Long-Term Exposure to Air Pollutants and Cognitive Function in Taiwanese Community-Dwelling Older Adults: A Four-Year Cohort Study

INTRODUCTION

As population ages rapidly, dementia has become an important health issue in older adults. Alzheimer’s disease (AD) is the most common cause of dementia and is characterized by noticeable cognitive and behavioral symptoms that are severe enough to affect daily life [1]. A previous survey study in Taiwan (2011–2013) reported in older adults, the age- and sex-adjusted prevalence for overall dementia and mild cognitive impairment (MCI) were 8.1% and 18.8%, respectively [2], which is higher than the prevalence observed in East Asia (4.98–6.99% in 2015) [3]. Because of the heavy burden on health care and aging society, it is important to prevent cognitive impairment before the clinical manifestation.

Cognitive performance could be assessed by various neuropsychological tests of different domains; these cognitive domains could be classified into memory, attention, language, or executive functioning [4]. During the induction period of dementia, different brain areas are involved in different cognitive functions [4], these changes could serve as potential precursors to dementia. Thus, a cohort study with repeated measure is helpful to clarify the disease progression and possible affected areas in the brain.

Although existing scientific consensuses acknowledge air pollution contributes to cardiovascular and respiratory diseases, studies on cognitive performance started to grow after 2010. Among different sizes of particulate matter (PM), PM_2.5 (PM ≤ 2.5 μm in diameter) mainly comes from primary combustion emission and secondary particles that were produced by gas-to-particle conversion [5]. PM_2.5 can penetrate the alveolar gas exchange region of the lungs, mainly deposits in the airway beyond larynx, and thus is known as respirable particles. PM_coarse (PM > 2.5–10μm in diameter) accounts for the majority of visible or obvious form of PM (e.g., black smoke or dust from road, soil, and construction sites) and contains sulfates, nitrates, organic and elemental carbon, sea salts, molds, dust mites pollen, and spores [5]. Existing research indicated the smaller the PM, the more harm to the brain [6]. Our previous case-control study found that long-term exposure to ozone and PM₁₀ (PM ≤ 10μm in diameter) at levels lower than the air quality standards of Taiwan Environmental Protection Administration (TEPA) was significantly associated with 2- to 3-fold increased risk of AD and vascular dementia [7]. Because of the heavy burden to dementia, we are hoping to clarify how exposure to air pollutants affect cognitive performance before disease onset.

The majority of previous studies found consistent associations between PM_2.5 exposure and the risk of cognitive impairment/decline (cohort studies: [8–14]; cross-sectional studies: [15–21]), dementia (AD and vascular dementia) (cohort studies: [11, 22–24]), or brain imaging outcomes such as larger white matter hyperintensity volume and risk of covert brain infarcts (cohort study: [25]). However, this association was not observed in another cross-sectional study [26]. Investigations on the associations between PM_coarse exposure and cognitive impairment/decline are limited and positive associations were observed in some studies (cohort study: [8]; cross-sectional study: [17]). Similarly, four cohort studies found positive associations between PM₁₀ exposure and the risk of cognitive impairment/decline [9, 27], and another cross-sectional study reported positive association between PM₁₀ and dementia [17]. However, these associations were not found in others (cohort studies: [22, 28]; cross-sectional studies: [18, 20]). Regarding NO₂, most of previous studies found that NO₂ exposure was associated with an increased risk of cognitive impairment/decline (cohort studies: [10, 14]; cross-sectional studies: [17]) or dementia (cohort studies: [23, 29]; case-control study: [30]), but such association was not observed in other cross-sectional studies [18, 26]. Taken together, the majority of previous studies found PM_2.5, PM_coarse, PM₁₀, and NO₂ exposure were associated with an increased risk of cognitive impairment/decline or dementia. Different cognitive domains and assessment tools; study design (cohort versus cross-sectional); exposure sources, duration and assessment tools; and geographic locations may explain the inconsistent findings.

OPEN IN VIEWER

Fig. 1

A conceptual timeline of the present study

Development of dementia may take decades; however, several previous studies 1) collected exposure data over a relatively short duration compared with the disease progression [16, 25], 2) focused on global cognition or limited cognitive domains [11, 21], or 3) used cross-sectional design (about 40% of studies). Because we are simultaneously exposed to multiple air pollutants, investigating co-exposure allows us to clarify the effect of how multipollutants affect cognitive performance in real-life settings. However, the majority of studies examine single pollutant in relation to a health outcome; limited data demonstrate how co-exposure to air pollutants affect cognitive performance over time [27]. In addition, air pollutants tend to vary across geographic locations, but studies in Asian countries are sparse. Therefore, the aim of this study was to explore the association between long-term (11 to 23 years) exposure to air pollutants (PM_2.5, PM_coarse, PM₁₀, and NO₂) and cognitive performance in community-dwelling non-demented older adults in a four-year cohort study. Moreover, this study explored the association of co-exposure to multiple pollutants with cognitive performance. In order to identify subpopulations that are more susceptible to the exposure of air pollutants, we further stratified these associations by apolipoprotein E (APOE) ɛ4 status to assess gene-environment interactions and stratified by other important covariates (e.g., sex and years of education) known to affect the risk of dementia [31].

METHODS

RESULTS

Spatial and temporal distributions of PM_2.5, PM _coarse , PM₁₀, and NO₂

The annual mean concentration of PM_2.5 (29.12μg/m³ for 11-year exposure) was higher than TEPA guidelines (15μg/m³), while the annual average level of PM₁₀ and NO₂ (50.42μg/m³ and 27.76 ppb for 23-year exposure, respectively) were below the air quality standards (65μg/m³ and 50 ppb, respectively; Supplementary Figure 1 and Supplementary Table 1). The annual mean level of PM_coarse was 24.14μg/m³ and TEPA does not have a corresponding standard for this pollutant. The spatial and temporal covariance models (Fig. 2) were used in BME framework, respectively. As the spatial and time lag increased, both the spatial and temporal covariance became smaller. The spatial and temporal covariances of PM_x are different. Among the spatial covariances, PM_2.5 has a wider spatial correlation range compared to PM_coarse and PM₁₀; spatial variability of PM₁₀ is smoother compared to PM_2.5 and PM_coarse. However, the smoother variability over time is present in PM_2.5 and PM_coarse, but not in PM₁₀. Elevated PM_2.5 level was observed in central western Taipei and Taoyuan Cities, possibly attributable to traffic and industrial emissions (Fig. 3).

OPEN IN VIEWER

Fig. 2

The spatial (upper)/temporal (lower) covariance model fitting used for Bayesian maximum entropy (BME) estimation: A) particulate matter≤2.5μm in diameter (PM_2.5); B) particulate matter > 2.5–10μm in diameter (PM_coarse); C) particulate matter≤10μm in diameter (PM₁₀); D) nitrogen dioxide (NO₂). X-axis indicates the spatial/temporal lag (km/years); Y-axis indicates the covariance of each air pollutant. Circles are estimated empirical covariance. Curved lines are fitted covariance models, which characterize the spatiotemporal dependence for the annual PM_2.5, PM_coarse, PM₁₀, and NO₂ exposure.

OPEN IN VIEWER

Fig. 3

Fitted covariance model used for Bayesian maximum entropy (BME) estimation in northern Taiwan: A) the average annual PM_2.5 exposure over 11 years, B) the average annual PM_coarse exposure over 11 years, C) the average annual PM₁₀ exposure over 23 years, and D) the average annual NO₂ exposure over 23 years. X-axis indicates the longitude of the study area; Y-axis indicates the latitude of the study area.

Stratified analyses by important covariates

For significant associations found between air pollutants and cognitive function over 4 years (Table 2), stratified analyses were performed by important covariates in Fig. 4 (sex, APOE ɛ4 status, and years of education). Although no significant interaction was observed between exposures to PM_2.5 or PM₁₀ and the aforementioned covariates on global or domain-specific cognition (Fig. 4), significant associations were found in some subgroups after stratification.

OPEN IN VIEWER

Fig. 4

Association between air pollutants and cognition over 4 years stratified by important variables (N = 360). A) PM_2.5 and global cognition by MoCA-T (binary), B) PM_2.5 and global cognition by MoCA-T (continuous), C) PM₁₀ and Trail Making Test A, D) PM_coarse and verbal fluency, and E) NO₂ and verbal fluency. Numbers in bold indicated statistically significant findings (p < 0.05)

For the association of PM_2.5 with global cognitive impairment, results became more evident in men (T3 versus T1: aOR = 5.72; p_trend =0.05, Fig. 4A), APOE ɛ4 non-carriers (T3 versus T1: aOR = 3.99; p_trend =0.03), and participants with education > 12 years (T3 versus T1: aOR = 8.10; p_trend =0.03). No significant associations were found in women, APOE ɛ4 carriers, and participants with education≤12 years. For continuous global cognition, results became more evident in women (T3 versus T1: β=–0.83, p_trend =0.02, Fig. 4B) and participants with education > 12 years (T3 versus T1: β=–0.72, p_trend =0.04). Associations did not reach statistical significance among men, both APOE ɛ4 carriers and non-carriers, and participants with education≤12 years.

For the association of PM₁₀ with poor performance of executive function, results became more evident in men (T2 versus T1: β = -0.22; T3 versus T1: β=–0.44; p_trend = 0.003, Fig. 4C), APOE ɛ4 non-carriers (T3 versus T1: β=–0.25; p_trend = 0.01), and participants with education≤12 years (T3 versus T1: β=–0.34; p_trend = 0.02). No significant associations were found in women, APOE ɛ4 carriers, and participants with education > 12 years. For the association of PM_coarse with poor performance of verbal fluency, results became more evident in women (T2 versus T1: β=–0.24; p_trend = 0.13, Fig. 4D), APOE ɛ4 non-carriers (T3 versus T1: β=–0.21; p_trend = 0.03), and participants with education > 12 years (T2 versus T1: β=–0.23; T3 versus T1: β=–0.28; p_trend = 0.02). No significant associations were found in men, APOE ɛ4 carriers, and participants with education≤12 years. For the association of NO₂ with better performance of verbal fluency, results became more evident in men (T2 versus T1: β=0.22; p_trend = 0.30, Fig. 4E), and participants with education > 12 years (T2 versus T1: β=0.15; p_trend = 0.66). No significant association was found in women, both APOE ɛ4 carriers and non-carriers, and participants with education≤12 years.

DISCUSSION

This study found that 11-year exposure to elevated PM_2.5 levels (annual mean higher than that established by TEPA) was associated with poor global cognition over four-year follow-up. An 11-year exposure to elevated PM_coarse levels or a 23-year exposure to elevated PM₁₀ levels (annual mean lower than that established by TEPA) was significantly associated with poor verbal fluency. These associations became more evident in women and APOE ɛ4 non-carriers, and participants with education > 12 years. Additionally, the co-exposure to PM_x and NO₂ was significantly associated with poor verbal fluency; co-exposure to PM₁₀ and NO₂ was significantly associated with poor executive function. Taken together, we found that PM₁₀, PM_coarse, PM_2.5, and NO₂ exposure affected different cognitive function (e.g., global or domain-specific cognition). Different cognitive domains were associated with PM_x exposure, which may be attributable to different sources of PM, adsorbed compounds presented on PM, geographic locations of sample collection, seasons, the deposit locations in the airway, and the ability of traverse traditional barriers in the lung or the blood-brain barrier [46]. In addition, this study found that the performance of verbal fluency was consistently affected by long-term PM_coarse, PM₁₀, and NO₂ exposure. Verbal fluency includes language, semantic memory, and executive function [47, 48] and is related to the prefrontal cortex (especially inferior frontal gyrus and middle frontal gyrus) of the brain [49, 50]. In sum, different air pollutants may be associated with either global or specific cognitive domains.

Air pollutants enter the human body through the olfactory bulb and nasal epithelial cells or via mechanical inhalation [6]. Some toxic compounds present on the surface of PM, such as manganese, polyaromatic hydrocarbons, and lipopolysaccharides, could affect the central nervous system through neurons directly or through the activation of microglial cells [46]. PM absorption could trigger the generation of proinflammatory cytokines and causes systematic or brain inflammation. The subsequent oxidative stress and its product, reactive oxygen species, can compromise the blood-brain barrier, which leads to the damage of neurons, the accumulation of Aβ, and the formation of senile plague, and subsequently affects cognitive function [46]. Instead of targeting specific cognitive domain(s), high-level PM_2.5 exposure may have moderate effect on each domain. However, the aggregated effect of PM_2.5 from all cognitive domains may explain the significant association between PM_2.5 exposure and the impairment of global cognition. In addition, Taiwan is an island country with subtropical climate, the high relative humidity may result in the adsorption of endotoxin on PM_coarse and PM₁₀ [51], which may partially explain our findings on long-term PM₁₀ or PM_coarse exposure affecting verbal fluency and executive function. Taken together, more studies are needed to clarify the toxicologic characteristics of air pollutants on different cognitive domains.

Our finding of 11-year PM_2.5 exposure with poor global cognition was consistent with previous studies (cohort studies: [8–14]; cross-sectional studies: [15–21]). In addition, we found that PM_coarse exposure was associated with poor verbal fluency. However, few studies explored verbal fluency in relation to PM_coarse exposure, and our finding was consistent with a previous cross-sectional study [17]. Furthermore, few studies explored executive function in relation to air pollution. We found that 23-year high-level PM₁₀ exposure was associated with poor executive function, which was consistent with a previous cohort study [28], but the other cohort study did not observe an association [10]. The significant findings we found were in Trail Making Test A but not Test B. It is possible that participants with high-level exposure to air pollutant had slower response in Trail Making Test A. In addition, after the assessment using Trail Making Test A, participants become more familiar with this type of test, which may also explain the non-significant finding in Trail Making Test B. In addition, this study found that exposure to elevated PM_coarse or co-exposure to elevated PM_x and NO₂ were associated poor verbal fluency or poor executive function (especially for PM₁₀ and NO₂ co-exposure), which should be attributable to the detrimental effect of PM_x exposure because NO₂ showed mild protective effect. Various brain areas have been related to the age-related decline in executive function (Trail Making Test), which include smaller grey matter volume in the dorsolateral and ventrolateral prefrontal cortex, medial prefrontal cortex, frontal pole, right inferior frontal gyrus, temporal lobe, insular cortex, caudate, globus pallidus, posterior parietal lobe, occipital cortex, and the cerebellum [52–54].

Furthermore, this study found that 23-year NO₂ exposure lower than TEPA standard was associated with better verbal fluency, which was inconsistent with limited previous studies (cross-sectional studies: [17, 20]). Among these studies, one study found inversed association (cross-sectional study: [17]) and two other found no association (cross-sectional studies: [18, 20]) between NO₂ exposure and verbal fluency. Previous animal studies revealed that exposure to NO₂ has shown to generate excitotoxicity, which affects the synaptic plasticity [55] and resulted in the accumulation and formation of Aβ and senile plaque via nitrification [56]; the subsequent neuronal apoptosis then affects cognitive function [56]. The protective finding in our study may be due to the long-term NO₂ exposure lower than the TEPA standard and the following rationale. NO₂ may participate with nitrotyrosination in the brain [57], which was observed in AD brain [58]. However, moderate nitrotyrosination has also been found in healthy people [59–61], which implies low-level nitration is needed to maintain physiological function [57]. As a whole, more experimental studies are needed to clarify the underlying mechanism of NO₂ exposure on human brain.

This study has the following strengths. First, since the progression of dementia may take decades, the collection of long-term exposure data started before the commencement of this study (from 2005–2015 for PM_2.5 and PM_coarse; from 1993–2015 for PM₁₀ and NO₂) and the design of cohort study allowed us to clarify the association between past exposure to air pollutants and cognitive performance over four-year follow-up. Secondly, few studies explored cognitive domains; assessment tools were also highly heterogeneous. This study assessed both global and domain-specific cognition (i.e., memory, executive function, verbal fluency, and attention), which allowed us to identify the effect of air pollutants on different cognitive domain(s). Thirdly, to account for the fact that multiple air pollutants coexist in the ambient environment, our analysis on the association between single-pollutant and cognitive function adjusted for other air pollutants. Fourthly, few studies explored the effect of co-exposure to air pollutants on cognition-related outcomes; we also assessed the co-exposure to these pollutants to clarify possible synergistic effect. Our exposure data were estimated from BME, which used both spatial and temporal data to estimate individual exposure over time; change in residency during the exposure period were noted to improve the accuracy of exposure estimation. Last but not least, this study included APOE ɛ4 status to assess important gene-environment interactions.

This study has some limitations; firstly, the study area was confined to Taipei metropolis, therefore, our findings may not be generalizable to other areas of Taiwan. Secondly, participation bias may occur because study participants were recruited from the health checkup program at baseline and tended to be healthier than the general population. However, after the four-year follow up, health status of these participants decreased and became similar to the general population. Therefore, we anticipate participation bias diminished over time in cohort studies [62, 63]. Thirdly, occupation before retirement may affect the association between exposure to air pollutants and the risk of cognitive impairment. After stratification by the occupational complexity, which was defined by Stern et al. [64] (low complexity: housewife, unskilled/semiskilled, skilled trade or craft, or clerical/office worker; high complexity: managerial, professional, or skilled workers), we found that the high-complexity group had similar findings with the whole population; the findings in the low-complexity group are less significant, especially in PM₁₀ and NO₂ exposure (data not shown). It is possible that almost all participants had been retired while they were recruited into this cohort during 2011–2013. Therefore, the effect of occupation may be attenuated as the number of years since retirement increased. Fourthly, this study used quantiles to categorize the levels of air pollutants in part because the results from the pollutant as a continuous variable was not significant, which may be attributable to the annual average exposure levels of PM₁₀, and NO₂ being lower than TEPA standards. It is also possible that the study participants, all of whom are community-dwelling older adults, have better overall health status compared to hospitalized patients; therefore, per unit increase in exposure concentration may not have a significant impact on their cognitive performance. In addition, chunking the continuous variable into increments of 10μg/m³ also resulted in the majority of the data in a single category. Therefore, this study adopted the quantile approach for grouping the exposure. Lastly, despite the estimation of long-term exposure to air pollutants, this study was unable to identify the critical time period in which cognitive function is most severely affected by exposure to air pollutants.

In conclusion, this four-year prospective study found that long-term (11 years) exposure to PM_2.5 levels higher than the current TEPA guideline values was significantly associated with cognitive impairment, signaling that the reduction of emission resources may be helpful to lower the risk of cognitive related outcome. In our single-pollutant exposure scenario, long-term (23 years) exposure to ambient PM₁₀ levels lower than the TEPA standards were associated poor cognitive function; co-exposure to both PM₁₀ and NO₂ revealed the same association. Notably, because these non-significant associations occurred at exposure concentrations lower than air quality guidelines established by TEPA, our results indicate that lowering the current national air quality standards for both PM₁₀ and NO₂ may be warranted in order to protect older adults from cognitive impairment. Equally important, the co-exposure of various air pollutants should be considered while evaluating the subsequent health outcomes in epidemiologic studies. Lastly, conducting prospective studies with long-term co-exposure period is suitable for exploring the association of air pollutants with cognitive function, which not only reflect the real-life exposure scenario to air pollutants but also allow us to clarify the effect of air pollution on the progression of cognitive impairment or dementia over time.