Characteristics of the study population
We included 41 TBI patients who completed the 6-month and 1-year follow-up. The median follow up time was 12.3 months. The majority were young (82.9%) with a median age of 32 (IQR:28) years and predominantly male (92.7%). The leading cause of TBI was road traffic accidents (82.9%), with the majority suffering from moderate to severe TBI (63.4%). More than three-quarters (75.6%) had 0 to 1 comorbidity, while alcohol consumption was infrequent (17.1%). Out of the 41 TBI patients, six patients developed PTE (14.63%), all within six months post-TBI, with median days to develop PTE 45 (IQR: 37) days (Table 1).
Table 1
Demographic and TBI characteristics of the study population (N = 41).
Characteristic
|
Frequency (n)
|
Percent (%)
|
Age group
|
Young (up to 59)
|
34
|
82.9
|
Old (≥ 60)
|
7
|
17.1
|
Gender
|
Male
|
38
|
92.7
|
Female
|
3
|
7.3
|
Ethnicity
|
Malay
|
19
|
46.3
|
Chinese
|
13
|
31.7
|
Indian
|
9
|
22.0
|
Alcohol consumption
|
Ever
|
7
|
17.1
|
Never
|
34
|
82.9
|
Prior history of head injury
|
Yes
|
4
|
9.8
|
No
|
37
|
90.2
|
Number of comorbidities
|
0 to 1
|
31
|
75.6
|
≥ 2
|
10
|
24.4
|
Etiology of TBI
|
RTA
|
34
|
82.9
|
|
Fall
|
6
|
14.6
|
|
Assault
|
1
|
2.4
|
TBI severity
|
Mild
|
15
|
36.6
|
|
Moderate to severe
|
26
|
63.4
|
Posttraumatic epilepsy
|
Yes
|
6
|
14.6
|
No
|
35
|
85.4
|
Serum HMGB1, IL-1β and TNF-α quantification by ELISA
The mean and median concentrations of HMGB1, IL-1β, and TNF-α at 0, 6, and 12 months for the entire cohort (N = 41) consistently decreased over the year-long observation period. The mean levels of HMGB1 decreased notably from 18.5 ng/mL at 0 months to 8.41 ng/mL at 12 months. Similarly, the mean levels of IL-1β and TNF-α decreased over one year post-TBI (Table 2).
Table 2
Mean (SD), median (IQR) level of HMGB1 (ng/mL), IL-1β (pg/mL) and TNF-α (pg/mL) at 0, 6 and 12 months in all patients (N = 41).
Biomarker and time (month)
|
Mean (SD)
|
Median (IQR)
|
HMGB1 at 0
|
18.5 (10.7)
|
16.6 (10.5
|
HMGB1 at 6
|
9.00 (5.01)
|
8.82 (4.96)
|
HMGB1 at 12
|
8.41 (6.54)
|
6.62 (7.21)
|
IL-1β at 0
|
83.7 (30.4)
|
83.2 (44.6)
|
IL-1β at 6
|
52.7 (8.43)
|
51.8 (5.47)
|
IL-1β at 12
|
50.6 (16.8)
|
51.7 (15.4)
|
TNF-α at 0
|
15.7 (5.63)
|
17.5 (5.90)
|
TNF-α at 6
|
4.62 (1.97)
|
4.03 (1.84)
|
TNF-α at 12
|
4.60 (3.39)
|
3.20 (1.64)
|
HMGB1 Levels Over Time:
The Friedman test confirmed a significant overall difference in HMGB1 levels across the three-time points (p < 0.001). Wilcoxon signed-rank tests revealed significant reductions in HMGB1 levels from baseline (0 months) to both six months (p < 0.001) and 12 months (p < 0.001).
IL-1β Levels Over Time:
Similar patterns were observed for IL-1β levels. The Friedman test underscored a significant overall difference in IL-1β levels across the three time points (p < 0.001). Wilcoxon signed-rank tests indicated significant reductions from baseline (0 months) to both six months (p < 0.001) and 12 months (p < 0.001). No significant difference was found between IL-1β levels at 6 and 12 months (p = 0.507).
TNF-α Levels Over Time:
For TNF-α levels, the Friedman test showed a significant overall difference in TNF-α levels across the three-time points (p < 0.001). Wilcoxon signed-rank tests demonstrated significant reductions from baseline (0 months) to both six months (p < 0.001) and 12 months (p < 0.001). Similarly, no significant difference was observed between TNF-α levels at 6 and 12 months (p = 0.210).
HMGB1, IL1-β and TNF-α levels association with PTE development:
Comparing PTE and non-PTE groups using the Mann-Whitney U test, there is a statistically significant difference in HMGB1 levels at 12 months, IL-1β levels at 6 months and TNF-α levels at 6 and 12 months between the PTE and non-PTE group (Table 3). The change in HMGB1 levels from 6 to 12 months was significantly different between PTE and non-PTE groups (p = 0.008). No significant differences in HMGB1 changes from 0 to 6 months and 0 to 12 months between groups. Figure 1 depicted HMGB1 and the changes in levels, as well as IL-1β and TNF-α levels between PTE and non-PTE groups over a year.
Table 3
Median (IQR) level of HMGB1 (ng/mL), IL-1β (pg/mL) and TNF-α (pg/mL) at 0, 6 and 12 months between PTE and non-PTE group.
Biomarker and time (month)
|
PTE (n = 6)
|
Non-PTE (n = 35)
|
p-value
|
HMGB1 at 0
|
19.2 (8.39)
|
15.5 (10.3)
|
0.199
|
HMGB1 at 6
|
9.47 (1.32)
|
8.06 (5.81)
|
0.347
|
HMGB1 at 12
|
11.2 (2.27)
|
6.14 (5.42)
|
*0.026
|
IL-1β at 0
|
87.5 (31.8)
|
83.2 (47.3)
|
0.592
|
IL-1β at 6
|
62.2 (8.52)
|
51.6 (3.30)
|
*0.023
|
IL-1β at 12
|
60.3 (21.1)
|
51.4 (17.1)
|
0.073
|
TNF-α at 0
|
17.7 (4.43)
|
17.1 (5.26)
|
0.685
|
TNF-α at 6
|
6.32 (3.76)
|
3.66 (1.65)
|
*0.003
|
TNF-α at 12
|
7.78 (5.92)
|
3.19 (1.32)
|
*0.004
|
HMGB change 0 to 6
|
10.21 (7.62)
|
6.37 (7.32)
|
0.097
|
HMBG changes 6 to 12
|
1.43(1.93)
|
1.94 (3.62)
|
*0.008
|
HMGB changes 0 to 12
|
8.09 (8.61)
|
7.66 (5.20)
|
0.580
|
*p = < 0.05 |
HMGB1, IL-1β, and TNF-α levels at different time points (0, 6, and 12 months) within PTE group (n = 6) vs non-PTE group (n = 35):
The Wilcoxon signed-rank test was used to evaluate whether differences between levels at 2 time points were significantly different within PTE and non-PTE group separately (see Supplementary 1). Within the PTE group, there were significant decreases in HMGB1 levels from baseline to 6 months (p = 0.028), 6 to 12 months (p = 0.028), and baseline to 12 months (p = 0.028). This indicates a continual reduction in systemic inflammation, as measured by HMGB1, over the first year in patients who develop post-traumatic epilepsy. There were no significant changes observed in IL-1β or TNF-α within the PTE group. In contrast, the non-PTE group demonstrated significant decreases across all three biomarkers. HMGB1 decreased from baseline to 6 months (p < 0.001), 6 to 12 months (p = 0.013), and baseline to 12 months (p < 0.001). Significant reductions were also seen for IL-1β from baseline to 6 months (p < 0.001) and baseline to 12 months (p < 0.001). Likewise, TNF-α decreased from baseline to 6 months (p < 0.001) and baseline to 12 months (p < 0.001) in these patients without PTE.
The cognitive performance scores, specifically ACE-III, CTMT, and WAIS-4, revealed a significant difference between the PTE and non-PTE groups at 6 and 12 months (Table 4). Patients in the PTE group exhibited significantly lower cognitive scores compared to the non-PTE group at both time points (Fig. 2).
Table 4
Cognitive performance scores between PTE and non-PTE group at 6 and 12 months.
Test
|
Cognitive score
|
PTE
|
Non-PTE
|
p-value
|
ACE-III
|
Mean (SD) at 6 mo
|
39.2 (26.4)
|
82.5 (11.3)
|
< 0.001
|
Mean (SD) at 12 mo
|
44.2 (21.4)
|
82.5 (11.5)
|
< 0.001
|
Median (IQR) at 6 mo
|
31.5 (16.8)
|
85 (14.5)
|
0.003
|
Median (IQR) at 12 mo
|
36 (8.75)
|
85 (13)
|
0.001
|
CTMT
|
Mean (SD) at 6 mo
|
103 (26.7)
|
158 (42)
|
0.004
|
Mean (SD) at 12 mo
|
115 (33.5)
|
161 (42.6)
|
0.016
|
Median (IQR) at 6 mo
|
90.5 (7.75)
|
155 (66)
|
0.003
|
Median (IQR) at 12 mo
|
105 (31.8)
|
160 (55.5)
|
0.017
|
WAIS-IV
|
Mean (SD) at 6 mo
|
66.8 (14.2)
|
81.6 (11.6)
|
0.008
|
Mean (SD) at 12 mo
|
62.2 (13.5)
|
82.9 (12.9)
|
0.025
|
Median (IQR) at 6 mo
|
63.5 (10.3)
|
84 (7.5)
|
< 0.001
|
Median (IQR) at 12 mo
|
60 (11)
|
81 (15)
|
0.005
|
*Mann-Whitney U test used for between-group comparisons of cognitive scores at each timepoint |
HMGB1, IL-1β and TNF-α levels correlation with cognitive impairment:
Correlation analysis highlighted significant relationships between changes in HMGB1, IL-1β, and TNF-α levels and alterations in cognitive scores (see Supplemetary 2). The r values provide a quantified measure of the strength and direction of the associations between biomarker level changes and cognitive score changes over the 6 months [24]. HMGB1 level changes and ACE-III score difference exhibit a moderate negative correlation of r = -0.375 (p < 0.05), between changes in HMGB1 levels and cognitive score changes. Similarly seen in IL-1β level changes that show a moderate negative correlation with ACE-III score changes (r=-0.337, p < 0.05). TNF-α levels have a strong negative correlation with ACE-III (r=-0.541, p < 0.001) and WAIS-IV (r=-0.460, p < 0.05).
HMGB1 relationship with IL-1β and TNF-α:
There is a moderately positive correlation between HMGB1 and IL-1β (r = 0.339, p < 0.05) and a moderately positive correlation between HMGB1 and TNF-α (r = 0.349, p < 0.05). The positive correlations indicate a general tendency for levels of HMGB1 to increase when levels of either IL1-β or TNF-α increase. So, higher inflammatory signaling via these cytokines seems associated with more HMGB1 release.
Association of cognitive function chnages with changes in HMGB1 levels and PTE
We conducted linear regression models to examine the association between changes in cognitive function (from 6 to 12 months), measured using the ACE-III scale, and two predictors: differences in HMGB1 levels (from 6 to 12 months) and the presence of PTE (Table 5). The ACE-III test was chosen for its comprehensive measure of cognitive function and reliability within the TBI population [20, 25].
Simple Regression Analysis:
We performed simple regression analysis to identify independent variables associated with changes in cognitive performance. Significant covariates with p < 0.2 for inclusion in multivariate models were PTE, ICH, GCS, HMGB1 changes, TNF-α changes, and IL-1β changes. No multicollinearity issues were observed.
Multiple linear regression Model 1:
The first multivariate model included the two independent variables: presence of PTE and HMGB1 changes. This model significantly predicted outcomes (p = 0.003) and explained 26.1% of variance (adjusted R2 = 0.222). Both PTE (B=-6.963, p = 0.014) and HMGB1 changes (B=-0.783, p = 0.003) were significant independent predictors, indicating presence of PTE and lesser HMGB1 decrease were associated with poorer cognitive outcomes.
Multivariate Model 2 (covariates added: Presence of ICH, GCS, Use of AED).
In Model 2, variables included were: PTE presence, HMGB1 changes, Presence of ICH, GCS, Use of AED and GCS. The multivariate regression analysis showed notable improvements in model fit compared to Model 1. The R2 value increased to 0.418, indicating that approximately 41.8% of the variance in the outcome variable was explained by the predictors in this model. Similarly, the adjusted R2 value increased to 0.335. The overall model was statistically significant (p = 0.001), and several predictors emerged as statistically significant contributors to the outcome variable, including use of AED (p = 0.035) and HMGB1 changes (p = 0.014).This suggests that changes in HMGB1 levels have a significant impact on the outcome variable which is cognitive level changes, even after accounting for other confounders.
Multivariate Model 3 (covariates added: TNF-α changes, and IL-1β changes).
Continuing the exploration, Model 3 further enhanced the explanatory power of the regression analysis. In Model 3, variables included were: PTE presence, HMGB1 changes, presence of ICH, GCS, use of AED, GCS, TNF-α changes, and IL-1β changes. The R2 value substantially increased to 0.665, indicating that approximately 66.5% of the variance in the outcome variable was accounted for by the predictors in this comprehensive model. The adjusted R2 value remained high at 0.594. The overall model was highly statistically significant (p < 0.001), underscoring its robustness. Notably, several predictors retained statistical significance including PTE presence (p = 0.002), HMGB1 changes (p = 0.038), and TNF-α changes (p < 0.001), further emphasizing their importance in predicting the outcome variable within the neuroscience context. HMGB1 changes remained a statistically significant predictor of the outcome suggests its significant impact on cognitive function, even after accounting for other relevant confounders like GCS, presence of ICH and the use of AED.
Table 5
Linear regression analysis of covariates association with change in cognitive performance from 6 to 12 months post TBI
|
Univariate regression
|
Multivariate regression
|
|
|
|
|
Model 1
|
Model 2
|
Model 3
|
|
|
|
R2
|
0.261
|
0.418
|
0.665
|
|
|
|
Adjusted R2
|
0.222
|
0.335
|
0.594
|
|
|
|
P value
|
0.003
|
0.001
|
< 0.001
|
|
|
|
F
|
6.702
|
5.031
|
9.377
|
Independent variables
|
|
|
|
|
|
|
B (SE B)
|
p-value
|
|
B (SE B)
|
p-value
|
B (SE B)
|
p-value
|
B (SE B)
|
p-value
|
Age
|
0.060 (-0.064, 0.184)
|
0.335
|
|
|
|
|
|
|
|
Gender
|
3.833 (-4.307, 11.973)
|
0.347
|
|
|
|
|
|
|
|
Ethnicity
|
-0.775 (-3.479, 1.929)
|
0.566
|
|
|
|
|
|
|
|
Alcohol
|
0.941 (-4.749, 6.632)
|
0.740
|
|
|
|
|
|
|
|
Depression score
|
-0.017 (-0.602, 0.567)
|
0.953
|
|
|
|
|
|
|
|
Year of education
|
-0.014 (-0.979, 0.951)
|
0.977
|
|
|
|
|
|
|
|
Had previous history of head injury
|
2.063 (-5.169, 9.294)
|
0.567
|
|
|
|
|
|
|
|
Presence of PTE
|
-4.943 (-10.795, 0.909)
|
0.095
|
|
-6.963 (-12.410, -1.515)
|
0.014
|
-4.376 (-10.039, 1.288)
|
0.126
|
-7.988 (-12.689, -3.287)
|
0.002
|
Presence of ICH
|
-5.60 (-11.389, 0.189)
|
0.058
|
|
|
|
-2.603 (-7.743, 2.536)
|
0.311
|
-1.050 (-5.123, 3.023)
|
0.603
|
GCS
|
0.605 (0.112, 1.098)
|
0.18
|
|
|
|
0.145 (-0.365, 0.654)
|
0.568
|
0.390 (0.023, 0.803)
|
0.064
|
Use of AED
|
-6.774 (-10.460, -3.087)
|
< 0.001
|
|
|
|
-4.315 (-8.299, -0.331)
|
0.035
|
-2.615 (-5.833, 0.603)
|
0.108
|
HMGB1 changes
|
-0.628 (-1.154, -0.103)
|
0.010
|
|
-0.783 (-1.289, -0.277)
|
0.003
|
-0.557 (-1.052, -0.053)
|
0.014
|
-0.364 (-0.768, 0.049)
|
0.038
|
TNF-α changes
|
-1.206 (-2.083, -0.329)
|
0.008
|
|
|
|
|
|
-1.590 (-2.318, -0.863)
|
< 0.001
|
IL-1β changes
|
-0.128 (-0.256, -0.001)
|
0.048
|
|
|
|
|
|
-0.071 (-0.161, 0.019)
|
0.117
|