Background. Air pollution has a negative influence on neurological and psychiatric disorders. However, findings concerning the impact of air pollution on depression remain inconclusive. A deeper insight into these associations is warranted.
Objectives. To evaluate the impact of long-term exposure to air pollution on the incidence of depression among residents of 13 counties in the Lower Silesia region of Poland.
Materials and methods. We used data on cases of depression from the National Health Fund (Narodowy Fundusz Zdrowia – NFZ) from 13 counties of Lower Silesia between January 1, 2010, and December 31, 2015. Patients with a confirmed diagnosis of depression were included. Data on air pollution levels were extracted from the Chief Inspectorate of Environmental Protection (Główny Inspektorat Ochrony Środowiska – GIOŚ), and demographic data were extracted from Statistics Poland (Główny Urząd Statystyczny – GUS).
Results. The percentage of people diagnosed with depression over the 6-year study period depended on the group of counties homogeneous in terms of air pollution exposure (p < 0.001). We showed statistically significant correlations between different depression diagnoses and exposure to air pollutants. Elevated concentration of airborne fine particles with a diameter less than 2.5 µm (PM2.5) and carbon monoxide (CO), and low benzo(a)pyrene (BaP), sulfur dioxide (SO2) and cadmium (Cd) levels were independent risk factors for major depressive episodes with psychotic symptoms (F32.3). There was a significant negative correlation between ozone (O3) levels and depression incidence.
Conclusions. Regions with heavy air pollution had a higher incidence of depression. There is a significant association between the exposure to air pollutants and different depression diagnoses.
Key words: depression, air pollution, particulate matter, ozone, nitrogen oxides
Depression is a mental disorder that affects 264 million people worldwide, and is a major cause of disability.1 Beck described depression as a combination of disturbed thinking, low mood and a negative perception of self, the surrounding world and the future.2 Depression is associated with excess mortality3 and increased suicide risk.4 There are many risk factors for depression,5 such as female gender, chronic illness, traumatic life events,6 unemployment, lower education, a lack of physical activity, and limited social support.7
In recent years, there has been an increasing interest in the environmental risk factors for depression.8 Lack of green spaces, poor quality of housing, chronic noise exposure, and air pollution are all strongly associated with risk of depression.9 Both chronic stress10 and gene–environment interactions11 contribute to risk for psychiatric illnesses, such as depression. Neuroinflammation is one of the main pathogenic mechanisms implicated in depression.12 A prominent environmental risk factor associated with neuroinflammation is air pollution.13 The exposure to intense air pollution may disrupt metabolic processes and thus lead to inflammation.14 For example, experimental studies have shown that high concentrations of air pollutants are associated with systemic inflammation and insulin resistance.15, 16
Long-term exposure to certain air pollutants leads to low-grade inflammation.17 In response, more anti-inflammatory cytokines are released, in turn causing immune tolerance. Chronic inflammation is also linked to the activation of the kynurenine (KYN) metabolic pathway of tryptophan, leading to the synthesis of various oxidant and immunomodulating molecules at the expense of the serotonin production.18 These metabolites modulate the immune system towards low-grade inflammation, strengthening the vicious cycle of the immune response.19 The KYN pathway plays a role in the pathogenesis of neurodegenerative and psychiatric diseases.18 The levels of quinolinic acid (a toxin produced by KYN pathway) are increased in patients with major depression.20 Moreover, experimental data indicate that one compound of air pollution, ozone (O3), causes increased concentration of kynurenine.21 These mechanisms could explain the adverse impact of air pollution on neuroinflammation and risk for mental illnesses.
Air pollution is a mixture of various particles and gases suspended in the air, such as particulate matter (coarse particles with an aerodynamic diameter between 2.5 µm and 10 µm (PM10), fine particles with a diameter less than 2.5 µm (PM2.5), and ultrafine particles with a diameter less than 0.1 µm (PM0.1)), e.g., O3, carbon monoxide (CO), sulfur dioxide (SO2), nitrogen oxides (NOx), nitrogen dioxide (NO2), and polycyclic aromatic hydrocarbons (PAHs). Their main sources are anthropogenic emissions from traffic, industry and coal stoves.22 The Lancet Commission highlighted that air pollution is one of the main environmental causes of disease and premature death.23 It is estimated that 8.79 million people worldwide died prematurely in 2019.24 Studies consistently indicate that increased mortality associated with exposure to air pollutants is largely due to cardiovascular and respiratory diseases.14 The health impact of air pollutants has been broadened to include neurological and psychiatric disorders such as anxiety, depression and dementia.25 Research on these associations is extremely important, since mental and neurological conditions are among the main causes of disability, including 30% of all strokes being related to air pollution.26
There are some proposed mechanisms whereby air pollutants can negatively affect the brain. For instance, particulate matter can reach the systemic circulation through the lungs or through olfactory neurons,27 and directly interact with the brain.28 Especially dangerous are ultrafine particles which, due to their small size and the toxic compounds bound to them, can cause pathophysiological alterations in the central nervous system.28 Another mechanism is alveolar inflammation, which leads to the production of cytokines29 that can penetrate the blood–brain barrier.30 The exposure to air pollutants triggers the activation of the hypothalamic–pituitary–adrenal (HPA) axis and the release of stress hormones,31 which is commonly associated with cardiovascular disease (CVD), type II diabetes mellitus, dementia, and depression.32 These mechanisms can explain the negative effect of air pollution on the central nervous system and mental health.33
Findings on the impact of air pollution on depression remain inconclusive. Two meta-analyses confirmed that a long-term exposure to fine particles is associated with increased incidence of depression34 and suicide risk.35 Another study showed that PM2.5 is significantly associated with the occurrence of depression for both long, and short-term exposure.36 Long-term exposure to PM2.5 and short-term exposure to PM10, NO2, SO2, and CO, are risk factors for depression.37 On the other hand, a meta-analysis of 22 studies showed a weaker relationship between air pollution and depression risk. This includes long-term exposure to PM2.5, PM10 and NO2, as well as short-term exposure to PM2.5, PM10, SO2, O3, and NO2.38 Moreover, current studies are based mostly on self-reports, and they do not specify the association between air pollution and the sub-diagnoses of depression.34, 37 No studies have examined the relationship between the exposure to air pollution and risk for depression with psychotic symptoms. There is also not enough evidence on long-term exposure to many different air pollutants and its impact on depression, especially at the sub-diagnosis level. Therefore, a deeper insight into these associations is important, and more data are needed to clarify the impact of air pollutants on depression risk.
The aim of this study was to evaluate the impact of long-term exposure to air pollution on depression risk among residents of 13 counties in Lower Silesia (Poland).
Materials and methods
Data on depression cases used in the study were derived from the National Health Fund (Narodowy Fundusz Zdrowia – NFZ) from counties (Polish: powiat; Głogów, Jelenia Góra, Kłodzko, Legnica, Lubań, Oława, Oleśnica, Polkowice, Złotoryja, Zgorzelec, Wałbrzych, Świdnica, and Wrocław) pertaining the period between January 1, 2010, and December 31, 2015. The inclusion criteria were based on the International Classification of Diseases, 10th Revision (ICD-10) code diagnosis: F32 (depressive episode), F32.0 (mild depressive episode), F32.1 (moderate depressive episode), F32.2 (severe depressive episode without psychotic symptoms), F32.3 (severe depressive episode with psychotic symptoms), F33 (recurrent depressive disorder), F33.0 (recurrent depressive disorder, current episode mild), F33.1 (recurrent depressive disorder, current episode moderate), F33.2 (recurrent depressive disorder, current episode severe without psychotic symptoms), and F33.3 (recurrent depressive disorder, current episode severe with psychotic symptoms). In total, data on depression diagnoses were available for 318,779 individuals per study year. Data on air pollution levels were extracted from the Chief Inspectorate of Environmental Protection (Główny Inspektorat Ochrony Środowiska – GIOŚ) and covered the period from 2010 to 2015. Considered pollutants included: PM10, PM2.5, SO2, NOx, benzo(a)pyrene (BaP)(PM10), O3, CO, and NO2. We used demographic data for the period 2010–2015 from Statistics Poland (Główny Urząd Statystyczny – GUS), including marriage and divorce rate, feminization rate, non-working population, working people, people >65 years old, medical help accessibility, beneficiaries of social welfare, unemployment, average salary, and total pollution emissions. Two statistical methods were used. Cluster analysis was used to distinguish between counties homogeneous in terms of air pollution. The 2nd method utilized all county data to perform univariate and multivariate regression analyses, and an analysis of variance (ANOVA). For multivariate and ANOVA analyses, data on marriage and divorce rate, feminization rate, non-working population, working people, people >65 years old, medical help accessibility, beneficiaries of social welfare, unemployment, average salary, total pollution emissions, and PM10, PM2.5, SO2, NOx, BaP(PM10), O3, CO, and NO2 concentrations were used. For univariate analyses, we used depression rate and its associations with PM10, PM2.5, SO2, NOx, BaP(PM10), O3, CO, and NO2 concentrations. The homogeneity of variance in the groups was verified using the Levene’s test. In the case of unequal variances in the compared groups, the Welch’s correction was applied. Normality of distribution was determined using the Kolmogorov–Smirnov and Shapiro–Wilk tests. For the construction of multivariate linear regression models, non-normal data were transformed using the Box–Cox method.
Table 1 presents data on air pollutant levels. Cluster analysis was used to distinguish homogeneous counties in terms of air pollution. It was possible to reduce a whole set to the average of individual groups. The Euclidean distance was used as a dissimilarity measure. Based on the hierarchical structure of the set of counties, they were divided into 4 clusters homogeneous in terms of air pollution:
• cluster 1: Lubań county;
• cluster 2: Głogów, Oława, Oleśnica, Polkowice, Złotoryja, Zgorzelec, and Wałbrzych counties;
• cluster 3: Jelenia Góra, Kłodzko, Legnica, and Świdnica counties;
• cluster 4: Wrocław county.
A graph of average values (Figure 1) for standardized air pollution parameters showed that the highest level of air pollution, except for lead (Pb) and SO2, was recorded in the Wrocław county (cluster 4). The next most affected counties in terms of the level of air pollution were Jelenia Góra, Kłodzko, Legnica, and Świdnica (cluster 3). The lowest level of pollution was recorded in the Lubań county (cluster 1).
The number of people diagnosed with depression in 2010–2015 was compared between individual counties (Table 2). The structure index (percentage) of people diagnosed with depression in the next 6 years of observation did not change significantly (p > 0.05). However, this depended significantly (p = 0.003) on the group of counties homogeneous in terms of air pollution (Figure 2). The average structure index of people diagnosed with depression between 2010 and 2015 across the 13 counties was 2.72 ±1.78% (women: 2.04 ±1.30%; men: 0.67 ±0.50%).
The linear relationships between depression diagnoses and air pollutant levels measured between 2010 and 2015 for all counties are summarized in Table 3. Since there were statistically significant correlations between the independent variables (predictors), the variance inflation factor (VIF) was estimated for each of them (Table 4). The strongest collinearity with PM10 was observed for NO2 and NOx (VIF > 10). Variables NO2 and NOx were excluded from the predictive model for estimating the depression rate. The concentration of PM10 (r = 0.385) and CO (r = 0.344) were significantly positively correlated with the frequency of severe depressive episodes with psychotic symptoms (F32.3). The concentration of O3 was significantly negatively correlated with depression risk (r = −0.283). There were statistically significant correlations between different depression diagnoses and the following pollutants: PM2.5 and F32 (B = 0.303), PM10 and F32, F32.0, F32.2, and F33; CO and F32, F32.0, F32.1, F32.2, F32.3, and F33; Pb(PM10) and F32, F32.0, and F32.2; BaP(PM10) and F32.0; arsenic (As)(PM10) and F32, F32.0 and F32.2; and cadmium (Cd)(PM10) and F32, F32.0 and F32.2. The O3 levels were significantly negatively correlated with all depression diagnoses except F33.
Mathematical model of the occurrence of major depressive episodes with psychotic symptoms (F32.3)
The percentage of people diagnosed with a F32.3 in all counties was 0.23 ±0.19%. Table 5 presents the results of uni- and multivariate regression analyses of predictors associated with the risk for F32.3. The PM2.5, CO, BaP, SO2, and Cd levels were independently associated with the occurrence of F32.3. The formula for evaluating the structure index of F32.3 can be expressed as:
F32.3 = −403.4 + 10.8 × PM2.5 + 379.8 × CO
+ 24.1 × Ni − 5.33 × BaP − 6.44 × SO2 + 25.8 × Cd
R = 0.803; F (degrees of freedom (df): 7,70) = 18.2; p < 0.001
The positive values of the regression coefficients indicated the effects of increased PM2.5, CO, Ni, and Cd on the occurrence of F32.3. All structural parameters of the model were statistically significant (p < 0.05), and the model fit was satisfactory (p < 0.001). In order to avoid the pseudoreplication problem, and since the measurements of depression rates were taken in each county annually, the correlation analysis in consecutive individual years between air pollution and the occurrence of F32.3 was performed. The PM10 correlated significantly with the F32.3 diagnosis in 2010 (rho = 0.560, p < 0.05), and both PM10 and NO2 concentrations were linked to this diagnosis across all years (rho = 0.386, p < 0.001; rho = 0.375, p < 0.001; Table 6; Figure 3, Figure 4).
Analysis including demographic
and social data
Over the entire analyzed period, the percentage of diagnosed cases of depression was stable at a level of 2.7 ±1.8%. The analyzed database lacked information on the depression incidence in the Wałbrzych county in 2010–2012. The missing data were supplemented with average values. Basic demographic data concerning depression cases (2010–2015) across the 13 counties of Lower Silesia are summarized in Table 6 and Table 7. Statistically significant differences were observed between depression incidence in individual counties (Figure 5) (p < 0.001). The highest incidence of depression was recorded in Jelenia Góra (2) and Legnica (4) counties (6.1%), while the lowest in Kłodzko (0.9%) and Oleśnica (1.1%) counties.. Table 8 shows the results of ANOVA concerning depression cases and Table 9 displays the results of post-hoc test (Turkey’s test).
The distribution of depression cases was non-normal (Figure 6). The analyzed counties were classified into 3 groups: group A – depression rate below 1.5% (Kłodzko, Lubań, Oleśnica, and Świdnica counties); group B – depression rate of 1.5–3.1% (Oława, Polkowice, Złotoryja, and Zgorzelec counties); group C – depression rate over 3.1% (Głogów, Jelenia Góra, Legnica, Wałbrzych, and Wrocław counties) (Figure 7). In group C, a high percentage of diagnosed depression cases was associated with higher concentrations of PM10 and PM2.5, as well as lower O3 concentrations. The level of BaP did not differ significantly between the individual counties. The average values of pollutants in groups of counties differing in the severity of diagnosed depression are summarized in Table 10 and Table 11, as well as Figure 8.
In counties with the highest percentage of diagnosed depression, there were significantly higher emissions of gaseous pollutants (1000 t/year/km2) (p = 0.003), SO2 (p = 0.011), NOx (p = 0.002), CO (p = 0.001), and CO2 (p = 0.003), as well as concentrations of PM10 (p = 0.039), PM2.5 (p = 0.008) and NOx (p = 0.025). The O3 levels were significantly inversely correlated with depression (p < 0.001; Table 10, Table 11, Figure 8).
In order to determine the independent predictors of depression incidence, univariate and multivariate linear regression analyses were performed (Table 12). Univariate regression analysis showed that PM10 and PM2.5 concentrations, as well as divorces (per 1000 inhabitants), demographic burden, demographic dependency rate for the elderly, percentage of people >65 years old, feminization rate, outpatient clinics per 1000 population, doctors (total working staff) per 10,000 population, number of beds in hospitals in relation to the population, working people rate, emission of gaseous pollutants, and average monthly gross salary were significantly positively correlated with the incidence of depression. On the other hand, O3 concentration, the number of marriages and the number of beneficiaries of social welfare were significantly negatively correlated with depression risk. Multivariate regression analysis showed that O3 concentration, demographic burden, feminization rate, number of doctors, number of hospital beds in relation to the population, number of employed people per 1000 population, and average monthly gross salary were independent predictors of the depression rate. Increased O3 concentrations were significantly associated with a lower depression risk (p = 0.004).
We showed that depression is associated with long-term exposure to higher concentrations of PM10, PM2.5 and NOx, and high emissions of SO2, NOx, CO, and CO2. Other studies have reported similar results. For example, research using the National Health Insurance Database demonstrated a positive association between PM2.5 and depression risk for both 1-year and 4-year exposure.39 Other studies also found this relationship for periods of 1 year,40 2 years and 5 years.41 In terms of PM10 long-term exposure, a positive relationship was found over a 1-year time period.42 A study by Vert et al. supports these results for both PM fractions.43 However, there was also a negative association for PM10.44 Furthermore, only a few studies evaluated the relationship between depression and NOx. These studies consistently suggest a positive correlation between depression and long-term exposure to NO2 and NOx.43, 45
We have observed a negative association between long-term exposure to O3 and depression, which contradicts some of the previous results.41 However, some studies found no relationship between O3 levels and depressive symptoms.46 Volatile organic compounds and NOx are precursors to O3 formation,47 and photochemical processes which depend on sunlight play a key role.48 This can explain our results, since sunlight exposure49 and vitamin D3 production50 have a preventive impact on depression. The O3 exposure and its health effects are more complicated, and possibly threshold-like.51 Finally, our study showed the important correlation between NOx, NO2 and depression risk, and these pollutants are known to decrease O3 production.52
We showed that high levels of PM2.5, CO, and low BaP, SO2 and Cd concentrations are independent risk factors for depressive episodes with psychotic symptoms (F32.3). To the best of our knowledge, there is no available research on this topic. However, some researchers attempted to find a connection between air pollution and depression severity. Pun et al. showed that PM2.5 levels are positively associated with depressive symptom severity.36 On the other hand, Wang et al. found no relationship between depressive symptoms and both long- and short-term exposure to air pollutants.13 Similarly, there is very little evidence regarding psychotic symptoms and air pollution. An analysis using machine learning found a connection between schizophrenia emergency room admissions and ambient PM2.5,53 which was confirmed by Gao et al.54
A multivariate analysis of demographic and pollution data showed that the number of hospital beds, number of welfare recipients, number of employed people, and the average monthly gross salary were all related to the incidence of depression. These results are inconsistent with other studies, which demonstrated that welfare recipients are more likely to suffer from depression.55 However, our results could represent reverse causality. For instance, we showed a positive correlation between depression and the number of hospital beds in the region, which could result from the fact that people with depression are more frequently hospitalized.13 Similarly, employment status and depression rates could be related to workplace conditions, burnout, stress, and bullying,56 which also translate into monthly salary. Depression accounts for up to 46% of days lost due to illness; therefore, it is more often diagnosed among the working population.56, 57 However, in our study, data concerning employment were not available.
Results of our analyses based on 318,779 cases showed that there is a positive association between air pollutant levels and the risk of depression. However, these results should be interpreted cautiously, because, as an observational study, despite being useful in explaining the exposure to an outcome, they carry out the risk of possible reverse causality or undetected bias.58 Other limitations are related to the use of semi-individual data, and potential underestimation of the prevalence of depression due to social stigmatization related to this diagnosis.59 Nevertheless, because of the large and standardized sample size, our study provides new insights into the relationship between air pollutant exposure and depression.
The analyzed database lacked information on depression incidence for the Wałbrzych county in 2010–2012. Since regression analysis was used, even if a strong correlation is evident, it does not imply causation. Moreover, some risk factors found in our study could simply reflect the indirect effects of other risk factors. Finally, it is crucial to remember that air pollutant concentrations depend on many variables, such as the direction and speed of the wind, atmospheric stability, solar radiation, or geography.60 This can cause spatial and temporal variations of pollutants levels, reaching dramatic changes, even over the course of several hours. Moreover, the stress response in the human population is extremely heterogeneous.61 Age, gender, existing disease status, and psychosocial stressors play a potential role in the biological response to air pollutants. Given the numerous risk factors of depression, demonstrating the impact of air pollution on depressive disorders is difficult. Therefore, obtaining large-scale data on depression cases, and relating this to demographic and pollution information can help in identifying environmental risk factors, which could be potentially modifiable.
Heavy air pollution is associated with a higher incidence of depression, while O3 levels were linked to the lower rate of depression. Elevated concentration of airborne PM2.5, as well as CO, and low BaP, SO2 and Cd levels are independent risk factors for the occurrence of major depressive episodes with psychotic symptoms (F32.3). A high percentage of people with depression could be related to higher concentrations of PM10 and PM2.5, lower O3 concentrations and higher emissions of gaseous pollutants, such as SO2, NOx, CO, and CO2.
The incidence of depression was associated with high concentrations of PM10 and PM2.5, as well as divorce rate per 1000 inhabitants, demographic burden, demographic dependency rate for the elderly, percentage of people >65 years old, feminization rate, outpatient clinics per 1000 population, doctors (total working staff) per 10,000 population, the number of hospital beds in relation to the population, working people rate, emission of gaseous pollutants, and average monthly gross salary. On the other hand, O3 concentration, the number of marriages, and the number of welfare beneficiaries were negatively correlated with depression risk. The independent predictors of depression were: O3 concentration, demographic burden, feminization rate, number of doctors, number of hospital beds in relation to the population, number of employed people per 1000 population, and the average monthly gross salary.
Availability of data and material
Data on air pollution are available on the website of the Chief Inspectorate of Environmental Protection: http://powietrze.gios.gov.pl/pjp/current. Demographic data are available on the website of the Statistics Poland: https://stat.gov.pl/.