This article was downloaded by: [New York University] On: 24 June 2015, At: 06:34 Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK
Journal of Clinical and Experimental Neuropsychology Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ncen20
Information processing speed and attention in multiple sclerosis: Reconsidering the Attention Network Test (ANT) a
a
b
Alexandra K. Roth , Douglas R. Denney & Sharon G. Lynch a
Department of Psychology, University of Kansas, Lawrence, KS, USA
b
Department of Neurology, University of Kansas Medical Center, Kansas City, KS, USA Published online: 26 May 2015.
Click for updates To cite this article: Alexandra K. Roth, Douglas R. Denney & Sharon G. Lynch (2015): Information processing speed and attention in multiple sclerosis: Reconsidering the Attention Network Test (ANT), Journal of Clinical and Experimental Neuropsychology, DOI: 10.1080/13803395.2015.1037252 To link to this article: http://dx.doi.org/10.1080/13803395.2015.1037252
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions
Journal of Clinical and Experimental Neuropsychology, 2015 http://dx.doi.org/10.1080/13803395.2015.1037252
Information processing speed and attention in multiple sclerosis: Reconsidering the Attention Network Test (ANT) Alexandra K. Roth1, Douglas R. Denney1, and Sharon G. Lynch2 1 2
Department of Psychology, University of Kansas, Lawrence, KS, USA Department of Neurology, University of Kansas Medical Center, Kansas City, KS, USA
Downloaded by [New York University] at 06:34 24 June 2015
(Received 14 October 2014; accepted 30 March 2015) Objective: The Attention Network Test (ANT) assesses attention in terms of discrepancies between response times to items that differ in the burden they place on some facet of attention. However, simple arithmetic difference scores commonly used to capture these discrepancies fail to provide adequate control for information processing speed, leading to distorted findings when patient and control groups differ markedly in the speed with which they process and respond to stimulus information. This study examined attention networks in patients with multiple sclerosis (MS) using simple difference scores, proportional scores, and residualized scores that control for processing speed through statistical regression. Method: Patients with relapsing-remitting (N = 20) or secondary progressive (N = 20) MS and healthy controls (N = 40) of similar age, education, and gender completed the ANT. Results: Substantial differences between patients and controls were found on all measures of processing speed. Patients exhibited difficulties in the executive control network, but only when difference scores were considered. When deficits in information processing speed were adequately controlled using proportional or residualized score, deficits in the alerting network emerged. The effect sizes for these deficits were notably smaller than those for overall information processing speed and were also limited to patients with secondary progressive MS. Conclusions: Deficits in processing speed are more prominent in MS than those involving attention, and when the former are properly accounted for, differences in the latter are confined to the alerting network. Keywords: Multiple sclerosis; Attention Network Test; Reaction time; Information processing speed; Alerting network; Executive control network.
Many investigators have pointed to a generalized slowing in information processing speed as the primary cognitive deficit associated with multiple sclerosis (MS; Chiavaralloti & DeLuca, 2008; DeLuca, Chelune, Tulsky, Lengenfelder, & Chiaravalloti, 2004; Denney, Gallagher, & Lynch, 2011; de Sonneville et al., 2002; Kail, 1998; Kujala, Portin, Revonsuo, & Ruutiainen, 1994; Lengenfelder et al., 2006; Macniven et al., 2008; Reicker, Tombaugh, Walker, & Freedman, 2007). Patients with MS complete substantially fewer items on rapid serial processing (RSP) tests such as
the Paced Auditory Serial Addition Test (PASAT: Kujala, Portin, Revonsuo, & Ruutiainen, 1995; Litvan, Grafman, Vendrell, & Martinez, 1988) and Symbol Digit Modalities Test (SDMT: Parmenter, Weinstock-Guttman, Garg, Munschauer, & Benedict, 2007) and also exhibit substantially longer response times on a variety of simple and choice reaction time (RT) measures (Bodling, Denney, & Lynch, 2012; de Sonneville et al., 2002; Hughes, Denney, & Lynch, 2011; Kujala et al., 1994; Reicker et al., 2007; Tombaugh, Berrigan, Walker, & Freedman, 2010).
This research was presented at the 29th Annual Congress of the European Committee for Treatment and Research in Multiple Sclerosis (ECTRIMS), in Copenhagen, Denmark, 4 October 2013. This project was unfunded, and the authors have no financial or other conflicts of interests to report. Address correspondence to: Douglas R. Denney, Department of Psychology, 1415 Jayhawk Blvd., Lawrence, KS 66045-7556, USA (E‑mail: [email protected]).
© 2015 Taylor & Francis
Downloaded by [New York University] at 06:34 24 June 2015
2
ROTH, DENNEY, LYNCH
Problems with attention are often cited as another feature of cognitive impairment in patients with MS (Kujala et al., 1995; McCarthy, Beaumont, Thompson, & Peacock, 2005; Paul, Beatty, Schneider, Blanco, & Hames, 1998). As others (Chiaravalloti & DeLuca, 2008; de Sonneville et al., 2002; Kujala et al., 1994, 1995; Macniven et al., 2008) have noted, however, measures of attention are often overlapped with information processing speed. Information processing speed involves how quickly an individual is able to perform a variety of cognitive tasks, often varying in the complexity of the cognitive operations required. On the other hand, attention is often evaluated by examining the discrepancy in performance on speeded tasks that differ in the burden they place on some facet of attention—for example, the difference between the time required to name letters that appear with or without a warning signal. A central concern of the present study is the way in which such discrepancies in performance are measured. The Attention Network Test (ANT; Fan, McCandliss, Sommer, Raz, & Posner, 2002) is a widely used instrument designed to assess three neural networks comprising Posner’s model of attention (Posner & Petersen, 1990): the alerting network that maintains a state of alertness; the orienting network that extracts spatial information from sensory input; and the executive control network that resolves conflicting information. The ANT measures choice RTs to a directional target that is accompanied by a variety of warning cues and flanker conditions. Network scores are based on differences between the RTs for different conditions. Thus, the alerting effect is measured by subtracting the average RT when a cue signals the imminent appearance of the target from the average RT when no such warning cue is provided. The orienting effect is measured by subtracting the average RT when the warning cue also provides information as to where the target will appear from the average RT when the cue is lacking in this spatial information. The executive control effect is measured by subtracting the average RT when the target is accompanied by congruent flankers from the average RT when the target is accompanied by incongruent flankers. One reason for focusing on the discrepancies in RTs is to distinguish the attention network effect from general information processing speed that affects RTs on both measures. However, assessing network effects on the basis of simple difference scores in studies comparing groups that differ substantially in their speed of information processing can lead to distorted findings because of the
disparity between the groups’ performance on whichever measure is serving as the baseline for the difference score. In other words, while providing an adequate control for within-group differences in information processing speed, we contend that simple difference scores are inadequate for controlling for between-group differences in this attribute. Several studies have compared MS patients and controls on the attention networks comprising Posner’s model (Crivelli et al., 2012; Ishigami, Fisk, Wojtowicz, & Klein, 2013; Omisade et al., 2012; Urbanek et al., 2010; Wojtowicz, Omisade, & Fisk, 2013). No significant differences in specific network effects have been found when small samples of patients and controls are compared (e.g., N = 11: Ishigami et al., 2013; N = 12: Omisade et al., 2012), probably due to these studies’ insufficient statistical power. However, when larger samples were examined, both Urbanek et al. (2010) and Crivelli et al. (2012) reported deficits in patients’ alerting network, whereas Wojtowicz et al. (2013) found impaired performance on the executive control network. The conflicting findings in these latter studies may stem from the way in which discrepancies in the RTs involved in each network score were treated. The network effects reported by Wojtowicz et al. (2013) were based on simple difference scores. Both Urbanek et al. (2010) and Crivelli et al. (2012) evaluated network effects with procedures designed to account for the differences in overall processing speed between the patients and controls. Urbanek et al. used proportional scores, dividing the difference scores by the participant’s overall average RT on all of the items of the ANT. Crivelli et al. regressed difference scores on an alternative measure of processing speed, the average RT on trials with no warning cue and neutral flankers, and then used the unstandardized residual scores to examine group differences in network effects. As in many other RT studies, differences between MS patients and controls in overall information processing speed are evident in practically all of the ANT studies, and, often these differences are more robust than the network effects themselves (Ishigami et al., 2013; Omisade et al., 2012; Wojtowicz et al., 2013). The principal aim of the present study was to show that such differences in processing speed distort the findings with respect to network effects on the ANT by actually diminishing the alerting effect and augmenting the executive control effect. We intend to show that, when appropriate controls are imposed on differences in processing speed, the conflicting findings
Downloaded by [New York University] at 06:34 24 June 2015
PROCESSING SPEED AND ATTENTION IN MS
concerning MS patients’ performance on the ANT are largely resolved. To do so, we examined network effects for MS patients and healthy controls first when these effects were assessed using difference scores and then when they were assessed using the proportional scores adopted by Urbanek et al. (2010) and the residualized scores adopted by Crivelli et al. (2012). We contend that a more accurate assessment of the impact of MS on attention networks emerges when the substantial between-group differences in processing speed are controlled using proportional or residualized scores. Others have recommended comparable corrections for information processing speed when examining attentional performance on the ANT (e.g., Faust & Balota, 1997; Fernández-Duque & Black, 2006) and similarly formatted measures (e.g., Capitani, Laiacona, Barbarotto, & Cossa, 1999; Denney & Lynch, 2009). All previous ANT studies of MS patients have been confined to patients with relapsing-remitting MS (RRMS). Although the course of MS is usually characterized by episodes of relapse separated by distinct periods of complete or partial remission, for most patients this pattern is eventually replaced by a general worsening of the disease without obviously distinguishable relapses, known as secondary progressive MS (SPMS). We included patients with both RRMS and SPMS in the present study. Although deficits in processing speed are evident in the earliest stage of MS (Bergendal, Fredrikson, & Almkvist, 2007; Bodling, Denney, & Lynch, 2009; DeLuca et al., 2004; Schulz, Kopp, Kunkel, & Faiss, 2006), these deficits generally worsen with disease progression, and it has also been suggested that impairment extends to other cognitive domains (de Sonneville et al., 2002). Our decision to include RRMS and SPMS patients was predicated on the assumption that attention might be one of those alternative domains with deficits more evident in secondary progressive patients. In summary, the present study examined information processing speed and attention in a samples of patients with RRMS and SPMS and healthy controls. Attention was evaluated on the basis of the discrepancies between RTs obtained under various stimulus conditions on the ANT, using difference scores, proportional scores, and residualized scores. The main hypothesis was that diminished processing speed would be evident in patients relative to controls, but that when these differences were sufficiently accounted for, the only differences in attention would be confined to the alerting network.
3
METHOD Participants Forty patients with clinically definite MS (Polman et al., 2011) and 40 healthy controls of comparable demographics (age, sex, education) were recruited for this study. The patients were diagnosed with RRMS (N = 20) or SPMS (N = 20) and were under the care of the same neurologist at a large university medical center. Patients ranged in age from 28 to 58 years (M = 47.1, SD = 8.4), and their years of education ranged from 12 to 30 (M = 16.6, SD = 4.5). Patients had been diagnosed with MS for 1 to 34 years (M = 14.7, SD = 8.0), and disability ratings determined by the Expanded Disability Status Scale (EDSS; Kurtzke, 1983) at the time of recruitment ranged from 1.0 to 8.5 (Mdn = 4.0). Exclusionary criteria for the patients were: neurological disorder other than MS; history of drug or alcohol abuse, premorbid psychological disorder, mental retardation, or head injury; current use of narcotics; visual acuity greater than 20/50 (corrected) or impaired color vision; disabling symptomatic involvement of the hands (e.g., paresthesia, paralysis); MS relapse within the past 30 days; or cognitive impairment of sufficient severity to interfere with comprehension of testing instructions. A convenience sample of healthy controls was recruited through personal contacts of the research staff or other staff members at the medical center. In addition to the exclusionary criteria applied to patients, controls were also excluded if they had any chronic medical condition or were taking any medications other than nutritional supplements, birth control, or low-dose aspirin. Controls ranged in age from 22 to 58 years (M = 46.2, SD = 8.7) and had between 14 and 21 years of education (M = 16.8, SD = 1.6).
Procedure The study was approved by the Human Subject Committee of the University of Kansas Medical Center, and all participants provided informed consent. Eligible patients were informed of the study during their regularly scheduled appointment at the MS clinic. After a patient consented to the study, a testing session was scheduled. During the testing session, all participants completed a battery of four tests administered in the same order: a 25-item test of simple reaction time; a 25-item test of choice reaction time; a computerized version of
4
ROTH, DENNEY, LYNCH
the Stroop test; and the ANT. Testing sessions were typically completed within 60 min; the time required for the first three tests was less than 10 min. Because the scores for the Stroop test were not directly relevant to the aims of the present study, they are not reported here.
Measures
Downloaded by [New York University] at 06:34 24 June 2015
Simple reaction time (SRT) This 25-item computerized task measured simple RT to a change in a visual stimulus (Hughes et al., 2011). A fixation stimulus (+) was presented for a variable duration (1500–4000 ms) before changing to the target stimulus (000), at which time the participant pressed a response button located below the touch pad on the laptop computer. The mean RT and standard deviation for all items were recorded. Choice reaction time (CRT) This 25-item computerized task was similar to the SRT (Hughes et al., 2011). Fixation stimuli (+) were displayed on both the left and the right sides of the computer screen. After a variable interval (1500–4000 ms), one of the two stimuli changed to the target stimulus (0), at which time the participant pressed the corresponding left or right
Figure 1. Design of the Attention Network Test.
response button beneath the touch pad. The overall percentage accuracy and the mean RT and standard deviation for correct items were recorded.
Attention Network Test (ANT) This study used the computerized ANT program described by Fan et al. (2002) and Fan (n.d.). The ANT measures RTs for determining whether a target stimulus (arrow) is pointing left or right. Each trial consisted of the following sequence (Figure 1). First, a fixation point (+) was presented in the center of the screen for a variable duration (400–1600 ms). Second, a warning cue (*) was presented for 100 ms. The warning cue provided either temporal information alone indicating that the target was about to appear (center cue, double cue), or temporal plus spatial information indicating that the target would be above or below the fixation point (spatial cue). On 25% of the trials, no warning cue was presented (no cue). Next, a short fixation period (400 ms) occurred, followed by the target stimulus. On two thirds of the trials, the target was accompanied by flankers consisting of four additional arrows—two on each side of the target—that pointed either in the same direction (congruent) or in the opposite direction (incongruent) to the target arrow. No flankers appeared with the target on the remaining 33% of the trials. The target
Downloaded by [New York University] at 06:34 24 June 2015
PROCESSING SPEED AND ATTENTION IN MS
remained on the screen until either the participant responded by pressing the left or right button beneath the touch pad or 1700 ms had elapsed. Any response occurring after 1700 ms was scored as incorrect. A brief posttarget fixation interval occurred before the start of the next trial. The task was introduced with a set of instructions, followed by an initial practice block of 24 trials accompanied by feedback regarding accuracy. If the participant appeared to struggle on the practice trials, the instructions were reviewed or testing was discontinued. The four warning cue conditions (no cue, double cue, center cue, spatial cue) were fully crossed with the three flanker conditions (no flankers, congruent flankers, incongruent flankers), and the resulting stimuli were presented in a randomized order with 96 trials in each of three blocks, for a total of 288 trials. Participants could take short breaks between each block. Participants’ median RTs for correct trials were computed for each of the 12 Warning Cue × Flanker conditions. Consistent with the method used by Fan et al. (2002), the mean of the median scores was then computed to combine the RTs pertaining to each overall warning cue or flanker condition. For example, an individual’s score for center cue was the mean of the median RTs for the center cue conditions that used no flankers, congruent flankers, and incongruent flankers. We also computed two additional scores replicating the measures of processing speed formulated by Urbanek et al. (2010) and Crivelli et al. (2012). The former consisted of the overall mean of the median scores for all 12 Warning Cue × Flanker conditions (RTall). The latter consisted of the median RT on the 24 trials in which there were no warning cues and no flankers accompanying the target (RTno).
5
Determination of network effects The formulations of the scores pertaining to the various attention networks are summarized in Table 1. Differences scores (dif) for each of the attention networks were computed by subtracting the score when the critical stimulus condition was in effect from the baseline score. Thus, the difference scores for the alerting network (Adif) were obtained by subtracting the RT for all double cue trials from the RT for all no cue trials. Similarly, the difference scores for the orienting network (Odif) were obtained by subtracting the RT for spatial cue trials from the RT for center cue trials. The difference scores for the executive control network (ECdif) were obtained by subtracting the RT for incongruent trials from the RT for congruent trials and then changing the sign to eliminate negative values. It must be noted that in the first two instances, larger scores indicate stronger alerting or orienting effects, whereas in this last instance, larger scores indicate weaker executive control. This disparity in direction comes about because the critical score is always subtracted from the baseline score; the stimulus condition that is critical to the alerting effect (i.e., double cue) or the orienting effect (spatial cue) is likely to reduce the RTs from those of the respective baseline conditions (no cue or center cue); but the stimulus condition that is critical to the executive control effect (i.e., incongruent flanker) is likely to increase the RTs from those of the baseline condition (congruent flanker). This same pattern of directionality was maintained for all network scores computed in this study. Proportional scores (pro) were determined by dividing the difference score for each network by the measure of information processing speed used by Urbanek et al. (2010): the mean of the median
TABLE 1 Formulation of attention network scores
Measure Difference scores (dif) Proportional scores (pro)
Alerting network (A)
Orienting network (O)
Executive control network (EC)
Adif = RTnc – RTdc Apro = (RTnc – RTdc)/RTall
Odif = RTcc – RTsc Opro = (RTcc – RTsc)/RTall
ECdif = –1(RTcf – RTif) ECpro = [–1(RTcf – RTif)]/ RTall
(Unstandardized residuals following the regression of Y on X) Residualized scores (res)
Ares: Y = RTnc – RTdc X = RTno
Ores: Y = RTcc – RTsc X = RTno
ECres: Y = –1(RTcf – RTif) X = RTno
Note. RT = reaction time; nc = no cue condition; dc = double cue condition; cc = center cue condition; sc = spatial cue condition; cf = congruent flanker condition; if = incongruent flanker condition; all = median reaction time on all trials of the ANT; no = median reaction time on ANT trials with no cues and no flankers.
6
ROTH, DENNEY, LYNCH
scores for all 12 Warning Cue × Flanker conditions (i.e., RTall). Residualized scores (res) were computed by regressing the difference score for each network on the measure of information processing used by Crivelli et al. (2012): the median RT for ANT trails with no warning cues and no flankers (i.e., RTno).
controls 2.5%). Outlier scores were replaced with the next highest value plus 1 (Field, 2009). When these corrections were performed on RTs, network effects were adjusted accordingly. Such adjustments affected only 0.8% of the network scores for patients and none of the network scores for controls. Analyses performed on data before and after adjustment for outliers resulted in virtually the same findings on all measures. The overall accuracy across all trials of the ANT was lower for patients than for controls (patients: 94.7 ± 6.4; controls: 97.5 ± 2.4), t(78) = 2.67, p = .010; η2 = .08. Paired comparisons following ANOVA showed that the difference was due to lower accuracy for the SPMS patients (93.4 ± 8.3; p = .006); the RRMS patients (96.0 ± 3.4) did not differ from controls (p = .69). All RTs for the ANT are based on correct trials.
Downloaded by [New York University] at 06:34 24 June 2015
Statistical analyses All analyses were performed using SPSS Version 20. Groups were compared with respect to gender using chi square. Differences on other demographic variables and performance on cognitive tasks were assessed using independent-samples t tests to compare patients versus controls and oneway analyses of variance (ANOVAs) to compare MS subtypes and controls. To maintain consistency with other studies, no correction for multiple comparisons was used in conjunction with the t tests. On the other hand, the post hoc paired comparisons following significant one-way ANOVAs were adjusted using Bonferroni’s procedure. Effect sizes between groups were assessed with eta-squared (η2).
Information processing speed Table 3 presents the measures of information processing speed on the SRT and the CRT, as well as those derived from the ANT, including the RT across all trials of the ANT (RTall), the median RT for trials with no warning cue and no flankers trials (RTno), and the average RT for each cue condition. Relative to controls, MS patients had longer RTs on all of these measures of processing speed. All differences were significant, with effect sizes (η2) ranging from .19 to .27.
RESULTS Demographic and disease-related information is presented in Table 2. Patients and controls did not differ in age, years of education, or gender distribution. Individual outliers for the six cue conditions of the ANT (i.e., no cue, double cue, center cue, spatial cue, congruent flanker, and incongruent flanker) were identified as any value beyond three standard deviations from their respective group mean. Outliers occurred rarely on the RTs (patients 1.7%; controls 0.2%) and on the overall accuracy scores for the ANT (patients 3.8%;
Attention networks Network effects for patients and controls based on the ANT are also presented in Table 3. When network effects were calculated as difference scores, the only differences between MS patients and controls occurred on the executive control network [ECdif: t(78) = 2.4, p = .019; η2 = .09], with
TABLE 2 Demographic and clinical characteristics of patient and control groups All patients (N = 40; 33F/7M) Variable Age (years) Education (years) Duration of MS (years) EDSS median
M
SD
47.1 16.6 14.7 4.0
8.4 4.6 8.0
RRMS (N = 20; 17F/3M)
Range
M
SD 8.9 4.3 7.8
1.0–8.5
43.5 16.7 10.9 3.0
SPMS (N = 2016; F/4M)
Range
M
SD 6.2 4.8 6.4
1.0–6.0
50.7 16.6 18.5 6.0
Range
Controls (N = 40; 31F/9M) M
SD
46.2 16.8
8.7 1.6
4.0–8.5
Note. MS = multiple sclerosis; RRMS = relapsing remitting multiple sclerosis; SPMS = secondary progressive multiple sclerosis; EDSS = Expanded Disability Status Scale; F = female; M = male.
PROCESSING SPEED AND ATTENTION IN MS
7
TABLE 3 Comparison of patients and controls on information processing speed and attention networks All patients (N = 40)
Downloaded by [New York University] at 06:34 24 June 2015
Measure Information processing speed SRT CRT ANT All trials No cue/no flanker trials Cue conditions No cue Double Center Spatial Congruent Incongruent Attention networks Alerting Adif Apro Ares Orienting Odif Opro Ores Executive control ECdif ECpro ECres
RRMS (N = 20)
SPMS (N = 20)
Controls (N = 40)
M
SD
M
SD
M
SD
M
SD
pa
ηb
577.4 620.5
163.2 147.5
577.3 587.7
196.9 143.8
577.5 653.4
126.0 147.3
459.8 508.5
66.6 52.7
Journal of Clinical and Experimental Neuropsychology Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ncen20
Information processing speed and attention in multiple sclerosis: Reconsidering the Attention Network Test (ANT) a
a
b
Alexandra K. Roth , Douglas R. Denney & Sharon G. Lynch a
Department of Psychology, University of Kansas, Lawrence, KS, USA
b
Department of Neurology, University of Kansas Medical Center, Kansas City, KS, USA Published online: 26 May 2015.
Click for updates To cite this article: Alexandra K. Roth, Douglas R. Denney & Sharon G. Lynch (2015): Information processing speed and attention in multiple sclerosis: Reconsidering the Attention Network Test (ANT), Journal of Clinical and Experimental Neuropsychology, DOI: 10.1080/13803395.2015.1037252 To link to this article: http://dx.doi.org/10.1080/13803395.2015.1037252
PLEASE SCROLL DOWN FOR ARTICLE Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content. This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions
Journal of Clinical and Experimental Neuropsychology, 2015 http://dx.doi.org/10.1080/13803395.2015.1037252
Information processing speed and attention in multiple sclerosis: Reconsidering the Attention Network Test (ANT) Alexandra K. Roth1, Douglas R. Denney1, and Sharon G. Lynch2 1 2
Department of Psychology, University of Kansas, Lawrence, KS, USA Department of Neurology, University of Kansas Medical Center, Kansas City, KS, USA
Downloaded by [New York University] at 06:34 24 June 2015
(Received 14 October 2014; accepted 30 March 2015) Objective: The Attention Network Test (ANT) assesses attention in terms of discrepancies between response times to items that differ in the burden they place on some facet of attention. However, simple arithmetic difference scores commonly used to capture these discrepancies fail to provide adequate control for information processing speed, leading to distorted findings when patient and control groups differ markedly in the speed with which they process and respond to stimulus information. This study examined attention networks in patients with multiple sclerosis (MS) using simple difference scores, proportional scores, and residualized scores that control for processing speed through statistical regression. Method: Patients with relapsing-remitting (N = 20) or secondary progressive (N = 20) MS and healthy controls (N = 40) of similar age, education, and gender completed the ANT. Results: Substantial differences between patients and controls were found on all measures of processing speed. Patients exhibited difficulties in the executive control network, but only when difference scores were considered. When deficits in information processing speed were adequately controlled using proportional or residualized score, deficits in the alerting network emerged. The effect sizes for these deficits were notably smaller than those for overall information processing speed and were also limited to patients with secondary progressive MS. Conclusions: Deficits in processing speed are more prominent in MS than those involving attention, and when the former are properly accounted for, differences in the latter are confined to the alerting network. Keywords: Multiple sclerosis; Attention Network Test; Reaction time; Information processing speed; Alerting network; Executive control network.
Many investigators have pointed to a generalized slowing in information processing speed as the primary cognitive deficit associated with multiple sclerosis (MS; Chiavaralloti & DeLuca, 2008; DeLuca, Chelune, Tulsky, Lengenfelder, & Chiaravalloti, 2004; Denney, Gallagher, & Lynch, 2011; de Sonneville et al., 2002; Kail, 1998; Kujala, Portin, Revonsuo, & Ruutiainen, 1994; Lengenfelder et al., 2006; Macniven et al., 2008; Reicker, Tombaugh, Walker, & Freedman, 2007). Patients with MS complete substantially fewer items on rapid serial processing (RSP) tests such as
the Paced Auditory Serial Addition Test (PASAT: Kujala, Portin, Revonsuo, & Ruutiainen, 1995; Litvan, Grafman, Vendrell, & Martinez, 1988) and Symbol Digit Modalities Test (SDMT: Parmenter, Weinstock-Guttman, Garg, Munschauer, & Benedict, 2007) and also exhibit substantially longer response times on a variety of simple and choice reaction time (RT) measures (Bodling, Denney, & Lynch, 2012; de Sonneville et al., 2002; Hughes, Denney, & Lynch, 2011; Kujala et al., 1994; Reicker et al., 2007; Tombaugh, Berrigan, Walker, & Freedman, 2010).
This research was presented at the 29th Annual Congress of the European Committee for Treatment and Research in Multiple Sclerosis (ECTRIMS), in Copenhagen, Denmark, 4 October 2013. This project was unfunded, and the authors have no financial or other conflicts of interests to report. Address correspondence to: Douglas R. Denney, Department of Psychology, 1415 Jayhawk Blvd., Lawrence, KS 66045-7556, USA (E‑mail: [email protected]).
© 2015 Taylor & Francis
Downloaded by [New York University] at 06:34 24 June 2015
2
ROTH, DENNEY, LYNCH
Problems with attention are often cited as another feature of cognitive impairment in patients with MS (Kujala et al., 1995; McCarthy, Beaumont, Thompson, & Peacock, 2005; Paul, Beatty, Schneider, Blanco, & Hames, 1998). As others (Chiaravalloti & DeLuca, 2008; de Sonneville et al., 2002; Kujala et al., 1994, 1995; Macniven et al., 2008) have noted, however, measures of attention are often overlapped with information processing speed. Information processing speed involves how quickly an individual is able to perform a variety of cognitive tasks, often varying in the complexity of the cognitive operations required. On the other hand, attention is often evaluated by examining the discrepancy in performance on speeded tasks that differ in the burden they place on some facet of attention—for example, the difference between the time required to name letters that appear with or without a warning signal. A central concern of the present study is the way in which such discrepancies in performance are measured. The Attention Network Test (ANT; Fan, McCandliss, Sommer, Raz, & Posner, 2002) is a widely used instrument designed to assess three neural networks comprising Posner’s model of attention (Posner & Petersen, 1990): the alerting network that maintains a state of alertness; the orienting network that extracts spatial information from sensory input; and the executive control network that resolves conflicting information. The ANT measures choice RTs to a directional target that is accompanied by a variety of warning cues and flanker conditions. Network scores are based on differences between the RTs for different conditions. Thus, the alerting effect is measured by subtracting the average RT when a cue signals the imminent appearance of the target from the average RT when no such warning cue is provided. The orienting effect is measured by subtracting the average RT when the warning cue also provides information as to where the target will appear from the average RT when the cue is lacking in this spatial information. The executive control effect is measured by subtracting the average RT when the target is accompanied by congruent flankers from the average RT when the target is accompanied by incongruent flankers. One reason for focusing on the discrepancies in RTs is to distinguish the attention network effect from general information processing speed that affects RTs on both measures. However, assessing network effects on the basis of simple difference scores in studies comparing groups that differ substantially in their speed of information processing can lead to distorted findings because of the
disparity between the groups’ performance on whichever measure is serving as the baseline for the difference score. In other words, while providing an adequate control for within-group differences in information processing speed, we contend that simple difference scores are inadequate for controlling for between-group differences in this attribute. Several studies have compared MS patients and controls on the attention networks comprising Posner’s model (Crivelli et al., 2012; Ishigami, Fisk, Wojtowicz, & Klein, 2013; Omisade et al., 2012; Urbanek et al., 2010; Wojtowicz, Omisade, & Fisk, 2013). No significant differences in specific network effects have been found when small samples of patients and controls are compared (e.g., N = 11: Ishigami et al., 2013; N = 12: Omisade et al., 2012), probably due to these studies’ insufficient statistical power. However, when larger samples were examined, both Urbanek et al. (2010) and Crivelli et al. (2012) reported deficits in patients’ alerting network, whereas Wojtowicz et al. (2013) found impaired performance on the executive control network. The conflicting findings in these latter studies may stem from the way in which discrepancies in the RTs involved in each network score were treated. The network effects reported by Wojtowicz et al. (2013) were based on simple difference scores. Both Urbanek et al. (2010) and Crivelli et al. (2012) evaluated network effects with procedures designed to account for the differences in overall processing speed between the patients and controls. Urbanek et al. used proportional scores, dividing the difference scores by the participant’s overall average RT on all of the items of the ANT. Crivelli et al. regressed difference scores on an alternative measure of processing speed, the average RT on trials with no warning cue and neutral flankers, and then used the unstandardized residual scores to examine group differences in network effects. As in many other RT studies, differences between MS patients and controls in overall information processing speed are evident in practically all of the ANT studies, and, often these differences are more robust than the network effects themselves (Ishigami et al., 2013; Omisade et al., 2012; Wojtowicz et al., 2013). The principal aim of the present study was to show that such differences in processing speed distort the findings with respect to network effects on the ANT by actually diminishing the alerting effect and augmenting the executive control effect. We intend to show that, when appropriate controls are imposed on differences in processing speed, the conflicting findings
Downloaded by [New York University] at 06:34 24 June 2015
PROCESSING SPEED AND ATTENTION IN MS
concerning MS patients’ performance on the ANT are largely resolved. To do so, we examined network effects for MS patients and healthy controls first when these effects were assessed using difference scores and then when they were assessed using the proportional scores adopted by Urbanek et al. (2010) and the residualized scores adopted by Crivelli et al. (2012). We contend that a more accurate assessment of the impact of MS on attention networks emerges when the substantial between-group differences in processing speed are controlled using proportional or residualized scores. Others have recommended comparable corrections for information processing speed when examining attentional performance on the ANT (e.g., Faust & Balota, 1997; Fernández-Duque & Black, 2006) and similarly formatted measures (e.g., Capitani, Laiacona, Barbarotto, & Cossa, 1999; Denney & Lynch, 2009). All previous ANT studies of MS patients have been confined to patients with relapsing-remitting MS (RRMS). Although the course of MS is usually characterized by episodes of relapse separated by distinct periods of complete or partial remission, for most patients this pattern is eventually replaced by a general worsening of the disease without obviously distinguishable relapses, known as secondary progressive MS (SPMS). We included patients with both RRMS and SPMS in the present study. Although deficits in processing speed are evident in the earliest stage of MS (Bergendal, Fredrikson, & Almkvist, 2007; Bodling, Denney, & Lynch, 2009; DeLuca et al., 2004; Schulz, Kopp, Kunkel, & Faiss, 2006), these deficits generally worsen with disease progression, and it has also been suggested that impairment extends to other cognitive domains (de Sonneville et al., 2002). Our decision to include RRMS and SPMS patients was predicated on the assumption that attention might be one of those alternative domains with deficits more evident in secondary progressive patients. In summary, the present study examined information processing speed and attention in a samples of patients with RRMS and SPMS and healthy controls. Attention was evaluated on the basis of the discrepancies between RTs obtained under various stimulus conditions on the ANT, using difference scores, proportional scores, and residualized scores. The main hypothesis was that diminished processing speed would be evident in patients relative to controls, but that when these differences were sufficiently accounted for, the only differences in attention would be confined to the alerting network.
3
METHOD Participants Forty patients with clinically definite MS (Polman et al., 2011) and 40 healthy controls of comparable demographics (age, sex, education) were recruited for this study. The patients were diagnosed with RRMS (N = 20) or SPMS (N = 20) and were under the care of the same neurologist at a large university medical center. Patients ranged in age from 28 to 58 years (M = 47.1, SD = 8.4), and their years of education ranged from 12 to 30 (M = 16.6, SD = 4.5). Patients had been diagnosed with MS for 1 to 34 years (M = 14.7, SD = 8.0), and disability ratings determined by the Expanded Disability Status Scale (EDSS; Kurtzke, 1983) at the time of recruitment ranged from 1.0 to 8.5 (Mdn = 4.0). Exclusionary criteria for the patients were: neurological disorder other than MS; history of drug or alcohol abuse, premorbid psychological disorder, mental retardation, or head injury; current use of narcotics; visual acuity greater than 20/50 (corrected) or impaired color vision; disabling symptomatic involvement of the hands (e.g., paresthesia, paralysis); MS relapse within the past 30 days; or cognitive impairment of sufficient severity to interfere with comprehension of testing instructions. A convenience sample of healthy controls was recruited through personal contacts of the research staff or other staff members at the medical center. In addition to the exclusionary criteria applied to patients, controls were also excluded if they had any chronic medical condition or were taking any medications other than nutritional supplements, birth control, or low-dose aspirin. Controls ranged in age from 22 to 58 years (M = 46.2, SD = 8.7) and had between 14 and 21 years of education (M = 16.8, SD = 1.6).
Procedure The study was approved by the Human Subject Committee of the University of Kansas Medical Center, and all participants provided informed consent. Eligible patients were informed of the study during their regularly scheduled appointment at the MS clinic. After a patient consented to the study, a testing session was scheduled. During the testing session, all participants completed a battery of four tests administered in the same order: a 25-item test of simple reaction time; a 25-item test of choice reaction time; a computerized version of
4
ROTH, DENNEY, LYNCH
the Stroop test; and the ANT. Testing sessions were typically completed within 60 min; the time required for the first three tests was less than 10 min. Because the scores for the Stroop test were not directly relevant to the aims of the present study, they are not reported here.
Measures
Downloaded by [New York University] at 06:34 24 June 2015
Simple reaction time (SRT) This 25-item computerized task measured simple RT to a change in a visual stimulus (Hughes et al., 2011). A fixation stimulus (+) was presented for a variable duration (1500–4000 ms) before changing to the target stimulus (000), at which time the participant pressed a response button located below the touch pad on the laptop computer. The mean RT and standard deviation for all items were recorded. Choice reaction time (CRT) This 25-item computerized task was similar to the SRT (Hughes et al., 2011). Fixation stimuli (+) were displayed on both the left and the right sides of the computer screen. After a variable interval (1500–4000 ms), one of the two stimuli changed to the target stimulus (0), at which time the participant pressed the corresponding left or right
Figure 1. Design of the Attention Network Test.
response button beneath the touch pad. The overall percentage accuracy and the mean RT and standard deviation for correct items were recorded.
Attention Network Test (ANT) This study used the computerized ANT program described by Fan et al. (2002) and Fan (n.d.). The ANT measures RTs for determining whether a target stimulus (arrow) is pointing left or right. Each trial consisted of the following sequence (Figure 1). First, a fixation point (+) was presented in the center of the screen for a variable duration (400–1600 ms). Second, a warning cue (*) was presented for 100 ms. The warning cue provided either temporal information alone indicating that the target was about to appear (center cue, double cue), or temporal plus spatial information indicating that the target would be above or below the fixation point (spatial cue). On 25% of the trials, no warning cue was presented (no cue). Next, a short fixation period (400 ms) occurred, followed by the target stimulus. On two thirds of the trials, the target was accompanied by flankers consisting of four additional arrows—two on each side of the target—that pointed either in the same direction (congruent) or in the opposite direction (incongruent) to the target arrow. No flankers appeared with the target on the remaining 33% of the trials. The target
Downloaded by [New York University] at 06:34 24 June 2015
PROCESSING SPEED AND ATTENTION IN MS
remained on the screen until either the participant responded by pressing the left or right button beneath the touch pad or 1700 ms had elapsed. Any response occurring after 1700 ms was scored as incorrect. A brief posttarget fixation interval occurred before the start of the next trial. The task was introduced with a set of instructions, followed by an initial practice block of 24 trials accompanied by feedback regarding accuracy. If the participant appeared to struggle on the practice trials, the instructions were reviewed or testing was discontinued. The four warning cue conditions (no cue, double cue, center cue, spatial cue) were fully crossed with the three flanker conditions (no flankers, congruent flankers, incongruent flankers), and the resulting stimuli were presented in a randomized order with 96 trials in each of three blocks, for a total of 288 trials. Participants could take short breaks between each block. Participants’ median RTs for correct trials were computed for each of the 12 Warning Cue × Flanker conditions. Consistent with the method used by Fan et al. (2002), the mean of the median scores was then computed to combine the RTs pertaining to each overall warning cue or flanker condition. For example, an individual’s score for center cue was the mean of the median RTs for the center cue conditions that used no flankers, congruent flankers, and incongruent flankers. We also computed two additional scores replicating the measures of processing speed formulated by Urbanek et al. (2010) and Crivelli et al. (2012). The former consisted of the overall mean of the median scores for all 12 Warning Cue × Flanker conditions (RTall). The latter consisted of the median RT on the 24 trials in which there were no warning cues and no flankers accompanying the target (RTno).
5
Determination of network effects The formulations of the scores pertaining to the various attention networks are summarized in Table 1. Differences scores (dif) for each of the attention networks were computed by subtracting the score when the critical stimulus condition was in effect from the baseline score. Thus, the difference scores for the alerting network (Adif) were obtained by subtracting the RT for all double cue trials from the RT for all no cue trials. Similarly, the difference scores for the orienting network (Odif) were obtained by subtracting the RT for spatial cue trials from the RT for center cue trials. The difference scores for the executive control network (ECdif) were obtained by subtracting the RT for incongruent trials from the RT for congruent trials and then changing the sign to eliminate negative values. It must be noted that in the first two instances, larger scores indicate stronger alerting or orienting effects, whereas in this last instance, larger scores indicate weaker executive control. This disparity in direction comes about because the critical score is always subtracted from the baseline score; the stimulus condition that is critical to the alerting effect (i.e., double cue) or the orienting effect (spatial cue) is likely to reduce the RTs from those of the respective baseline conditions (no cue or center cue); but the stimulus condition that is critical to the executive control effect (i.e., incongruent flanker) is likely to increase the RTs from those of the baseline condition (congruent flanker). This same pattern of directionality was maintained for all network scores computed in this study. Proportional scores (pro) were determined by dividing the difference score for each network by the measure of information processing speed used by Urbanek et al. (2010): the mean of the median
TABLE 1 Formulation of attention network scores
Measure Difference scores (dif) Proportional scores (pro)
Alerting network (A)
Orienting network (O)
Executive control network (EC)
Adif = RTnc – RTdc Apro = (RTnc – RTdc)/RTall
Odif = RTcc – RTsc Opro = (RTcc – RTsc)/RTall
ECdif = –1(RTcf – RTif) ECpro = [–1(RTcf – RTif)]/ RTall
(Unstandardized residuals following the regression of Y on X) Residualized scores (res)
Ares: Y = RTnc – RTdc X = RTno
Ores: Y = RTcc – RTsc X = RTno
ECres: Y = –1(RTcf – RTif) X = RTno
Note. RT = reaction time; nc = no cue condition; dc = double cue condition; cc = center cue condition; sc = spatial cue condition; cf = congruent flanker condition; if = incongruent flanker condition; all = median reaction time on all trials of the ANT; no = median reaction time on ANT trials with no cues and no flankers.
6
ROTH, DENNEY, LYNCH
scores for all 12 Warning Cue × Flanker conditions (i.e., RTall). Residualized scores (res) were computed by regressing the difference score for each network on the measure of information processing used by Crivelli et al. (2012): the median RT for ANT trails with no warning cues and no flankers (i.e., RTno).
controls 2.5%). Outlier scores were replaced with the next highest value plus 1 (Field, 2009). When these corrections were performed on RTs, network effects were adjusted accordingly. Such adjustments affected only 0.8% of the network scores for patients and none of the network scores for controls. Analyses performed on data before and after adjustment for outliers resulted in virtually the same findings on all measures. The overall accuracy across all trials of the ANT was lower for patients than for controls (patients: 94.7 ± 6.4; controls: 97.5 ± 2.4), t(78) = 2.67, p = .010; η2 = .08. Paired comparisons following ANOVA showed that the difference was due to lower accuracy for the SPMS patients (93.4 ± 8.3; p = .006); the RRMS patients (96.0 ± 3.4) did not differ from controls (p = .69). All RTs for the ANT are based on correct trials.
Downloaded by [New York University] at 06:34 24 June 2015
Statistical analyses All analyses were performed using SPSS Version 20. Groups were compared with respect to gender using chi square. Differences on other demographic variables and performance on cognitive tasks were assessed using independent-samples t tests to compare patients versus controls and oneway analyses of variance (ANOVAs) to compare MS subtypes and controls. To maintain consistency with other studies, no correction for multiple comparisons was used in conjunction with the t tests. On the other hand, the post hoc paired comparisons following significant one-way ANOVAs were adjusted using Bonferroni’s procedure. Effect sizes between groups were assessed with eta-squared (η2).
Information processing speed Table 3 presents the measures of information processing speed on the SRT and the CRT, as well as those derived from the ANT, including the RT across all trials of the ANT (RTall), the median RT for trials with no warning cue and no flankers trials (RTno), and the average RT for each cue condition. Relative to controls, MS patients had longer RTs on all of these measures of processing speed. All differences were significant, with effect sizes (η2) ranging from .19 to .27.
RESULTS Demographic and disease-related information is presented in Table 2. Patients and controls did not differ in age, years of education, or gender distribution. Individual outliers for the six cue conditions of the ANT (i.e., no cue, double cue, center cue, spatial cue, congruent flanker, and incongruent flanker) were identified as any value beyond three standard deviations from their respective group mean. Outliers occurred rarely on the RTs (patients 1.7%; controls 0.2%) and on the overall accuracy scores for the ANT (patients 3.8%;
Attention networks Network effects for patients and controls based on the ANT are also presented in Table 3. When network effects were calculated as difference scores, the only differences between MS patients and controls occurred on the executive control network [ECdif: t(78) = 2.4, p = .019; η2 = .09], with
TABLE 2 Demographic and clinical characteristics of patient and control groups All patients (N = 40; 33F/7M) Variable Age (years) Education (years) Duration of MS (years) EDSS median
M
SD
47.1 16.6 14.7 4.0
8.4 4.6 8.0
RRMS (N = 20; 17F/3M)
Range
M
SD 8.9 4.3 7.8
1.0–8.5
43.5 16.7 10.9 3.0
SPMS (N = 2016; F/4M)
Range
M
SD 6.2 4.8 6.4
1.0–6.0
50.7 16.6 18.5 6.0
Range
Controls (N = 40; 31F/9M) M
SD
46.2 16.8
8.7 1.6
4.0–8.5
Note. MS = multiple sclerosis; RRMS = relapsing remitting multiple sclerosis; SPMS = secondary progressive multiple sclerosis; EDSS = Expanded Disability Status Scale; F = female; M = male.
PROCESSING SPEED AND ATTENTION IN MS
7
TABLE 3 Comparison of patients and controls on information processing speed and attention networks All patients (N = 40)
Downloaded by [New York University] at 06:34 24 June 2015
Measure Information processing speed SRT CRT ANT All trials No cue/no flanker trials Cue conditions No cue Double Center Spatial Congruent Incongruent Attention networks Alerting Adif Apro Ares Orienting Odif Opro Ores Executive control ECdif ECpro ECres
RRMS (N = 20)
SPMS (N = 20)
Controls (N = 40)
M
SD
M
SD
M
SD
M
SD
pa
ηb
577.4 620.5
163.2 147.5
577.3 587.7
196.9 143.8
577.5 653.4
126.0 147.3
459.8 508.5
66.6 52.7
