Accelerated Brain Aging in Mild Traumatic Brain Injury: Longitudinal Pattern Recognition with White Matter Integrity

Abstract

Mild traumatic brain injury (mTBI) initiating long-term effects on white matter integrity resembles brain-aging changes, implying an aging process accelerated by mTBI. This longitudinal study aims to investigate the mTBI-induced acceleration of the brain-aging process by developing a neuroimaging model to predict brain age. The brain-age prediction model was defined using relevance vector regression based on fractional anisotropy from diffusion tensor imaging of 523 healthy individuals. The model was used to estimate the brain-predicted age difference (brain-PAD) between the chronological and estimated brain age in 116 acute mTBI patients and 63 healthy controls. Fifty patients were followed for 6 ∼ 12 months to evaluate the longitudinal changes in brain-PAD. We investigated whether brain-PAD was greater in patients of older age, post-concussion complaints, and apolipoprotein E (APOE) ɛ4 genotype, and whether it had the potential to predict neuropsychological outcomes. The brain-age prediction model predicted brain age accurately (r = 0.96). The brains of mTBI patients in the acute phase were estimated to be “older,” with greater brain-PAD (2.59 ± 5.97 years) than the healthy controls (0.12 ± 3.19 years) (p < 0.05), and remained stable 6–12 month post-injury (2.50 ± 4.54 years). Patients who were older or who had post-concussion complaints, rather than APOE ɛ4 genotype, had greater brain-PADs (p < 0.001, p = 0.024). Additionally, brain-PAD in the acute phase predicted information processing speed at the 6 ∼ 12 month follow-up (r = -0.36, p = 0.01). In conclusion, mTBI accelerates the brain-aging process, and brain-PAD may be capable of evaluating aging-associated issues post-injury, such as increased risks of neurodegeneration.

Introduction

Mild traumatic brain injury (mTBI), accounting for 80 ∼ 90% of all TBI,¹ is typically associated with white matter integrity loss that persists chronically after the initial injury.² The long-term sequelae of mTBI initiate risks for neurodegeneration and age-associated conditions,³ making it a clinically important topic. A large-scale (n = 2,794,852) observational study with a 15-year follow-up showed that chronic mTBI was associated with a higher risk of dementia (hazard ratio: 1.17).⁴ However, neuroimaging biomarkers that evaluate individual-based risks of age-associated or neurodegenerative deterioration in the brain for mTBI patients are still scarce.

MRI metrics strongly relate to chronological age,^5,6 making brain age predictable. The brain-predicted age difference (brain-PAD = predicted brain age - chronological age) reflects how far a brain deviates from the normal aging trajectory under a given disease.^7

–10 Increases in brain-PAD indicate that the brain-aging process accelerated and imply the risks of neurodegeneration and other age-related issues,¹¹ which have been widely revealed in psychiatric disorders and neurological diseases,^12

–15 including TBI. Recently, Cole and coworkers observed a 4.66-year increase in brain-PAD estimated by morphometrics in white matter and a 5.97-year increase estimated by gray matter morphometrics in chronic moderate/severe TBI patients compared with mild cases.¹⁵ However, these results provide limited understanding of how mTBI accelerates the brain-aging process, because only 17 patients with mTBI were included in these authors' research. The majority of TBIs are within the mild range, and sustain insults from diffuse axonal injury.¹⁶ They usually manifest as a brain-wide distribution of abnormalities on diffusion tensor imaging (DTI) metrics, such as fractional anisotropy (FA) value decreases.^17,18 Such deterioration in white matter integrity is thought to be responsible for further neurodegenerative and aging-related issues.^19
–21 Therefore, we hypothesized that mTBI might accelerate the aging process by decreased integrity in white matter microstructure.

Herein, relevance vector regression²² was applied to build the brain-age prediction model based on DTI metrics to estimate brain age in mTBI patients and healthy controls (HC). The patients were followed into their chronic phase to evaluate the longitudinal changes in brain-PAD. We also investigated whether brain-PAD was prone to be greater in patients of older age, post-concussion complaints, and apolipoprotein E4 genotype,^3,23
–25 and whether it could predict neuropsychological outcomes. In addition, we hypothesized that white matter tracts contributing to brain-PAD in mTBI patients were primarily located in degenerative-vulnerable regions.

Methods

Participants

Training set

DTI scans from 523 anonymized healthy individuals were used to build the brain-age prediction model (details in Table S1). Exclusion criteria for DTI scans included poor quality leading to image processing failure. Ethical approval and informed consent were obtained from each original study.

Test sets

Ethics approval was obtained from the Second Affiliated Hospital of Wenzhou Medical University, and data collection was conducted between May 2016 and April 2018. The flow of participants through the study is shown in Figure 1. There were 116 acute mTBI patients with negative head computed tomography (CT) scans and Glasgow Coma Scale (GCS) scores of 15 recruited from the local emergency department. Sixty-four patients consented to participate in the follow-up investigations, and 50 of them were ultimately followed up 6 ∼ 12 month post-injury into their chronic phase. Screening for mTBI was based on the World Health Organization's Collaborating Centre for Neurotrauma Task Force ²⁶ (details of inclusion and exclusion criteria listed in Table 1). The retrospective baseline health conditions of the patients were recorded at their first visit. HCs (n = 63) were recruited through advertisements and carefully screened for any neurological or psychiatric disorders. In addition, an independent cohort including 70 patients with acute mTBIs, with imaging from a different magnetic resonance imaging (MRI) scanner was included for replication use (details in Supplementary Methods). All subjects gave written, informed consent in person, and the study was approved by the local institutional review board and conducted in accordance with the Declaration of Helsinki.

FIG. 1.

Flow diagram shows the inclusion process for participants of test sets. brain-PAD, brain-predicted age difference; mTBI, mild traumatic brain injury; HCs, healthy controls; DTI, diffusion tensor imaging.

Table 1.

The Inclusion and Exclusion Criteria for mTBI

Inclusion and exclusion criteria	Items
Inclusion criteria	1. Glasgow Coma Score of 13–15
	2. One or more of the following:a. Loss of consciousness (if present) <30 min;
	b. Post-traumatic amnesia (if present) <24 h;
	c. Other transient neurological abnormalities such as focal signs and seizure
	3. No contraindications to MRI
Exclusion criteria
	1. History of neurological disease
	2. History of psychiatric disorder
	3. History of brain injury
	4. History of substance or alcohol abuse
	5. History of intubation and/or presence of a skull fracture and administration of sedatives on arrival in the emergency department
	6. Spinal cord injury
	7. Manifestation of mTBI for other reasons:a. Because of medications for other injuries (e.g., systemic injuries, facial injuries, or intubation);b. Caused by other problems (e.g., psychological trauma, language barrier, or coexisting medical conditions)c. Caused by penetrating craniocerebral injury

mTBI, mild traumatic brain injury; MRI, magnetic resonance imaging.

Clinical symptom and neuropsychological assessments

Clinical symptoms were assessed at the first patient visit (n = 116). The post-concussive symptoms (PCS) were evaluated by the International Classification of Diseases, 10th edition (ICD-10) clinical criteria.²⁷ Patients with three or more symptoms were classified as having PCS (PCS+) or not having (PCS-). According to the anterograde amnesia (AA) duration, patients were divided into those with AA (AA+) and those without AA (AA-).

The information processing speed (IPS) was assessed by the Digit Symbol Coding test²⁸ and the Trail-Making Test (Part A)²⁹ for HCs and patients at both the first and follow-up visits. The Z-score calculated from these two tests summarized IPS performance.

Apolipoprotein E (APOE) genotyping

Two milliliters of peripheral blood was collected from each of 81 mTBI patients for APOE genotype analysis. DNA was extracted from peripheral blood samples using Omega D3494-01 Blood DNA Midi Kit (Omega Bio-tek, Inc., United States). Two single-nucleotide polymorphisms (SNPs) (rs429358 and rs7412) were genotyped to identify APOE genotypes composed of the APOE-ɛ2, -ɛ3, and -ɛ4 alleles using Assay Design Suite v2.0 (Agena Bioscience, Inc., San Diego, CA). All genotypes containing the ɛ4 allele (ɛ4/ɛ4; ɛ4/ɛ3; ɛ4/ɛ2) were combined as the composite APOE ɛ4+, and others (ɛ2/ɛ2; ɛ2/ɛ3; ɛ3/ɛ3) were combined as the composite APOE ɛ4-.

Neuroimaging data acquisition and pre-processing

The training data sets were collected from multi-centers, by using different MRI scanners (Table S1). The intraclass correlation coefficient (ICC) as the homogeneity index for training data sets was calculated on FA after pre-processing, showing “excellent” homogeneity (ICC = 0.84) for multi-center data sets.^30,31

The DTI scans for test sets were acquired from a 3T MRI scanner (GE750) by a single-shot, spin echo-based and diffusion-weighted echo planar imaging sequence (repetition time [TR] = 8000 ms, echo time [TE] = 68 ms, flip angle = 90 degrees, thickness = 2 mm, slices = 75, field of view [FOV] = 256 mm × 256 mm, matrix size = 128 × 128, two averages, voxel size = 2 mm × 2 mm × 2 mm). DTI scans (b = 1000 sec/mm²) were acquired with 30 diffusion gradient directions, and diffusion imaging without weighting (b = 0) was repeated five times.

All raw DTI data were quality controlled and preprocessed using the FSL5.0.9 software package (http://www.fmrib.ox.ac.uk/fsl/index.html) following the standard procedure³² Briefly, the DICOM files were converted into NIFTI images first. Skull and non-brain tissues were removed by using the BET tool in FSL. After correcting eddy current distortion and head motion artifacts, the native FA image was estimated by fitting the diffusion tensor model at each voxel. The FA image was then registered to FA template (FMRIB58_FA) with 1 mm isotropic resolution by using FNIRT. This FA template was derived from HCs in the Montreal Neurological Institute 152 (MNI152) standard space. Fifty white matter regions were selected based on the rICBM_DTI_81_WMPM_FMRIB58 1 mm atlas; subsequently, the mean FA value of each tract was calculated for feature selection.

Brain-age prediction modeling procedure

The brain-age prediction was performed by defining a relevance vector regression (RVR) to model the relationship between FA values and chronological age in the training set, using Scikit-learn Python library. RVR derives from the Relevance Vector Machine based on Bayesian formulation, which extends linear regression models into probabilistic computation. Specifically, in the current study, the regression model for brain-age prediction was built by estimating the relevance vectors and their corresponding weight distributions through the type-II maximum likelihood algorithm. The relevance vectors were derived from the input data (i.e., FA values), which represented the prototypical examples associated with the regression target instead of solely separating attributes. The training process was visualized in Figure 2A–D (details in Supplementary Methods). The prediction accuracy was assessed using the Pearson correlation (r) between chronological age and predicted brain age, the proportion of the variation (R² ) was explained by the trained model, the mean absolute error (MAE), and root mean squared error (RMSE) of the brain-PAD scores.

FIG. 2.

Overview of brain-age prediction model. (A) The model was built in a given segment of the age span. (a) In the training set, the age span (11 ∼ 87 years) was partitioned into three segments (phase 1: 11 ∼ 32 years, n = 186; phase 2: 26 ∼ 48 years, n = 156; phase 3: 42 ∼ 87 years, n = 272). Two adjacent age phases in the training set were slightly overlapped (i.e., 6-year overlap) to enhance the tolerance toward the inter-individual variability of brain age prediction. (b) In the test sets, each age span segment corresponded with the age phase in the training set, in which the overlapped part of two adjacent phases was averaged (phase 1: 20 ∼ 29 years; phase 2: 29 ∼ 45 years; phase 3: > 45 years). (B) Feature selections contained two steps: (a) Pearson correlation coefficient (r) was calculated (p < 0.001) between the mean fractional anisotropy (FA) value of each WMT and age in each age segment of the training set, and WMT with the absolute value of r > 0.25 was kept. (b) Dimensionality of features was then reduced by principle components analysis (PCA). The components with the top 90% variance were kept as the input features. (C) The supervised learning process was conducted using relevance vector regression (RVR) to define the relationship between the chronological ages and input features. Internal validation was assessed by running 10-fold cross-validation on the shuffled training set with the same error calculation as the test set. (D) The trained RVR brain-age model was applied to all test sets. This panel is a schematic drawing showing the linear regression of individual chronological age and brain-predicted age in each test set. Color image is available online.

Statistical analysis

All statistical analyses were conducted in SPSS (version 21.0; IBM, Armonk, NY). To examine the increased brain age in the test sets, we compared the brain-PAD scores with zero using the Wilcoxon Signed-rank test. Between-group comparisons in brain-PAD scores for mTBI samples versus HCs were conducted by analysis of covariance (ANCOVA), adjusting for sex, education, and age. Longitudinal comparison analysis was performed on brain-PAD between the acute and chronic phases with linear mixed-effects (LME) model analysis. Between-group comparisons in brain-PAD scores for mTBI PCS subgroups (PCS+ vs. PCS-), APOE genotypes (APOE ɛ4+ vs. APOE ɛ4-), AA subgroups (AA+ vs. AA-), and chronological age subgroups (youth: ≤30 years, middle-aged: 30 ∼ 50 years, elderly: ≥50 years) were also conducted by ANCOVA, adjusting for sex, education, and/or age. Spearman rank-order correlation (r) was conducted to examine the associations of brain-PAD scores with the post-injury time and Z-score for IPS performance. For the mTBI cohort, the r between brain-PAD and FA values of white matter tracts (WMTs) involved in model training was calculated to evaluate the related WMTs contributing to increased brain age. The results were considered significant at p < 0.05 (two-tailed). Effect sizes for ANCOVA and LME analysis were quantified using partial η² ; effect sizes for the Wilcoxon signed-rank test were quantified by the rank biserial correlation coefficient (r_rb).

Results

Demographic and clinical characteristics of mTBI patients and HCs

Demographic and clinical information for the test sets is reported in Table 2. The mTBI cohort included 116 acute patients (58 males, 39.7 ± 11.9 years of age; education: 7.9 ± 3.9 years; first-visit post-injury: 3.5 ± 3.4 days; all presented as mean ± standard deviation [SD]). There were eight patients with a history of health issues (i.e., hyperglycemia, hypertension, and hyperlipidemia). Sixty-four patients consented to participate in the follow-up investigations and 50 of them (78.1%) were ultimately followed 6–12 months to their chronic phase (216.2 ± 117.8 days). Sixty-three HCs were enrolled (31 males, 36.7 ± 12.4 years of age; education: 10.9 ± 6.6 years). Patients showed worse performance on IPS tests than HCs (p < 0.001).

Table 2.

Demographic and Clinical Characteristics of mTBI Patients and Healthy Controls

Demographic and clinical characteristics	mTBI cohort(n = 116)	HCs(n = 63)	p value
Age (Years)	39.7 ± 11.9	36.7 ± 12.4	0.101
Sex (Male, female)	58, 58	31, 32	0.919
Education (Years)	7.9 ± 3.9	10.9 ± 6.6	0.001^*
Race
Asian-Chinese	100%	100%	-
First-visit post-injury (days)	3.5 ± 3.4	-	-
Injury cause
Traffic accident (%)	57.8	-	-
Fall (%)	20.7	-	-
Assault (%)	11.2	-	-
Other (%)	10.3	-	-
Injury severity
PCS (±)	91/25	-	-
Coma duration (min)	11.3 ± 10.8	-	-
AA duration (h)	0.4 ± 1.7	-	-
IPS DSCT TMT_A (sec)	31.3 ± 15.972.3 ± 49.8	45.7 ± 18.144.8 ± 30.9	< 0.001^* < 0.001^*

P values were derived from with the χ² test or Mann–Whitney U test, as appropriate.

P value indicates a significant difference.

mTBI, mild traumatic brain injury; HCs, healthy controls; PCS +, with post-concussion symptom complaints; PCS -, without post-concussion symptom complaints; AA duration, anterograde amnesia duration; IPS, information processing speed; DSCT, Digit Symbol Coding Test; TMT_A = Trail-Making Test Part A

Chronological age predicted by FA values

The RVR model could accurately predict individual chronological age for both the training set and the HC test set. For the training set, age was accurately predicted by FA values (r = 0.96, R ² = 0.93, MAE = 3.74, RMSE = 5.03). The mean PAD for the training group was -0.18 (± 5.03) years. For the HC test set, brain age was also accurately predicted (r = 0.97, R ² = 0.94, MAE = 2.57, RMSE = 3.16), and the mean PAD was 0.12 (± 3.19) years.

mTBI accelerated brain-aging process

The predicted brain age significantly correlated with chronological age in both HCs (R ² = 0.93, p < 0.001) and mTBIs (R ² = 0.87, p < 0.001) (Fig. 3A). mTBI patients had predicted brain age older than their chronological age in the acute phase (PAD = 2.59 ± 5.97 years; W = 4854, p < 0.001, r_rb = 0.431), which was replicated in an independent mTBI cohort (PAD = 3.26 ± 4.55 years; W = 2158, P < 0.001, r_rb = 0.737). Acute mTBI patients had higher brain-PAD scores (2.59 ± 5.97 years) than HCs (0.12 ± 3.19 years) (F _{1, 173} = 7.23, p = 0.008, partial η ² = 0.041; Fig. 3B). Additional analyses showed that the educational level and baseline health conditions of patients had no confounding effects on brain-PAD in the acute phase (Details in Supplementary Results).

FIG. 3.

Mild traumatic brain injury (mTBI) increased the brain-aging process. (A) Scatter plot and linear regression of individual chronological age and brain-predicted age in each test set derived from the relevance vector regression (RVR) prediction model. Points and triangles indicated mTBI individuals in the acute phase and healthy controls (HCs) respectively, and lines were regression lines for each group (mTBI = red; HC = green), with 95% confidence interval displayed. The regression lines of both chronological age and brain-predicted age were fitted well (HCs: R² = 0.94, p < 0.001; acute mTBI: R² = 0.87, p < 0.001). The solid black line (black) was the line of identity (y = x). (B) The cross-sectional comparison (with error bar) of brain-predicted age difference (brain-PAD) scores between acute mTBI patients and HCs. The brain-PAD scores in acute mTBI individuals (2.59 ± 5.97 years) were significantly higher than both zero (p < 0.001) and the brain-PAD scores in HCs (0.12 ± 3.19 years, p = 0.008). (C) The longitudinal comparison of brain-PAD scores between the acute and chronic phases in patients with mTBI. The brain-PAD scores in mTBI patients were still significantly higher than zero in their chronic phase (2.50 ± 4.54 years, p < 0.001), but not significantly higher than scores in the acute phase (1.87 ± 5.66 years, p = 0.51). Color image is available online.

mTBI patients had relatively stable brain-PADs (2.50 ± 4.54 years) in their chronic phase. Their brain-PAD scores (2.50 ± 4.54 years) in the chronic phase were also higher than those for HCs (0.12 ± 3.19 years) (F _1,108 = 8.16, p = 0.005, partial η ² = 0.07). Longitudinal comparison of brain-PADs between the two phases showed no statistical differences (F _{1, 112} = 0.436, p = 0.51; Fig. 3C). In addition, brain-PAD scores were not associated with post-injury days in either the acute or chronic phase (acute phase: r = -0.045, p = 0.634; chronic phase: r = -0.015, p = 0.920). Moreover, brain-PADs derived from the acute phase were inversely related to IPS performance in both the acute phase (r = -0.294, p = 0.001; Fig. 4A) and the 6–12 month follow-up (r = -0.356, p = 0.011; Fig. 4B).

FIG. 4.

Predicted age difference (PAD) in the acute phase predicted neuropsychological outcomes. (A) In the acute phase, the increased brain-PAD related to the decreased information processing speed (IPS) (p = 0.001). (B) Patients with greater PAD in the acute phase still performed more slowly on the IPS in the chronic phase. (p = 0.011). r = Spearman rank-order correlation coefficient. Color image is available online.

Neurodegeneration-vulnerable tracts associated with brain-PAD scores

The r between brain-PAD and FA values of WMTs involved in model training (n = 33) was calculated to evaluate the related WMT contribution to increased brain age. The FA value in 16 WMTs was significantly correlated with the brain-PAD scores in the acute phase. They were mainly located in the commissural, association, and projection fibers (r: -0.305 ∼ -0.473, p < 0.05 after familywise error (FWE) correction for multiple comparisons) (Table 3), primarily including the corpus callosum, fornix, corona radiate, and superior longitudinal fasciculus. In the chronic phase, the FA value in the body of the fornix was still significantly related to the brain-PAD at 6 ∼ 12 month follow-up, after FWE correction for multiple comparisons (r = -0.434, p < 0.05) (Table 3).

Table 3.

Correlation between FA value of Age-Sensitive WMTs and Brain-PAD in Acute and Chronic Phases

Age-sensitive tracts	r-acute	p value	r-chronic	p value
Middle cerebellar peduncle	-0.256	0.006	-	-
Genu of corpus callosum	-0.473	<0.001^a	-0.374	0.008
Body of corpus callosum	-0.34	<0.001^a	-0.326	0.021
Splenium of corpus callosum	-0.249	0.007	-	-
Body of fornix	-0.332	<0.001^a	-0.434	0.0016^a
Inferior cerebellar peduncle (R)Inferior cerebellar peduncle (L)Superior cerebellar peduncle (R)Superior cerebellar peduncle (L)	-0.008-0.288-0.186-0.247	0.9310.0020.0460.008	----	----
Cerebral peduncle (R)	-0.305	<0.001^a	-0.207	0.15
Cerebral peduncle (L)	-0.351	<0.001^a	-0.17	0.238
Posterior limb of internal capsule (R)Posterior limb of internal capsule (L)	-0.205-0.169	0.0270.069	--	--
Anterior corona radiata (R)	-0.426	<0.001^a	-0.365	0.01
Anterior corona radiata (L)	-0.419	<0.001^a	-0.378	0.007
Superior corona radiata (L)	-0.315	<0.001^a	-0.056	0.7
Posterior thalamic radiation (R)	-0.303	<0.001^a	-0.377	0.007
Posterior thalamic radiation (L)	-0.385	<0.001^a	-0.23	0.108
Sagittal stratum (R)	-0.381	<0.001^a	-0.236	0.1
Sagittal stratum (L)	-0.338	<0.001^a	-0.127	0.377
External capsule (R)	-0.250	0.007	-	-
External capsule (L)	-0.135	0.149	-	-
Cingulate gyrus (R)	-0.204	0.028	-	-
Hippocampus (L)	-0.140	0.134	-	-
Fornix crus (R)	-0.229	0.013	-	-
Fornix crus (L)	-0.374	<0.001^a	-0.254	0.075
Superior longitudinal fasciculus (R)	-0.308	<0.001^a	-0.021	0.887
Superior longitudinal fasciculus (L)	-0.261	0.005	-	-
Superior fronto-occipital fasciculus (R)	-0.355	<0.001^a	-0.31	0.029
Superior fronto-occipital fasciculus (L)	-0.371	<0.001^a	-0.244	0.088
Inferior fronto-occipital fasciculus (R)	-0.004	0.966	-	-
Inferior fronto-occipital fasciculus (L)	-0.043	0.648	-	-
Tapetum (L)	-0.236	0.011	-	-

Statistical significance after familywise error (FWE) correction for multiple comparisons (p < 0.05).

WMT, white matter tract; brain-PAD, brain-predicted age difference.

Factors affecting brain-PAD scores

The patients' chronological age at mTBI affected their PAD derived from the acute phase (F _{2, 111} = 10.05, p < 0.001, partial η ² = 0.15), after adjusting for sex and education (Table 4). The brain-PAD scores in the elderly patients (n = 27, 6.74 ± 5.61 years) were higher than those in the middle-aged (n = 57, 1.98 ± 5.44 years) and young patients (n = 32, 0.18 ± 5.66 years) (multiple comparisons for Bonferroni correction, p < 0.05) (Fig. 5A).

FIG. 5.

Individual differences in predicted age difference (PAD). (A) Patients' chronological age when mild traumatic brain injury (mTBI) occurred affected their PAD in the acute phase. The elderly patients (≥ 50 years of age) had greater PAD (6.74 ± 1.08 years) than middle-aged (30 ∼ 50 years) (1.98 ± 0.72 years) or young (≤ 30 years) (0.18 ± 1.00 years) patients. (B) The post-concussion symptom (PCS) severity post-injury affected the PAD in the acute phase. Patients with severer complaints on PCS had greater PAD (3.262 ± 6.350 years) than those without PCS complaints (0.159 ± 3.488 years). PCS+ = with post-concussion symptom complaints; PCS- = without post-concussion symptom complaints. Color image is available online.

Table 4.

Individual Differences in Brain-PAD Scores Post-Injury

	Brain-PAD (Years)	F value	p value	partial η²
Chronological age		10.05	< 0.001^a	0.15
Youth (n = 32)	0.18 ± 5.66
Adults (n = 57)	1.98 ± 5.44
Elderly (n = 27)	6.74 ± 5.61
PCS		5.26	0.02^a	0.05
PCS+ (n = 91)	3.21 ± 5.44
PCS- (n = 25)	0.34 ± 5.50
AA		0.86	0.35	-
AA+ (n = 15)	3.84 ± 5.57
AA- (n = 101)	2.41 ± 5.52
APOE		0.46	0.49	-
ɛ4+ (n = 17)	3.75 ± 6.31
ɛ4- (n = 64)	2.57 ± 6.24

Statistical significance after multiple comparisons for Bonferroni correction (p < 0.05).

PCS, post-concussion symptoms; PCS +, with PCS complaints; PCS -, without PCS complaints; AA, anterograde amnesia; AA +, with AA; AA -, without AA; APOE, apolipoprotein E; ɛ4+, APOE ɛ4 carriers; ɛ4- , APOE non-ɛ4 carriers

The PCS severity post-injury also affected the PAD in the acute phase (F _{1, 111} = 5.26, p = 0.02, partial η ² = 0.05), after adjusting for sex, education, and age. Patients with PCS complaints (i.e., PCS+) had higher brain-PAD scores (n = 91, 3.21 ± 5.44 years) than the PCS- group (n = 25, 0.34 ± 5.50 years) (Fig. 5B) (Table 4).

The AA and APOE genotypes did not affect PAD in the acute phase (AA: F _1,111 = 0.86, p = 0.35; APOE: F _{1, 111} = 0.46, P = 0.49), after adjusting for sex, education, and age (Table 4).

Discussion

Our study revealed that mTBI triggered an accelerated aging process in the brain, deviating from its chronological trajectory. Increased brain age was observed in the early acute phase and remained stable within a 6 ∼ 12 month follow-up. The WMTs relevant to the accelerated-aging process were primarily located in the connecting aging-related and neurodegeneration-vulnerable regions. Also, the brain-PAD of patients measured in the acute phase could predict the individual performance of IPS at the 6 ∼ 12 month follow-up. Patients with older chronological age and more severe PCS complaints were more vulnerable to having “older” brain age.

Unlike one recent study about increased brain-PAD observed in chronic patients,³³ the observations in our research took place during a relatively early phase of mTBI. The results showed that the discrepancy between predicted brain age and chronological age (i.e., brain-PAD >2 years) was detectable even at the acute phase post-injury. Moreover, the increased brain age cannot simply be explained as a transient physiological pattern in response to traumatic events. Instead, the current longitudinal observation of brain-PAD further demonstrated that such an accelerated brain-aging process was a stable change. As described in neurological and neuropsychiatric studies, increases in brain age have been considered to be a surrogate neurobiomarker of vulnerability to neurodegeneration.^7,34 Likewise, the brain-PAD score may be a candidate biomarker for mTBI. No significant correlations between the brain-PAD score and time since injury in either the acute or chronic phase were observed, which further supplemented the explanation that such increases in brain age might reflect vulnerability to an accelerated aging process.

For both the acute phase and the follow-up, the increases in brain age were inversely related to the integrity of the fornix, implying that the persistent brain-aging process accelerated by mTBI might be caused by the deteriorated integrity of the fornix. The fornix, connecting the hippocampal formation in the limbic system, is responsible for neurodegeneration progression.³⁵ Its integrity alterations were observed in mild cognitive impairment and chronological aging processes.^36,37 Moreover, a recent theory about the Alzheimer's disease (AD) pathway suggests that transmission initiates in the hippocampus regions, spreading latterly to the rest of the brain,^38,39 implying that the fornix may serve as a common origin of neurodegeneration in the transmission pathway. Therefore, based on the current findings on the steady association between the integrity of the fornix and increased brain age, we can infer that the brain-PAD score derived from WMT integrity was informative in evaluating the potential neurodegeneration effects.

We explored factors affecting brain-PAD in patients with mTBI, which could explain the individual differences in the acceleration of brain-aging post-injury. A relative higher variance of PAD was presented in the present study as well as the previous studies^8,10,15 (e.g., PAD of HCs: 0.07 ± 7.41 years; PAD of TBIs: 4.66 ± 10.8 years in Cole and coworkers¹⁵), which may be the result of individual differences affecting the PAD. The present results showed that patients with older chronological age and more severe PCS complaints were more vulnerable to developing greater brain-PAD. These results, on the one hand, were consistent with some retrospective studies' findings that post-concussion syndrome and older age increased the risks of developing dementia for patients with mTBI.^3,23 On the other hand, these results revealed the possibility that age at exposure to TBI and PCS severity may explain such high variance of PAD. Therefore, PAD may potentially be used as an index to reflect the individual differences of outcome in mTBIs. In our study, the APOE genotypes and AA showed non-significant effects on brain-PAD following mTBI. Both APOE ɛ4 and AA are generally considered to be risk factors for AD.^24,25 Those results provide some indications that the neurodegeneration-vulnerable aging process accelerated by mTBI may be different from AD pathology.⁴⁰ However, there still need to be further studies to explore the common and dissociated factors contributing to neurodegeneration in mTBI and AD.

Our model explained the variations in the chronological aging process as aging-related changes in WMTs. Unlike previous studies using macroscopic morphometrics to predict brain age in moderate to severe TBI patients,^15,33 we adopted subtler microstructural features, considering that mTBI usually causes diffuse axonal injury¹⁶ and long-distance white matter disconnections.¹⁹ This damage, thereafter, results in degradation in communication efficiency among brain regions, and contributes to cognitive dysfunctions during the brain-aging process, commonly reflected as a slow speed of information processing and transferring.⁴¹ Moreover, our results demonstrated that the brain-PAD of the acute phase could predict the individual profiles of IPS in the chronic phase. A greater brain-PAD was relevant to poorer IPS performance, suggesting that white matter integrity loss underlying the accelerated aging process contributed to the cognitive impairments following mTBI.

There are limitations to the present study. First, our brain-age prediction model was trained with DTI metrics with respect to injury characteristics of mTBI. The brain-aging process is biologically complex, which may be explained further by the multi-modality of neuroimaging. Second, the follow-up loss and insufficient APOE ɛ4 samples limited our understandings of the association between the accelerated brain-aging process following mTBI and neurodegenerative diseases. Third, the increased brain-PAD in our study only provided the evidence in neuroimaging to support the potentially accelerated brain aging post-injury. The accelerated brain aging may also involve a series of complex processes, such as Wallerian degeneration of white matter,⁴² deposition of amyloid-β proteins in the brain,⁴³ and increases in neurofilament light chain caused by diffuse axonal injury.⁴⁴

Conclusion

In summary, acceleration of the brain-aging process underlying deterioration in WMTs is detectable even in the acute phase of mTBI, and persists into the chronic phase, which provides insights into the risks of mTBI for neurodegeneration-vulnerable conditions. Moreover, the increases in brain age are related to clinical characteristics and are capable of predicting neuropsychological outcomes.

Footnotes

Acknowledgments

We thank all patients and healthy volunteers who participated in the study. We also acknowledge the data sharing from IXI (http://brain-development.org/ixi-dataset/), CoRR (http://fcon_1000.projects.nitrc.org/indi/CoRR/html/), and Cyclotron Research Centre (University of Liège) (https://www.nitrc.org/projects/parktdi/).

Funding Information

This study was funded by the National Natural Science Foundation of China (Grant Nos. 82071993, 81771914, 81871331, 61971451, 81671671), Key projects in Hunan Province (Grant Nos. 2020SK4001, 2019SK-2131), Natural Science Foundation of Zhejiang Province (Nos. LY19H180003), Wenzhou Science and Technology Bureau in China (No. Y20180112), Fundamental Research Funds for the Central Universities (Grant Nos. Xjj2018229, xzy022019045).

Author Disclosure Statement

No competing financial interests exist.

Supplementary Material

Supplementary Table S1

Supplementary Text (includes Methods and Results)

References

Levin

H.S.

, and Diaz-Arrastia

R.R.

(2015). Diagnosis, prognosis, and clinical management of mild traumatic brain injury. Lancet Neurol. 14, 506–517.

Mac Donald

C.L.

, Barber

, Andre

, Panks

, Zalewski

, and Temkin

(2019). Longitudinal neuroimaging following combat concussion: sub-acute, 1 year and 5 years post-injury. Brain commun. 1, fcz031.

Gardner

R.C.

, Burke

J.F.

, Nettiksimmons

, Kaup

, Barnes

D.E.

, and Yaffe

(2014). Dementia risk after traumatic brain injury vs nonbrain trauma: the role of age and severity. JAMA Neurol. 71, 1490–1497.

Fann

J.R.

, Ribe

A.R.

, Pedersen

H.S.

, Fenger-Grøn

, Christensen

, Benros

M.E.

, and Vestergaard

(2018). Long-term risk of dementia among people with traumatic brain injury in Denmark: a population-based observational cohort study. Lancet Psychiatry, 5, 424–431.

Cox

S.R.

, Ritchie

S.J.

, Tucker-Drob

E.M.

, Liewald

D.C.

, Hagenaars

S.P.

, Davies

, Wardlaw

J.M.

, Gale

C.R.

, Bastin

M.E.

, and Deary

I.J.

(2016). Ageing and brain white matter structure in 3,513 UK Biobank participants. Nat. Commun. 7, 1–13.

Good

C.D.

, Johnsrude

I.S.

, Ashburner

, Henson

R.N.

, Friston

K.J.

, and Frackowiak

R.S.

(2001). A voxel-based morphometric study of ageing in 465 normal adult human brains. NeuroImage, 14, 21–36.

Wang

, Knol

M.J.

, Tiulpin

, Dubost

, de Bruijne

, Vernooij

M.W.

, Adams

H.H.

, Ikram

M.A.

, Niessen

W.J.

, and Roshchupkin

G.V.

(2019). Gray matter age prediction as a biomarker for risk of dementia. Proc. Natl. Acad. Sci. 116, 21,213–21,218.

Schnack

H.G.

, Van Haren

N.E.

, Nieuwenhuis

, Hulshoff Pol

H.E.

, Cahn

, and Kahn

R.S.

(2016). Accelerated brain aging in schizophrenia: a longitudinal pattern recognition study. Am. J. Psychiatry, 173, 607–616.

Cole

J.H.

, Annus

, Wilson

L.R.

, Remtulla

, Hong

Y.T.

, Fryer

T.D.

, Acosta-Cabronero

, Cardenas-Blanco

, Smith

, and Menon

D.K.

(2017). Brain-predicted age in Down syndrome is associated with beta amyloid deposition and cognitive decline. Neurobiol. Aging, 56, 41–49.

10.

Cole

J.H.

, Underwood

, Caan

M.W.A.

, Francesco

D.D.

, Zoest

R.A.V.

, Leech

, Wit

F.W.N

.M., Portegies

, Geurtsen

G.J.

, and Schmand

B.A.

(2017). Increased brain-predicted aging in treated HIV disease. Neurology, 88, 1349–1357.

11.

Cole

J.H.

and Franke

(2017). Predicting age using neuroimaging: innovative brain ageing biomarkers. Trends Neurosci. 40, 681–690.

12.

Gaser

, Franke

, Klöppel

, Koutsouleris

, Sauer

, and Alzheimers Disease Neuroimaging Initiative (2013). BrainAGE in mild cognitive impaired patients: predicting the conversion to Alzheimer's disease. PloS One 8, e67346.

13.

Koutsouleris

, Davatzikos

, Borgwardt

, Gaser

, Bottlender

, Frodl

, Falkai

, Riecher-Rössler

, Möller

H.-J.

, and Reiser

(2014). Accelerated brain aging in schizophrenia and beyond: a neuroanatomical marker of psychiatric disorders. Schizophr. Bull. 40, 1140–1153.

14.

Liang

, Zhang

, and Niu

(2019). Investigating systematic bias in brain age estimation with application to post-traumatic stress disorders. Hum. Brain Mapp. 40, 3143–3152.

15.

Cole

J.H.

, Leech

, Sharp

D.J.

, and Alzheimer's Disease Neuroimaging

Initiative

(2015). Prediction of brain age suggests accelerated atrophy after traumatic brain injury. Ann. Neurol. 77, 571–581.

16.

Smith

D.H.

, Meaney

D.F.

, and Shull

W.H.

(2003). Diffuse axonal injury in head trauma. J. Head Trauma Rehabil. 18, 307–316.

17.

Mac Donald

C.L.

, Johnson

A.M.

, Cooper

, Nelson

E.C.

, Werner

N.J.

, Shimony

J.S.

, Snyder

A.Z.

, Raichle

M.E.

, Witherow

J.R.

, Fang

, Flaherty

S.F.

, and Brody

D.L.

(2011). Detection of blast-related traumatic brain injury in U.S. military personnel. N. Engl. J. Med. 364, 2091–2100.

18.

Lipton

M.L.

, Gellella

, Lo

, Gold

, Ardekani

B.A.

, Shifteh

, Bello

J.A.

, and Branch

C.A.

(2008). Multifocal white matter ultrastructural abnormalities in mild traumatic brain injury with cognitive disability: a voxel-wise analysis of diffusion tensor imaging. J. Neurotrauma, 25, 1335–1342.

19.

Sharp

D.J.

, Scott

, and Leech

(2014). Network dysfunction after traumatic brain injury. Nat. Rev. Neurol. 10, 156.

20.

Smith

D.H.

, Johnson

V.E.

, and Stewart

(2013). Chronic neuropathologies of single and repetitive TBI: substrates of dementia?. Nat. Rev. Neurol. 9, 211.

21.

Rosenfeld

J.V.

, McFarlane

A.C.

, Bragge

, Armonda

R.A.

, Grimes

J.B.

, and Ling

G.S.

(2013). Blast-related traumatic brain injury. Lancet. Neurol. 12, 882–893.

22.

Tipping

M.E.

(2001). Sparse Bayesian learning and the relevance vector machine. J. Mach. Learn. Res. 1, 211–244.

23.

Carroll

, Cassidy

J.D.

, Peloso

, Borg

, Von Holst

, Holm

, Paniak

, and Pépin

(2004). Prognosis for mild traumatic brain injury: results of the WHO Collaborating Centre Task Force on Mild Traumatic Brain Injury. J. Rehabil. Med. 36, 84–105.

24.

Jagust

(2013). Vulnerable neural systems and the borderland of brain aging and neurodegeneration. Neuron, 77, 219–234.

25.

Nicoll

J.A.

, Roberts

G.W.

, and Graham

D.I.

(1995). Apolipoprotein E ɛ4 allele is associated with deposition of amyloid β-protein following head injury. Nat. Med. 1, 135–137.

26.

Holm

, David Cassidy

, Carroll

, and Borg

(2005). Summary of the WHO collaborating centre for neurotrauma task force on mild traumatic brain injury. J. Rehabil. Med. 37, 137–141.

27.

World Health Organization (1992). The ICD-10 Classification of Mental and Behavioural Disorders: Clinical Descriptions and Diagnostic Guidelines. World Health Organization: Geneva.

28.

Ryan

J.J.

, and Ward

L.C.

(1999). Validity, reliability, and standard errors of measurement for two seven-subtest short forms of the Wechsler Adult Intelligence Scale—III. Psychol. Assess. 11, 207.

29.

Bowie

C.R.

, and Harvey

P.D.

(2006). Administration and interpretation of the Trail Making Test. Nat. Protoc. 1, 2277.

30.

Merisaari

, Tuulari

J.J.

, Karlsson

, Scheinin

N.M.

, Parkkola

, Saunavaara

, Lähdesmäki

, Lehtola

S.J.

, Keskinen

, and Lewis

J.D.

(2019). Test–retest reliability of Diffusion Tensor Imaging metrics in neonates. NeuroImage, 197, 598–607.

31.

Bright

M.G.

, and Murphy

(2013). Reliable quantification of BOLD fMRI cerebrovascular reactivity despite poor breath-hold performance. NeuroImage, 83, 559–568.

32.

Jenkinson

, Beckmann

C.F.

, Behrens

T.E.J.

, Woolrich

M.W.

, and Smith

S.M.

(2012). FSL. NeuroImage, 62, 782–790.

33.

Savjani

R.R.

, Taylor

B.A.

, Acion

, Wilde

E.A.

, and Jorge

R.E.

(2017). Accelerated changes in cortical thickness measurements with age in military service members with traumatic brain injury. J. Neurotrauma, 34, 3107–3116.

34.

Cole

J.H.

, and Franke

(2017). Predicting age using neuroimaging: innovative brain ageing biomarkers. Trends Neurosci. 40, 681–690.

35.

D'Esposito

, Verfaellie

, Alexander

M.P.

, and Katz

D.I.

(1995). Amnesia following traumatic bilateral fornix transection. Neurology, 45, 1546–1550.

36.

Jacquemont

, Fallani

F.D.V.

, Bertrand

, Epelbaum

, Routier

, Dubois

, Hampel

, Durrleman

, Colliot

, and Initiative, A.s.D.N. (2017). Amyloidosis and neurodegeneration result in distinct structural connectivity patterns in mild cognitive impairment. Neurobiol. Aging, 55, 177–189.

37.

de Groot

, Cremers

L.G.

, Ikram

M.A.

, Hofman

, Krestin

G.P.

, van der Lugt

, Niessen

W.J.

, and Vernooij

M.W.

(2015). White matter degeneration with aging: longitudinal diffusion MR imaging analysis. Radiology, 279, 532–541.

38.

Iba

, Guo

J.L.

, McBride

J.D.

, Zhang

, Trojanowski

J.Q.

, and Lee

V.M.-Y.

(2013). Synthetic tau fibrils mediate transmission of neurofibrillary tangles in a transgenic mouse model of Alzheimer's-like tauopathy. J. Neurosci. 33, 1024–1037.

39.

Ahmed

, Cooper

, Murray

T.K.

, Garn

, McNaughton

, Clarke

, Parhizkar

, Ward

M.A.

, Cavallini

, and Jackson

(2014). A novel in vivo model of tau propagation with rapid and progressive neurofibrillary tangle pathology: the pattern of spread is determined by connectivity, not proximity. Acta Neuropathol. 127, 667–683.

40.

Weiner

M.W.

, Harvey

, Hayes

, Landau

S.M.

, Aisen

P.S.

, Petersen

R.C.

, Tosun

, Veitch

D.P.

, Jack Jr

C.R.

, and Decarli

(2017). Effects of traumatic brain injury and posttraumatic stress disorder on development of Alzheimer's disease in Vietnam veterans using the Alzheimer's Disease Neuroimaging Initiative: preliminary report. Alzheimers Dement. 3, 177–188.

41.

O'Sullivan

, Jones

D.K.

, Summers

, Morris

, Williams

, and Markus

(2001). Evidence for cortical “disconnection”. as a mechanism of age-related cognitive decline. Neurology, 57, 632–638.

42.

Meythaler

J.M.

, Peduzzi

J.D.

, Eleftheriou

, and Novack

T.A.

(2001). Current concepts: diffuse axonal injury–associated traumatic brain injury. Arch. Phys. Med. Rehabil. 82, 1461–1471.

43.

Johnson

V.E.

, Stewart

, and Smith

D.H.

(2012). Widespread tau and amyloid-beta pathology many years after a single traumatic brain injury in humans. Brain Pathol. 22, 142–149.

44.

Petzold

(2005). Neurofilament phosphoforms: surrogate markers for axonal injury, degeneration and loss. J. Neurol. Sci. 233, 183–198.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.01 MB

0.02 MB