1
Data Supplement:Hemispatial neglect: Subtypes, Neuroanatomy, and Disability by L. Buxbaum et al.
Electronic mail:
METHODS:
Subjects: Inclusion and exclusion criteria, Testing Schedules
Test description and scoring information: Bells Test, Letter Cancellation, Picture Scanning, Letter Reading, Line Bisection, Sensory Examination, Sustained Attention to Response Task (SART), Dual Task, Lateralized Target Test, Lateralized Response Test, Anosognosia Questionnaire, Functional Independence Measure (FIM)
Determination of Neglect
DATA ANALYSIS: Comment
RESULTS:
Motor function
Subtypes
Attention and sensorimotor response speed
Clinical implications of the neglect syndrome: FIM, Caregiver burden
Reliability study (neuroanatomic data)
Categorical Modeling details
Lesion patterns for neglect subtypes
Role of temporal lobe involvement
Regression analyses
Subjects
Inclusion and exclusion criteria:
Subjects with a history of previous head injury, left hemispheric stroke, or other neurologic disorder were excluded, as were those for whom review of the medical chart indicated a DSM-IV Axis I disorder (e.g., major depression, psychosis, dementia; American Psychiatric Association, 1994) at the time of the study. Patients with previous RCVA (n = 5) were not excluded. Inclusion was also based on three behavioral criteria: (a) language comprehension adequate to understand task instructions; (b) visual and auditory acuity and motor function sufficient to perform the testing protocol; (c) attention and behavioral control adequate for a 90-minute testing session.
Testing schedules:
Acute patients were tested within seven days of admission to the rehabilitation hospital. Acute patients whose initial performance was consistent with operationally-defined neglect (see (E)Methods-Determination of neglect) were tested weekly until discharge or until they reached 6 weeks post-stroke (Mn = 2 sessions, s.d. = 0.89, range = 1-4 sessions). Chronic patients were assessed once.
Test description and scoring information
Bells Test:
Participant is required to cancel 35 bells presented in a dense array with 282 other stimuli on 8.5 x 11” paper. A left-right difference score was calculated.
Letter cancellation:
Participant is required to cancel the letters E and R (20 each) presented among 150 other letters on 8.5 x 11” paper. The difference between number of targets canceled on the left and right was calculated.
Picture scanning:
Participant is required to name objects in an 11” x 14” photograph of items used in grooming and self-care. There are two versions of the subtest; one was assigned to each patient for the duration of testing. We calculated the percent of targets named on the left and right of the total possible (three on the left in both versions; four on the right in version A, and five on the right in version B), and then calculated the difference between the left and right percentage scores.
Menu reading:
Participant required to read 24 words presented in four columns (two versions used). A left-right difference score was calculated.
Line bisection:
Participant is required to denote the center of each of three lines, each 20.5 cm long. The mean distance of responses (in mm.) from the true midpoint was calculated.
Sensory examination scoring:
For both visual and tactile examinations, participants who reported less than four unilateral left stimuli were characterized as exhibiting a visual field defect or tactile sensory loss. Patients who reported three or four unilateral left stimuli, and who reported at least one fewer left stimuli on bilateral trials, were characterized as exhibiting extinction. Note that there is no consensus for a particular cut-off score defining extinction, as it is likely to be a continuous rather than discrete phenomenon. A recent review (2)presents data from patients whose extinctionranges from15% to 80% of left sided targets on bilateral trials depending on stimulus characteristics.
Sustained Attention to Response Test (SART):
Participants viewed 225 single digits centered on a computer screen over a 4.3-minute period. Each digit was presented for 250-msec., followed by a 900-msec mask. Participants responded with a mouse key press to all digits except the number ‘3’, which appeared 25 times in a quasi-randomized fashion. Trials with response latencies > 10,000 ms were considered “missed” trials (14% of trials). Trials with response latencies <150 ms were trimmed (3.7% of trials). Mean response latency to target numbers was recorded (SART Response Time). The number of correct responses to ‘non-3’s’ (“SART Go”, max. = 200) and correctly withheld responses to ‘3’s were also tallied (“SART No-go”, maximum=25).
Dual Tasktest:
In the baseline condition, participants responded with a key press to a black dot 1cm in diameter appearing in randomized order in far left, left, center, near right, and far right locations on a computer screen at intervals of 500-2000 msecs over 64 trials. Mean response latencies to right-sided and left-sided targets in the Baseline condition were used as measures of attention and sensorimotor response time to targets in ipsilesional and contralesional hemispace. This study is not the first to use DT Base as a measure of “simple” attention (as distinguished from “executive” attention; see (5). Additional unpublished analyses with data from that study indicate that DTBase predicts performance on a complex naturalistic action task (NAT) (r = -.67, p < .002) even in left hemisphere stroke patients who do not exhibit neglect (as assessed by above-cutoff performance of the Star Cancellation test from the BIT); thus, a non-lateralized component of attention appears to contribute to DT Base scores. Nevertheless, it is clear that lateralized attention deficits (i.e., neglect) may also be expected to contribute to performance of this task.
In the dual task condition (DT Dual), participants performed the baseline task in conjunction with an oral digit repetition task conducted at participants’ span. Decrements in response to lateralized targets under Dual Task load was measured by subtracting mean response latencies to contralesional and ipsilesional targets in DT Base from DT Dual (Left Dual - Left Base and Right Dual – Right Base). Trials with response latencies <150 ms and > 10,000 ms. were trimmed, resulting in a loss of 1% of the data. Trials with response latencies <150 ms and > 10,000 ms. were trimmed, resulting in a loss of 1% of the data.
Lateralized Target test:
In this test of “perceptual neglect”, participants viewed three horizontally-arrayed 5 mm. dots separated by 4 cm. on a computer monitor. They were asked to fixate on the center dot, and fixation was monitored by the examiner. After 2000 ms., one of the three dots was replaced by a downward-pointing arrow. Participants responded by pressing the spacebar on the computer keyboard when a left- or right- sided arrow was detected.
Lateralized Response test:
In this test of “motor neglect” was designed to be sensitive both to “intentional neglect” (i.e., impairment in capacity to direct action plans into contralateral hemispace) and “directional hypokinesia” (i.e., slowed movement into contralateral hemispace) but because it measures response time to move into contralalesional space, it does not distinguish them (see (1, 6)). In our task, participants viewed three horizontally-arrayed dots 5 mm. dots separated by 4 cm. They fixated on the center dot, and fixation was monitored. After 2000 ms., the center dot was replaced by a left-, right-, or downward-pointing arrow. Participants began each trial with the right forefinger touching a dot on the table top positioned at body and keyboard midline, and responded by pressing a key on the partially-shielded keyboard with left (keys Q,W,E,A,S, and D), central (keys I, O, P, K,L, and ;), and right (keys 7, 8, 9, 4, 5, and 6) response areas. In both tasks, nine practice trials were followed by 20 trials each with left, right and center targets or responses in randomized order. Mean difference in response time for left versus right trials was calculated for each participant.
Anosognosia Questionnaire:
There were five questions. Two questions addressed awareness of sensorimotor impairment (Is there anything wrong with either of your arms?; …legs?), and three addressed general awareness of deficit (Why are you in the hospital? Do you have problems seeing? Did you have any difficulties with the tests today?). For each question, there is a more specific follow-up problem if the patient is assessed to be in denial of deficit(s). Three points are given if the patient denies any problem despite a follow-up question, two points if a problem is admitted on the follow-up query, and one point if the primary question is answered correctly (max. = 15). Deviation scores are then calculated as follows: for the questions concerning use of extremities, the rater assigns an impairment score (1 = no active limb movement, 2 = more difficulty moving left than right, 3 = good movement bilaterally. The rater’s score is then subtracted from the patient’s awareness score. For the three test items assessing general awareness of deficit, the deviation score is calculated by subtracting 1 from the patient’s awareness score. For each item, the deviation score may range from 0 (no anosognosia) to 2 (moderate-severe anosognosia). The total anosognosia score is the sum of the deviation scores for each item (range 0 – 10).
Functional Independence Measure (FIM):
This widely-used measure of clinical severity is organized into 4 subscales on a physical dimension (self-care, mobility, locomotion, and sphincter control) and 2 on a cognitive dimension (communication and social cognition). Each item is rated from 1 (total assistance required) to 7 (complete independence), allowing for a maximum total score of 126. Scores were transformed into normalized ratio scales (mean 50, range 0-100) for use in parametric analyses based on the method of Heinemann et al (3). The FIM’s reliability was reviewed in a meta-analysis of 11 publications (4). Test-retest reliability ranged from .84 - .93 and inter-rater reliability ranged from .83 - .99 across the studies.
Determination of neglect
The presence of neglect was defined by performance below cut-off scores on at least one of the 5 “clinical” neglect tests, i.e., the Bells Test, Letter Cancellation, Line Bisection, Menu Reading, and Picture Scanning tests. For the Bells Test, Letter Cancellation, and Menu Reading, patients were considered to exhibit neglect if their left-right difference scores were greater than 20% of the total number of items on each side of the array. As there were two versions of the Picture Scanning task, patients were considered to exhibit neglect on this task if the difference between the percent of items named on each side of the array was greater than 20 percentage points. Finally, the cut-off score for line bisection was determined by scoring guidelines provided for the Behavioral Inattention Test (7).
For the tasks designed to identify patients with subtypes of neglect (Motor, Perceptual, Personal, and Peripersonal) one of our goals was to identify patients with dissociations in performance. This required standardizing the scoring of the subtypes tasks so that across-task comparisons could be made. Consequently, we defined deficient performance as that below the 20th percentile for the study population of RCVARCVA patients. Patients with reaction times below the 20th percentile on the Lateralized Target or Lateralized Response tasks were characterized as exhibiting perceptual or motor neglect. A small number of patients (n = 4) missed sufficient (> 50%) left sided targets on the Lateralized Target task that their left-sided reaction time data were unreliable; these patients were characterized as exhibiting perceptual neglect on the basis of the high miss rate. Patients with performance below the 20th percentile on one of the five clinical tests or below the 20th percentile on the Fluff test were characterized as exhibiting peripersonal or personal neglect. Of course, performance could be below the 20th percentile on any combination of the subtypes tasks.
DATA ANALYSIS
All data were examined to determine whether they met statistical assumptions for parametric testing, and unless otherwise noted, it can be assumed that this was the case. If not, we pursued one of two strategies. In some cases, data were transformed as indicated. Nonparametric testing was performed when transformation was not sufficient to normalize the distribution in question.
In some instances, multiple comparisons (e.g., correlational analyses) were made with the same data set. We recognize the wisdom of performing Bonferroni corrections under these circumstances. However, given that this is an exploratory study examining a large number of different patterns, a concern was that we would have had insufficient power to detect differences of potential interest. Consequently, the observed patterns of interest should be confirmed in future investigations.
RESULTS
Motor Function:
Mann Whitney Tests were performed to assess Grip strength in neglect and non-neglect patients. A+ were significantly weaker than A- for both left hand (Z= -2.1, p = .03) and right hand (z = -3.7, p < .002), but the difference between left and right hand strength was comparable in the two groups (z= -0.6, p = .5). The same pattern was observed in the Chronic patients: C+ were weaker than C- for both left (Z = -1.9, p < .05) and right hand (z = -2.5, p < .01), but the left-right difference score was comparable in the two groups (z = -1.0, p = .29).
Subtypes:
A small number of patients (n = 4) missed sufficient (> 50%) left sided targets on the Lateralized Target task that their left-sided reaction time data were unreliable; these patients were characterized as exhibiting perceptual neglect on the basis of the high miss rate. Patients with performance below the 20th percentile on one of the five clinical tests or below the 20th percentile on the Fluff test were characterized as exhibiting peripersonal or personal neglect. Of course, performance could be below the 20th percentile on any combination of the subtypes tasks.
One possibility is that some of the patients exhibiting an apparently pure neglect subtype had scores clustering near one another, straddling the 20th percentile cut-off on the two tasks of interest (e.g., 19th percentile on the Motor neglect task, and 21st percentile on the Perceptual neglect task). Thus, the subtype designation might be an artifact of our transformation of continuous into categorical data. We assessed the number of patients characterized as exhibiting a neglect subtype (< 20th percentile on the relevant task) whose score was above average(> 50th percentile) on the contrasting relevant task. Of the Motor neglect patients presented in Table 4 eight Acute and six Chronic patients performed above the 50th percentile on Perceptual neglect testing. Conversely, 11 Acute and four Chronic patients with Perceptual neglect performed above the 50th percentile on Motor neglect testing. Sixteen Acute and 19 Chronic patients with Peripersonal neglect performed above the 50th percentile on Personal neglect testing. One patient showed the reverse pattern.
Attention and Sensorimotor response speed
In both Acute and Chronic patients, Average Neglect Percentile was correlated with performance on a measure requiring a motor response to simple visual targets in both ipsilesional and contralesional hemispace (DT Base Right: Acute patients: r = -.41, p < .0003, Chronic patients: r = -.44, p < .0001; DT Base Left: Acute patients: r = -.51, p < .0001, Chronic patients: r = -.46, p < .0001 ). Average neglect percentile also correlated with measures of sustained attention (SART Go: Acute patients: r = .49, p < .001; Chronic patients: r = .43, p < .0001; SART Response Time: Acute patients: r = -.47, p < .0001, Chronic patients: r = - .47, p < .0001). On the other hand, there were no correlations of Average Neglect Percentile with measures frequently assumed to assess “executive function”. Thus, Average Neglect Percentile did not correlate with the decrement in response to lateralized or non-lateralized targets in dual as compared to single task conditions (Left Dual - Left Base; Dual-Base Dec). Nor were there correlations of Average Neglect Percentile with a measure of response inhibition (SART No-Go) (-1.5< r < 1.5, p > .2 for all of these comparisons).
Clinical Implications of the neglect syndrome
FIM: Admission FIM data was obtained for 80 Acute patients (see (E)T-7 ). There was a correlation between Average Neglect Percentile scores and FIM Total scores at admission (r = .32, p = .004). We also obtained discharge FIM data for 64 Acute patients. To assess whether A+ patients exhibited a clinical recovery pattern that differed from A-, we performed two repeated measures ANOVAs with converted FIM Physical and Cognitive (C-FIM Phys and C-FIM Cog) as the dependent variables, Session (Admission, Discharge) as the within subjects factor, and Group (A+, A-) as the between subject factor. In both analyses there were main effects of Group (A+ more impaired than A-, F(1,62) > 5.2, p < .02) and Session (admission more impaired than discharge, F(1,62) > 42.2, p < .0001), but no interaction (F < 0.4). These results suggest that patients with neglect have poorer clinical outcome than those without neglect, as assessed by a widely used measure of functional disability. On the other hand, FIM gains made during rehabilitation are not significantly different in the two groups.
It is possible that patients with moderate to severe RCVA are likely to exhibit neglect, and that it is the severity of stroke, rather than neglect per se, that predicts lowered FIM scores. While this possibility can not be discounted, an additional analysis to be described below suggests that neglect is a more potent predictor of disability than is lesion size.
Family Burden. Scores on the Family Burden Questionnaire were more elevated for C+ (Mn. = 5.2, s.d. = 2.1) than C- patients (Mn. = 2.9, s.d. = 2.1, t=-3.9, p<.001). To assess whether neglect contributed to the prediction of family burden beyond overall clinical severity, we performed two regression analyses, both with Average neglect percentile as independent variables and Family Burden Questionnaire as dependent variables. The first regression used Admission FIM Total score as an additional independent variable and the second regression used Discharge FIM Total as an additional independent variable.
Adjusted R squared for the first regression was .28 (F = 8.5, p = .0009). Average neglect percentile was a strong independent predictor of burden (Std. Coefficient = -3.8, p = .0004). On the other hand, Admission FIM did not contribute to the prediction of burden (Std. Coefficient = -.56, p = .57).
Adjusted R squared for the second regression was .43 (F = 11.2, p = .0003). Discharge FIM Total contributed to the prediction of family burden (Std. Coefficient = -2.2, p = .04). Importantly, Average neglect percentile made an independent contribution to the prediction of family burden (Std. Coefficient = -2.1, p = .039).
We also assessed whether Chronic patients characterized as exhibiting a neglect subtype (Personal, Peripersonal, Motor, or Perceptual neglect) were more of a burden to families than those without these syndromes. When we entered scores from subtypes tasks (Average Neglect Percentile, Fluff test, Lateralized Target test, Lateralized Response test were used in this analysis) into a multiple regression model with Family Burden Questionnaire as the dependent variable, the multiple adjusted R-squared was .19 (F = 3.9, p = .008), and the only significant independent predictor of performance was the Average Neglect Percentile, which in this analysis was the measure of peripersonal neglect (std. Coefficient = -.36, p < .02). Thus, it does not appear that the presence or absence of specific neglect subtypes has a substantial impact on family burden.