No significant associations between history of head injury and Alzheimer’s disease fluid biomarkers in older adults

Abstract

Background:

Concussions are gaining attention as a risk factor for Alzheimer’s disease (AD). Previous reports suggest concussion, also called head injury (HI), may be associated with changes to AD biomarkers, including amyloid and tau. However, there has been little characterization of biofluid biomarkers in older adults with remote history of HI.

Objective:

We investigated whether aging participants at risk for AD with self-reported HI history would demonstrate alterations to cerebrospinal fluid (CSF) and blood plasma biomarkers of AD.

Methods:

Using two-way ANCOVAs and linear mixed effects models, we examined both baseline cross-sectional and longitudinal associations between HI history, cognition, and AD biofluid biomarkers in 100 participants with HI history compared to 2411 without HI history from the ADNI dataset.

Results:

On baseline analysis, participants with HI history had higher CSF Aβ_42/40 ratios than non-HI participants. There were no other baseline differences in biomarkers between HI and non-HI participants, nor were there any main effects of HI upon longitudinal analysis. We observed consistent main effects of age and cognitive impairment that suggested a pattern of worsened AD biomarker signatures in impaired participants with increasing age.

Conclusions:

Our findings do not support an association between self-reported HI history and AD fluid biomarkers in older adults from the ADNI dataset. Further characterization of fluid biomarker trajectories both in the acute post-HI period and in participants with remote HI is needed to understand the temporal dynamics of fluid biomarkers after HI and the implications of HI for AD risk.

Introduction

Alzheimer’s disease (AD) is the most common form of dementia, and over 150 million people worldwide are predicted to be diagnosed with AD by 2050.¹ AD is a continuum of increasing pathological and symptom severity that incorporates preclinical and prodromal stages, such as subjective cognitive decline (SCD) and mild cognitive impairment (MCI), before the dementia stage.² AD is pathologically defined by the presence of extracellular amyloid plaques and intracellular tau neurofibrillary tangles that deposit in characteristic Thal and Braak stages with increasing disease progression.^3,4

A multitude of well-characterized biological and potentially modifiable lifestyle-related risk factors affect one’s individual risk for AD. Biological risk factors include advanced age, female sex, and the presence of one or more APOE ϵ4 allele(s).^2,5 Head injury (HI), also known as concussion, is a type of traumatic brain injury (TBI) that is a potentially modifiable risk factor for AD, and there would be an estimated 3% reduction in dementia prevalence if head injuries were completely eliminated.⁶ However, the mechanisms mediating the link between HI and AD are not well understood. Molecular changes in the brain after HI are complex and can lead to neuronal dys-function.^7,8 Acute changes in the brain following HI include rapid, uncontrolled neuronal depolarization, release of excitatory neurotransmitters, alterations to ionic concentrations, diminished cerebral blood flow, and an increased demand for glucose.^7,8 These physiological changes may be associated with neuropsychiatric symptoms in the injured individual, such as headache, memory interruptions, sleep disruption, depression, and anxiety.^7,8 Additionally, upregulation of proteins associated with neurodegeneration, such as amyloid-β (Aβ) and hyperphosphorylated tau, have been observed in both animal models and human TBI patients, suggesting HI may induce neurodegenerative processes in the brain that contribute to long-term dementia risk.^9-14 Further, individuals who experience repetitive injuries may be at risk for different long-term outcomes than individuals with only one injury. Repetitive injuries, particularly those observed in professional athletes such as American football players or boxers, have been strongly linked to the development of chronic traumatic encephalopathy (CTE), a neurodegenerative disease characterized by behavioral dysregulation, changes in memory, attention, and executive functioning, and a unique pattern of hyperphosphorylated tau deposition in sulcal depths.^15-20 While there may be differential effects of single injuries versus multiple injuries on AD or CTE risk, there is little data available to examine this question.

Prior studies have reported that HI is associated with increased risk for AD, earlier disease onset, and increased deposition of both Aβ and tau on positron emission tomography (PET).^21-28 However, there is a lack of consensus in the literature, as other studies found no relationships.^29-36 Potential reasons for this lack of clarity may include separate strategies for HI ascertainment in different cohorts, large variations in sample size, and/or samples consisting of young versus older adults, among other reasons. As such, greater investigation into the possibility of HI-associated biomarker changes in older adults at risk for AD is needed to clarify whether such associations exist and thereby represent a mechanistic explanation behind HI as a risk factor for AD.

Phosphorylated tau (pTau) and Aβ are AD-specific biomarkers that can be precisely measured in biofluid samples, including cerebrospinal fluid (CSF) and blood plasma, thus offering an in-vivo method for detecting and tracking changes associated with AD.³⁷ With the advantages of being more accessible and less expensive than neuroimaging biomarkers like PET, fluid biomarkers also exhibit characteristic changes in patients with AD and can be used for diagnosis and disease staging.³⁷ However, it is important to consider that plasma levels of Aβ may reflect peripheral alongside central nervous system (CNS) amyloid processes.³⁸ As such, plasma Aβ levels may not exclusively reflect CNS Aβ deposition. The canonical AD biomarker signature is characterized by decreased Aβ and increased pTau in both CSF and plasma, and this signature is robust for detecting and diagnosing AD.³⁹ In addition to AD pathology-specific biomarkers, other biomarkers can be used to capture a more complete picture of ongoing pathology. For example, neurofilament light chain (NfL) is a non-specific neurodegeneration biomarker that is elevated in patients with AD.^40,41 Additionally, use of CSF and plasma biomarker ratios such as Aβ_42/40 is becoming commonplace, as these measures improve diagnostic accuracy and have high concordance with PET findings.^37,42,43

Intriguingly, HI has been linked to both acute and chronic changes in AD biomarkers such as Aβ and tau,^44-47 though fluid biomarker evidence from older adults is limited as most previous reports concentrated on the relatively acute post-injury phase or studied younger adults who were not at substantial risk for AD due to advanced age.^48-52 Additionally, there is little data examining HI and longitudinal AD fluid biomarkers.^29,31 Therefore, the goal of this study was to characterize AD biomarker patterns in CSF and blood plasma in individuals with self-reported history of HI from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset. We hypothesized participants with HI would show a worsened AD-like biomarker signature, and the associations between HI and biomarkers would be most pronounced in individuals with cognitive impairment.

Methods

ADNI participants

Data used in the preparation of this article were obtained from ADNI (http://adni.loni.usc.edu). The ADNI was launched in 2003 as a public-private partnership, led by Principal Investigator Michael W. Weiner, MD. The primary goal of ADNI has been to test whether serial magnetic resonance imaging, PET, other biological markers, and clinical and neuropsychological assessment can be combined to measure the progression of MCI and early AD. For up-to-date information, see www.adni-info.org. See supporting information, http://adni.loni.usc.edu, and previous reports^53-61 for more details. Written informed consent was obtained from all participants in ADNI.

To identify participants with HI, we searched the medical history database (“RECMHIST”) for entries including the key words/phrases “hit head,” “TBI,” “traumatic brain injury,” “head injury,” and “concussion,” as described previously.²² These search terms returned 159 instances of HI from a total of 102 participants. 57 of the reported injuries were the result of participants reporting more than one HI. There were therefore 102 participants who self-reported HI history. The most common causes of HI were sports/athletic activities, falls, and motor vehicle accidents. Two participants had no data past the screening visit and were removed, leaving 100 participants with HI in the sample. There were 2411 participants who did not report a HI, for a total of 2511 participants. The mean number of participants with at least one biomarker measurement for each biomarker modality was 1,178, and participants had an average of 2 measurements per biomarker (Supplemental Table 1). Follow-up for all biomarkers was calculated according to the participant’s first-ever study visit. For example, if a participant had only CSF data collected at baseline, followed by both CSF and blood plasma at 36 months, the blood plasma data would be recorded as the 36-month measurement. Further, the sample included participants from ADNI 1, ADNI GO, ADNI 2, and ADNI 3.

Participants were grouped based on HI history (1 = history of HI, 2 = no HI) and cognitive status (1 = cognitively normal (CN), 2 = cognitively impaired). In the ADNI study, a clinical diagnosis/cognitive status is obtained based on cognitive symptoms, cognitive scores, and clinical assessment.⁶² For this study, participants with subjective cognitive decline/complaints were included in the CN group. Patients with a diagnosis of MCI or AD were included in the cognitively impaired group. As cognitive status could change between study visits, this grouping was repeated for every study visit. Due to limited sample size particularly in the HI group, we did not conduct additional stratification between impaired participants (e.g., mild cognitive impairment (MCI) versus AD).

Biomarker assays

Biomarker collection and sample processing procedures have been described previously.⁵⁵ From cerebrospinal fluid samples, we analyzed data from the Roche Elecsys immunoassays of Aβ₄₀, Aβ₄₂, total tau, and pTau-181. We also calculated and analyzed the Aβ_42/40 ratio from these data. In blood plasma, we analyzed the raw Aβ_42/40 ratio calculated by the Bateman group,⁶³ as well as the Simoa assays of pTau-181 and NfL.^64,65 One outlying data point, defined as a data point falling more than three standard deviations from the mean of all other data points, was removed for the analysis of plasma pTau-181.

Data organization

Other potential risk factors included common demographic variables (sex, race, ethnicity, years of education, and APOE genotype (ϵ4 carrier versus non-carrier)). We calculated participants’ precise age at each study visit by subtracting the date of the visit from participants’ date of birth and dividing it by 365.25. As ADNI reports only birth month and year, age may be miscalculated by ∼30 days. If there was no recorded date for a visit, any biomarker measurements from that visit were excluded (Supplemental Table 2). If a study visit had biomarker measurements but no cognitive status, we inferred cognitive status based on previous and subsequent visits (if available). 27 cognitive statuses were inferred (Supplemental Table 3), none of which represented a diagnostic conversion (e.g., the participant was either previously CN and remained as such or was impaired and remained as such). Biomarker measurements outside the limit of detection (those that contained < or >) were also removed (Supplemental Table 4).

Statistical analyses

Demographic and neuropsychological variables were compared between groups via Chi-square tests (for dichotomous variables) or two-way ANCOVAs using SPSS 29.0.1.0. Age, sex, and years of education were included as covariates where appropriate. A p-value < 0.05 was considered significant for all comparisons. Cross-sectional analyses were performed using two-way ANCOVAs where biomarker measures were dependent variables and HI and cognitive status, and the interaction between HI and cognitive status, were independent variables. As the data were not normally distributed for all biomarkers except the CSF and plasma Aβ_42/40 ratios (Supplemental Figure 1), a log transformation was performed (Supplemental Figure 2). Covariates included age, sex, race, ethnicity, education, and APOE ϵ4 carrier status. For longitudinal analyses, random slope and random intercept linear mixed effects models (LMEMs) were generated via the nlme package in R 2023.12.0.⁶⁶ The biomarker measurement was the dependent variable, and fixed effects were age, HI (yes or no), and the interaction between age and HI. Covariates included cognitive status, sex, race, ethnicity, years of education, and APOE genotype (ϵ4 carrier versus non-carrier). Random effects were age and participant ID number (RID). LMEMS were optimal for this analysis due to their robust nature in the presence of missing data and their flexibility to incorporate multiple fixed effects, random effects, and covariates.^67,68 Post-hoc tests for significant main/interaction effects or covariates were performed using the emmeans() function with Tukey adjustment or via separate linear regressions outside the mixed effects model for continuous variables.

Results

Demographics, clinical measures, and cognitive performance

Table 1 presents the baseline demographics and neuropsychological performance of the included participants. Impaired participants had fewer years of education than CN participants (p = 0.015), higher frequency of APOE ϵ4 positivity (p < 0.001) and performed worse on all baseline cognitive measures than CN participants (p < 0.001 for all cognitive tests) (Table 1). Sex and ethnicity (proportion of non-Hispanic white participants) were significantly different between groups (Table 1). The CN-non-HI group had a lower proportion of non-Hispanic white participants. The impaired groups had lower proportions of female participants than expected, while the CN-non-HI group had a higher proportion of female participants. There were no other significant differences based on HI history, nor any significant interactions between cognitive status and HI history (Table 1). The average number of study visits for participants with HI was 6.96 (SD +/− 4.41), and the average number of study visits for participants without HI was 5.45 (SD +/− 3.55).

Table1.

Baseline demographics and neuropsychological performance of included participants (HI: head injury; DX: diagnosis; SD: standard deviation; SE: standard error; n: number of participants; APOE: Apolipoprotein E; CN: cognitively normal; SCD: subjective cognitive decline; EMCI: early mild cognitive impairment; LMCI: late mild cognitive impairment; AD: Alzheimer’s disease; CDR: Clinical Dementia Rating; MoCA; Montreal Cognitive Assessment; MMSE: Mini-Mental State Examination; RAVLT: Rey’s Auditory Verbal Learning Test; ANOVA: analysis of variance).

	Head injury (n = 100)		No head injury (n = 2411)		Total n participants	HI p-value	DX p-value	DX by HI p-value
Baseline characteristics	CN (n = 33)	Cognitively impaired (n = 67)	CN (n = 896)	Cognitively impaired (n = 1515)		-	-	-
Age (SD)	71.96 (6.57)	72.40 (7.55)	72.23 (6.71)	73.45 (7.79)		0.414	0.300	0.624
Years of education (SE)	16.85 (2.65)	16.16 (2.97)	16.49 (2.50)	15.75 (2.79)		0.186	0.015	0.934
% female (n)*	42.4 (14)	29.9 (20)	57.5 (5 15)	42.9 (650)		<0.001 (Pearson Chi-Square = 57.792, likelihood ratio = 58.181)
% Non-Hispanic White (n)*	97.0 (32)	100.0 (67)	93.5 (832)	95.8 (1443)	2497 (99.44%), 14 missing (6 CN-non-HI, 8 Impaired-non-HI)	0.032 (Pearson Chi-Square = 8.800, likelihood ratio = 11.814)
% APOE ϵ4 positive (n)*	31.0 (9)	51.7 (30)	30.5 (236)	54.7 (740)	2215 (88.21%), 296 missing (4 CN-HI, 9 Impaired-HI, 122 CN-non-HI, 161 Impaired-non-HI)	<0.001 (Pearson Chi-Square = 119.531, likelihood ratio = 122.000)
Diagnostic status (CN, SCD, EMCI, LMCI, AD)*	23 CN, 10 SCD	26 EMCI, 31 LMCI, 10 AD	534 CN, 360 SCD	424 EMCI, 665 LMCI, 415 AD	2498 (99.48%), 13 missing (2 CN-non-HI, 9 Impaired-non-HI)	<0.001 (Pearson Chi-Square = 2511.710, likelihood ratio = 3303.266)
CDR – Global (SE)+	0.00 (0.00)	0.54 (0.17)	0.002 (0.03)	0.57 (0.20)	2510 (99.96%), 1 missing (Impaired-non-HI)	0.540	<0.001	0.496
CDR – Sum of Boxes (SE)+	0.02 (0.09)	1.93 (1.63)	0.04 (0.16)	2.31 (1.76)	2510 (99.96%), 1 missing (Impaired-non-HI)	0.262	<0.001	0.269
MoCA Total Score (SE)+	26.41 (1.97)	21.77 (3.89)	25.92 (2.59)	21.78 (4.40)	1585 (63.12%), 926 missing (16 CN-HI, 23 Impaired-HI, 248 CN-non-HI, 639 Impaired-non-HI)	0.992	<0.001	0.792
MMSE Total Score (SE)+	28.97 (0.81)	27.18 (2.68)	29.07 (1.13)	26.38 (2.79)	2480 (98.77%), 31 missing (14 CN-non-HI, 17 Impaired-non-HI)	0.270	<0.001	0.077
Trails Making A [seconds] (SE)+	34.27 (10.95)	42.77 (23.98)	33.34 (11.77)	47.92 (27.42)	2398 (95.50%), 113 missing (3 CN-HI, 5 Impaired-HI, 29 CN-non-HI, 76 Impaired-non-HI)	0.480	<0.001	0.265
Trails Making B [seconds] (SE)h+	85.13 (44.95)	123.92 (78.21)	80.78 (39.90)	136.31 (80.09)	2353 (93.71%), 158 missing (3 CN-HI, 7 Impaired-HI, 32 CN-non-HI, 116 Impaired-non-HI)	0.874	<0.001	0.316
Animal Fluency Score (SE)+	22.07 (5.56)	17.03 (5.02)	20.98 (5.40)	15.88 (5.60)	2407 (95.86%), 104 missing (3 CN-HI, 5 Impaired-HI, 29 CN-non-HI, 67 Impaired-non-HI)	0.135	<0.001	0.970
RAVLT – Immediate Recall (SE)+	46.87 (10.12)	32.00 (11.64)	45.64 (10.12)	31.34 (11.07)	2402 (95.66%), 109 missing (3 CN-HI, 5 Impaired-HI, 32 CN-non-HI, 69 Impaired-non-HI)	0.285	<0.001	0.655
RAVLT – Delayed Recall (SE)+	8.07 (3.67)	3.37 (3.46)	7.79 (4.00)	3.04 (3.71)	2405 (95.78%), 106 missing (3 CN-HI, 5 Impaired-HI, 31 CN-non-HI, 67 Impaired-non-HI)	0.420	<0.001	0.947

^*Chi-square test.

⁺Age, sex, and years of education covaried.

Baseline cross-sectional biomarker analyses

There were no significant differences between HI and non-HI participants for any biomarker (panel a of Figures 1, 2 and 4-8) besides CSF Aβ_42/40 ratios, where HI participants had higher ratios than non-HI participants (p = 0.004) (Figure 3(a)) (Supplemental Tables 5-7). CN participants had higher CSF Aβ₄₂ levels (p < 0.001) and CSF Aβ_42/40 ratios (p = 0.004) than impaired participants, while impaired participants had higher CSF pTau-181, CSF total tau, and plasma NfL levels (p < 0.001 for all comparisons). Age was a significant covariate for all biomarkers except plasma Aβ_42/40 ratios. Sex was a significant covariate for CSF Aβ₄₀, CSF pTau-181, CSF total tau, plasma NfL, and plasma pTau-181. APOE ϵ4 status was a significant covariate for CSF Aβ₄₂, CSF Aβ_42/40, CSF pTau-181, CSF total tau, and plasma pTau-181. Education was a significant covariate for CSF pTau-181, CSF total tau, and plasma pTau-181. Ethnicity was a significant covariate for plasma Aβ_42/40 ratios. All log-transformed biomarkers had similar results (Supplemental Table 6).

Figure 1.

Cerebrospinal fluid (CSF) amyloid-β₄₀ (Aβ₄₀) analyzed cross-sectionally at baseline (a) and longitudinally (b) (red/left = HI, blue/right = non-HI; red crossbar = mean). There were 568 participants with at least one measurement, the average number of measurements per participant was 1.62, and the total number of datapoints was 921.

Figure 2.

Cerebrospinal fluid (CSF) amyloid-β₄₂ (Aβ₄₂) analyzed cross-sectionally at baseline (a) and longitudinally (b) (red/left = HI, blue/right = non-HI; red crossbar = mean). There were 1660 participants with at least one measurement, the average number of measurements per participant was 1.89, and the total number of datapoints was 3153.

Figure 4.

Cerebrospinal fluid (CSF) phosphorylated tau (pTau)-181 ratios analyzed cross-sectionally at baseline (a) and longitudinally (b) (red/left = HI, blue/right = non-HI; red crossbar = mean). There were 1650 participants with at least one measurement, the average number of measurements per participant was 1.89, and the total number of datapoints was 3133.

Figure 8.

Blood plasma phosphorylated tau (pTau)-181 analyzed cross-sectionally at baseline (a) and longitudinally (b) (red/left = HI, blue/right = non-HI; red crossbar = mean). There were 1190 participants with at least one measurement, the average number of measurements per participant was 3.11, and the total number of datapoints was 3700.

Figure 3.

Cerebrospinal fluid (CSF) amyloid-β₄₂/amyloid-β₄₀ (Aβ_42/40) ratios analyzed cross-sectionally at baseline (a) and longitudinally (b) (red/left = HI, blue/right = non-HI; red crossbar = mean). There were 568 participants with at least one measurement, the average number of measurements per participant was 1.62, and the total number of datapoints was 918.

As we observed that sex was a significant covariate for multiple biomarkers, we performed two-way ANCOVAs that tested biomarker levels as function of HI and sex with covariates of age, cognition, race, ethnicity, education, and APOE ϵ4 carriership (Supplemental Table 7). Female participants had higher CSF pTau-181 (p = 0.01) and CSF total tau (p = 0.005) than male participants, but there were no other significant differences by sex, and there were no significant interactions between HI and sex.

Longitudinal biomarker analyses

There were no main effects of HI history, nor any interactions between HI history and age, for any biomarker (panel b of Figures 1-8) (Supplemental Tables 8-17, Supplemental Figure 3). Age was positively associated with CSF pTau-181, CSF total tau, plasma NfL, and plasma pTau-181 levels (p < 0.001 for all comparisons), and negatively associated with CSF Aβ₄₂ levels and CSF Aβ_42/40 ratios (p < 0.001 for all comparisons). CN participants had higher CSF Aβ₄₀ (p = 0.0734), CSF Aβ₄₂ (p < 0.0001), and CSF Aβ_42/40 (p < 0.0001) ratios than impaired participants, while impaired participants had higher CSF pTau-181, CSF total tau, plasma NfL, and plasma pTau-181 than CN participants (p < 0.0001 for all comparisons). Female participants had higher CSF Aβ₄₀ (p = 0.0314), CSF Aβ₄₂ (p = 0.0156), CSF pTau-181 (p = 0.0057), CSF total tau (p = 0.0001), and plasma NfL (p = 0.0008) than male participants, while male participants had higher plasma pTau-181 (p = 0.0191) than female participants. Non-Hispanic/Latino participants trended towards higher CSF pTau-181 (p = 0.0786) and CSF total tau (p = 0.0871) than Hispanic/Latino participants, while Hispanic/Latino participants had higher plasma Aβ_42/40 ratios (p = 0.0086) than non-Hispanic/Latino participants. Educational attainment was negatively associated with CSF pTau-181 (p < 0.001) and CSF total tau (p = <0.001). APOE ϵ4 noncarriers had higher CSF Aβ₄₂ (p < 0.0001) and CSF Aβ_42/40 ratios (p < 0.0001) than ϵ4 carriers, while ϵ4 carriers had higher CSF pTau-181 (p < 0.0001), CSF total tau (p < 0.0001), plasma NfL (p = 0.0006), and plasma pTau-181 (p < 0.0001).

As we observed that sex was a significant covariate for multiple biomarkers, we ran additional linear mixed effects models to test for significant interactions of HI and sex. Age, HI, and sex were fixed effects with covariates of cognition, race, ethnicity, years of education, and APOE ϵ4 carriership (Supplemental Table 9). Random effects were age and participant ID number (RID). There was an interaction between age and sex only on plasma NfL levels (p = 0.0355), but there were no significant interactions between HI and sex on biomarker levels, nor were there any three-way interactions between age, HI, and sex.

Discussion

Our findings do not support an association between AD biofluid biomarkers and self-reported history of HI in aging participants from the ADNI cohort. Consistent main effects of age and cognitive impairment indicated a pattern of worsened AD-like biomarker signatures in impaired participants with increasing age.

Previous studies have shown that biofluid Aβ levels drop in AD,³⁷ whereas phosphorylated tau isoforms, including pTau-181, increase.^69,70 NfL levels also tend to increase in AD, though NfL is not an AD-specific marker.^25,26 In our analyses, age was positively associated with CSF pTau-181, CSF total tau, plasma NfL, and plasma pTau-181 levels, but negatively associated with CSF Aβ₄₂ and CSF Aβ_42/40 ratios. Additionally, CN participants had higher CSF Aβ₄₀, CSF Aβ₄₂, and CSF Aβ_42/40 ratios than impaired participants, while impaired participants had higher CSF pTau-181, CSF total tau, plasma NfL, and plasma pTau-181 than CN participants. These findings are concordant with prior literature and indicate a pattern of worsened AD-like biomarker signatures with increasing age and in cognitively impaired participants relative to those without impairment.

We only observed one significant difference in biomarker levels based on HI, where HI participants had higher CSF Aβ_42/40 ratios than non-HI participants. This finding was in the opposite direction that we anticipated, as higher Aβ_42/40 ratios are indicative of less AD pathology, suggesting HI participants have lower deposited amyloid than participants without HI. However, as the remainder of the analyses were nonsignificant, our findings do not support a main effect of HI on AD fluid biomarkers. In the context of prior literature, this result is relatively unsurprising. One study examining plasma biomarkers in veterans with history of concussion also did not observe any changes in plasma amyloid levels,⁵³ and numerous reports found no association between cognition and plasma tau in individuals with HI history.^71-73 Oppositely, some studies observed alterations to fluid pTau levels in participants with history of concussion^4474–76 but those samples included more severe injuries⁷⁴ or former professional athletes who experienced repetitive concussions during their careers.^44,75 This raises the question of whether elevated pTau in individuals with history of repetitive injuries may be more likely to indicate chronic traumatic encephalopathy (CTE) as opposed to AD.^77,78 Alternatively, as numerous studies have noted associations between HI and elevated amyloid and/or tau deposition on PET,^{22,26,2779–82} HI may be associated with an increase in cortical amyloid and/or tau deposition during the post-injury period.^10,2583–87 Similarly, there is mixed data from sports concussion studies regarding whether NfL levels are altered either during acute recovery from concussion, or in the years following concussion. One study found no evidence for NfL changes in the hours/days after concussion⁴⁸; however, other studies suggested NfL is elevated in the immediate post-injury period⁵² and may remain elevated even 5 years after injury.⁵¹ However, two reports that examined the association of TBI with NfL studied professional athletes with high risk for repetitive concussion, which again raises the question of whether elevated NfL in these populations is related to CTE pathology rather than AD,^51,52 and also whether a single remote incidence of HI would be sufficient to result in chronic changes to NfL.

In our initial analyses where participant sex was included as a covariate, we observed that sex was a significant covariate for multiple biomarkers both at baseline and upon longitudinal analysis. Specifically, female participants had higher longitudinal CSF Aβ₄₀, Aβ₄₂, pTau-181, total tau, and plasma NfL, while male participants had higher plasma pTau-181. However, when we performed subsequent analyses testing for main effects of sex in combination with HI on biomarker levels both cross-sectionally and longitudinally, we found no significant interactions between sex and HI for any biomarker. Extensive literature suggests women are at higher risk for AD,^88-90 though the presence of sex differences in biofluid and imaging-based AD biomarkers is not definitive.⁹⁰ Our results are similarly unclear, as female participants had more favorable CSF Aβ levels, but less favorable CSF tau levels. Additionally, male participants had less favorable plasma pTau levels compared to female participants. Interpretation of these findings is challenging, as there is no clear pattern to implicate either female or male sex as being associated with poorer AD biofluid biomarker profiles in this analysis.

Prior literature has also suggested educational attainment may be protective against AD (see⁹¹ for review). Our results indicated that educational attainment was negatively associated with both CSF pTau-181 and total tau levels, which is concurrent with prior reports of a potential protective effect of educational attainment against AD biomarkers. However, no such relationship between educational attainment and biomarker levels was observed for any Aβ biomarker, nor for any plasma biomarker. Hence, our data suggest higher educational attainment may be associated with lower CSF tau, but our findings do not definitively support a protective effect of educational attainment on all AD biomarkers.

We also observed associations between ethnicity and AD biofluid biomarkers. Non-Hispanic/Latino participants trended towards higher CSF pTau-181 and total tau, while Hispanic/Latino participants had higher plasma Aβ_42/40 ratios. This pattern suggests non-Hispanic/Latino participants have worsened AD biomarker signatures compared to those of Hispanic/Latino backgrounds. There have been few reports investigating AD fluid biomarkers in Hispanic/Latino populations with which to compare our findings, but work from the Health & Aging Brain Study—Health Disparities (HABS-HD) cohort⁹² has suggested AD biomarkers do indeed differ between Mexican-Americans and non-Hispanic Whites.^92-96 However, much more work is needed to characterize AD fluid biomarkers in diverse populations to establish whether true differences exist between participants of different ethnic backgrounds, as in our study the Hispanic population did not solely contain Mexican-American individuals.

Similarly, additional research into the temporal kinetics of biofluid biomarker levels after brain injury will be necessary to clarify if/how biomarkers change in response to HI, how long these changes persist, and their implications for long-term AD/dementia risk.^49,97 For example, levels of GFAP, NfL, and UCH-L1 may be increased in athletes following concussion,^{46,48,49,51,52,7597–99} but it is unclear how long these biomarkers remain elevated before returning to pre-injury levels. Our findings indicate that AD-specific biomarkers may not be significantly elevated in adults with far-remote injuries. Additionally, future studies specifically examining important covariates such as sex, educational attainment, and ethnicity will also be necessary to clarify whether these covariates are associated with any/all biofluid biomarkers in the context of HI.

Limitations and future directions

Our study has a number of limitations, chief of which is low power due to limited availability of participants with HI. The ADNI medical history battery does not ask about HI history, so probable underreporting of HI incidence likely limited the number of participants available for analysis. This is evidenced by prior work that indicates up to 25% of American adults may have experienced a concussion in their lifetime, whereas only approximately 4% of participants in ADNI self-reported history of HI.¹⁰⁰ An additional consequence of not directly asking about HI is a lack of clinical detail if a participant does report HI. Participants may not offer information about when the HI occurred, the mechanism of injury, or whether consciousness was lost. This precludes analysis of time since injury or mechanism of injury as a factor, and participants also cannot be stratified based on loss of consciousness as an indicator of injury severity. As the severity of an injury could plausibly affect the extent and/or duration of any biomarker changes, this lack of clinical detail is a hindrance towards improving our understanding of the association of HI with dementia biomarkers.¹¹ Furthermore, reliance on self-report of HI introduces the possibility of recall bias, as participants who have experienced HI but did not volunteer that information would erroneously be included in the non-HI group. In the future, this analysis should be replicated using participants from datasets with better HI ascertainment, like the Indiana Memory and Aging Study (IMAS).

Another limitation is the lack of extensive longitudinal data for some participants. Future replications using datasets with greater longitudinal follow-up will be beneficial. Furthermore, lack of diversity in the cohort is a limitation. As ADNI is comprised mainly of participants who are highly educated and of European descent, our findings may not generalize to other populations.

Furthermore, it is possible that there are differences between participants who remain in the study for an extended period and those who leave the study early due to mortality, voluntary withdrawal, or some other reason. Such potential group differences may bias the results, and future studies with larger longitudinal samples and consistent biomarker measurements are warranted. In a separate sensitivity analysis comparing participants with only a single baseline biomarker measurement to those with two or more measurements, some differences were observed in the proportion of participants who were cognitively normal versus impaired, as well as the proportion of participants with and without HI. However, the differences were inconsistent. For example, the group with two or more measurements had a higher proportion of cognitively normal individuals for baseline analyses of CSF Aβ₄₀, the CSF Aβ_42/40 ratio, CSF total tau, and plasma NfL. However, the group with only one measurement had a higher proportion of cognitively normal participants for baseline analyses of CSF Aβ₄₂, CSF pTau, and plasma pTau. The group with two or more measurements also had a higher proportion of HI participants than the single measurement group for baseline analyses of plasma pTau and NfL. Given these mixed findings, it is challenging to estimate the level of bias that differences due to drop out may have had on our findings. As we did not observe significant relationships between HI and biomarkers in our principal analysis, these group differences are not likely causing false positivity in this sample. However, the participants may have other fundamental differences that the data cannot capture, and this is a limitation of our analysis.

Finally, though studies from sports medicine have been helpful in illuminating acute post-concussion changes in relevant fluid biomarker levels, the majority of reports suggest that fluid biomarker levels return to their pre-injury baseline in a matter of days to weeks.^{46,48,49,51,52,7597–99} The time elapsed between HI and biomarker data capture in the participants included in our analysis is therefore a significant limitation of our study, as it is plausible that no biofluid biomarkers remain elevated long enough to be captured in aging cohorts such as ADNI. This may represent a potential explanation for why we did not observe significant changes in biomarker levels between participants with and without HI. Future studies measuring fluid biomarkers in older adults immediately following incidents of HI, such as those incurred by falls, could prove useful in illuminating the temporal dynamics of fluid biomarker changes after HI and their relevance to risk for developing dementia.

Conclusion

We did not find strong evidence to suggest history of HI is associated with AD fluid biomarkers in aging participants from the ADNI cohort. In line with prior literature, we consistently observed that both increasing age and cognitive impairment were associated with poorer biomarker profiles. We also noted intriguing associations between sex, ethnicity, and educational attainment on biomarker levels. However, further research is needed to more clearly determine whether remote history of HI is related to fluid biomarkers of AD.

Supplementary Material

Supplemental material for this article is available online.

Figure 5.

Cerebrospinal fluid (CSF) total tau analyzed cross-sectionally at baseline (a) and longitudinally (b) (red/left = HI, blue/right = non-HI; red crossbar = mean). There were 1653 participants with at least one measurement, the average number of measurements per participant was 1.89, and the total number of datapoints was 3145.

Figure 6.

Blood plasma amyloid-β₄₂/amyloid-β₄₀ (Aβ_42/40) ratios analyzed cross-sectionally at baseline (a) and longitudinally (b) (red/left = HI, blue/right = non-HI; red crossbar = mean). There were 550 participants with at least one measurement, the average number of measurements per participant was 1.52, and the total number of datapoints was 836.

Figure 7.

Blood plasma neurofilament light chain (NfL) analyzed cross-sectionally at baseline (a) and longitudinally (b) (red/left = HI, blue/right = non-HI; red crossbar = mean). There were 1583 participants with at least one measurement, the average number of measurements per participant was 2.70, and the total number of datapoints was 4278.

Data availability statement

Data used in this project are owned by the Alzheimer’s Disease Neuroimaging Initiative (ADNI). Data are publicly and freely available on the ADNI website (https://adni.loni.usc.edu/data-samples/adni-data/#AccessData) upon sending a request to the ADNI Data Sharing and Publications Committee that includes the investigator’s institutional affiliation and the proposed uses of the ADNI data.