Veering in Hemi-Parkinson’s Disease: Primacy of Visual over Motor Contributions

Abstract

Veering while walking is often reported in individuals with Parkinson’s disease (PD), with potential mechanisms being vision-based (asymmetrical perception of the visual environment) or motoric (asymmetry in stride length between relatively affected and non-affected body side). We examined these competing hypotheses by assessing veering in 13 normal control participants (NC) and 20 non-demented individuals with PD: 9 with left-side onset of motor symptoms (LPD) and 11 with right-side onset (RPD). Participants walked in a corridor under three conditions: eyes-open, egocentric reference point (ECRP; walk toward a subjectively perceived center of a target at the end of the corridor), and vision-occluded. The visual hypothesis predicted that LPD participants would veer rightward, in line with their rightward visual-field bias, whereas those with RPD would veer leftward. The motor hypothesis predicted the opposite pattern of results, with veering toward the side with shorter stride length. Results supported the visual hypothesis. Under visual guidance, RPD participants significantly differed from NC, veering leftward despite a shorter right- than left-stride length, whereas LPD veered rightward (not significantly different from NC), despite shorter left- than right-stride length. LPD participants showed significantly reduced rightward veering and stride asymmetry when they walked in the presence of a visual landmark (ECRP) than in the eyes-open condition without a target. There were no group differences in veering in the vision-occluded condition. The findings suggest that interventions to correct walking abnormalities such as veering in PD should incorporate vision-based strategies rather than solely addressing motor asymmetries, and should be tailored to the distinctive navigational profiles of LPD and RPD.

1. Introduction

Parkinson’s disease (PD) is a neurodegenerative disorder, the typical motor symptoms of which include resting tremor, bradykinesia, postural instability, freezing of gait, shuffling gait pattern, rigidity in the trunk and limbs, reduced pelvis rotation, and lack of arm swing, all of which put people with PD at a high risk of falling (Bloem, Boers et al. 2001; Schaafsma, Balash et al. 2003; Schaafsma, Giladi et al. 2003; Wood, Bilclough et al. 2002). Non-motor features of the disease have also been identified. In the visual domain, these include changes in basic visual functions such as contrast sensitivity, motion and optic flow perception, color discrimination and visuospatial perception (Archibald, Clarke et al. 2011; Bodis-Wollner, Marx et al. 1987; Bodis-Wollner 1990; Brandies and Yehuda 2008; Davidsdottir, Cronin-Golomb et al. 2005; Davidsdottir, Wagenaar et al. 2008; Harris, Calvert et al. 1990; Uc, Rizzo et al. 2005).

A current view is that the role of vision in spatial navigation includes not only perceiving the layout of the world, but also, importantly, controlling one’s movement. Absence of proper visual inputs has long been acknowledged as a critical risk factor for falls especially for people with visual impairment due to neurological disorders or normal aging (e.g., Lee and Scudds 2003; Hafström, Fransson et al. 2002; Perrin, Jeandel et al. 1997). This proposition has not typically been applied to PD, because the disease was traditionally characterized as a motor disorder rather exclusively, with the focus of rehabilitation research directed at interventions targeting the motor symptoms. Davidsdottir and colleagues reported that visual and visuospatial impairments were prevalent in a sample of 81 individuals with PD, with visual hallucinations, double vision and contrast sensitivity deficits being associated with freezing of gait (Davidsdottir, Cronin-Golomb et al. 2005). Although visual processing is impaired, there is increased dependence on vision in PD for postural control (Azulay, Mesure et al. 2002) and for gait regulation while walking (Morris, Iansek et al. 2005). Therefore, advancing our understanding of the non-motor symptoms of PD such as deficits in visuospatial processing, as well as their potential contribution to locomotive disability, is a pressing need in the field.

PD almost always has unilateral onset due to the underlying hemispheric neuropathology, and this laterality is reflected in the difficulties that people with PD commonly endorse in regard to navigating in space (Davidsdottir, Cronin-Golomb et al. 2005). During spatial navigation tasks, veering (lateral deviation from a straight or intended path) in PD has been measured quantitatively; persons with LPD veered rightward in the presence of visual input, whereas persons with RPD veered leftward (Davidsdottir, Wagenaar et al. 2008; Young, Wagenaar et al. 2010). This finding echoes the different profiles that LPD and RPD display on visual perception tasks, including horizontal line bisection (Davidsdottir, Wagenaar et al. 2008; Laudate, Neargarder et al. 2013; Lee, Harris et al. 2001), copying and drawing tasks (Shelton, Bowers et al. 1990; Vallar 1998), self-report of daily visual function (Davidsdottir, Cronin-Golomb et al. 2005), reaching and grasping tasks (Rossit, McIntosh et al. 2012), body-scaled aperture estimation (Lee, Harris et al. 2001), and size perception comparison in two hemi-spaces (Harris, Atkinson et al. 2003; Milner and Harvey 1995). Overall, individuals with LPD exhibit a rightward spatial bias, perceiving stimuli as shorter or smaller on the left than the right. By contrast individuals with RPD perceive visual stimuli more like healthy control adults, who have been reported to bisect lines slightly to the left (“pseudoneglect”) (Jewell and McCourt 2000). It appears that the consequences of right hemisphere damage (LPD) contribute to more severe visuospatial impairments than damage to the left hemisphere (RPD), as the right hemisphere mediates more visuospatial processing than the left in the general population and also in PD (Cronin-Golomb 2010).

Asymmetry of symptoms in PD also influences the dynamics of sensorimotor coordination (Boonstra, van der Kooij et al. 2008; Frazzitta, Pezzoli et al. 2013; Lin, Wagenaar et al. 2014; Nanhoe-Mahabier, Snijders et al. 2011; Plotnik, Giladi et al. 2005; Yogev, Plotnik et al. 2007). Individuals with PD typically have less stable and more asymmetric gait patterns during locomotion, with shorter stride length on the initially affected body side than on the secondarily affected body side (Lin, Wagenaar et al. 2014; Plotnik, Giladi et al. 2005; Young, Wagenaar et al. 2010). Although no conclusive association has been drawn between motor asymmetry and veering, the difference in stride length between body sides has been offered as an explanation (Guth and Laduke 1994). Previous veering studies indicated that those with LPD veered rightward, whereas those with RPD veered leftward during normal walking, corresponding to the hemisphere with presumed lower dopamine levels and greater neuropathology (Davidsdottir, Wagenaar et al. 2008; Young, Wagenaar et al. 2010).

Whether the source of veering in PD is more attributed to errors in visuospatial perception or to asymmetry of motor features has not been addressed directly. These two potential mechanisms provide contradictory predictions for veering direction. If veering is primarily driven by asymmetrical walking patterns expected in PD between the relatively affected and relatively non-affected body side, a tendency to veer towards the side of body that has relatively shorter step length would be observed regardless of whether they walked with eyes open or vision occluded, i.e., LPD would veer leftward, whereas RPD would veer rightward. On the other hand, if veering is driven by visuospatial bias (as seen in mild hemineglect), veering should be shifted in the opposite direction, with LPD veering rightward and RPD veering leftward, as reported in the studies of Davidsdottir and colleagues (2008), and Young and colleagues (2010)—but these studies did not example stride asymmetry. The visuospatial bias might be observed especially when participants were asked to walk towards the self-perceived center of a horizontal line placed at the end of the corridor. The resulting visuospatial shift of the egocentric midline in PD would come into play: LPD would generate rightward error on perceiving the center of the bar, resulting in a rightward veering trajectory, and a similar (but leftward) effect would be expected in RPD, although the size of the bias would be expected to be smaller because the influence of right hemisphere dysfunction on visuospatial perception is greater than that of the left. Our goal was to assess directly whether visuospatial bias or motor bias accounts better for lateral drift in individuals with LPD and RPD.

2. Method

2.1. Participants

The study included 20 non-demented individuals who had been diagnosed with idiopathic PD (11 men, 9 women) and 13 normal control adults (NC; 4 men, 9 women). The distribution of men and women did not differ between the PD and NC groups (χ² = 1.87, p = 0.17). The PD participants were recruited from the Parkinson’s Disease Clinic at the Boston Medical Center and from the Fox Foundation Trial Finder. The NC group was recruited from the Fox Trial Finder and the local community. Participants underwent health history screening prior to taking part in the study. Exclusion criteria included the inability to ambulate independently or history of musculoskeletal impairments or pain condition; lower extremity impairments that prevented free movement; use of walking assistive devices; coexistence of serious chronic medical illness; history of traumatic brain injury or stroke; psychiatric or neurological diagnoses (besides PD, in the PD group); surgery affecting the thalamus, basal ganglia, or other brain regions; history of alcoholism or other drug abuse; use of psychoactive medication except antidepressants or anxiolytics in the PD group; use of any psychoactive medication in the control group; presence of clinically significant eye disease, or corrected binocular acuity poorer than 20/40. Participants were screened for acuity binocularly at a distance of 10 ft using a Snellen chart; Snellen scores were converted to logMAR scores for the analysis. Mean acuity was −0.01 (20/16 Snellen) (SD = 0.07) for the PD group, and −0.09 (20/16 Snellen) (SD = 0.03) for the NC group. There was a significant group difference with NC showing better acuity (t[26.1] = 4.21, p = 0.001, η² = 0.29) but this is probably not of clinical significance, as both groups’ acuity was very good. Initial analysis showed no effect of acuity on veering, and accordingly it was not considered in further analyses.

All participants were right handed except three of the PD group and one of the NC group, all of whom were left handed. We conducted separate veering analyses with and without individuals who were left handed and found that the results were not affected; therefore handedness was not considered further in the analyses. All participants were native English speakers. All were non-demented as indexed by their scores on the modified Mini-Mental State Exam (mMMSE; Stern, Sano, Paulson & Mayeux, 1987), each obtaining 26.45 or better on conversion to standard MMSE scoring.

The PD group reflected mild to moderate stages of the disorder (stages 1–3 on the Hoehn and Yahr scale) (Hoehn and Yahr, 1967) (Table 1). The average disease duration was 4.7 years (SD = 4.0). Disease severity was determined with the use of the Unified Parkinson’s Disease Rating Scale (UPDRS, 4 sections; Fahn & Elton, 1987; Levy et al., 2005). The PD group had a mean UPDRS total of 35.5 (SD = 14.5) denoting mild-moderate disease severity, with a mean motor score of 21.2 (SD = 10.1). All participants were taking medication for their parkinsonian symptoms and at the time of testing were in their “on” period. Levodopa equivalent dosage (LED) mean was 457.7 (SD = 335.5) mg/day for LPD, 486.4 (SD = 318.4) mg/day for RPD. There was no significant difference in LED between these groups (t[18] = 0.20, p = 0.85).

Table 1.

Participant Characteristics

Measure	LPD	RPD	NC	p value
Sample size	9	11	13
Men:Women	6:3	5:6	4:9	0.25, ns
Age (years)	67.3 (7.6)	66.9 (5.8)	62.3 (5.5)	0.11, ns
Education (years)	17.0 (2.4)	17.9 (1.5)	17.8 (2.3)	0.60, ns
MMSE	28.2 (1.1)	28.9 (1.0)	29.0 (0.9)	0.21, ns
BDI-II	4.6 (2.4)	6.1 (4.7)	3.5 (4.7)	0.35, ns
BAI	4.8 (3.5)	6.3 (6.0)	1.2 (1.5)	0.01
H & Y	2 (1.5–3)	2 (1–3)	NA	0.52, ns
UPDRS total score	36.7 (12.5)	34.5 (16.5)	NA	0.75, ns
UPDRS motor score	22.3 (6.9)	20.2 (12.4)	NA	0.65, ns
UPDRS motor asymmetry score	−0.3 (0.3)	0.3 (0.4)	NA	0.004
LED	457.7 (335.5)	486.4 (318.4)	NA	0.85, ns

Note. Univariate Analysis of Variance tests were conducted comparing LPD (left-onset Parkinson’s disease), RPD (right-onset Parkinson’s disease) and NC groups (normal control).

BDI-II = Beck Depression Inventory – II; BAI = Beck Anxiety Inventory; H & Y = Hoehn & Yahr stage; UPDRS = Unified Parkinson’s Disease Rating Scale; LED = levodopa equivalent dosage. Values presented are means (standard deviations) except for Hoehn and Yahr which is median and range.

The PD group was further characterized by side of motor symptom onset: nine with LPD (6 men and 3 women) and 11 with RPD (5 men and 6 women) (Table 1). The mean disease duration for LPD was 3.7 years (SD 2.9) and for RPD was 5.6 years (SD 4.7), with no difference between groups, t(17.1) = 1.10, p = 0.29. The distribution of men and women did not differ between the two groups (χ² = 0.9, p = 0.34). The LPD group included one in stage 1.5, five in stage 2, two in stage 2.5 and one in stage 3 (median 2, range 1.5 – 3). The RPD group included one in stage 1, six in stage 2, one in stage 2.5 and three in stage 3 (median 2, range 1 – 3). The distribution across stages did not differ between the two groups (χ² = 3.26, p = 0.52). The initial side of onset was identified using self-report and through review of neurology records. The current side and extent of motor severity were assessed using the UPDRS motor score (entire motor section of UPDRS: items 18–44). The mean motor score was 22.3 (SD = 6.9) for LPD and 20.2 (SD = 12.4) for RPD, with no significant difference between groups (t[18] = 0.46, p = 0.65).

We also examined motor function for the two body sides separately, using the relevant subset of UPDRS motor items (20–39; tremor, rigidity, finger taps, hand movements, leg agility). The score on the left body side was 9.1 (SD = 3.1) for LPD and 6.2 (SD = 5.6) for RPD; on the right body side it was 6.4 (SD = 3.1) for LPD and 7.7 (SD = 4.3) for RPD. There was no group difference for either the left side (t[18] = 1.40, p = 0.18) or the right side (t[18] = 0.75, p = 0.46). Comparing left-side and right-side motor scores within-group, there was a significant difference for LPD (left side score higher, meaning more severe; t[8] = 2.87, p = 0.021). The right-side score was higher than the left-side score for RPD but the difference was not significant (t[10] = 1.78, p = 0.11). The extent of motor asymmetry was expressed as (Right score – Left score) / (Right score + Left score). The asymmetry index may range from −1 to 1, where scores closer to 1 indicate more extensive and severe symptoms on the right side of the body and scores closer to −1 indicate more extensive and severe symptoms on the left side of the body. The group means for LPD and RPD were −0.3 (SD = 0.3) and 0.3 (SD = 0.4), respectively. The difference was significant (t[18] = 3.25, p = 0.004, η² = 0.37) in regard to direction of asymmetry, though the extent of the asymmetry was the same for the two groups (absolute value of 0.3). Although most individuals did not display strong and obvious motor asymmetry at the time of the study, there is evidence that the hemispheric asymmetry of brain lesions in PD remains well after motor symptoms have progressed from unilateral to bilateral (Rinne, Laihinen et al. 1993). Hence, we would expect the impact of the hemispheric asymmetry on veering to be maintained in our sample.

We compared the LPD, RPD, and NC groups on demographic and other characteristics potentially pertinent to the study. As shown in Table 1, the groups did not differ on the following variables: age, F(2,30) = 2.34, p = 0.11, education, F(2,30) = 0.53, p = 0.60, numbers of men and women, χ² = 2.76, p = 0.25, or MMSE F(2,30) = 1.65, p = 0.21. Mood was assessed for all participants using the Beck Depression Inventory II (BDI-II) and Beck Anxiety Inventory (BAI) (Beck & Steer, 1993; Beck, Steer, & Brown, 1996). There were no group differences on the BDI-II (F[2, 30] = 1.10, p = 0.35). There was a significant effect of group on the BAI (F[2, 30] = 5.24, p = 0.01, η² = 0.26). Specifically, the RPD group had a significantly higher mean BAI than did the NC group (p = 0.01). There was no significant difference on BAI between the RPD and LPD groups (p = 0.69) or between the LPD and NC groups (p = 0.11). We used BAI as a covariate in the veering analyses and found that it did not affect any of the results; therefore it was not considered further in the analyses.

2.2. Apparatus

The over-ground walking assessment was implemented in a corridor (3.7m wide, 2.6m high, 10.4m long) constructed in the laboratory using black curtains on both sides. The room was well lit and the sounds from surroundings were strictly controlled. Participants were allowed to take a break between walks along the corridor as needed. An experimenter was immediately behind the participant at all times to ensure safety.

Three-dimensional kinematics

Three-dimensional kinematic data were collected using an Optotrak 3020 System (Northern Digital Inc., Waterloo, ON, Canada), with a spatial resolution of 0.1 mm. Three position sensors were placed at the end of the walkway in left, right and middle positions facing the participant’s direction of walking. The placement allows for an environmental reference plane to capture bilateral locomotor movements for at least eight strides. The sensors were calibrated and the mean error was accepted when the value was 0.7 mm or less. Infrared light-emitting diodes (IREDs) were applied as position markers on the participant’s chin (lower mandible) and bilaterally on the ankle (lateral calcaneus), knee (patella), hip (anterior superior iliac spine), wrist (radiocarpal joint), shoulder (humeral head), cheek (2 cm below zygomatic arch). The instantaneous position of each IRED was sampled during walking trials at a rate of 100 Hz and stored to disk for further analysis.

2.3. Procedure

The study was carried out in accordance with the Code of Ethics of the World Medical Association (Declaration of Helsinki). Informed consent was obtained.

Participants started each trial by aligning each foot with a marker placed at the center of, and perpendicular to the edge of the corridor's start line. A practice session was provided to each of the participants to enable them to become acclimated to the walking environment and choose their preferred walking speed. After the practice session, the actual data collection began. No feedback on walking speed was provided. Participants were first asked to walk a straight line down the middle of the walkway with eyes open, at a comfortable speed, until they reached the end (eyes-open condition) (3 trials). Then they were instructed to repeat the task wearing a blindfold, which was similar to a sleeping mask with an elastic band (3 trials). The experimenter told them to stop either when they reached the end or were about to bump into the side of the walkway. The lights in the room were turned off during this condition, then put back on for the rest of the session. In the third condition, which assessed veering behavior in relation to visuospatial bias, a 0.05m wide and 1.53m long rectangle (“line”) colored in bright yellow was placed horizontally at eye level on the wall at the end of the walkway. The examiner stood next to the line and moved a crossbar along it slowly from one end to the other (randomized direction across participants). The participants were instructed to stand straight facing forward and stated when the crossbar fell in the perceived center of the horizontal line. This test is similar to the Landmark test of line bisection which has been used to detect lateral biases in allocentric spatial perception in PD (Davidsdottir, Wagenaar et al. 2008; Laudate, Neargarder et al. 2013; Lee, Harris et al. 2001), but at a longer viewing distance. The crossbar was then removed, though the line remained. The participants were then instructed to walk towards their self-perceived center of the horizontal line (egocentric reference point condition, ECRP) (3 trials). Lateral deviation of the judgment of the center of the horizontal line was recorded and used for data analysis. The line was not used in the other conditions. We expected that if there were a visuospatial bias, PD participants would perceive the center of the line as off true center, compared to NC, and consequentially would engage in veering in the direction dictated by the bias.

2.4. Data Reduction

The kinematic data were filtered using a zero-lag, fourth order Butterworth low-pass filter with a cut-off frequency of 5 Hz. Angular positions of the arms and legs in the sagittal plane were defined by the orientations of vectors from shoulder to wrist markers and from hip to ankle markers, respectively, measured relative to laboratory vertical (i.e., to the gravity axis). Positive angle values indicate forward wrist or ankle positions. Stride cycles for each leg were identified by two consecutive maxima from the angular position data of the corresponding leg. All the gait variables were computed using MatLab (MathWorks, Inc., Natick, MA) employing only the middle strides (excluding the first and last strides) in order to avoid acceleration and deceleration variations at the beginning and at the end of the distance walked.

2.5. Dependent Variables

2.5.1. Veering

The midpoint between left and right hip position data was calculated, and veering was defined for each trial as the difference in medio-lateral position of this midpoint between the beginning of the first and last of the middle strides during walking. Positive drift values indicate rightward veering and negative values indicate leftward veering. Since, as noted in the literature, veering could be accounted for by undetected body orientation errors at the starting point (Guth and Laduke 1994; Kallie, Schrater et al. 2007), we calculated the hip angle relative to the starting line using left and right hip positional data in the anteroposterior and mediolateral direction, tan⁻¹ ((R_AP − L_AP)|(R_ML − L_ML)), then tested whether there was any misalignment before initiating walking and its relation with veering. Analysis of variance (ANOVA) showed that there were no significant differences in hip angle by group or condition (all p>.16), meaning that initial body orientation would not account for any group differences in veering.

2.5.2. Stride Parameters

Participants walked a total of 10 meters. Data from only the middle strides were analyzed for each leg, as the first stride reflected reaching a comfortable walking pattern, and the last stride slowing down and stopping at the end of the corridor. In this study, the number of consecutive strides of the left and right legs that were covered ranged from four to six. The following stride parameters were computed for each of the middle strides and then averaged across strides: walking speed, stride length, and stride asymmetry (calculated as the difference in the average stride lengths between the left and right legs), all of which may impact veering behavior (Guth and Laduke 1994). For example, higher walking speeds have been associated with smaller amounts of veering (Cicinelli 1989; Klatzky, Loomis et al. 1990). Additionally, if veering in PD is driven only by the motoric factor of stride length asymmetry (one of our hypotheses to be tested), the direction of veering should be toward the body side with the shorter stride length.

Average walking speed (m/s) was determined by dividing the linear displacement of the chin marker (between the times of left heel strike for the first and last strides) by the time elapsed between these heel strike events. The chin marker displacement was calculated according to:

$${{\mathrm{D}}_{\mathrm{T}}}^2={{\mathrm{D}}_{\text{AP}}}^2+{{\mathrm{D}}_{\text{ML}}}^2$$

where D_T represents the total linear displacement (Euclidean distance) of the chin marker, and D_AP and D_ML are its displacements in the anteroposterior and mediolateral directions, respectively. Stride lengths (in meters) of the left and right legs were calculated for each trial by the anteroposterior displacements over the middle strides by the left and right ankle markers, respectively, divided by the number of the middle strides. Considering that variation in leg lengths among participants might have an impact on the results, we normalized stride lengths of each leg by dividing them by the individual's leg length, measured as the distance between hip and ankle markers on the side of the respective leg. These normalized stride lengths were used in the data analysis.

2.6. Data Analysis

Statistical analyses were performed using SPSS 18.0 (SPSS, Inc., Chicago, IL).

Separate mixed design ANOVAs were performed to examine the effects of group (LPD, RPD and NC) and condition (eyes-open, vision-occluded and egocentric reference point [ECRP]) on veering and the stride parameters of normalized stride length and stride asymmetry. The analyses for all of the parameters were based on the average of three trials per condition. We used age as a covariate in the stride analyses because even though the three groups did not significantly differ in age, previous literature suggests that there are age-related changes in gait for stride length, at least (Himann, Cunningham et al. 1988; Prince, Corriveau et al. 1997).

A series of a priori between groups t-tests (or ANOVAs if a covariate was included), were performed to examine the differences between LPD and RPD, LPD and NC, and RPD and NC under each vision condition. A priori within groups t-tests were used to examine differences on the eyes-open and vision-occluded conditions, eyes-open and ECRP conditions, and vision-occluded and ECRP conditions within each group. In addition, we used Spearman correlations to examine the relation between veering (direction and extent) and stride asymmetry during walking in each condition for each group. We predicted that those individuals in each group with higher stride asymmetry scores would demonstrate more veering. We used one-tailed tests to examine these directional hypotheses.

3. Results

3.1. Walking speed

An analysis of variance (ANOVA) was conducted to examine whether the three groups (LPD, RPD, and NC) differed in walking speed across conditions, as this could impact veering behavior. Results revealed significant group differences in walking speed (F[2, 30] = 6.23, p = 0.005, η² = 0.29), with the LPD and RPD groups each walking significantly more slowly than the NC group based on Tukey’s post hoc test (LPD vs. NC: p = 0.021; RPD vs. NC: p = 0.011). The LPD group had a mean walking speed of 1.1 m/s (SD = 0.1) in the eyes-open condition, 0.7 m/s (SD = 0.2) in the vision-occluded condition, and 1.2 m/s (SD = 0.2) in the ECRP condition. The RPD group had a mean walking speed of 1.2 m/s (SD = 0.1) in the eyes-open condition, 0.8 m/s (SD = 0.2) in the vision-occluded condition, and 1.1 m/s (SD = 0.1) in the ECRP condition. The NC group had a mean walking speed of 1.3 m/s (SD = 0.2) in the eyes-open condition, 1.0 m/s (SD = 0.2) in the vision-occluded condition, and 1.3 m/s (SD = 0.2) in the ECRP condition. Walking speed did not significantly affect the results on veering, however, so it was not considered further.

3.2. Veering

The mean veering score for LPD, RPD, and NC, respectively, were as follows for each condition, with negative numbers indicating leftward veering and positive numbers indicating rightward veering: eyes-open 53.4 mm [SD = 154.0] LPD, −52.0 mm [SD = 135.8] RPD, 114.8 mm [SD = 84.4] NC; vision-occluded −183.1 mm [SD = 571.2] LPD, 63.7 mm [SD = 494.4] RPD, 36.9 mm [SD = 336.1] NC; and ECRP 0.8 mm [SD = 148.3] LPD, −25.4 mm [SD = 96.4] RPD, 70.8 mm [SD = 81.2] NC. A mixed design ANOVA on veering showed no significant effects of group (F[2,30] = 1.16, p = 0.33), condition (F[1.1,33.3] = 0.53, p = 0.49), or interaction between group and condition (F[2.2,33.3] = 1.26, p = 0.30). Age was not a significant covariate for veering F(1,29)=1.31, p=0.26. We conducted one-sample t-tests (2-tailed) to examine whether LPD or RPD exhibited significant deviation from the true center. Neither of the PD groups veered significantly from center for any of the conditions (all p’s > 0.23).

3.2.1. Between-groups comparisons

A series of planned contrasts between groups t-tests revealed a significant group difference between RPD (leftward veering) and NC (rightward veering) in the eyes-open condition (t[22] = 3.67, p = 0.001, η² = 0.38) and in the ECRP condition (RPD: leftward veering, NC: rightward veering) (t[22] = 2.66, p = 0.014, η² = 0.24) (Fig. 1). These two groups did not differ in veering in the vision-occluded condition (t[22] = 0.16, p = 0.88). There was no significant difference between the LPD group and either the RPD or NC group in any of the conditions (all p’s > 0.12).

Fig. 1.

Veering (in mm) during walking under three conditions: eyes-open, vision-occluded and ECRP. 9 LPD, 11 RPD and 13 NC. Negative values represent veering toward the left whereas positive values represent veering toward the right. Horizontal lines represent standard error of the mean.

On the perceptual test of line bisection that was given at the beginning of the ECRP condition, all groups bisected the line rightward of the true center (LPD: 31.6 mm [31.7]; RPD: 34.2 mm [50.6]; NC: 16.2 mm [25.5]. There was no significant correlation between deviation from true center on the perceptual task and veering (deviation from true center on the walking task) of the same condition or other the other visual (eyes-open) condition; nor was there a significant correlation between deviation from true center on the perceptual task and UPDRS motor asymmetry (all p’s > 0.10).

3.2.2. Within-groups comparisons

A series of a priori within group t-tests revealed that for the LPD group, there was significantly less veering in the ECRP condition (0.8 mm [SD = 148.3]) than in the eyes-open condition (53.4 mm [SD = 153.9]), with rightward veering in both conditions, t(8) = 2.32, p = 0.049, η² = 0.40 (Fig. 1). With vision occluded, mean veering was leftward (−183.0) but the difference between veering under this condition and under either the eyes-open condition or the ECRP condition was not significant (all p’s > 0.26). The NC group on average showed rightward veering in all walking conditions. Like the LPD group, the NC group showed significantly less veering in the ECRP condition (70.8 mm [SD = 81.2]) than in the eyes-open condition (114.8 mm [SD = 84.4]) t(12) = 2.19, p = 0.049, η² = 0.29; no other conditions significantly differed (mean for vision-occluded condition 70.8 mm; all p’s > 0.37). For the RPD group, on average veering was leftward in both the eyes-open (−52.0 mm) and ECRP conditions (−25.4 mm) and rightward in the vision-occluded condition (63.7 mm), but the differences between conditions were not significant (all p’s > 0.24).

We examined potential gender effects because of reported differences in performance of men and women on navigational tasks (Davidsdottir et al., 2008). We found a trend for an effect of gender only for the RPD group in the ECRP condition. Men with RPD showed leftward veering (mean −80.9, SD 29.6 mm) whereas women with RPD veered rightward (mean 20.9, SD 110.6 mm) (t[9] = 1.98, p = 0.08, η² = 0.30). There were no differences between men and women in the RPD group for the other two conditions (eyes open and eyes closed), nor for the LPD or NC groups in any condition (all p’s > 0.12).

3.3. Normalized stride length

A preliminary ANOVA with the three groups was conducted and showed no differences between the normalized stride length computed based on the left leg time series and that computed based on the right leg time series (F[1,30] = 0.06, p = 0.80). An ANCOVA with age included as a covariate revealed only a trend toward an effect of group (F[2,29] = 3.03, p = 0.06), a significant main effect of condition (F[1.3,40.0] = 109.33, p = 0.001, η² = 0.79), and no interaction between group and condition (F[2.7,40.0] = 1.76, p = 0.18). Age was a significant covariate (F[1,30]= 8.10, p = 0.008, η² = 0.21) in the analysis. We also compared stride length for the more affected body side for the PD groups, i.e. left leg for LPD, right leg for RPD. There was no effect of group, F(1,17) = 0.09, p = 0.78, a significant effect of condition, F(1.41,25.39) = 73.2, p = 0.001, η² = 0.80, and no interaction, F(1.41, 25.39) = 0.59, p = 0.51. Age was a significant covariate F(1,17) = 9.92, p = 0.006, η² = 0.37. Because we found no significant effect of group whether analyzing left vs. right leg or more-affected vs. less-affected leg, the normalized stride length based on the left leg time series was used in further analyses, as per convention (e.g., Lin et al., 2014; Morris et al., 2005; Young et al., 2010).

3.3.1. Between-groups comparison

A series of a priori t-tests and/or ANCOVAs (when age was a significant covariate) demonstrated that compared to the NC group, the LPD group exhibited significantly shorter stride length in the eyes-open condition (t[20] = 2.70, p = 0.014, η² = 0.27) and in the vision-occluded condition (F[1,19] = 8.42, p = 0.009, η² = 0.31). In the latter, age was a significant covariate (F[1,19] = 5.62, p = 0.029, η² = 0.23) (Fig. 2). There was trend for a difference between these two groups for the ECRP condition (t[20] = 1.78, p = 0.09, η² = 0.14). RPD had significantly shorter stride length than NC in the vision-occluded condition (t[13.5] = 2.96, p = 0.011, η² = 0.31); the groups did not differ in the eyes-open condition (F[1,21] = 1.00, p = 0.33) or the ECRP condition (F[1,21] = 2.21, p = 0.15). Age was a significant covariate in the two latter comparisons (F[1,21] = 5.46, p = 0.029, η² = 0.21 and F[1,21] = 4.38, p = 0.049, η² = 0.17, respectively). There were no significant differences between the LPD and RPD groups in any of the conditions (all p’s > 0.38), with age as a significant covariate (all p’s < 0.036).

Fig. 2.

Normalized stride length (SL) on the left leg under three walking conditions: eyes-open, vision-occluded and ECRP. 9 LPD, 11 RPD and 13 NC. Vertical lines represent standard error of the mean. The unit is in m/m as the values are the stride length after normalization by individual’s leg length.

3.3.2. Within-groups comparison

A series of a priori within group t-tests showed that stride length was significantly shorter for all groups when walking with vision occluded than either with eyes open or in the ECRP condition (Fig. 2). For LPD, eyes-open vs. vision-occluded: t(8) = 6.55, p = 0.001, η² = 0.84; vision-occluded vs. ECRP: t(8) = 5.71, p = 0.001, η² = 0.80. For RPD, eyes-open vs. vision-occluded: t(10) = 7.10, p = 0.001, η² = 0.83; vision-occluded vs. ECRP: t(10) = 6.79, p = 0.001, η² = 0.82. For NC, eyes-open vs. vision-occluded: t(12) = 6.12, p = 0.001, η² = 0.76; vision-occluded vs. ECRP: t(12) = 5.39, p = 0.001, η² = 0.71). There was a trend for stride length to be longer in the ECRP condition than in eyes-open condition for LPD (t[8] = 1.95, p = 0.087, η² = 0.32), and shorter for RPD for ECRP than for eyes-open (t[10] = 2.08, p = 0.064, η² = 0.30), with no difference for NC (t[12] = 0.49, p = 0.64).

3.4. Stride Asymmetry

In regard to stride asymmetry, calculated as the difference in stride lengths between the left and right legs, an ANOVA revealed no significant main effects of group, F(2,30) = 0.20, p = 0.82, or condition, F(2,60) = 0.61, p = 0.55, or interaction between group and condition, F(4,60) = 0.58, p = 0.68. Age was not a significant covariate for stride asymmetry F(1,29)=1.96, p=0.17.

3.4.1. Between-groups comparison

A series of a priori between group t-tests showed that there was a significant group difference, as expected, between LPD and RPD in the eyes-open condition, t(18) = 3.37, p = 0.003, η² = 0.39, with LPD showing shorter stride length on the left body side and RPD showing shorter stride length on the right body side (see Fig. 3). The two groups did not differ in the other two conditions (all p’s > 0.14). Neither the LPD nor RPD group’s stride asymmetry was significantly different from that of NC in any of the conditions (all p’s > 0.45).

Fig. 3.

Difference in stride length between left and right body side (SL_diff) under three walking conditions: eyes-open, vision-occluded and ECRP. 9 LPD, 11 RPD and 13 NC. Negative values represent shorter stride length on the left body side whereas positive values represent shorter stride length on the right body side. Horizontal lines represent standard error of the mean. The unit is m/m as the values are the stride length after normalization by individual’s leg length. The large variance in the NC group was driven by two individuals, one with shorter stride length for the right leg and the other with shorter stride length for the left leg.

3.4.2. Within-groups comparison

Stride asymmetry in the LPD group was significantly less in the ECRP condition, in which participants walked towards a self-perceived center of the horizontal line in front of them, than in the baseline eyes-open condition (t[8] = 2.34, p = 0.048, η² = 0.41). There was also a trend for stride asymmetry to be less in the vision-occluded condition than in the eyes-open condition (t[8] = 1.94, p = 0.089, η² = 0.32) (see Fig. 3), and there was no difference between the vision-occluded and ECRP conditions (t[8] = 1.34, p = 0.22). For neither RPD nor NC were any significant differences observed between any two of the conditions (all p’s > 0.17).

3.5. Correlations

We examined correlations between veering (direction and extent) and stride length asymmetry. A negative veering value indicates leftward drift (positive indicating rightward) and a negative value of stride asymmetry indicates shorter strides with the left leg (positive indicating shorter right-leg strides).

A significant correlation between less leftward veering and less stride length asymmetry (shorter strides with left leg) was found for the LPD group in the ECRP condition, ρ = 0.61, p =0.04, and a trend in the same direction was found for the vision-occluded condition, ρ = −0.54, p = 0.07. There was no correlation for the eyes-open condition, ρ = 0.47, p = 0.10. For the RPD group, the correlation between veering and stride length asymmetry was not significant for any of the conditions (all ρ’s < 0.07, p’s > 0.42). For the NC group, there was a trend for a correlation between veering and stride asymmetry for the eyes-open condition (ρ = 0.40, p = 0.09). There were no significant correlations for the NC group for either the vision-occluded condition or the ECRP condition (all ρ’s < 0.12, p’s > 0.35).

4. Discussion

The results of the present study support the hypothesis that visual dysfunction, rather than motor dysfunction, is the predominant driver of veering in PD relative to healthy adults. This study also provides quantitative evidence for the existence of distinct patterns of veering and stride asymmetry that are specific to side of motor symptom onset in PD, under conditions with visual input.

LPD have been reported to have a tendency toward mild left spatial hemineglect that produces a rightward shift of egocentric midline, whereas RPD tend to have (if any bias) a slight right spatial hemineglect that produces a leftward shift of egocentric midline (Davidsdottir, Wagenaar et al. 2008; Laudate, Neargarder et al. 2013; Lee, Harris et al. 2001). Based on the findings of previous visuospatial studies (Davidsdottir et al., 2008; Young et al., 2010), one would expect people with PD to veer in the direction of the lateral shift of the egocentric midline. By contrast, from a biomechanics point of view, PD veering should be influenced by motoric asymmetry between the relatively more affected body side and the relatively less affected body side. This would predict results opposite to those based on the visuospatial hypothesis: individuals with PD should veer towards the side with a shorter stride length.

Our findings mainly support the former prediction under conditions of visual guidance. When participants were instructed to walk straight ahead in the eyes-open condition, LPD veered rightward and RPD veered leftward relative to NC, consistent with our earlier studies on veering (Davidsdottir, Wagenaar et al. 2008; Young, Wagenaar et al. 2010), despite shorter stride length on the more affected body side (i.e., on the left side for LPD and on the right side for RPD). In the ECRP condition, when participants were asked to walk toward a subjectively perceived center of a target at the end of the corridor (reflecting their egocentric reference point), they veered in the same direction as seen in baseline eyes-open condition. When the task was performed with vision occluded, group differences were not significant, though it should be noted that the direction was in fact opposite that seen under visual guidance; that is, the direction predicted by biomechanics alone: LPD veered to the left, corresponding to the body side with shorter stride length, and RPD veered to the right, likewise corresponding to the body side with shorter stride length. We conclude that under conditions of visual guidance that mirror everyday life, the mechanism underlying veering is predominantly vision-based instead of motoric.

Comparing our results to those of other studies reveals some inconsistencies in regard to the vision-occluded condition. As noted above, we found that RPD and NC veered to the body side that had shorter stride length, which was the right side; for the same reason, LPD veered to the left. In the study by Young and colleagues (Young, Wagenaar et al. 2010), the expected stride asymmetry was found between the initially-affected side and secondarily-affected side for PD, with LPD having a shorter stride length on the left and RPD on the right; NC had shorter stride length on the left than right. Despite the different directions of stride asymmetry between groups, all participants showed leftward veering in the vision-occluded condition. A possible explanation for the difference across studies was in regard to body orientation upon onset of walking. As the initial orientation of the body could be responsible for the trajectory of veering (Guth and Laduke 1994; Kallie, Schrater et al. 2007), it is important to guarantee that the alignment of the body axis relative to the true midline of the walkway is consistent across groups throughout the experiment. In the present study we tested body alignment using the angle between left and right hip markers before walking was initiated and showed that there were no significant differences in hip angle across all groups and conditions. Hence, we were able to rule out the possibility that initial body orientation could account for group differences in the direction of veering. This information was not provided in the previous studies, leaving open this possibility in accounting for the different results reported.

Our findings underscore the dominant role of vision in controlling the direction of veering, but we also found evidence for the influence of certain motor characteristics of PD. There was a significant correlation of veering with stride asymmetry for LPD, under the ECRP condition. Although LPD veering and stride asymmetry were minimal in the ECRP condition, the correlation between the two variables was significant in a positive direction, meaning that the less the asymmetry in stride length (caused by shorter strides on the left than the right body side), the less leftward veering. We expect that the asymmetry of motor symptoms may have some impact on veering in PD, though it may not be powerful enough to overrule the effect of vision. Further study with larger samples and a wider range of disease severity will be required to examine this possibility.

It is noteworthy that age seemed to have no impact on the observed group differences in veering and stride asymmetry, in contrast to age being a significant covariate for group differences in stride length. As aging has been associated with reduced stride length (Prince, Corriveau et al. 1997), examining a range of age groups to further examine the effect of age on veering and stride asymmetry would be of interest as a future research direction. In regard to gender, we found a trend toward a difference between men and women only for the RPD group in the ECRP condition. Men with RPD showed leftward veering whereas women with RPD veered rightward. The result with men accords with that of Davidsdottir et al. (2008), who reported that men with RPD had a leftward bias on a line bisection task (which is similar to line bisection task but at a farther distance), whereas women with RPD had almost no bias. The gender difference found on ECRP was not found for the eyes-open condition, on which participants were asked to walk in a straight line though without being able to refer to the horizontal line at the end of the walking corridor that was used in the ECRP condition, suggesting that the self-perceived center may play a role in distinguishing men from women in veering, and specifically for those with RPD, though we do not wish to over-interpret this finding as it was only a trend.

It is well accepted that visual cues are critical for gait improvement for people with PD (Azulay, Mesure et al. 1999; Lebold and Almeida 2011; Lewis, Byblow et al. 2000; Morris, Iansek et al. 1994; Spaulding, Barber et al. 2013; Vitorio, Lirani-Silva et al. 2014). These studies used traditional cueing methods such as stripes placed on the ground. Recently, Vitorio and colleagues reported that persons with PD could regulate stride length regardless of whether or not they were looking at their lower limbs while walking—that is, exproprioceptive information (from the lower limbs) is not crucial for gait improvements generated by visual cues (Vitorio, Lirani-Silva et al. 2014). The ECRP condition in the present study is similar in that participants gazed at the self-perceived center of the horizontal line at eye level, and accordingly did not focus on their lower limbs' movement during walking. For LPD, a trend toward longer stride length was observed in the ECRP condition than in the baseline eyes-open condition, which was consistent with the findings of longer stride lengths achieved with visual cues reported by Vitorio et al. We also found less veering and decreased stride asymmetry under the ECRP condition compared to the eyes-open condition. LPD benefited more than RPD from cuing; the effects of the visual cue were significant for LPD but not for RPD. We have reported that individuals with LPD appear to be more visually dependent than those with RPD (Davidsdottir et al., 2008), which may explain the greater ability of LPD to benefit from conditions that provide visual cuing. These findings point to a potential role of explicit visual landmarks to guide locomotion in PD. It would be interesting for future research to examine other gait characteristics, such as stride-to-stride adjustments/corrections in body orientation angles that might be associated with veering during walking with a target present in front.

Our results suggest that individuals with LPD, with presumed predominant right hemisphere pathology, demonstrated patterns of gait disturbances that were visually influenced, as shown by differences in the extent of veering between the eyes-open and ECRP conditions. Within this subgroup, common parkinsonian gait disturbances such as veering, stride asymmetry, and to some extent stride length were amenable to amelioration by visual guidance, mainly focusing on self-perceived center (ECRP condition). The effect of directing attention to perceived center had significant effects on navigating the environment, raising the possibility of attentional or environmental strategies for intervention. Targeting visual attention and related aspects of cognition presents a potential but to date underexamined avenue of treatment (Doruk, Gray et al. 2014; Paris, Saleta et al. 2011; Sinforiani, Banchieri et al. 2004). These interventions hold promise particularly when combined with action observation-based (internal) strategies for intervention that have been shown to improve gait and walking in PD (Pelosin, Avanzino et al. 2010; Pelosin, Bove et al. 2013). Interventions to improve visual attention may prove to be a reasonable strategy to improve locomotion in PD, especially for individuals with left-side onset of symptoms. A further possibility suggested by the almost complete lack of veering in the ECRP condition is that distortion of visuospatial processing in hemiPD may be corrected by the use of objective environmentally-anchored landmarks, which may serve as targets to provide appropriate locomotion paths and guide locomotor trajectories. Alternatively, it may be possible to teach individuals with PD to visualize a landmark in the distance (such as a line bisected down the middle) and walk toward it. The potential value of the imaged line would need to be assessed through further research.

In conclusion, the existence of the distinct directions of veering for LPD and RPD observed in this study supports the primacy of the visual control of navigation over the role of motor function as measured by kinematic data. This finding suggests that information on veering may be of importance in the management of PD. In particular, interventions for gait disorders in PD should emphasize vision, visual attention, and environmental modification as means to rehabilitate veering problems, as this strategy may be more effective than focusing solely on motor symptoms as targets for treatment.

Highlights.

Direction of veering in Parkinson’s disease (PD) depends on the side of PD onset.

Veering in PD corresponds to the side of spatial bias, not motor bias.

Vision-based rather than or in addition to motor-based strategies may aid PD walking.