Does brain damage caused by stroke versus trauma have different neuropsychological outcomes? A lesion-matched multiple case study

Abstract

Traumatic brain injury (TBI) and stroke both have the potential to cause significant damage to the brain, with resultant neuropsychological impairments. How these different mechanisms of injury influence cognitive and behavioral changes associated with brain damage, however, is not well understood. Moreover, previous research directly comparing TBI and stroke has not accounted carefully for lesion location and size. Here, using a detailed lesion-matching approach that was used previously to compare neuropsychological outcomes in stroke versus tumor, we compared the neuropsychological profiles of 14 patients with focal lesions caused by TBI to those of 27 lesion-matched patients with stroke. Each patient with TBI was matched to two patients with stroke, based on lesion location and size (except 1 TBI case where only 1 stroke match was available). Demographic attributes (age, gender, handedness, education) were also matched in the TBI: stroke triplets, as much as possible. The patients with TBI versus stroke had similar performances across all cognitive and behavioral measures, with no significant or clinically meaningful differences. A supplemental analysis on developmental- versus adult-onset TBI cases (with their respective stroke matches) also yielded non-significant results, with TBI and stroke groups being statistically indistinguishable. Our results suggest that focal lesions caused by TBI versus stroke have similar neuropsychological outcomes in the chronic recovery phase, when location and size of lesion are comparable across TBI versus stroke mechanisms of injury.

Introduction

Traumatic brain injuries (TBI) and strokes are acquired forms of brain damage that have different pathophysiologies that can affect neural integrity. TBI is an insult to the brain that can be the result of blunt force trauma (closed head injury) or an object penetrating the skull (open head injury) (Ferrell & Tanev, 2002; Iaccarino et al., 2018). In many of these cases, diffuse axonal injury (DAI) occurs when the brain rapidly moves during injury (Payne et al., 2021; Smith et al., 2003). Approximately, 69 million people globally and 1.7 million people in the United States sustain a TBI each year (Dewan et al., 2018; Georges & Booker, 2020). Of this population, about 20% have a moderate to severe injury (Iaccarino et al., 2018).

Strokes are caused by an occlusion (ischemic) or rupture (hemorrhagic) of a blood vessel in the brain, resulting in a depletion of oxygen and resultant damage to the surrounding brain tissue (Musuka et al., 2015; Moskowitz & Iadecola, 2010). Approximately 15 million people world-wide, including 795,000 individuals in the United States, suffer from strokes annually (Virani et al., 2021). An estimated 80% of strokes are ischemic (Boehme et al., 2017; Virani et al., 2021). The majority of patients with strokes are women, whereas most patients with TBIs are men (Zhang et al., 2016). Additionally, the modal age of onset for stroke is 65 years old; for TBI it is 35.4 years old (Hukkelhoven et al., 2003; Li et al., 2016; Roy-O'Reilly & McCullough, 2018).

Cognitive and behavioral changes are frequent debilitating sequela following both neurologic events. Previous studies have found that long-term cognitive impairment was prevalent in up to 50% of patients with stroke and 65% with moderate/severe TBI and was associated with worse rehabilitative outcomes (Rabinowitz & Levin, 2014; Parker et al., 2018; Schaapsmeerders et al., 2013; Sexton et al., 2019; Tang et al., 2018; Zinn et al., 2004). Cognitive deficits can be chronic. Deficits tend to stabilize after about 3 months, and are partly related to demographic attributes (del Ser et al., 2005; Douiri et al., 2013; Patel et al., 2003; Shretlen & Shapiro, 2003; Skandsen, et al., 2010). Importantly and not surprisingly, cognitive outcomes are strongly influenced by lesion characteristics such as size and location of injury.

Another factor is mechanism of injury, per se. For instance, Anderson et al. (1990) compared the neuropsychological outcomes of patients with lesions caused by stroke versus tumor, and carefully matched the groups on lesion images from the participants’ brain scans (Anderson et al., 1990). The researchers found that patients with strokes had more focal and more severe cognitive deficits than patients with tumors. These results demonstrate that cognitive outcomes in patients with brain damage may be influenced by the mechanism of injury per se, even when lesion size and location are controlled for (Anderson et al., 1990).

An important and unanswered question is whether mechanism of injury is important when comparing outcomes in patients with focal lesions caused by stroke versus TBI. Both stroke and TBI can result in focal lesions, although focal lesions are more common after stroke than after TBI, as the latter is more often accompanied by diffuse damage (e.g., DAI as mentioned earlier). Studies have reported that up to 80% of patients with strokes and 28% of patients with moderate/severe TBI have purely focal lesions (Ferrell & Tanev, 2002; Vos, 2011). Zhang et al. (2016) found that patients with TBI scored statistically significantly lower than patients with strokes on orientation and memory recall subtests from mental status screening tests (the Mini Mental State Exam (MMSE) and Montreal Cognitive Assessment (MoCA)) (Zhang et al., 2016). However, the patients with TBI versus stroke were not matched by lesion size and location. To our knowledge, there are no studies that have investigated cognitive outcomes in stroke versus TBI, in patients with focal lesions that were matched for size and location between the two mechanisms of injury.

The Current Study

Here, using a similar methodology to one used previously by our group (Anderson et al., 1990), we compared the cognitive profiles of patients in the chronic recovery stage, who had focal brain damage caused by either TBI or stroke. The aim of our study was to determine whether these different mechanisms of injury – TBI versus stroke – were associated with different cognitive and behavioral outcomes. As summarized above, previous studies comparing cognitive outcomes in patients with TBI versus stroke did not account for lesion size and location. To isolate mechanism of injury, we compared outcomes in TBI versus stroke when accounting for lesion location and size and other key demographic attributes. We hypothesized that there would be no statistically significant differences in the neuropsychological outcomes between the TBI versus stroke groups, when lesion location and size are controlled for.

Materials and Methods

Participants

The data used in this current study were collected with approval from the Institutional Review Board. We used data from 14 and 27 patients with TBI and stroke, respectively, from the Iowa Neurological Patient Registry. Patients who are enrolled in the Registry have focal, stable lesions, with no confounding premorbid factors such as previous neurological events, alcohol/drug abuse, psychiatric disease, or developmental disorders. The Registry is comprised mainly of patients with stroke, owing to the inclusion/exclusion criteria which include the requirement for a focal, stable lesion (excluding most TBI except for select cases of focal contusions leading to a single, focal lesion). Each TBI case was matched to two stroke cases, to increase sample size. There was one exception, where a TBI patient was matched to just one stroke patient, yielding 27 distinct patients with stroke. We were unable to find a suitable additional stroke match with the appropriate lesion volume and location, and demographics, similar to this particular TBI case. The TBI case had a small left polar frontal lesion that was difficult to find a stroke match for, and we opted to use just one stroke match in this case. The TBI sample consisted of a total of 14 (12 males, 2 females) participants and the stroke sample included 27 (17 males, 10 females) participants. The average years of education for patients with TBI was 13.64 and for patients with strokes was 12.96. The TBI and stroke samples were comprised of 12 and 26 Caucasian patients, respectively. Two patients with TBI and 1 with stroke identified their race as “Other”. The majority of the sample population was right- handed, with 2 ambidextrous patients with TBI and 1 left-handed patient with stroke. There was a statistical difference in age of lesion onset for TBI and Stroke cases, 23.25 and 50.82 years old, respectively, (p<.000). There was no difference in the education level, sex distribution, and handedness between the pathology groups (p’s >.05).

For the TBI sample, we pulled from our Patient Registry all cases in which the patient had a focal, stable brain lesion and comprehensive neuroimaging and neuropsychological data. This included 5 patients with TBI who had incurred their lesion onset during childhood, before the age of 16. This led to a mismatch in “chronicity” (time between age of lesion onset and testing) and “plasticity” between our groups, with chronicity being 14.56 years for patients with TBI and 3.03 years for patients with stroke. Importantly, all participants were studied as adults, for both neuroimaging and neuropsychological data, and to address the difference in chronicity, we reran all of the main analyses after removing the 5 patients with childhood-onset lesions due to TBI, in order to determine whether the chronicity and plasticity factors were playing a role in our results.

Measures

Table 1 includes a summary of the neuropsychological tests that were analyzed in this study. Each assessment is included in the standard battery of tests that is administered to patients who are enrolled in the Iowa Patient Registry. We chose to focus on tests that assessed cognitive domains that are typically impacted after neurological events such as TBI or stroke. All of the data used in this study have been screened for validity concerns, as patient data are excluded from the Iowa Patient Registry if performance and/or symptom validity tests are failed. Of note, patients were administered the most current version of the WAIS and WRAT at the time of their brain scan (neuropsychological and neuroimaging data were generally contemporaneous). As a result, we used various versions of these tests in the current study, such that we used data from the WRAT-R or WRAT-4, and data from the WAIS-III or WAIS-IV. Previous research has shown that different versions of these instruments are comparable and valid for measuring academic achievement and cognitive outcomes in neurological samples (Mross et al., 2020; Peterson et al., 2019; Robbins, 2014; Westmacott et al., 2009).

Table 1.

Neuropsychological Tests

Functional Domain	Neuropsychological Assessments
Attention/ Working Memory	Wechsler Adult Intelligence Scale (WAIS): Digit Span, Arithmetic, Letter-Number Sequencing
Processing speed	Trail Making Test A (TMT-A), WAIS: Symbol Search, Coding
Language	WAIS: Similarities, Information, Vocabulary; Boston Naming Test (BNT)
Visuospatial	WAIS: Block Design, Matrix Reasoning, Visual Puzzles, Picture Completion;
Executive Functioning	Trail Making Test B (TMT-B), Controlled Oral Word Association Test (COWA)
Academic Achievement/IQ	Wide Range Achievement Test- Reading; WAIS: Full Scale IQ
Memory	Rey Auditory Verbal Learning Test (Rey AVLT); Rey Complex Figure Test (RCFT)-Recall; Benton Visual Retention Test (BVRT)
Orientation	Test of Orientation: Time; Place; and Personal Information

Procedure

Lesion location and size

Stroke and TBI participants were matched based on the size and location of their lesion. For instance, we matched and grouped a TBI patient that had a unilateral lesion in the frontal lobe to 2 patients with stroke that had lesions in a similar area. For bilateral lesions, we matched 2 patients with stroke with bilateral lesions in a similar area (or in a few instances, 1 stroke patient with a unilateral lesion in one matched side and another stroke patient with a lesion in the other matched side). Participants were matched on lesion size and location using a combination of neuroanatomical expertise and computerized voxel-based matching software in Excel. Stroke matches were identified for each TBI patient by calculating the percent of voxels in the stroke lesion mask, for every stroke case in the Patient Registry, that were located within the volume of the TBI lesion mask. The number of voxels in each stroke lesion mask was divided by the TBI case’s volume to produce a percentage. Greater percentages indicated that all voxels within the stroke lesion mask fell within the bounds of the TBI lesion mask. Next, we used FSL's 3D visualization tools to visually compare the size and neuroanatomical location of the stroke and TBI lesions on standard coronal brain slices until two matches were selected. If one of the stroke cases with a higher percentage of overlap had a lesion volume that was more expansive than the TBI comparison, a second stroke match with a smaller lesion volume was selected.

Nearly half of our sample was comprised of patients with lesions within or overlapping with the frontal lobe (n=19, 46%). We included unilateral (n = 33) and bilateral lesion cases (n=8). Within the TBI sample 10 had unilateral lesions and 4 had bilateral lesions. The stroke sample had a distribution of 23 unilateral lesions and 4 bilateral lesions. Descriptions of lesion locations and volumes for each of the 14 Matched Sets (9 adult-onset TBI cases and 5 developmental onset cases and their stroke matches) can be found in Table 2. The mapped lesion images for all Matched Sets are presented in Figure 1. Twelve patients with strokes (44%) had hemorrhagic strokes and 15 (56%) suffered from ischemic strokes. The TBI and stroke groups had an average lesion volume size of 77,261 and 84,716 voxel units, respectively, which was not statistically different, p>0.05. While lesion location and the availability of neuropsychological test data were prioritized in determining which stroke matches were selected, an effort was also made to match participants on demographic attributes such as gender and education, in so far as possible.

Table 2.

Lesion locations and volumes for TBI cases and matched stroke cases

Matched Set (4-digit ID number from the Iowa Patient Registry)	Lesion Location	Lesion Volume (voxel units)
#1
TBI 1: 0376	Left basal ganglia-parietal	20652
Stroke 1a: 3589	Left basal ganglia	3569
Stroke 1b: 1001	Left frontal-temporal-parietal	205704
#2
TBI 2: 1033	Left temporal-parietal	42192
Stroke 2a: 1808	Left temporal-occipital	13441
Stroke 2b: 3053	Left temporal-parietal-occipital	113457
#3
TBI 3: 1704	Left temporal	40939
Stroke 3a: 3570	Left temporal	17986
Stroke 3b: 3476	Left temporal	48020
#4
TBI 4: 2286	Bilateral prefrontal	193449
Stroke 4a: 1983	Bilateral prefrontal	98009
Stroke 4b: 2025	Right prefrontal	48444
#5
TBI 5: 2530	Left temporal-frontal	23324
Stroke 5a: 3134	Left frontal-parietal	33693
Stroke 5b: 3348	Left temporal	10964
#6
TBI 6: 2553	Left prefrontal	7935
Stroke 6: 2313	Bilateral prefrontal	92310
#7
TBI 7: 2762	Left temporal-frontal	99757
Stroke 7a: 1038	Left temporal-frontal-basal ganglia (occipital)	205048
Stroke 7b: 1211	Left frontal	80609
#8
TBI 8: 3382	Right prefrontal-basal ganglia (small left basal ganglia)	76319
Stroke 8a: 2118	Right prefrontal	59188
Stroke 8b: 2224	Right prefrontal	125105
#9
TBI 9: 3508	Right occipital-temporal	133457
Stroke 9a: 1512	Right occipital-temporal	78279
Stroke 9b: 1658	Bilateral occipital	24409
#10
TBI 10: 2280	Right temporal-parietal	16924
Stroke 10a: 3343	Right temporal-basal ganglia-insular	66135
Stroke 10b: 0747	Right temporal-frontal-parietal-basal ganglia-insular	180146
#11
TBI 11: 2683	Right prefrontal	17214
Stroke 11a: 3886	Right prefrontal	47560
Stroke 11b: 1986	Right prefrontal	69510
#12
TBI 12: 2990	Right prefrontal	63474
Stroke 12a: 1331	Right prefrontal/frontal-basal ganglia	99565
Stroke 12b: 1725	Right prefrontal/frontal-basal ganglia	168238
#13
TBI 13: 3027	Bilateral occipital-temporal	128211
Stroke 13a: 0983	Left occipital-temporal	157008
Stroke 13b: 1103	Right occipital-temporal	142773
#14
TBI 14: 3041	Left frontal-basal ganglia	217809
Stroke 14a: 3157	Left frontal-insular	58600
Stroke 14b: 3471	Left frontal-parietal	39563

Figure 1. Lesion mapping of each Matched Set.

Note. For the patients with TBI and stroke in the 14 Matched Sets, horizontal (transverse) sections are provided, at the z-coordinates indicated in the Figure. The color red was used for mapping TBI lesions, and blue was used for mapping stroke lesions. In each Matched Set, the TBI case is shown in the first row, and the two stroke matches are shown in the rows below (Matched Set #6 was an exception, with only one stroke match available). R= right hemisphere; L= left hemisphere.

Analysis

For this study we used age adjusted scores for most neuropsychological tests, as outlined in Table 3. In each Matched Set, the scores of the two stroke matches were averaged. Then, the average (or simply just the one stroke case for the matched set mentioned previously) was compared to the respective TBI patient’s scores, using paired sample t-tests in SPSS. A Matched Set was not included in an analysis if the TBI or both stroke matches were missing a score for a particular test. We included Matched Sets of one TBI case and one stroke match if they both had a score for a particular test (i.e., if only one stroke match had a score, and the TBI patient had a score, we included the test). This resulted in varying numbers for the neuropsychological tests. This approach was taken in order to keep the sample size and power as high as possible. A formal power analysis was conducted, to determine – given our N’s – the power for detecting large effect sizes. We focused on large effect sizes based on previous work (as found in Huertas Hoyas et al. (2015) and because our interest was in effect sizes that would be clinically meaningfully and consequential to patient functioning (we were not concerned with small differences that would be of little consequence to the day-to-day functioning of a person). We used the pwr package in R software (https://cran.r-project.org/web/packages/pwr/pwr.pdf) (Huertas Hoyas et al., 2015). It was determined that our study was well powered to detect large effect sizes, with the majority (specifically, 26 out of 30) of subtests having a power estimate over 90%.

Table 3.

Neuropsychological Test Outcomes

	TBI		Stroke
Test	N	Mean (SD; Range)	N	Mean (SD; Range)	P- Value	Cohen’s d	95% CI [LL, UL]
WAIS: FSIQ	11	101.0 (21.4; 74-131)	18	100.1 (13.0; 82-124)	.894	.041	[−.551, .631]
Block Design	14	11.4 (3.3; 6-10)	27	9.5 (2.5; 5.5-16)	.056	.560	[−.014, 1.117]
Similarities	13	9.9 (3.4; 5-16)	24	10.0 (2.6; 6-15.5)	.882	.042	[−.585, .503]
Digit Span	13	8.9 (3.1; 5-15)	24	9.5 (1.5; 7-11.5)	.591	.153	[−.697, .397]
Matrix Reasoning	9	11.0 (4.2; 4-17)	14	9.4 (2.1; 6-13)	.306	.364	[−.322, 1.031]
Vocabulary	12	10.1 (3.8; 5-16)	16	10.9 (3.0; 5-15.5)	.401	.252	[−.822, .328]
Arithmetic	12	10.1 (3.4; 6-16)	22	10.8 (2.8; 7-17)	.605	.154	[−.720, .419]
Symbol Search	9	7.3 (2.1; 4-11)	13	8.5 (4.0; 4-15)	.402	.295	[−.955, .382]
Information	13	10.2 (3.9; 4-18)	25	10.4 (2.7; 7-14.5	.793	.074	[−.617, .471]
Coding	14	7.7 (3.3; 3-14)	26	7.7 (2.0; 4-11)	.990	.000	[−.524, .524]
Letter-Number Seq.	7	7.3 (3.8; 2-12)	11	10.0 (2.9; 5.5-15)	.229	.506	[−1.280, .304]
Comprehension	9	10.1 (4.2; 6-17)	14	11.8 (3.1; 5.5-16)	.334	.342	[−1.006, .341]
Picture Completion	11	9.5 (3.2; 5-15)	18	9.1 (2.5; 5.5-13)	.740	.103	[−.492, .693]
Visual Puzzles	13	10.2 (3.3; 5-15)	25	9.3 (2.1; 5.0-13)	.086	.519	[−.072, 1.090]
WRAT Reading	12	91.7 (23.6; 45-123)	20	94.4 (16.5; 76.5-116)	.646	.136	[−.702, .435]
Rey AVLT:
Trial 1	12	5.5 (1.4; 4-8)	22	5.6 (2.0; 3-9)	.824	.066	[−.631, .502]
Sum Trials 1-5	12	45.3 (12.9; 28-67)	22	42.6 (11.1; 22.5-60.5)	.566	.171	[−.403, .737]
Delayed Recall	12	8.3 (5.0; 0-15)	22	8.2 (3.2; 3.5-14)	.918	.030	[−.536, .596]
Delayed Recog	12	26.3 (5.5; 15-30)	22	27.3 (2.7; 22.5-30)	.433	.235	[−.804, .344]
BVRT:
# Correct	14	7.0 (2.8; 3-10)	26	6.4 (2.0; 3-9)	.557	.161	[−.369, .685]
# Errors	14	4.8 (4.8; 0-14)	26	5.8 (3.7; 1-14)	.576	.153	[−.677, .377]
Complex Figure Test:
Copy	14	31.6 (4.2; 22-36)	26	30.5 (2.8; 25-35.5)	.419	.039	[−.485, .563]
Delayed Recall	14	15.9 (8.6; 0-32)	26	15.4 (7.6; 2.5-30)	.885	.223	[−.312, .750]
Boston Naming Test	13	47.9 (15.9; 7-60)	23	49.7 (15.2; 4-59)	.565	.164	[−.708, .387]
COWA	14	33.9 (19.8; 13-82)	27	31.6 (12.4; 8-51.5)	.673	.115	[−.412, .639]
Trail Making Test:
Part A	11	41.6 (32.7; 20-120)	19	44.1 (19.0; 21-82)	.831	.066	[−.656, .527]
Part B	11	106.9 (92.4; 40-300)	19	106.6 (68.5; 39.5-279)	.993	.003	[−.588, .594]
Orientation:
Time	12	−2.7 (6.9; −24-0)	21	−0.3 (1.3; −4-0)	.269	.336	[−.912, .253]
Place	12	1.8 (0.6; 0-2)	21	2.0 (0.0; 2-2)	.191	.402	[−.984, .196]
Personal Info	12	3.9 (0.3; 3-4)	21	4.0 (0.0; 4-4)	.339	.289	[−.861, .296]

Note. Results were calculated using the raw score for each neuropsychological test listed in the Table, except for the WAIS-IV subtests (Age-Corrected Scaled Scores) and WRAT Reading (Standard Scores). The WAIS-IV subtests are ordered following the order in the Scoring Booklet. SD= standard deviation; WAIS IV-FSIQ = Wechsler Adult Intelligence Scale-Fourth Edition, Full Scale Intelligence Quotient; Letter Num Seq= Letter Number Sequencing; WRAT= Wide Range Achievement Test; BVRT= Benton Visual Retention Test; Delayed Recog= Delayed Recognition; COWA= Controlled Oral Word Association Test; Personal Info= personal Information; Rey AVLT= Rey Auditory-Verbal Learning Test: Sum trials1-5= the sum of raw scores across Trials 1, 2, 3, 4, and 5; CI [LL, UP]= Confidence Intervals [ Lower limit, Upper limit]

Results

Neuropsychological profiles

We conducted paired sample t-tests on the scores of each neuropsychological test for both groups (TBI v. stroke), using a Dunn–Šidák correction to the alpha level to correct for multiple comparisons (p= 0.0015). None of the outcomes was significant, and for most of the tests, the group means were very close: e.g., average FSIQ was 101.0 in the TBI group and 100.1 in the stroke group (Table 3). This example, in fact (with the mean FSIQ scores being virtually identical in the two groups) is representative of almost all of the contrasts between the groups. Nor was there a consistent pattern of outcomes favoring one group versus the other – looking across the various tests in Table 3, there are some where the mean scores were slightly better in the TBI group and some where the mean scores were slightly better in the stroke group (although as noted above, not statistically different in any of these cases). Moreover, all of the confidence intervals for the between-group comparisons included zero (Table 3). Finally, the effect sizes for the between-group contrasts were almost all in the small range (as determined by Cohen’s d’s reported in Table 3), although there were a few exceptions. These findings provide strong and consistent support for the conclusion that the TBI and stroke groups did not have different neuropsychological test performances.

We took an additional approach to address whether we could be missing anything that might differ between the TBI and stroke groups. One issue is whether the groups differ in ways that would be clinically meaningful (putting aside statistical significance per se). We had a senior board-certified clinical neuropsychologist (SWA), blinded to the study design and hypotheses, establish clinically meaningful difference scores for each neuropsychological measure used in our study (e.g., 10 points for the FSIQ scores, 3 points for the ACSS for the WAIS subtests, etc.). Using these data, equivalence testing was conducted with the two-one-sided test (TOST) procedure in order to test the equivalence of average outcomes between groups for the various neuropsychological measures. The results indicated that for the majority of the neuropsychological measures, the performances of patients with TBI versus stroke were statistically equivalent. There were a few tests that did not meet the standards of equivalence, but no consistent pattern favoring one group versus the other was found (see Table 4). Further, for most of the tests that did not meet the statistical criterion for equivalence, the difference scores were right at the edge of the confidence interval of equivalence (see Table 4). The results from the equivalence testing, in addition to the effect sizes for the between-group contrasts, support the conclusion that patients with TBI and stroke do not differ significantly on cognitive measures.

Table 4.

Equivalence test data

Test	Clinically Meaningful Difference Interval	Observed Difference 90% Confidence Interval
FSIQ*	[−10, 10]	[−10.00, 9.45]
Block Design*	[−3, 3]	[−3.47, −0.57]
Similarities	[−3, 3]	[−1.15, 1.79]
Digit Span	[−3, 3]	[−0.77, 1.84]
Matrix Reasoning*	[−3, 3]	[−3.19, 1.05]
Vocabulary*	[−3, 3]	[−0.47, 3.06]
Arithmetic	[−3, 3]	[−1.02, 2.12]
Symbol Search	[−3, 3]	[−0.37. 2.93]
Information	[−3, 3]	[−1.40, 1.97]
Coding	[−3, 3]	[−1.29, 1.47]
Letter-Number Seq.*	[−3, 3]	[0.15, 4.74]
Comprehension*	[−3, 3]	[−1.19, 3.26]
Picture Completion	[−3, 3]	[−1.93, 1.24]
Visual Puzzles	[−3, 3]	[−2.43, 0.52]
WRAT Reading*	[−10, 10]	[−8.67, 13.33]
RAVLT Trial 1	[−3, 3]	[−0.59, 1.05]
RAVLT Total 1-5*	[−8, 8]	[−8.07, 3.85]
RAVLT Delayed Recall	[−3, 3]	[−2.48, 1.91]
RAVLT Delayed Recog.	[−4, 4]	[−2.15, 0.12]
BVRT Correct	[−2, 2]	[−1.75, 0.46]
BVRT Errors*	[−3, 3]	[−0.83, 3.01]
CFT Copy	[−5, 5]	[−2.72, 0.85]
CFT Delayed Recall*	[−4, 4]	[−3.52, 3.97]
Boston Naming Test*	[−3, 3]	[−3.80, 10.21]
COWA*	[−9, 9]	[−10.35, 5.45]
Trails A*	[−7, 7]	[−13.16, 15.86]
Trails B*	[−15, 15]	[−40.56, 46.84]
Orientation Time	[−3, 3]	[−0.36, 5.15]
Orientation Place	[−1, 1]	[0.01, 0.53]
Orientation Pers. Info	[−2, 2]	[−0.03, 0.20]

Note. An asterisk (*) indicates the test that did not meet the standards of equivalence means.

We reran all of the main analysis with just the 9 adult-onset TBI cases (and their matched Stroke cases). All results were non-significant after the application of the Dunn–Šidák correction. Similarly, we reran all of the main analysis with just the 5 childhood-onset TBI cases (and their matched Stroke cases), which also yielded entirely non-significant results. The results for both supplemental analyses did include a few relatively large effect sizes. However, again, all confidence intervals included 0, and our study is sufficiently powered to detect large effects, which decreases the likelihood that our null findings are the result of Type 2 error.

Discussion

Traumatic brain injuries and strokes have different pathophysiological mechanisms, but both have the potential to cause deficits in cognitive and behavioral functioning. Previous studies have shown that the mechanism of injury might play a role in the neuropsychological outcomes of patients. Consequentially, this study aimed to compare chronic outcomes in patients with stroke versus TBI, using a battery of neuropsychological tests. To isolate the effect of mechanism of injury, we carefully controlled for lesion location and size, and to the extent possible, for various demographic attributes. The results indicated that patients with TBI and stroke performed similarly on all of the cognitive and behavioral measures. There were no significant between-group differences. This was true not only statistically speaking (all p values non-significant, all confidence intervals included 0, and most effect sizes were small), but also from a clinical perspective: the group differences were trivial and not of a magnitude that would translate into meaningful real-world differences in cognitive or behavioral aptitude.

These findings are consistent with results from Zhang et al. (2016), which showed that patients with TBI and stroke did not perform statistically differently on most neuropsychological tests (Zhang et al., 2016). The current study is the first to match the lesion size and location between patients with TBI and stroke. Not controlling for these variables introduces the possibility that etiology-based differences could be explained by lesion location and size, rather than mechanism of injury. In our study, we could isolate mechanism of injury as the critical factor differentiating the TBI and stroke groups, as we matched on lesion location and size.

We had patients in our sample with lesions in diverse brain regions, which reduces the likelihood that an etiological factor that is skewed to a specific brain area may be accounting for the findings. For instance, if most of the sample had lesions located predominately in the orbito-frontal region, there would be a possibility that a pathological factor isolated to this area may be influencing these patients to perform similarly on neuropsychological tests. Consequentially, our findings may not generalize to patients who have lesions in other parts of the brain. Thus, it is important that we included patients who had damage to varied brain regions, to ensure that we are investigating the effect of the mechanism of injury on behavior more broadly. Thus, we can be reasonably confident that the similarities we see in the performances of patients with TBI and stroke can be taken as evidence that mechanism of injury is not a critical factor in neuropsychological outcome, provided that lesion size and location are comparable between these two pathological groups.

One limitation in our study is the small sample size. We were limited by the TBI material in our Patient Registry, which is restricted to focal contusions that created a focal, stable brain lesion, and thus, the N for available TBI cases was small (14). Also, for some tests we had incomplete data. It is important to address whether low sample size might explain some of the non-significant results. However, the data in Table 3 are very compelling, and overall, simply do not support the notion that the TBI and Stroke groups have different levels of neuropsychological performance. As previously stated, the p values across all measures were non-significant. Even though moderate effect sizes were present for a few subtests, all confidence intervals included a value of 0. These findings indicate that we are likely to find non-significant results if we repeated the study. Further, a power analysis showed that our study was adequately powered to detect large effect sizes, which also decreases the likelihood of Type 2 error. An additional and separate approach, using equivalence testing, indicated that for the majority of the neuropsychological measures, the performances of patients with TBI versus stroke were statistically equivalent. Of note, clinically meaningful scores for the equivalence testing were determined by a senior, board-certified clinical neuropsychologist (blinded to the hypothesis of our study). Ultimately, the fact that the study is powered to detect large effects, in addition to this equivalence testing approach, indicates that TBI and stroke patients did not perform differently.

Future researchers may replicate this study with a larger sample size with power to detect smaller effect sizes and may find statistically significant differences in performance between these pathology groups. However, being powered to detect very small effects or minor differences between the groups may not have clinically significant implications. In any event, the results from the current study convincingly and consistently show that there are no statistically or clinically significant differences in performance across patients with TBI and strokes.

It is important to address the issue of diffuse axonal injury in the TBI patients. DAI is common in moderate and severe TBI, and we cannot exclude the possibility that this was present in some of our TBI cases. However, it was not apparent on the fine-grained structural imaging that we used (3T MRI). Also, since DAI would be expected to cause more widespread and prominent cognitive deficits in patients with TBI, this would predict greater impairment in the TBI patients compared to their matched stroke cases. We found no differences in the performance of patients with TBI versus stroke. Therefore, even if present, the effects of DAI did not appear to impact our results. Another limitation is that we did not have information regarding initial severity of injury in TBI and stroke cases. We should reiterate that patients included in our Patient Registry had to have a focal, stable lesion (an inclusion requirement); thus, it is safe to assume that the TBI cases were at least moderate in severity. Most of the stroke cases would also likely fall in the range of moderate severity per the NIH stroke scale, but again, this information was not routinely collected, and we cannot comment with certainty on the initial severity.

There were some caveats to the process we used to match patients based on lesion size and location. We used neuroanatomical expertise and computerized voxel-based matching software in Excel to determine the percent of a stroke lesion mask voxels that appear within the boundaries of the TBI lesion mask. We opted to use this method in order to prioritize lesion size over lesion location to preserve specificity in the lesion matching process. Still, with the current method, there were approximately 7 cases where large portions of the TBI or Stroke cases lesion remained unaccounted for. Further, 4 of these cases included patients with stroke or TBI with bilateral lesions. In order to address these issues, we added a second stroke case with a similar lesion volume and location to compensate for that unaccounted lesion space. We acknowledge that this method is not precise and serves as a limitation to our study, as patients with larger lesion volume tend to have worse cognitive outcomes than those with smaller/circumscribed lesions (e.g., Thye & Mirman, 2018). However, the current study findings did not show any statistically significant difference in the performance between patients with TBI or stroke on any measures.

We also attempted to match the participants based on their demographics. There were some limitations to this process as well. In some instances, participants were not perfectly matched on age, which leaves open the question of the potential impact of age differences on the performances of the patients with stroke and TBI. As discussed in the analysis section, we largely addressed this concern by analyzing age adjusted scores for most of the neuropsychological tests. Previous research conducted by Hartshorne and Germine (2015) also showed that there is no age where age of peak performance (AoPP) plateaus (Hartshorne & Germine, 2015). In other words, those findings did not support the notion that there is an age where AoPP was at highest for all cognitive tasks or most cognitive tasks. The implication of these findings is that cognitive abilities vary among individuals and are developing and changing throughout the lifespan. Therefore, categorizing cognitive abilities by age may not be as straightforward as is often believed. This finding is important and relevant for the current study, since we could not age-match perfectly, as it implies that matching based on other factors may be of more relevance than age (especially across an adult-aged sample). Previous researchers found that education was a stronger determinant of neuropsychological performance in comparison to both age and gender (Tripathi et al., 2014). In the current study, education levels between the stroke and TBI groups were very similar.

We attempted to match patients with TBI and stroke based on gender, but this matching was imperfect, and this is a limitation. TBI’s tend to impact men at a higher rate, whereas strokes affect women more often. The epidemiology of TBI and stroke, in combination with our being limited to a small subset of patients from our Patient Registry, made it impossible to match every TBI case to two stroke cases of the same gender. However, as previously stated, studies have found that education was a stronger determinant of cognitive performance in comparison to gender (Tripathi et al., 2014). Although it remains an empirical question, we would expect that the finding that patients with TBI and stroke performed comparably across neuropsychological test (when lesion location and size are controlled for) would hold for both men and women patients.

Childhood-onset cases with TBI were included in our study. Previous research has suggested that outcomes of brain injury may differ between childhood-onset and adult-onset groups, e.g., children may exhibit more neuroplasticity post-injury than adults. However, findings from this literature are very mixed, and some studies have reported that patients with adult-onset TBI fare better in terms of long-term functional outcomes than patients with developmental-onset TBI (Niedzwecki et al., 2008). Our follow-up analyses on the adult-onset cases with TBI versus their matches, and on the childhood-onset cases versus their stroke matches, indicated that in both cases, the groups with TBI and stroke performed similarly. These findings are consistent with the conclusion that the developmental TBI cases did not account for the non-significant results.

Despite these limitations, the current study provides evidence to support the hypothesis that broadly, the mechanism of injury does not produce differences in the neuropsychological outcomes of patients with focal lesions in the chronic recovery stage following TBI versus stroke when lesion location and volume are matched. We should acknowledge that the stroke and TBI samples used in our study may not be representative of the general populations of such patients. To be clear, our priority was to isolate mechanism of injury, and as such, the TBI sample is not typical of moderate/severe TBI patients, and our findings may not generalize to a more typical moderate/severe TBI population. Nonetheless, our findings suggest that in patients with TBI who develop a focal, stable lesion, it may be acceptable to combine such patients with patients with stroke to study brain-behavior relationships. It is notable that the current findings differ from those reported by Anderson et al. (1990), which indicated that patients with tumor and stroke had different cognitive outcomes and should not be combined in samples when conducting neuropsychological research.

The vast majority of the sample in the current study was Caucasian, which also serves as a potential limitation on the generalizability of the findings. For example, research has shown that non-Caucasian individuals often perform worse on neuropsychological measures, in comparison to Caucasian patients, before the application of group specific norms (Werry & Bergström, 2019). Replicating our study with a more racially/ethnically diverse sample is important.

Conclusion

In conclusion, the findings in our study were notable for entirely non-significant results, with TBI and stroke groups being statistically indistinguishable in their performance across all cognitive measures. The findings suggest that the mechanism of injury does not influence the neuropsychological outcomes in TBI and stroke patients with focal lesions, when location and size of lesion are comparable across these pathology groups. Our study was made possible by the availability of detailed neuropsychological and neuroanatomical data for patients with TBI and stroke, from our Patient Registry, which allowed the lesion matching procedure and the experimental isolation of mechanism of injury in regard to cognitive and behavioral outcomes. Future studies could replicate this research across different pathology groups, use larger sample sizes, and include more racially diverse participants. Future research could also investigate personality changes and address gaps in the literature involving differences in functional abilities between pathology groups. Overall, in order to ensure that resources are being allocated correctly and appropriate treatment is being applied, it is crucial to understand whether the mechanism of injury affects cognitive performance and outcome.