3.02 (0.62,5.48) 4.64 (2.16,7.19) 5.14 (2.67,7.68) 1.81 (−0.52,4.19) −0.16 (−2.49,2.23)
3.23 (0.20,6.35) 6.25 (2.94,9.66) 6.73 (3.04,10.55) 5.66 (2.45,8.96) 5.61 (2.37,8.96) 6.34 (3.07,9.72) 1.93 (−0.96,4.92) −0.04 (−2.93,2.94)
3.20 (0.20,6.29) 6.16 (2.90,9.53) 6.64 (3.02,10.39) 5.58 (2.42,8.84) 5.58 (2.38,8.89) 6.31 (3.08,9.64) 1.93 (−0.94,4.88) −0.04 (−2.9,2.9)
3.17 (0.21,6.21) 6.04 (2.83,9.34) 6.52 (3.00,10.17) 5.47 (2.36,8.67) 5.54 (2.39,8.79) 6.26 (3.08,9.52) 1.93 (−0.9,4.83) −0.05 (−2.86,2.85)
0 1 2 3 4 5 6 7
(0.99,5.88) (0.9,5.86) (0.49,6.29) (0.72,5.53) (0.88,5.69) (0.60,5.34) (−0.96,3.63) (−1.78,2.93)
(1.00,5.80) (0.90,5.78) (0.52,6.23) (0.73,5.46) (0.90,5.64) (0.64,5.32) (−0.90,3.62) (−1.74,2.89)
(1.02,5.66) (0.91,5.64) (0.58,6.10) (0.73,5.31) (0.95,5.53) (0.73,5.27) (−0.77,3.62) (−1.66,2.83)
Table 5 Percentage changes with per IQR change in PM10 (95% Confidence Interval) for two pollutants models without and with interaction (within the first, second and third quantiles of the two-pollutants). PM10 Co-pollutant
Without interaction
With interaction First quartile
Second quantile
Third quantile
SO2 Lag 0 Lag 1 Lag 2 Lag 3 Lag 4 Lag 5 Lag 6 Lag 7
3.25 (1.05,5.5) 3.23 (0.99,5.51) 3.51 (0.84,6.26) 2.15 (−0.05,4.39) 1.91 (−0.22,4.1) 1.94 (−0.19,4.12) 0.86 (−1.27,3.03) −0.43 (−2.55,1.73)
3.73 (0.97,6.56) 3.57 (0.8,6.41) 3.31 (−0.01,6.74) 2.74 (0.03,5.53) 1.92 (−0.7,4.61) 1.07 (−1.49,3.71) −0.49 (−3.02,2.12) −0.74 (−3.35,1.93)
3.71 (0.99,6.51) 3.56 (0.82,6.37) 3.31 (0.03,6.71) 2.72 (0.04,5.48) 1.92 (−0.67,4.58) 1.1 (−1.45,3.71) −0.45 (−2.96,2.13) −0.73 (−3.32,1.92)
3.68 (1.01,6.43) 3.54 (0.85,6.3) 3.33 (0.1,6.65) 2.69 (0.05,5.39) 1.92 (−0.62,4.53) 1.15 (−1.36,3.71) −0.37 (−2.85,2.16) −0.72 (−3.25,1.88)
NO2 Lag 0 Lag 1 Lag 2 Lag 3 Lag 4 Lag 5 Lag 6 Lag 7
1.59 (−1.04,4.29) 2.23 (−0.47,5) 3.81 (0.74,6.98) 1.24 (−1.42,3.97) 1.79 (−0.76,4.41) 0.42 (−2.11,3.02) −1.53 (−4.05,1.07) −1.37 (−3.9,1.24)
2.57 (−1.82,7.16) 5.07 (0.48,9.87) 5.59 (−0.09,11.59) 2.95 (−1.44,7.54) 0.69 (−3.47,5.03) −2.71 (−6.73,1.48) −6.16 (−10.14,-2) −4.11 (−8.16,0.12)
2.55 (−1.79,7.08) 5.01 (0.48,9.74) 5.55 (−0.05,11.46) 2.91 (−1.42,7.44) 0.71 (−3.39,4.99) −2.64 (−6.61,1.49) −6.06 (−9.99,−1.95) −4.05 (−8.05,0.12)
2.52 (−1.73,6.96) 4.93 (0.49,9.56) 5.5 (0.02,11.28) 2.86 (−1.39,7.3) 0.75 (−3.28,4.94) −2.55 (−6.45,1.51) −5.93 (−9.8,−1.89) −3.97 (−7.9,0.13)
CO Lag Lag Lag Lag Lag Lag Lag Lag
2.76 (0.15,5.44) 3.93 (1.27,6.65) 4.65 (1.82,7.55) 2.42 (−0.09,4.99) 2.43 (0,4.93) 2.46 (0.03,4.95) −0.09 (−2.5,2.38) −1.22 (−3.63,1.25)
3.51 (0.07,7.07) 6.16 (2.53,9.92) 6.75 (2.32,11.38) 4.38 (1,7.87) 2.62 (−0.6,5.94) 2.5 (−0.72,5.83) −1.34 (−4.43,1.84) −1.93 (−5.03,1.27)
3.5 (0.09,7.03) 6.11 (2.51,9.83) 6.71 (2.33,11.28) 4.34 (0.99,7.79) 2.62 (−0.57,5.9) 2.5 (−0.69,5.79) −1.31 (−4.37,1.84) −1.91 (−4.98,1.26)
3.48 (0.11,6.96) 6.04 (2.5,9.71) 6.65 (2.35,11.14) 4.28 (0.98,7.69) 2.61 (−0.54,5.85) 2.5 (−0.65,5.75) −1.28 (−4.3,1.84) −1.89 (−4.93,1.24)
0 1 2 3 4 5 6 7
120
Environment International 114 (2018) 115–122
F. Wang et al.
observed effect modification by age and comorbidities, with significantly stronger associations in the elderly and in CVD patients. Future work should focus on elucidating the underlying biological mechanisms.
from deeper respiratory deposition of PM2.5, leading to a longer period of impairment. In this study, the effect of PM2.5 on lag day 0 showed a trend to be more pronounced in females, while other estimates did not show significant sex difference. The number of females in our sample was larger, which is in line with the sex-specific prevalence of major depression in China (Gu et al., 2013). Higher effects of air pollution on respiratory health have been observed among females than among males, and the possible sources of difference in air pollution response include life stage, work-related coexposures, hormonal status, and other factors (Clougherty, 2010). However, few studies have explored the effects of PM on depression in the context of sex-specific differences. Szyszkowicz et al. (2016) reported that exposure to PM2.5 was associated with increased risk in males one day after the exposure, while no such association was observed in females. This was not with the case in our study, and the reasons could be differences in the data sources, composition and relative concentrations of pollutants, or variations in patient characteristics. Further study is required to determine the effects of sex on the association between PM levels and hospitalizations for depression. Depressive symptoms are highly prevalent in the elderly (Thielke et al., 2010), yet few studies to date have looked at the association between PM2.5 exposure and depression in the elderly. Kioumourtzoglou et al. (2017) reported a possible association between long-term PM2.5 exposure and the onset of depression among older women. Another cohort study found that increases in PM10 were significantly associated with increased depressive symptoms in the elderly (Lim et al., 2012). In our study, the effects of both PM2.5 and PM10 appeared more pronounced in the elderly. The findings suggest that the elderly should get more protection from the air pollution as they might be more susceptible to ambient PM-related depression. Depressive disorder is frequently comorbid with other conditions, such as CVD (Musselman et al., 1998) and diabetes (Stuart and Baune, 2012). We found that CVD patients, when exposed to higher levels of PM, had a significantly higher risk of hospitalization for depression than non-CVD patients. This finding is consistent with that of a study on ambient PM that reported an increased risk of suicide among subjects with pre-existing CVD (Kim et al., 2010). Exposure to ambient PM has been shown to adversely affect vascular endothelial function, leading to increased plasma viscosity, a risk factor of blood clotting, thrombosis, and ischemic stroke (Gurgueira et al., 2002; Lucking et al., 2011; Liu et al., 2017). CVD has been linked to depression via systemic inflammation (Ridker et al., 1997). Air pollutants could therefore trigger depressive symptoms, especially among CVD patients. CVD can be both a mediator and a modifier and future studies are needed to figure it out. A previous study found an association between PM and depression in patients with diabetes or COPD (Kim et al., 2016), but we did not, perhaps owing to the smaller sample in our study. This study has some limitations. First, our data could not discriminate between emergent hospitalization and scheduled admission, making it difficult to determine an association between PM exposure and symptoms of depression. Second, using citywide air-pollutant mass concentrations as proxies for personal exposure may result in measurement error, underestimating the effects of the pollutants (Goldman et al., 2011). In addition, potential exposure and outcome misclassification should be considered, as the ICD-10 codes were the only data we could use to define patients with depression. In China, air pollution levels are much higher than in the US and Canada, the sources of most studies on air pollution and depression, resulting in limited comparability and generalizability of the findings. Moreover, we mainly discussed the findings of a linear model, although the non-linear findings have been added to the supplementary material; more research on non-linear associations is needed. In summary, our findings indicate that short-term elevations in the levels of PM2.5 and PM10 are significantly associated with an increased risk of hospitalization for depression in 26 Chinese cities. We also
Appendix A. Supplementary data Supplementary data to this article can be found online at https:// doi.org/10.1016/j.envint.2018.02.012. References Altman, D.G., Bland, J.M., 2003. Interaction revisited: the difference between two estimates. BMJ 326, 219. Bell, M.L., Samet, J.M., Dominici, F., 2004. Time-series studies of particulate matter. Annu. Rev. Public Health 25, 247–280. Block, M.L., Wu, X., Pei, Z., et al., 2004. Nanometer size diesel exhaust particles are selectively toxic to dopaminergic neurons: the role of microglia, phagocytosis, and NADPH oxidase. FASEB J. 18, 1618–1620. Brites, D., Fernandes, A., 2015. Neuroinflammation and depression: microglia activation, extracellular microvesicles and microRNA dysregulation. Front. Cell. Neurosci. 9, 476. Calderon-Garciduenas, L., Solt, A.C., Henriquez-Roldan, C., et al., 2008. Long-term air pollution exposure is associated with neuroinflammation, an altered innate immune response, disruption of the blood–brain barrier, ultrafine particulate deposition, and accumulation of amyloid beta-42 and alpha-synuclein in children and young adults. Toxicol. Pathol. 36, 289–310. Campbell, A., Oldham, M., Becaria, A., et al., 2005. Particulate matter in polluted air may increase biomarkers of inflammation in mouse brain. Neurotoxicology 26, 133–140. Carbajal-Arroyo, L., Miranda-Soberanis, V., Medina-Ramón, M., Rojas-Bracho, L., Tzintzun, G., Solís-Gutiérrez, P., Méndez-Ramírez, I., Hurtado-Díaz, M., Schwartz, J., Romieu, I., 2011. Effect of PM(10) and O(3) on infant mortality among residents in the Mexico city metropolitan area: a case-crossover analysis, 1997–2005. J. Epidemiol. Community Health 65, 715–721. Carracedo-Martinez, E., Taracido, M., Tobias, A., Saez, M., Figueiras, A., 2010. Casecrossover analysis of air pollution health effects: a systematic review of methodology and application. Environ. Health Perspect. 118, 1173–1182. Center for Disease Control and Prevention, 2010. Current depression among adults – United States, 2006 and 2008. Morb. Mortal. Wkly Rep. 59, 1229–1235. Clougherty, J.E., 2010. A growing role for gender analysis in air pollution epidemiology. Environ. Health Perspect. 118 (2), 167–176. Cuijpers, P., Beekman, A.T., Reynolds, C.F. rd, 2012. Preventing depression: a global priority. JAMA 307 (10), 1033–1034. Goldberg, M.S., Gasparrini, A., Armstrong, B., Valois, M.F., 2011. The short-term influence of temperature on daily mortality in the temperate climate of Montreal, Canada. Environ. Res. 111, 853–860. Goldman, G.T., Mulholland, J.A., Russell, A.G., Strickland, M.J., Klein, M., Waller, L.A., et al., 2011. Impact of exposure measurement error in air pollution epidemiology: effect of error type in time-series studies. Environ. Health 10, 10–61. Gu, L., Xie, J., Long, J., Chen, Q., Chen, Q., et al., 2013. Epidemiology of major depressive disorder in mainland China: a systematic review. PLoS One 8 (6), e65356. Gurgueira, S.A., Lawrence, J., Coull, B., Murthy, G.G., Gonzalez-Flecha, B., 2002. Rapid increases in the steady-state concentration of reactive oxygen species in the lungs and heart after particulate air pollution inhalation. Environ. Health Perspect. 110, 749–755. Hasler, G., 2010. Pathophysiology of depression: do we have any solid evidence of interest to clinicians? World Psychiatry Official Journal of the World Psychiatric Association 9 (3), 155–161. Hu, T.W., He, Y., Zhang, M., Chen, N., 2007. Economic costs of depression in China. Soc. Psychiatry Psychiatr. Epidemiol. 42 (2), 110–116. Janes, H., Sheppard, L., Lumley, T., 2005. Case-crossover analyses of air pollution exposure data: referent selection strategies and their implications for bias. Epidemiology 16 (6), 717–726. Kan, H., Chen, R., Tong, S., 2012. Ambient air pollution, climate change, and population health in China. Environ. Int. 42 (1), 10. Kim, C., Jung, S.H., Kang, D.R., Kim, H.C., Moon, K.T., Hur, N.W., Shin, D.C., Suh, I., 2010. Ambient particulate matter as a risk factor for suicide. Am. J. Psychiatry 167 (9), 1100–1107. Kim, K.N., Lim, Y.H., Bae, H.J., Kim, M., Jung, K., Hong, Y.C., 2016. Long-term fine particulate matter exposure and major depressive disorder in a community-based urban cohort. Environ. Health Perspect. 124 (10), 1547–1553. Kioumourtzoglou, M.A., Power, M.C., Hart, J.E., et al., 2017. The association between air pollution and onset of depression among middle-aged and older women. Am. J. Epidemiol. 27 (10), 1. Kleine-Budde, K., Müller, R., Kawohl, W., Bramesfeld, A., Moock, J., Rössler, W., 2013. The cost of depression — a cost analysis from a large database. J. Affect. Disord. 147 (1–3), 137–143. Lim, Y.H., Kim, H., Kim, J.H., Bae, S., Park, H.Y., Hong, Y.C., 2012. Air pollution and symptoms of depression in elderly adults. Environ. Health Perspect. 120 (7), 1023–1028. Liu, H., Tian, Y., Xu, Y., Zhang, J., 2017. Ambient particulate matter concentrations and hospitalization for stroke in 26 Chinese cities: a case-crossover study. Stroke 48 (8),
121
Environment International 114 (2018) 115–122
F. Wang et al.
Sarnat, S.E., Chang, H.H., Weber, R.J., 2016. Ambient PM2.5 and health: does PM2.5 oxidative potential play a role? Am. J. Respir. Crit. Care Med. 194, 530–531. Stuart, M.J., Baune, B.T., 2012. Depression and type 2 diabetes: inflammatory mechanisms of a psychoneuroendocrine co-morbidity. Neurosci. Biobehav. Rev. 36 (1), 658–676. Szyszkowicz, M., 2007. Air pollution and emergency department visits for depression in Edmonton, Canada. Int. J. Occup. Med. Environ. Health 20 (3), 241–245. Szyszkowicz, M., Rowe, B.H., Colman, I., 2009. Air pollution and daily emergency department visits for depression. Int. J. Occup. Med. Environ. Health 22 (4), 355–362. Szyszkowicz, et al., 2016. Air pollution and emergency department visits for depression: a multicity case-crossover study. Environmental Health Insights 10, 155–161. Thielke, S.M., Diehr, P., Unutzer, J., 2010. Prevalence, incidence, and persistence of major depressive symptoms in the cardiovascular health study. Aging Ment. Health 14 (2), 168–176. Veronesi, B., Makwana, O., Pooler, M., Chen, L.C., 2005. Effects of subchronic exposures to concentrated ambient particles. VII. Degeneration of dopaminergic neurons in Apo E−/− mice. Inhal. Toxicol. 17, 235–241. Winquist, A., Kirrane, E., Klein, M., Strickland, M., Darrow, L.A., Sarnat, S.E., Gass, K., Mulholland, J., Russell, A., Tolbert, P., 2014. Joint effects of ambient air pollutants on pediatric asthma emergency department visits in Atlanta, 1998–2004. Epidemiology 25 (5), 666–673.
2052–2059. Lucking, A.J., Lundback, M., Barath, S.L., Mills, N.L., Sidhu, M.K., Langrish, J.P., et al., 2011. Particle traps prevent adverse vascular and prothrombotic effects of diesel engine exhaust inhalation in men. Circulation 123, 1721–1728. Moscicki, E.K., 2001. Epidemiology of completed and attempted suicide: toward a framework for prevention. Clin. Neurosci. Res. 1 (5), 310–323. Musselman, D.L., Evans, D.L., Nemeroff, C.B., 1998. The relationship of depression to cardiovascular disease: epidemiology, biology, and treatment. Arch. Gen. Psychiatry 55 (7), 580–592. Raison, C.L., Capuron, L., Miller, A.H., 2006. Cytokines sing the blues: inflammation and the pathogenesis of depression. Trends Immunol. 27 (1), 24–31. Rashid, T., Heider, I., 2008. Life events and depression. Ann. Punjab Med. Coll. 2 (1), 128–149. Ridker, P.M., Cushman, M., Stampfer, M.J., Tracy, R.P., Hennekens, C.H., 1997. Inflammation, aspirin, and the risk of cardiovascular disease in apparently healthy men. N. Engl. J. Med. 336 (14), 973–979. Rustad, J.K., Musselman, D.L., Nemeroff, C.B., 2011. The relationship of depression and diabetes: pathophysiological and treatment implications. Psychoneuroendocrinology 36 (9), 1276–1286. Saravane, D., Feve, B., Frances, Y., et al., 2009. Drawing up guidelines for the attendance of physical health of patients with severe mental illness. L'Encéphale 35 (4), 330–339.
122