Person-specific digital measurements of air pollutant exposure and biomarkers of Alzheimer's disease: Findings from a pilot study

Abstract

Background

Air pollution has been almost exclusively measured by region-level ambient exposure of particulate matter (PM) from government agencies (e.g., EPA), which does not capture the potentially large within-region variation and takes no consideration of human mobility in daily life. The use of region-level PM may have systematically biased the estimated associations with Alzheimer disease (AD).

Objective

To obtain pilot data on person-specific digital measurements of exposure to PM and volatile organic compounds (VOC) and to correlate them with AD biomarkers, including plasma and cerebrospinal fluid (CSF) analytes and amyloid and tau positron emission tomography (PET) signal, and to generate hypotheses about these correlations.

Methods

Fifty nine older participants underwent a 7-day digital collection of exposure to PM and VOC by Atmotube Pro, a validated wearable air quality sensor. Person-specific levels of PM and VOC were correlated with AD biomarkers. Sample sizes for future studies to detect the observed correlations were estimated.

Results

The first functional principal components (FFPC) of 7-day exposure to PM and VOC captured 75% to 86% of the total variation. The FFPC of PM₁, PM_2.5, PM₁₀ was correlated with CSF total tau (r = 0.63, 0.60, and 0.60, respectively; p < 5%). The FFPC of VOC was marginally correlated with amyloid biomarkers (r = −0.41 with CSF Aβ_42/40, r = 0.45 with CSF Aβ₄₀, and r = 0.25 with PET amyloid centiloid).

Conclusions

The pilot study generated the hypothesis that person-level air pollutant exposure to PM and VOC is associated with AD biomarkers, but larger studies are needed to test the hypothesis.

Keywords

Alzheimer's disease amyloid biomarkers particulate matter tau biomarkers volatile organic compounds

Introduction

Many environmental factors are associated with Alzheimer’a disease (AD) and related dementia (ADRD).^1–3 In 2019 the World Health Organization (WHO) estimated that ∼99% of the world population was living in regions where the levels of air pollutants exceeded the WHO air quality guidelines. Fine particulate matter (PM) is a mixture of solid particles and liquid droplets found in the air. Recent studies have found that higher exposure to the PM with diameters of 2.5 micrometers or smaller (PM_2.5) was associated with lower concentrations of Aβ₄₂ in cerebrospinal fluid (CSF),¹ higher likelihood of a positive amyloid PET scan,² more amyloid plaques in autopsied brains,³ and worse cognition with increased incidence of ADRD.^4,5 Higher levels of volatile organic compounds (VOC) were also found in AD patients compared with normal controls.^6,7 However, all existing research linking air pollutant exposure with AD biomarkers is based on regional, not person-specific, measurements of outdoor ambient air quality by government agencies (e.g., US Environmental Protection Agency [EPA] and European Environment Agency [EEA]). Whereas region-level measures of air pollutant exposure are important and scalable, they do not account for potentially large variability of outdoor air quality within the same regions or human factors due to mobility in daily life (e.g., time spent indoor vs. outdoor).⁸ The use of region-level measurements of PM exposure in associational analyses with person-level measurements of AD biomarkers may hence lead to significant statistical bias in the estimated effect size and loss of statistical power.

The objectives of this pilot study were to assess the feasibility to obtain person-specific data on air pollutant exposure to PM and VOC, and to generate hypotheses about their associations with person-specific AD biomarkers. These biomarkers include those from the molecular imaging of amyloid and tau pathology with positron emission tomography (PET), plasma and cerebrospinal fluid (CSF) analytes,^9–12 brain volumes as measured by magnetic resonance imaging (MRI), and cognitive outcomes.

Methods

Participants

This study included older community-dwelling individuals recruited from the greater St Louis, Missouri, metropolitan area and enrolled in the longitudinal studies of memory and aging at the Washington University (WU) Knight Alzheimer Disease Research Center (ADRC). Details of recruitment have been described.^13,14 Participants were predominantly cognitively normal at the time of the cross-sectional study, but some had early symptomatic AD (mild cognitive impairment, very mild dementia, or mild dementia with a typical AD dementia syndrome). Individuals were excluded if they had illnesses that could prevent participation in neuroimaging, or adversely affect cognition (e.g., metastatic cancer). The inclusion criteria for the current study were (1) availability of data of cognitive measures and AD biomarkers for at least one of the following biomarker modalities: plasma or CSF biomarkers; amyloid PET; structural MRI; and (2) normal cognition or early symptomatic AD. Potential participants meeting inclusion criteria were referred by Knight ADRC Clinical Core to the study and then contacted by the study coordinator for their willingness to participate. All participants provided written informed consent at recruitment. The WU Human Research Protection Office approved procedures.

Demographics and social determinants of health

Age, sex, years of education, and self-identified race were collected. Social determinants of health (SDOH) were assessed by a comprehensive battery that was recently developed at the Knight ADRC, referred to as the Social and Structural Determinants Influencing Aging and Dementia battery (SS-DIAD).¹⁵

Clinical and cognitive assessments

The Knight ADRC's clinical and cognitive assessment protocols are consistent with that of the National Alzheimer Coordinating Center (NACC) Uniform Data Set (UDS).¹⁶ The NACC UDS includes standard diagnostic criteria for dementia and its differential diagnoses,¹⁷ and uses the global Clinical Dementia Rating^® (CDR^®)¹⁸ to operationalize the presence or absence of dementia, and when present, the severity of dementia. The UDS cognitive battery includes measures of episodic memory, working memory, semantic knowledge, executive function and attention, and visuospatial ability. All scales were oriented so that a larger score indicates better cognition. Z-scores were computed by subtracting the mean from the individual test scores and dividing the difference by the standard deviation (SD). A global cognitive composite score was constructed by averaging Z-scores across tests, as previously described.¹⁹

APOE genotypes

Details of the APOE genotyping protocols have been described previously.²⁰ We dichotomized APOE ε4 carrier status as positive or negative, indicating presence of one or two APOE ε4 alleles, or none.

Digital collection of person-specific air pollutant exposure

Digital measurements of person-specific air pollutant exposure were obtained by a commercially available and wearable air quality sensor, Atmotube Pro, by Atmotube.²¹ The device is designed with a carabiner so that it may be worn easily and can be paired with a smartphone for pinging GPS coordinates (Figure 1). Participants were asked to wear the device continuously for 7 days as they navigated their daily life. PM₁, PM_2.5, PM₁₀, and VOC were measured about once every 5 min. Data were uploaded to the Cloud storage and then downloaded into our local secure servers for analyses after participants mailed their device back.

Figure 1.

Atmotube Pro linked with a smart phone.

Measurement properties of person-specific air pollutant exposure by Atmotube Pro have previously been established. Field and laboratory evaluations done by the South Coast Air Quality Monitoring District (AQMD) and its Air Quality-Sensor Performance Evaluation Center (AQ-SPEC) program,²² an independent government agency, showed very strong correlations of PM₁ and PM_2.5 from Atmotube Pro sensors with EPA-approved Federal Equivalent Method (FEM) GRIMM in the laboratory evaluations (r > 0.99), and strong correlations with the same reference instrument in the field evaluation (0.79 <r²< 0.90; see the South Coast AQMD website). Test-retest reliability was provided in Section (Test-retest reliability of person-specific air pollutant exposure by Atmotube Pro) below.

CSF and plasma sample collection and analysis

Blood was collected at the time of lumbar puncture (LP) or clinical assessment. CSF samples (20–30 mL) were collected at 8 AM after overnight fasting by gravity drip, briefly centrifuged at low speed, and aliquoted into polypropylene tubes prior to freezing at −80°C. All plasma samples (fasted and non-fasted) were analyzed at C₂N Diagnostics with the PrecivityAD1™ assay for Aβ_42/40 and the PrecivityAD2™ for p-tau217, which has previously been described.^23,24 An automated immunoassay (LUMIPULSE G1200, Fujirebio, Malverne, PA) was used to measure CSF concentrations of Aβ₄₀, Aβ₄₂, total tau (t-tau), p-tau181, and NfL.^25,26 Plasma NfL and GFAP were measured with Quanterix assay kits at WU on a HD-X analyzer.^27,28

Brain imaging scan collection and processing

The WU MRI protocol was implemented on 3T scanners for a total of approximately 60 min, and provided high gray-white matter contrast (1 mm×1 mm×1 mm) MPRAGE T1-weighted volume acquisitions.²⁹ PET PiB or Florbetapir (AV45) imaging and tau Flortaucipir (AV1451) imaging were conducted as detailed previously.^29,30 Structural MRI processing steps include motion correction, if applicable, averaging across scans, and atlas transformation. Regional volumes and cortical thickness were obtained via the FreeSurfer image analysis suite.³¹ Determination of the regions-of-interest (ROIs) and pipelines were previously describe.³² FreeSurfer quality control (QC) criteria include (1) dural inclusion, gray matter exclusion, sulcus inclusion, cerebellum inclusion; (2) cerebellum, subcortical and hippocampus segmentation exclusion; (3) white matter exclusion, and (4) lateral ventricle segmentation. Analyses were performed on adjusted volumetric measures after regressing for the effect of scanner platforms and with adjustment for head size.³² Amyloid and tau deposition in the ROIs was determined using FreeSurfer, and a standardized uptake value ratio (SUVR) with correction for partial volume effects was calculated.³² The cerebellum was chosen as the reference region. The mean amyloid cortical SUVR was calculated from FreeSurfer regions within the prefrontal cortex, precuneus, and temporal cortex.²⁹ A centiloid scale to harmonize PET amyloid data from PiB and AV45³³ was developed, following the same level-2 calibration procedure in the original Centiloid Working Group paper.³⁴ Tau PET scans were collected and processed as previously described.¹²

Statistical analyses

We first estimated the test-retest reliability of person-specific measurements of PM and VOC exposure over 7-days. i.e., the intraclass correlation coefficient (ICC),³⁵ using within-person consecutive measurements every 15 min. We then demonstrated that person-specific digital measurements of air pollutant exposure can capture between-individual variation in the same region and also within-individual variation in the same day by showing that participants from the same region do not share the same levels of PM exposure at the same time point, and that significant within-individual variation also existed across time. We next conducted a simulation study to assess the potential statistical bias and loss of statistical power when region-level measures of air pollutant exposure were used in the correlational analyses with AD biomarkers, in comparison to the person-specific measures. Finally, we analyzed 7-day air pollutant exposure data to PM and VOC by extracting main features from the time series, and then correlated these features with the AD biomarkers. The first feature was the average daily exposure. The second feature was the proportion of time in 7 days that person-specific PM exposure levels were over the limit as recommended by the WHO (15 µg/m3 for PM₁ and PM_2.5, 45 µg/m3 for PM₁₀, and 0.05 ppm for VOC). The third feature was the first principal component (FFPC) after performing a functional principal component analysis (FPCA)³⁶ of 7-day air pollutant exposure for each index.

The association analysis was conducted to correlate each of the features from the 7-day air pollutant exposure with all AD biomarkers and cognitive outcomes that were obtained prior to the digital assessments of air pollutant exposure. Because not all participants had data on all biomarker modalities, each pairwise correlation of an exposure index with each biomarker was estimated using the participants for whom the bivariate data were available. These pairwise correlations were tested for statistical significance by a standard normal test after the Fisher's Z-transformation.³⁷ This test was chosen because Fisher's Z-transformation can be approximated by a normal distribution regardless of the sample sizes, assuring the validity of our tests for a large number of pairwise correlations estimated from a wide range of sample sizes. Because of the preliminary nature of the pilot study with a primary goal to generate possible hypotheses for a future definite study, no rigorous multiplicity adjustment was performed. Estimates of correlations were adjusted for the effect of the time interval from the collection of AD biomarkers/cognition to the digital collection of air pollutant exposure as well as the seasons of the exposure, and additionally adjusted for the effects of other covariates: age, race, cognitive status, APOE ε4 status, years of education, sex, and medical comorbidities.

Finally, after generating a set of hypotheses about the correlations between person-specific air pollutant exposure and AD biomarkers, we performed a power analysis to decide the sample sizes needed for future studies to rigorously test these hypotheses.

Results

A total of 59 individuals from the WU Knight ADRC participated in the pilot study. The participants had an average age of 76.28 years (SD = 5.28 years), an average of 16.29 (SD = 2.00) years of education, and were mostly female (58%) and predominantly White (85%) and cognitively normal (88%). About 39% of the participants were APOE ε4 carriers. In addition to digital assessment of air pollutant exposure over 7 days, all 59 participants in our pilot study had NACC UDS.¹⁶ A subset had data on CSF Aβ_42/40, CSF total tau, p-tau181 (n = 20) and multiple other CSF tau species (217,153,175, 205, 231; n = 14) concentrations, plasma Aβ_42/40 (n = 54) and p-tau217 (n = 22), cortical thickness and hippocampal volume (n = 41), amyloid PET (n = 35), and tau PET (n = 33). Data collection of AD biomarkers and cognition occurred mostly prior to the data collection of air pollutant exposure, with a median time interval of between 0.9 years (cognition) and 5.9 years (CSF). Table 1 presents the basic demographics of the study participants. A total of 57 participants had data on CSF Aβ_42/40 and/or amyloid PET, and 28 (49%) were amyloid positive by either CSF or PET amyloid measures.

Table 1.

Demographics, clinical characteristics, and air pollutant exposure of study participants.

Characteristic	N = 59^a
Amyloid positivity (CSF or amyloid PET)
Negative	29 (51%)
Positive	28 (49%)
CDR
0	52 (88%)
0.5	7 (12%)
Race
Black	9 (15%)
White	50 (85%)
Sex
F	34 (58%)
M	25 (42%)
APOE ε4 status
Negative	36 (61%)
Positive	23 (39%)
Education (y)	16.29 (2.00)
Age (y)	76.28 (5.28)
Daily mean of PM₁ (µg/m³)	7.11 (13.92)
Daily mean of PM_2.5 (µg/m³)	8.93 (14.93)
Daily mean of PM₁₀ (µg/m³)	10.72 (15.47)
Daily mean of VOC (ppm)	0.37 (0.24)

n (%) or Mean (SD).

Test-retest reliability of person-specific air pollutant exposure measured by atmotube pro

From the 59 participants who underwent 7-day person-specific measurements of PM exposure, the within-person consecutive measurements every 15 min were used to estimate the test-retest reliability of PM₁, PM_2.5, PM₁₀, and VOC. Figure 2 presents the estimated intraclass correlation coefficient (ICC) as a function of time during a randomly selected day of 24 h starting from midnight across participants, indicating very good test-retest reliability of the person-specific measurements of PM and VOC by the device. Not surprisingly, ICCs were lower around the early morning and lunch times, reflecting changes in the daily routine when air pollutant exposure was subject to sudden changes.

Figure 2.

Intraclass correlation coefficient (ICC) of air pollutant exposure over the course of a day. The ICC (y-axis) was computed for each index (PM₁, PM_2.5, PM₁₀, and VOC) between every possible pair of adjacent time windows of 15 min each (starting and ending at midnight), and shown as a function of the starting time. A smoothed ICC curve for each index was obtained by the method of Locally Estimated Scatterplot Smoothing (LOESS).

Person-specific digital measurements of air pollutant exposure captured between-individual variation in the same region and within-individual variation in the same day

Among the 59 participants who underwent 7-day digital collection of their air pollutant exposure, eight from the same St Louis region were measured on the same day (12/01/2023). Figure 3 presents the time course of PM_2.5 levels from the 8 participants from midnight to midnight, demonstrating that participants from the same region did not share the same levels of PM exposure at the same time point (as EPA's region-level data would imply), and that significant within-individual variation also existed across the 24 h.

Figure 3.

Between- and within-individual variations of PM_2.5 levels from 8 participants of the same region and in the same 24 h on 12/01/2023 (different colors represent different participants).

Statistical bias and loss of power when region-level measurements of air pollutant exposure were used in the correlational analyses with AD biomarkers

To demonstrate the importance of person-specific measurements of air pollutant exposure, we simulated 5000 independent data sets to assess the statistical bias and loss of statistical power when a region-level PM exposure was correlated with a person-level AD biomarker. Each data set was originally simulated at the person level for both the PM exposure and biomarker over 100 participants from a bivariate normal distribution with a true correlation (r) of 0.15 or 0.45. We then created the region-level PM exposure for a randomly selected subset of 100p% (0 < p < 1) participants, using the mean of the simulated person-specific exposures in the subset. Finally, we estimated the correlation between the PM exposure and biomarker and tested its statistical significance. Figure 4 shows the mean bias (defined as the difference between the mean of the 5000 estimated correlations and the true correlations: left vertical axis, red color; true r = 0.15 = solid line, true r = 0.45 = dashed line) and the mean loss of statistical power (defined as the difference between the empirical power of the 5000 tests and the true power: right vertical axis; green color) as functions of p, the proportion of subjects with only region-level data of PM exposure, demonstrating increased bias and loss of statistical power as p increases. Note that, when person-level data were used in PM exposure, i.e., p = 0, no bias (=0, left vertical axis) and no loss of statistical power (=0, right vertical axis) were observed.

Figure 4.

Bias (left vertical axis, red color) and loss of statistical power (right vertical axis, green color) for the estimated correlation between an air pollutant exposure index and an AD biomarker when a proportion of participants (p) had only region-level data on air pollutant exposure (two true correlations were plotted: 0.15 in solid line and 0.45 in dashed line).

Correlation of person-specific exposure to PM and VOC with AD biomarkers

The estimated Spearman correlations of person-specific daily PM₁, PM_2.5, and PM₁₀ exposure (averaged over 7 days) with CSF p-tau181 and total tau showed a strong statistical trend and ranged from 0.35 to 0.59. PM₁₀ exposure was also correlated with CSF p-tau153 (r = 0.82, p = 0.044) and showed a trend with the tau PET burden (r = 0.22) and plasma p-tau217 (r = 0.17). There was also a trend towards statistical significance in the correlation of VOC exposure with plasma p-tau217 (r = 0.51, p = 0.052), p-tau181 (r = 0.41, p = 0.134), and CSF p-tau217 (r = 0.69, p = 0.089). The estimated correlation of person-specific daily VOC level was −0.37 (p = 0.210) with CSF Aβ_42/40, −0.34 (p = 0.019) with plasma Aβ₄₂, −0.16 (p = 0.278) with plasma Aβ_42/40, and 0.20 (p = 0.3) with amyloid PET centiloid.

The FFPC derived from a functional principal component analysis (FPCA)³⁶ of 7-day air pollutant exposure for each index captured most of the variation across subjects (86.49%, 86.28%, 86.11%, and 74.80% for PM₁, PM_2.5, PM₁₀ and VOC, respectively), and may be interpreted as person-specific ‘total exposure’. The correlation matrix (Table 2) showed positive correlations of the FFPC for both PM and VOC with almost all tau biomarkers, and some were statistically significant (r = 0.63, 0.60, and 0.60 between PM₁, PM_2.5, PM₁₀ and CSF total tau, p < 5%). The FFPC for VOC was marginally correlated with amyloid biomarkers (r = −0.41 with CSF Aβ_42/40, 0.45 with CSF Aβ₄₀, and 0.25 with PET amyloid). Analyses after adjusting for the time interval from collections of AD biomarkers to the digital collection of air pollutant exposure and the seasons of the exposure revealed largely consistent estimates: the FFPC of PM₁, PM_2.5, PM₁₀ were significantly correlated with CSF total tau (r = 0.88, 0.83, and 0.82, respectively; p < 5%), and the FFPC for VOC was marginally correlated with amyloid biomarkers (r = −0.34 with CSF Aβ_42/40, 0.35 with CSF Aβ₄₀, and 0.26 with PET amyloid). No correlations of air pollutant exposure were found with cognitive composite.

Table 2.

Correlations of the FFPC of air pollutant exposure (PM, VOC) with biomarkers of AD across modalities and their 95% CIs (*indicates raw p-value < 5%).

A/T/N Biomarker		PM₁	PM_2.5	PM₁₀	VOC
A	Amyloid PET centiloid	−0.05 (−0.42, 0.33)	−0.07 (−0.43, 0.32)	−0.10 (−0.46, 0.28)	0.25 (−0.14, 0.57)
	CSF Aβ_42/40	−0.18 (−0.67, 0.41)	−0.13 (−0.63, 0.46)	−0.09 (−0.61, 0.48)	−0.41 (−0.78, 0.19)
	CSF Aβ₄₀	0.26 (−0.34, 0.71)	0.23 (−0.37, 0.69)	0.22 (−0.37, 0.69)	0.45 (−0.13, 0.80)
	Plasma Aβ_42/40	0.06 (−0.23, 0.34)	0.07 (−0.22, 0.35)	0.09 (−0.21, 0.36)	−0.09 (−0.37, 0.20)
T	Tau PET burden	0.26 (−0.14, 0.59)	0.30 (−0.10, 0.61)	0.30 (−0.10, 0.62)	−0.02 (−0.41, 0.37)
	CSF p-tau181	0.38 (−0.22, 0.77)	0.37 (−0.23, 0.76)	0.37 (−0.23, 0.76)	0.49 (−0.08, 0.82)
	CSF total tau (t-tau)	0.63* (0.12, 0.88)	0.60* (0.07, 0.86)	0.60* (0.08, 0.87)	0.19 (−0.41, 0.67)
	Plasma p-tau217	0.28 (−0.27, 0.69)	0.33 (−0.22, 0.72)	0.33 (−0.21, 0.72)	0.30 (−0.25, 0.70)
N	Hippocampal volume	−0.06 (−0.39, 0.28)	−0.02 (−0.35, 0.32)	−0.07 (−0.39, 0.28)	−0.02 (−0.36, 0.32)
N	Cortical thickness	0.18 (−0.16, 0.49)	0.04 (−0.30, 0.37)	−0.07 (−0.40, 0.27)	−0.10 (−0.43, 0.24)

Correlations of person-specific time exposed to PM levels over WHO's recommended limit correlated with AD biomarkers

Using WHO's recommended limits for daily PM and VOC exposure, we computed the proportion of time over 7 days that person-specific PM exposure levels were over the limit for each participant. Times of excessive exposure to PM₁, PM_2.5, and PM₁₀ were all significantly correlated with CSF total tau (r = 0.85, 0.82, and 0.72, p = 0.0002, 0.0005, and 0.006, respectively), and showed a statistical trend with CSF p-tau181 (r = 0.36, 0.38, and 0.31, respectively). Time of excessive exposure to VOC was also marginally correlated with CSF Aβ₄₂ (r = −0.2), plasma NfL (r = 0.25) and cortical thickness (r = −0.18).

Correlation of person-specific exposure to PM and VOC with SDOH

Although not significant, we found a statistical trend indicating that people who reported living in very littered neighborhoods were exposed to higher levels of PM₁₀ (r = 0.24, p = 0.131), and those feeling a strong sense of belonging to their neighborhoods were exposed to less VOC (r = −0.34 with the FFPC, p = 0.03).

Sample sizes needed in a future study to power the test of the observed correlations

Due to the small sample size in our pilot study, very few of these correlations reported above were statistically significant after adjusting for multiple comparisons. However, multiple AD biomarkers had correlations with pollutants in the same direction (great biomarker abnormality was associated with higher levels of pollutants), suggesting a relationship between AD and pollution. More specifically, the statistical trends and relatively large magnitude for multiple of the estimated correlations support the hypothesis that both daily exposure levels to PM and VOC and duration of excessive exposure over the WHO's recommended limits are associated with tau biomarkers (plasma p-tau217, CSF p-tau217, CSF p-tau181, and tau PET burden) and that VOC exposure is also associated with amyloid and neurodegeneration biomarkers (plasma/CSF Aβ_42/40, amyloid PET amyloid burden, CSF NfL, and cortical thickness). To rigorously test these hypotheses, larger studies are needed. We hence performed a power analysis to determine the sample sizes needed to test these hypotheses for future studies. Figure 5 presents the sample sizes (n) needed for a future study with 80% statistical power to test a single correlation at 5% level, and multiple correlations of multiple air pollutant exposure measures with multiple AD biomarkers of AD at 5% level with rigorous Bonferroni's adjustment. These power analyses were based on a two-sided standard normal test after the Fisher's Z-transformation.³⁷ The direction of the correlation was assumed positive, but the results apply when the direction of the correlation is negative as well.

Figure 5.

Sample sizes (n) needed with 80% power for a future study to test a single correlation or multiple correlations (with Bonferroni's adjustment) between person-specific air pollutant exposure and AD biomarkers at 5% significance level.

Discussion

Associations of air pollutant exposure with AD in general, and AD biomarkers in particular, remain poorly understood. In a recently updated meta-analysis,³⁸ the authors concluded that “strong conclusions remain elusive, although the weight of the evidence suggests an adverse association between PM_2.5 and cognitive decline”, and further noted “a continued need to confront methodological challenges in this line of research.” A major methodological challenge is that all existing research linking air pollutant exposure with AD biomarkers is based on regional, not person-specific, measurements from government agencies (EPA and EEA) outdoor ambient air quality monitoring. For example, only ∼15 monitoring stations of PM are currently used in the entire State of Missouri, each providing a single measurement of PM levels to a large neighboring region at each time. Our pilot study indicated that the use of region-level measurements of outdoor ambient PM exposure in associational analyses with person-level measurements of AD biomarkers leads to significant bias in the estimated effect size, and reduces the statistical power. One explanation is that these data ignore the potentially large within-region variation of outdoor air quality and indoor air quality. More specifically, we found that a commercially available and wearable air quality sensor provides reliable and valid person-specific measurements of both outdoor and indoor air pollutant exposure, which further demonstrated that substantial between-individual variation exists in person-specific measurements of air pollutant exposure from participants within the same region at the same time. We also found that person-specific measures of air pollutant exposure led to less biased estimates to the correlations between air pollutant exposure and AD biomarkers in comparison to region-level data of air pollution provided by the government agencies.

Perhaps most importantly, our pilot study suggested that person-specific exposure to PM, and time exposed to excessive levels of PM over WHO's recommended limits, correlated with tau biomarkers, and the correlations with amyloid biomarkers seemed to be relatively weak. These preliminary findings are not entirely consistent with the current and limited literature that mostly reported associations of PM_2.5 with CSF Aβ or amyloid PET biomarkers. The inconsistency may be due to differences in study design, populations under study, and methodologies. Particularly, because almost all studies in the literature were based on region-level PM exposure, given our finding that region-level exposure data may lead to substantial bias in association analyses with AD biomarkers, the inconsistency highlights the need to obtain person-specific air pollutant exposure in future large correlational studies between air pollution and AD biomarkers. Further, because biomarker changes in tau are ‘closer’ (than amyloid) to changes in cognition,³⁹ our findings are hence consistent with the reported effect of air pollutant exposure on cognition.^40,41 However, the biological mechanism of a stronger association with tau (than amyloid) pathology remains an open question, more research is needed in this direction.

Interestingly, we found that person-specific exposure to VOC is not only correlated with tau biomarkers, but also with amyloid biomarkers. Given that concentrations of VOC are up to ten times higher indoors than outdoors,⁴² our pilot data suggest that it is important to measure both person-specific indoor and outdoor air quality, as they may influence AD biomarkers. We also found that the amount of time exposed to excessive levels of VOC may be correlated with MRI-based neurodegeneration biomarkers including structural outcomes, implying that air pollution may contribute to the neurodegeneration that is associated with AD. Finally, our pilot data seem to confirm that person-specific air pollutant exposure may depend on the characteristics of neighborhoods people live in and the associated SDOH. These results may imply that in future large studies to associate person-specific air pollutant exposure with AD biomarkers, SDOH such as littering in the neighborhoods and how people feel about their neighborhoods are important factors to consider in the design and analyses of these studies.

Our pilot study focused on the associations of person-specific air pollutant exposure with AD biomarkers among individuals who were mostly cognitively normal or at very early symptomatic stages of AD. Because of converging scientific evidence that indicates neurodegenerative processes associated with AD may begin at middle age^39,43 and years or even decades prior to symptom onset when the disease is clinically at the early latent stage,^44–46 biomarkers from biofluid and imaging have guided recent clinical trials on AD by allowing trialists to confirm AD pathology in impaired and cognitively normal individuals (for treatment and prevention trials, respectively) and to conduct efficacy analyses on AD pathology. Hence, biomarkers of AD have played a central role in the recent success of disease-modifying therapies of AD that targeted amyloid pathology.^47,48 However, it remains unknown whether the treatment effects may be confounded with air pollutant exposure because these trials did not obtain person-specific measures of air pollutant exposure. Our findings, if confirmed by large and representative future longitudinal studies, may inform design and analysis of future natural history studies and clinical trials for prevention and treatment of AD to adequately account for possible contributions of air pollutant exposure to AD biomarkers and design these studies to appropriately control the effect of air pollution in comparative analyses.

The main strengths of this pilot study include the person-specific measurements of air pollutant exposure including both indoor and outdoor exposures, and availability of AD biomarkers from all modalities, including plasma, CSF, amyloid PET, tau PET, and MRI. Main limitations of the pilot study include small sample size, short term (7 days) and non-cumulative measurements of air pollutant exposure, lack of mechanism assessment between air pollution and AD biomarkers, and lack of longitudinal data. Another major limitation is the relatively large interval (0.9 to 5.9 years) between the retrospectively obtained biomarker data and the prospectively collected data on air pollution exposure. Whereas this is necessary to leverage existing biomarker data in our pilot study, it prevents a causal inference to be made on the association between AD biomarkers and air pollutant exposures. Future large and longitudinal studies are needed to obtain both person-specific air pollutant exposure and AD biomarkers, as well as potentially thousands of different exposure chemicals and metabolites in blood that may be linked with both air pollutant exposure and AD biomarkers. These future studies must be designed to have a much shorter time interval between collections of biomarker and air pollution exposure data, a much larger sample size to allow stratified analyses by major AD risk factors including APOE genotype (for an assessment of gene-environment interaction), a balanced measurement of air pollutant exposure across all seasons, and an assessment of person-specific daily activities both indoors and outdoors (in addition to air pollutant exposure) to allow analyses of possible differential relationships of indoor and outdoor air pollution with AD biomarkers.

Footnotes

Acknowledgements

We would like to extend our gratitude to the participants from WU Knight ADRC who generously volunteered their time to contribute to this study.

Ethical considerations

The study was approved by the WU Human Research Protection Office approved procedures.

Consent to participate

All participants in this study provided written informed consent prior to participation and any study procedures.

Author contributions

Chengjie Xiong: Conceptualization, Data curation, Writing – original draft, Writing – review & editing.

Ruijin Lu: Formal analysis.

Quoc Bui: Formal analysis.

Brent Popp: Data curation.

Suzanne E Schindler: Conceptualization, Investigation.

Leah P Shriver: Conceptualization, Investigation.

Carlos Cruchaga: Conceptualization, Investigation.

Jason Hassenstab: Conceptualization, Investigation.

Tammie LS Benzinger: Conceptualization, Investigation.

Folasade Agboola: Data curation.

Emily Gremminger: Project administration.

Marissa L Streitz: Project administration.

Charles Hall: Conceptualization, Investigation.

Krista L Moulder: Project administration.

John C Morris: Conceptualization, Investigation, Writing – review & editing.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Declaration of conflicting interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Drs. Xiong, Schindler, Shriver, Benzinger, and Morris all have received research funding from the National Institute on Aging of the National Institutes of Health that was made to their institutions.

Dr Hall has received research funding from the National Institute for Occupational Safety and Health of the Centers for Disease Control that was made to his institution.

Dr Morris is funded by NIH grants # P30 AG066444; P01AG003991; P01AG026276;

Neither Dr Morris nor his family owns stock or has equity interest (outside of mutual funds or other externally directed accounts) in any pharmaceutical or biotechnology company.

Dr Xiong consults for Diadem and Intune Bio. There are no conflicts.

Drs. Xiong and Benzinger and Hall served on Data Safety Monitoring Board or Advisory Boards for FDA and/or NIH-funded studies. There are no conflicts.

Dr Schindler has served on advisory boards or received speaker fees from Eisai, Eli Lilly, and Novo Nordisk.

Dr Benzinger has had consulting agreements with Biogen, Eisai, Merck, and Jansen. There are no conflicts.

Dr Benzinger received payment or honoraria for lectures, presentations, speakers bureaus, manuscript writing, or educational events from Biogen, Lilly, Peer View, and Medscape. There are no conflicts.

Dr Benzinger has had research grants from Siemens Healthineers and Hyperfine.

Avid Radiopharmacueticals (a wholly owned subsidiary of Eli Lilly), LMI and Lantheus have provided reagents and technology transfer agreements to Dr Benzinger's institution for the production of radiopharmaceuticals.

Dr Lu, Dr Bui, Dr Moulder, Mr Popp, Ms. Agboola, and Ms. Gremminger have nothing to disclose.

Data availability statement

The data supporting the findings of this study are available on request from the corresponding author.

References

Casey

Liang

, et al. Association between fine particulate matter exposure and cerebrospinal fluid biomarkers of Alzheimer's disease among a cognitively healthy population-based cohort. Environ Health Perspect 2024; 132: 47001.

Iaccarino

La Joie

Lesman-Segev

, et al. Association between ambient air pollution and amyloid positron emission tomography positivity in older adults with cognitive impairment. JAMA Neurol 2021; 78: 197–207.

Christensen

Liang

, et al. Association of PM_2.5 exposure and Alzheimer disease pathology in brain bank donors-effect modification by APOE genotype. Neurology 2024; 102: e209162.

Zhang

Weuve

Langa

, et al. Comparison of particulate air pollution from different emission sources and incident dementia in the US. JAMA Intern Med 2023; 183: 1080–1089.

Shi

Steenland

, et al. A national cohort study (2000–2018) of long-term air pollution exposure and incident dementia in older adults in the United States. Nat Commun 2021; 12: 6754.

Tiele

Wicaksono

Daulton

, et al. Breath-based non-invasive diagnosis of Alzheimer's disease: a pilot study. J Breath Res 2020; 14: 026003.

Mazzatenta

Pokorski

Sartucci

, et al. Volatile organic compounds (VOCs) fingerprint of Alzheimer's disease. Respir Physiol Neurobiol 2015; 209: 81–84.

Jung

Bernabé

Moors

, et al. Effects of floor level and building type on residential levels of outdoor and indoor polycyclic aromatic hydrocarbons, black carbon, and particulate matter in New York City. Atmosphere (Basel) 2011; 2: 96–109.

Klunk

Engler

Nordberg

, et al. Imaging brain amyloid in Alzheimer's disease with Pittsburgh Compound-B. Ann Neurol 2004; 55: 306–319.

10.

Xiong

Luo

Wolk

, et al. Baseline levels and longitudinal changes in plasma Aβ42/40 among Black and white individuals. Nat Commun 2024; 15: 5539.

11.

Xiong

Jasielec

Weng

, et al. Longitudinal relationships among biomarkers for Alzheimer disease in the adult children study. Neurology 2016; 86: 1499–1506.

12.

Mishra

Gordon

, et al. AV-1451 PET imaging of tau pathology in preclinical Alzheimer disease: defining a summary measure. Neuroimage 2017; 161: 171–178.

13.

Luo

Agboola

Grant

, et al. Sequence of Alzheimer disease biomarker changes in cognitively normal adults: a cross-sectional study. Neurology 2020; 95: e3104–e3116.

14.

Morris

Schindler

McCue

, et al. Assessment of racial disparities in biomarkers for Alzheimer disease. JAMA Neurol 2019; 76: 264–273.

15.

Streitz

Denny

Xiong

, et al. Development of a comprehensive battery to collect social and structural determinants of health (SSDOH) in cognitively normal or very mildly impaired persons. Alzheimer Dis Assoc Disord 2022; 36: 97–102.

16.

Morris

Weintraub

Chui

, 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.

17.

McKhann

Knopman

Chertkow

, et al. The diagnosis of dementia due to Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement 2011; 7: 263–269.

18.

Morris

. The clinical dementia rating (CDR) - current version and scoring rules. Neurology 1993; 43: 2412–2414.

19.

Bonomi

Schindler

, et al. Relationships of cognitive measures with cerebrospinal fluid but not imaging biomarkers of Alzheimer disease vary between black and white individuals. Ann Neurol 2024; 95: 495–506.

20.

Luo

Agboola

Grant

, et al. Accelerated longitudinal changes and ordering of Alzheimer disease biomarkers across the adult lifespan. Brain 2022; 145: 4459–4473.

21.

Atmotube Pro. https://www.aqmd.gov/aq-spec/sensordetail/atmotube—pro (2023, accessed 1 February 2024).

22.

AQ-SPEC. http://www.aqmd.gov/docs/default-source/aq-spec/summary/atmotube-pro—summary-report.pdf?sfvrsn = 8 (2023, accessed 1 February 2024).

23.

Kirmess

Meyer

, et al. Assessment of a plasma amyloid probability score to estimate amyloid positron emission tomography findings among adults with cognitive impairment. JAMA Netw Open 2022; 5: e228392.

24.

West

Kirmess

Meyer

, et al. A blood-based diagnostic test incorporating plasma Aβ42/40 ratio, ApoE proteotype, and age accurately identifies brain amyloid status: findings from a multi cohort validity analysis. Mol Neurodegener 2021; 16: 30.

25.

Alcolea

Pegueroles

Muñoz

, et al. Agreement of amyloid PET and CSF biomarkers for Alzheimer's disease on Lumipulse. Ann Clin Transl Neurol 2019; 6: 1815–1824.

26.

Barthélemy

Saef

, et al. CSF Tau phosphorylation occupancies at T217 and T205 represent improved biomarkers of amyloid and tau pathology in Alzheimer's disease. Nat Aging 2023; 3: 391–401.

27.

Yen

Hendrix

, et al. Timing of biomarker changes in sporadic Alzheimer's disease in estimated years from symptom onset. Ann Neurol 2024; 95: 951–965.

28.

Schindler

Karikari

Ashton

, et al. Effect of race on prediction of brain amyloidosis by plasma Aβ42/Aβ40, phosphorylated tau, and neurofilament light. Neurology 2022; 99: e245–e257.

29.

Gordon

Blazey

, et al. Longitudinal β-Amyloid deposition and hippocampal volume in preclinical Alzheimer disease and suspected non-Alzheimer disease pathophysiology. JAMA Neurol 2016; 73: 1192–1200.

30.

Brier

Gordon

Friedrichsen

, et al. Tau and Aβ imaging, CSF measures, and cognition in Alzheimer's disease. Sci Transl Med 2016; 8: 338ra66.

31.

Fischl

. Freesurfer. Neuroimage 2012; 62: 774–781.

32.

D'Angelo

Vlassenko

, et al. Quantitative analysis of PiB-PET with FreeSurfer ROIs. PLoS One 2013; 8: e73377.

33.

Flores

Wang

, et al. Comparison of Pittsburgh compound B and florbetapir in cross-sectional and longitudinal studies. Alzheimers Dement (Amst) 2019; 11: 180–190.

34.

Klunk

Koeppe

Price

, et al. The Centiloid Project: standardizing quantitative amyloid plaque estimation by PET. Alzheimers Dement 2015; 11: 1–15.e1-4.

35.

Weir

. Quantifying test-retest reliability using the intraclass correlation coefficient and the SEM. J Strength Cond Res 2005; 19: 231–240.

36.

Wang

Chiou

. Müller HG. Functional data analysis. Ann Rev Stat Application 2016; 3: 257–295.

37.

Graybill

. Theory and application of the linear model. North Scituate, MA: Duxbury, 1976.

38.

Weuve

Bennett

Ranker

, 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.

39.

Bateman

Xiong

Benzinger

, et al. Dominantly inherited Alzheimer network. Clinical and biomarker changes in dominantly inherited Alzheimer's disease. N Engl J Med 2012; 367: 795–804.

40.

Kulick

Wellenius

Boehme

, et al. Long-term exposure to air pollution and trajectories of cognitive decline among older adults. Neurology 2020; 94: e1782–e1792.

41.

Zhang

Chen

Zhang

. The impact of exposure to air pollution on cognitive performance. Proc Natl Acad Sci U S A 2018; 115: 9193–9197.

42.

EPA. https://www.epa.gov/indoor-air-quality-iaq/what-are-volatile-organic-compounds-vocs (2023, accessed 1 February 2024).

43.

Braak

. Neuropathological stageing of Alzheimer-related changes. Acta Neuropathol 1991; 82: 239–259.

44.

Price

Wade

, et al. Neuron number in the entorhinal cortex and CA1 in preclinical Alzheimer disease. Arch Neurol 2001; 58: 1395–1402.

45.

Sperling

Aisen

Beckett

, et al. Toward defining the preclinical stages of Alzheimer's disease: recommendations from the National Institute on Aging-Alzheimer's Association workgroups on diagnostic guidelines for Alzheimer's disease. Alzheimers Dement 2011; 7: 280–292.

46.

Jack

Jr Knopman

Weigand

, et al. An operational approach to National Institute on Aging-Alzheimer's Association criteria for preclinical Alzheimer disease. Ann Neurol 2012; 71: 765–775.

47.

van Dyck

Swanson

Aisen

, et al. Lecanemab in early Alzheimer's disease. N Engl J Med 2023; 388: 9–21.

48.

Sims

Zimmer

Evans

, et al. Donanemab in early symptomatic Alzheimer disease: the TRAILBLAZER-ALZ 2 randomized clinical trial. JAMA 2023; 330: 512–527.