Association of White Matter Rarefaction, Arteriolosclerosis, and Tau With Dementia in Chronic Traumatic Encephalopathy

Key Points

Question

What are the contributions of white matter rarefaction and cerebrovascular disease to dementia in older, deceased individuals who had played football and developed chronic traumatic encephalopathy?

Findings

In this cross-sectional study of 180 deceased individuals older than 40 years who had played football and had chronic traumatic encephalopathy, the number of years of football play (a proxy for repetitive head impacts) was associated with worse white matter rarefaction and greater dorsolateral frontal cortex neurofibrillary tangles. White matter rarefaction and neurofibrillary tangles were associated with dementia; arteriolosclerosis was not associated with the number of years of play, but it contributed to dementia.

Meaning

In chronic traumatic encephalopathy, dementia is likely a result of neuropathologic changes associated with repetitive head impacts, including white matter rarefaction and phosphorylated tau, in addition to nonhead trauma–associated pathologic changes, such as arteriolosclerosis.

This cross-sectional study investigates association of white matter rarefaction and cerebrovascular disease with dementia in deceased men older than 40 years who had played football and been found to have chronic traumatic encephalopathy.

Abstract

Importance

Chronic traumatic encephalopathy (CTE) is a neurodegenerative disease associated with repetitive head impacts, including those from US football, that presents with cognitive and neuropsychiatric disturbances that can progress to dementia. Pathways to dementia in CTE are unclear and likely involve tau and nontau pathologic conditions.

Objective

To investigate the association of white matter rarefaction and cerebrovascular disease with dementia in deceased men older than 40 years who played football and had CTE.

Design, Setting, and Participants

This cross-sectional study involves analyses of data from the ongoing Understanding Neurologic Injury and Traumatic Encephalopathy (UNITE) Study, which is conducted via and included brain donors from the Veterans Affairs–Boston University–Concussion Legacy Foundation brain bank between 2008 and 2017. An original sample of 224 men who had played football and were neuropathologically diagnosed with CTE was reduced after exclusion of those younger than 40 years and those missing data.

Exposures

The number of years of football play as a proxy for repetitive head impacts.

Main Outcomes and Measures

Neuropathological assessment of white matter rarefaction and arteriolosclerosis severity (on a scale of 0-3, where 3 is severe); number of infarcts, microinfarcts, and microbleeds; and phosphorylated tau accumulation determined by CTE stage and semiquantitative rating of dorsolateral frontal cortex (DLFC) neurofibrillary tangles (NFTs) (none or mild vs moderate or severe). Informant-based retrospective clinical interviews determined dementia diagnoses via diagnostic consensus conferences.

Results

A total of 180 men were included. The mean (SD) age of the sample at death was 67.9 (12.7) years. Of 180, 120 [66.7%]) were found to have had dementia prior to death. Moderate to severe white matter rarefaction (84 of 180 [46.6%]) and arteriolosclerosis (85 of 180 [47.2%]) were common; infarcts, microinfarcts, and microbleeds were not. A simultaneous equations regression model controlling for age and race showed that more years of play was associated with more severe white matter rarefaction (β, 0.16 [95% CI, 0.02-0.29]; P = .03) and greater phosphorylated tau accumulation (DLFC NFTs: β, 0.15 [95% CI, 0.004-0.30]; P = .04; CTE stage: β, 0.27 [95% CI, 0.14-0.41]; P < .001). White matter rarefaction (β, 0.16 [95% CI, 0.02-0.29]; P = .03) and DLFC NFTs (β, 0.16 [95% CI, 0.03-0.28]; P = .01) were associated with dementia. Arteriolosclerosis and years of play were not associated, but arteriolosclerosis was independently associated with dementia (β, 0.21 [95% CI, 0.07-0.35]; P = .003).

Conclusions and Relevance

Among older men who had played football and had CTE, more years of football play were associated with more severe white matter rarefaction and greater DLFC NFT burden. White matter rarefaction, arteriolosclerosis, and DLFC NFTs were independently associated with dementia. Dementia in CTE is likely a result of neuropathologic changes, including white matter rarefaction and phosphorylated tau, associated with repetitive head impact and pathologic changes not associated with head trauma, such as arteriolosclerosis.

Introduction

Exposure to repetitive head impacts (RHI)¹ is associated with chronic traumatic encephalopathy (CTE).^2,3,4,5,6 This condition has been diagnosed neuropathologically in athletes who played contact sports, particularly US football players, using criteria developed by a consensus conference sponsored by National Institute of Neurological Disorders and Stroke and National Institute of Biomedical Imaging and Bioengineering.^2,3,4,7 Chronic traumatic encephalopathy clinically presents with cognitive, behavioral, and mood symptoms that can progress to dementia.^2,3,5,8,9,10 The contribution of phosphorylated tau pathologic changes, comorbid neurodegenerative disease, or other pathologic conditions to dementia in CTE remains unclear.^11,12,13,14

The late effects of RHI might contribute to dementia in CTE via tau and nontau pathways, including cerebrovascular disease (CBVD). Cerebrovascular disease often coexists with neurodegenerative diseases, especially Alzheimer disease (AD), and contributes to cognitive decline.^{15,16,17,18,19,20,21,22} Arteriolosclerosis, infarcts, and microinfarcts are common with aging and cardiovascular disease (CVD), and CVD is frequent in individuals who had previously played US football.^23,24 In the setting of CTE, CBVD may also encompass white matter (WM) degeneration secondary to RHI. In living men who had previously played for the National Football League, WM alterations are present on magnetic resonance imaging,^25,26,27 are associated with RHI,²⁵ and affect executive function.²⁵ In vivo imaging studies^{27,28,29,30,31} in athletes who played contact sports (eg, US football players, boxers) show cerebral blood flow alterations and blood-brain barrier disruption. These brain changes may be associated with or result in chronic WM degeneration and affect cognitive outcomes.³²

Neuropathological studies^2,33,34,35 show that WM degeneration and CBVD are common comorbidities in CTE, but their role in the clinical manifestations of CTE is poorly understood. We investigated the association of WM pathologic changes and CBVD with dementia in older men with CTE who were deceased and had played football by examining arteriosclerosis, infarcts, microinfarcts, microbleeds, and WM rarefaction. White matter rarefaction might reflect both CBVD pathologic conditions associated with CVD³⁶ and pathologic conditions (eg, axonal loss) that are secondary to RHI.^12,13 We hypothesized that WM rarefaction would be associated with years of football (a proxy for RHI exposure), that WM rarefaction would contribute to dementia, and that other markers of CBVD pathologic changes associated with CVD (eg, arteriolosclerosis) would independently contribute to dementia.

Methods

Participants

The sample included deceased men who had been football players and were neuropathologically diagnosed with CTE, from the Understanding Neurologic Injury and Traumatic Encephalopathy (UNITE) Study and Veterans Affairs–Boston University–Concussion Legacy Foundation brain bank.^8,37 The next of kin contacted the CTE Center to arrange brain donation near the time of or after the individual’s death, in some cases. Other brain donors were referred by medical examiners, recruited by the Concussion Legacy Foundation, or agreed to donation during life. To be eligible, donors needed to have a history of RHI (eg, from contact sports or military service). Brain donors were excluded for a prolonged postmortem interval or poor tissue quality. Antemortem symptomatic status was not an inclusion or exclusion criterion. Institutional review board approvals for brain donation were obtained through the Boston University Medical Campus. Approval for postmortem clinical record review, interviews with informants, and neuropathological evaluation was obtained through the Boston University Medical Campus institutional review board. The next of kin provided written informed consent.

Neuropathological Diagnoses

Neurodegenerative Diseases

Methods for pathological processing and evaluation have been published elsewhere and follow established procedures.^38,39 Neuropathologists (T.D.S., B.R.H., and A.C.M.) were blinded to clinical data. Non-CTE neurodegenerative diseases were diagnosed using established neuropathological diagnostic criteria.⁷ The neuropathological diagnosis of CTE was made using criteria defined by the National Institutes of Neurological Disease and Stroke and National Institute of Brain Imaging and Behavior consensus panel.⁷ The CTE stage and severity of NFT burden in the dorsolateral frontal cortex (DLFC) served as semiquantitative scales of phosphorylated tau severity. The pathological stage of each CTE case was graded using the 4-stage classification scheme^{2,3,8,9,10,11,13} (eMethods 1 in the Supplement contains a description of the CTE staging scheme). Phosphorylated tau density of the dorsolateral frontal cortex (DLFC) (on a scale of 0 to 3 points; 0 indicates no neurofibrillary tangles [NFTs], while 3 indicates severe NFTs) was examined because it is an initial site of phosphorylated tau deposition in CTE that becomes severely affected with disease advancement, and phosphorylated tau severity in this region has been shown to be associated with dementia in CTE.^2,3,13

White Matter Rarefaction and CBVD

The neuropathological examination included semiquantitative assessment of WM rarefaction and arteriolosclerosis, as well as quantitative assessment of infarcts, microinfarcts, and microbleeds.⁴⁰ White matter rarefaction and arteriolosclerosis were assessed using slides stained with luxol fast blue and hemotoxylin-eosin. White matter rarefaction is a measure of overall WM integrity and defined by the degree of myelin loss, extent of tissue attenuation or vacuolization around small blood vessels, and density of reactive astrocytes. Arteriolosclerosis was defined as the concentric hyaline thickening of the media of arterioles. Overall ratings of WM rarefaction and arteriolosclerosis severity were made by the neuropathologists (T.D.S., B.R.H., and A.C.M.) using a scale of 0 to 3 points, with 0 indicating none and 3 indicating severe changes, and these findings were based on the evaluation of the subcortical WM of the middle frontal, inferior parietal, superior temporal, and occipital cortices, as well as deep WM of the basal ganglia. The presence or absence and number of remote (not acute) infarcts, microinfarcts, and microbleeds were evaluated in the cerebral cortex, subcortical WM, deep WM, basal ganglia, brainstem, and cerebellum.

Informant-Based Retrospective Clinical Evaluation

Clinical data were obtained through online surveys and retrospective telephone clinical interviews with informants.^3,8,9,37 A neurologist and/or neuropsychologist (M.L.A., R.A.S., and J.M.) with expertise in neurodegenerative disorders conducted telephone clinical interviews with informants to obtain medical and clinical histories, including the presence, nature, and timeline of symptoms associated with cognition, behavior or mood, and daily functioning. A consensus diagnosis of dementia was adjudicated based on modified Diagnostic and Statistical Manual of Mental Disorders (Fourth Edition, Text Revision) criteria. eMethods 2 in the Supplement provides further details.

Statistical Analyses

The DLFC NFT burden was recoded into the categories none to mild vs moderate to severe. The CTE diagnoses were regrouped into stages I and II vs III and IV. These variables were recoded to optimize equal group sample sizes and limit ambiguity regarding the severity of CTE pathologic changes (eg, between CTE stages I and II). Arteriolosclerosis and WM rarefaction were not dichotomized because of a relatively equal distribution across scales of 0 to 3 points; the model was repeated with them as binary variables to determine potential differences in model fit. Years of US football play served as a proxy for RHI.^2,9,13 The level of play (ie, youth or high school, college, semiprofessional, or professional) was not analyzed because it is highly correlated with the number of years of play (r = 0.64).

As described in eMethods 3 in the Supplement, proportional odds and logistic regression models were performed to guide selection of variables into a simultaneous equations regression model and determine model pathways among years of football play, WM rarefaction, CBVD, phosphorylated tau pathologic characteristics, and dementia. A P value of .20 or less served as the selection criterion into the model. Age and race were included as covariates. The model tested for direct (ie, from independent variable to outcome pathways) and indirect (ie, partial or complete mediation) results. The model fit was evaluated by the Akaike information criteria, standardized root mean square residual, goodness of fit index, Bentler-Bonett normed fit index, and root mean square error of approximation.⁴¹ Standardized coefficients (β’s) are presented. For model factors associated with dementia, a binary logistic regression model that controlled for age and race was used to report on their association with odds for dementia. Statistical significance was defined by P values less than .05. Data analysis was completed from February 2018 to February 2019 with SAS version 9.4 (SAS Institute Inc).

Results

Participants included 224 deceased men who had played US football; this was reduced to 180 individuals after exclusion of individuals missing data on independent variables and outcome variables (n = 14) and participants younger than 40 years at death (n = 30). Younger participants were excluded because dementia and neuropathological outcomes are associated with age. Only 2 individuals younger than 40 years had dementia; CTE, arteriolosclerosis, and WM rarefaction severity were mild or absent in participants younger than 40 years (Table 1). Spline analyses (not shown) showed that odds for dementia and severity of CTE, arteriolosclerosis, and WM rarefaction pathologic changes tended to increase starting at age 40 years.

Table 1. Dementia and Pathologic Characteristics in Those Younger Than 40 Years Compared With Those 40 Years and Older.

Characteristic	Individuals, No. (%)		P Value
	<40 y (n = 30)a	≥40 y (n = 180)a
Dementia	2 (6.7)	120 (66.7)	<.001
Chronic traumatic encephalopathy stage
I/II	30 (100.0)	35 (19.4)	<.001
III/IV	0	145 (80.6)
Dorsolateral frontal cortex neurofibrillary tangle density rating
None or mild	20 (66.7)	50 (27.8)	<.001
Moderate or severe	10 (33.3)	130 (72.2)
Arteriolosclerosisb
None	26 (86.7)	54 (30.0)	<.001
Mild	4 (13.3)	41 (22.8)
Moderate	0	67 (37.2)
Severe	0	18 (10.0)
White matter rarefaction, mean (SD)b
None	16 (53.3)	22 (12.2)	<.001
Mild	7 (23.3)	74 (41.1)
Moderate	4 (13.3)	58 (32.2)
Severe	3 (10.0)	26 (14.4)
Remote microinfarcts	1 (3.3)	49 (27.2)	.002

^aIndividuals who were younger than 40 years were excluded from analyses because of low pathologic burden and minimal presence of dementia (remote infarcts and microbleeds are not shown because of low burden in the overall sample). The final sample size of 180 individuals was finalized after exclusion of individuals for missing data and those younger than 40 years. Analyses were completed after exclusion of the 14 individuals who had missing data.

^bOrdinal regression was conducted for arteriolosclerosis and white matter rarefaction. Fisher exact tests were used to examine differences between the 2 groups for all other variables because of small cell sizes.

Sample Characteristics

Table 2 and Table 3 present sample characteristics. When the DLFC NFT burden was recoded into 2 categories, 50 individuals were in the none-to-mild group vs 130 in the moderate-to-severe group. Thirty-five individuals had CTE at stages I and II, and 145 had CTE at stages III and IV.

Table 2. Sample Characteristics of the Deceased US Football Players.

Characteristic	Patients, No. (%)
Total, No.	180a
Demographic Factors
Age at death, mean (SD), y	67.9 (12.7)
African American	34 (18.9)
Education level
Less than high school or some high school	1 (0.6)
High school or graduate equivalency diploma	2 (1.1)
Some college	34 (18.9)
College degree	94 (52.2)
More than college	10 (5.6)
Graduate degree	39 (21.7)
Athletic participation time variables, mean (SD), y
Football play	14.9 (5.7)
Since retirement	40.9 (12.7)
Age at first exposure to football	12.1 (2.8)
Highest level played
Youth or high school	7 (3.9)
College	45 (25.0)
Semiprofessional	7 (3.9)
Professional	121 (67.2)
Football primary position
Lineman (offensive or defensive)	83 (46.1)
Linebacker	22 (12.2)
Defensive back or safety	18 (10.0)
Running back	27 (15.0)
Quarterback	13 (7.2)
Special teams	1 (0.6)
Other or multiple	9 (5.0)
Wide receiver	2 (1.1)
Unknown	5 (2.8)
Any other contact sport history	28 (15.6)
Military historyb	53 (29.4)
Combat exposure	7 (3.9)
Neuropathologic Factors
Chronic traumatic encephalopathy
I	12 (6.7)
II	23 (12.8)
III	86 (47.8)
IV	59 (32.8)
Dorsolateral frontal cortex neurofibrillary tangles density rating
None	5 (2.8)
Mild	45 (25.0)
Moderate	40 (22.2)
Severe	90 (50.0)
Lewy body disease
Brainstem predominant	9 (5.0)
Limbic or neocortical	29 (16.1)
Frontotemporal lobar degeneration
Tau	8 (4.4)
TDP-43	8 (4.4)
Motor neuron disease	10 (5.6)
Prion disease	2 (1.1)
Alzheimer disease neuropathologic changes
Not Alzheimer disease	90 (50.0)
Alzheimer disease neuropathologic change severity
Low	25 (13.9)
Intermediate	41 (22.8)
High	24 (13.3)
Consortium to Establish a Registry for Alzheimer Disease neuritic plaque score
None	101 (56.1)
Sparse	50 (27.8)
Moderate	17 (9.4)
Frequent	12 (6.7)
Braak stage
0-III	142 (80.2)
IV-VI	35 (19.8)

Abbreviation: TDP-43, transactive response DNA-binding protein 43 kDa.

^aOf the 180 participants, samples sizes were reduced to 179 individuals for combat history and 177 individuals for Braak stage because of missing data.

^bMilitary history refers to any history of service in the military, regardless of combat exposure.

Table 3. Cerebrovascular Disease Characteristics.

Characteristic	Patients, No. (%)			P Valueb
	Total Sample (N = 180)	CTE Only (n = 101)	CTE and Comorbidities (n = 79)a
White matter rarefaction
None	22 (12.2)	16 (15.8)	6 (7.6)	.91
Mild	74 (41.1)	39 (38.6)	35 (44.3)
Moderate	58 (32.2)	32 (32.7)	25 (31.6)
Severe	26 (14.4)	13 (12.9)	13 (16.5)
Atherosclerosis
None	102 (60.0)	58 (59.2)	44 (58.7)	.30
Mild	34 (20.0)	19 (20.0)	15 (20.0)
Moderate	25 (14.7)	11 (11.6)	14 (18.7)
Severe	9 (5.3)	7 (7.4)	2 (2.7)
Arteriolosclerosis
None	54 (30.0)	29 (28.7)	25 (31.6)	.04
Mild	41 (22.8)	20 (19.8)	21 (26.6)
Moderate	67 (37.2)	43 (42.6)	24 (30.4)
Severe	18 (10.0)	9 (8.9)	9 (11.4)
Remote microinfarcts	49 (27.2)	29 (28.7)	20 (25.3)	.09c
Cerebral cortex, No.
1	8	4	4
2	10	5	5
≥3	10	9	1
Subcortical and periventricular white matter, No.
1	7	4	3
2	4	3	1
≥3	8	3	5
Subcortical gray matter, No.
1	5	1	4
2	5	4	1
≥3	5	3	2
Brainstem and cerebellum, No.
1	7	7	0
2	4	2	2
≥3	4	3	1
Remote infarcts	30 (16.9)	18 (18.0)	12 (15.6)	.39c
Cerebral cortex, No.
1	11	9	2
2	4	2	2
≥3	2	1	1
Deep cerebral gray matter and internal capsule, No.
1	6	3	3
2	1	1	0
≥3	0	0	0
Subcortical and periventricular white matter, No.
1	7	3	4
2	0	0	0
≥3	1	1	0
Brainstem and cerebellum, No.
1	4	2	2
2	1	1	0
≥3	0	0	0
Remote microbleeds	5 (2.8)	2 (2.0)	3 (3.8)	.62c
Cerebral cortex, No.
1	2	1	1
2	1	0	1
≥3	2	1	1
Subcortical and periventricular white matter, No.
1	0	0	0
2	0	0	0
≥3	2	1	1
Subcortical gray matter, No.
1	0	0	0
2	1	0	1
≥3	1	1	0
Brainstem and cerebellum, No.
1	0	0	0
2	0	0	0
≥3	2	1	1

Abbreviation: CTE, chronic traumatic encephalopathy.

^aParticipants with CTE and other comorbid neurodegenerative disease diagnoses, whereas CTE only includes those with CTE and who did not have comorbid neurodegenerative disease diagnoses.

^bOrdinal regression controlling for age compared differences between participants with CTE only and those with CTE and comorbid diagnoses for white matter rarefaction, atherosclerosis, and arteriolosclerosis. Logistic regression controlling for age was used to examine differences between the participant groups on all remaining variables.

^cCompared absence vs presence. Of the 180 participants, sample sizes were reduced because of missing data to 170 participants for atherosclerosis and 177 participants for remote infarcts. There are variable sample sizes across the regional assessments because of missing data.

Of the 180 men who donated their brains, 120 were determined to have had antemortem dementia. Microinfarcts were examined globally (ie, as an overall absence or presence) because of the insufficient sample size at the level of brain region. Because of the small number of participants who had infarcts and microbleeds, these variables were not examined.

Simultaneous Equations Regression Model Results

Results of the proportional odds and logistic regression analyses are presented in eMethods 4 in the Supplement. Results from these models remain unchanged when controlling for the football position(s) played and age at first exposure to US football. These were not included the final model. Figure 1 depicts a summary of the primary findings from the simultaneous equations regression model. The model fit was excellent (Akaike information criteria, 77.56; standardized root mean square residual, 0.02; goodness of fit index, 0.99; root mean square error of approximation <0.01; Bentler-Bonett normed fit index, 0.98; χ² = 3.56; P = .83), and findings were similar when WM rarefaction and arteriolosclerosis were recoded as none to mild vs moderate to severe (Akaike information criteria, 76.38; standardized root mean square residual, 0.02; goodness of fit index, 0.99; Bentler-Bonett normed fit index, 0.99; root mean square error of approximation <0.01; χ² = 2.38; P = .94). Increasing age was associated with more severe WM rarefaction (β, 0.37 [95% CI, 025-0.50]; P < .001), high CTE stage (β, 0.32 [95% CI, 0.19-0.45]; P < .001), and increased likelihood for dementia (β, 0.27 [95% CI, 0.13-0.42]; P < .001).

Figure 1. Cross-Sectional Model on the Contributions of White Matter Rarefaction, Arteriolosclerosis, and Phosphorylated Tau to Dementia in Chronic Traumatic Encephalopathy (CTE).

Simultaneous equations regression models tested the association of pathological markers of cerebrovascular disease with dementia in older deceased individuals who had played football and had CTE. The Figure provides an illustrative conceptual summary of the primary significant findings. Not all variables or pathways are displayed for ease of presentation. See eMethods 3 and 4 in the Supplement for a summary of all variables included in the model and associated pathways estimated. The β values shown are standardized estimates, and each path shown is significant at an α less than .05. Standardized estimates are interpreted as changes in SD units. For instance, there was a 0.16-SD change in dementia per unit SD increase in white matter rarefaction and dorsolateral frontal cortex neurofibrillary tangles of hyperphosphorylated tau. Standardized estimates permit direct comparison of the effect sizes across pathways. The dashed line between years of football play and dementia denotes a significant indirect effect in which the association of years of football play with dementia was mediated by dorsolateral frontal cortex neurofibrillary tangles and severity of white matter rarefaction. Arteriolosclerosis and white matter rarefaction were rated on a scale of 0 to 3 points, with 3 indicating severe pathologic changes. Dorsolateral frontal cortex neurofibrillary tangles was a binary variable, with 0 indicating a low burden of neurofibrillary tangles and 1 a high burden.

The association between age and dementia was partially mediated by WM rarefaction (indirect effect: β, 0.05 [95% CI, −0.005 to 0.11]; P = .07). There were no significant associations of race with WM rarefaction, CTE stage, DLFC NFT burden, or dementia.

More years of play was associated with increased likelihood for more severe WM rarefaction (β, 0.16 [95% CI, 0.02-0.29]; P = .02), as well as moderate to severe DLFC NFT burden (β, 0.15 [95% CI, 0.004-0.30]; P = .04) and high CTE stage (β, 0.27 [95% CI, 0.14-0.41]; P < .001). More severe WM rarefaction (β, 0.16 [95% CI, 0.02-0.29]; P = .03) and greater DLFC NFT burden (β, 0.16 [95% CI, 0.03-0.28]; P = .01) subsequently corresponded to increased likelihood for having dementia. As shown by the standardized β values, WM rarefaction and DLFC NFT burden were equally associated with dementia. The association of years of play with dementia was mediated by WM rarefaction and DLFC NFT burden (indirect effect: β, 0.05 [95% CI, 0.006-0.09]; P = .03). Although arteriolosclerosis was not associated with years of football play, it was independently associated with dementia (β, 0.21 [95% CI, 0.07-0.35]; P = .003). The simultaneous equations regression model accounted for all variance in the model; therefore, the association of WM rarefaction and arteriolosclerosis with dementia were independent of all variables, including phosphorylated tau severity. Figure 2 shows exemplar images of arteriolosclerosis, WM rarefaction, and phosphorylated tau pathologic changes.

Figure 2. White Matter Rarefaction, Arteriolosclerosis, and Dorsolateral Frontal Cortex Tau Pathologic Changes in Participants With Chronic Traumatic Encephalopathy.

A, Luxol fast blue with hemotoxylin-eosin histochemical staining shows robust myelin staining (blue) in a former US football college player in his early 40s who was neuropathologically diagnosed with chronic traumatic encephalopathy (CTE) (stage I/II) who was not determined to have had antemortem dementia. B and C, In a man who had played professional US football, was in his mid-80s, had been neuropathologically diagnosed with CTE (stage III/IV), and was determined by consensus to have dementia, there was severe loss (3+) of myelinated fibers (B), as well as marked arteriolosclerosis with hyalinization of the vessel wall (C, arrowhead). D-F, Immunohistochemical staining for tau pathologic changes (antibody AT8) in the dorsolateral frontal cortex shows perivascular accumulations of abnormal tau at the sulcal depths within neurons and cell processes (asterisks indicate the lumen of cortical vessels; arrowheads, examples of pretangles and tangles within neurons). The scale bar for A-C and F is 100 μm; for D and E, 150 μm.

Sensitivity analyses were conducted to clarify whether the associations of WM rarefaction and arteriolosclerosis with dementia were independent of comorbid neurodegeneration. Overall model fit and significant associations of WM rarefaction and arteriolosclerosis with dementia remained when AD Neuropathologic Change was added as a covariate, as well as when individuals diagnosed with prion disease and motor neuron disease were excluded from the sample. The model was repeated in participants who had CTE alone (n = 101). In this reduced sample, the effect sizes of WM rarefaction (β, 0.13 [95% CI, −0.07 to 0.32]) and arteriolosclerosis (β, 0.25 [95% CI, 0.06-0.43]) on dementia were similar to those in the entire sample.

White Matter Rarefaction and Arteriolosclerosis and Dementia: Logistic Regression

A logistic regression controlling for age and race showed that participants who had more severe WM rarefaction were at 1.69 times greater odds for dementia (odds ratio [OR], 1.69 [95% CI, 1.08-2.62]; P = .03). Similar outcomes were observed for arteriolosclerosis (OR, 1.81 [95% CI, 1.23-2.67]; P = .003). The ORs for WM rarefaction (OR, 1.68 [95% CI, 1.08-2.62]; P = .02) and arteriolosclerosis (OR, 1.84 [95% CI, 1.23-2.75]; P = .003) remained unchanged when DLFC NFTs were included as a covariate. A logistic regression showed the dementia outcomes associated with WM rarefaction (OR, 1.59 [95% CI, 1.02-2.47]; P = .04) and arteriolosclerosis (OR, 1.72 [95% CI, 1.16-2.55]; P = .007) remained stable after controlling for AD Neuropathologic Change (and age and race). Compared with participants with a low DLFC NFT burden, those with a high DLFC NFT burden had 2.65 times the odds of having dementia (OR, 2.65 [95% CI, 1.24-5.70]; P = .01), after controlling for age and race.

Sensitivity Analyses: Associations With Atherosclerosis and Cardiovascular Disease

Cumulative logit proportional odds (ordinal outcomes) and logistic (binary outcomes) regression models that controlled for age and race examined whether WM rarefaction and arteriolosclerosis reflect downstream sequelae from large-vessel disease (eg, atherosclerosis). The sample size for these analyses was 170 individuals because of missing atherosclerosis data. Large-vessel atherosclerosis increased the odds for more severe small-vessel arteriolosclerosis (OR, 1.62 [95% CI, 1.16-2.27]; P = .005) and microinfarcts (OR, 2.27 [95% CI, 1.48-3.48]; P < .001). Atherosclerosis was not associated with WM rarefaction (data not shown). Statistically significant associations were not found between WM rarefaction with arteriolosclerosis and microinfarcts.

In the full sample minus the individuals excluded because of missing data, 81 of 133 men (60.9%) had a reported history of hypertension, and 29 of 134 men (21.6%) had a reported history of diabetes. Logistic regression showed years of football play was not associated with hypertension or diabetes (data not shown). Simple proportional odds regression revealed that hypertension was associated with more severe arteriolosclerosis (OR, 1.26 [95% CI, 1.03-2.38]; P < .001) and atherosclerosis (OR, 2.51 [95% CI, 1.20-3.25]; P = .004), but not WM rarefaction. A history of diabetes was not associated with severity of atherosclerosis, arteriolosclerosis, or WM rarefaction (data not shown). Binary logistic regressions showed no significant associations between a history of hypertension and diabetes with dementia (data not shown). All sensitivity analyses controlled for age and race.

Discussion

In this convenience sample of deceased men older than 40 years who had played US football and were neuropathologically diagnosed with CTE, longer duration of football play was associated with more severe WM rarefaction and greater DLFC NFT burden. Worse WM rarefaction severity and greater DLFC NFT burden subsequently corresponded to an increased likelihood of dementia. Arteriolosclerosis was also independently associated with increased odds for having dementia. Dementia in older individuals with CTE may be a result of neuropathologic conditions associated with RHI, including WM rarefaction and DLFC NFTs, as well as non-RHI–associated pathologic conditions, such as arteriolosclerosis.

Arteriolosclerosis is a marker of cerebral small-vessel disease associated with aging and CVD. As expected, it was associated with hypertension and atherosclerosis in this sample, but not years of play. The causative mechanisms of WM rarefaction are multifaceted and might reflect CBVD associated with CVD.³⁶ White matter rarefaction can also represent demyelination and axonal loss.⁴² White matter rarefaction was not associated with atherosclerosis, arteriolosclerosis, or CVD risk factors. Instead, more years of football, a surrogate metric of RHI exposure,^2,9,13 was a corollary of more severe WM rarefaction. In CTE, WM rarefaction may encompass diffuse, long-term degeneration (eg, axonal and myelin loss, astrocytosis, neuroinflammation) secondary to RHI from football.^12,13 Acute WM injuries^{6,43,44,45,46} from RHI might persist over time^{2,3,25,47,48,49} and worsen with age. Studies that include direct CVD and RHI metrics and refined measures of WM integrity are needed to improve understanding of the pathogenesis of WM rarefaction and cerebral small vessel changes in CTE.

White matter rarefaction and arteriolosclerosis were associated with increased odds for dementia independent of phosphorylated tau; they were not associated with phosphorylated tau severity. White matter rarefaction and arteriolosclerosis thus had independent or additive associations with dementia. These pathologic conditions also independently affect the clinical expression of AD.^{19,50,51,52,53,54} Cases of CTE often present with various cognitive and neuropsychiatric changes that can progress to dementia.^3,8,55 The extent to which these symptoms are accounted for by phosphorylated tau is unclear, especially behavioral and mood changes that occur early in life.^3,8,9,56 Nontau pathologic changes, including WM degeneration associated with RHI and arteriolosclerosis associated with aging or CVD, may affect the clinical expression of CTE and contribute to clinical hetereogeneity.⁴⁶

The DLFC is an initial region of phosphorylated tau accumulation in CTE, and DLFC NFTs burden contributed to dementia in this sample of participants with CTE and other comorbid neurodegenerative pathologic characteristics. These findings are similar to data linking plaques and NFTs from AD with clinical decline.^19,50,57 Dementia in AD is a result of mixed neuropathologic changes (not only phosphorylated tau),^{11,58,59,60,61,62} and the current findings support a similar phenomenon in CTE. In 2013, Stern et al⁸ reported a dementia rate of approximately 30% (compared with 67% in this sample) among a small sample of men (n = 36) who had played US football, were neuropathologically diagnosed with CTE, and were without comorbid neurodegenerative disease. That smaller sample was younger (mean age, 57 years; compared with 68 years in the current sample). The lower rate of dementia might reflect the lack of comorbid neurodegenerative disease in the Stern et al study.⁸

Limitations

The current findings are limited by ascertainment bias that reduces the generalizability of the findings to the broader population of individuals who play US football. Population-based studies are needed to overcome potential biases from sample selection. Ascertainment bias may have had minimal influence on the estimated associations. The accuracy of the estimated associations would only be influenced by ascertainment bias if both the independent variables and outcomes were differentially associated with brain bank selection. It is theoretically unlikely that CBVD had an independent differential association with selection into the brain bank. Recent research showed that the associations of age at first exposure to football on clinical outcomes remained unchanged after adjusting (through inverse-probability weighting) for some factors shown to affect brain bank selection among older adults (eg, dementia, depression, age, race).^9,63 Continued studies that elucidate potential biases associated with sample selection into the Understanding Neurologic Injury and Traumatic Encephalopathy study will be important to place the current and future findings in context.

The DLFC NFTs were examined based on a priori hypotheses and the extant data.^2,3,13 Regional evaluation of phosphorylated tau was beyond the scope of this study and not examined to avoid an overidentified model, decreased statistical power attributable to the number of pathways to be estimated, violation of statistical assumptions associated with multicollinearity, and decreased sample size attributable to missing data across the regions. A separate study that uses advanced statistical modeling and refined quantitative measures of regional phosphorylated tau will clarify the associations between regional phosphorylated tau severity, RHI, and dementia. Granular examinations of microvasculopathy were not conducted, which may be more sensitive to the associations of head trauma with CTE.^6,34 Future research should examine other markers of vascular injury, such as C-reactive protein, vascular cell adhesion molecule 1, intercellular adhesion molecule 1, and measures of vessel tortuosity. Inclusion of a neurodegenerative disease comparison group, such as individuals with AD, would provide insight into distinctive patterns of CBVD severity in CTE, as well as comparative contributions of CBVD to dementia. Dementia was determined by consensus diagnostic conferences based on information obtained through retrospective telephone interviews with informants. This approach might have affected the accuracy of the estimated outcomes because of subjectivity and recall biases. Corroboration of in-life dementia diagnoses through medical records could be an added source of supporting evidence. Data from medical records collected as part of the larger study of this sample are not yet ready for analysis. In addition, APOE genotyping is ongoing and not included here. The exact role of apolipoprotein E in CTE is unclear,^2,8 but apolipoprotein E status might have influenced many of the associations examined.

Conclusions

In this convenience sample of older football players with neuropathologically confirmed CTE, more years of football play were associated with more severe WM rarefaction and greater phosphorylated tau accumulation. White matter rarefaction, DLFC NFTs, and arteriosclerosis were independently associated with dementia. Increased understanding of comorbid contributors to the pathogenesis and symptom profile of CTE is imperative to fully diagnose, treat, and prevent this progressive brain disease.

Supplement.

eMethods 1. CTE stage classification scheme

eMethods 2. Dementia diagnostic procedures

eMethods 3. Statistical analyses

eMethods 4. Results of proportional odds and logistic regression models

eReferences.