Wisconsin Card Sorting Test as a Measure of Executive Function Impairments in Stroke Patients

APPLIED NEUROPSYCHOLOGY, 17: 267 277, 2010 Copyright # Taylor & Francis Group, LLC ISSN: 0908-4282 print=1532-4826 online DOI: 10.1080/09084282.2010.525104 Wisconsin Card Sorting Test as a Measure of Executive Function Impairments in Stroke Patients Krzysztof Jodzio and Daria Biechowska Institute of Psychology, University of Gdańsk, Gdańsk, Poland The Wisconsin Card Sorting Test (WCST) is among the most frequently administered neuropsychological tests. It is assumed that successful completion of this test requires engagement of executive functions (EF). One of the most common origins of EF impairments is ischemic stroke. The present study intends to evaluate the diagnostic use of the WCST as a measure of these impairments in poststroke patients. Forty-four patients (8 women and 36 men) who had recent unilateral stroke (22 left hemisphere, 22 right hemisphere) participated in the study. The overall accuracy of the WCST in classifying stroke survivors as having executive disorders was poor. Nevertheless, statistical analysis revealed its negative predictive power to be greater than positive predictive power (i.e., normal scores on the WCST reliably indicated the absence of executive disorders in 8 or more out of 10). Performance on the WCST is clearly influenced by severity of the executive disorders. Namely, patients with severe impairment of EF (as measured by go=no-go, fluency, and other EF tests) performed more poorly on the WCST than patients with lesser impairment or those with no impairment at all, the latter group s results being indistinguishable. In addition, this study highlights a three-factor solution to the WCST, which accounted for 90.3% of the variance. The scores that most strongly loaded on Factors 1 to 3 were, in order: percentage of conceptual-level responses, number of trials to complete the first category, and failures to maintain the set of responses. Finally, an analysis using multivariate analysis of variance, with the anterior versus posterior site and left versus right side of the lesion as independent variables, revealed a relatively weak effect of lesion location on the WCST performance. In particular, with respect to all test scores, there is only one significant interaction between the site and side of lesion was obtained (F (1,24) ¼ 4.12; p <.05; i.e., the number of categories achieved was significantly smaller after damage to the frontal lobe on the left than on the right side, whereas the laterality effect was not significant after nonfrontal lesions). In conclusion, to ascertain the cerebral substrates of poststroke executive dysfunction, there is a need to apply more accurate tests than the WCST. The study highlights the importance of a multicomponent approach to executive functioning in stroke patients. Key words: control, executive function, stroke, WCST INTRODUCTION The concept of executive functions (EF) attained a position of considerable significance in contemporary Address correspondence to Krzysztof Jodzio, Institute of Psychology, University of Gdańsk, Pomorska 68, 80-343 Gdańsk, Poland. E-mail: psydb@ug.edu.pl; psykj@ug.edu.pl neuropsychology. However, neither EF nor executive impairments have been well defined. Lezak, Howieson, and Loring (2004) describe EF as functional abilities that enable a person to program, initiate, and control one s own behaviors in an independent, purposive, and conscious way. In other words, EF enable an individual to set goals, to form plans, to initiate actions, and to regulate and evaluate behavior according to plan and

268 JODZIO & BIECHOWSKA to situational demands. In contrast to Lezak et al., EF are also considered to be a product of coordinated operation of various processes to accomplish a particular goal in a flexible manner (Funahashi, 2001). Despite the apparent uncertainty concerning the exact nature of executive (dys-)functions (for review, see Goldberg, 2001; Royall et al., 2002), it is generally accepted that EF are higher-level functions that integrate and control more basic cognitive processes. Although impairment in EF tends to show up globally, affecting various cognitive domains, it typically involves a cluster of symptoms with one or two being predominant. It has been shown, for instance, that approximately 50% of first-time stroke patients experience executive problems, which are heterogeneous with respect to severity and specificity (Jodzio, Biechowska, & Gąsecki, 2008; Zinn, Bosworth, Hoenig, & Swartzwelder, 2007). Originally, the term dysexecutive syndrome was introduced in the 1980s to describe EF impairments and to replace the common, yet imprecise concept of frontal syndrome (Jodzio, 2008). The frontal syndrome in neuropsychiatric taxonomies emphasized alterations in personality, as well as abnormalities of emotional and social behavior, while in neuropsychological taxonomies, frontal syndrome emphasized associated cognitive rather than emotional disorders. Patients with damage to the dorsolateral prefrontal cortex (DLPFC) most often exhibited executive cognitive deficits, while patients with orbitofrontal damage revealed personality and emotional changes. DLPFC pathology is often associated with deficits in set shifting, sequencing, planning, and inhibition. In contrast, orbitofrontal pathology is most often associated with deficits in emotion recognition, changes in evaluation of social information, and impulse disinhibition (Cummings & Mega, 2003; Rankin, 2007). Although both dorsolateral and orbitofrontal pathology are associated with distinct executive control syndromes, it is clear that EF impairments have been discovered in patients with nonfrontal lesions involving, for instance, the temporal lobes, some subcortical structures (e.g., the basal ganglia, thalamus), and the cerebellum (Godefroy & Stuss, 2007; Royall et al., 2002). The term dysexecutive impairment(s), which is commonly used as shorthand to refer to symptoms of frontal-lobe pathology (Funahashi, 2001; Goldberg, 2001), can therefore be misleading. In fact, the dysexecutive syndrome is a relatively new diagnostic entity that includes a variety of symptoms of disorganized behavior, such as loss of ability to program, initiate, and monitor ongoing actions. One of the most common origins of dysexecutive impairment, regardless of its mechanism, is ischemic stroke. This common neurological event consists of a disruption of the vascular supply to the brain, of rapid onset with neurological symptoms persisting longer than 24 hours, caused by arterial occlusion through progressive atherosclerosis or embolism. A high frequency of executive dysfunction has been observed in patients who sustain infarcts in the territory of the middle and=or anterior cerebral arteries and in the anterior thalamus (Jodzio, 2008; Vataja et al., 2003; Zinn et al., 2007). Other stroke-related sources of executive dysfunction should be taken into account as well (e.g., subarachnoid hemorrhage, lacunar infarcts). The majority of researchers argue (e.g., Benge, Caroselli, & Temple, 2007; Godefroy & Stuss, 2007; Ownsworth & Shum, 2008; Royall et al., 2002) that EF impairments represent a core deficit of poststroke disability in instrumental activities of daily living because the presence of executive disorders on psychometric tests is associated with a poor functional outcome. These disorders may be particularly indicated by tests that require cognitive flexibility (i.e., rule detection, maintenance, and shifting of rules). On formal testing, many patients who had a stroke commonly fail to perform well on the Wisconsin Card Sorting Test (WCST) and other similar tests (for description, see Lezak et al., 2004) that require subjects to shift response set in relation to external cues (i.e., set switching). Other possible reasons of defective performance on the WCST are inability to sustain attention and maintain set due to stimulus interference (i.e., set maintenance), concept formation deficit, impaired working memory, and lack of problem-solving capability. These multiple potential sources of performance deficits on the WCST suggest that the test itself requires application of several distinct cognitive skill sets for its successful performance. There is evidence for a multifractional dimensional structure of the WCST; however, this thesis raises controversy (Barceló & Knight, 2002). The potential multidimensional structure of the WCST makes it difficult to identify the exact cognitive and executive abilities that are necessary for successful performance. In addition, most researchers who employ the WCST in their studies of brain-damaged populations report data for only few (from two to four) WCST subscales, thus rendering it incomplete (Royall et al., 2002). Surprisingly, the WCST perhaps is the best-described EF test, because it has been validated in many lesion and neuroimaging studies. It is also unclear whether poor performance on the WCST is linked only to prefrontal dysfunction. A classical study done more than 40 years ago by Brenda Milner (1963; according to Burgess et al., 2006) showed that patients with DLPFC lesions performed worse on the test than those with either orbitofrontal or posterior cortex lesions. Specifically, only patients with damage to the DLPFC evidenced set shifting and perseverative errors. However, the specificity of the WCST as a marker of frontal-lobe pathology remains controversial in light of the more recent findings, because EF impairments have

WCST IN STROKE PATIENTS 269 also been reported following nonfrontal lesions and diffuse cerebral disease, such as the degenerative dementias (Baudic et al., 2006; Duke & Kaszniak, 2000; Harciarek & Jodzio, 2005). For example, almost 50% of patients with temporal lobe epilepsy have been reported to display impaired performance on the WCST (Kim, Lee, Yoo, Kang, & Lee, 2007). Davidson, Gao, Mason, Winocur, and Anderson (2008) found no correlation between lesion locus and performance on the WCST in neurosurgical patients. Similarly, frontal-lobe substrates for executive functioning measured by WCST seem questionable in light of recent findings published by Hashimoto, Uruma, and Abo (2008). These researchers examined changes in circulation of hemoglobin in patients with traumatic brain injury (TBI) using near-infrared spectroscopy. They detected abnormality of frontal-lobe activation in TBI patients who nevertheless displayed normal WCST scores. Finally, studies of patients with stroke have suggested that performance on the WCST is susceptible to frontal-striatal circuit lesions, rather than frontal dysfunction per se. Eslinger and Grattan (1993), for instance, observed poststroke executive impairment following selective subcortical damage. Frontal-lobe and striatum damage each caused a similar degree of impairment of WCST performance, although both groups performed at a significantly lower level than patients with damage to the posterior cortex and healthy individuals. This finding is partially consistent with data from positron emission tomography activation studies in patients with Parkinson s disease and functional neuroimaging studies in healthy adults, indicating that at least some executive processes (e.g., planning, motor control) are associated with frontal-striatal circuit functions (for review, see Saint-Cyr, 2003; see also Dagher, Owen, Boecker, & Brooks, 2001). Further evidence that the WCST is a nonspecific measure of executive dysfunction in ischemic stroke is provided by studies which failed to demonstrate significant differences on WCST measures between frontal and nonfrontal stroke survivors (e.g., Alvarez & Emory, 2006; Leskelä et al., 1999). It is also worth noting that executive disorders are often found in patients whose infarct lesion does not affect the frontal lobes (Vataja et al., 2003). Despite the evidence that executive impairments may occur without direct damage to the prefrontal lobes, it is nevertheless true that whenever prefrontal sites are impaired, one is likely to see deficits in EF, including impairments of one kind or another on the WCST. What needs to be clarified is the application of the WCST and its subscales as a measure of EF impairments in patients with vascular unilateral brain damage. We evaluated the accuracy of this test in classifying stroke survivors as having executive disorders. We also sought to discover whether selected clinical parameters (i.e., severity of executive disorders and days postonset) have an effect on the WCST performance. In addition, we investigated the factor structure of the WCST using all test parameters. Finally, the present study sought to investigate the effect of lesion side (left vs. right) and site (anterior vs. posterior) on the WCST scores. Participants METHODS Forty-four patients (8 women and 36 men), all of whom had previously incurred unilateral ischemic stroke, have been examined for the purpose of this study. The patients were aged 20 to 79 years, with an average age of 56 years (SD ¼ 15). Patients had an average of 12 years of education (SD ¼ 3); all were right-handed and native speakers of Polish. All patients met the following selection criteria: (1) recent stroke, as clinically determined by a neurologist; (2) a single cerebral infarction restricted only to one hemisphere; (3) no history of cerebral disease or disorder prior to the current stroke; (4) no signs of achromatopsia; (5) hearing adequate for completion of neuropsychological tasks; and (6) no more than 30 days postonset at the time of inclusion in the study (in days, M ¼ 10, SD ¼ 5; range 3 25). In addition, patients had no known history of other significant medical disease such as psychiatric disorder, progressive dementia, substance abuse, or additional neurological events (e.g., head injury). Subjects with aphasia were excluded as well. In all cases, the lesion location was established on the basis of computed tomography (CT) and=or magnetic resonance imaging (MRI) scans. Noticeably, most patients had a common infarction in both cortical and subcortical areas. Twenty-two patients had right-sided and 22 had left-sided lesions. Of the patients, 13 had solely frontal (i.e., anterior) and 15 had solely nonfrontal (i.e., posterior) lesions. The remainder of the patients stroke was caused by extensive damage throughout the large portion of the left or right hemisphere, involving both anterior and posterior areas. A more detailed description of the lesion sites was not possible on the basis of the available data. All individuals were recruited from the Department of Neurology at the Medical University of Gdańsk in Poland between December 2005 and November 2008. Before testing, all participants gave written consent following detailed explanation of the procedure. Each person was tested individually in a quiet testing room. Twenty-four patients presented with EF impairments of ranging severity. In particular, 8 subjects had mild impairment, whereas 16 patients presented with moderate or severe executive deficits. The methods used for

270 JODZIO & BIECHOWSKA detecting the executive impairments are described in detail in the report by Jodzio et al. (2008). To outline, evaluation of severity of dysexecutive symptoms was based on performance on three specific tasks (i.e., the Word-Fluency Test [phonetic criterion], the Trail- Making Test, part B, derived from the Halstead-Reitan Battery, and the Go=no-go task [for description, see Lezak et al., 2004]). For the purpose of this study, these tasks were only used to classify the patients as having impaired or normal EF (see Table 1). In other words, performance on these tasks represented an objective diagnostic criterion for evaluation of the executive functioning. The presence of EF impairments was determined if the patient received at least one abnormal score on any test mentioned above. Materials The assessment was carried out using the WCST. Its Polish edition, which is based on the revised and expanded manual by Heaton, Chelune, Talley, Kay, and Curtiss (1993), was put together by Jaworowska (2002), who included a manual consisting of detailed scoring instructions and average results for ages 21 to 79 years for the full pack (128 cards) with the four stimulus cards set up. To our knowledge, the present study is the first one to evaluate the usefulness of the Polish version of the WCST in patients with brain damage, given the fact that Jaworowska s study involved healthy participants only. Briefly, the WCST test consists of 128 paper cards containing geometric designs that vary in color, form, and number. The subject is given four cards and then asked to sort the remaining deck of cards by color, form, or number but is not instructed as to how to do that. Thus, the subject is required to infer the correct sorting principles from limited feedback from the examiner, who only tells the participant whether the sorting is correct or incorrect. Subjects examined for our purposes TABLE 1 The WCST Performance Comparison Between Stroke Patients (N ¼ 44) Differ on Executive Functioning and Probabilities Indicating the Diagnostic Efficiency of the Test Executive Functions WCST Parameters (Scores) Impaired (N ¼ 24) Normal (N ¼ 20) Chi-square=p Value PPP NPP CCR Total Errors: -abnormal 5 a 5 0.11=0.74 21 75 45 -normal 19 15 b Perseverative Errors: -abnormal 4 a 5 17 75 43 -normal 20 15 b Nonperseverative Errors: -abnormal 4 a 2 17 90 50 -normal 20 18 b Categories Achieved: -abnormal 6 a 2 25 90 54 -normal 18 18 b Failure to Maintain Set: -abnormal 5 a 3 20 85 50 -normal 19 17 b Perseverative Responses: -abnormal 4 a 4 17 80 45 -normal 20 16 b Percentage of Conceptual-Level Responses: -abnormal 9 a 5 0.79=0.37 37 75 54 -normal 15 15 b Trials to Complete the First Category: -abnormal 7 a 1 29 95 59 -normal 17 19 b Learning to Learn: -abnormal 9 a 7 0.03=0.86 37 65 50 -normal 15 13 b Mean: 24 81 50 Note. WCST ¼ Wisconsin Card Sorting Test. Values in the second and third columns from the left refer to the number of patients in each subgroup. Two values marked a or b superscript were used to calculate probabilities presented in columns on the right PPP (Positive Predictive Power), NPP (Negative Predictive Power), and CCR (overall correct classification rates indicating average percentage of patients diagnosed correctly as having impaired or normal executive functions), according to the formulas: PPP ¼ (value a =24) 100; NPP ¼ (value b =20) 100; CCR ¼ [(value a þ value b )=44] 100.

TABLE 2 Performance on the WCST of Stroke Patients as a Function of Severity of Executive Disorders: Mean Values (Raw Scores) and SDs Executive Functions WCST IN STROKE PATIENTS 271 Mild Impairment (N ¼ 8) Moderate-to-Severe Impairment (N ¼ 15) Normal (N ¼ 20) WCST Parameters M SD M SD M SD F=p Value Total Errors 40.75 a 22.78 62.47 b 23.26 42.20 a 23.94 3.79=.03 Perseverative Errors 23.25 a 15.45 35.40 a 20.19 24.55 a 21.46 1.54=.23 Nonperseverative Errors 17.50 a 9.39 27.00 a 16.97 17.55 a 8.90 2.84=.07 Categories Achieved 4.50 a 1.69 2.13 b 2.29 4.40 a 1.90 6.34=.004 Failure to Maintain Set 0.87 a 1.36 1.13 a 1.35 1.10 a 1.44 0.10=.91 Perseverative Responses 27.37 a 17.97 41.53 a 26.63 27.80 a 27.79 1.40=.26 Percentage of Conceptual-Level Responses 52.87 a 18.86 33.40 b 23.01 53.40 a 21.81 4.06=.02 Trials to Complete the First Category 25.62 a 21.92 62.00 b 53.21 19.75 a 26.24 5.80=.006 Learning to Learn 1.65 a 11.50 1.85 a 9.49 6.49 a 9.02 0.97=.39 Note. WCST ¼ Wisconsin Card Sorting Test. Differences between means marked a superscript and means marked b superscript are significant at p <.05. completed the WCST according to standardized instructions. The scores we used included the raw scores of: (1) total errors, (2) perseverative errors, (3) nonperseverative errors, (4) the total categories achieved, (5) failures to maintain the set of responses, (6) perseverative responses, (7) percentage of conceptual-level responses, (8) number of trials to complete the first category, and (9) learning to learn (see Tables 1 and 2). These scores are recommended for clinical interpretation (Lezak et al., 2004). RESULTS The Diagnostic Efficiency and Predictive Power of the WCST On the basis of the performance on selected executive tests (see Participants section) and the WCST, we divided the patients into groups as having either normal or abnormal EF. The results of our analysis for each of the major WCST subscale parameters are shown in Table 1. Unfortunately, small sample sizes in the four subgroups prohibited statistical analysis, with the exception of total errors, percentage of conceptual-level responses, and learning to learn. Correlations between these WCST scores and the prevalence of executive impairment did not reach statistical significance (ps >.37). This is to say, the chi-square test on the contingency table was nonsignificant, indicating that the groups (i.e., patients with impaired EF vs. patients with retained EF) were not significantly different in their frequency of patients having abnormal and normal WCST scores. Percentage of patients who obtained abnormal scores for the WCST ranged from 14% to 36%. Because problems appeared in the exploratory study, additional analysis was completed taking into account the diagnostic efficiency statistics known as positive and negative predictive power (PPP, NPP). These probabilities evaluate the usefulness of a given test in clinical practice for making diagnostic decisions about patients (Grodzinsky & Barkley, 1999; see also Duff et al., 2008; Elwood, 1993). According to Grodzinsky and Barkley, PPP refers to the probability of having the diagnosis given the presence of an abnormal score on the test and NPP indicates the probability of not having the diagnosis given the presence of a normal score on the test. We used these conditional probabilities in conjunction with the overall classification rate separately for each WCST score. The results set out in Table 1 show that the average NPP was 81% (ranging from 65% to 95%). Most patients with normal WCST scores were correctly classified as having preserved executive abilities. However, because the PPP was surprisingly poor (reaching only 24%), an overall classification rate was not different from level of chance (i.e., 50%). Correspondence Between the WCST Performance and Selected Clinical Data Next, we analyzed selected predictors for performance on the WCST (i.e., severity of executive dysfunctions and duration of illness [days postonset]). For this purpose, WCST scores were treated as separate dependent (quantitative) variables. Average results and standard deviations calculated for each WCST parameter (score) are shown in Table 2. Average group differences were assessed with multivariate analysis of variance (MAN- OVA), followed by one-way analysis of variance (ANOVA) considering statistical significance when p <.05. MANOVA revealed a tendency toward significance (F ¼ 2.06; p ¼.06; i.e., differences in the WCST performance between patients who differed in terms of severity of dysexecutive impairment). Separate univariate analysis revealed that the effect of severity given as

272 JODZIO & BIECHOWSKA an independent variable (between three subgroups) was significant for the following WCST parameters: total errors (F (2,40) ¼ 3.79; p <.03), the total categories achieved (F (2,40) ¼ 6.34; p <.004), percentage of conceptual-level responses (F (2,40) ¼ 4.06; p <.02), and number of trials to complete the first category (F (2,40) ¼ 5.80; p <.006). Analysis of the number of nonperseverative errors yielded a tendency toward significance (F (2,40) ¼ 2.84; p ¼.07). Moreover, a similar pattern of overall performance (determined on the basis of the between-group comparisons) emerged in the analysis of these scores taken together. That is, only patients with moderate-to-severe impairment of EF performed more poorly on the WCST than patients with only mild impairment and those without any executive impairment, who did not differ from each other. In consequence, none of the WCST scores yielded reliable estimates of sensitivity to detect more subtle executive changes. Importantly, there were no significant group differences for age (F (2,40) ¼ 2.61; p ¼.09) and years of education (F (2,40) ¼ 1.86; p ¼.17). To compare three age- and education-matched groups, one patient had to be excluded from the analysis. With reference to the hypothetical association between performance on the WCST and duration of illness (days postonset), correlation analysis (Pearson s r) was used separately for each test score. No significant correlations between WCST scores and days postonset were found. Factor Structure of the WCST Eight WCST scores were submitted to a principal components analysis with varimax rotation: total errors, perseverative errors, nonperseverative errors, the total categories achieved, failures to maintain the set of responses, perseverative responses, percentage of conceptual-level responses, and number of trials to complete the first category. The last parameter (i.e., learning to learn), was excluded from the analysis, as this parameter could not be computed for some patients. As shown in Table 3, the analysis revealed a three-factor solution accounting for 90.3% of the variance. The amount of variance accounted for by each factor was 59.5%, 16.5%, and 14.2%, respectively. Factor loadings of 0.50 or greater were considered significant. A similar cutoff point had been adopted in previous studies on the WCST, for example, concerning persons with Parkinson s disease (Paolo, Tröster, Axelrod, & Koller, 1995). The scores that most strongly loaded on Factors 1 to 3 were, in order: percentage of conceptual-level responses, number of trials to complete the first category, and failures to maintain the set of responses. The identical factor structure was revealed using an oblique Promax rotation as well. This additional analysis TABLE 3 Results of the Factor Analysis (Varimax Rotation) for Stroke Patients (the Full Sample) Using the WCST Scores was introduced because factors of the WCST, as has been noted in previous studies (e.g., Benge et al., 2007), might significantly correlate with one another. Neuroanatomical Correlates of the WCST Performance after Stroke Factor WCST Score (variable) 1 2 3 Percent Variance 59.5 16.5 14.2 Eigenvalue 3.95 2.13 1.14 Percentage of Conceptual-Level 0.92 a 0.37 0.01 Responses Total Errors 0.91 a 0.35 0.06 Perseverative Errors 0.88 a 0.32 0.16 Perseverative Responses 0.87 a 0.33 0.18 Nonperseverative Errors 0.79 a 0.24 0.33 Trials to Complete the First Category 0.22 0.91 a 0.12 Categories Achieved 0.26 0.87 a 0.23 Failure to Maintain Set 0.03 0.07 0.95 a Note. WCST ¼ Wisconsin Card Sorting Test. Values marked a superscript indicate a statistically significant (.5) loading for this variable. A two-way (2 2) MANOVA, with the anterior versus posterior site and left versus right side of the lesion as between-subjects factors, was computed on dependent variables (i.e., the WCST scores). For this purpose, only 28 patients (classified into four subgroups) were selected for analysis of the effect of lesion location on the WCST performance (see Table 4), because stroke in other patients was caused by extensive damage involving both anterior and posterior parts of the left or right hemisphere (see Participants section). In the frontal group, 7 patients had right-sided and 6 patients had left-sided lesions, and there was no evidence of additional damage to posterior cortical areas. The results revealed that the common hypothesis that patients with frontal lesions tend to perform worse on the WCST than those with nonfrontal lesions has, in fact, limited backing. MANOVA showed that the patients with damage restricted to the frontal lobes did not differ significantly from those with damage to the posterior parts of the brain on any WCST parameter (.17 < ps <.98). Similarly, the side of the brain damage alone did not affect performance on the test of stroke patients (.27 < ps <.94). Although both main effects (i.e., site and side) did not appear, a significant interaction between the site and side was found (F (1,24) ¼ 4.12; p <.05): The number of categories achieved was significantly smaller after damage to the frontal lobe on the left than on the right side, whereas the difference

WCST IN STROKE PATIENTS 273 TABLE 4 Subject Characteristics for Anterior and Posterior Lesion Groups Subject Age (Years) Sex Education (Years) Days Postonset Lesion Side and Site Anterior lesion of the left hemisphere (frontal lobes) L. K. 68 M 9 7 LH MDL E. P. 49 F 14 14 LH MDL P. C. 68 M 7 9 LH MDL P. L. 56 M 10 19 LH MDL N. P. 79 M 16 8 LH MDL S. C. 75 M 11 6 LH MDL M ¼ 65.8 11.2 10.5 (SD) 11.4 3.3 5.0 Anterior lesion of the right hemisphere (frontal lobes) P. M. 37 M 18 8 RH MDL G. W. 65 F 7 9 RH DLT & MDL J. C. 71 M 14 6 RH DLT & MDL J. S. 56 M 8 8 RH DLT & MDL K. O. 76 F 12 8 RH DLT B. M. 46 M 14 14 RH ORB C. W. 41 M 11 14 RH ORB M ¼ 56.0 12.0 9.6 (SD) 15.2 3.8 3.1 Posterior lesion of the left hemisphere J. C. 72 M 9 8 LH PA P. L. 20 M 13 7 LH PA K. A. 56 M 10 7 LH PA & TE A. M. 65 M 9 7 LH TE & OC J. D. 63 M 14 16 LH TE & OC J. C. 66 M 12 10 LH OC M ¼ 57.0 11.2 9.2 (SD) 18.8 2.1 3.5 Posterior lesion of the right hemisphere I. C. 51 M 8 12 RH PA R. W. 58 M 7 16 RH PA K. P. 60 M 10 10 RH PA B. L. 79 M 12 19 RH PA M. B. 46 M 13 12 RH TE J. D. 54 M 13 10 RH TE A. K. 49 M 11 10 RH TE & OC K. C. 20 F 14 14 RH OC A. H. 33 F 13 7 RH OC M ¼ 50 11.2 12.2 (SD) 16.7 2.4 3.6 Note. LH ¼ left hemisphere; RH ¼ right hemisphere; DLT ¼ dorsal lateral; MDL ¼ medial; ORB ¼ orbital; PA ¼ parietal; TE ¼ temporal; OC ¼ occipital. FIGURE 1 Interaction of the lesion side (left vs. right) and site (anterior vs. posterior) in Categories Achieved on the Wisconsin Card Sorting Test. between hemispheres was no more significant after posterior lesions (Figure 1). However, there were no significant interactions between lesion locations for the other WCST scores (.17 < ps <.83). DISCUSSION The present study was designed to investigate the predictive power of the WCST (i.e., its diagnostic accuracy in classifying stroke patients as having executive disorders). In addition, we examined the relationship between the WCST performance and selected clinical parameters (i.e., severity of executive disorders and duration of illness). This study also sought to establish the factor structure of the test. Finally, we investigated the neuroanatomical correlates of the WCST performance of stroke patients. The Diagnostic Efficiency and Predictive Power of the WCST To determine whether the WCST may be useful in exploring and elucidating executive disorders in stroke patients, diagnostic efficiency statistics (i.e., PPP and NPP) have been computed. Surprisingly, our results show that the PPP rates for the WCST scores were definitely less than the NPP rates. Hence, interpretation of abnormal scores on this test as indication of presence of EF impairments after stroke must be attempted with caution. However, normal scores can be readily interpreted as suggesting the absence of the EF disorders in more than 8 out of 10 cases (the average NPP rates reaching 81%). The WCST score, which is known as number of trials to complete the first category, is particularly relevant because normal score could reliably rule out any possible diagnosis of executive dysfunction. Namely, we found the NPP for this score rising to a very impressive 95% (see Table 1). It appears that particular WCST scores present dissimilar predictive power values and consequently divergent diagnostic potential with respect to stroke patients. Our findings are to some extent inconsistent with the findings of Grodzinsky and Barkley (1999), who had examined hyperactive children and found that both the PPP and NPP rates were less than satisfactory for all three WCST scores they used (i.e., percentage of perseverative errors, failures to maintain set, and number of trials to complete the first category [ranging from 50 to 71 for the PPP, and from 49 to 52 for the NPP]). In contrast to the above, good rates of the NPP for these scores (ranging from 75% to 95%) in our study contrasted with poor PPP rates (ranging from 17% to 29%). However, groups examined for the purpose of our study and those of Grodzinsky and Barkley s in

274 JODZIO & BIECHOWSKA general did not differ significantly in number of subjects having abnormal scores on any of the scores calculated from this test. A replication is clearly desirable, although the difference between our results and those published by Grodzinsky and Barkley may relate to the variation in age and etiologies of the patients. Irrespective of the potential importance of the age and pathogenesis in examining executive impairment on the basis of the WCST performance (for more extensive discussion, see for example, Davidson et al., 2008; Huizinga & van der Molen, 2007; Kim et al., 2007; Rhodes, 2004), this test should not be employed in clinical practice as the sole method of diagnosing poststroke executive disorders, because no test score in Table 1 exceeded an overall correct classification rate (CCR) of 55%. The CCR parameter is often used as evidence for discriminant validity (Elwood, 1993), which unfortunately in light of the present study seemed insufficient. An average CCR did not exceed the base rate, or in a word, chance. Indeed, in the case of the majority of patients in routine clinical practice, some additional tests are required to complete their assessment for example, further assessment of intelligence, attention, or praxis. Thus, results of the current study contrast with the common presumption about the role of the WCST as a gold standard EF measure. Nonetheless, Royall et al. (2002) argue that because the WCST is a complex test in a psychological sense, it may be the most proven of all putative EF measures. Moreover, multiple EF could be ascribed to the various WCST subtests, but this assertion is difficult to verify empirically. Correspondence Between the WCST Performance and Selected Clinical Data To determine the relationship between the WCST performance and severity of executive disorders and duration of illness, two types of analyses were performed. First, MANOVA was applied to investigate any possible effect of severity of dysexecutive impairment on the WCST performance. The multivariate analysis revealed significant differences between patients with retained EF, those suffering from mild executive dysfunction, and patients with moderate or severe EF dysfunction (see Table 2). Subsequent between-group comparisons of the average differences across nine test scores revealed that the patients who exhibited moderate or severe executive impairment performed more poorly than other groups on four out of the nine WCST parameters (i.e., the total errors, the total categories achieved, percentage of conceptual-level responses, and number of trials to complete the first category). However, results of this study also indicate that the persons with milder executive problems and those with normal EF have comparable difficulty performing the WCST. Consequently, these findings indicate that the WCST does not reliably detect more subtle executive problems observed in stroke survivors. This is indirectly congruent with the proposition expressed above that prevalence of executive disorders after stroke can be underestimated whenever the WCST is the only diagnostic measure used to probe EF. Unlike other studies (e.g., Barceló & Knight, 2002; Paolo et al., 1995; for review, see also, Jaworowska, 2002; Lezak et al., 2004), the current study also demonstrated that perseverative errors most widely used WCST score in the diagnosis of executive disorders following frontal-lobe damage did not distinguish the groups, including patients varied with respect to degree of EF impairments. According to Lezak et al. (2004), the perseverative errors score is useful for detecting abnormalities in such cognitive domains as concept formation, profiting from correction, and conceptual flexibility. Second, with regard to the relationship between the WCST performance and duration of illness (i.e., time from the onset), r-pearson s correlations were computed separately for each test score. Although correlations did not reach statistical significance, we cannot entirely exclude the possibility that study of a more homogeneous group of patients, including either patients examined after a longer time from the onset (for example, several months rather than days) would lead to a different conclusion. Because only patients with recent stroke (i.e., no more than 30 days postonset) were included in the study, there is a need to take practical limitations of our design into consideration. Indeed, some symptoms resolve quite well on their own within a few days or weeks, but others may persist, evolving into chronic stage (for review, see Worthington, 2003). For this reason, future work will need to investigate the practical implications of the time-course analysis of various executive disorders in the stroke population. Prognosis for these disorders is still uncertain. Moreover, EF deficits appear to be among the most persistent consequences of acquired brain damage. Factor Structure of the WCST While the factor structure of the WCST has been widely studied, no consensus structure has yet emerged. Despite the controversy, the results of this study support repeated factor-analytic studies (e.g., Benge et al., 2007; Greve, Bianchini, Hartley, & Adams, 1999; Paolo et al., 1995). It is clear that the WCST does not tap a hypothetical single (executive) function or ability but generally measures more complex domains of functioning (i.e., multiple dimensions of EF). Therefore, considerable evidence for a multifactorial structure of the WCST is clear in light of this study. Specifically, the

WCST IN STROKE PATIENTS 275 results generally confirmed the previously reported three-factor structure of the test (see Table 3). The first factor, referred to as concept formation, consisted of percentage of conceptual-level responses, total errors, perseverative errors, perseverative responses, and nonperseverative errors. This component appears to represent an overall conceptualization factor and is similar to the concept formation=perseveration= problem-solving factor found in elderly persons with normal cognitive functions (Greve et al., 1999) and those with Parkinson s disease (Paolo et al., 1995). The second factor, labeled as set switching (or shifting), included trials to complete the first category and categories achieved. This component relates strongly not only to learning as such but also to the ability to display flexibility in the face of changing schedules of verbal reinforcement (i.e., feedback information from the examiner). The remaining factor, referred to as set maintenance, was composed of only one WCST score (i.e., failure to maintain set). The conceptualization factor can be ascribed to fundamental concept formation processes. The other two are closely related to particular attentional functions or abilities: to shift response set and to sustain mental set, respectively. Our results are in keeping with the idea that sustained attention may be one of the executive processes involved in the maintenance of error awareness, which obviously plays a critical role in the WCST performance. It is well documented that feedback on error may enhance patients performance also on the other tests of executive attention (McAvinue, O Keeffe, McMackin, & Robertson, 2005). In addition, the existence of two relatively distinct attentional factors on the structure of the WCST is confirmed by developmental studies, showing age-group differences in switching and maintenance on the test (e.g., Huizinga & van der Molen, 2007). However, it is unclear whether the creation of factor scores may be necessary for clinical use. In other words, to the extent that these separate factors are discriminable aspects of the test, they ought to provide additional clinically relevant information. For instance, Benge et al. (2007) demonstrated the relationship of WCST factor scores with functional outcomes in a group of individuals with severe TBI. Although this finding was partially replicated in a sample of stroke patients (Greve et al., 1999), there is still little known about the similarities and differences in WCST factor structure in certain varieties of stroke (e.g., arterial infarct, hemorrhage, vascular dementia, etc.) associated with the dysexecutive problems (Godefroy & Stuss, 2007). It is likely, as Paolo et al. (1995) have convincingly argued, that the factor structure of the WCST and what the scores represent may vary according to sample used and type of neurological impairment. Neuroanatomical Correlates of the WCST Performance after Stroke In general, the results of our lesion study do not substantiate application of the WCST to frontal-lobe lesions. There was no difference regarding most WCST measures between subjects with frontal and several nonfrontal lesion profiles. Some subjects, who had been shown by CT or MRI scans to have focal lesions in the occipital, temporal, or parietal regions, performed the test just as poorly as frontal patients. It is already known that impairments on the WCST are not caused only by frontal cortex lesions (for examples, see Alvarez & Emory, 2006; Daum & Mayes, 2000). Instead, there is now evidence that damage to nonfrontal brain regions may cause deficits on some executive tests that may prove hard to distinguish qualitatively as well as quantitatively from the deficits caused by frontal cortex lesions (Lezak et al., 2004). Hence, the relationship between lesion location and the pattern of the WCST performance is relatively weak. This notion is supported by ANOVA, with the anterior versus posterior site and left versus right side of the lesion as between-subjects factors. However, there was one very interesting and important exception to the finding that frontal and posterior lesions had similar effects on the WCST performance. That is, patients with damage restricted to the left frontal lobe were severely impaired compared with the patients with lesions elsewhere at the subscale considered very crucial to set shifting and cognitive flexibility (i.e., categories achieved score). In other words, the interaction of the anterior versus posterior site and side of the lesion was significant, indicating existence of a relationship between inflexibility and damage to left frontal neocortical networks. From this finding, we have some indication that subjects performance on the WCST may depend on the side of the lesion (i.e., a laterality effect), which can be an equally important factor or in some circumstances an even more significant factor than lesion site (i.e., anterior vs. posterior) within hemisphere. It is also worth noting that this difference with respect to lesion lateralization did not arise from differences in the nature of the lesions or other patient variables, because the groups statistically compared were homogeneous in terms of several clinical and demographic parameters (for details, see the Participants section and Table 4). However, according to Lezak et al. (2004), WCST findings with respect to lesion lateralization lack consistency. Lezak et al. cite, for instance, an older piece by Taylor who observed that more neurosurgical patients with left-sided lesions displayed serious impairments on this test after an operation than did those with right-sided lesions. In contrast, some other researchers cited by Lezak et al. reported just the opposite (i.e., a

276 JODZIO & BIECHOWSKA tendency for patients with right frontal damage to display executive deficits on the WCST). Two further points should be stressed. First, we would like to underscore the need for caution in generalizing these results given the relatively small sample of patients in this study. And second, long-term neuropsychological sequelae of the stroke were not discussed because our study focused on its early period. It is possible that the clinical picture initially observed and its neuroanatomical correlates were at least partially due to rapid hemodynamic disturbances and=or diaschisis, which are usually of lesser significance but nevertheless can negatively influence undamaged structures in the brain. Interestingly, a similar effect was observed in patients after TBI (Wu et al., 2004). Although the present study is insufficient to clarify this issue, it is commonly observed that the area of cerebral hypoperfusion caused by cerebrovascular accident is larger than the structural lesion, in particular in patients with recent stroke (Jodzio, Drumm, Nyka, Lass, & Gąsecki, 2005). Furthermore, the relationship between lesion location and executive impairments on the WCST may be not at all localizable in the traditional sense and stable in terms of its strength. For example, Funahashi (2001) emphasizes the need to consider dynamic and flexible changes in neural interaction among frontal neurons, especially within the DLPFC, as well as between frontal lobes and other cerebral areas depending upon the characteristics or the temporal context of the task (for review, see Miller & Cummings, 2007). In conclusion, this study suggests that more sensitive tests than the WCST are needed to establish a relationship between a specific type of executive dysfunction and a particular lesion location in people with recent unilateral stroke. ACKNOWLEDGEMENTS This work was supported by the University of Gdańsk grant number BW=7400-5-0060-8 to the authors. The authors are grateful to Dr. Dariusz Gąsecki from the Department of Neurology at the Medical University of Gdańsk, Poland, for his help with the anatomical analysis and assistance with the subject recruitment for this study. REFERENCES Alvarez, J. A., & Emory, E. (2006). Executive function and the frontal lobes: A meta-analytic review. Neuropsychology Review, 16, 17 42. Barceló, F., & Knight, R. T. (2002). Both random and perseverative errors underlie WCST deficits in prefrontal patients. Neuropsychologia, 40, 349 356. Baudic, S., Dalla Barba, G., Thibaudet, M. C., Smagghe, A., Remy, P., & Traykov, L. (2006). Executive function deficits in early Alzheimer s disease and their relations with episodic memory. Archives of Clinical Neuropsychology, 21, 15 21. Benge, J. F., Caroselli, J. S., & Temple, R. O. (2007). Wisconsin Card Sorting Test: Factor structure and relationship to productivity and supervision needs following severe traumatic brain injury. Brain Injury, 21, 395 400. Burgess, P. W., Alderman, N., Forbes, C., Costello, A., Coates, L. M.-A., Dawson, D. R., et al. (2006). The case for the development and use of ecologically valid measures of executive function in experimental and clinical neuropsychology. Journal of the International Neuropsychological Society, 12, 194 209. Cummings, J. L., & Mega, M. S. (2003). Neuropsychiatry and behavioral neuroscience. New York: Oxford University Press. Dagher, A., Owen, A. M., Boecker, H., & Brooks, D. J. (2001). The role of the striatum and hippocampus in planning: A PET activation study in Parkinson s disease. Brain, 124, 1020 1032. Daum, I., & Mayes, A. R. (2000). Memory and executive function impairments after frontal or posterior cortex lesions. Behavioral Neurology, 12, 161 173. Davidson, P. S. R., Gao, F. Q., Mason, W. P., Winocur, G., & Anderson, N. D. (2008). Verbal Fluency, Trail Making, and Wisconsin Card Sorting Test performance following right frontal lobe tumor resection. Journal of Clinical and Experimental Neuropsychology, 30, 18 32. Duff, K., Clark, J. D. H., O Bryant, S. E., Mold, J. W., Schiffer, R. B., & Sutker, P. B. (2008). Utility of the RBANS in detecting cognitive impairment associated with Alzheimer s disease: Sensitivity, specificity, and positive and negative predictive powers. Archives of Clinical Neuropsychology, 23, 603 612. Duke, L. M., & Kaszniak, A. W. (2000). Executive control functions in degenerative dementias: A comparative review. Neuropsychology Review, 10, 75 99. Elwood, R. W. (1993). Clinical discriminations and neuropsychological tests: An appeal to Bayes Theorem. The Clinical Neuropsychologist, 7, 224 233. Eslinger, P. J., & Grattan, L. M. (1993). Frontal-lobe and frontal-striatal substrates for different forms of human cognitive flexibility. Neuropsychologia, 31, 17 28. Funahashi, S. (2001). Neuronal mechanisms of executive control by the prefrontal cortex. Neuroscience Research, 39, 147 165. Godefroy, O., & Stuss, D. (2007). Dysexecutive syndromes. In O. Godefroy & J. Bogousslavsky (Eds.), The behavioral and cognitive neurology of stroke (pp. 369 406). New York: Cambridge University Press. Goldberg, E. (2001). The executive brain: Frontal lobes and the civilized mind. New York: Oxford University Press. Greve, K. W., Bianchini, K. J., Hartley, S. M., & Adams, D. (1999). The Wisconsin Card Sorting Test in stroke rehabilitation: Factor structure and relationship to outcome. Archives of Clinical Neuropsychology, 14, 497 509. Grodzinsky, G. M., & Barkley, R. A. (1999). Predictive power of frontal-lobe tests in the diagnosis of attention-deficit hyperactivity disorder. The Clinical Neuropsychologist, 13, 12 21. Harciarek, M., & Jodzio, K. (2005). Neuropsychological differences between frontotemporal dementia and Alzheimer s disease: A review. Neuropsychology Review, 15, 131 145. Hashimoto, K., Uruma, G., & Abo, M. (2008). Activation of the prefrontal cortex during the Wisconsin Card Sorting Test (Keio version) as measured by two-channel near-infrared spectroscopy in patients with traumatic brain injury. European Neurology, 59, 24 30. 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, Inc.