Associations between systemic inflammation, neighborhood tree canopy, and white matter hyperintensities: The healthy brain initiative

Abstract

Background

Systemic inflammation is associated with greater white matter hyperintensities (WMH), increasing risk of Alzheimer's disease. Neighborhood greenspaces (e.g., parks) have been associated with reduced inflammation and dementia risk and fewer WMH.

Objective

To examine whether systemic inflammation and neighborhood % tree canopy are independently associated with and interact to influence WMH.

Methods

We used cross-sectional data on 161 participants with magnetic resonance imaging (MRI) in University of Miami's Healthy Brain Initiative who were ≥ 50-years old with no to mild cognitive impairment. U.S. Forest Service tree canopy data (2021) were used to derive tree canopy percentage per Census tract. High-sensitivity C-reactive protein (hs-CRP), a marker of systemic inflammation, was categorized as low/moderate (0–3 mg/L) and high (>3 mg/L). Total WMH volume was derived from structural MRI. Multivariable linear regression examined independent associations of hs-CRP and WMH and % tree canopy and WMH, and the association between a four-category hs-CRP-tree canopy variable (based on median cut points) and WMH.

Results

Participants were 68.7 ± 9.3 years old and 70% were female. Higher hs-CRP was associated with greater WMH, and greater neighborhood % tree canopy was associated with fewer WMH. Compared to those with higher hs-CRP and lower % tree canopy, those with higher hs-CRP and greater % tree canopy had fewer WMH.

Conclusions

Higher systemic inflammation and greater neighborhood % tree canopy were independently associated with WMH, and our findings suggest that they interact to influence WMH (e.g., in those with higher inflammation, having greater neighborhood tree canopy associated with lower WMH).

Introduction

Alzheimer's disease (AD) is the seventh leading cause of death globally, with approximately 10 million new cases each year and over 7 million Americans living with AD.^1,2 Although pharmacological treatments are available to slow progression and reduce symptom burden, there is currently no cure for AD. Chronic, systemic inflammation (i.e., affecting the entire body) and vascular inflammation (i.e., affecting the blood vessels) increase the risk for dementia and cognitive decline, as well as other age-related diseases in older adults. ³ In addition, prolonged or chronic inflammation has been associated with the production of amyloid-β (Aβ) plaques, one of the pathological hallmarks of AD,^4–6 and thus, inflammation is thought to be a significant contributor to the development of AD. Vascular inflammation and vascular risk factors such as hypertension (high blood pressure) and diabetes mellitus have been associated with the development of Aβ plaques, ⁷ and is considered to have a bidirectional relationship with inflammation contributing to the accumulation of Aβ plaques and Aβ plaques contributing to further inflammatory damage.^8–10 Chronic and sustained inflammation, both neuroinflammation and systemic inflammation, have been implicated in the pathogenesis and progression of AD. ⁶

Inflammation that increases over time or is persistently elevated has been associated with lower white matter volume. ¹¹ White matter hyperintensity (WMH) volume is a measure of white matter damage (i.e., demyelination and axonal degeneration), ascertained via magnetic resonance imaging (MRI), that has been associated with greater risk of cognitive impairment and AD dementia, independently of vascular risk factors.^12,13 Blood-based markers of systemic inflammation (e.g., high-sensitivity C-reactive protein (hs-CRP)) and vascular inflammation (e.g., lipoprotein A (Lp(a)) have been associated with greater WMH volume in several studies.^14–17 Individuals with AD have been shown to have elevated levels of systemic, or peripheral, pro-inflammatory markers such as hs-CRP, IL-6, IL-2, etc. ¹⁸ Thus, inflammation biomarkers have been associated with WMH, a risk factor for AD, as well as greater risk of AD pathology and diagnosis.

Beyond these known risk factors for AD, inflammation and WMH, several protective factors have been identified that improve brain health and reduce AD risk. The 2024 Lancet Commission on Dementia Prevention, Intervention, and Care noted that up to 45% of AD cases are attributed to modifiable risk factors including lifestyle behaviors such as physical inactivity, social isolation, and depression. ¹⁹ Addressing these modifiable risk factors, through both individual-level and community-level interventions, can help delay the onset of AD and/or prevent its clinical symptoms. Indeed, studies have shown that better mental health, ²⁰ greater physical activity, ²¹ and more social engagement, ²² as well as interventions to promote these, are associated with lower AD risk and slower cognitive decline. It has been reported that some older individuals, despite having high levels of inflammatory markers and AD neuropathology identified post-mortem (e.g., Aβ plaques) exhibit resistance to overt signs and symptoms (i.e., participants did not display cognitive decline) of AD and related dementias. ²³ The possible mechanisms by which this could occur include genetics, ²⁴ anti-inflammatory physiologic processes, ²⁵ and lifestyle or environmental factors including cognitive reserve. ²⁶ Thus, there are likely protective processes or mechanisms that promote resilience and resistance to dementia.

Evidence suggests that health-promoting community characteristics, including greenspaces (e.g., parks, tree cover), are associated with increases in healthy behaviors and lower risk for AD.^27–29 Thus, community-level interventions to increase greenspace exposure may be a useful target in risk reduction and AD prevention. The mechanisms by which greenspace is associated with a reduction in AD risk are likely many, as greenspace has been shown to affect health in many ways. For instance, greenspace has been associated with increased physical activity, ³⁰ increased social engagement and cohesion, ³¹ and improved mental health. ³² Additionally, greater greenspace exposure has been associated with fewer WMH; a study of older adults in the Cardiovascular Health Study found that individuals living in lower-income neighborhoods with less greenspace had greater WMH worsening over 5 years compared to those in higher-income neighborhoods with more greenspace. ³³ Overall, a significant and growing number of studies demonstrate that individuals living in greener neighborhoods and who spend more time in greenspaces have better brain health outcomes, including reduced WMH and AD risk.^33–36

Although there have been studies that have examined associations between systemic inflammatory markers and WMH, as well as between greenspace and WMH, to our knowledge no study has looked at whether systemic inflammation and greenspace interact to influence WMH. Based on the scientific gaps in the literature, this study's aims are twofold. First, we aimed to determine whether systemic inflammation and neighborhood percentage of tree canopy are each independently associated with WMH. We hypothesized that greater systemic inflammation will be associated with greater WMH, and greater neighborhood tree canopy will be associated with fewer WMH. Second, we aimed to determine whether systemic inflammation and neighborhood percentage of tree canopy interact to influence WMH. We hypothesized that among individuals with greater inflammation, those living in neighborhoods with more tree canopy (versus less) will have fewer WMH. Our study concentrates on hs-CRP, a systemic inflammatory marker, as it is a widely accepted and validated biomarker recognized by the American College of Cardiology for its superior clinical utility and predictive value in assessing cardiovascular disease risk. ³⁷ Additionally, hs-CRP is found to be associated with greater WMH volume, ¹⁶ a marker of increased risk of cerebrovascular disease and dementia.^13,38 We investigate neighborhood tree canopy because it can be targeted via small-scale community interventions such as tree planting initiatives. Lastly, we focus on WMH because it is an early marker of biological brain changes that is associated with increased AD risk. Determining whether community-level greenspace exposure buffers against neurological damage, such as WMH, could aid in future intervention and policy development and further the mission of prevention and brain health promotion of AD and related dementias.

Methods

Sample

This cross-sectional, observational study includes a total of 161 participants living in the South Florida region, primarily Palm Beach and Broward counties. The study is a secondary analysis of the data collected as part of the Healthy Brain Initiative (HBI), established by University of Miami's Comprehensive Center for Brain Health (CCBH) in Boca Raton, Florida. HBI began recruiting participants on March 24, 2022. At the annual follow-up visits, demographics and medical history are collected, physical and neurological assessments are conducted, and blood-based biomarkers such as high-sensitivity C-reactive protein (hs-CRP) are measured. The HBI cohort receives brain MRI scans at baseline and every 2 years after the baseline visit. This study includes adults ≥ 50 years old who are medically eligible to undergo an MRI scan and have no cognitive impairment to mild cognitive impairment. Participants who have missing hs-CRP values and/or have incomplete MRI data were excluded from the study. The University of Miami Institutional Review Board approved the HBI study, and participants provided informed consent.

Neighborhood greenspace

Environmental greenspace exposure was derived from maps and satellite imagery (i.e., neighborhood % tree canopy). Tree canopy cover is defined as the amount of area of leaves, needles, branches, and stems covering the ground as seen from an aerial view. Tree canopy maps indicate a percentage of tree cover from 0–100% for each 30 × 30 meter cell, represented by a single pixel. Data from these maps were derived from the United States Department of Agriculture (USDA) Forest Service (USFS) Geospatial Technology and Applications Center (GTAC) 2021 Tree Canopy Cover product suite. Census tracts are the geographic regions representing the smallest statistical area encompassing available demographic data. Florida Census tract shapefiles were obtained from the U.S. Census TIGER/Line Geodatabase website for the purpose of defining the boundaries measuring mean tree canopy percentage within HBI participants’ neighborhoods. Figure 1 presents the mean neighborhood % tree canopy values for the geographical location of the participants.

Figure 1.

Mean neighborhood % tree canopy values varying by geographical location of the participant.

Biomarkers of inflammation

Hs-CRP, a marker of systemic inflammation, is typically categorized based on the relative risk of cardiovascular disease in categories of low: 0 to <1 mg/L, moderate: 1–3 mg/L, high: >3 mg/L, as recommended by the American Heart Association and the Centers for Disease Control and Prevention. ³⁹ For our primary analysis, participants were stratified into two groups: low/moderate hs-CRP and high hs-CRP to compare the clinically relevant high hs-CRP to lower levels of hs-CRP. Compared to other systemic inflammatory biomarkers, hs-CRP has been shown to be associated with WMH as well as WMH progression. ⁴⁰ Hs-CRP was derived via blood samples from that were sent to Cleveland Heart Lab for laboratory analysis. While hs-CRP was the primary inflammation biomarker of interest, we also included lipoprotein A (Lp(a) in exploratory analyses. Although there is no generalized consensus on Lp(a) clinical stratification, according to the American College of Cardiology/American Heart Association (ACC/AHA) guidelines, individuals with a level of ≥125 nmol/L are considered at greater risk for atherosclerotic cardiovascular disease including conditions such as heart attack as well as structural heart disease.^41,42

Combined hs-CRP-tree canopy exposure

We created a combined hs-CRP-tree canopy exposure variable to assess whether systemic inflammation and neighborhood % tree canopy interact to influence WMH volume. The CRP-tree canopy variable contained four permutations of high and low values for both CRP and % tree canopy, where percent tree canopy was dichotomized based on the median cut point of 10.2% (low ≤10.2%; high >10.2%) to maximize the sample size in each category (no clinically significant cut point has been established for tree canopy). The four categories included high hs-CRP with low % tree canopy, high hs-CRP with high % tree canopy, low/moderate hs-CRP with low % tree canopy, and low/moderate hs-CRP with high % tree canopy. In our primary analyses, the reference group was defined as those with high hs-CRP and low % tree canopy, as we hypothesize that this group would have the greatest WMH volume and the remaining categories would be associated with reduced WMH comparatively. We did not create a similar combined variable with Lp(a) and WMH as the exploratory analyses revealed it was not a significant independent predictor for WMH.

WMH

Non-contrast MRI was performed at the Sperling Medical Group in Delray Beach, FL on all participants at their baseline study visits. All participants have brain MRI scans using a GE 3 Tesla 750 W scanner at baseline and every 2 years using a standardized protocol, ⁴³ with MRI sequences that are modified from the ADNI-3 canonical protocol. ⁴⁴ Post-processing of all MRI images is performed according to ADNI procedures, including correction for geometric distortion of gradient nonlinearity, field inhomogeneity correction, histogram-peak sharpening, and correction for site-specific scanner features/field strength. Cortical, subcortical, ventricular, and WMH volumes are analyzed using Combinostics^® cNeuro Suite, an FDA-cleared quantitative pipeline.^45,46 Combinostics cMRI suite is comparable to manual segmentation with Pearson coefficients ranging from 0.84 to 0.99. The WMH percentage was obtained by dividing WMH volume (mL) by intracranial volume (mL), which was normalized in this way to account for head size differences, and log transformed prior to entry into the regression models due to skew.

Covariates

Baseline demographics (as shown in Table 1) included age, sex, self-identified racial group (White, Black, African American, and other which consisted of Asian and those who identified as two or more races), Hispanic ethnicity, completed years of education, and living situation (living alone or with others (e.g., spouse/ partner, child, relatives, friends). A comorbidity index score was constructed based on self-reported conditions including diabetes, stroke, heart disease, hypertension, and hypercholesteremia (one point for each; score range: 0–5). Additional covariates included obesity (body mass index (BMI) ≥ 30 kg/m²), frailty (i.e., CHSA Clinical Frailty Scale ⁴⁷ ) and diagnosis of cognitively normal (versus cognitive impairment, CDR>0) determined at the HBI visit (yes versus no). The prevalence of self-reported arthritis and other autoimmune diseases are reported to describe the sample as they are conditions that can cause systemic inflammation (e.g., high hs-CRP), but because they can be upstream causes of elevated hs-CRP, they are not included as covariates in the regression analyses. The Area Deprivation Index (ADI), obtained from the University of Wisconsin's Neighborhood Atlas, was used to quantify neighborhood-level socioeconomic disadvantage based on participants’ residential addresses, incorporating variables such as income, education, employment, and housing quality. Higher ADI scores indicate greater socioeconomic disadvantage (ranging from 1 to 100). Population density (Census tract level) was defined as people per square mile.

Table 1.

Study sample characteristics.

Characteristica	Total sample	Low/moderate hs-CRP (≤3 mg/L)	High hs-CRP (>3 mg/L)	pf
Sample, N	161	131	30	-
Age, mean (SD)	68.7 (9.3)	69.0 (9.0)	66.0 (10.3)	0.09
Sex, n (%)
Female	112 (69.6%)	93 (71%)	19 (63.3%)	0.41
Male	49 (30.4%)	38 (29%)	11 (36.7%)	Ref.
Race, n (%)
White	142 (88.8%)	121 (93.1%)	21 (70.0%)	Ref.
Black/African American	13 (8.1%)	6 (4.6%)	7 (23.3%)	0.09
Other	5 (3.1%)	3 (2.3%)	2 (6.7%)	0.68
Hispanic, n (%)	22 (13.7%)	16 (12.2%)	6 (20.0%)	0.27
Education (years), mean (SD)	16.3 (2.8)	16.2 (2.9)	16.6 (2.5)	0.49
Living Situation, n (%)
Lives alone	47 (29.2)	42 (32.1%)	5 (16.7%)	0.10
Otherc	114 (70.8)	89 (67.9%)	25 (83.3%)	Ref.
Comorbidities, n (%)
Diabetes	14 (8.7%)	9 (6.9%)	5 (16.7%)	0.10
Stroke	6 (3.7%)	5 (3.8%)	1 (3.3%)	0.89
Heart Disease	18 (11.3%)	16 (12.3%)	2 (6.7%)	0.39
Hypertension	56 (35%)	41 (31.5%)	15 (50%)	0.06
Hypercholesterolemia	85 (53.1%)	69 (53.1%)	16 (53.3%)	0.98
BMI ≥ 30 kg/m2 (obesity)	43 (26.9%)	26 (19.8%)	17 (58.6%)	<0.0001
Arthritis	63 (39.1%)	50 (38.2%)	13 (43.3%)	0.67
Autoimmune disease	20 (12.4%)	16 (12.2%)	4 (13.3%)	0.89
Cognitive impairmentb	49 (30.4%)	37 (28.2%)	12 (40.0%)	0.24
CHSA clinical frailty scale, mean (SD)d	2.9 (0.89)	2.8 (0.9)	3.2 (0.8)	0.02
Area deprivation index, mean (SD)	34.8 (26.2)	34.9 (26.5)	34.7 (25.1)	0.97
WMH (%)e, mean (SD)	0.49 (0.82)	0.48 (0.82)	0.50 (0.81)	0.95
Lp(a) (nmol/L), mean (SD)	67.9 (90.9)	67.3 (89.0)	70.9 (100.1)	0.84

hs-CRP: high sensitivity C reactive protein; BMI: body mass index; WMH: white matter hyperintensities; CHSA: Canadian Study of Health and Aging; Lp(a): lipoprotein A; ^a Missing data: racial group, n = 1; living situation, n = 1; diabetes, n = 1, stroke, n = 1, heart disease, n = 1, hypertension, n = 1; CHSA clinical frailty scale, n = 1; BMI, n = 1; area deprivation index, n = 3; dementia status, n = 3. ^b mild cognitive impairment or dementia diagnosis; ^c includes those who live with spouse/ partner, child, relatives, friends, or other; ^d Scored as 1 = Very fit; 2 = Well, without active disease; 3 = Well, with treated comorbid disease; 4 = Apparently vulnerable; 5 = Mildly frail; 6 = Moderately frail; 7 = Severely frail. ^e Displayed as a WMH volume mL/intracranial volume in mL; ^f p-value from unadjusted logistic regression.

Analyses

We describe characteristics of the overall sample as well as when stratified by hs-CRP (low/moderate hs-CRP and high hs-CRP). Multivariable linear regression models, which used generalized estimating equations accounting for clustering by US Census tract, tested the independent associations between the dichotomized hs-CRP measure and WMH and between the continuous percent tree canopy measure and WMH (Aim 1). Multivariable models controlled for variables that were determined a priori to be potential confounders based on prior associations with the exposures and outcome, including age, sex, education, race, ethnicity, living alone, ADI, population density, and normal cognition diagnosis. A similar multivariable linear regression analysis was conducted to investigate the association between the combined CRP-tree canopy variable and WMH (Aim 2). Although our primary analyses focused on a dichotomized measure of hs-CRP based on an established, clinically meaningful cut point (>3 mg/L), in sensitivity analyses, we ran the fully adjusted Aim 1 model after replacing the dichotomous hs-CRP measure with a continuous measure. As part of our sensitivity analyses, we reran this updated model after adding a multiplicative interaction term between the continuous hs-CRP and % tree canopy measures (i.e., hs-CRP×% tree canopy) to detect potential effect modification. Statistical analyses were conducted in SAS v9.4 and statistical significance was determined based on p < 0.05.

Results

Participants (n = 161) were 68.7 years old on average (standard deviation [SD] = 9.3) and had 16.3 years of education (SD = 2.8) (Table 1). Seventy percent were women; 8% identified as Black, 89% White, as 3% other racial group; and 14% identified as Hispanic. The mean WMH percentage was 0.49% (SD = 0.89%). Participants had a mean hs-CRP value of 2.08 mg/L (SD = 2.24), with approximately 19% having high hs-CRP (>3 mg/L). Mean Lp(a) was 67.9 nmol/L (SD = 90.9) and hs-CRP and Lp(a) values were not correlated (r = 0.11). On average, the sample lived in neighborhoods with higher SES (mean ADI=34.8, SD = 26.2). In women compared to men, 83% versus 78% had low/moderate hs-CRP and 17% versus 22% had high hs-CRP (Supplemental Table 1). The majority (85%) of those identifying as White had low/moderate hs-CRP, while a majority (53.8%) of those identifying as Black had high hs-CRP (Supplemental Table 2).

Few significant differences were found between the low/moderate hs-CRP and high-hs-CRP group in terms of comorbidities (Table 1). A greater percentage of participants (59%) in the high hs-CRP group had a BMI ≥ 30 kg/m² (i.e., obese) compared to the low/moderate hs-CRP (20%). In addition, individuals with high hs-CRP scored higher (frailer) on the CHSA clinical frailty scale (mean=3.2, SD = 0.8) compared to those with low/moderate hs-CRP (mean=2.8, SD = 0.9).

Table 2 outlines both the objectively measured and self-reported neighborhood characteristics and participants’ neighborhood-based behaviors. Mean neighborhood % tree canopy coverage was 10.5% for the overall sample (SD = 4.8). Almost half (49.6%) of the total sample reported spending ≥2.5 h walking in their neighborhood per week, and almost half (45%) agreed (somewhat agree to strongly agree) that they spent a lot of time in their neighborhood greenspace relaxing, exercising and/or socializing. When asked whether their neighborhood has a lot of greenspaces (grass, shrubs, trees, and other vegetation), participants responded with a mean score of 5.6 (SD = 1.8) on a 7-point Likert scale (1 = strongly disagree to 7 = strongly agree). When asked to rate whether their neighborhood has many shade trees, participants reported a mean score of 5.0 (SD = 1.8), also on the 7-point Likert scale.

Table 2.

Objective neighborhood characteristics and self-reported neighborhood behaviors.

Characteristica	Total sample	Low/moderate hs-CRP (≤3 mg/L)	High hs-CRP (>3 mg/L)	pd
Neighborhood % tree canopy, mean (SD)	10.5 (4.8)	10.6 (4.9)	10.0 (4.2)	0.56
Population density (people/mi2), mean (SD)	4728 (2994)	4831 (3116)	4268 (2359)	0.37
Time spent walking in neighborhood per week, n (%)
Up to 30 min	18 (13.5%)	4 (17.4%)	14 (12.7%)	0.56
At least 30 min but less than 1.5 h	25 (18.8%)	2 (8.7%)	23 (20.9%)
At least 1.5 h but less than 2.5 h	24 (18.1%)	5 (21.7%)	19 (17.3%)
2 and ½ hours or more	66 (49.6%)	12 (52.2%)	54 (49.1%)
Reported spending lot of time in neighborhood greenspaces (relaxing, exercising, and/or socializing), n (%) b,c	72 (45.0%)	60 (46.2%)	12 (40.0%)	0.54
Rating their neighborhood has a lot of green space (grass, shrubs, trees, and other vegetation), mean (SD)c	5.6 (1.8)	5.6 (1.8)	5.5 (1.9)	0.84
Rating their neighborhood has many shade trees, mean (SD)c	5.0 (1.8)	5.1 (1.9)	4.9 (1.8)	0.65

hs-CRP: high sensitivity C reactive protein.

^a Missing data: tree canopy, n = 7; population density, n = 7; time spent walking outdoors per week, n = 28; Spend a lot of time relaxing, exercising, and/or socializing in greenspace, n = 1; ^b Calculated from 3 questions on whether the participant spends a lot of time i) relaxing, ii) exercising, or iii) socializing in parks, gardens, and other greenspaces in the neighborhood (yes = answered somewhat agree to strongly agree for at least one of these questions); ^c 7-point scale from 1 = strongly disagree to 7 = strongly agree; ^d p-value from unadjusted logistic regression.

Table 3 provides multivariable regression findings for the analyses examining hs-CRP and % tree canopy independently (Aim 1). In the unadjusted model including hs-CRP and % tree canopy as independent predictors, high hs-CRP (versus low/moderate) was associated with greater WMH (estimate=1.50, 95% CI = 1.17, 1.82), and greater neighborhood % tree canopy was associated with fewer WMH (estimate=–0.24, 95% CI = -0.26, −0.22) (Table 3, Model 1). In the model that additionally controlled for all covariates, we observed attenuation in the estimates but the associations for hs-CRP (estimate=0.62, 95% CI = 0.12, 1.12) and % tree canopy remained significant (estimate=-0.12, 95% CI = -0.17, −0.08) (Table 3, Model 2). In exploratory analyses, Lp(a) was not independently associated with WMH in the fully-adjusted model (estimate=0.28, 95% CI = -0.08, 0.65; data not shown).

Table 3.

Associations between CRP, neighborhood tree canopy, and white matter hyperintensity volume.

	Characteristic	Model 1	Model 2
		Estimate (95% CI)	Estimate (95% CI)
Primary predictors	Hs-CRP (>3 mg/L versus ≤3 mg/L)	1.50 (1.17, 1.82)****	0.62 (0.12, 1.12)*
	Neighborhood % tree canopy	−0.24 (−0.26, −0.22)****	−0.12 (−0.17, −0.08)****
Covariates	Age (years)	-	0.06 (0.03, 0.09)****
	Women (versus men)	-	0.44 (0.01, 0.87)*
	Education (years)	-	−0.06 (−0.12, 0.01)
	Nonwhite (versus White)	-	−0.56 (−1.16, 0.03)
	Hispanic ethnicity (versus non-Hispanic)	-	0.05 (−0.78, 0.89)
	Living alone (versus with others)	-	0.15 (−0.32, 0.62)
	Area deprivation index (ADI)	-	0.01 (0.00, 0.02)**
	Population density (1000 people/mi2)	-	−0.10 (−0.17, −0.03)**
	Normal cognition diagnosis (yes versus no)	-	0.10 (−0.34, 0.54)

*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. hs-CRP: high sensitivity C reactive protein; CI: confidence interval.

Table 4 provides the multivariable regression findings for the combined CRP-tree canopy variable. Compared to those with high hs-CRP and low % tree canopy, fewer WMH were observed for those with high CRP and high % tree canopy (estimate=-1.16, 95% CI = -2.12, −0.20), those with low/moderate hs-CRP and low % tree canopy (estimate=-1.17, 95% CI = -1.86, −0.48), and those with low/moderate hs-CRP and high % tree canopy (estimate=-2.07, 95% CI = -2.83, −1.31).

Table 4.

Interaction between neighborhood tree canopy and hs-CRP and white matter hyperintensity volume.

Hs-CRP and neighborhood tree canopy combined valuesa	Estimate (95% CI)b
High hs-CRP and low % tree canopy	Ref.
High hs-CRP and high % tree canopy	−1.16 (−2.12, −0.20)*
Low/moderate hs-CRP and low % tree canopy	−1.17 (−1.86, −0.48)***
Low/moderate hs-CRP and high % tree canopy	−2.07 (−2.83, −1.31)****

hs-CRP: high sensitivity C reactive protein; Ref: reference/comparison group; CI: confidence interval.

^a High hs-CRP: >3 mg/L, low/moderate hs-CRP: ≤3 mg/L; low tree canopy: ≤10.2%; high tree canopy: >10.2%.

^b Controlling for age, sex, education, race, Hispanic ethnicity, area deprivation index, living alone, population density, and normal cognition diagnosis.

*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001.

Sensitivity analyses

In the fully adjusted Aim 1 model that replaced dichotomized hs-CRP with the continuous measure, hs-CRP was not associated with WMH (estimate: 0.06, 95% CI: −0.04, 0.16), while greater % tree canopy remained associated with fewer WMH (estimate: −0.14, 95% CI: −0.18, −0.10) (data not shown). In that model, an interaction was not detected after including the interaction term between the continuous hs-CRP and % tree canopy variables (p = 0.59). In sensitivity analyses, we also controlled for variables that could be either confounders or upstream causes of inflammation (i.e., comorbidity index score, obesity, and frailty) in all the multivariable regression models, including the analyses with the continuous hs-CRP measure. These analyses resulted in no substantive difference in findings and therefore data are not shown.

Discussion

In this study we shed light on the relationship of systemic inflammation (hs-CRP) and greenspace (neighborhood % tree canopy) with WMH volume. We found that both systemic inflammation and neighborhood % tree canopy were independently associated with WMH volume. In addition, we found that compared to those with higher hs-CRP and lower % tree canopy, those with higher hs-CRP and higher % tree canopy had fewer WMH. Overall, these findings suggest that living in neighborhoods with more tree canopy may help protect against WMH development, including amongst those with high systemic inflammation who are at greater risk for WMH.

Consistent with our findings, previous studies have found that systemic inflammation that increases over time or is persistently elevated is associated with increased WMH volume.^11,15 In our analysis, higher levels of CRP were associated with greater WMH volume, while no significant association was observed for Lp(a), suggesting that not all inflammatory biomarkers may contribute equally to white matter pathology. Future research should aim to clarify how systemic, vascular, and neuroinflammatory processes contribute to WMH development, further elaborating on a model of systemic-to-vascular-to-brain inflammation in relation to WMH. For example, blood-based markers of neuroinflammation such as glial fibrillary acidic protein (GFAP) could provide further insight. ⁴⁸ Data on GFAP are becoming available for this cohort, so future studies can investigate this marker further. Longitudinal studies incorporating these inflammatory markers may help elucidate how different inflammatory pathways contribute to white matter injury and cognitive decline both synergistically and independently.

Our findings corroborate those from two prior studies of older adults. The first found that a greater density of parks within a 20-min walk of home was associated with fewer WMH. ⁴⁹ The second found higher WMH volumes in individuals living in neighborhoods with lower % total greenspace combined with lower neighborhood median income (versus higher neighborhood greenspace and income). ³³ While the current study did not combine measures of neighborhood income and greenspace as done in the prior study, ³³ this should be investigated in future studies given that these variables can be highly correlated (e.g., findings for each independent variable are difficult to disentangle from the other without stratification). Overall, the extant literature suggest that multiple types of greenspaces may benefit brain health by reducing risk of white matter damage. However, provision of tree canopy represents a potential micro-scale community intervention via tree plantings that may be more feasible for most communities (versus overall greening or finding space for new parks).

Categories of systemic inflammation and neighborhood % tree canopy differentially influenced WMH volume. All three permutations of hs-CRP and tree canopy exposure were statistically significantly different from those with high hs-CRP and low % tree canopy. Notably, among those with high systemic inflammation, having more neighborhood tree canopy was associated with lower WMH volume. Higher neighborhood % tree canopy may be a protective factor (possibly conferring resistance) against WMH volume worsening. Those with higher hs-CRP may receive a greater benefit from the lifestyle factors (e.g., physical activity and social engagement) promoted when living in neighborhoods with more tree canopy. In other words, neighborhoods with more tree canopy may promote health in ways that buffer against further WMH worsening amongst those at greater risk. To our knowledge, no prior studies have examined the interaction of % tree canopy with systemic inflammation in relation to WMH volume. The implications of this, if replicated in other studies, is that greater exposure to neighborhood tree canopy may reduce WMH among those who are already at greater risk for WMH worsening (e.g., with high systemic inflammation), suggesting avenues for brain health promotion via individual- and community-level interventions to increase greenspace exposure.

A strength of this study is our novel approach of investigating the combination of systemic inflammation and greenspace exposure in influencing WMH volume. Several limitations must be noted. Our smaller sample size may have limited detection of an interaction between hs-CRP and % tree canopy, which may require a larger sample for sufficient statistical power. Although we did not find an association with WMH when focused on the continuous hs-CRP variable, we did observe that high risk hs-CRP levels were associated with greater WMH when focused on a clinically established cut point (hs-CRP >3 versus ≤3 mg/L). Additional studies with larger sample sizes will be needed to further investigate potential non-linearity of the association between hs-CRP and WMH. We did not have enough participants with data on GFAP, a neuroinflammation biomarker. GFAP will need to be examined once HBI data accrues further. In addition, our analyses are cross-sectional. New studies could assess whether sustained changes in inflammation impact WMH and examine the interplay of systemic inflammation, greenspace exposure, and WMH volume changes over time, as some evidence suggests that those with high hs-CRP that declined over time did not have an increased risk of WMH worsening and those with increasing hs-CRP over time had the highest WMH volumes. ¹⁵ Also, we defined greenspace as neighborhood % tree canopy; however, additional studies would benefit from objective measures of time spent in their neighborhood greenspaces to more accurately depict greenspace exposure. Although this study was intended to shed light on the association of systemic inflammation and % tree canopy with WMH volume, independently and combined, future studies are needed to investigate the mechanisms that explain how systemic inflammation and greenspace alter WMH volume. Lastly, additional studies could also examine if the protective effects of greenspace on brain health are in part explained by a reduction in inflammation. In prior studies, greater exposure to greenspace and time spent outdoors was linked to lower levels of systemic inflammation, supporting earlier findings that environmental factors such as greenspace can influence or modify inflammatory processes and in turn affect health outcomes. ²⁷

Overall, we found that high systemic inflammation was associated with greater WMH, high neighborhood % tree canopy was associated with lower WMH, and among those with high systemic inflammation, those living in greener neighborhoods (> % tree canopy) had lower WMH. If these findings are replicated and if plausible causal mechanisms are elucidated, planning and policy changes to increase neighborhood tree canopy could help decrease WMH in older adults and vulnerable groups (i.e., with high systemic inflammation), thus serving as a population-level intervention to reduce AD and related dementias and cerebrovascular risk.

Data availability statement

The data supporting the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.