Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Nov 26.
Published in final edited form as: Nat Aging. 2021 May 3;1(5):430–437. doi: 10.1038/s43587-021-00060-4

Short-term air pollution, cognitive performance, and nonsteroidal anti-inflammatory drug use in the Veterans Affairs Normative Aging Study

Xu Gao 1,2,*, Brent Coull 3, Xihong Lin 3, Pantel Vokonas 4, Avron Spiro III 4,5,6, Lifang Hou 7, Joel Schwartz 8, Andrea A Baccarelli 2
PMCID: PMC8622756  NIHMSID: NIHMS1695634  PMID: 34841262

Abstract

Air pollution, especially the fine particulate matter (PM2.5), may impair cognitive performance1–3, but its short-term impact remains poorly understood. We investigated the short-term associations of PM2.5 with the cognitive performances of 954 white males measured as the global cognitive function (GCF) and Mini-Mental State Examination (MMSE) scores, and further explored whether taking nonsteroidal anti-inflammatory drugs (NSAIDs) could modify their relationships. Higher short-term exposure to PM2.5 demonstrated non-linear negative associations with cognitive function. Compared with the lowest quartile of the 28-day average PM2.5 concentration, the 2nd, 3rd, and 4th quartiles were associated with 0.378-, 0.376-, and 0.499-unit decreases in GCF score, 0.484-, 0.315-, and 0.414-unit decreases in MMSE score, and 69%, 45%, and 63% greater odds of low MMSE scores (≤25), respectively. Such adverse effects were attenuated among NSAIDs users compared to non-users. This study elucidates the short-term impacts of air pollution on cognition and warrants further investigations on the modifying effects of NSAIDs.

The global shift toward an older population and the dramatically increased aging pace across the globe have sparked a growing interest in “successful aging”, the capacity of elderly people to live without major diseases and disability throughout their later years4. Premature decline in cognitive function threatens successful aging, and the link between long-term PM2.5 (PM with an aerodynamic diameter of <2.5μm) exposure and impaired cognitive performance in the aging population is well-established. Reported effects include reduced brain volume, cognitive decrements, and dementia development1,3. For instance, older women had nearly twice the risk of dementia if the average PM2.5 levels exceeded 12 μg/m3 in the preceding three years in a U.S.-wide cohort5. Another study based on a national sample of older U.S. adults found that older adults living in areas with high annual concentrations of PM2.5 had a 1.5-time-greater error rate in working memory and orientation tests than those exposed to lower concentrations6. Importantly, over the past few decades, annual average PM2.5 levels have been substantially lowered through large-scale emissions control policies7, which could moderate some of the adverse health effects of air pollution. Average annual ambient concentrations of black carbon (BC), a key component of PM2.5 and a tracer of vehicular traffic, have also been associated with poor cognition of children and adults8,9.Yet, even in regions with air pollution levels within the yearly mandated average standards, short-term (days to weeks) peaks of air pollution are frequently reported, along with adverse health consequences10,11. Harms caused by short-term air pollution on cognitive health are expected to be worse at locations worldwide with poorer air quality. Therefore, there is a dearth of research examining the impact of short-term PM2.5 exposure on cognitive performance in aging individuals.

Cost-effective approaches to prevent the potential detrimental impacts of air pollution on cognitive aging are critical with high public health value12. One of the most commonly accepted mechanisms of pollution-related cognitive decline is neuroinflammation resulting from inhaled airborne particles that directly enter the brain via nerve connections in the olfactory system13,14. It may trigger vascular/endothelial dysfunction15,16 and may promote cognitive impairment17–19. Many studies have investigated nonsteroidal anti-inflammatory drugs (NSAIDs), especially aspirin, as a potential treatment regimen for cognitive dysfunction and dementia, but mostly yielded null and inconclusive findings20,21. However, the use of NSAIDs as an intervention approach to moderate the influences of exogenous exposures, including air pollution, on cognitive health has not yet been investigated.

To address these knowledge gaps, we investigated the associations of short-term (≤28 days) ambient PM2.5 and BC levels with cognitive performances measured by global cognitive function (GCF) and Mini-Mental State Examination (MMSE) scores. We further sought to test whether NSAID use modified the PM–cognitive performance relationships. This investigation was undertaken in the Normative Aging Study (NAS), a cohort of older men from the Greater Boston area.

The mean participants’ age (±standard deviation) of the 954 participants at the initial visit was 69.22±7.12 years (Table S1). About 64% of individuals were former smokers, and 31% were never smokers. Most participants were obese or overweight, consumed <2 alcoholic drinks per day, received ≥13 years of education, were native English speakers, and had computer experience. A total of 165 participants (~17%) had a low MMSE score (≤25). Only 3.9% of individuals used corticosteroids. Over 75% of all participants were NSAIDs users, most of whom took aspirin only. NSAID users had higher GCF and MMSE scores, as well as lower odds of low MMSE scores, than non-users of NSAIDs. A total of 2551 visits from the 954 participants were employed to perform the following analyses. Binary 28-day average PM2.5 level was highly related with seasons of visits but not with the use of NSAIDs. Average PM concentrations over the 28-day exposure window remained stable (Table S2). The 28-day average levels were 10.77±3.04 μg/m3 (IQR=4.06) for PM2.5 and 0.92±0.35 μg/m3 (IQR=0.43) for BC. PM2.5 and BC levels were highly correlated with each other across exposure windows (all p-values <0.0001).

We first determined the associations of ambient PM levels with GCF score, MMSE score, and odds of low MMSE score by treating the PM levels as continuous variables. Higher PM2.5 levels at each exposure window were associated with lower GCF and MMSE scores and greater odds of low MMSE score (Table 1 and Figure 1). Associations of PM2.5 with GCF score across the exposure windows were weaker than the associations of PM2.5 with MMSE and low MMSE score odds. Specifically, an IQR increase in the 28-day average PM2.5 concentration was significantly associated with an 0.169-unit decreases in MMSE score and 21% greater odds of low MMSE scores, but was not significantly associated with reduced GCF score. With restricted cubic spline regression (Figure 2, Extended data figure 1), we found that the PM2.5–cognitive performance relationships were not linear. At lower PM2.5 levels (≤~10 μg/m3), GCF and MMSE scores dropped dramatically with increased PM2.5 levels, but the change remained stable at relatively higher PM2.5 levels (>~10 μg/m3). Further, increased odds of low MMSE scores were observed mostly in response to lower PM2.5 levels.

Table 1.

Associations of PM2.5 levels with GCF score, MMSE score, and odds of low MMSE scoresa

Exposure window PM2.5 GCF score MMSE score Low MMSE scores
Coefficients (SE) p-value Coefficients (SE) p-value Nvisit / NLow MMSE Odds ratio (95% CI) p-value
Same day Per IQR change −0.005 (0.062) 0.94 −0.024 (0.038) 0.53 2,551 / 490 1.09 (0.98 – 1.21) 0.12
Quartiles Q1 Ref Ref 637 / 124 Ref
Q2 −0.272 (0.163) 0.10 −0.128 (0.101) 0.20 639 / 118 1.04 (0.78 – 1.38) 0.81
Q3 −0.287 (0.166) 0.08 −0.212 (0.103) 0.039 637 / 126 1.30 (0.98 – 1.73) 0.07
Q4 −0.042 (0.176) 0.81 −0.089 (0.109) 0.42 638 / 122 1.25 (0.92 – 1.69) 0.16
7-day Per IQR change −0.117 (0.080) 0.14 −0.172 (0.050) 0.0005 2,551 / 490 1.23 (1.07 – 1.41) 0.0030
Quartiles Q1 Ref Ref 638 / 123 Ref
Q2 −0.216 (0.163) 0.19 −0.137 (0.101) 0.18 636 / 108 1.01 (0.75 – 1.38) 0.93
Q3 −0.294 (0.167) 0.08 −0.217 (0.103) 0.035 639 / 133 1.44 (1.07 – 1.95) 0.017
Q4 −0.286 (0.178) 0.11 −0.394 (0.110) 0.0003 638 / 126 1.52 (1.10 – 2.11) 0.011
14-day Per IQR change −0.097 (0.086) 0.26 −0.181 (0.053) 0.0007 2,551 / 490 1.23 (1.06 – 1.43) 0.0060
Quartiles Q1 Ref Ref 638 / 116 Ref
Q2 −0.280 (0.165) 0.09 −0.290 (0.102) 0.0046 636 / 118 1.23 (0.92 – 1.66) 0.16
Q3 −0.458 (0.169) 0.0073 −0.419 (0.105) <0.0001 639 / 139 1.81 (1.35 – 2.42) <0.0001
Q4 −0.330 (0.186) 0.07 −0.414 (0.115) 0.0003 638 / 117 1.55 (1.12 – 2.17) 0.0091
21-day Per IQR change −0.138 (0.091) 0.13 −0.196 (0.056) 0.0005 2,551 / 490 1.26 (1.08 – 1.48) 0.0038
Quartiles Q1 Ref Ref 637 / 117 Ref
Q2 −0.360 (0.165) 0.029 −0.351 (0.102) 0.0006 638 / 122 1.37 (1.03 – 1.83) 0.032
Q3 −0.394 (0.171) 0.022 −0.347 (0.106) 0.0011 639 / 124 1.50 (1.11 – 2.03) 0.0086
Q4 −0.492 (0.190) 0.0095 −0.464 (0.117) <0.0001 637 / 127 1.77 (1.27 – 2.48) 0.0008
28-day Per IQR change −0.145 (0.092) 0.12 −0.169 (0.057) 0.0029 2,551 / 490 1.21 (1.03 – 1.42) 0.019
Quartiles Q1 Ref Ref 638 / 113 Ref
Q2 −0.378 (0.166) 0.023 −0.484 (0.103) <0.0001 637 / 135 1.69 (1.27 – 2.26) 0.0004
Q3 −0.376 (0.172) 0.028 −0.315 (0.106) 0.0029 638 / 122 1.45 (1.07 – 1.97) 0.016
Q4 −0.499 (0.188) 0.0080 −0.414 (0.116) 0.0004 638 / 120 1.63 (1.17 – 2.28) 0.0040
Open in a new tab
a:

Analyzed with mixed linear regression (GCF and MMSE scores) and logistic regression (odds of low MMSE scores) models with random participant-specific intercepts. Models adjusted for age (years), BMI (underweight or normal weight/overweight/obese), smoking status (current/former/never smoker), alcohol intake (<2 drinks per day or ≥2 drinks per day), hypertension, coronary heart disease, diabetes, education (≤12 years, 13–16 years, >16 years), English as first language (yes/no), computer experience (yes/no), ambient temperature (°C) and relative humidity (%). P-values were two-sided and multiple comparisons were not adjusted. Bolded p-values are <0.05.

Figure 1. Associations of PM2.5 levels with GCF score, MMSE score, and odds of low MMSE scores in the exposure window of 28 days.

Figure 1

Open in a new tab

Analyses were performed based on the data of 2551 medical visits from 954 participants. Data are presented as point estimates of effects +/− 1.96 standard errors (i.e. 95% confidence intervals). Central dots of error bar: point estimates of effects; corresponding lines: 95% confidence levels; * the point estimates in 2.a, and 3.a which were statistically significant (a two-sided p-value <0.05, multiple comparisons were not adjusted). Detailed estimates and p-values please see Tables 1 and S5.

Figure 2. Best-fitting models for the relationships of 28-day average PM2.5 levels with GCF score, MMSE score, and odds of low MMSE score.

Figure 2

Open in a new tab

Analyses were performed based on the data of 2551 medical visits from 954 participants. Solid line: point estimates; dash line: confidence intervals; dot: knots; green line: reference line

Therefore, we evaluated PM2.5–cognitive performance relationships using the quartiles of PM2.5 concentrations. Compared with the lowest quartile (≤8.65 μg/m3) of the 28-day average PM2.5 level, the 2nd (8.65~10.35 μg/m3), 3rd (10.35~12.71 μg/m3), and 4th quartiles (>12.71 μg/m3) were robustly associated with 0.378-, 0.376-, and 0.499-unit decreases in GCF score, 0.484-, 0.315-, and 0.414-unit decreases in MMSE score, and 69%, 45%, and 63% greater odds of low MMSE scores, respectively (Table 1, all p-values <0.05). Analyses of BC–cognitive performance relationships yielded similar patterns (Figures S1–S2, Table S3). Increased BC levels up to 28 days before the visits were robustly associated with lower GCF scores, but had much weaker associations with MMSE scores and lower odds of low MMSE scores compared with increased PM2.5 levels.

After adjusting for potential covariates, none of the measures of cognitive performance were associated with use of NSAIDs or aspirin (Table S4). However, NSAID users were less affected by adverse impacts of short-term ambient PM levels on cognitive performance than non-users. Higher PM2.5 levels were associated with lower GCF and MMSE scores, and greater odds of low MMSE scores in non-users of NSAIDs, but these effects were vigorously attenuated among NSAID users (Figures 1, 3 & Extended data figure 2). In non-users of NSAIDs, compared with the lowest quartile of 28-day average PM2.5 levels, the 2nd, 3rd, and 4th quartiles were statistically associated with 0.889-, 0.987-, and 1.416-unit decreases in GCF score, 0.938-, 0.868-, and 0.894-unit decreases in MMSE score, and 131%, 210%, and 128% greater odds of low MMSE scores, respectively (Table S5). However, the same quartiles were associated with 0.267-, 0.252-, and 0.292-unit decreases in GCF score, 0.347-, 0.144-, and 0.237-unit decreases in MMSE score, and 59%, 16%, and 44% greater odds of low MMSE scores among NSAID users (Table S5). Although the p-values of interaction terms varied, similar patterns were observed across the exposure windows. The modifying effects were robust for GCF and MMSE scores, but relatively weaker for the odds of low MMSE scores. Similar modifying effects were also observed in analyses for BC (Figures S1 & S3, Table S6).

Figure 3. Best-fitting models for the relationships of 28-day average PM2.5 levels with GCF score, MMSE score, and odds of low MMSE score, by NSAID use.

Figure 3

Open in a new tab

Analyses were performed based on the data of 2551 medical visits from 954 participants. Solid line: point estimates; dash line: confidence intervals; dot: knots; green line: reference line

Looking specifically at aspirin use (Figure S4), the “non-aspirin NSAIDs only” and “aspirin+non-aspirin NSAIDs” subgroups showed somewhat modifying effects on most associations between PM levels and cognitive performance, with much wider CIs than the “aspirin only” subgroup. The “aspirin-only” group showed similar trends to NSAID users, suggesting that the modifying effect of NSAID use in our study was mainly driven by aspirin.

Sensitivity analyses with inverse probability weights (IPWs) further validated the reliability of our main findings. Weighted primary analyses and subgroup analyses yielded essentially unchanged estimates of the PM-cognitive performance relationships (Table S7), suggesting that our primary findings were not biased by loss to follow-up or survivor bias. In another sensitivity analysis with an additional adjustment of the main analysis models for 1-year PM2.5 average levels, we observed that the associations between short-term ambient PM2.5 concentrations and cognitive performance were only slightly attenuated but still robust, especially for the MMSE score and the odds of low MMSE scores (Table S8). This indicates that the longer-term average PM levels may influence our findings to a limited extent and our main findings may essentially reflect the short-term impact of PM exposure on cognitive performance. In a sensitivity analysis using a propensity score for NSAIDs use, we found that the effects of categorized PM2.5 levels on cognitive performance by NSAID use (Table S9) were slightly enhanced, robust, and mostly showing patterns similar to the corresponding main results (Table S5). This suggests that residual bias, including prescription bias, may affect our primary findings only to a limited degree.

Although evidence linking impaired cognitive function with long-term air pollution exposure is accumulating rapidly, the critical exposure windows in the air pollution-cognitive performance relationships remain uncertain3. Most previous studies were based on exposures over one month to 5 years before cognitive testing and found consistent inverse associations of PM2.5 and BC with cognitive function1,3. However, short-term peaks of air pollution also harm cognitive function. Our investigation provided more evidence on this point of view by demonstrating the change of cognitive function under short-term PM2.5 and BC exposures in a relatively large population. Our findings are consistent with an experimental study that found robust declines in MMSE and Ruff 2 & 7 tests scores of 63 participants after 1-hour-long exposures of candle burning and outdoor commuting22. Marginal evidence for the general relationship between air pollution and brain health also highlighted the critical role of short-term air pollution spikes. Even though PM2.5 data were not available, Lo et al. observed a trend of increased severe cognitive impairment under short-term exposures to PM10 and other air pollutants in a longitudinal survey in Asians23. In other reports, PM2.5 exposure in the 2 days before evaluation was related to an increased risk of hospitalization for Parkinson’s disease (PD)24, and PM2.5 exposure within 7 days before evaluations aggravated PD in 391 patients25. Along with our findings of the sensitivity analysis with annual average PM concentrations, these studies suggest that the effects of short-term PM exposures on cognitive ability and brain health may be in the same direction as those from long-term exposures. Additionally, our results underscore that such short-term cognitive function impairment may happen under PM2.5/BC concentrations (PM2.5 <~10 μg/m3; BC <~1 μg/m3) below the levels that regulators consider acceptable. Notably, the observed BC–cognitive function relationships were slightly different from those of PM2.5, though with similar patterns. Considering that BC is a major component of air pollution in our study area but only contributed to ~7% of the PM2.5 mass at Boston26, this difference we observed in this study could be explained by other components of the PM2.5 mixture27, such as sulfates, sodium, and calcium, which may distort the overall estimates of PM2.5. Future studies with available PM component data will be worthwhile to validate our findings and dissect the effect of each PM component on cognitive performance.

Intriguingly, our study showed that aging men taking NSAIDs experienced fewer adverse short-term impacts of PM exposures on cognitive health than non-users, though we found no direct associations between recent NSAID use and cognitive performance. The lack of a direct association is consistent with a recent review that summarized four relevant randomized clinical trials and found no evidence supporting the use of low-dose aspirin or other NSAIDs of any class for the prevention of dementia20. For instance, in the ASPREE trial, low-dose aspirin did not prolong disability-free survival or reduce all-cause dementia over 5 years but increased the risk of major hemorrhage28. Collectively, we suggest that NSAID use may mainly affects the inverse associations between PM levels and cognitive performance as an effect modifier. There are two potential explanations for this modifying effect. First, NSAIDs, especially aspirin, may moderate the neuroinflammation triggered by PM inhalation. Specifically, aspirin may dampen the inflammatory response resulting from PM exposures and reduce levels of inflammatory biomarkers, including C-reactive protein and interleukin-629,30. Elevated levels of the two inflammatory markers were negatively associated with a composite score of executive function and processing speed31. Hence, the impact of PM on cognitive performance may be attenuated in NSAID users through a decrease in the inflammatory response in the nervous system. However, our paper provides no evidence about inflammation as a potential mechanism linking PM exposure to cognitive impairment, nor about its role in mediating NSAID’s moderating effects. Another possibility is that NSAIDs may moderate PM-induced disturbances in cerebrovascular hemodynamics32, which are suspected to be related to cognitive impairment33. A study in rats showed that aspirin may reduce cerebral hemodynamic disturbances induced by global ischemia34. However, the biological mechanisms underlying PM-associated cognitive dysfunction remain poorly understood2,3. The potential alleviation of PM-related cognitive dysfunction in NSAID users observed in our study may facilitate the identification of these underlying biological mechanisms.

Major strengths of our study include detailed information on a broad range of covariates and multiple measures of ambient PM concentrations and cognitive function. Several limitations are notable when interpreting the results. First, the NAS is an aging cohort but was not specifically designed for pharmacoepidemiology. We collected information on recent NSAID use but not the details of the dose and duration of NSAID use. This may curb the generalization of our findings to long-term NSAID use and to their cumulative dose, as it limited our capacity to explore whether the modifying effects of NSAIDs varied by the duration of NSAID use. Residual bias, including that deriving from prescription bias should also be acknowledged as individuals using NSAIDs may be inherently different from those who did not. One of our sensitivity analyses addressed this issue by adding propensity scores of NSAID use to the main analysis models. These models, which we used in the attempt of minimizing the influences of residual bias to the greatest extent, confirmed our findings, suggesting that the influence of residual bias was very limited in our main findings. Furthermore, although the overall sample size was relatively large, some of the non-significant results with relatively large CIs may be attributable to limited statistical power in the relatively smaller subgroups. This specifically restricted our capacity to fully understand the effect of non-aspirin NSAIDs on the PM–cognitive function relationships. Multiple comparisons may also induce false-positive findings, but we observed consistent significant associations between certain PM levels and cognitive function across each time window and drug use subgroups, which may indicate that our findings were not incidental. Additionally, Boston has its unique chemical components of air pollution for its geographical location and the effects of those components on cognitive function are not yet evaluated. Both may limit the generalization of our findings in other areas with different sources of pollutants. We also acknowledge that our analysis is subject to measurement bias in that the levels of PM obtained from a single site in Boston that we utilized may differ from that at the participants’ address and/or their personal exposure. Nevertheless, a previous Boston study found that the correlations of ambient PM2.5 concentrations and the corresponding personal PM exposure were fairly high35, and such measurement bias from using the data of a single site will result in primarily Berkson-type measurement error, which may bias the standard errors but not the estimated associations of our primary findings36. Given the majority of NAS participants are retired and spend most of their time at their residencies, we believe that the discrepancies between the city-average and personal PM data are likely to be non-differential and to bias results toward the null, rather than causing the observed associations. Lastly, participants in this study were older white men, which suggests the possibilities that the results might not be generalized to other racial/ethnic groups or/and women.

Despite the fact that air pollutant emissions have been regulated by governments and regulatory agencies for decades, short-term spikes of air pollution remain frequent and may impair health. Our study indicates that short-term air pollution exposure may be related to short-term alterations in cognitive function and that NSAIDs may modify this relationship. Nevertheless, we were not able to fully distinguish the potential residual impact of longer-term PM exposure on cognitive health in this study, thus future analyses to investigate whether cognitive impairments are transient or persistent in over the years would be of high scientific significance. Our findings are also important for other locations around the world where air quality is poorer than in the United States and the impact of PM exposure on cognitive health is expected to be heavier. Our study warrants future investigations on NSAID use to assist the prevention of aging-related health outcomes resulting from exogenous exposures (e.g. air pollution, smoking) instead of as a treatment regimen. Given the cross-sectional nature and subclinical findings of our study, our analyses with propensity score may not be able to completely rule out prescription bias and other unmeasured confounding. Therefore, multidisciplinary studies based on larger cohorts with more detailed NSAID usage information, as well as randomized clinical trials of NSAID use are strongly required to confidently validate the short-term PM–cognitive function relationships identified by our study and further elucidate the modifying effect of NSAIDs we identified in our study.

Methods

Study design and population

Established in 1963, the NAS is a cohort of 2,280 men from the Greater Boston area, where air pollution data have been collected since 199537,38. Participants were free of known chronic medical conditions at initial health screening (1961–1970), and completed detailed on-site physical examinations and questionnaires every 3–5 years on a rolling basis. Due to the small proportion of non-white participants in the NAS (N=29), we analyzed only white participants to increase statistical power. A total of 2,551 visits from 954 participants without a stroke history from 1995–2012 with available air pollution, cognitive function test, and NSAID use data, and lived in the Greater Boston area during the study period were eligible for the present study. The NAS was approved by the Institutional Review Boards of the VA Boston Healthcare System and Columbia University, and each participant provided written informed consent. This report followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Data collection

Participants were asked to provide detailed information about their lifestyles, activity levels, and demographic factors37. Body-mass-index (BMI [kg/m2]) was calculated from height and weight. Hypertension was defined as 1) a systolic blood pressure ≥160 mm Hg; 2) a diastolic blood pressure ≥95 mm Hg; and/or 3) use of any hypertension medication. Coronary heart disease (CHD) included physician diagnosed myocardial infarction and angina pectoris. Diabetes was defined as 1) physician diagnosis; 2) a fasting blood glucose >126 mg/dl, and/or 3) use of any diabetes medication. Recent use of NSAIDs over the past month (“no”/“yes”) was recorded at each visit with questionnaires (Table S10) and further classified into four groups based on aspirin use (“no”/“yes, non-aspirin NSAIDs only”/“yes, aspirin only”/“yes, aspirin+non-aspirin NSAIDs”). Corticosteroid use (“no”/“yes”) was also obtained at each visit to in an attempt to account for potential confounding from other anti-inflammatory medications.

Cognitive testing

Cognitive tests in the NAS began in 19938. To estimate GCF, we used four individual general cognitive tests administered at most cognitive assessments: the word list memory task test (including 2 unique scores), the digit span backward test, the verbal fluency test, and the sum of drawings test. Sums of the z-scored five scores were the GCF score, with positive scores indicating better cognitive performance. We also conducted the MMSE, a widely used screening test for dementia8. The MMSE assesses overall cognition by testing several domains including memory, visuospatial ability, attention, language, and orientation. This test includes 30 questions, but the maximum MMSE score in this study was 29 owing to exclusion of a county identification question with poor participant performance39. Because of the change of the maximum MMSE score, we dichotomized the MMSE scores to normal (>25) and low (≤25), as in previous analyses of the NAS cognitive data39,40.

Exposure assessments

We used PM2.5 (μg/m3) and BC (μg/m3) levels on the day of each visit and mean values at 7, 14, 21, and 28 days before each visit. PM2.5 and BC concentrations were measured hourly at the Harvard University supersite located on the Countway Library of Medicine, which is 50 m from the nearest street, ~1 km from the examination site, and with a median distance of ~20 km with participants’ residencies in the Greater Boston area41, using tapered element oscillation microbalances (Model 1400A, Rupprecht and Pastashnick) and aethalometers (Magee Scientific Co., Model AE-16). Given the longitudinal correlations between daily personal PM exposure and daily ambient concentrations are high in our study area35, we assumed that the measures of ambient PM concentrations could serve as surrogates of PM exposures of participants at their home addresses. We also obtained temperature and relative humidity data from the National Weather Service Station at Logan Airport (Boston, MA, USA), located approximately 12 km from the examination center41.

Statistical analysis

Descriptive statistics were used to summarize sociodemographic, lifestyle factor, and cognitive function measures for all visits and for visits by NSAID use.

We first evaluated the associations of PM levels with GCF and MMSE scores using linear mixed-effects regression and the odds of low MMSE scores using logistic regression. We employed random participant-specific intercepts accounting for the correlation of repeated measures of individuals across the study period in all models and treated each GCF score as a repeated measure of underlying global cognition8. We adjusted for basic covariates in all models including: age (years), BMI (underweight or normal weight/overweight/obese), smoking status (current/former/never smoker), alcohol intake (<2 drinks or ≥2 drinks per day), hypertension, CHD, diabetes, education (≤12 years, 13–16 years, >16 years), English as the first language (yes/no), and computer experience (yes/no). We also included ambient temperature (degree Celsius), relative humidity (%), and seasons of study visit in regression models because: 1) PM concentrations are season-related: PM levels are higher in the cold season and negatively related to relative humidity42, and 2) temperature and humidity may also be associated with cognitive performance according to previous studies43. Corresponding PM2.5 concentrations (for BC models only) were also included in models for this evaluation to adjust for the effects from other unmeasured components of PM2.5 other than BC. Estimates of the effects of PM concentrations on cognitive performance were reported as changes per interquartile range (IQR) increase in PM levels. Dose-response curves and linearity tests for PM–cognitive performance relationships were further assessed by restricted cubic spline regression44. Models were adjusted for the previously described covariates and the 25th, 50th, and 75th percentiles were selected as knots. If non-linear relationships were recognized, models using the quartiles of ambient PM levels as predictors were used to further explore the PM–cognitive performance relationships.

Similar models were then applied to examine whether there were direct associations between NSAIDs/aspirin use and cognitive performances (GCF score, MMSE score, and the odds of low MMSE). Models were adjusted for the basic covariates plus the use of corticosteroids.

Finally, we examined the modifying effects of NSAID/aspirin use on the PM–cognitive performance relationships. Models included the main effect terms of PM (PM2.5 or BC) and drug (NSAIDs or aspirin) use, as well as their interaction term “PM*drug use”, which was used to test the modifying effects of using NSAIDs or aspirin. Models were further adjusted for the basic covariates plus corticosteroids use, corresponding ambient temperature (degree Celsius), relative humidity (%), and PM2.5 concentrations (for BC models only). Subgroup analyses stratified by NSAID or aspirin use were then conducted to estimate the effects of ambient PM levels (per IQR change) on cognitive performance in each subgroup. Dose-response curves and linearity tests in subgroups were also performed with restricted cubic spline regression.

Sensitivity analyses

We performed three sensitivity analyses to test the robustness of our primary findings. 1) Because healthier study participants are more likely to participate in subsequent examinations over time, to evaluate the validity of the missing at random assumption and to assess the impact of potential selection bias caused by non-random unavailability for follow-up, we used inverse probability weighting as a sensitivity analysis to correct for potential survival bias45. Weighted models simultaneously adjusting for the inverse probability weights (IPWs) and the previously introduced covariates were conducted to validate the robustness of our primary findings. 2) Furthermore, to test whether the associations of short-term PM and cognitive function were biased by the effects of longer-term PM concentrations, we estimated the 1-year average PM2.5 level of each visit and then categorized and added to the main regression model a binary variable obtained using the cutoff of 12 μg/m3, the annual standard of PM2.5 concentrations mandated by the U.S. Environmental Protection Agency. 3) Finally, to further controlling for confounding from residual bias including potential prescription bias, we estimated the propensity scores of NSAID use at each visit among our participants. We then added the propensity score to the main regression model used to test the modifying effect of NSAIDs on the PM2.5-cognitive performance relationship. The estimation of the propensity score is shown in Table S6.

SAS version 9.4 TS1M5 (SAS Institute Inc., Cary, NC, USA) was used to perform data cleaning and all analyses. A two-sided p-value <0.05 was considered statistically significant.

Extended Data

Extended Data Fig. 1. Best-fitting models for the relationships of PM2.5 levels with GCF score, MMSE score, and odds of low MMSE score in the exposure window of 21 days.

Extended Data Fig. 1

Open in a new tab

Analyses were performed based on the data of 2551 medical visits from 954 participants. Solid line: point estimates; dash line: confidence intervals; dot: knots; green line: reference line

Extended Data Fig. 2. Best-fitting models for the relationships of PM2.5 levels with GCF score, MMSE score, and odds of low MMSE score in the exposure window of 21 days, by NSAID use.

Extended Data Fig. 2

Open in a new tab

Analyses were performed based on the data of 2551 medical visits from 954 participants. Solid line: point estimates; dash line: confidence intervals; dot: knots; green line: reference line

Supplementary Material

1695634_Supp_Material
1695634_Reporting_Summary

Acknowledgments

This work was supported by the National Institute of Environmental Health Sciences (grants P30ES009089, R01ES021733, R01ES025225, R01ES015172, and R01ES027747). The VA Normative Aging Study is supported by the Cooperative Studies Program/Epidemiology Research and Information Center of the U.S. Department of Veterans Affairs and is a component of the Massachusetts Veterans Epidemiology Research and Information Center, Boston, Massachusetts. Dr. Spiro was supported by a Senior Research Career Scientist award from the Clinical Science R&D Service of the U.S. Department of Veterans Affairs.

Footnotes

Competing interests statement

The authors declare that they have no competing interests.

Code availability: Codes of SAS version 9.4 TS1M5 for statistical analysis are available upon request from Dr. Xu Gao.

Data availability: The data that support the findings of this study are available on reasonable request from Dr. Andrea A. Baccarelli. The data are not publicly available due to restrictions of ethics approval requirements for this study.

References

  • 1.Power MC, Adar SD, Yanosky JD & Weuve J Exposure to air pollution as a potential contributor to cognitive function, cognitive decline, brain imaging, and dementia: A systematic review of epidemiologic research. Neurotoxicology 56, 235–253, doi: 10.1016/j.neuro.2016.06.004 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Griffiths CJ & Mudway IS Air pollution and cognition. Bmj 363, k4904, doi: 10.1136/bmj.k4904 (2018). [DOI] [PubMed] [Google Scholar]
  • 3.Schikowski T & Altug H The role of air pollution in cognitive impairment and decline. Neurochem Int 136, 104708, doi: 10.1016/j.neuint.2020.104708 (2020). [DOI] [PubMed] [Google Scholar]
  • 4.World Health Organization. World report on ageing and health. Geneva: World Health Organization; (2015). [Google Scholar]
  • 5.Cacciottolo M et al. Particulate air pollutants, APOE alleles and their contributions to cognitive impairment in older women and to amyloidogenesis in experimental models. Transl Psychiatry 7, e1022, doi: 10.1038/tp.2016.280 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Ailshire JA & Clarke P Fine particulate matter air pollution and cognitive function among U.S. older adults. J Gerontol B Psychol Sci Soc Sci 70, 322–328, doi: 10.1093/geronb/gbu064 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.World Health Organization. Air quality guidelines: global update 2005: particulate matter, ozone, nitrogen dioxide, and sulfur dioxide. (World Health Organization,, 2006). [PubMed] [Google Scholar]
  • 8.Power MC et al. Traffic-related air pollution and cognitive function in a cohort of older men. Environmental health perspectives 119, 682–687, doi: 10.1289/ehp.1002767 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Suglia SF, Gryparis A, Wright RO, Schwartz J & Wright RJ Association of black carbon with cognition among children in a prospective birth cohort study. American journal of epidemiology 167, 280–286, doi: 10.1093/aje/kwm308 (2008). [DOI] [PubMed] [Google Scholar]
  • 10.Atkinson RW et al. Short-term exposure to traffic-related air pollution and daily mortality in London, UK. J Expo Sci Environ Epidemiol 26, 125–132, doi: 10.1038/jes.2015.65 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Guo B et al. Using rush hour and daytime exposure indicators to estimate the short-term mortality effects of air pollution: A case study in the Sichuan Basin, China. Environ Pollut 242, 1291–1298, doi: 10.1016/j.envpol.2018.08.028 (2018). [DOI] [PubMed] [Google Scholar]
  • 12.Laumbach R, Meng Q & Kipen H What can individuals do to reduce personal health risks from air pollution? Journal of thoracic disease 7, 96–107, doi: 10.3978/j.issn.2072-1439.2014.12.21 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Calderon-Garciduenas L et al. Brain inflammation and Alzheimer’s-like pathology in individuals exposed to severe air pollution. Toxicol Pathol 32, 650–658, doi: 10.1080/01926230490520232 (2004). [DOI] [PubMed] [Google Scholar]
  • 14.Calderon-Garciduenas L et al. DNA damage in nasal and brain tissues of canines exposed to air pollutants is associated with evidence of chronic brain inflammation and neurodegeneration. Toxicol Pathol 31, 524–538, doi: 10.1080/01926230390226645 (2003). [DOI] [PubMed] [Google Scholar]
  • 15.Delfino RJ et al. Circulating biomarkers of inflammation, antioxidant activity, and platelet activation are associated with primary combustion aerosols in subjects with coronary artery disease. Environmental health perspectives 116, 898–906, doi: 10.1289/ehp.11189 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Bind MA et al. Air pollution and markers of coagulation, inflammation, and endothelial function: associations and epigene-environment interactions in an elderly cohort. Epidemiology 23, 332–340, doi: 10.1097/EDE.0b013e31824523f0 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Vendemiale G, Romano AD, Dagostino M, de Matthaeis A & Serviddio G Endothelial dysfunction associated with mild cognitive impairment in elderly population. Aging Clin Exp Res 25, 247–255, doi: 10.1007/s40520-013-0043-8 (2013). [DOI] [PubMed] [Google Scholar]
  • 18.De Silva TM & Faraci FM Microvascular Dysfunction and Cognitive Impairment. Cell Mol Neurobiol 36, 241–258, doi: 10.1007/s10571-015-0308-1 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Block ML, Zecca L & Hong JS Microglia-mediated neurotoxicity: uncovering the molecular mechanisms. Nat Rev Neurosci 8, 57–69, doi: 10.1038/nrn2038 (2007). [DOI] [PubMed] [Google Scholar]
  • 20.Jordan F et al. Aspirin and other non-steroidal anti-inflammatory drugs for the prevention of dementia. Cochrane Database Syst Rev 4, CD011459, doi: 10.1002/14651858.CD011459.pub2 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Veronese N et al. Low-Dose Aspirin Use and Cognitive Function in Older Age: A Systematic Review and Meta-analysis. Journal of the American Geriatrics Society 65, 1763–1768, doi: 10.1111/jgs.14883 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Shehab MA & Pope FD Effects of short-term exposure to particulate matter air pollution on cognitive performance. Scientific reports 9, 8237, doi: 10.1038/s41598-019-44561-0 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Lo YC, Lu YC, Chang YH, Kao S & Huang HB Air Pollution Exposure and Cognitive Function in Taiwanese Older Adults: A Repeated Measurement Study. Int J Environ Res Public Health 16, doi: 10.3390/ijerph16162976 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Zanobetti A, Dominici F, Wang Y & Schwartz JD A national case-crossover analysis of the short-term effect of PM2.5 on hospitalizations and mortality in subjects with diabetes and neurological disorders. Environ Health 13, 38, doi: 10.1186/1476-069X-13-38 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Lee H et al. Short-term air pollution exposure aggravates Parkinson’s disease in a population-based cohort. Scientific reports 7, 44741, doi: 10.1038/srep44741 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 26.Masri S, Kang CM & Koutrakis P Composition and sources of fine and coarse particles collected during 2002–2010 in Boston, MA. J Air Waste Manag Assoc 65, 287–297, doi: 10.1080/10962247.2014.982307 (2015). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Laden F, Neas LM, Dockery DW & Schwartz J Association of fine particulate matter from different sources with daily mortality in six U.S. cities. Environmental health perspectives 108, 941–947, doi: 10.1289/ehp.00108941 (2000). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.McNeil JJ et al. Effect of Aspirin on Disability-free Survival in the Healthy Elderly. N Engl J Med 379, 1499–1508, doi: 10.1056/NEJMoa1800722 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Berk M et al. Aspirin: a review of its neurobiological properties and therapeutic potential for mental illness. BMC Med 11, 74, doi: 10.1186/1741-7015-11-74 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Fond G et al. Effectiveness and tolerance of anti-inflammatory drugs’ add-on therapy in major mental disorders: a systematic qualitative review. Acta Psychiatr Scand 129, 163–179, doi: 10.1111/acps.12211 (2014). [DOI] [PubMed] [Google Scholar]
  • 31.Tegeler C et al. The inflammatory markers CRP, IL-6, and IL-10 are associated with cognitive function--data from the Berlin Aging Study II. Neurobiol Aging 38, 112–117, doi: 10.1016/j.neurobiolaging.2015.10.039 (2016). [DOI] [PubMed] [Google Scholar]
  • 32.Wellenius GA et al. Ambient fine particulate matter alters cerebral hemodynamics in the elderly. Stroke 44, 1532–1536, doi: 10.1161/STROKEAHA.111.000395 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Novak V Cognition and Hemodynamics. Curr Cardiovasc Risk Rep 6, 380–396, doi: 10.1007/s12170-012-0260-2 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Xu L et al. Effect of enoxaparin and aspirin on hemodynamic disturbances after global cerebral ischemia in rats. Resuscitation 81, 1709–1713, doi: 10.1016/j.resuscitation.2010.07.018 (2010). [DOI] [PubMed] [Google Scholar]
  • 35.Sarnat JA, Brown KW, Schwartz J, Coull BA & Koutrakis P Ambient gas concentrations and personal particulate matter exposures: implications for studying the health effects of particles. Epidemiology 16, 385–395, doi: 10.1097/01.ede.0000155505.04775.33 (2005). [DOI] [PubMed] [Google Scholar]
  • 36.Zeger SL et al. Exposure measurement error in time-series studies of air pollution: concepts and consequences. Environmental health perspectives 108, 419–426, doi: 10.1289/ehp.00108419 (2000). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Gao X et al. Nonsteroidal Antiinflammatory Drugs Modify the Effect of Short-Term Air Pollution on Lung Function. American journal of respiratory and critical care medicine 201, 374–378, doi: 10.1164/rccm.201905-1003LE (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Bell B, Rose CL & Damon A The Normative Aging Study: an interdisciplinary and longitudinal study of health and aging. The International Journal of Aging and Human Development 3, 5–17 (1972). [Google Scholar]
  • 39.Weisskopf MG et al. Cumulative lead exposure and prospective change in cognition among elderly men: the VA Normative Aging Study. American journal of epidemiology 160, 1184–1193, doi: 10.1093/aje/kwh333 (2004). [DOI] [PubMed] [Google Scholar]
  • 40.Colicino E et al. Telomere Length, Long-Term Black Carbon Exposure, and Cognitive Function in a Cohort of Older Men: The VA Normative Aging Study. Environmental health perspectives 125, 76–81, doi: 10.1289/EHP241 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Gao X et al. Impacts of air pollution, temperature, and relative humidity on leukocyte distribution: An epigenetic perspective. Environ Int 126, 395–405, doi: 10.1016/j.envint.2019.02.053 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 42.Pateraki S, Asimakopoulos DN, Flocas HA, Maggos T & Vasilakos C The role of meteorology on different sized aerosol fractions (PM(1)(0), PM(2).(5), PM(2).(5)-(1)(0)). Sci Total Environ 419, 124–135, doi: 10.1016/j.scitotenv.2011.12.064 (2012). [DOI] [PubMed] [Google Scholar]
  • 43.Tian X, Fang Z & Liu W Decreased humidity improves cognitive performance at extreme high indoor temperature. Indoor Air, doi: 10.1111/ina.12755 (2020). [DOI] [PubMed] [Google Scholar]
  • 44.Desquilbet L & Mariotti F Dose-response analyses using restricted cubic spline functions in public health research. Statistics in medicine 29, 1037–1057, doi: 10.1002/sim.3841 (2010). [DOI] [PubMed] [Google Scholar]
  • 45.Seaman SR & White IR Review of inverse probability weighting for dealing with missing data. Stat Methods Med Res 22, 278–295, doi: 10.1177/0962280210395740 (2013). [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1695634_Supp_Material
NIHMS1695634-supplement-1695634_Supp_Material.pdf (915KB, pdf)
1695634_Reporting_Summary
NIHMS1695634-supplement-1695634_Reporting_Summary.pdf (2.3MB, pdf)

RESOURCES