Abstract
Background:
Some studies suggest that ambient particulate air pollution is associated with cognitive decline. However, the findings are mixed, and there is no relevant research examining the influences of ultrafine particles (UFP), which may have more toxicity than larger particles. We therefore conducted this study to investigate whether residential UFP exposure is associated with cognitive decline using data from the Alzheimer’s Disease Research Centers in the United States.
Methods:
This is a longitudinal study of participants who were aged 65 years and older and had normal cognitive status at baseline. Residential UFP exposure, expressed as particle number concentrations (PNC), was assessed in 2016–2017 using a nationwide land use regression model, and was assigned to each participant using their 3-digit residential ZIP codes. Cognitive functions including memory, attention, language, executive function, and global function were assessed annually using 15 neuropsychological tests from March 2015 to February 2022. Linear mixed-effects models were used to examine the associations after adjustment for covariates including baseline age, sex, APOE ε4 status, race, education, smoking status, history of diabetes, quartiles of neighborhood median household income, and interaction terms of follow-up time with each covariate.
Results:
This study included 5646 participants (mean age 76 years, 65% female). On average, each participant had 4 annual visits. When PNC was treated as a continuous variable, there were no statistically or clinically significant changes in annual decline of each cognitive function in relation to an interquartile range elevation in PNC (4026 particles/cm3). Similarly, when PNC was treated as a categorical variable including five exposure groups, there were no linear exposure-response trends in annual decline of each cognitive function across the five exposure groups.
Conclusions:
This study found no meaningful associations between residential UFP exposure and cognitive decline in global and domain-specific functions. There is a need for further research that assigns UFP exposure at a finer geographic scale.
Keywords: Aged, Air Pollution, Ultrafine Particulate Matter, Cognitive Dysfunction, Dementia, Longitudinal Studies
1. Introduction
Age-related cognitive decline and dementia is an emerging public health problem worldwide, imposing enormous health, social, and economic costs [1]. Dementia is a major cause of disability and dependency among older people age 65 and older [2]. The World Health Organization estimated that currently more than 55 million people live with dementia globally, and about 10 million new cases are diagnosed every year [1,2]. With the increase in life expectancy and the older population, age-related cognitive decline and dementia will become increasingly important. It is projected that worldwide the number of people with dementia will increase to 153 million by 2050 [3]. The risk of dementia increases with age, and there is currently no effective cure for the condition. However, increasing evidence has suggested that it is plausible to prevent or delay cognitive decline and dementia by targeting various modifiable risk factors including environmental risk factors [4–6].
Ambient particulate air pollution is a major environmental risk for public health [7], and has been associated with multiple chronic conditions including cardiovascular diseases [8–11], chronic lung diseases [12,13], diabetes [14,15], and chronic inflammation and oxidative stress [10,11,16,17]. Some longitudinal studies have found that ambient particulate air pollution is associated with cognitive decline [18–21]. However, it is notable that the previous findings are heterogeneous [22], some longitudinal studies did not find the association [23–26]. Furthermore, the previous studies have assessed exposure to particulate air pollution using particle mass concentration.
To our knowledge, there is no existing research examining the association between ultrafine particles (particles smaller than 100 nm, UFP) and cognitive decline. Some evidence has shown that UFP may have more toxicity than larger particles due to their greater ability in penetration and deposition, and larger surface area to interact with cell membranes [27,28]. Therefore, we conducted this longitudinal analysis to investigate whether residential exposure to higher levels of UFP is associated with accelerated decline in global and domain-specific cognitive functions using a sample of participants from the Alzheimer’s Disease Research Centers (ADRCs) in the United States [29].
2. Material and methods
2.1. Study participants
In collaboration with ADRCs across the United States, the National Alzheimer’s Coordinating Center (NACC) at the University of Washington has regularly collected standardized clinical and neuropathological research data from 2005 using the Uniform Data Set (UDS) forms [30]. The participants were recruited by each ADRC from multiple sources including clinician referral, self-referral by participants or family members, and active recruitment in the community organizations. There were two groups of participants in the sample: one group included individuals with impaired cognition ranging from mild cognitive impairment to dementia; the other group included individuals who had normal cognitive function and were used as healthy controls for studying age-related cognitive decline and transitions to dementia. During annual visits at ADRC clinics, participants’ demographics, medical history, neurological examination, and neuropsychological assessment data were collected by trained clinic personnel [31]. This was an open cohort, participants could dynamically enter or leave the cohort at different points of time. We acquired data from the NACC, which cover a follow-up period from 2005 until February 2022. This study was conducted using publicly accessible de-identified data, thus institutional ethics approval was not required.
In March 2015, the ADRCs implemented the UDS version 3 [32], which included substantial revisions on the neuropsychological test battery [33]. We thus restricted our analysis to participants who used the UDS version 3 or later. Starting from March 2015, the first visit was treated as baseline visit for each participant. During the follow-up period from March 2015 to February 2022, we selected participants and the data records according to the following criteria: (1) aged 65 years and older at baseline; (2) had normal cognitive status at baseline (clinical diagnosis of cognitive status was normal [NACCUDSD = 1] and the sum of boxes of the Clinical Dementia Rating was 0 [CDRSUM = 0]); (3) had complete residential ZIP codes and the exposure data (lived in the contiguous US); (4) had complete data on covariates; and (5) had at least 2 visits and relevant neuropsychological tests during the follow-up period (Figure 1).
Figure 1.
Data flow diagram for the selection of participants from the NACC sample during the study period from March 2015 to February 2022.
2.2. Study design
Our analysis is longitudinal with repeated measures of cognitive functions. Residential UFP exposure was assessed in 2016–2017 using a nationwide land use regression model (LUR) [34]. The exposure estimates were assigned to participants using their 3-digit residential ZIP codes. Global and domain-specific cognitive functions were assessed every year using 15 neuropsychological tests during the follow-up period from March 2015 to February 2022 [33]. Linear mixed-effects models were used to assess the associations of UFP exposure with annual decline of each cognitive function [35].
2.3. Ultrafine particles assessment
UFP are too small for mass-based measurement to appropriately reflect the levels of UFP in the air, they are better measured using particle number concentrations (PNC). Residential UFP exposure was assessed using a nationwide LUR model [34], which was developed in 2016–2017 to estimate annual average outdoor PNC at census block level in the contiguous United States. The model was able to explain 77% of spatial variability of annual average PNC (R2 = 0.77, root-mean-square error = 2400 particles/cm3) [34]. We obtained the census-block-level PNC and population data, assigned the data to the corresponding 3-digit ZIP code areas, and calculated population-weighted average annual PNC at each 3-digit ZIP code area:
where is the number of people living in a census block, is census-block-level PNC. The exposure data were assigned to each participant using their 3-digit residential ZIP codes. The population-weighted average PNC at each 3-digit ZIP code area took into account both the PNC and the number of people at each census block, and thus could more accurately (vs arithmetic mean PNC) reflect human UFP exposure at each 3-digit ZIP code area.
2.4. Cognitive function assessment
Numerous neuropsychological tests were administered for participants during annual visits at ADRC clinics between March 2015 and February 2022 [33]. Following the previous studies [18–21], we assessed four domain-specific cognitive functions including memory (4 tests), attention (4 tests), language (5 tests), and executive function (2 tests) using 15 tests available in the UDS version 3 or later (Table 1). For each test, we calculated standardized z scores using the mean and standard deviation (SD) of the original measures at baseline. For the Trail Making Test Part A and Part B, the distributions of the original scores were right skewed, thus log-transformed data were used to calculate z scores. Furthermore, we calculated means of z scores of all 15 tests or relevant tests in each domain to assess global and domain-specific cognitive functions at the time of each visit (Table 1).
Table 1.
Neuropsychological test battery (version 3) in the Uniform Data Set (UDS)
| Domain and neuropsychological test | Variable in UDS |
|---|---|
|
| |
| Memory | |
| Craft Story 21 Recall (Immediate): total story units recalled, verbatim scoring | CRAFTVRS |
| Craft Story 21 Recall (Immediate): total story units recalled, paraphrase scoring | CRAFTURS |
| Craft Story 21 Recall (Delayed): total story units recalled, verbatim scoring | CRAFTDVR |
| Craft Story 21 Recall (Delayed): total story units recalled, paraphrase scoring | CRAFTDRE |
| Memory on average: mean of the above 4 tests in z scores | Not available |
|
| |
| Attention | |
| Number Span Test: forward, number of correct trials | DIGFORCT |
| Number Span Test: forward, longest span | DIGFORSL |
| Number Span Test: backward, number of correct trials | DIGBACCT |
| Number Span Test: backward, longest span | DIGBACLS |
| Attention on average: mean of the above 4 tests in z scores | Not available |
|
| |
| Language | |
| Multilingual Naming Test (MINT): total score | MINTTOTS |
| Number of correct F-words generated in 1 minute | UDSVERFC |
| Number of correct L-words generated in 1 minute | UDSVERLC |
| Animals: total number of animals named in 1 minute | ANIMALS |
| Vegetables: total number of vegetables named in 1 minute | VEG |
| Language on average: mean of the above 5 tests in z scores | Not available |
|
| |
| Executive function | |
| Trail Making Test Part A: total number of seconds to complete | TRAILA |
| Trail Making Test Part B: total number of seconds to complete | TRAILB |
| Executive function on average: mean of the above 2 tests in z scores | Not available |
|
| |
|
Global function Mean of the above 15 tests in z scores |
Not available |
2.5. Covariates
We retrieved baseline information on the potential confounding factors including age, sex (male, female), apolipoprotein E (APOE) ε4 status defined as the presence of at least one ε4 allele (yes, no), race/ethnicity (white, black, other), education (years), smoking status (current, former, never), and history of diabetes (yes, no). In addition, we added quartiles of neighborhood median household income at 3-digit ZIP code areas to reflect neighborhood socioeconomic status (SES), which were developed using the 2010 census-tract-level median household income data [36].
2.6. Statistical analyses
We divided study participants into quintiles according to their average PNC levels during the follow-up period, and compared personal and neighborhood characteristics at baseline and average cognitive scores during the follow-up period between the five exposure groups. We used linear mixed-effects models to examine the associations of UFP exposure with annual decline of each cognitive function after accounting for the repeated measures correlation [35]. Exposure-related change in annual decline of each cognitive function was expressed in the model using an interaction term between follow-up time and the exposure variable. We first treated PNC as a continuous variable to examine linear relationships, we examined changes in annual decline of each cognitive function in relation to an interquartile range (IQR) elevation in PNC during the follow-up period. Furthermore, we treated PNC as a categorical variable including five exposure groups to examine potential nonlinear relationships and exposure-response trends. We calculated differences in annual decline of each cognitive function during the follow-up period for each exposure group by using the group with the lowest exposure as the reference group.
We used three models to examine how covariates would influence effect estimates. Model 1 was an unadjusted model including participant-specific random intercept, participant-specific random slop for follow-up time, PNC, follow-up time, and an interaction term between PNC and follow-up time. Model 2 was based on model 1 and additionally adjusted for baseline age, sex, APOE ε4 status, and 3 interaction terms of follow-up time with each covariate. Model 3 was based on model 2 and additionally adjusted for race, education, smoking status, history of diabetes, quartiles of neighborhood median household income, and 5 interaction terms of follow-up time with each covariate. All statistical tests were 2-sided, and were conducted using the R statistical computing environment, version 4.2.2 (R Foundation for Statistical Computing, Vienna, Austria). The lme4 package (version 1.1–31) for linear mixed-effects models was used in the main statistical analysis [35].
3. Results
3.1. Characteristics of study participants
During the study period, there were 8685 participants who were aged 65 years and older and had normal cognitive status at baseline, and had complete PNC exposure data. After excluding those with missing data on covariates and those with less than 2 visits and neuropsychological tests (n = 3039), leaving 5646 participants for the analysis of global function (the sample size was 4939 for memory, 4953 for attention, 5645 for language, and 5422 for executive function) (Figure 1). Compared with excluded participants, included participants were more likely to be White, and had better global and domain-specific cognitive functions at baseline (eTable 1).
These participants lived in 451 different 3-digit ZIP code areas during the follow-up period (Figure 2); the median of surface areas was 3310 km2 (IQR 754 – 7495 km2) (1278 mi2, 291 – 2894 mi2). There were a total of 22,517 visits; each participant, on average, had 4 visits; the average follow-up time was 3.5 years (SD 1.6 years). Based on individual visits, the average PNC was 8196 particles/cm3 (IQR 6305 – 10,331 particles/cm3, range 2342 – 14,663 particles/cm3).
Figure 2.
Geographic distribution of study participants by 3-digit residential ZIP codes. Each small polygon area represents a 3-digit ZIP code area.
At baseline, the average age was 76.0 years (SD 7.3 years), average education was 16.2 years (SD 2.8 years). The five exposure groups had similar average age and education (Table 2). Overall, 65% of the participants were female, 81% were White, 29% carried at least one APOE ε4 allele, and 16% had diabetes. Compared with those in the lowest exposure quintile, participants in the highest exposure quintile were less likely to be White or be an APOE ε4 carrier, and were more likely to be female, Black, and live in a higher SES neighborhood. The distributions of smoking status and history of diabetes were similar across the five exposure groups (Table 2). For the five cognitive functions, the highest exposure quintile consistently had worse global function and domain-specific functions including memory, attention, language, and executive function compared with the lowest exposure quintile (Table 2).
Table 2.
Baseline characteristics of participants by quintiles of residential exposure to ultrafine particlesa
| Quintiles of average PNC (range, particles/cm3) |
||||||
|---|---|---|---|---|---|---|
| Characteristic | Lowest 2525 – 5580 (n = 1130) |
Second 5581 – 7198 (n =1129) |
Third 7199 – 9117 (n =1156) |
Fourth 9118 – 10,698 (n =1119) |
Highest 10,699 – 14,663 (n =1112) |
Overall (n = 5646) |
|
| ||||||
| Age (SD), years | 75.3 (6.7) | 76.4 (7.4) | 75.5 (7.3) | 76.9 (7.7) | 76.1 (7.3) | 76.0 (7.3) |
| Sex | ||||||
| Male | 445 (39.4) | 386 (34.2) | 419 (36.2) | 412 (36.8) | 306 (27.5) | 1,968 (34.9) |
| Female | 685 (60.6) | 743 (65.8) | 737 (63.8) | 707 (63.2) | 806 (72.5) | 3,678 (65.1) |
| APOE ε4 status | ||||||
| No | 787 (69.6) | 770 (68.2) | 814 (70.4) | 804 (71.8) | 820 (73.7) | 3,995 (70.8) |
| Yes | 343 (30.4) | 359 (31.8) | 342 (29.6) | 315 (28.2) | 292 (26.3) | 1,651 (29.2) |
| Race | ||||||
| White | 1,037 (91.8) | 919 (81.4) | 977 (84.5) | 896 (80.1) | 765 (68.8) | 4,594 (81.4) |
| Black | 56 (5.0) | 171 (15.1) | 104 (9.0) | 135 (12.1) | 240 (21.6) | 706 (12.5) |
| Otherb | 37 (3.3) | 39 (3.5) | 75 (6.5) | 88 (7.9) | 107 (9.6) | 346 (6.1) |
| Education (SD), years | 16.1 (2.6) | 16.2 (2.8) | 16.2 (2.9) | 16.5 (2.6) | 16.2 (3.0) | 16.2 (2.8) |
| Smoking status | ||||||
| Never | 615 (54.4) | 629 (55.7) | 618 (53.5) | 628 (56.1) | 564 (50.7) | 3,054 (54.1) |
| Former | 474 (41.9) | 464 (41.1) | 502 (43.4) | 458 (40.9) | 510 (45.9) | 2,408 (42.6) |
| Current | 41 (3.6) | 36 (3.2) | 36 (3.1) | 33 (2.9) | 38 (3.4) | 184 (3.3) |
| History of diabetes | ||||||
| No | 960 (85.0) | 940 (83.3) | 963 (83.3) | 969 (86.6) | 907 (81.6) | 4,739 (83.9) |
| Yes | 170 (15.0) | 189 (16.7) | 193 (16.7) | 150 (13.4) | 205 (18.4) | 907 (16.1) |
| Quartiles of neighborhood median household income | ||||||
| Lowest ($22,750 – $47,266) |
432 (38.2) | 97 (8.6) | 332 (28.7) | 313 (28.0) | 247 (22.2) | 1,421 (25.2) |
| Second ($47,267 – $56,351) |
407 (36.0) | 572 (50.7) | 149 (12.9) | 257 (23.0) | 221 (19.9) | 1,606 (28.4) |
| Third ($56,352 – $74,776) |
212 (18.8) | 272 (24.1) | 204 (17.6) | 272 (24.3) | 244 (21.9) | 1,204 (21.3) |
| Highest ($74,777 – $180,263) |
79 (7.0) | 188 (16.7) | 471 (40.7) | 277 (24.8) | 400 (36.0) | 1,415 (25.1) |
|
| ||||||
| Cognitive function z scores (SD)c | ||||||
| Memory (n = 4939) | 0.056 (0.879) | −0.077 (0.882) | 0.081 (0.880) | 0.076 (0.880) | −0.074 (0.918) | 0.012 (0.891) |
| Attention (n = 4953) | 0.006 (0.769) | −0.065 (0.779) | −0.067 (0.810) | 0.037 (0.801) | −0.009 (0.835) | −0.021 (0.800) |
| Language (n = 5645) | 0.002 (0.690) | −0.021 (0.729) | −0.008 (0.714) | −0.015 (0.761) | −0.027 (0.758) | −0.014 (0.731) |
| Executive function (n = 5422)d | −0.056 (0.867) | 0.129 (0.957) | −0.013 (0.875) | 0.052 (0.905) | 0.153 (0.959) | 0.052 (0.916) |
| Global function (n = 5646) | 0.013 (0.613) | −0.057 (0.653) | −0.003 (0.645) | −0.024 (0.679) | −0.058 (0.682) | −0.026 (0.655) |
Abbreviations: APOE, apolipoprotein E; PNC, particle number concentration; SD, standard deviation.
Data are presented as the number of participants (column proportion, %) for categorical variables, and mean (SD) for continuous variables.
The other group includes American Indian or Alaska Native, Native Hawaiian or Pacific Islander, Asian, multiracial, and unknown or ambiguous persons.
Domain-specific z scores were calculated for each participant by averaging all available tests in a specific domain, global z scores were calculated for each participant by averaging all available tests. Lower scores indicate worse functions, unless otherwise indicated.
Lower scores indicate better functions.
3.2. Ultrafine particles and cognitive decline
When PNC was treated as a continuous variable and adjusted for the covariates (Table 3), there were no significant changes in annual decline of memory, attention, language, and global function in relation to an IQR elevation in PNC (4026 particles/cm3). Executive function was significantly improved, but the magnitude was small and not clinically meaningful (Table 3). In addition, APOE ε4 status and age were consistently associated with faster annual declines of each domain-specific and global cognitive function. History of diabetes was associated with faster annual decline in language. Current smokers had faster annual decline in memory and executive function, former smokers had slightly faster decline in language. Whereas living in the highest or third quartile of neighborhood median household income was associated with slower annual decline in executive function (Table 3).
Table 3.
Associations of annual cognitive decline with PNC and covariatesa
| Variable | Memory | Attention | Language | Executive functionb | Global function |
|---|---|---|---|---|---|
|
| |||||
| PNC × Timecd | −0.0001 (−0.0087 to 0.0086) |
−0.0053 (−0.0121 to 0.0014) |
−0.0002 (−0.0054 to 0.0050) |
−0.0089 (−0.0166 to −0.0011) |
−0.0008 (−0.0053 to 0.0036) |
| Age × Timed | −0.0043 (−0.0052 to −0.0035) |
−0.0024 (−0.0031 to −0.0017) |
−0.0038 (−0.0043 to −0.0033) |
0.0062 (0.0054 to 0.0069) |
−0.0041 (−0.0046 to −0.0037) |
| Sex (female vs male) × Timed |
−0.0065 (−0.0188 to 0.0057) |
0.0066 (−0.0029 to 0.0161) |
−0.0006 (−0.0080 to 0.0068) |
0.0072 (−0.0037 to 0.0181) |
−0.0030 (−0.0094 to 0.0034) |
| APOE ε4 (yes vs no) × Timed |
−0.0243 (−0.0372 to −0.0114) |
−0.0131 (−0.0231 to −0.0031) |
−0.0285 (−0.0363 to −0.0208) |
0.0303 (0.0190 to 0.0417) |
−0.0272 (−0.0339 to −0.0205) |
| Race (White vs other) × Timed |
0.0065 (−0.0186 to 0.0316) |
−0.0150 (−0.0347 to 0.0048) |
−0.0001 (−0.0146 to 0.0144) |
0.0162 (−0.0056 to 0.0379) |
−0.0011 (−0.0137 to 0.0114) |
| Race (Black vs other) × Timed |
0.0021 (−0.0277 to 0.0319) |
0.0030 (−0.0204 to 0.0263) |
0.0215 (0.0040 to 0.0389) |
−0.0004 (−0.0264 to 0.0257) |
0.0165 (0.0013 to 0.0316) |
| Education × Timed | −0.0001 (−0.0022 to 0.0022) |
−0.0014 (−0.0031 to 0.0003) |
−0.0001 (−0.0014 to 0.0012) |
0.0004 (−0.0016 to 0.0023) |
−0.0011 (−0.0022 to 0.0001) |
| Smoking status (former vs never) × Timed |
−0.0010 (−0.0127 to 0.0108) |
−0.0009 (−0.0100 to 0.0082) |
−0.0074 (−0.0145 to −0.0004) |
−0.0029 (−0.0133 to 0.0075) |
−0.0029 (−0.0090 to 0.0033) |
| Smoking status (current vs never) × Timed |
−0.0530 (−0.0864 to −0.0196) |
0.0136 (−0.0125 to 0.0398) |
−0.0185 (−0.0386 to 0.0015) |
0.0311 (0.0014 to 0.0607) |
−0.0172 (−0.0346 to 0.0002) |
| Diabetes (yes vs no) × Timed |
−0.0056 (−0.0216 to 0.0104) |
0.0027 (−0.0098 to 0.0151) |
−0.0105 (−0.0201 to −0.0008) |
0.0128 (−0.0013 to 0.0270) |
−0.0012 (−0.0095 to 0.0072) |
| Income quartile (2 vs 1) × Timed |
0.0067 (−0.0092 to 0.0226) |
−0.0001 (−0.0125 to 0.0123) |
−0.0022 (−0.0117 to 0.0072) |
0.0117 (−0.0022 to 0.0255) |
−0.0017 (−0.0099 to 0.0066) |
| Income quartile (3 vs 1) × Timed |
0.0030 (−0.0143 to 0.0202) |
0.0069 (−0.0065 to 0.0203) |
0.0039 (−0.0063 to 0.0141) |
−0.0158 (−0.0309 to −0.0007) |
−0.0006 (−0.0094 to 0.0083) |
| Income quartile (4 vs 1) × Timed |
0.0074 (−0.0093 to 0.0241) |
−0.0037 (−0.0167 to 0.0093) |
0.0079 (−0.0022 to 0.0179) |
−0.0208 (−0.0356 to −0.0060) |
0.0062 (−0.0025 to 0.0149) |
| PNC (IQR increase) |
−0.0061 (−0.0392 to 0.0269) |
0.0356 (0.0052 to 0.0659) |
−0.0151 (−0.0371 to 0.0068) |
0.0559 (0.0277 to 0.0841) |
−0.0113 (−0.0298 to 0.0073) |
| Time (year)d |
0.3420 (0.2591 to 0.4249) |
0.2031 (0.1386 to 0.2677) |
0.2633 (0.2148 to 0.3118) |
−0.4034 (−0.4756 to −0.3312) |
0.3132 (0.2712 to 0.3553) |
| Age (year) |
−0.0374 (−0.0408 to −0.0340) |
−0.0155 (−0.0187 to −0.0122) |
−0.0266 (−0.0289 to −0.0243) |
0.0486 (0.0457 to 0.0514) |
−0.0303 (−0.0323 to −0.0283) |
| Sex (female vs male) |
0.1792 (0.1295 to 0.2289) |
−0.0461 (−0.0933 to 0.0011) |
0.2028 (0.1671 to 0.2386) |
−0.0495 (−0.0932 to −0.0058) |
0.1320 (0.1017 to 0.1624) |
| APOE ε4 (yes vs no) |
−0.1023 (−0.1535 to −0.0510) |
−0.0033 (−0.0520 to 0.0453) |
0.0091 (−0.0277 to 0.0459) |
0.0562 (0.0112 to 0.1012) |
−0.0336 (−0.0649 to −0.0024) |
| Race (White vs other) |
0.2244 (0.1248 to 0.3241) |
0.3724 (0.2780 to 0.4667) |
0.3026 (0.2323 to 0.3729) |
−0.3344 (−0.4208 to −0.2480) |
0.2918 (0.2320 to 0.3516) |
| Race (Black vs other) |
−0.0635 (−0.1806 to 0.0536) |
0.0543 (−0.0567 to 0.1653) |
−0.0892 (−0.1721 to −0.0062) |
0.2154 (0.1135 to 0.3173) |
−0.0833 (−0.1538 to −0.0129) |
| Education (year) |
0.0614 (0.0527 to 0.0701) |
0.0638 (0.0556 to 0.0721) |
0.0719 (0.0657 to 0.0781) |
−0.0665 (−0.0741 to −0.0589) |
0.0689 (0.0637 to 0.0742) |
| Smoking status (former vs never) |
0.0696 (0.0221 to 0.1171) |
0.0754 (0.0302 to 0.1205) |
0.0420 (0.0079 to 0.0761) |
0.0081 (−0.0336 to 0.0498) |
0.0536 (0.0246 to 0.0826) |
| Smoking status (current vs never) |
0.0262 (−0.1048 to 0.1573) |
0.0921 (−0.0327 to 0.2169) |
−0.0101 (−0.1049 to 0.0848) |
0.1133 (−0.0025 to 0.2292) |
0.0034 (−0.0773 to 0.0840) |
| Diabetes (yes vs no) |
−0.0419 (−0.1068 to 0.0230) |
−0.1374 (−0.1990 to −0.0757) |
−0.1339 (−0.1803 to −0.0875) |
0.2188 (0.1621 to 0.2755) |
−0.1371 (−0.1766 to −0.0976) |
| Income quartile (2 vs 1) |
−0.0275 (−0.0922 to 0.0371) |
0.0172 (−0.0442 to 0.0785) |
−0.0001 (−0.0459 to 0.0456) |
0.0008 (−0.0551 to 0.0566) |
−0.0085 (−0.0474 to 0.0304) |
| Income quartile (3 vs 1) |
−0.0098 (−0.0791 to 0.0595) |
0.0304 (−0.0354 to 0.0962) |
0.0749 (0.0259 to 0.1239) |
−0.0167 (−0.0768 to 0.0433) |
0.0417 (0.0000 to 0.0834) |
| Income quartile (4 vs 1) |
0.0529 (−0.0132 to 0.1189) |
0.1160 (0.0533 to 0.1787) |
0.0920 (0.0442 to 0.1397) |
−0.0619 (−0.1203 to −0.0036) |
0.0911 (0.0505 to 0.1317) |
Abbreviations: APOE, apolipoprotein E; IQR, interquartile range; PNC, particle number concentrations.
Results are derived from the fully adjusted model (model 3), lower values indicate worse cognitive functions, unless otherwise indicated.
Lower values indicate better cognitive functions.
First row (PNC × Time) indicates changes in annual decline of each cognitive function (z-scores) in relation to an IQR elevation in PNC (4026 particles/cm3) during the follow-up period.
Follow-up time (year).
When PNC was treated as a categorical variable including five exposure groups (Table 4), there were no linear exposure-response trends in differences of domain-specific and global cognitive declines across the five exposure groups. In addition, the magnitudes of the differences in annual cognitive declines were small, and mostly not statistically significant.
Table 4.
Adjusted differences (95% CIs) in annual decline of cognitive functions (z-scores) during the follow-up period by quintiles of particle number concentrations (PNC)a
| Quintiles of PNC | Memory | Attention | Language | Executive functionb | Global function |
|---|---|---|---|---|---|
|
| |||||
| Lowest (2525 – 5580 particles/cm3) |
0.0000 (Referent) |
0.0000 (Referent) |
0.0000 (Referent) |
0.0000 (Referent) |
0.0000 (Referent) |
| Second (5581 – 7198 particles/cm3) |
−0.0136 (−0.0317 to 0.0045) |
0.0037 (−0.0104 to 0.0177) |
0.0138 (0.0028 to 0.0248) |
−0.0020 (−0.0181 to 0.0142) |
0.0026 (−0.0070 to 0.0122) |
| Third (7199 – 9117 particles/cm3) |
−0.0131 (−0.0321 to 0.0058) |
0.0119 (−0.0028 to 0.0267) |
−0.0038 (−0.0152 to 0.0076) |
−0.0142 (−0.0309 to 0.0025) |
−0.0030 (−0.0129 to 0.0069) |
| Fourth (9118 – 10,698 particles/cm3) |
0.0038 (−0.0157 to 0.0234) |
0.0007 (−0.0146 to 0.0159) |
0.0038 (−0.0074 to 0.0150) |
−0.0228 (−0.0396 to −0.0059) |
0.0041 (−0.0056 to 0.0138) |
| Highest (10,699 – 14,663 particles/cm3) |
−0.0089 (−0.0277 to 0.0099) |
−0.0088 (−0.0234 to 0.0058) |
0.0056 (−0.0058 to 0.0170) |
−0.0130 (−0.0300 to 0.0039) |
−0.0032 (−0.0131 to 0.0067) |
The results are derived from the interaction term between follow-up time and quintiles of particle number concentrations in the fully adjusted model (model 3). Lower values indicate worse cognitive functions, unless otherwise indicated.
Lower values indicate better functions.
3.3. Sensitivity analyses
The relationship of PNC with each specific neuropsychological test in a domain was not completely consistent with the relationship between PNC and the domain-specific function (eTable 2). When cognitive functions were measured using the original scores, the results were similar to those based on standardized z scores (eTable 3). Furthermore, the results did not substantially change across the three statistical models with different adjustments for covariates (eTable 2, eTable 3).
4. Discussion
In this longitudinal analysis of older people with normal cognitive status at baseline, we used a nationwide LUR model to estimate residential UFP exposure, and investigated whether UFP exposure was associated with cognitive decline. We found no meaningful associations between residential UFP exposure and annual decline in global and domain-specific cognitive functions.
Our findings are consistent with some previous longitudinal studies [23–26], which also suggested null associations between particulate air pollution and cognitive decline. Based on the same cohort and similar study methods, Cleary et al. did not find significant associations between residential fine particulate matter (PM2.5) and cognitive decline [23]. Similarly, two large longitudinal studies in the US [24] and UK [25] found no significant associations between residential PM2.5 and cognitive decline.
In addition, some longitudinal studies reported no significant associations of cognitive decline with traffic-related air pollution indicated by traffic proximity [19], black carbon [20], nitrogen dioxide [20], and nitrogen oxides [26]. In urban areas, UFP mainly originates from motor vehicle exhaust, PNC concentrations decrease exponentially with increasing distance from major roadways and reach background levels within about 200 m [37]. These previous findings on traffic-related air pollution are somewhat consistent with the null associations observed in our study.
However, there are some longitudinal studies showing that particulate air pollution is associated with cognitive decline [18–21]. Scrutinizing these findings may be useful to better understand the discrepancies between individual studies. In the Nurses’ Health Study [18], Weuve et al. found that coarse particulate matter (PM2.5–10) and PM2.5 were associated with faster decline in global cognition. The association was stronger for PM2.5–10 than PM2.5 in terms of effect sizes and linear exposure-response trends [18]. These findings are not consistent with the general consensus that PM2.5 is more strongly associated with chronic health outcomes than coarse particles [10].
In a longitudinal study including two cohorts in New York City [19], Kulick et al. found that in one cohort, residential PM2.5 was associated with cognitive decline in global function as well as memory, language, and executive function. However, when PM10 was used as the exposure, the associations were largely null (except for global cognition). Furthermore, these significant associations were not observed in another cohort with smaller sample size [19].
In the French Three-City Cohort Study [20], Duchesne et al. found that residential PM2.5 was associated with slightly faster decline in global cognition, but not with domain-specific cognition [20]. Recently, Younan et al. found that greater reduction in residential PM2.5 was associated with slower decline in global cognition and episodic memory among older women, and the associations remained stable after adjustment for covariates [21]. Although these previous studies suggest significant associations, it is notable that the findings are not consistent across the individual studies, and there is currently no sufficient evidence to support a causal relationship between particulate air pollution and cognitive decline.
Our study has some strengths. First, we included a comprehensive neuropsychological battery, which quantitatively characterized annual changes in global and four domain-specific cognitive functions during the follow-up period. In addition, we used both standardized z scores and original scores as the outcome variables in the statistical analysis, and the results are consistent. Second, previous studies have used particle mass concentrations to assess exposure to particulate air pollution. To our knowledge, the current study is the first investigating the associations between UFP exposure and cognitive decline. Finally, the study participants were recruited by ADRCs across the US, making the findings broadly generalizable.
This study has some limitations to be considered. First, our UFP exposure was assessed using a nationwide LUR model rather than actual measurement, the LUR model was able to explain 77% of spatial variation in UFP [34]. In addition, because of privacy protections, only the first 3 digits of the residential ZIP codes were available to assign the exposure data. A 3-digit ZIP code may cover a large area, especially in rural and remote regions. The PNC levels at 3-digit ZIP code areas would miss near roadway scale elevation of UFP [37]. Second, our exposure assessment aimed to approximately reflect exposure levels outside the residences rather than actual personal exposure. Many factors such as air infiltration, individual mobility, and outdoor activity might affect actual personal UFP exposure [38]. It is possible that exposure assessment errors were nondifferential, which would lead to underestimation of the true associations [39,40]. Finally, higher levels of UFP may indicate higher levels of urbanization, which may in turn indicate potential better education, higher-level occupations, higher income, and improved access to social and health services. These protective factors may play a role in slowing the process of age-related cognitive decline. Our multivariable analysis may not be able to disentangle the influences of these factors. This limitation may partly explain the null association in our study if the association is truly existent.
5. Conclusions
In this 4-year longitudinal analysis, we found no meaningful associations between residential UFP exposure and cognitive decline in global and domain-specific functions. In addition, it is notable that the previous findings on the associations between particulate air pollution (e.g., black carbon, PM2.5, PM2.5–10, and PM10) and cognitive decline are mixed. Even among studies reporting significant associations, the findings are not consistent across the individual studies. Given the potential importance of ambient UFP with respect to cognitive decline, more research is needed to clarify the relationship using finer grain exposure assessment that captures near highway UFP gradients.
Supplementary Material
Highlights.
APOE ε4 carriers had faster annual decline in every cognitive function.
Smoking was associated with faster decline in memory and executive function.
Higher income neighborhoods had greater exposure to ultrafine particles.
No meaningful association was found between ultrafine particles and cognitive decline.
Acknowledgments
Dr. Doug Brugge is funded by the National Institute of Environmental Health Sciences (R01ES030289). The NACC database is funded by NIA/NIH Grant U24 AG072122. NACC data are contributed by the NIA-funded ADRCs: P30 AG062429 (PI James Brewer, MD, PhD), P30 AG066468 (PI Oscar Lopez, MD), P30 AG062421 (PI Bradley Hyman, MD, PhD), P30 AG066509 (PI Thomas Grabowski, MD), P30 AG066514 (PI Mary Sano, PhD), P30 AG066530 (PI Helena Chui, MD), P30 AG066507 (PI Marilyn Albert, PhD), P30 AG066444 (PI John Morris, MD), P30 AG066518 (PI Jeffrey Kaye, MD), P30 AG066512 (PI Thomas Wisniewski, MD), P30 AG066462 (PI Scott Small, MD), P30 AG072979 (PI David Wolk, MD), P30 AG072972 (PI Charles DeCarli, MD), P30 AG072976 (PI Andrew Saykin, PsyD), P30 AG072975 (PI David Bennett, MD), P30 AG072978 (PI Neil Kowall, MD), P30 AG072977 (PI Robert Vassar, PhD), P30 AG066519 (PI Frank LaFerla, PhD), P30 AG062677 (PI Ronald Petersen, MD, PhD), P30 AG079280 (PI Eric Reiman, MD), P30 AG062422 (PI Gil Rabinovici, MD), P30 AG066511 (PI Allan Levey, MD, PhD), P30 AG072946 (PI Linda Van Eldik, PhD), P30 AG062715 (PI Sanjay Asthana, MD, FRCP), P30 AG072973 (PI Russell Swerdlow, MD), P30 AG066506 (PI Todd Golde, MD, PhD), P30 AG066508 (PI Stephen Strittmatter, MD, PhD), P30 AG066515 (PI Victor Henderson, MD, MS), P30 AG072947 (PI Suzanne Craft, PhD), P30 AG072931 (PI Henry Paulson, MD, PhD), P30 AG066546 (PI Sudha Seshadri, MD), P20 AG068024 (PI Erik Roberson, MD, PhD), P20 AG068053 (PI Justin Miller, PhD), P20 AG068077 (PI Gary Rosenberg, MD), P20 AG068082 (PI Angela Jefferson, PhD), P30 AG072958 (PI Heather Whitson, MD), P30 AG072959 (PI James Leverenz, MD).
Footnotes
Declaration of interests
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Credit author statement
Wenqi Gan: Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization, Project administration.
Kevin Manning: Methodology, Validation, Investigation, Writing – review & editing.
Ekaterina Cleary: Conceptualization, Validation, Investigation, Writing – review & editing.
Richard Fortinsky: Methodology, Validation, Investigation, Writing – review & editing.
Doug Brugge: Conceptualization, Methodology, Validation, Investigation, Writing – review & editing, Supervision, Project administration.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Data availability
The data presented in this study are available upon request from the National Alzheimer’s Coordinating Center (NACC) at the University of Washington.
References
- 1.World Health Organization. Global status report on the public health response to dementia. Available at: https://apps.who.int/iris/bitstream/handle/10665/344701/9789240033245-eng.pdf. Accessed: May 16, 2023.
- 2.World Health Organization. Dementia. Available at: https://www.who.int/news-room/fact-sheets/detail/dementia. Accessed: May 16, 2023.
- 3.Collaborators GBDDF. Estimation of the global prevalence of dementia in 2019 and forecasted prevalence in 2050: an analysis for the Global Burden of Disease Study 2019. Lancet Public Health 2022;7:e105–e125. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Livingston G, Huntley J, Sommerlad A, Ames D, Ballard C, Banerjee S, et al. Dementia prevention, intervention, and care: 2020 report of the Lancet Commission. Lancet 2020;396:413–446. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Tariq S, Barber PA. Dementia risk and prevention by targeting modifiable vascular risk factors. J Neurochem 2018;144:565–581. [DOI] [PubMed] [Google Scholar]
- 6.Montero-Odasso M, Ismail Z, Livingston G. One third of dementia cases can be prevented within the next 25 years by tackling risk factors. The case “for” and “against”. Alzheimers Res Ther 2020;12:81. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Cohen AJ, Brauer M, Burnett R, Anderson HR, Frostad J, Estep K, et al. Estimates and 25-year trends of the global burden of disease attributable to ambient air pollution: an analysis of data from the Global Burden of Diseases Study 2015. Lancet 2017;389:1907–1918. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Gan WQ, Koehoorn M, Davies HW, Demers PA, Tamburic L, Brauer M. Long-term exposure to traffic-related air pollution and the risk of coronary heart disease hospitalization and mortality. Environ Health Perspect 2011;119:501–507. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Gan WQ, Davies HW, Koehoorn M, Brauer M. Association of long-term exposure to community noise and traffic-related air pollution with coronary heart disease mortality. Am J Epidemiol 2012;175:898–906. [DOI] [PubMed] [Google Scholar]
- 10.Brook RD, Rajagopalan S, Pope CA 3rd, Brook JR, Bhatnagar A, Diez-Roux AV, et al. Particulate matter air pollution and cardiovascular disease: An update to the scientific statement from the American Heart Association. Circulation 2010;121:2331–2378. [DOI] [PubMed] [Google Scholar]
- 11.Ohlwein S, Kappeler R, Kutlar Joss M, Kunzli N, Hoffmann B. Health effects of ultrafine particles: a systematic literature review update of epidemiological evidence. Int J Public Health 2019;64:547–559. [DOI] [PubMed] [Google Scholar]
- 12.Gan WQ, Fitzgerald JM, Carlsten C, Sadatsafavi M, Brauer M. Associations of ambient air pollution with chronic obstructive pulmonary disease hospitalization and mortality. Am J Respir Crit Care Med 2013;187:721–727. [DOI] [PubMed] [Google Scholar]
- 13.Kurt OK, Zhang J, Pinkerton KE. Pulmonary health effects of air pollution. Curr Opin Pulm Med 2016;22:138–143. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Eze IC, Hemkens LG, Bucher HC, Hoffmann B, Schindler C, Kunzli N, et al. Association between ambient air pollution and diabetes mellitus in Europe and North America: systematic review and meta-analysis. Environ Health Perspect 2015;123:381–389. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Yang BY, Fan S, Thiering E, Seissler J, Nowak D, Dong GH, et al. Ambient air pollution and diabetes: A systematic review and meta-analysis. Environ Res 2020;180:108817. [DOI] [PubMed] [Google Scholar]
- 16.Lane KJ, Levy JI, Scammell MK, Peters JL, Patton AP, Reisner E, et al. Association of modeled long-term personal exposure to ultrafine particles with inflammatory and coagulation biomarkers. Environ Int 2016;92–93:173–182. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Corlin L, Woodin M, Hart JE, Simon MC, Gute DM, Stowell J, et al. Longitudinal associations of long-term exposure to ultrafine particles with blood pressure and systemic inflammation in Puerto Rican adults. Environ Health 2018;17:33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Weuve J, Puett RC, Schwartz J, Yanosky JD, Laden F, Grodstein F. Exposure to particulate air pollution and cognitive decline in older women. Arch Intern Med 2012;172:219–227. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Kulick ER, Wellenius GA, Boehme AK, Joyce NR, Schupf N, Kaufman JD, et al. Long-term exposure to air pollution and trajectories of cognitive decline among older adults. Neurology 2020;94:e1782–e1792. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Duchesne J, Gutierrez LA, Carriere I, Mura T, Chen J, Vienneau D, et al. Exposure to ambient air pollution and cognitive decline: Results of the prospective Three-City cohort study. Environ Int 2022;161:107118. [DOI] [PubMed] [Google Scholar]
- 21.Younan D, Wang X, Millstein J, Petkus AJ, Beavers DP, Espeland MA, et al. Air quality improvement and cognitive decline in community-dwelling older women in the United States: A longitudinal cohort study. PLoS Med 2022;19:e1003893. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Weuve J, Bennett EE, Ranker L, Gianattasio KZ, Pedde M, Adar SD, et al. Exposure to Air Pollution in Relation to Risk of Dementia and Related Outcomes: An Updated Systematic Review of the Epidemiological Literature. Environ Health Perspect 2021;129:96001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Cleary EG, Cifuentes M, Grinstein G, Brugge D, Shea TB. Association of Low-Level Ozone with Cognitive Decline in Older Adults. J Alzheimers Dis 2018;61:67–78. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Loop MS, Kent ST, Al-Hamdan MZ, Crosson WL, Estes SM, Estes MG Jr., et al. Fine particulate matter and incident cognitive impairment in the REasons for Geographic and Racial Differences in Stroke (REGARDS) cohort. PLoS One 2013;8:e75001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Cullen B, Newby D, Lee D, Lyall DM, Nevado-Holgado AJ, Evans JJ, et al. Cross-sectional and longitudinal analyses of outdoor air pollution exposure and cognitive function in UK Biobank. Sci Rep 2018;8:12089. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Oudin A, Forsberg B, Lind N, Nordin S, Oudin Astrom D, Sundstrom A, et al. Is Long-term Exposure to Air Pollution Associated with Episodic Memory? A Longitudinal Study from Northern Sweden. Sci Rep 2017;7:12789. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Health Effects Institute. Understanding the health effects of ambient ultrafine particles. Available at: https://www.healtheffects.org/system/files/Perspectives3.pdf. Accessed: May 16, 2023.
- 28.Schraufnagel DE. The health effects of ultrafine particles. Exp Mol Med 2020;52:311–317. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.National Alzheimer’s Coordinating Center. About NACC Data. Available at: https://naccdata.org/requesting-data/nacc-data. Accessed: May 16, 2023.
- 30.Morris JC, Weintraub S, Chui HC, Cummings J, Decarli C, Ferris S, et al. The Uniform Data Set (UDS): clinical and cognitive variables and descriptive data from Alzheimer Disease Centers. Alzheimer Dis Assoc Disord 2006;20:210–216. [DOI] [PubMed] [Google Scholar]
- 31.Beekly DL, Ramos EM, Lee WW, Deitrich WD, Jacka ME, Wu J, et al. The National Alzheimer’s Coordinating Center (NACC) database: the Uniform Data Set. Alzheimer Dis Assoc Disord 2007;21:249–258. [DOI] [PubMed] [Google Scholar]
- 32.Besser L, Kukull W, Knopman DS, Chui H, Galasko D, Weintraub S, et al. Version 3 of the National Alzheimer’s Coordinating Center’s Uniform Data Set. Alzheimer Dis Assoc Disord 2018;32:351–358. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Weintraub S, Besser L, Dodge HH, Teylan M, Ferris S, Goldstein FC, et al. Version 3 of the Alzheimer Disease Centers’ Neuropsychological Test Battery in the Uniform Data Set (UDS). Alzheimer Dis Assoc Disord 2018;32:10–17. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Saha PK, Hankey S, Marshall JD, Robinson AL, Presto AA. High-Spatial-Resolution Estimates of Ultrafine Particle Concentrations across the Continental United States. Environ Sci Technol 2021;55:10320–10331. [DOI] [PubMed] [Google Scholar]
- 35.Bates D, Mächler M, Bolker B, Walker S. Fitting linear mixed-effects models using lme4. J Stat Softw 2015;67:1–48. [Google Scholar]
- 36.Rozzi GC. zipcodeR: Advancing the analysis of spatial data at the ZIP code level in R. Software Impacts 2021;9:100099. [Google Scholar]
- 37.Karner AA, Eisinger DS, Niemeier DA. Near-roadway air quality: synthesizing the findings from real-world data. Environ Sci Technol 2010;44:5334–5344. [DOI] [PubMed] [Google Scholar]
- 38.Lane KJ, Levy JI, Scammell MK, Patton AP, Durant JL, Mwamburi M, et al. Effect of time-activity adjustment on exposure assessment for traffic-related ultrafine particles. J Expo Sci Environ Epidemiol 2015;25:506–516. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Van Roosbroeck S, Li R, Hoek G, Lebret E, Brunekreef B, Spiegelman D. Traffic-related outdoor air pollution and respiratory symptoms in children: the impact of adjustment for exposure measurement error. Epidemiology 2008;19:409–416. [DOI] [PubMed] [Google Scholar]
- 40.Dionisio KL, Chang HH, Baxter LK. A simulation study to quantify the impacts of exposure measurement error on air pollution health risk estimates in copollutant time-series models. Environ Health 2016;15:114. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The data presented in this study are available upon request from the National Alzheimer’s Coordinating Center (NACC) at the University of Washington.