The Contribution of Cognition and Spasticity to Driving Performance in Multiple Sclerosis

Abstract

Objective

To examine the independent and combined impact of cognitive dysfunction and spasticity on driving tasks involving high cognitive workload and lower-limb mobility in individuals with multiple sclerosis (MS).

Design

Single-visit cohort study.

Setting

Clinical research center.

Participants

Seventeen drivers with MS and 14 normal controls. The MS group exhibited a broad range of cognitive functioning and disability. Eight MS patients had significant spasticity in the knee proximal to the pedals (based on the Modified Ashworth Scale).

Interventions

Not applicable.

Main Outcome Measures

A brief neuropsychologic test battery and 2 driving simulations. Simulation 1 required participants to maintain a constant speed and lane position while attending to a secondary task. Simulation 2 required participants to adjust their speed to accelerations and decelerations of a lead car in front of them.

Results

MS patients demonstrated greater variability in lane position (effect size g=1.30), greater difficulty in maintaining a constant speed (g=1.25), and less ability to respond to lead car speed changes (g=1.85) compared with controls. Within the MS group, in a multivariate model that included neuropsychologic and spasticity measures, cognitive functioning was the strongest predictor of difficulty in maintaining lane position during the divided attention task and poor response time to lead car speed changes, whereas spasticity was associated with reductions in accuracy of tracking the lead car movements and speed maintenance.

Conclusions

In this preliminary study, cognitive and physical impairments associated with MS were related to deficits in specific components of simulated driving, and assessment of these factors may help guide the clinician regarding the types of driving behaviors that would put MS patients at increased risk for a crash.

Neurologic sequelae associated with MS can affect everyday functioning, including the patient’s ability to safely drive an automobile. For example, MS patients have been found to have a higher rate of automobile crashes than normal people,^1,2 and to have greater difficulty on a driving simulator.³ The factors that can potentially affect driving skills include the obvious neuromotor symptoms (weakness, sensory disturbance, coordination problems, spasticity), as well as MS-related neurocognitive disturbance. Thirty to seventy percent of all MS patients experience some form of cognitive dysfunction during their lifetime.⁴ Memory, attention, processing speed, and executive functioning can be affected, with intellectual functioning and language skills typically preserved.⁵ Schultheis⁶ reported that MS participants with documented cognitive impairment performed more poorly than cognitively intact patients on measures associated with driving performance, including the Neurocognitive Driving Test and the Useful Field of View,⁷ a test of driving-related visual attention.⁸

Spasticity affects 40% to 75% of MS patients,⁹ and frequently contributes to MS-associated disability by affecting both the gross motor activity needed for ambulation and more refined motor skills.¹⁰ To focus on the potential impact of cognitive impairment on driving, many studies have opted to exclude MS participants with greater than minimal physical disability (including spasticity). This potentially excludes an important, and large, subgroup of MS patients who, while they may experience physical disabilities, might also continue to drive. In addition, excluding such patients may affect the external validity of study findings, especially if the effects of physical disabilities outweigh the effects of cognitive impairment.

The goal of the present study was to examine the contribution of both physical and cognitive sequelae to driving performance in drivers with MS-associated spasticity and cognitive dysfunction. To assess different components of driving performance, we utilized simulations that measured (1) lane tracking under high cognitive workload (limited emphasis on lower-limb mobility), and (2) response to lead car speed changes, emphasizing lower-limb mobility under limited cognitive load. We hypothesized that poor cognitive functioning would be associated with worse performance on driving behaviors with a high cognitive demand, and that driving maneuvers involving pedal movements would be most affected by lower-limb spasticity.

METHODS

Participants

Participants consisted of 17 persons with multiple sclerosis (MS) and 14 normal control participants. The MS participants were enrolled in a clinical trial for the treatment of MS-related spasticity, and only pretreatment data were used in these analyses. To be included in this study, participants had to have routinely driven an automobile, including a minimum of 400km (250 miles) in the past year. MS participants were included if they: (1) had definite or probable MS (diagnosed by a neurologist [JC-B] using the Poser criteria), (2) had complaints of spasticity and at least moderate increase in tone, (3) were between the ages of 18 and 65 years, (4) were fluent in English, and (5) were on a stable dose of either baclofen (Lioresal) or tizanidine for at least 3 months. Participants were excluded if they: (1) had significant Axis I psychiatric disorder (eg, major depressive disorder, bipolar disorder) or neurologic disease other than MS, (2) had recent active substance abuse, (3) had an unstable medical problem, (4) required benzodiazepines to control spasticity, or (5) frequently used sedatives or high doses of analgesic medications. A total of 29 participants were initially contacted regarding participating in the study. Six were unwilling to make the time commitment, 3 were physically unable to use the simulator, and an additional 3 were not currently driving. Control participants were recruited from the community using advertisements and flyers, had to be current drivers, and were excluded if there was a history of any central nervous system disorder that might affect cognition, evidence of a head injury with loss of consciousness more than 30 minutes, or a history of drug abuse. The project was approved by the University of California San Diego Institutional Review Board, and all participants signed a consent form approved by the institutional review board.

The 2 groups were similar with respect to age, education, and sex (table 1), and had relatively similar total years of driving experience as well as recent driving experience. Nine of the MS participants had reduced their driving, but this occurred 2 or more years prior to this study, so kilometers (miles) driven in the past year remains a reasonable representation of recent driving habits.

Table 1.

Demographic and Driving Characteristics of the Control and MS Participants

Demographic Variables	Control (n=14)	MS (n=17)	P
Age (y)	47.7±11.7	49.5±7.9	.611
Education (y)	14.4±1.8	14.5±2.1	.890
Female, n (%)	9 (64)	11 (65)	.981
Kilometers driven in past year†	9569 (5344–17197)	6293 (3521–9938)	.081*

^*Wilcoxon rank sum test.

^†Median (interquartile range)

The MS group evidenced a range of disability as assessed by the Kurtzke EDSS.¹¹ The median score for the group was 6.0 (intermittent or unilateral constant assistance to walk about 100m; EDSS range, 3.0–7.5). On the MAS¹² (see below), 8 (47%) of 17 of the group had a spasticity score of greater than 2 in the knee used to manipulate the accelerator and brake pedal, while 9 (53%) of 17 did not. Visual function was assessed via the visual function section of the EDSS (none of the participants had a scotomoa), as well as self-report (MSQLI). For this study, we focused on MSQLI complaints regarding near vision (reading letters, printed materials, dials), because this would most likely relate to simulator performance. Seven (41%) of the MS participants reported that viewing these items was somewhat difficult, whereas the remaining participants reported no difficulties.

Procedures

Neuropsychologic testing

MS participants completed a brief neuropsychologic test battery designed to assess domains frequently affected by MS. The battery assessed motor functioning and dexterity (Grooved Pegboard Test^13,14), speed of information processing (TMT Part A^13,15,16, Wechsler Adult Intelligence Scale—Third Edition digit symbol^17,18), attention and working memory (Paced Auditory Serial Addition Test^19,20), executive functioning (TMT Part B [TMT-B]^15,16,21), and learning and memory (HVLT-R^22,23). Control subjects completed a subset of these tests (TMT, digit symbol, HVLT-R) to ensure that there were no significant, undiagnosed disorders. Due to time constraints relating to the clinical trial, additional measures that might be related to driving (eg, visuospatial or visual memory measures), were not administered. Overall cognitive functioning was estimated using the GDS approach, which considers both the number and the severity of deficits in a person’s performance throughout the test battery and weights impaired performances on cognitive tests (based on demographically corrected T scores) more heavily than average and above average performance.^24,25 Test performance in the normal range yields a GDS of 0, a GDS score of 1 would indicate that on average the person was 1 SD below expected on the cognitive measures (ie, a higher GDS indicates greater impairment), and severe impairment would be assigned a score of 5. This approach has been shown to be sensitive to cognitive dysfunction for a variety of disorders.²⁴

Spasticity assessment

MS participants were assessed for level of spasticity using the MAS,¹² an examiner-administered, independent measure of spasticity (scale range, 0–4). Spasticity was rated bilaterally for the elbow, hip, and knee. Because the focus of this study was on the effect that spasticity might have on pedal performance, we used the score for the knee used to manipulate the pedals. Consistent with how they drive their cars, three participants used their left foot for pedal manipulation. These participants did not perform any differently than the other participants on the driving assessments, that is, their scores were in the middle of the distribution for key outcome measures.

Driving simulation

All participants completed a 15-minute driving simulation. The simulation was presented on a Pentium III personal computer using a 17-inch monitor at 1280×1024 resolution, and running STISIM drive software.^a Hardware included a steering wheel, turn signal, and brake and accelerator pedals.²⁶ The system provided auditory feedback (eg, engine noise). Two separate tasks were embedded within 1 simulation. To minimize the novelty of the tests, participants completed a pretest training session of approximately 10 minutes to familiarize them with the hardware and the tasks (see below) that they were to encounter during the simulations.

The lane tracking simulation task, lasting approximately 7 minutes, required participants to maintain lane position while holding a constant speed (88kph [55mph]). Cognitive workload, and thus the sensitivity of the measure, was increased by having participants attend to divided attention tasks in the upper corners of the monitor; they were to respond to a triangle in the left or right corners of the monitor by pressing the left or right turn signal, respectively, and to a picture of a horn in either corner by pressing a button on the console. The scenery consisted of a straight, 2-lane roadway with mountains and clouds on the horizon. Subjects were not required to negotiate any curves or turns, and there were no traffic, pedestrians, or traffic control devices (eg, traffic lights). There are 3 primary components to this test, and these were the outcomes of interest: lane position, speed, and response to the divided attention stimuli. The SDLP was used to assess how well participants maintained their lane position. SDLP has been shown to have good test-retest reliability,²⁷ and previous studies have shown performance on versions of this simulator to be sensitive to driving impairments.^28–30

The car following portion of the simulation lasted approximately 8 minutes. Participants were to follow a lead vehicle at a safe and constant distance. The lead vehicle changed speed following a sinusoidal wave with amplitude of 9.6kph (6mph). Each change in speed occurred over a period of 1 minute. The mean velocity of the lead vehicle was 88kph (55mph). Rather than maintaining a constant speed, as in the lane tracking task, participants needed to repeatedly vary their speed. The primary outcomes were (1) coherence between the participant and lead cars (a general correlation [0–1] of the participant’s ability to accurately track the speed variations of the lead car); (2) time delay (or the reaction time to changes in the lead car’s speed); and (3) modulus (the average ratio of the following vehicle’s speed to the lead vehicle’s speed. This measures the degree to which participants overcompensate [>1] or undercompensate [<1] their separation distance from the lead car at any point in the time series. For example, if the participant’s vehicle’s speed is 96kph [60mph] and the lead vehicle’s speed is 48kph [30mph], then the modulus at this point in time would be 2).

Within the MS group, we hypothesized that cognition would most strongly be associated with the SDLP component of the lane tracking task, because it does not require pedal movements and is thus least likely to be affected by lower-limb spasticity. Spasticity, on the other hand, was predicted to impact the speed maintenance component of the lane tracking task, as well as all of the outcomes on the car following task, because pedal movements (accelerations and decelerations) are central to the task.

Statistical Analyses

Variables were assessed for the presence of outliers and non-normal distributions prior to analysis. When appropriate, log or Box-Cox transformations were undertaken to achieve normality. For SDLP, normality was achieved by modifying 2 deviant scores (1 MS, 1 control; operationalized as 3.0 SDs from the group mean³²) to be 1 measurement unit further from the mean than the next most deviant score in the distribution.^32,33 The outlier thus remains usable for analysis and the extreme score in the distribution, but exerts less influence on the analyses. Independent t tests were used for continuous variables.

We employed nonparametric analyses if the distributions failed to meet the assumptions of normality and homogeneity of variance. Wilcoxon rank-sum tests were used for 2 independent group analyses where normality was in question but the assumption of homogeneous variance remained tenable. For categorical variables, a chi-square test of independence was employed. When expected cell counts were less than 5, we used the Fisher exact test. Raw scores are shown in tables to aid interpretation; all statistical analyses utilized the transformed variables. Because significance levels are in part dependent on sample size, we also report effect sizes using a Hedge bias correction, which provides an estimate of the magnitude of the group differences. Effect sizes were classified as small (.20), medium (.50), and large (.80).³⁴

RESULTS

The initial analyses involved a comparison of the MS and control participants (see table 2). The MS group drove significantly faster than the control group on the lane tracking simulation, and had a greater variability in speed maintenance. The MS patients also exhibited a greater deviation in lane position (SDLP), the primary variable of interest. There was a large effect size for all group differences. Because SDLP can be affected by speed (higher speeds usually result in greater lane deviations), we entered both speed and group membership in a regression model predicting SDLP. The model was significant (P=.004, adjusted R²=.28), with group status still predicting SDLP (P=.002); speed was not a significant contributor to the model (P=.91). To rule out the possibility that upper extremity spasticity affected SDLP (eg, through reduced steering control), we examined whether there was a relationship between upper-extremity MAS scores and SDLP. There was no significant relationship between the 2, whether spasticity was assessed as a continuous or dichotomous (marked [n=4], or not) variable (P>.60). There was a trend for more MS patients to miss at least one of the divided attention targets compared with the control participants (58.8% vs 28.6%; P=.09).

Table 2.

Simulator Performance of the Control and MS Participants

Simulator Variables	Control (n=14)	MS (n=17)	Hedge Effect Size	P
Lane tracking
Average speed (kph)*	88.4±14.6	99.5±13.7	0.81	.030
Average speed deviation (kph)*	2.9±1.6	5.5 ±2.9	1.25	.002
SDLP	1.05±0.30	1.57±0.45	1.30	.001
Divided attention task
Missed at least 1 target, n (%)	4 (28.6)	10 (58.8)†		.029
Car following
Coherence†	0.88±0.09	0.61±0.22	1.85	<.001
Time delay (s)*‡	0.72 (0.32–1.18)	1.56 (0.38–3.24)	0.65	.074
Modulus‡§	1.10 (1.05–1.29)	1.12 (1.04–1.22)	0.05	.984

^*Effect size and P value based on log transformation.

^†Effect size and P value based on Box-Cox transformation.

^‡Median (interquartile range).

^§Wilcoxon rank-sum test.

On the car following task, MS participants had greater difficulty than controls in tracking the movements of the lead car (coherence correlation, .61 vs .88; large effect size) and were slower to respond to changes in the lead car speed (median delay, 0.72s vs 1.56s; medium effect size). There were no significant group differences in the degree to which the groups over- or undercompensated the distance from the lead car (modulus). The modulus and time delay findings were similar when 2 MS participants with a coherence less than 0.3 (indicating particularly inconsistent/incorrect performance on the task) were removed from the analyses.

We next sought to determine whether, within the MS group, cognitive functioning (based on the GDS) and/or spasticity indicators were predictors of simulator performance, and whether visual difficulties also contributed to the outcome. In a correlational analysis, cognitive functioning was most strongly associated with variations in lane position, and delay in responding to lead car speed changes (table 3). There was a weaker association between neuropsychologic performance and average speed, average speed deviation, coherence, and modulus. With respect to individual neuropsychologic tests, SDLP was most strongly associated with T scores on TMT-B time (r=−.49, P=.052), digit symbol (r=−.51, P=.037), and HVLT total words (r=−.63, P=.0063). Time delay was associated with performance on grooved pegboard dominant (r=−.52, P=.038) and nondominant hand (r=−.49, P=.047), digit symbol (r=−.47, P=.059), and HVLT delayed recall (r=−.53, P=.028).

Table 3.

Correlations Between Neuropsychologic Functioning (GDS) and Simulator Performance Within the MS Group

Variable	r	P
Lane tracking
SDLP	.49	.045
Average speed	.33	.195
Average speed deviation	.26	.320
Car following
Coherence	.29	.254
Time delay	.42	.093
Modulus	.27	.289

Spasticity, on the other hand, was most strongly associated with driving parameters involving pedal movement (changing and maintaining speed) (table 4). When comparing those with and without significant spasticity, the spastic subjects exhibited lower coherence and modulus on the car following task, with large effect sizes, while the differences were associated with a medium effect size for time delay and average speed deviation. There was a small effect size with respect to speed differences and SDLP.

Table 4.

Relationship Between Spasticity in the Knee Used to Operate the Pedals and Simulation Performance.

Simulator Variables	Nonspastic (n=9)	Spastic (n=8)	Hedge Effect Size	P
Lane tracking
SDLP	1.54±0.47	1.61±0.45	.15	.759
Average speed (kph)*	101.23±11.27	97.53±11.27	.26	.546
Average speed deviation (kph)*	4.67±2.41	6.60±3.22	.68	.167
Car following
Coherence†	0.71±0.14	0.50±0.25	.99	.049
Time delay (s)*‡	1.37 (0.06–2.69)	1.85 (1.12–6.23)	.70	.152
Modulus	1.22±0.16	1.06±0.11	1.04	.040

^*Based on log values.

^†Based on Box-Cox transformation.

^‡Median (interquartile range).

To determine the independent versus combined impact of cognitive function and spasticity on driving performance, we ran stepwise regressions using GDS and spasticity (yes, no) as predictors, and key driving variables as the outcomes. The correlation between the GDS and spasticity was .49 (P=.043). As would be expected, only GDS (P=.0453) entered the model predicting SDLP (adjusted R²=.19). GDS was also the only predictor in a model predicting time delay (P=.093, adjusted R²=.12). Only spasticity entered into models predicting coherence (P<.049, adjusted R²=.18), modulus (P<.040, adjusted R²=.20), and average speed deviation (P=.167, adjusted R²=.07).

Vision complaints (yes, no) were only associated with worse SDLP (P=.04), and greater speed deviation (P=.09). In a hierarchical regression, the addition of visual information to the model incorporating cognitive functioning increased the amount of SDLP variance explained from 19% to 27%. In the model for average speed deviation, the adjusted R² increased to .15, from .07, when adding visual function along with spasticity.

DISCUSSION

In the present study, drivers with MS exhibited reduced driving skills relative to normal controls across a range of driving behaviors, as assessed on driving simulations. These included greater difficulty in maintaining lane position under high cognitive demand, missing a secondary stimulus (suggesting reduced attention), and driving at a higher speed and having larger speed variability than control participants. These deficits did not appear to be the result of the MS participants prioritizing one task over the other (ie, directing more attention to the lane tracking/speed control task vs divided attention task), because performance was worse on all tasks. In addition, MS participants were slower to respond to a lead car’s accelerations and decelerations. On average, the MS group was about 1 second slower to respond to changes in lead car speed. Although this is not directly related to accident avoidance, because the changes in lead car speed were not abrupt, at a speed of 88kph (55mph) the MS group would on average take an additional 24m (80ft) to notice and respond to slowing traffic.

One goal of the present study was to identify the specific neurologic characteristics that put MS participants at increased risk for poor driving performance. Within the MS group, worse cognitive functioning was most strongly associated with difficulties in attending to multiple stimuli (maintaining lane position/speed and responding to the divided attention measure) as well as a delayed reaction in responding to lead car changes. This is consistent with, and expands on, the findings of Schultheis et al⁶ who found that cognitively impaired MS patients had slower reaction times than cognitively intact MS participants. Our study demonstrates that this performance decrement holds true in a semi-naturalistic driving environment (a fully interactive simulation in which the measurements are taken repeatedly over a period of time), and that attentional/lane tracking abilities are significantly impacted by multitasking. Clinicians might consider advising MS patients with cognitive impairments that they limit possible distractions such as diverting attention to radios, CD changers, cell phones, et cetera. A recent naturalistic study of normal drivers found that momentary inattention (eg, reaching for an object in a car) was the most common cause of crashes,³⁵ so this is good advice for all drivers. But such maneuvers may be even more risky for MS patients with cognitive decline.

As noted earlier, most MS studies have excluded people with spasticity. In this study, spasticity, rather than cognition, was strongly associated with poor performance on driving components involving pedal movements. Participants with spasticity had greater difficulty tracking the lead car speed changes, reacting to lead car changes, and maintaining a constant speed. This is not to suggest that cognitive performance is irrelevant to these activities, but does highlight the fact that spasticity can significantly impact specific driving abilities, and perhaps overshadow the relationship between cognition and driving in certain situations. This could also put individuals at increased risk for a crash. Interestingly, the MS patients showed a lower level of compensating for lead car changes on the car following task compared with the control group. This could suggest that MS patients were adjusting to their condition by being more conservative in their speed adjustments when there was a car in front of them.

Also of note, time delay (response time to lead car changes) was the variable most associated with both cognition and spasticity. Although in a multiple regression model only cognition predicted time delay, it seems likely that a driver’s ability to respond to high-risk situations would be impacted by one’s ability to anticipate the situation, divide attention, respond quickly, and make the motor movements necessary to press on the brake or accelerator. Given the relatively strong univariate relationship between spasticity and time delay, it may be that the small sample size left us underpowered to fully parcel out the joint contribution, or interaction, between spasticity and cognition on this task.

This study demonstrates both a reduction in the ability to divide attention in some MS participants, and slower foot movements. MS patients with either cognitive impairment or spasticity in the limb used to manipulate the pedals, and especially both conditions, would be advised to provide greater gaps between cars in front of them. In addition, this recommendation would likely hold true for situations such as left turns in which they must cross in front of oncoming traffic, one of the most common accident scenarios.

Patients with MS often experience visual difficulties. Although in the present study visual dysfunction only related to select simulator performance variables (lane deviation, speed deviation), inclusion of visual information did improve on the statistical models predicting these outcomes, and clinicians should remain cognizant that visual impairments, perhaps even mild ones, may affect driving abilities.

Study Limitations

A limitation to this study is, of course, the small sample size, and thus the interpretation of these findings can only be considered preliminary. Another limitation is the brevity of the simulator evaluations. The simulations were intentionally brief in order to be practical for use in clinical trials, but there is likely a cost with respect to sensitivity and specificity of the findings. In addition, the MS group was comprised of individuals recruited for a clinical trial, and thus had less than satisfactory response to traditional treatments. This may limit the generalizability of the results.

CONCLUSIONS

It should be noted that, based on the present findings, and similar to other studies of MS drivers, one cannot say that the MS drivers are impaired drivers and should not be on the road. Although other studies have found simulator performance (using different scenarios) to be predictive of on-road driving,^36,37 consistent evidence supporting a strong relationship between various simulations and on-road performance is lacking. The reduced performance in the MS group likely places them at higher risk for a crash (previous studies in other cohorts found MS patients to have a higher accident rate than controls^1,2), but it is not known whether the increase is of such significance that it outweighs the individual and societal advantages of the independence in these patients. Although it has its shortcomings, an on-road evaluation remains the gold standard for determining driving competence. Nonetheless, the present study does provide new data regarding the impact that various MS conditions may have on a person’s ability to safely drive an automobile. Driving rehabilitation specialists may want to emphasize strategies that assist MS patients with compensating for these specific weaknesses (eg, potential effects of cognitive and physical impairments, proper foot positioning), while clinicians may want to consider discussing the potential driving conditions (eg, close car following, multitasking) that may put particular patients at risk for automobile crashes. Of course, it is recommended that patients, caregivers, and health care professionals be included in this discourse.

Future studies should examine larger cohorts that present with a wider range of cognitive and physical impairments. More comprehensive simulations, assessing routine driving, accident avoidance, and navigational skills would also better inform the scientific and clinical community regarding the breadth of driving skills that are, and are not, affected by MS. And, last, relating performance on these different measures to on-road assessments would provide important data regarding the degree to which assessments in the laboratory translate to real world performance.

List of Abbreviations

EDSS

Expanded Disability Status Scale

GDS

global deficit score

HVLT-R

Hopkins Verbal Learning Test–Revised

MAS

Modified Ashworth Scale

MS

Multiple sclerosis

MSQLI

Multiple Sclerosis Quality Of Life Index

SDLP

Standard deviation of lateral position

TMT

Trail-Making Test