Abstract:
Dementia has been considered a major global public health priority (WHO, 2012). It is very common in older adults (Alzheimer’s Association, 2015b; Seshadri & Wolf, 2007; Weuve et al., 2014) and characterized by the progressive and irreversible loss of memory and mental abilities (Alzheimer’s Association, 2008). Those affected often experience other comorbidities, disability and early death (Alzheimer’s Association, 2019; Arrighi et al., 2010). No cure currently exists for progressive dementias, and the associated healthcare costs exceed those of other age-related conditions (Alzheimer’s Association, 2015a). Recently, animal and human studies have begun reporting on air pollution neurotoxicity (J. L. Allen et al., 2017), including dementia (Carey et al., 2018; Chang et al., 2014; Chen, Kwong, Copes, Hystad, et al., 2017; Chen, Kwong, Copes, Tu, et al., 2017; Kilian & Kitazawa, 2018; Oudin et al., 2016; Power et al., 2016). Traffic-related air pollutants (TRAP) such as nitrogen dioxide (NO2), ultrafine particulates (UFP) and black carbon (BC) are important components of community air pollution that can vary substantially over space and time (Fujita et al., 2014; Liu et al., 2018). TRAP exposure has been shown to be associated with neurotoxicity (Block & Calderón-Garcidueñas, 2009; Calderón-Garcidueñas et al., 2002, 2003; Calderón-Garcidueñas, Mora-Tiscareño, et al., 2008; Calderón-Garcidueñas, Solt, et al., 2008) and pathologies such as Alzheimer’s Disease (AD) in animals (Yan et al., 2015, 2016) as well as cognitive deficits, including late-life dementia, though the evidence has been stronger for some pollutants than others (Carey et al., 2018; Chang et al., 2014; Chen, Kwong, Copes, Hystad, et al., 2017; Chen, Kwong, Copes, Tu, et al., 2017; Kilian & Kitazawa, 2018; Oudin et al., 2016; Power et al., 2011, 2016; X. Yu et al., 2020). In particular, research indicates that UFPs may play an important role in the adverse health effects associated with particulate matter (Brown et al., 2001; Donaldson, 2001; N. Li et al., 2003; Lundborg et al., 2001; Manigrasso et al., 2017; Oberdorster et al., 1994; Seaton et al., 1995; Stone et al., 2000). Still, epidemiologic studies investigating dementia and long-term TRAP exposure are limited due to the absence of models that appropriately capture long-term human exposure to TRAP (Rivera et al., 2012). This study addresses this gap in the literature through three specific aims: In Aim 1, we use fine-scale, long-term NO2 exposure as well as road proximity to assess the association between TRAP and late-life all-cause and AD dementia incidence in a community-based prospective cohort study. This study was conducted using the Adult Changes in Thought (ACT) cohort, a well-characterized, Seattle-based, prospective cohort study of aging and the brain among elderly individuals (65+ years) that has been ongoing since 1994 (Kukull et al., 2002; L. Wang et al., 2006). Participants were assigned long-term NO2 exposure based on a spatiotemporal model that incorporates decades of local air quality monitoring data based on residential history. Our primary analyses indicated that for every additional 5 ppb increase in 10-year average NO2 exposure, the hazard of all-cause and AD dementia is estimated to be 1% (HR: 1.01, 95% CI: 0.91, 1.11) and 2% (HR: 1.02, 95% CI: 0.91, 1.13) greater, respectively, after adjusting for important potential confounders. Sensitivity and secondary analyses investigating the impact of different exposure windows, model adjustments, exposure quality and more were in agreement, supporting the robustness of our results. These findings are in line with the literature and a recent meta-analysis indicating that there is no evidence of an association between NO2 and dementia incidence. In Aim 2, we leverage a highly innovative mobile monitoring campaign specifically designed to assess spatially-granular, long-term TRAP exposure for the ACT cohort (Blanco et al., 2019; Stanley, 2019) to characterize otherwise unavailable annual-average UFP and BC exposure. We calculate weighted UFP and BC averages from repeated short-term monitoring samples and use these to build universal kriging models with partial least squares regression to summarize hundreds of geographic covariate predictors. The hold-out model validation results indicated low model bias and high precision (RMSE: 933 pt UFP/cm3, 58 ng BC/m3; R2: 0.87 for UFP, 0.85 for BC). Predicted annual average UFP and BC exposure for ACT cohort locations had a median (IQR) of 6,782 (1,788) pt/cm3 and 525 (134) ng/m3, respectively. Similar to past studies, predicted concentration were highest near the downtown, industrial and airport areas as well as along major highways. Sensitivity analyses taking different approaches for dealing with extreme observations, calculating annual averages and building models all resulted in very similar results, strengthening the robustness of these exposure models. These findings support the use of these prediction models for future epidemiologic investigations of TRAP exposure in the ACT cohort. Aim 3 extends the exposure surfaces developed in Aim 2 for 2019 back to 1995 in order to characterize otherwise unavailable, spatially granular, long-term BC and UFP exposure for the ACT cohort. We use time-varying values of emission indicators (highway emissions) and surrogates (population density and green space; hereafter referred to jointly as “indicators”) known to be strongly associated with TRAP along with observations of air pollution trends over time to extrapolate model predictions back in time. We validate models against historical observations at air monitoring sites. Results from these models showed that annual average BC and UFP exposure estimates for the ACT cohort were generally higher and more variable for earlier years. Locations near Seattle and along major roadways saw the sharpest drops in BC levels, while locations near the Sea-Tac Airport saw the sharpest drops in UFP levels over time. Models captured overall spatial and temporal pollution trends, though they were conservative and underpredicted observed concentrations at AQS sites. These models provide an understanding of how these otherwise poorly characterized pollutants may have changed over time in the Puget Sound, an important gap in the field. Until now, investigations of TRAP exposure have been largely limited to short-term human exposure and animal studies despite the growing body of evidence linking some TRAPs to brain health. In one of the first truly long-term epidemiologic studies of TRAP exposure, we found no evidence that elevated levels of long-term NO2 exposure is associated with an increased risk of late-life dementia incidence. Furthermore, we are one of the first to build annual-average UFP and BC exposure models from a novel and extensive mobile monitoring campaign specifically designed to assess exposure in a long-standing, community-based, prospective cohort study of aging and the brain. These models can be used to further advance the field and support epidemiologic investigations of dementia incidence and long-term TRAP exposure, including UFPs and pollutant mixtures.