Archives of Clinical Neuropsychology 28 (2013) 798–807
Heterogeneity in Trail Making Test Performance in OEF/OIF/OND Veterans with Mild Traumatic Brain Injury Nicholas S. Thaler1,*, John F. Linck1, Daniel J. Heyanka 1, Nicholas J. Pastorek2, Brian Miller2, Jennifer Romesser3, Anita Sim 4, Daniel N. Allen 5 1
*Corresponding author at: UCLA Semel Institute for Neuroscience and Human Behavior, 760 Westwood Plaza, C8-746, Los Angeles, CA 90095. Tel.: 310-478-3711 ext 43963; fax: 310-206-8525. E-mail address: [email protected] (N. S. Thaler). Accepted 24 September 2013
Abstract This study used cluster analysis to examine variability in Trail Making Test (TMT) performance in a sample of Operation Enduring Freedom/ Operation Iraqi Freedom/Operation New Dawn (OEF/OIF/OND) veterans referred for mild traumatic brain injury (mTBI). Three clusters were extracted, two of which were characterized by level of performance and the third with a unique performance pattern characterized by slow performance on the TMT B (Low B). Clusters did not differ on demographic or psychiatric variables. The Above Average cluster had better performance on measures of processing speed, working memory, and phonemic fluency compared with the Low B cluster. Results suggest that a subset of patients with mTBI perform poorly on TMT B, which subsequently predicts poorer cognitive functioning on several other neuropsychological measures. This subset may be vulnerable to cognitive changes in the context of mTBI and multiple comorbidities while a number of other patients remain cognitively unaffected under the same circumstances. Keywords: Mild cognitive impairment; Statistical methods; Assessment
The Trail Making Test (TMT; Reitan & Wolfson, 1993) is routinely administered during the neuropsychological assessment of brain injury (Lange, Iverson, Zakrzewski, Ethel-King, & Franzen, 2005; O’Bryant, Hilsabeck, Fisher, & McCaffrey, 2003). There is consistent evidence that individuals who sustain moderate-to-severe traumatic brain injuries (TBIs) exhibit substantial impairments in both TMT Part A (TMT A) and TMT Part B (TMT B) (Heled, Hoofien, Margalit, Natovich, & Agranov, 2012; Lange et al., 2005; Thaler et al., 2012), reflecting the associated cognitive dysfunction often characteristic of these populations (Roebuck-Spencer & Sherer, 2012). However, TBI is heterogeneous and mechanism/location of injury, premorbid functioning, secondary complications, and numerous other factors can influence neurocognitive performance (Reitan & Wolfson, 1993). As neurocognitive performance is a robust indicator of current functioning and prognostic outcome (Hanks et al., 2008; Spitz, Ponsford, Rudzki, & Maller, 2012), it is critical to identify subgroups of patients who have distinct cognitive profiles that, in turn, can assist in treatment planning and patient care. This can be empirically accomplished using cluster analysis, which is a multivariate classification technique that allows for a statistical grouping of like cases into homogeneous subsets (or clusters) based on their similarity across one or more characteristics. Cluster analysis is particularly useful when there is substantial heterogeneity within a particular group of individuals on characteristics that are of interest. This statistical technique has been applied to investigate neuropsychological heterogeneity in a number of populations including TBI (Allen & Goldstein, 2013) and has confirmed distinct IQ, memory, and neuropsychological clusters are present among TBI populations which differ in functional capacity and outcome across subgroups (Allen et al., 2010; DeJong & Donders, 2010; Goldstein, Allen, & Caponigro, 2010; Malec, Machulda, & Smigielski, 1993). # The Author 2013. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: [email protected]. doi:10.1093/arclin/act080 Advance Access publication on 20 October 2013
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
Oklahoma City Department of Veteran Affairs Medical Center, Oklahoma City, OK, USA Michael E. DeBakey Department of Veteran Affairs Medical Center, Houston, TX, USA 3 George E. Wahlen Department of Veteran Affairs Medical Center, Salt Lake City, UT, USA 4 Minneapolis Department of Veteran Affairs Medical Center, Minneapolis, MN, USA 5 University of Nevada, Las Vegas, NV, USA 2
N.S. Thaler et al. / Archives of Clinical Neuropsychology 28 (2013) 798–807
799
Methods Participants Participants in this study included 78 veterans (mean age 30.1 years, SD ¼ 7.0; 93.6% men) referred for neuropsychological evaluation following completion of a Comprehensive TBI Evaluation (mean months since injury ¼ 49.6, SD ¼ 25.4) due to least one mTBI (mean injuries ¼ 2.2, SD ¼ 1.8) sustained while on active duty during an OEF/OIF/OND deployment. Participants were selected from a consecutive series of 169 cases presenting with mTBI from the Functional Outcomes Research Team (FORT) study. Participating sites included VA hospitals in Midwestern, Southern, and Western United States. An mTBI was defined based on criteria from Centers for Disease Control and Prevention (2003): an occurrence of injury to the head with at least one of the following: (i) any period of confusion, disorientation, or impaired consciousness, (ii) any period of memory
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
While most existing cluster studies have examined neurocognitive heterogeneity in moderate-to-severe TBI, relatively few have addressed such heterogeneity in mild TBI (mTBI). Neuropsychological outcome in mTBI remains an area of interest, and research examining the neurocognitive sequelae of this population has concluded that on average, mTBIs are not associated with neuropsychological impairment after a 3-month recovery period (Babikian et al., 2011; Belanger, Curtiss, Demery, Lebowitz, & Vanderploeg, 2005; Iverson & Lange, 2011). However, there has also been criticism that even when injuries are classified as mild, individual patients nonetheless exhibit heterogeneous performance on cognitive tests and meta-analytic studies obscure a subset of patients who do suffer impairment past in the 3-month period (Bigler et al., 2013; Iverson, 2010; Pertab, James, & Bigler, 2009). Rohling and colleagues (2011) and Larrabee, Binder, Rohling, and Ploetz (2013) recently responded to these criticisms by reanalyzing the data and, after controlling for statistical and methodological differences, demonstrated negligible effect size differences between patients with mTBI and controls. Larrabee and colleagues (2013) also suggested that any observed neuropsychological differences may be related to comorbid psychiatric conditions and Rohling and colleagues (2011) noted several other alternative explanations for abnormal test performances. Regardless, the presence or absence of long-term residual neurocognitive impairment in a subset of mTBI samples remains controversial in the literature. Research on TMT performance in mTBI populations has indicated that TMT B is sensitive to generalized neurocognitive impairment (Demery, Larson, Dixit, Bauer, & Perlstein, 2010; Nelson, Yoash-Gantz, Pickett, & Campbell, 2009; Sroufe et al., 2010). This is likely because TMT B assesses attentional switching and cognitive flexibility (Sa´nchez-Cubillo et al., 2009; Thaler et al., 2012), among other abilities that are often affected by neurological insult (Bogdanova & Verfaellie, 2012). The TMT’s broad sensitivity to brain impairment makes it an attractive measure to investigate with cluster analysis and identify subtypes of neurocognitive functioning. TMT B may also potentially address the criticism that some of the neuropsychological measures analyzed in prior meta-analyses may not have sufficient sensitivity to detect subtle neuropsychological impairments following mTBI (Bigler et al., 2013). Cluster analysis with the TMT has been accomplished in a sample of children who were originally classified with moderate-to-severe TBI at the time of injury and referred for evaluation 1 year later (Allen et al., 2013). Three clusters emerged which represented distinct levels of severity that subsequently predicted neuropsychological and academic outcome. The cluster associated with the most impairment was notable for exhibiting a distinctly slow TMT B completion time that deviated from the TMT A to TMT B pattern observed in the other clusters. Results indicated that patients who are initially classified with moderate-to-severe TBI at the time of injury using the Glasgow Coma Scale (GCS) had differential outcomes across academic and neuropsychological measures, although the GCS was insufficient in explaining these outcomes over time. Specifically, the GCS did not statistically differ on many variables, and visual inspection of mean scores indicated substantial overlap in performance between the moderate and severe cases. However, neuropsychological outcomes were representatively reclassified using TMT clusters, with one cluster in particular capturing patients with persistent deficits. There has yet to be a similar analysis examining subtypes of neurocognitive functioning within an mTBI sample. Such a study is warranted given the TMT’s value in identifying neurocognitive impairments in processing speed, sustained attention, and executive functions. Based on the study by Allen and colleagues (2013), it was hypothesized that clusters will emerge that differ on the level of performance, with poorer performing clusters indicative of poorer neuropsychological outcomes. If certain clusters are linked to poorer neuropsychological outcomes, then these clusters may represent subtypes of patients with mTBI that are particularly vulnerable to changes in performance in the presence of common co-morbidities given the overlapping nature of symptoms in conditions such as mTBI, chronic pain, and posttraumatic stress disorder (PTSD; Otis, Fortier, & Keane, 2012). It was also hypothesized that Trails B would best differentiate cluster performance. It is uncertain if the clusters would differ on other relevant variables, including length of injury, number of previous injuries, age, or co-morbid symptoms including PTSD and head pain. Identifying such differences would help pinpoint key variables that influence neurocognitive outcome in the poorest performing group. Alternatively, if no such differences were identified, it would suggest that a subset of patients is more vulnerable to cognitive changes regardless of other factors.
800
N.S. Thaler et al. / Archives of Clinical Neuropsychology 28 (2013) 798–807
dysfunction around the time of injury, (iii) loss of consciousness lasting ≤30 min, or (iv) neurological or neuropsychological dysfunction. Exclusion criteria for this study included missing TMT A and B data (n ¼ 1), outlier performance on the TMT A and B defined as 3.0 or more standard deviations from the sample mean (n ¼ 5), and evidence of insufficient effort toward testing on any of three symptom validity measures including the TOMM Trial 2 (n ¼ 27), the WMT Immediate Recognition Trial (n ¼ 81), and the Rey 15 test (n ¼ 11). There were no significant differences between participants who were excluded from those included on age, gender, ethnicity, years of education, or service-connected percentage. This study was approved by the respective Institutional Review Boards and VA Research and Design Committees at each VA site. Measures
Data Analysis Hierarchical cluster analysis was used with raw total seconds of the TMT A and B tests serving as attributes. A three-cluster solution was specified as this was identified as the best solution in a previous study (Allen, Thaler, Cross, & Mayfield, 2013), though two- and four-cluster solutions were also examined for comparison. Hierarchical methods join clusters sequentially through a specified clustering method that mathematically defines cluster distance based on a distance coefficient. This study used the Squared Euclidean distance coefficient, as this measure provides a measure of distance between points based in Euclidean space, and is sensitive to level and pattern of neuropsychological data (Allen et al., 2010). Ward’s method was selected as the clustering method as it is consistent with cluster methodology in other studies and is resistant to outliers (Morris, Blashfield, & Satz, 1981). The two-, three-, and four-cluster solutions were examined for stability. This was accomplished by using an alternative K-means partitional method on the cases and then comparing the hierarchical and partitional methods using Cohen’s kappa. The hierarchical cluster means served as cluster centers for the K-means method. The final cluster solution was selected based on cluster stability and the relevance and interest of additional clusters (i.e., additional clusters with few participants or similar performance levels and patterns to previous clusters were not retained). Once the final cluster solution was established, cluster membership then served as an independent variable that was compared across demographic and clinical variables. Significant differences for any such variables were then included as covariates for subsequent analyses. A cluster by TMT mixed-model ANOVA using TMT T-scores examined between- and within-subject effects of the cluster and TMT variables. Cluster effects are expected given that raw TMT scores were used as the determinants of cluster membership, and so would substantially differ on subsequent analysis. Significant main effects for TMT would be an unexpected albeit interesting finding as this would indicate that the overall means of the standardized scores significantly differed. If this were the case, there may be unexpected influences on TMT scores that affect the sample as a whole. However, significant cluster by TMT
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
Several neurocognitive measures served as external variables for the TMT clusters. Simple attention was measured with the Wechsler Adult Intelligence Scale-Fourth Edition (WAIS-IV; Wechsler, 2008) Digits Forward score, while complex attention/ working memory was measured with the WAIS-IV Digits Backward and Digits Sequence score, and the Paced Auditory Serial Addition Test (PASAT; Diehr et al., 2003; Gronwall, 1977) Trial 1 and 2 scores. Information processing speed was measured with the WAIS-IV Symbol Search and Coding subtests and the Stroop Color and Word Test Color Naming and Word Reading Trials (SCWT; Golden, Freshwater, & Golden, 2002) trials. Immediate and delayed verbal memory was examined using the California Verbal Learning Test-Second Edition (CVLT-II; Delis, Kramer, Kaplan, & Ober 2000) Trials 1 – 5, Short Delay Free Recall, and Long Delay Free Recall scores. Premorbid functioning was assessed with the Wide Range Achievement Test—4th Edition (WRAT-4; Wilkinson & Robertson, 2006), reading subtest while verbal fluency was assessed with the Controlled Oral Word Association Test (COWAT; Benton, 1967). Finally, executive functioning were measured with the Stroop Interference trial and the Wisconsin Card Sorting Test Perseverative Errors (WCST; Heaton, Chelune, Talley, Kay, & Curtiss, 1993). Current PTSD symptoms were assessed with the Posttraumatic Stress Disorder Checklist (PCL; Weathers et al., 1993). The severity of current headache pain was assessed with the Headache Impact Test (HIT-6; Bayliss et al., 2003), while symptom severity and exaggeration were assessed with the Minnesota Multiphasic Personality Inventory-2 (MMPI-2; Butcher et al., 2001), F, Fp, and fake bad scale (FBS) subscales. Effort was assessed with the Test of Memory Malingering (TOMM; Tombaugh, 1996), the Word Memory Test (Green, 2003), and the Rey-15 item test (Lezak, 1995). As stated above, failure on one or more of these measures was used to determine if participants demonstrated adequate effort to be included in the study. This criterion was used in concordance with findings that using specific subtests of effort measures are an appropriate method of screening out poor effort in research protocols (Bauer et al., 2007; O’Bryant et al., 2007).
N.S. Thaler et al. / Archives of Clinical Neuropsychology 28 (2013) 798–807
801
interaction effects would be of interest, as they would indicate that cluster membership impacts the relationship between TMT A and B scores. Clusters were then compared across neurocognitive and self-report measures using MANOVA with scores grouped by cognitive domain followed by one-way ANOVA and post-hoc Tukey tests to identify the clusters’ neuropsychological and psychiatric profiles. Results Cluster Analysis
Demographic and Psychiatric Variables The clusters did not differ in age, education, pre-morbid functioning, gender, ethnicity, months since last injury, or total number of injuries. There were no differences among the clusters across the PCL, F(2,73) ¼ 1.2, p ¼ .32 or the HIT-6 Inventory, F(2,75) ¼ 0.6, p ¼ .55. There were also no differences among the clusters across any of the MMPI-2 symptom validity raw scores, including the raw F scale, F(2,75) ¼ 1.5, p ¼ .24, the raw FBS scale, F(2,75) ¼ 0.3, p ¼ .74, or the raw Fp scale, F(2,75) ¼ 1.7, p ¼ .19 (Table 1). Neuropsychological Data A mixed-model ANOVA with TMT as the within-subjects factor and cluster as the between-subjects factor found a significant main effect for cluster, F(2,75) ¼ 47.5, p , .01, partial h 2 ¼ 0.56 but not for TMT, F(1,75) ¼ 0.6, p ¼ .42. A significant TMT by cluster interaction effect was found, F(2,75) ¼ 5.4, p , .01, partial h 2 ¼ 0.13. See Fig. 1 for a plot of the interaction. As seen in Fig. 1, the Average and Above-Average clusters have similar levels of TMT A and B performance. However, the low B cluster exhibited a notable drop in TMT B performance that explains the interaction effect. Main effects for the TMT indicated that the low B and Average cluster performed poorer on the TMT A score than the Above-Average cluster, while the low B cluster performed poor on the TMT B score compared with the Average cluster, which in turn performed poorer than the Above-Average cluster. A derived B-A score was also calculated to ascertain the degree of impairment associated with TMT B independent of TMT A. A one-way ANOVA confirmed that the Low B cluster performed much poorer than the other clusters on this derived score. MANOVAs were next run on the neuropsychological standardized scores that were not included in the original analyses. WAIS-IV subtests were converted using the WAIS-IV manual. The most recent Golden norms (Golden, Freshwater, &
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
The two-cluster solution extracted one cluster (n ¼ 51) with average performance on both TMT A and B (Average; TMT A raw ¼ 22.5 seconds, mean T-score ¼ 52.2; TMT B raw ¼ 50.5 s, mean T-score ¼ 54.2) and a second cluster (n ¼ 27) with low average performance on TMT A (Low B; raw ¼ 28.9 s, mean T-score ¼ 44.0) and well-below average performance on TMT B (raw ¼ 86.4 s, mean T-score ¼ 36.5). The three-cluster solution split the average cluster into a cluster (n ¼ 20) with above-average performance on TMT A and B (above average; TMT A raw ¼ 18.2 s, mean T-score ¼ 58.6); TMT B raw ¼ 39.4 s, T-score ¼ 61.1, and a new average cluster (n ¼ 31) that had average performance on both trails (TMT A raw ¼ 25.2 s, mean T-score ¼ 48.1; TMT B ¼ 57.8 s, mean T-score ¼ 49.8). The four-cluster solution split the cluster with well-below average performance on TMT B. One cluster (n ¼ 21) had low-average performance on TMT A (raw ¼ 29.6 s, mean T-score ¼ 43.4) and well-below average performance on TMT B (raw ¼ 81.5 s, mean T-score ¼ 36.7). The new cluster (n ¼ 6) had average performance on TMT A (raw ¼ 26.5 s, mean T-score ¼ 46.0) and well-below average performance on TMT B (raw ¼ 103.8 s, mean T-score ¼ 35.8). When compared with their partitional equivalents, all three solutions demonstrated excellent classification rates (two-cluster kappa ¼ 0.92; three-cluster kappa ¼ 0.90; four-cluster kappa ¼ 0.91). The four-cluster solution extracted an additional cluster that had only six participants, representing ,8% of the sample. The new cluster also did not differ in pattern of performance with the cluster from which it split. The three-cluster solution had clusters that each comprised a substantial proportion of the sample (40, 35, 25%), and the split of an average cluster into an above-average and average cluster is theoretically relevant. Based on these findings and those from Allen and colleagues (2013), the three-cluster solution was retained for further analyses. See Fig. 1 for raw scores and T-scores of the three-cluster solution. One-way ANOVA with post hoc Tukey tests indicated that the Above-Average cluster completed the TMT A in significantly fewer seconds than the Average or low B clusters, F(2,75) ¼ 14.6, p , .01, partial h 2 ¼ 0.28. The Above-Average cluster completed the TMT B significantly faster than the Average cluster, which in turn completed the TMT B significantly faster than the low B cluster, F(2,75) ¼ 216.9, p , .01, partial h 2 ¼ 0.85.
802
N.S. Thaler et al. / Archives of Clinical Neuropsychology 28 (2013) 798–807
Table 1. Demographic and clinical data
Years age Years education Number of injuries Months since last injurya %Service connected MMPI-2 F raw score MMPI-2 FBS raw score PCL total score HIT-6 total score
Above-average (n ¼ 20; mean [SD])
Average (n ¼ 31; mean [SD])
27.7 (4.2) 13.2 (1.3) 2.3 (2.0) 54.6 (23.5) 53.1 (34.5) 11.8 (6.9) 20.2 (4.7) 43.8 (15.8) 56.8 (11.1)
32.4 (8.0) 13.6 (1.5) 2.4 (1.8) 51.5 (27.7) 51.6 (31.8) 13.4 (8.9) 21.5 (6.7) 50.4 (17.1) 58.6 (8.9)
Low B (n ¼ 27; mean [SD]) 30.7 (6.9) 13.1 (1.9) 2.0 (1.6) 43.3 (24.0) 60.5 (33.8) 10.0 (5.5) 21.2 (6.5) 49.5 (14.7) 60.0 (10.2)
F 2.9 0.8 0.3 1.1 0.4 1.5 0.3 1.2 0.6
p .07 .47 .33 .40 .65 .24 .74 .32 .55
x2 %Male %Applying for increase %Ethnicity African American Caucasian Hispanic Other
95 31.4
87 31.4
100 37.1
4.1 1.0
.13 .61
5.0 65.0 20.0 10.0
12.9 67.7 12.9 6.5
14.8 63.0 14.8 7.4
1.7
.95
Notes: PCL ¼ Posttraumatic Stress Disorder Checklist. HIT-6 ¼ Headache Impact Test. MMPI-2 ¼ Minnesota Multiphasic Personality Inventory-2. FBS ¼ Fake Bad Scale. a Months since injury available for 65 (83.3%) of the sample. No differences emerged for incidence of missing data among clusters, x 2(2) ¼ 0.9, p ¼ .65.
Golden, 2002) were used for the Stroop raw scores, the modified PASAT short-version norms were used for the PASAT score (Diehr et al., 2003), and Heaton norms were used for the TMT and COWAT measures. See Table 2 for a comparison of scores. All measures of processing speed significantly differed, with the Above-Average cluster outperforming the other two clusters on the WAIS-IV subtests and all Stroop trials. The Above-Average cluster also outperformed the Low B cluster on the PASAT (trials 1 and 2) and on the COWAT FAS (Fig. 2). The Above-Average cluster had above-average performance on some variables (Symbol Search, Stroop Color/Word) and average performance on other variables. The Average cluster performed 0.8– 1.0 standard deviations below the mean on the Stroop Color and Word Reading trials, respectively, but was within normal limits across the other measures. The Low B cluster consistently performed between 0.3 and 1.3 standard deviations from the mean (mean range of T scores ¼ 36.7 – 46.2), with performance on the Stroop trials, the PASAT, and the COWAT FAS considered below average.
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
Fig. 1. Three-cluster performance of the TMT.
N.S. Thaler et al. / Archives of Clinical Neuropsychology 28 (2013) 798–807
803
Table 2. Neuropsychological outcomes Average (n ¼ 31; mean [SD])
Low B (n ¼ 27; mean [SD])
58.6 (8.4) 61.1 (6.8) 20.4 (0.5)
48.1 (11.7) 49.8 (6.5) 20.3 (1.0)
44.0 (.96) 36.5 (9.9) 0.6 (1.2)
9.2 (1.8)
8.9 (2.7)
8.5 (2.3)
9.1 (2.0) 11.3 (3.6) 49.5 (8.9)
9.9 (3.1) 10.1 (2.3) 47.6 (12.8)
9.2 (2.0) 9.4 (2.5) 40.9 (9.2)
12.4 (2.9) 10.8 (2.1) 50.0 (11.0) 50.4 (9.5)
10.1 (2.5) 9.3 (2.6) 40.0 (11.0) 42.3 (10.4)
8.9 (2.8) 8.2 (2.2) 36.7 (13.1) 38.0 (11.1)
52.1 (14.3) 0.6 (0.9) 0.5 (0.8)
51.2 (13.2) 0.4 (1.1) 0.3 (1.1)
54.8 (7.5) 0.6 (1.0) 0.2 (0.9
102.0 (10.7)
98.2 (12.5)
95.0 (10.4)
48.8 (6.9) 50.0 (10.8)
45.2 (9.5) 50.1 (10.4)
40.4 (9.2) 47.9 (11.1)
57.1 (8.5) 63.2 (11.1)
47.7 (10.8) 51.8 (10.0)
43.9 (10.0) 57.2 (15.7)
F
p
Partial h 2
Tukey
19.3 12.1 56.7 9.3
,.01 ,.01 ,.01 ,.01
0.36 0.24 0.60 0.20
LB, A , AA LB , A , AA LB , A, AA
0.5 2.6 0.9 2.5 4.4 3.6 10.1 6.5 7.5 8.1 0.8 0.6 0.2 0.7
.62 ,.02 .43 .09 ,.02 ,.01 ,.01 ,.01 ,.01 ,.01 .58 .51 .79 .50
2.1 2.7 5.4 0.4 6.7 10.3 5.1
.12 ,.04 ,.01 .71 ,.01 ,.01 ,.01
0.09
LB , AA
0.11 0.17 0.21 0.15 0.17 0.18
LB, A , AA LB , AA LB, A , AA LB, A , AA
0.07 0.13
LB , AA
0.16 0.22 0.12
LB, A , AA A , AA
Notes: LB ¼ Low TMT B cluster; A ¼ Average cluster; AA ¼ Above-Average Cluster; WCST ¼ Wisconsin Card Sorting Test; COWAT ¼ Controlled Oral Word Association Test; WRAT-4 ¼ Wide Range Achievement Test-4th Edition; CVLT-II ¼ California Verbal Learning Test-Second Edition; WAIS-IV ¼ Wechsler Adult Intelligence Scale-Fourth Edition.
Fig. 2. T-scores of significant neurocognitive variables.
Discussion The TMT remains an integral part of standard neuropsychological assessment for its brevity, established research base, and sensitivity to brain dysfunction. TMT B in particular is an established broad measure that is sensitive if not specific to generalized cognitive impairment (Lange et al., 2005; Sa´nchez-Cubillo et al., 2009; Thaler et al., 2012). This study is one of the first to cluster analyze TMT scores into distinct profiles that reflect general levels of cognitive functioning for mTBI, and provides results
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
Trail Making Test Trail Making A (t-score) Trail Making B (t-score) B2A (z-score) Short Term Attention WAIS-IV Digits Forward (scaled) Working Memory WAIS-IV Digits Backwards (scaled) WAIS-IV Digits Sequencing (scaled) PASAT Trials 1 + 2 (t-score) Processing Speed WAIS-IV Symbol Search (scaled) WAIS-IV Coding (scaled) Stroop Word Trial (t-score) Stroop Color Trial (t-score) Verbal Memory CVLT-II Total (t-score) CVLT-II SDFR (z-score) CVLT-II LDFR (z-score) Premorbid IQ WRAT-4 Reading (standard score) Verbal Fluency COWAT FAS (t-score) COWAT Animals (t-score) Executive Functions Stroop Color/Word (t-score) WCST Perseverative Errors (t-score)
Above-average (n ¼ 20; mean [SD])
804
N.S. Thaler et al. / Archives of Clinical Neuropsychology 28 (2013) 798–807
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
consistent with findings from a moderate-to-severe pediatric TBI sample by demonstrating that heterogeneity in TMT performance also exists in mTBI (Allen, Thaler, Cross, & Mayfield, 2013). In the current sample, three TMT clusters were identified, two are defined by similar levels of performance and a third by a deviated TMT B score. The Above-Average cluster represented individuals who scored above expectations on measures of processing speed and executive functions, as in part demonstrated by their overall better performance on the WCST compared with the Average cluster. This cluster consisted of 26% of the sample with a mean TMT A score at the 75th percentile and a mean TMT B score at the 86th percentile, which might be reasonably expected on a normal distribution curve. The Average cluster generally performed at mean levels, suggesting that this represents individuals with unremarkable neurocognitive functioning. However, patients in the Low B cluster score exhibited additional weaknesses on measures of processing speed, working memory, and verbal fluency. Patients in this cluster represented 35% of the total sample but obtained TMT B scores that were on average at the 8th percentile, a four-fold increase of prevalence than would be expected in a normal curve. Neuropsychologists often use the TMT to obtain an estimate of general neuropsychological functioning and the obtained profiles here may provide a useful reference of TMT scores that might be expected in similar populations. Patients with scores that are similar to those observed in this study may in turn have similar expected above-average, average, or below-average performance on other neuropsychological measures. From a clinical perspective, neuropsychologists may therefore note that a below-average TMT B score and comparatively spared TMT A score is a useful indicator of weaknesses in other neuropsychological tasks. For example, individuals in this cluster were on average at the 16th percentile for the COWAT FAS and the 18th percentile for the PASAT. As these tests measure more than simple processing speed, there is support that individuals in the Low TMT B cluster have deficits in other neurocognitive domains. Researchers may also find the associated weaknesses in the Stroop, WAIS-IV PSI, PASAT, and COWAT FAS as further evidence that these measure covary with the TMT B in some patients presenting with neuropsychological complaints. As these patients were assessed on average 2 –6 years after their last injury, results suggest that a subset of patients who have sustained at least one mTBI demonstrate some lower than expected neuropsychological functioning even after demonstrating good effort on three symptom validity tests. This issue has been previously raised (Iverson, 2010). This study provides some of the first evidence that TMT profiles can characterize functioning in other neurocognitive domains. As the clusters did not differ on relevant demographic, clinical, or psychiatric variables, findings suggest that the TMT and particularly the TMT B can provide valuable clinical information for veterans presenting with mTBI. In addition, there are clear subsets of patients who demonstrate average to above-average performance on nearly all neuropsychological variables, confirming what has been repeatedly shown in the literature on mTBI, that many patients presenting with mTBI years after injury demonstrate no residual cognitive deficits (Babikian et al., 2011; Belanger et al., 2005). What is less clear, however, is why a subset of individuals with no differences in demographic variables, injury characteristics, psychiatric symptoms, or ratings on a screening for headaches would perform significantly differently than others in that cohort. This may in part be a reflection of the complications involved in assessing mTBI, which incorporates retrospective interviews regarding the loss of consciousness, amnesia, and transitory neurologic symptoms that are cardinal signs of mTBI (Ruff & Jamora, 2009). Results raise several possible hypotheses regarding cognitive changes following mTBI. It is possible that the Low B cluster may represent individuals who exhibit natural variability of neuropsychological functioning independent of injury history, and so happen to have lower than average performance on constructs assessed by the TMT B. Though this is certainly possible, the unique pattern observed with a spared TMT A and deviated score on TMT B and the B-A derived score, the relatively large proportion of participants in this cluster, as well as the comparable estimated premorbid IQ across groups, do suggest that factors beyond simple normal variation are involved. Another possible hypothesis is that these patients are more vulnerable to their polytrauma comorbidities; that is, though no differences in severity of psychiatric impairments or headaches were observed, these patients respond more negatively to their comorbid headaches and/or PTSD which in turn affect their performance. There are complex relationships between deployment-related factors and current health status (Vanderploeg et al., 2012) and some patients may be more cognitively vulnerable to such factors (Larrabee et al., 2013). In line with this, it is possible that additional variables, such as medication effects, substance abuse, and comorbid medical conditions, have an interaction effect with the TMT B and associated neuropsychological functioning. Finally, the TMT B has been noted to increase visual interference and have increased response distance among stimuli, which in turn affect individuals with naturally slower psychomotor speed or visual scanning (Strauss, Sherman, & Spreen, 2006). There remains the possibility that a subset of patients continue to have residual cognitive impairment following multiple uncomplicated mTBI (Iverson, 2010). This statement must be couched with significant caution, as the deficits observed in the Low B cluster were mild and may have minimal real-world implications. Other studies have found that such residual cognitive weaknesses are not substantial in nature, if present at all (Belanger et al., 2005; Schretlen & Shapiro, 2003) though these papers have been criticized as obfuscating individual differences through group analyses and neglecting a portion of patients who do have persisting weaknesses (Bigler et al., 2013; Iverson, 2010; Pertab, James, & Bigler, 2009). This study then may address
N.S. Thaler et al. / Archives of Clinical Neuropsychology 28 (2013) 798–807
805
Funding This material is based upon work supported by the Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development; B6812C (PI- Levin HS) VA RR&D Traumatic Brain Injury Center of Excellence, Neurorehabilitation: Neurons to Networks. Conflict of Interest None declared.
References Allen, D. N., & Goldstein, G. (2013). Cluster analysis in neuropsychological research: Recent applications. New York: Springer. Allen, D. N., Leany, B. D., Thaler, N. S., Cross, C., Sutton, G. P., & Mayfield, J. (2010). Memory and attention profiles in pediatric traumatic brain injury. Archives of Clinical Neuropsychology, 25 (7), 618– 633. Allen, D. N., Thaler, N. S., Cross, C., & Mayfield, J. (2013). Classification of traumatic brain injury severity: A neuropsychological approach. In D. N. Allen, & G. G. Goldstein (Eds.), Cluster analysis in neuropsychological research: Recent applications. New York: Springer. Babikian, T., Satz, P., Zaucha, K., Light, R., Lewis, R. S., & Asarnow, R. F. (2011). The UCLA longitudinal study of neurocognitive outcomes following mild pediatric traumatic brain injury. Journal of the International Neuropsychological Society, 17 (5), 886– 895. Bauer, L., O’Bryant, S. E., Lynch, J. K., McCaffrey, R. J., & Fisher, J. M. (2007). Examining the Test of Memory Malingering trial 1 and word memory test immediate recognition as screening tools for insufficient effort. Assessment, 14 (3), 215–222. Bayliss, M. S., Dewey, J. E., Dunlap, I. I., Batenhorst, A. S., Cady, R. R., Diamond, M. L., et al. (2003). A study of the feasibility of Internet administration of a computerized health survey: The headache impact test. Quality of Life Research: An International Journal of Quality of Life Aspects of Treatment, Care and Rehabilitation, 12 (8), 953– 961. Belanger, H. G., Curtiss, G., Demery, J. A., Lebowitz, B. K., & Vanderploeg, R. D. (2005). Factors moderating neuropsychological outcomes following mild traumatic brain injury: A meta-analysis. Journal of the International Neuropsychological Society, 11 (3), 215–227. Benton, A. L. (1967). Problems of test construction in the field of aphasia. Cortex, 3, 32–58. Bigler, E. D., Farrer, T. J., Pertab, J. L., James, K., Petrie, J., & Hedges, D. W. (2013). Reaffirmed limitations of meta-analytic methods in the study of mild traumatic brain injury: A response to Rohling et al. The Clinical Neuropsychologist, 27 (2), 176–214. Bogdanova, Y., & Verfaellie, M. (2012). Cognitive sequelae of blast-induced traumatic brain injury: Recovery and rehabilitation. Neuropsychology Review, 22 (1), 4– 20.
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
such a limitation, as cluster analysis classifies cases into smaller homogeneous clusters, some which may represent lower functioning subtypes that do not emerge when combined with higher functioning groups. Rohling, Larrabee, and Millis (2012) recently argued that the effect sizes of any such subgroup would be negligible. Of interest, the differences in our groups indicate moderate-to-large effect sizes on some neuropsychological variables, particularly on the Stroop and PSI measures. Regardless, further research is required to better understanding and characterize this Low B subgroup. This cluster should be replicated across other samples of patients presenting with mTBI and externally validated with relevant variables useful for functional capacity and outcome, to confirm or refute that the observed cognitive differences provide meaningful long-term outcomes. If this cluster does represent patients who exhibit residual cognitive deficits of executive functions, these patients may exhibit weaknesses on experimental laboratory and event related potential measures. Establishing a tie between this cluster and such measures may provide evidence that cognitive dysfunction secondary to injury is present in a subset of mTBI patients despite a lack of structural neuroimaging findings. This study was limited by the unavailability of additional data such as medication use and comorbid medical conditions. As mentioned, it is possible that these may in part influence cluster membership, though the lack of differences on available demographic and clinical variables among clusters suggests any such influence is minor. Another limitation is that the TMT and particularly the TMT B is a very broad measure of cognitive functioning and not specific in the nature and severity of cognitive impairment. The aforementioned replication and external validity studies are therefore all the more crucial before any definitive statements about the nature and predictive outcome of the Low B cluster are made. Cluster analysis itself is a descriptive and datadriven statistical tool that alone cannot explain heterogeneity, but rather must be based on theory or existing studies (Cross, 2013). Continued replication of the identified TMT clusters is therefore requisite for further interpretation. Despite these limitations, these findings provide preliminary evidence that a subtype of patients referred for mTBI present with poorer performance on the TMT B, which in turn associates with other neuropsychological measures of processing speed, fluency, and executive functions. These patients may represent individuals with particular vulnerability to cognitive dysfunction, which may be related to associated comorbid conditions, natural variation, or residual impairment resultant of the brain injury.
806
N.S. Thaler et al. / Archives of Clinical Neuropsychology 28 (2013) 798–807
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
Butcher, J. N., Graham, J. R., Ben-Porath, Y. S., Tellegen, A., & Dahlstrom, G. W. (2001). Minnesota Multiphasic Personality Inventory-2 (MMPI-2): Manual for administration and scoring. Minneapolis: University of Minnesota Press. Cross, C. L. (2013). Statistical and methodological considerations when using cluster analysis in neuropsychological research. In D. N. Allen, & G. G. Goldstein (Eds.), Cluster analysis in neuropsychological research (pp. 13–35). New York: Springer. DeJong, J., & Donders, J. (2010). Cluster subtypes on the California Verbal Learning Test-Second Edition (CVLT-II) in a traumatic brain injury sample. Journal of Clinical and Experimental Neuropsychology, 32 (9), 953– 960. Delis, D. C., Kramer, J. H., Kaplan, E., & Ober, B. A. (2000). California Verbal Learning Test–Second Edition. San Antonio, TX: Psychological Corporation. Demery, J. A., Larson, M. J., Dixit, N. K., Bauer, R. M., & Perlstein, W. M. (2010). Operating characteristics of executive functioning tests following traumatic brain injury. The Clinical Neuropsychologist, 24 (8), 1292– 1308. Diehr, M. C., Cherner, M., Wolfson, T. J., Miller, S. W., Grant, I., & Heaton, R. K., & HIV Neurobehavioral Research Center, T. (2003). The 50 and 100-item short forms of the Paced Auditory Serial Addition Task (PASAT): Demographically corrected norms and comparisons with the full PASAT in normal and clinical samples. Journal of Clinical and Experimental Neuropsychology, 25 (4), 571– 585. Golden, C. J., Freshwater, S., & Golden, Z. (2002). Stroop color and word test. Chicago, IL: Stoelting. Goldstein, G., Allen, D. N., & Caponigro, J. M. (2010). A retrospective study of heterogeneity in neurocognitive profiles associated with traumatic brain injury. Brain Injury, 24 (4), 625–635. Green, P. (2003). Manual for the Word Memory Test. Edmonton, AB: Green’s Publishing (Revised 2005). Gronwall, D. M. (1977). Paced auditory serial-addition task: A measure of recovery from concussion. Perceptual and Motor Skills, 44 (2), 367– 373. Hanks, R. A., Millis, S. R., Ricker, J. H., Giacino, J. T., Nakese-Richardson, R., Frol, A. B., et al. (2008). The predictive validity of a brief inpatient neuropsychologic battery for persons with traumatic brain injury. Archives of Physical Medicine Rehabilitation, 89, 950–957. Heaton, R. K., Chelune, G. J., Talley, J. L., Kay, G. G., & Curtiss, G. (1993). Wisconsin Card Sorting Test manual: Revised and expanded. Odessa, FL: Psychological Assessment Resources. Heled, E., Hoofien, D., Margalit, D., Natovich, R., & Agranov, E. (2012). The Delis– Kaplan executive function system sorting test as an evaluative tool for executive functions after severe traumatic brain injury: A comparative study. Journal of Clinical and Experimental Neuropsychology, 34 (2), 151– 159. Iverson, G. L. (2010). Mild traumatic brain injury meta-analyses can obscure individual differences. Brain Injury, 24, 1256–1255. Iverson, G. L., & Lange, R. T. (2011). Mild traumatic brain injury. In M. R. Schoenberg, & J. G. Scott (Eds.), The little black book of neuropsychology: A syndromebased approach (pp. 697– 719). New York: Springer. doi:10.1007/978-0-387-76978-3_22. Lange, R. T., Iverson, G. L., Zakrzewski, M. J., Ethel-King, P. E., & Franzen, M. D. (2005). Interpreting the trail making test following traumatic brain injury: Comparison of traditional time scores and derived indices. Journal of Clinical and Experimental Neuropsychology, 27 (7), 897–906. Larrabee, G. J., Binder, L. M., Rohling, M. L., & Ploetz, D. M. (2013). Meta-analytic methods and the importance of non-TBI factors related to outcome in mild traumatic brain injury: Response to Bigler et al. (2013). The Clinical Neuropsychologist, 27 (2), 215–237. Lezak, M. D. (1995). Neuropsychological assessment (3rd ed.). New York: Oxford University Press. Malec, J. F., Machulda, M. M., & Smigielski, J. S. (1993). Cluster analysis of neuropsychological test results among patients with traumatic brain injury (TBI): Implications for a model of TBI-related disability. Clinical Neuropsychologist, 7 (1), 48–58. Morris, R., Blashfield, R. K., & Satz, P. (1981). Neuropsychology and cluster analysis: Potentials and problems. Journal of Clinical Neuropsychology, 3, 79–99. Nelson, L. A., Yoash-Gantz, R. E., Pickett, T. C., & Campbell, T. A. (2009). Relationship between processing speed and executive functioning performance among OEF/OIF veterans: Implications for postdeployment rehabilitation. The Journal of Head Trauma Rehabilitation, 24 (1), 32– 40. O’Bryant, S. E., Engel, L. R., Kleiner, J. S., Vasterling, J. J., & Black, F. W. (2007). Test of memory malingering (TOMM) trial 1 as a screening measure for insufficient effort. The Clinical Neuropsychologist, 21 (3), 511– 521. O’Bryant, S. E., Hilsabeck, R. C., Fisher, J. M., & McCaffrey, R. J. (2003). Utility of the Trail Making Test in the assessment of malingering in a sample of mild traumatic brain injury litigants. The Clinical Neuropsychologist, 17 (1), 69–74. Otis, J. D., Fortier, C. B., & Keane, T. M. (2012). In J. Vasterling (Ed.). PTSD and mild traumatic brain injury (pp. 105– 123). New York, NY: Guilford Press. Pertab, J. L., James, K. M., & Bigler, E. D. (2009). Limitations of mild traumatic brain injury meta-analyses. Brain Injury, 23 (6), 498– 508. Reitan, R. M., & Wolfson, D. (1993). The Halstead–Reitan neuropsychological test battery: Theory and clinical interpretations. Tucson, AZ: Neuropsychology Press. Roebuck-Spencer, T., & Sherer, M. (2012). Moderate and severe traumatic brain injury. In G. J. Larrabee (Ed.), Forensic neuropsychology: A scientific approach (2nd ed., pp. 260 –280). New York: Oxford University Press. Rohling, M. L., Binder, L. M., Demakis, G. J., Larrabee, G. J., Ploetz, D. M., & Langhinrichsen-Rohling, J. (2011). A meta-analysis of neuropsychological outcome after mild traumatic brain injury: Re-analyses and reconsiderations of Binder et al. (1997), Frencham et al. (2005), and Pertab et al. (2009). The Clinical Neuropsychologist, 25 (4), 608– 623. Rohling, M. L., Larrabee, G. J., & Millis, S. R. (2012). The “miserable minority” following mild traumatic brain injury: Who are they and do meta-analyses hide them? The Clinical Neuropsychologist, 26 (2), 197– 213. Ruff, R. M., & Jamora, C. W. (2009). Myths and mild traumatic brain injury. Psychological Injury and the Law, 2, 34– 42. Sa´nchez-Cubillo, I. I., Peria´n˜ez, J. A., Adrover-Roig, D. D., Rodrı´guez-Sa´nchez, J. M., Rı´os-Logo, M. M., Tirapu, J. J., et al. (2009). Construct validity of the Trail Making Test: Role of task-switching, working memory, inhibition/interference control, and visuomotor abilities. Journal of the International Neuropsychological Society, 15 (3), 438–450. Schretlen, D. J., & Shapiro, A. M. (2003). A quantitative review of the effects of traumatic brain injury on cognitive functioning. International Review of Psychiatry, 15, 341–349. Spitz, G., Ponsford, J. L., Rudzki, D., & Maller, J. J. (2012). Association between cognitive performance and functional outcome following traumatic brain injury: A longitudinal multilevel examination. Neuropsychology, 26 (5), 604– 612. Strauss, E., Sherman, E. M. S., & Spreen, O. (2006). A Compendium of Neuropsychological Tests: Administration, Norms, and Commentary. New York, NY: Oxford University Press. Sroufe, N. S., Fuller, D. S., West, B. T., Singal, B. M., Warschausky, S. A., & Maio, R. F. (2010). Postconcussive symptoms and neurocognitive function after mild traumatic brain injury in children. Pediatrics, 125 (6), e1331–e1339.
N.S. Thaler et al. / Archives of Clinical Neuropsychology 28 (2013) 798–807
807
Thaler, N. S., Allen, D. N., Hart, J. S., Boucher, J. R., McMurray, J. C., & Mayfield, J. (2012). Neurocognitive correlates of the Trail Making Test for older children in patients with traumatic brain injury. Archives of Clinical Neuropsychology, 27 (4), 446–452. Tombaugh, T. N. (1996). Test of Memory Malingering (TOMM). Toronto, ON: Multi-Health Systems. Vanderploeg, R. D., Belanger, H. G., Horner, R. D., Spehar, A. M., Powell-Cope, G., Luther, S. L., et al. (2012). Health outcomes associated with military deployment: Mild traumatic brain injury, blast, trauma, and combat associations in the Florida National Guard. Archives of Physical Medicine and Rehabilitation, 93 (11), 1887– 1895. Weathers, F. W., Litz, B. T., Herman, D. S., Huska, J. A., & Keane, T. M. (1993). The PTSD Checklist (PCL): Reliability, validity, and diagnostic utility. Paper presented at the Annual Meeting of the International Society for Traumatic Stress Studies, San Antonio, TX. Wechsler, D. (2008). Wechsler Adult Intelligence Scale-Fourth Edition: Administration and scoring manual. San Antonio, TX: Psychological Corporation. Wilkinson, G. S., & Robertson, G. (2006). WRAT4: Wide Range Achievement Test professional manual (4th ed.). Lutz, FL: Psychological Assessment Resources.
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
Heterogeneity in Trail Making Test Performance in OEF/OIF/OND Veterans with Mild Traumatic Brain Injury Nicholas S. Thaler1,*, John F. Linck1, Daniel J. Heyanka 1, Nicholas J. Pastorek2, Brian Miller2, Jennifer Romesser3, Anita Sim 4, Daniel N. Allen 5 1
*Corresponding author at: UCLA Semel Institute for Neuroscience and Human Behavior, 760 Westwood Plaza, C8-746, Los Angeles, CA 90095. Tel.: 310-478-3711 ext 43963; fax: 310-206-8525. E-mail address: [email protected] (N. S. Thaler). Accepted 24 September 2013
Abstract This study used cluster analysis to examine variability in Trail Making Test (TMT) performance in a sample of Operation Enduring Freedom/ Operation Iraqi Freedom/Operation New Dawn (OEF/OIF/OND) veterans referred for mild traumatic brain injury (mTBI). Three clusters were extracted, two of which were characterized by level of performance and the third with a unique performance pattern characterized by slow performance on the TMT B (Low B). Clusters did not differ on demographic or psychiatric variables. The Above Average cluster had better performance on measures of processing speed, working memory, and phonemic fluency compared with the Low B cluster. Results suggest that a subset of patients with mTBI perform poorly on TMT B, which subsequently predicts poorer cognitive functioning on several other neuropsychological measures. This subset may be vulnerable to cognitive changes in the context of mTBI and multiple comorbidities while a number of other patients remain cognitively unaffected under the same circumstances. Keywords: Mild cognitive impairment; Statistical methods; Assessment
The Trail Making Test (TMT; Reitan & Wolfson, 1993) is routinely administered during the neuropsychological assessment of brain injury (Lange, Iverson, Zakrzewski, Ethel-King, & Franzen, 2005; O’Bryant, Hilsabeck, Fisher, & McCaffrey, 2003). There is consistent evidence that individuals who sustain moderate-to-severe traumatic brain injuries (TBIs) exhibit substantial impairments in both TMT Part A (TMT A) and TMT Part B (TMT B) (Heled, Hoofien, Margalit, Natovich, & Agranov, 2012; Lange et al., 2005; Thaler et al., 2012), reflecting the associated cognitive dysfunction often characteristic of these populations (Roebuck-Spencer & Sherer, 2012). However, TBI is heterogeneous and mechanism/location of injury, premorbid functioning, secondary complications, and numerous other factors can influence neurocognitive performance (Reitan & Wolfson, 1993). As neurocognitive performance is a robust indicator of current functioning and prognostic outcome (Hanks et al., 2008; Spitz, Ponsford, Rudzki, & Maller, 2012), it is critical to identify subgroups of patients who have distinct cognitive profiles that, in turn, can assist in treatment planning and patient care. This can be empirically accomplished using cluster analysis, which is a multivariate classification technique that allows for a statistical grouping of like cases into homogeneous subsets (or clusters) based on their similarity across one or more characteristics. Cluster analysis is particularly useful when there is substantial heterogeneity within a particular group of individuals on characteristics that are of interest. This statistical technique has been applied to investigate neuropsychological heterogeneity in a number of populations including TBI (Allen & Goldstein, 2013) and has confirmed distinct IQ, memory, and neuropsychological clusters are present among TBI populations which differ in functional capacity and outcome across subgroups (Allen et al., 2010; DeJong & Donders, 2010; Goldstein, Allen, & Caponigro, 2010; Malec, Machulda, & Smigielski, 1993). # The Author 2013. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: [email protected]. doi:10.1093/arclin/act080 Advance Access publication on 20 October 2013
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
Oklahoma City Department of Veteran Affairs Medical Center, Oklahoma City, OK, USA Michael E. DeBakey Department of Veteran Affairs Medical Center, Houston, TX, USA 3 George E. Wahlen Department of Veteran Affairs Medical Center, Salt Lake City, UT, USA 4 Minneapolis Department of Veteran Affairs Medical Center, Minneapolis, MN, USA 5 University of Nevada, Las Vegas, NV, USA 2
N.S. Thaler et al. / Archives of Clinical Neuropsychology 28 (2013) 798–807
799
Methods Participants Participants in this study included 78 veterans (mean age 30.1 years, SD ¼ 7.0; 93.6% men) referred for neuropsychological evaluation following completion of a Comprehensive TBI Evaluation (mean months since injury ¼ 49.6, SD ¼ 25.4) due to least one mTBI (mean injuries ¼ 2.2, SD ¼ 1.8) sustained while on active duty during an OEF/OIF/OND deployment. Participants were selected from a consecutive series of 169 cases presenting with mTBI from the Functional Outcomes Research Team (FORT) study. Participating sites included VA hospitals in Midwestern, Southern, and Western United States. An mTBI was defined based on criteria from Centers for Disease Control and Prevention (2003): an occurrence of injury to the head with at least one of the following: (i) any period of confusion, disorientation, or impaired consciousness, (ii) any period of memory
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
While most existing cluster studies have examined neurocognitive heterogeneity in moderate-to-severe TBI, relatively few have addressed such heterogeneity in mild TBI (mTBI). Neuropsychological outcome in mTBI remains an area of interest, and research examining the neurocognitive sequelae of this population has concluded that on average, mTBIs are not associated with neuropsychological impairment after a 3-month recovery period (Babikian et al., 2011; Belanger, Curtiss, Demery, Lebowitz, & Vanderploeg, 2005; Iverson & Lange, 2011). However, there has also been criticism that even when injuries are classified as mild, individual patients nonetheless exhibit heterogeneous performance on cognitive tests and meta-analytic studies obscure a subset of patients who do suffer impairment past in the 3-month period (Bigler et al., 2013; Iverson, 2010; Pertab, James, & Bigler, 2009). Rohling and colleagues (2011) and Larrabee, Binder, Rohling, and Ploetz (2013) recently responded to these criticisms by reanalyzing the data and, after controlling for statistical and methodological differences, demonstrated negligible effect size differences between patients with mTBI and controls. Larrabee and colleagues (2013) also suggested that any observed neuropsychological differences may be related to comorbid psychiatric conditions and Rohling and colleagues (2011) noted several other alternative explanations for abnormal test performances. Regardless, the presence or absence of long-term residual neurocognitive impairment in a subset of mTBI samples remains controversial in the literature. Research on TMT performance in mTBI populations has indicated that TMT B is sensitive to generalized neurocognitive impairment (Demery, Larson, Dixit, Bauer, & Perlstein, 2010; Nelson, Yoash-Gantz, Pickett, & Campbell, 2009; Sroufe et al., 2010). This is likely because TMT B assesses attentional switching and cognitive flexibility (Sa´nchez-Cubillo et al., 2009; Thaler et al., 2012), among other abilities that are often affected by neurological insult (Bogdanova & Verfaellie, 2012). The TMT’s broad sensitivity to brain impairment makes it an attractive measure to investigate with cluster analysis and identify subtypes of neurocognitive functioning. TMT B may also potentially address the criticism that some of the neuropsychological measures analyzed in prior meta-analyses may not have sufficient sensitivity to detect subtle neuropsychological impairments following mTBI (Bigler et al., 2013). Cluster analysis with the TMT has been accomplished in a sample of children who were originally classified with moderate-to-severe TBI at the time of injury and referred for evaluation 1 year later (Allen et al., 2013). Three clusters emerged which represented distinct levels of severity that subsequently predicted neuropsychological and academic outcome. The cluster associated with the most impairment was notable for exhibiting a distinctly slow TMT B completion time that deviated from the TMT A to TMT B pattern observed in the other clusters. Results indicated that patients who are initially classified with moderate-to-severe TBI at the time of injury using the Glasgow Coma Scale (GCS) had differential outcomes across academic and neuropsychological measures, although the GCS was insufficient in explaining these outcomes over time. Specifically, the GCS did not statistically differ on many variables, and visual inspection of mean scores indicated substantial overlap in performance between the moderate and severe cases. However, neuropsychological outcomes were representatively reclassified using TMT clusters, with one cluster in particular capturing patients with persistent deficits. There has yet to be a similar analysis examining subtypes of neurocognitive functioning within an mTBI sample. Such a study is warranted given the TMT’s value in identifying neurocognitive impairments in processing speed, sustained attention, and executive functions. Based on the study by Allen and colleagues (2013), it was hypothesized that clusters will emerge that differ on the level of performance, with poorer performing clusters indicative of poorer neuropsychological outcomes. If certain clusters are linked to poorer neuropsychological outcomes, then these clusters may represent subtypes of patients with mTBI that are particularly vulnerable to changes in performance in the presence of common co-morbidities given the overlapping nature of symptoms in conditions such as mTBI, chronic pain, and posttraumatic stress disorder (PTSD; Otis, Fortier, & Keane, 2012). It was also hypothesized that Trails B would best differentiate cluster performance. It is uncertain if the clusters would differ on other relevant variables, including length of injury, number of previous injuries, age, or co-morbid symptoms including PTSD and head pain. Identifying such differences would help pinpoint key variables that influence neurocognitive outcome in the poorest performing group. Alternatively, if no such differences were identified, it would suggest that a subset of patients is more vulnerable to cognitive changes regardless of other factors.
800
N.S. Thaler et al. / Archives of Clinical Neuropsychology 28 (2013) 798–807
dysfunction around the time of injury, (iii) loss of consciousness lasting ≤30 min, or (iv) neurological or neuropsychological dysfunction. Exclusion criteria for this study included missing TMT A and B data (n ¼ 1), outlier performance on the TMT A and B defined as 3.0 or more standard deviations from the sample mean (n ¼ 5), and evidence of insufficient effort toward testing on any of three symptom validity measures including the TOMM Trial 2 (n ¼ 27), the WMT Immediate Recognition Trial (n ¼ 81), and the Rey 15 test (n ¼ 11). There were no significant differences between participants who were excluded from those included on age, gender, ethnicity, years of education, or service-connected percentage. This study was approved by the respective Institutional Review Boards and VA Research and Design Committees at each VA site. Measures
Data Analysis Hierarchical cluster analysis was used with raw total seconds of the TMT A and B tests serving as attributes. A three-cluster solution was specified as this was identified as the best solution in a previous study (Allen, Thaler, Cross, & Mayfield, 2013), though two- and four-cluster solutions were also examined for comparison. Hierarchical methods join clusters sequentially through a specified clustering method that mathematically defines cluster distance based on a distance coefficient. This study used the Squared Euclidean distance coefficient, as this measure provides a measure of distance between points based in Euclidean space, and is sensitive to level and pattern of neuropsychological data (Allen et al., 2010). Ward’s method was selected as the clustering method as it is consistent with cluster methodology in other studies and is resistant to outliers (Morris, Blashfield, & Satz, 1981). The two-, three-, and four-cluster solutions were examined for stability. This was accomplished by using an alternative K-means partitional method on the cases and then comparing the hierarchical and partitional methods using Cohen’s kappa. The hierarchical cluster means served as cluster centers for the K-means method. The final cluster solution was selected based on cluster stability and the relevance and interest of additional clusters (i.e., additional clusters with few participants or similar performance levels and patterns to previous clusters were not retained). Once the final cluster solution was established, cluster membership then served as an independent variable that was compared across demographic and clinical variables. Significant differences for any such variables were then included as covariates for subsequent analyses. A cluster by TMT mixed-model ANOVA using TMT T-scores examined between- and within-subject effects of the cluster and TMT variables. Cluster effects are expected given that raw TMT scores were used as the determinants of cluster membership, and so would substantially differ on subsequent analysis. Significant main effects for TMT would be an unexpected albeit interesting finding as this would indicate that the overall means of the standardized scores significantly differed. If this were the case, there may be unexpected influences on TMT scores that affect the sample as a whole. However, significant cluster by TMT
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
Several neurocognitive measures served as external variables for the TMT clusters. Simple attention was measured with the Wechsler Adult Intelligence Scale-Fourth Edition (WAIS-IV; Wechsler, 2008) Digits Forward score, while complex attention/ working memory was measured with the WAIS-IV Digits Backward and Digits Sequence score, and the Paced Auditory Serial Addition Test (PASAT; Diehr et al., 2003; Gronwall, 1977) Trial 1 and 2 scores. Information processing speed was measured with the WAIS-IV Symbol Search and Coding subtests and the Stroop Color and Word Test Color Naming and Word Reading Trials (SCWT; Golden, Freshwater, & Golden, 2002) trials. Immediate and delayed verbal memory was examined using the California Verbal Learning Test-Second Edition (CVLT-II; Delis, Kramer, Kaplan, & Ober 2000) Trials 1 – 5, Short Delay Free Recall, and Long Delay Free Recall scores. Premorbid functioning was assessed with the Wide Range Achievement Test—4th Edition (WRAT-4; Wilkinson & Robertson, 2006), reading subtest while verbal fluency was assessed with the Controlled Oral Word Association Test (COWAT; Benton, 1967). Finally, executive functioning were measured with the Stroop Interference trial and the Wisconsin Card Sorting Test Perseverative Errors (WCST; Heaton, Chelune, Talley, Kay, & Curtiss, 1993). Current PTSD symptoms were assessed with the Posttraumatic Stress Disorder Checklist (PCL; Weathers et al., 1993). The severity of current headache pain was assessed with the Headache Impact Test (HIT-6; Bayliss et al., 2003), while symptom severity and exaggeration were assessed with the Minnesota Multiphasic Personality Inventory-2 (MMPI-2; Butcher et al., 2001), F, Fp, and fake bad scale (FBS) subscales. Effort was assessed with the Test of Memory Malingering (TOMM; Tombaugh, 1996), the Word Memory Test (Green, 2003), and the Rey-15 item test (Lezak, 1995). As stated above, failure on one or more of these measures was used to determine if participants demonstrated adequate effort to be included in the study. This criterion was used in concordance with findings that using specific subtests of effort measures are an appropriate method of screening out poor effort in research protocols (Bauer et al., 2007; O’Bryant et al., 2007).
N.S. Thaler et al. / Archives of Clinical Neuropsychology 28 (2013) 798–807
801
interaction effects would be of interest, as they would indicate that cluster membership impacts the relationship between TMT A and B scores. Clusters were then compared across neurocognitive and self-report measures using MANOVA with scores grouped by cognitive domain followed by one-way ANOVA and post-hoc Tukey tests to identify the clusters’ neuropsychological and psychiatric profiles. Results Cluster Analysis
Demographic and Psychiatric Variables The clusters did not differ in age, education, pre-morbid functioning, gender, ethnicity, months since last injury, or total number of injuries. There were no differences among the clusters across the PCL, F(2,73) ¼ 1.2, p ¼ .32 or the HIT-6 Inventory, F(2,75) ¼ 0.6, p ¼ .55. There were also no differences among the clusters across any of the MMPI-2 symptom validity raw scores, including the raw F scale, F(2,75) ¼ 1.5, p ¼ .24, the raw FBS scale, F(2,75) ¼ 0.3, p ¼ .74, or the raw Fp scale, F(2,75) ¼ 1.7, p ¼ .19 (Table 1). Neuropsychological Data A mixed-model ANOVA with TMT as the within-subjects factor and cluster as the between-subjects factor found a significant main effect for cluster, F(2,75) ¼ 47.5, p , .01, partial h 2 ¼ 0.56 but not for TMT, F(1,75) ¼ 0.6, p ¼ .42. A significant TMT by cluster interaction effect was found, F(2,75) ¼ 5.4, p , .01, partial h 2 ¼ 0.13. See Fig. 1 for a plot of the interaction. As seen in Fig. 1, the Average and Above-Average clusters have similar levels of TMT A and B performance. However, the low B cluster exhibited a notable drop in TMT B performance that explains the interaction effect. Main effects for the TMT indicated that the low B and Average cluster performed poorer on the TMT A score than the Above-Average cluster, while the low B cluster performed poor on the TMT B score compared with the Average cluster, which in turn performed poorer than the Above-Average cluster. A derived B-A score was also calculated to ascertain the degree of impairment associated with TMT B independent of TMT A. A one-way ANOVA confirmed that the Low B cluster performed much poorer than the other clusters on this derived score. MANOVAs were next run on the neuropsychological standardized scores that were not included in the original analyses. WAIS-IV subtests were converted using the WAIS-IV manual. The most recent Golden norms (Golden, Freshwater, &
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
The two-cluster solution extracted one cluster (n ¼ 51) with average performance on both TMT A and B (Average; TMT A raw ¼ 22.5 seconds, mean T-score ¼ 52.2; TMT B raw ¼ 50.5 s, mean T-score ¼ 54.2) and a second cluster (n ¼ 27) with low average performance on TMT A (Low B; raw ¼ 28.9 s, mean T-score ¼ 44.0) and well-below average performance on TMT B (raw ¼ 86.4 s, mean T-score ¼ 36.5). The three-cluster solution split the average cluster into a cluster (n ¼ 20) with above-average performance on TMT A and B (above average; TMT A raw ¼ 18.2 s, mean T-score ¼ 58.6); TMT B raw ¼ 39.4 s, T-score ¼ 61.1, and a new average cluster (n ¼ 31) that had average performance on both trails (TMT A raw ¼ 25.2 s, mean T-score ¼ 48.1; TMT B ¼ 57.8 s, mean T-score ¼ 49.8). The four-cluster solution split the cluster with well-below average performance on TMT B. One cluster (n ¼ 21) had low-average performance on TMT A (raw ¼ 29.6 s, mean T-score ¼ 43.4) and well-below average performance on TMT B (raw ¼ 81.5 s, mean T-score ¼ 36.7). The new cluster (n ¼ 6) had average performance on TMT A (raw ¼ 26.5 s, mean T-score ¼ 46.0) and well-below average performance on TMT B (raw ¼ 103.8 s, mean T-score ¼ 35.8). When compared with their partitional equivalents, all three solutions demonstrated excellent classification rates (two-cluster kappa ¼ 0.92; three-cluster kappa ¼ 0.90; four-cluster kappa ¼ 0.91). The four-cluster solution extracted an additional cluster that had only six participants, representing ,8% of the sample. The new cluster also did not differ in pattern of performance with the cluster from which it split. The three-cluster solution had clusters that each comprised a substantial proportion of the sample (40, 35, 25%), and the split of an average cluster into an above-average and average cluster is theoretically relevant. Based on these findings and those from Allen and colleagues (2013), the three-cluster solution was retained for further analyses. See Fig. 1 for raw scores and T-scores of the three-cluster solution. One-way ANOVA with post hoc Tukey tests indicated that the Above-Average cluster completed the TMT A in significantly fewer seconds than the Average or low B clusters, F(2,75) ¼ 14.6, p , .01, partial h 2 ¼ 0.28. The Above-Average cluster completed the TMT B significantly faster than the Average cluster, which in turn completed the TMT B significantly faster than the low B cluster, F(2,75) ¼ 216.9, p , .01, partial h 2 ¼ 0.85.
802
N.S. Thaler et al. / Archives of Clinical Neuropsychology 28 (2013) 798–807
Table 1. Demographic and clinical data
Years age Years education Number of injuries Months since last injurya %Service connected MMPI-2 F raw score MMPI-2 FBS raw score PCL total score HIT-6 total score
Above-average (n ¼ 20; mean [SD])
Average (n ¼ 31; mean [SD])
27.7 (4.2) 13.2 (1.3) 2.3 (2.0) 54.6 (23.5) 53.1 (34.5) 11.8 (6.9) 20.2 (4.7) 43.8 (15.8) 56.8 (11.1)
32.4 (8.0) 13.6 (1.5) 2.4 (1.8) 51.5 (27.7) 51.6 (31.8) 13.4 (8.9) 21.5 (6.7) 50.4 (17.1) 58.6 (8.9)
Low B (n ¼ 27; mean [SD]) 30.7 (6.9) 13.1 (1.9) 2.0 (1.6) 43.3 (24.0) 60.5 (33.8) 10.0 (5.5) 21.2 (6.5) 49.5 (14.7) 60.0 (10.2)
F 2.9 0.8 0.3 1.1 0.4 1.5 0.3 1.2 0.6
p .07 .47 .33 .40 .65 .24 .74 .32 .55
x2 %Male %Applying for increase %Ethnicity African American Caucasian Hispanic Other
95 31.4
87 31.4
100 37.1
4.1 1.0
.13 .61
5.0 65.0 20.0 10.0
12.9 67.7 12.9 6.5
14.8 63.0 14.8 7.4
1.7
.95
Notes: PCL ¼ Posttraumatic Stress Disorder Checklist. HIT-6 ¼ Headache Impact Test. MMPI-2 ¼ Minnesota Multiphasic Personality Inventory-2. FBS ¼ Fake Bad Scale. a Months since injury available for 65 (83.3%) of the sample. No differences emerged for incidence of missing data among clusters, x 2(2) ¼ 0.9, p ¼ .65.
Golden, 2002) were used for the Stroop raw scores, the modified PASAT short-version norms were used for the PASAT score (Diehr et al., 2003), and Heaton norms were used for the TMT and COWAT measures. See Table 2 for a comparison of scores. All measures of processing speed significantly differed, with the Above-Average cluster outperforming the other two clusters on the WAIS-IV subtests and all Stroop trials. The Above-Average cluster also outperformed the Low B cluster on the PASAT (trials 1 and 2) and on the COWAT FAS (Fig. 2). The Above-Average cluster had above-average performance on some variables (Symbol Search, Stroop Color/Word) and average performance on other variables. The Average cluster performed 0.8– 1.0 standard deviations below the mean on the Stroop Color and Word Reading trials, respectively, but was within normal limits across the other measures. The Low B cluster consistently performed between 0.3 and 1.3 standard deviations from the mean (mean range of T scores ¼ 36.7 – 46.2), with performance on the Stroop trials, the PASAT, and the COWAT FAS considered below average.
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
Fig. 1. Three-cluster performance of the TMT.
N.S. Thaler et al. / Archives of Clinical Neuropsychology 28 (2013) 798–807
803
Table 2. Neuropsychological outcomes Average (n ¼ 31; mean [SD])
Low B (n ¼ 27; mean [SD])
58.6 (8.4) 61.1 (6.8) 20.4 (0.5)
48.1 (11.7) 49.8 (6.5) 20.3 (1.0)
44.0 (.96) 36.5 (9.9) 0.6 (1.2)
9.2 (1.8)
8.9 (2.7)
8.5 (2.3)
9.1 (2.0) 11.3 (3.6) 49.5 (8.9)
9.9 (3.1) 10.1 (2.3) 47.6 (12.8)
9.2 (2.0) 9.4 (2.5) 40.9 (9.2)
12.4 (2.9) 10.8 (2.1) 50.0 (11.0) 50.4 (9.5)
10.1 (2.5) 9.3 (2.6) 40.0 (11.0) 42.3 (10.4)
8.9 (2.8) 8.2 (2.2) 36.7 (13.1) 38.0 (11.1)
52.1 (14.3) 0.6 (0.9) 0.5 (0.8)
51.2 (13.2) 0.4 (1.1) 0.3 (1.1)
54.8 (7.5) 0.6 (1.0) 0.2 (0.9
102.0 (10.7)
98.2 (12.5)
95.0 (10.4)
48.8 (6.9) 50.0 (10.8)
45.2 (9.5) 50.1 (10.4)
40.4 (9.2) 47.9 (11.1)
57.1 (8.5) 63.2 (11.1)
47.7 (10.8) 51.8 (10.0)
43.9 (10.0) 57.2 (15.7)
F
p
Partial h 2
Tukey
19.3 12.1 56.7 9.3
,.01 ,.01 ,.01 ,.01
0.36 0.24 0.60 0.20
LB, A , AA LB , A , AA LB , A, AA
0.5 2.6 0.9 2.5 4.4 3.6 10.1 6.5 7.5 8.1 0.8 0.6 0.2 0.7
.62 ,.02 .43 .09 ,.02 ,.01 ,.01 ,.01 ,.01 ,.01 .58 .51 .79 .50
2.1 2.7 5.4 0.4 6.7 10.3 5.1
.12 ,.04 ,.01 .71 ,.01 ,.01 ,.01
0.09
LB , AA
0.11 0.17 0.21 0.15 0.17 0.18
LB, A , AA LB , AA LB, A , AA LB, A , AA
0.07 0.13
LB , AA
0.16 0.22 0.12
LB, A , AA A , AA
Notes: LB ¼ Low TMT B cluster; A ¼ Average cluster; AA ¼ Above-Average Cluster; WCST ¼ Wisconsin Card Sorting Test; COWAT ¼ Controlled Oral Word Association Test; WRAT-4 ¼ Wide Range Achievement Test-4th Edition; CVLT-II ¼ California Verbal Learning Test-Second Edition; WAIS-IV ¼ Wechsler Adult Intelligence Scale-Fourth Edition.
Fig. 2. T-scores of significant neurocognitive variables.
Discussion The TMT remains an integral part of standard neuropsychological assessment for its brevity, established research base, and sensitivity to brain dysfunction. TMT B in particular is an established broad measure that is sensitive if not specific to generalized cognitive impairment (Lange et al., 2005; Sa´nchez-Cubillo et al., 2009; Thaler et al., 2012). This study is one of the first to cluster analyze TMT scores into distinct profiles that reflect general levels of cognitive functioning for mTBI, and provides results
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
Trail Making Test Trail Making A (t-score) Trail Making B (t-score) B2A (z-score) Short Term Attention WAIS-IV Digits Forward (scaled) Working Memory WAIS-IV Digits Backwards (scaled) WAIS-IV Digits Sequencing (scaled) PASAT Trials 1 + 2 (t-score) Processing Speed WAIS-IV Symbol Search (scaled) WAIS-IV Coding (scaled) Stroop Word Trial (t-score) Stroop Color Trial (t-score) Verbal Memory CVLT-II Total (t-score) CVLT-II SDFR (z-score) CVLT-II LDFR (z-score) Premorbid IQ WRAT-4 Reading (standard score) Verbal Fluency COWAT FAS (t-score) COWAT Animals (t-score) Executive Functions Stroop Color/Word (t-score) WCST Perseverative Errors (t-score)
Above-average (n ¼ 20; mean [SD])
804
N.S. Thaler et al. / Archives of Clinical Neuropsychology 28 (2013) 798–807
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
consistent with findings from a moderate-to-severe pediatric TBI sample by demonstrating that heterogeneity in TMT performance also exists in mTBI (Allen, Thaler, Cross, & Mayfield, 2013). In the current sample, three TMT clusters were identified, two are defined by similar levels of performance and a third by a deviated TMT B score. The Above-Average cluster represented individuals who scored above expectations on measures of processing speed and executive functions, as in part demonstrated by their overall better performance on the WCST compared with the Average cluster. This cluster consisted of 26% of the sample with a mean TMT A score at the 75th percentile and a mean TMT B score at the 86th percentile, which might be reasonably expected on a normal distribution curve. The Average cluster generally performed at mean levels, suggesting that this represents individuals with unremarkable neurocognitive functioning. However, patients in the Low B cluster score exhibited additional weaknesses on measures of processing speed, working memory, and verbal fluency. Patients in this cluster represented 35% of the total sample but obtained TMT B scores that were on average at the 8th percentile, a four-fold increase of prevalence than would be expected in a normal curve. Neuropsychologists often use the TMT to obtain an estimate of general neuropsychological functioning and the obtained profiles here may provide a useful reference of TMT scores that might be expected in similar populations. Patients with scores that are similar to those observed in this study may in turn have similar expected above-average, average, or below-average performance on other neuropsychological measures. From a clinical perspective, neuropsychologists may therefore note that a below-average TMT B score and comparatively spared TMT A score is a useful indicator of weaknesses in other neuropsychological tasks. For example, individuals in this cluster were on average at the 16th percentile for the COWAT FAS and the 18th percentile for the PASAT. As these tests measure more than simple processing speed, there is support that individuals in the Low TMT B cluster have deficits in other neurocognitive domains. Researchers may also find the associated weaknesses in the Stroop, WAIS-IV PSI, PASAT, and COWAT FAS as further evidence that these measure covary with the TMT B in some patients presenting with neuropsychological complaints. As these patients were assessed on average 2 –6 years after their last injury, results suggest that a subset of patients who have sustained at least one mTBI demonstrate some lower than expected neuropsychological functioning even after demonstrating good effort on three symptom validity tests. This issue has been previously raised (Iverson, 2010). This study provides some of the first evidence that TMT profiles can characterize functioning in other neurocognitive domains. As the clusters did not differ on relevant demographic, clinical, or psychiatric variables, findings suggest that the TMT and particularly the TMT B can provide valuable clinical information for veterans presenting with mTBI. In addition, there are clear subsets of patients who demonstrate average to above-average performance on nearly all neuropsychological variables, confirming what has been repeatedly shown in the literature on mTBI, that many patients presenting with mTBI years after injury demonstrate no residual cognitive deficits (Babikian et al., 2011; Belanger et al., 2005). What is less clear, however, is why a subset of individuals with no differences in demographic variables, injury characteristics, psychiatric symptoms, or ratings on a screening for headaches would perform significantly differently than others in that cohort. This may in part be a reflection of the complications involved in assessing mTBI, which incorporates retrospective interviews regarding the loss of consciousness, amnesia, and transitory neurologic symptoms that are cardinal signs of mTBI (Ruff & Jamora, 2009). Results raise several possible hypotheses regarding cognitive changes following mTBI. It is possible that the Low B cluster may represent individuals who exhibit natural variability of neuropsychological functioning independent of injury history, and so happen to have lower than average performance on constructs assessed by the TMT B. Though this is certainly possible, the unique pattern observed with a spared TMT A and deviated score on TMT B and the B-A derived score, the relatively large proportion of participants in this cluster, as well as the comparable estimated premorbid IQ across groups, do suggest that factors beyond simple normal variation are involved. Another possible hypothesis is that these patients are more vulnerable to their polytrauma comorbidities; that is, though no differences in severity of psychiatric impairments or headaches were observed, these patients respond more negatively to their comorbid headaches and/or PTSD which in turn affect their performance. There are complex relationships between deployment-related factors and current health status (Vanderploeg et al., 2012) and some patients may be more cognitively vulnerable to such factors (Larrabee et al., 2013). In line with this, it is possible that additional variables, such as medication effects, substance abuse, and comorbid medical conditions, have an interaction effect with the TMT B and associated neuropsychological functioning. Finally, the TMT B has been noted to increase visual interference and have increased response distance among stimuli, which in turn affect individuals with naturally slower psychomotor speed or visual scanning (Strauss, Sherman, & Spreen, 2006). There remains the possibility that a subset of patients continue to have residual cognitive impairment following multiple uncomplicated mTBI (Iverson, 2010). This statement must be couched with significant caution, as the deficits observed in the Low B cluster were mild and may have minimal real-world implications. Other studies have found that such residual cognitive weaknesses are not substantial in nature, if present at all (Belanger et al., 2005; Schretlen & Shapiro, 2003) though these papers have been criticized as obfuscating individual differences through group analyses and neglecting a portion of patients who do have persisting weaknesses (Bigler et al., 2013; Iverson, 2010; Pertab, James, & Bigler, 2009). This study then may address
N.S. Thaler et al. / Archives of Clinical Neuropsychology 28 (2013) 798–807
805
Funding This material is based upon work supported by the Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development; B6812C (PI- Levin HS) VA RR&D Traumatic Brain Injury Center of Excellence, Neurorehabilitation: Neurons to Networks. Conflict of Interest None declared.
References Allen, D. N., & Goldstein, G. (2013). Cluster analysis in neuropsychological research: Recent applications. New York: Springer. Allen, D. N., Leany, B. D., Thaler, N. S., Cross, C., Sutton, G. P., & Mayfield, J. (2010). Memory and attention profiles in pediatric traumatic brain injury. Archives of Clinical Neuropsychology, 25 (7), 618– 633. Allen, D. N., Thaler, N. S., Cross, C., & Mayfield, J. (2013). Classification of traumatic brain injury severity: A neuropsychological approach. In D. N. Allen, & G. G. Goldstein (Eds.), Cluster analysis in neuropsychological research: Recent applications. New York: Springer. Babikian, T., Satz, P., Zaucha, K., Light, R., Lewis, R. S., & Asarnow, R. F. (2011). The UCLA longitudinal study of neurocognitive outcomes following mild pediatric traumatic brain injury. Journal of the International Neuropsychological Society, 17 (5), 886– 895. Bauer, L., O’Bryant, S. E., Lynch, J. K., McCaffrey, R. J., & Fisher, J. M. (2007). Examining the Test of Memory Malingering trial 1 and word memory test immediate recognition as screening tools for insufficient effort. Assessment, 14 (3), 215–222. Bayliss, M. S., Dewey, J. E., Dunlap, I. I., Batenhorst, A. S., Cady, R. R., Diamond, M. L., et al. (2003). A study of the feasibility of Internet administration of a computerized health survey: The headache impact test. Quality of Life Research: An International Journal of Quality of Life Aspects of Treatment, Care and Rehabilitation, 12 (8), 953– 961. Belanger, H. G., Curtiss, G., Demery, J. A., Lebowitz, B. K., & Vanderploeg, R. D. (2005). Factors moderating neuropsychological outcomes following mild traumatic brain injury: A meta-analysis. Journal of the International Neuropsychological Society, 11 (3), 215–227. Benton, A. L. (1967). Problems of test construction in the field of aphasia. Cortex, 3, 32–58. Bigler, E. D., Farrer, T. J., Pertab, J. L., James, K., Petrie, J., & Hedges, D. W. (2013). Reaffirmed limitations of meta-analytic methods in the study of mild traumatic brain injury: A response to Rohling et al. The Clinical Neuropsychologist, 27 (2), 176–214. Bogdanova, Y., & Verfaellie, M. (2012). Cognitive sequelae of blast-induced traumatic brain injury: Recovery and rehabilitation. Neuropsychology Review, 22 (1), 4– 20.
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
such a limitation, as cluster analysis classifies cases into smaller homogeneous clusters, some which may represent lower functioning subtypes that do not emerge when combined with higher functioning groups. Rohling, Larrabee, and Millis (2012) recently argued that the effect sizes of any such subgroup would be negligible. Of interest, the differences in our groups indicate moderate-to-large effect sizes on some neuropsychological variables, particularly on the Stroop and PSI measures. Regardless, further research is required to better understanding and characterize this Low B subgroup. This cluster should be replicated across other samples of patients presenting with mTBI and externally validated with relevant variables useful for functional capacity and outcome, to confirm or refute that the observed cognitive differences provide meaningful long-term outcomes. If this cluster does represent patients who exhibit residual cognitive deficits of executive functions, these patients may exhibit weaknesses on experimental laboratory and event related potential measures. Establishing a tie between this cluster and such measures may provide evidence that cognitive dysfunction secondary to injury is present in a subset of mTBI patients despite a lack of structural neuroimaging findings. This study was limited by the unavailability of additional data such as medication use and comorbid medical conditions. As mentioned, it is possible that these may in part influence cluster membership, though the lack of differences on available demographic and clinical variables among clusters suggests any such influence is minor. Another limitation is that the TMT and particularly the TMT B is a very broad measure of cognitive functioning and not specific in the nature and severity of cognitive impairment. The aforementioned replication and external validity studies are therefore all the more crucial before any definitive statements about the nature and predictive outcome of the Low B cluster are made. Cluster analysis itself is a descriptive and datadriven statistical tool that alone cannot explain heterogeneity, but rather must be based on theory or existing studies (Cross, 2013). Continued replication of the identified TMT clusters is therefore requisite for further interpretation. Despite these limitations, these findings provide preliminary evidence that a subtype of patients referred for mTBI present with poorer performance on the TMT B, which in turn associates with other neuropsychological measures of processing speed, fluency, and executive functions. These patients may represent individuals with particular vulnerability to cognitive dysfunction, which may be related to associated comorbid conditions, natural variation, or residual impairment resultant of the brain injury.
806
N.S. Thaler et al. / Archives of Clinical Neuropsychology 28 (2013) 798–807
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
Butcher, J. N., Graham, J. R., Ben-Porath, Y. S., Tellegen, A., & Dahlstrom, G. W. (2001). Minnesota Multiphasic Personality Inventory-2 (MMPI-2): Manual for administration and scoring. Minneapolis: University of Minnesota Press. Cross, C. L. (2013). Statistical and methodological considerations when using cluster analysis in neuropsychological research. In D. N. Allen, & G. G. Goldstein (Eds.), Cluster analysis in neuropsychological research (pp. 13–35). New York: Springer. DeJong, J., & Donders, J. (2010). Cluster subtypes on the California Verbal Learning Test-Second Edition (CVLT-II) in a traumatic brain injury sample. Journal of Clinical and Experimental Neuropsychology, 32 (9), 953– 960. Delis, D. C., Kramer, J. H., Kaplan, E., & Ober, B. A. (2000). California Verbal Learning Test–Second Edition. San Antonio, TX: Psychological Corporation. Demery, J. A., Larson, M. J., Dixit, N. K., Bauer, R. M., & Perlstein, W. M. (2010). Operating characteristics of executive functioning tests following traumatic brain injury. The Clinical Neuropsychologist, 24 (8), 1292– 1308. Diehr, M. C., Cherner, M., Wolfson, T. J., Miller, S. W., Grant, I., & Heaton, R. K., & HIV Neurobehavioral Research Center, T. (2003). The 50 and 100-item short forms of the Paced Auditory Serial Addition Task (PASAT): Demographically corrected norms and comparisons with the full PASAT in normal and clinical samples. Journal of Clinical and Experimental Neuropsychology, 25 (4), 571– 585. Golden, C. J., Freshwater, S., & Golden, Z. (2002). Stroop color and word test. Chicago, IL: Stoelting. Goldstein, G., Allen, D. N., & Caponigro, J. M. (2010). A retrospective study of heterogeneity in neurocognitive profiles associated with traumatic brain injury. Brain Injury, 24 (4), 625–635. Green, P. (2003). Manual for the Word Memory Test. Edmonton, AB: Green’s Publishing (Revised 2005). Gronwall, D. M. (1977). Paced auditory serial-addition task: A measure of recovery from concussion. Perceptual and Motor Skills, 44 (2), 367– 373. Hanks, R. A., Millis, S. R., Ricker, J. H., Giacino, J. T., Nakese-Richardson, R., Frol, A. B., et al. (2008). The predictive validity of a brief inpatient neuropsychologic battery for persons with traumatic brain injury. Archives of Physical Medicine Rehabilitation, 89, 950–957. Heaton, R. K., Chelune, G. J., Talley, J. L., Kay, G. G., & Curtiss, G. (1993). Wisconsin Card Sorting Test manual: Revised and expanded. Odessa, FL: Psychological Assessment Resources. Heled, E., Hoofien, D., Margalit, D., Natovich, R., & Agranov, E. (2012). The Delis– Kaplan executive function system sorting test as an evaluative tool for executive functions after severe traumatic brain injury: A comparative study. Journal of Clinical and Experimental Neuropsychology, 34 (2), 151– 159. Iverson, G. L. (2010). Mild traumatic brain injury meta-analyses can obscure individual differences. Brain Injury, 24, 1256–1255. Iverson, G. L., & Lange, R. T. (2011). Mild traumatic brain injury. In M. R. Schoenberg, & J. G. Scott (Eds.), The little black book of neuropsychology: A syndromebased approach (pp. 697– 719). New York: Springer. doi:10.1007/978-0-387-76978-3_22. Lange, R. T., Iverson, G. L., Zakrzewski, M. J., Ethel-King, P. E., & Franzen, M. D. (2005). Interpreting the trail making test following traumatic brain injury: Comparison of traditional time scores and derived indices. Journal of Clinical and Experimental Neuropsychology, 27 (7), 897–906. Larrabee, G. J., Binder, L. M., Rohling, M. L., & Ploetz, D. M. (2013). Meta-analytic methods and the importance of non-TBI factors related to outcome in mild traumatic brain injury: Response to Bigler et al. (2013). The Clinical Neuropsychologist, 27 (2), 215–237. Lezak, M. D. (1995). Neuropsychological assessment (3rd ed.). New York: Oxford University Press. Malec, J. F., Machulda, M. M., & Smigielski, J. S. (1993). Cluster analysis of neuropsychological test results among patients with traumatic brain injury (TBI): Implications for a model of TBI-related disability. Clinical Neuropsychologist, 7 (1), 48–58. Morris, R., Blashfield, R. K., & Satz, P. (1981). Neuropsychology and cluster analysis: Potentials and problems. Journal of Clinical Neuropsychology, 3, 79–99. Nelson, L. A., Yoash-Gantz, R. E., Pickett, T. C., & Campbell, T. A. (2009). Relationship between processing speed and executive functioning performance among OEF/OIF veterans: Implications for postdeployment rehabilitation. The Journal of Head Trauma Rehabilitation, 24 (1), 32– 40. O’Bryant, S. E., Engel, L. R., Kleiner, J. S., Vasterling, J. J., & Black, F. W. (2007). Test of memory malingering (TOMM) trial 1 as a screening measure for insufficient effort. The Clinical Neuropsychologist, 21 (3), 511– 521. O’Bryant, S. E., Hilsabeck, R. C., Fisher, J. M., & McCaffrey, R. J. (2003). Utility of the Trail Making Test in the assessment of malingering in a sample of mild traumatic brain injury litigants. The Clinical Neuropsychologist, 17 (1), 69–74. Otis, J. D., Fortier, C. B., & Keane, T. M. (2012). In J. Vasterling (Ed.). PTSD and mild traumatic brain injury (pp. 105– 123). New York, NY: Guilford Press. Pertab, J. L., James, K. M., & Bigler, E. D. (2009). Limitations of mild traumatic brain injury meta-analyses. Brain Injury, 23 (6), 498– 508. Reitan, R. M., & Wolfson, D. (1993). The Halstead–Reitan neuropsychological test battery: Theory and clinical interpretations. Tucson, AZ: Neuropsychology Press. Roebuck-Spencer, T., & Sherer, M. (2012). Moderate and severe traumatic brain injury. In G. J. Larrabee (Ed.), Forensic neuropsychology: A scientific approach (2nd ed., pp. 260 –280). New York: Oxford University Press. Rohling, M. L., Binder, L. M., Demakis, G. J., Larrabee, G. J., Ploetz, D. M., & Langhinrichsen-Rohling, J. (2011). A meta-analysis of neuropsychological outcome after mild traumatic brain injury: Re-analyses and reconsiderations of Binder et al. (1997), Frencham et al. (2005), and Pertab et al. (2009). The Clinical Neuropsychologist, 25 (4), 608– 623. Rohling, M. L., Larrabee, G. J., & Millis, S. R. (2012). The “miserable minority” following mild traumatic brain injury: Who are they and do meta-analyses hide them? The Clinical Neuropsychologist, 26 (2), 197– 213. Ruff, R. M., & Jamora, C. W. (2009). Myths and mild traumatic brain injury. Psychological Injury and the Law, 2, 34– 42. Sa´nchez-Cubillo, I. I., Peria´n˜ez, J. A., Adrover-Roig, D. D., Rodrı´guez-Sa´nchez, J. M., Rı´os-Logo, M. M., Tirapu, J. J., et al. (2009). Construct validity of the Trail Making Test: Role of task-switching, working memory, inhibition/interference control, and visuomotor abilities. Journal of the International Neuropsychological Society, 15 (3), 438–450. Schretlen, D. J., & Shapiro, A. M. (2003). A quantitative review of the effects of traumatic brain injury on cognitive functioning. International Review of Psychiatry, 15, 341–349. Spitz, G., Ponsford, J. L., Rudzki, D., & Maller, J. J. (2012). Association between cognitive performance and functional outcome following traumatic brain injury: A longitudinal multilevel examination. Neuropsychology, 26 (5), 604– 612. Strauss, E., Sherman, E. M. S., & Spreen, O. (2006). A Compendium of Neuropsychological Tests: Administration, Norms, and Commentary. New York, NY: Oxford University Press. Sroufe, N. S., Fuller, D. S., West, B. T., Singal, B. M., Warschausky, S. A., & Maio, R. F. (2010). Postconcussive symptoms and neurocognitive function after mild traumatic brain injury in children. Pediatrics, 125 (6), e1331–e1339.
N.S. Thaler et al. / Archives of Clinical Neuropsychology 28 (2013) 798–807
807
Thaler, N. S., Allen, D. N., Hart, J. S., Boucher, J. R., McMurray, J. C., & Mayfield, J. (2012). Neurocognitive correlates of the Trail Making Test for older children in patients with traumatic brain injury. Archives of Clinical Neuropsychology, 27 (4), 446–452. Tombaugh, T. N. (1996). Test of Memory Malingering (TOMM). Toronto, ON: Multi-Health Systems. Vanderploeg, R. D., Belanger, H. G., Horner, R. D., Spehar, A. M., Powell-Cope, G., Luther, S. L., et al. (2012). Health outcomes associated with military deployment: Mild traumatic brain injury, blast, trauma, and combat associations in the Florida National Guard. Archives of Physical Medicine and Rehabilitation, 93 (11), 1887– 1895. Weathers, F. W., Litz, B. T., Herman, D. S., Huska, J. A., & Keane, T. M. (1993). The PTSD Checklist (PCL): Reliability, validity, and diagnostic utility. Paper presented at the Annual Meeting of the International Society for Traumatic Stress Studies, San Antonio, TX. Wechsler, D. (2008). Wechsler Adult Intelligence Scale-Fourth Edition: Administration and scoring manual. San Antonio, TX: Psychological Corporation. Wilkinson, G. S., & Robertson, G. (2006). WRAT4: Wide Range Achievement Test professional manual (4th ed.). Lutz, FL: Psychological Assessment Resources.
Downloaded from http://acn.oxfordjournals.org/ at University of Saskatchewan on June 21, 2015
