AD prevalence was acquired from Medicare data (Figure 1A) and average nighttime light intensity was generated from satellite acquired data (Figure 1B) and the data were averaged for years 2012–2018. States were ranked according to average nighttime light intensity and were divided into five groups representing those with the lowest average light intensity (i.e., 1/5, darkest) to the highest average light intensity (i.e., 5/5, brightest) (Figure 1C). Analysis revealed a statistically significant difference in AD prevalence between groups [F(4, 43) = 13.500, p < 0.001]. Multiple comparisons testing found statistical differences between states with the darkest average nighttime light intensity and states with brightest average nighttime light intensity (Figure 1D). Pearson correlation analysis similarly demonstrated a relationship between light intensity and AD prevalence [r(46) = 0.557, p < 0.001, Figure 1E]. This relationship also existed when examining those above the age of 65 [ANOVA: F(4, 43) = 14.560, p < 0.001; Correlation: r(46) = 0.490, p < 0.001] and those under the age of 65 [ANOVA: F(4, 43) = 6.950, p < 0.001; Correlation: r(46) = 0.550, p < 0.001] (Supplementary Figure S1). Each year was also assessed individually, and the same positive relationship was observed between nighttime light intensity and AD prevalence for each year independently (Supplementary Table S1).

A linear mixed model was applied to examine the overall effect of light intensity on AD prevalence (from 2012 to 2018) taking account of within-state correlation due to repeated measures. When examining all individuals, the overall regression was statistically significant (Chi-square = 1081.9, df = 1, p < 0.0001). Average nighttime light intensity was significantly associated with AD prevalence (regression coefficient = 0.283, SE = 0.102, 95% CI = 0.082–0.485, p = 0.006, R² = 0.382, Table 1). This effect was also observed in those over the age of 65, under the age of 65, in both men and women, and in all races examined except Asian Pacific Islander (Table 1). It was noted that stronger regression coefficients were associated with certain ethnic groups compared to others (e.g., Native Americans). The factors contributing to this observation are not clear but could reflect conditions in particular geographic areas and should be investigated further.

Co-variates known (or proposed to be) risk factors for AD including alcohol abuse, atrial fibrillation, chronic kidney disease, depression, diabetes, heart failure, hyperlipidemia, hypertension, obesity, and stroke were subsequently included in the model. Data for the covariates were obtained from Medicare Chronic Conditions data or from the CDC. When looking at all individuals, average nighttime light intensity was associated with AD prevalence even when accounting for alcohol abuse, chronic kidney disease, depression, heart failure, and obesity (Table 2). These data suggest that nighttime light intensity has a stronger influence on AD prevalence than these conditions. However, other covariates were more strongly associated with AD than light intensity including atrial fibrillation, diabetes, hyperlipidemia, hypertension, and stroke (Table 2) indicating that nighttime light exposure had a more subtle effect than these disease covariates. Similar results were obtained when examining men, women, and those over the age of 65 with each having a unique profile (Table 2). However, for those under the age of 65 average light intensity was associated with AD prevalence even when considering all the covariates (estimate = 0.174, SE = 0.064, p = 0.007) (Table 2) suggesting that those under the age of 65 may be particularly sensitive to the effects of exposure to light at night.

Each state is heterogeneous including urban, suburban, and rural areas each with different exposures (e.g., dietary habits/patterns, activity level, air pollution, light pollution). Therefore, we examined the relationship between average nighttime light intensity and AD prevalence in counties since these are smaller in size and more homogeneous than the state level data. To do so, the largest city in each state was identified and the county in which this city resided was noted and average nighttime light intensity was determined for that county and compared to AD prevalence from Medicare Chronic Conditions data at the county level. Using this approach, data was retrieved for 45 counties and the District of Columbia. The data were subsequently clustered based on average nighttime light intensity representing those counties with the lowest average nighttime light intensity (i.e., 1/4, darkest) to the highest average nighttime light intensity (i.e., 4/4, brightest) (Figure 2A). Analysis revealed a statistically significant difference in AD prevalence between groups [F(3, 42) = 10.750, p < 0.001]. Multiple comparisons testing found that states with the darkest average nighttime light intensity were statistically different from states with brightest average nighttime light intensity (Figure 2B). Pearson correlation analysis similarly demonstrated a relationship between light intensity and AD prevalence [r(44) = 0.599, p < 0.001, Figure 2C]. This relationship also existed when examining those over the age of 65 [ANOVA: F(3, 42) = 10.840, p < 0.001; Correlation: r(44) = 0.612, p < 0.001] and those under the age of 65 [ANOVA: F(3, 42) = 8.424, p < 0.001; Correlation: r(44) = 0.514, p < 0.001] (Supplementary Figure S2).

A linear mixed effects model was applied to examine the overall effect of average light intensity on AD prevalence (from 2012 to 2018) considering within-county correlation due to repeated measures. When examining all individuals, average nighttime light intensity was significantly association with AD prevalence (regression coefficient = 0.040, SE = 0.003, 95% CI = 0.034–0.047, p < 0.001, R² = 0.304) including in those over the age of 65 (regression coefficient = 0.052, SE = 0.004, 95% CI = 0.043–0.060, p < 0.001, R² = 0.304) and under the age of 65 (regression coefficient = 0.019, SE = 0.002, 95% CI = 0.015–0.023, p < 0.001, R² = 0.480). In summary, the results observed on the state level are recapitulated on the county level bolstering support for the positive association between AD prevalence and nighttime light exposure.

The analyses reveal that greater average nighttime light intensity (i.e., light pollution) was associated with higher AD prevalence. This was true for 2012–2018 average and each year examined individually, and in those over and under the age of 65 (i.e., 65+, <65), in both sexes, and in each race (except Asian Pacific Island which may be related to power). This finding was observed when examining data on the state level as well as on the county level.

Average nighttime light intensity was more strongly associated with AD prevalence than some diseases and conditions reported or suspected to be risk factors for AD including alcohol abuse (Anstey et al., 2009; Andrews et al., 2020; Nallapu et al., 2023; Wang et al., 2023), chronic kidney disease (Zhang et al., 2020; Tang et al., 2022), depression (Saiz-Vazquez et al., 2021), heart failure (Qiu et al., 2006; Cermakova et al., 2015), and obesity (Alford et al., 2018; Al-Kuraishy et al., 2023). However, other covariates were more strongly associated with AD than average nighttime light intensity including atrial fibrillation (Giannone et al., 2022), diabetes, hyperlipidemia, hypertension, and stoke. This finding is perhaps not surprising as the effects of nighttime light exposure would be expected to be more subtle than factors robustly associated with AD. However, in individuals under the age of 65, nighttime light intensity was associated with AD prevalence to a greater degree than all other disease risk factors, suggesting that younger individuals may be particularly sensitive to the detrimental effects of light at night. While it is challenging to speculate about why those under the age of 65 would be particularly vulnerable to the effects of nighttime light exposure, literature demonstrates that there are individual differences in light sensitivity (Chellappa, 2021). Indeed, APOE genotype—a factor influencing early onset AD—impacts response to biological stressors (e.g., oxidative stress) and this could account for the increased vulnerability to the effects of nighttime light exposure (Jofre-Monseny et al., 2008; Dose et al., 2016). However, more research on how light at night may influence AD is needed.

Dementia is not a modern phenomenon. Dementia is mentioned by Pythagoras, Hippocrates, Plato, and Shakespeare (Yang et al., 2016) and yet since the development of diagnostic criteria and AD incidence and prevalence have been reported, rates of AD have risen dramatically. This increase is attributed to factors like the aging population, emergence of lifestyle related diseases like obesity, high blood pressure, and diabetes, and a myriad of environmental factors related to urbanization and the industrial revolution. One such example is air pollution which is associated with cognitive decline (Franz et al., 2023), dementia (Gong et al., 2023), faster cognitive decline in those with AD (Lee et al., 2023), and incident neurodegenerative disease (Zhu et al., 2023). Exposure to light at night could similarly be impacting the development or progression of dementia and neurodegenerative disease.

To illustrate this possibility, we can examine Amish communities (who intentionally limit their use of modern technology and electricity and may consequently be exposed to less light at night). Amish communities have a lower prevalence of cognitive impairment in individuals of advanced age compared to the general population (Pericak-Vance et al., 1996; Johnson et al., 1997; Holder and Warren, 1998). Numerous factors other than light exposure may contribute to this phenomenon (e.g., emphasis on family/community, traditional farming, genetics), but that being said, exposure of Drosophila to dim light during the dark period promotes neurodegeneration (Kim et al., 2018), and data from mice demonstrate that dim light during the dark period alters neuronal dendrites in AD-relevant brain regions (cortex, hippocampus) (Delorme et al., 2022), and long-term exposure to outdoor light at night is associated with increased risk of cognitive impairment in humans (Chen et al., 2022), and there is a positive correlation between light pollution and Parkinson’s disease (another neurodegenerative disease) (Romeo et al., 2013). Additional studies carefully evaluating indoor and outdoor light exposure and mechanistic evaluations are needed to fully understand the impact of nighttime light exposure and light pollution on AD.

The analyses in this report encompass years 2012 through 2018 during which time there was a trend for states to have a decrease in nighttime light (2012 = 100,000, 2018 = 96.890, p = 0.054). The pandemic further decreased nighttime light exposure which was at the lowest level in 2020; however, despite the fact that 17 states have legislation designed to reduce light pollution nighttime light was the highest in 2022 (latest data available) (2012 = 100,000, 2022 = 103,500, p = 0.007) (Supplementary Figure S3). Given the association between nighttime light exposure and AD prevalence the increased nighttime light could potentially impact AD incidence and prevalence.

There are multiple mechanisms by which exposure to light at night may influence AD development or progression. First, there is a plethora of literature demonstrating that exposure to light at night is associated with sleep disruption. Sleep disruption can activate microglia and astrocytes, promote inflammation, negatively alter the clearance of amyloid beta (potentially via the glymphatic system), is associated with loss of neurons in the cortex, hippocampus, and locus coeruleus, and hippocampal atrophy (Kang et al., 2009; Xia et al., 2017; Atrooz et al., 2019; Reddy and van der Werf, 2020; Xiao et al., 2020; Owen et al., 2021; Zamore and Veasey, 2022). A meta-analysis extrapolating from 27 observational studies reports that individuals with sleep problems have a higher risk of cognitive impairment and AD, compared to individuals without sleep problems (Bubu et al., 2017). These observations share many features that are common with AD. Second, exposure to light at night may disrupt circadian rhythms (Raap et al., 2015) and sleep is one output of circadian rhythms perhaps suggesting an upstream disruption in circadian rhythms. Circadian rhythm disruption is widely reported by our group and others to have detrimental effects on health. Changes in circadian rhythms often precede symptoms of AD in humans (Colwell, 2021) and it is interesting to consider that exposure to light at night (via circadian disruption) may contribute to AD pathogenesis (Musiek, 2015; Musiek, 2017). Additionally, it is noteworthy that disruption of circadian rhythms is associated with increased risk of diseases that are risk factors for AD including obesity, diabetes, and depression, just to name a few (Fishbein et al., 2021). Lastly, on the biochemical level, exposure of mice to dim light increases the production of the pro-inflammatory cytokine interleukin 1β (IL-1β) (Walker et al., 2020), which is a feature associated with AD and exposure of mice to dim light during the dark period decreases levels of the neurotrophic factor brain derived neurotrophic factor (BDNF) in a brain region relevant for AD (i.e., the hippocampus) (Walker et al., 2020). This finding is intriguing as our group has demonstrated that low levels of BDNF are a feature that precede cognitive impairment (Voigt et al., 2021). Taken together there are numerous mechanisms by which exposure to light at night can impact AD, the relative contributions of these mechanisms will require additional investigation.

There are limitations associated with this study related to the curation of AD prevalence and light exposure data. First, the Medicare data are limited to those enrolled in Medicare Part A and Part B. Those with Medicare Advantage, only Part A, or only Part B are excluded; thus, the data presented in this report are not comprehensive (although Medicare data are reported to be reasonably reliable) (Grodstein et al., 2022). Second, the Medicare data report current residences of individuals which may not reflect life-long residence (important to understand causality). Third, this study evaluated AD prevalence and not incidence. Fourth, this study used state and county average light intensity data from satellites. No indoor light data were available although indoor light exposure (e.g., televisions, computers, phones) is critically important and should be evaluated in future studies. However, living in an area with more intense outdoor light at night is associated with shorter sleep duration, increased daytime sleepiness, and dissatisfaction with sleep quality suggesting that outdoor light can meaningfully impact biology providing validity to the notion that outdoor light exposure is important (Ohayon and Milesi, 2016; Xiao et al., 2020). However, the totality of outdoor and indoor nighttime light exposure is important to consider to fully understand the impact of nighttime light on AD.

De-identified data were obtained from Medicare reports of disease prevalence. All available data were included inclusive of both sexes.

Nighttime light exposure data was generated by the www.lightpollutionmap.info platform (Jurij Stare)³ which maps radiance data from NASA’s VIIRS/NPP Lunar BRDF-Adjusted Nighttime Lights Yearly composites (AllAngle_Composite_Snow_Free).

The Day/Night Band (DNB) sensor of the Visible Infrared Imaging Radiometer Suite (VIRRS), on board the Suomi-National Polar-orbiting Partnership (S-NPP) and Joint Polar Satellite System (JPSS) satellite platforms, provide daily measurements of nocturnal visible and near-infrared (NIR) light. NASA developed a suite of products of nighttime lights (NTL) applications including NASA’s Black Marble product suite (VNP46/VJ146). NASA’s Black Marble nighttime lights product, at 15 arc-second spatial resolution, is available from January 2012-present with data from the VIIRS DNB sensor. The VNP46/VJ146 product suite includes the daily at-sensor top of atmosphere (TOA) nighttime lights (NTL) product (VNP46A1/VJ146A1), daily moonlight-adjusted nighttime lights product (VNP46A2/VJ146A2), monthly moonlight-adjusted nighttime lights product (VNP46A3/VJ146A3), and yearly moonlight-adjusted nighttime lights product (VNP46A4/VJ146A4). To remove residual background noise, NASA’s VIIRS/NPP Lunar BRDF-Adjusted NTL composite values with radiances less than 0.5 nW.cm-2.sr-1 are set to zero. The retrieval algorithm, developed and implemented for routine global processing at NASA’s Land Science Investigator-led Processing System (SIPS), utilizes all high-quality, cloud-free, atmospheric-, terrain-, vegetation-, snow-, lunar-, and stray light-corrected radiance to estimate daily nighttime lights and other intrinsic surface optical properties. Details about algorithms, operational processing, evaluation, and validation are found in the user guide (Roman et al., 2021). Data for each year include from January 1st through December 31.

VIRRS data were overlaid on a map and areas of interest (i.e., states, counties) were outlined in triplicate (three independent measurements) and values were averaged to develop the average nighttime light intensity for each area of interest. Radiance data are provided as 10⁻⁹ W/cm^{2 *} sr. World Atlas images were developed using Falchi et al. (2016, 2017).

Average nighttime light intensity data were generated for each state (excluding Alaska and Hawaii) or county and data were analyzed as an average of 2012–2018 as well as each year independently for each state. States were divided into five groups and county data divided into four groups according to average nighttime light intensity and an ANOVA was conducted followed by a post hoc Tukey for pairwise comparisons (GraphPad Prism 10.0.2) to compare AD prevalence among the groups. Correlation analysis was also conducted with state and county data to evaluate the relationship between AD prevalence and average nighttime light exposure (GraphPad Prism 10.0.2).

A linear mixed model was then applied to examine the relationship with all data together taking account of within-state or within-county correlation due to repeated measures. The state and county assessments included age, sex, and race in the model. For the state data, biological covariates (atrial fibrillation, chronic kidney disease, depression, diabetes, heart failure, hyperlipidemia, hypertension, obesity, and stroke) were considered and subgroup analyses were performed including: age (65+, <65), sex (men, women), and race (Asian Pacific Islander, Hispanic, Native American, Non-Hispanic Black, Non-Hispanic White). Analyses were conducted using SAS (v 9.4).