Acute Effects of a Lateral Postural Assist on Voluntary StepInitiation in Patients with Parkinson’s Disease
Marie-Laure Mille, PhD,
Marjorie Johnson Hilliard, MS, PT,
Katherine M. Martinez, MA, PT,
Tanya Simuni, MD,
and Mark W. Rogers, PhD, PT
 Department of Physical Therapy and Human Movement Sciences, Feinberg School of Medicine, Northwestern University,Chicago, Illinois, USA
 Department of Neurology, Feinberg School of Medicine, Northwestern University, Chicago, Illinois, USA
 Department of Physical Medicine and Rehabilitation, Feinberg School of Medicine, Northwestern University,Chicago, Illinois, USA
 Anticipatory postural adjustments (APAs) for lat-eral weight transfer and stability precede and accompany vol-untary stepping. Patients with Parkinson’s disease (PD) showdelays in step initiation with altered APA characteristics thatmay reflect impaired interactions between posture and locomo-tion. The purpose of this study was to examine the influence of a lateral postural assist on step initiation in patients with earlystage PD while off medication and healthy controls. Subjectsperformed self-paced rapid forward steps. In one condition(ASSIST), the APA was assisted at onset with a lateral pullapplied to the pelvis by a motor-driven robotic system. Groundreaction forces and whole body kinematics were recorded tocharacterize the APA and step characteristics. Overall, PDsubjects had a longer APA duration (
0.01) and longer firststep duration (
 0.027) than Control subjects. With theASSIST, the APA duration for both groups was shorter (
0.001), the step onset time was earlier (
 0.001), and thespeed of the first step became faster for PD subjects. Posturalassistance affecting the interaction between posture and loco-motion may have therapeutic potential for improving move-ment function in patients with PD. © 2006 Movement DisorderSociety
Key words:
 Parkinson’s disease; postural control; stepping;coordination; rehabilitation
Disorders of postural balance and gait are commonoccurrences for many people with Parkinson’s disease(PD) and are among the most debilitating symptoms thatcharacterize the progression of the disease.
Injuriousfalls are a frequent consequence associated with posturalinstability and disturbances of gait.
Although these axialsymptoms may initially improve with pharmacologicaltreatment due to general improvements in bradykinesiaand rigidity, difficulties with posture and gait often re-spond only modestly to medications.
Because of therelative resistance of axial symptoms to medication treat-ment, physical therapy interventions for improving pos-tural balance and gait play a central role in the clinicalcare of individuals with PD.
However, the evidence insupport of the effectiveness of such rehabilitation ap-proaches in PD is inconclusive.
Traditionally, therapeutic interventions for posture andgait impairments have focused on separate aspects of theproblems such as posture and balance training
or gaittraining.
However, the control of posture and loco-motion are interdependent at several levels of the centralnervous system (CNS).
Therefore, it is conceivablethat impaired posture and gait component interactionscould contribute to deficits in locomotion, not only due tobiomechanical limitations, but also because of adaptivechanges in neural control.Anecdotal information indicates that difficulties withinitiating locomotion movements in PD, such as risingfrom a chair and gait initiation, may be improved if thenormally automatic anticipatory postural adjustments
*Correspondence to: Dr. Mark W. Rogers, Department of PhysicalTherapy and Human Movement Sciences, Feinberg School of Medi-cine, Northwestern University, 645 North Michigan Avenue, Suite1100, Chicago, IL 60611. E-mail: m-rogers@northwestern.eduReceived 28 November 2005; Revised 9 May 2006; Accepted 30June 2006
Published online 6 November 2006 in Wiley InterScience ( DOI: 10.1002/mds.21139
 Movement Disorders
Vol. 22, No. 1, 2007, pp. 20–27© 2006 Movement Disorder Society
(APAs) that precede and accompany such movements(e.g., forward lean in chair rising and lateral weight shiftin stepping) are modestly assisted manually by the cli-nician.
This affect may involve the known influence of external sensory cueing on improving movement perfor-mance in PD by serving as a timing cue for enhancing thesequencing of motor subtasks that comprise a complexplan of action.
Alternatively, the postural assis-tance could enable the prerequisite postural state of change from bipedal to single limb support that may beanticipated by the CNS before the release of the stepcycle is initiated.
In this form of motor prediction, theneural circuits for initiating stepping could be activelydelayed until APAs have reached a state condition asso-ciated with lateral weight transfer and postural stabili-ty.
Difficulties with satisfying the postural prerequi-sites for stepping may therefore contribute to gaitinitiation problems in PD. Thus in addition to specificproblems with postural control (e.g., regulating posturalmuscle tone) or with gait (e.g., stride length regulation,shuffling steps, festinating gait, gait freezing), there mayalso be a disruption of the normal sequencing and inter-action of posture and locomotion components. If thatwere the case, then intervention approaches should targetways to improve the coordinative interactions betweenthe postural balance and gait components of locomotion.The transition from stationary standing to steppingduring gait initiation is frequently impaired in PD. TheAPAs for lateral weight transfer and stability duringforward stepping have been characterized by an abnor-mally prolonged duration and reduced amplitude, with adelay in the first step onset timing and execution.
External sensory cues have been shown to increase thenatural self-selected speed and distance of first step ini-tiation,
but not the maximum speed of gait initia-tion.
The effects of direct external assistance with pro-ducing APAs on step initiation are presently unknown.The purpose of this study was to determine whetherexternal robotic assistance with the performance of theAPA could alter the temporal sequencing and other char-acteristics of posture and gait components during rapidstep initiation in patients with PD and healthy controlsubjects. We hypothesized that, compared with baselineperformance, the application of a lateral postural assistwould (1) result in an earlier first step onset time, and (2)increase the maximum execution speed of the first step.
A total of 16 community dwelling adults participatedin this study; 8 patients with PD (7 men, 1 woman) and8 age- and gender-matched healthy Control subjects(CS). Individuals with PD were recruited from the Move-ment Disorders Clinic at Northwestern Memorial Hospi-tal. Patient inclusion criteria included being diagnosed bya movement disorders neurologist (T.S.) to have idio-pathic PD with a modified Hoehn & Yahr Stage of 1.5 to2.5.
One subject with Stage 3 was included in theanalysis. Control subjects were recruited from two sourc-es: (1) through the Buehler Center on Aging ResearchRegistry, and (2) from the university faculty/staff popu-lation. CS inclusion criteria included lack of a history of PD and within 3-year age match of 1 of the PD subjectsof the same gender. Exclusion criteria for both groupsincluded a history of significant cardiovascular, pulmo-nary, musculoskeletal; metabolic, or other neurologicaldisorders, and a score of 24 or below on the Mini-MentalState Examination.
All subjects signed an informed consent form ap-proved by the Northwestern University Institutional Re-view Board before participation. A physical therapist orresearch assistant from the Movement Disorders Clinicevaluated participants with PD on the same day as thelateral assist protocol using the Unified Parkinson’s Dis-ease Rating Scale (UPDRS) Motor subsection outcomemeasure. The subjects were assessed in the defined med-ications
 state (12 hours off PD medications). UPDRSserved as a clinical measure of the level of motor im-pairment and functional limitation. Overall, the cohort of PD patients had early stage PD: only 1 subject wastreated with levodopa, 3 with dopamine agonists, and 2were not receiving any PD medications. One subject hadmore advanced PD (Hoehn & Yahr scale Stage 3) andwas intolerant of dopaminergic therapy so was treatedwith amantadine. The group characteristics are summa-rized in Table 1. Due to the heterogeneous nature of themedication regimen of the test subjects, data analysiswas performed only for the medications
Experimental Protocol and Data Collection
Before trials were recorded, the subject practicedwalking with the maximum possible speed. Foot tracingswere taken on a sheet of dark paper by chalk dust appliedbeneath the shoes to serve as a guide for stepping duringthe trials. The tracings of the initial foot position weresecured over two force platforms (Advanced MechanicalTechnology Incorporated, Newton, MA), and the step-ping pattern tracing was extended in front of the subjectalong the walkway. During recorded trials subjectsplaced each foot on a separate force platform that re-corded the ground reaction forces at a sampling fre-quency of 500 Hz for 7 seconds. The kinematics of 17passive markers were recorded using a six-camera infra-
 Movement Disorders, Vol. 22, No. 1, 2007 
red motion analysis system (Motus Peak PerformanceSystem, Peak Performance Co., Englewood, CO), sam-pled at 60 Hz. The markers were placed symmetricallyon the skin and/or clothing, at the foot, ankle, knee, hip,wrist, elbow, shoulder, and on a headband covering thecrown of the head and both ear canals.
To determine the preferred limb to initiate stepping,each subject was asked to indicate with which foot theywould kick a ball placed in front of them. For each trial,the subjects were asked to stand naturally with theirweight evenly distributed on both feet. Subjects wereinstructed to take three steps forward as fast as theycould, beginning with the preferred foot at any self-selected time following a ready instruction from one of the investigators, who simultaneously commenced datacollection. They were told that they might feel a pull atany time on a belt placed around their pelvis, but to reactnaturally and continue to step as fast as possible. Step-ping instructions were repeated intermittently throughoutthe blocks of trials. Rest breaks were offered as needed.Subjects performed the task under four conditionseach consisting of six consecutive trials. The baseline(BASE) condition was always measured first. In theASSIST condition, pulling the subjects laterally at thepelvis level assisted the APAs. The ASSIST conditionwas always followed by a NONE condition (withoutassistance of the APA) to identify any short-term effectof the ASSIST on performance. Finally, in the TUGcondition, a slight translation (25% of the displacementused in the ASSIST) insufficient to mechanically helpthe performance of the APA was provided. There weretwo possible testing orders of the conditions: BASE–ASSIST–NONE–TUG or BASE–TUG–ASSIST–NONE.The presentation of these two orders was alternatelyrotated between and across the subjects and was imple-mented to lessen the possibility of an ordering effect of the ASSIST condition on the TUG condition.The lateral translations were provided by a motordriven robotic system with a cable attachment hookedonto the first single stance limb side of a belt around thesubject’s pelvis (Fig. 1). The details of the system havebeen more fully described in a previous report.
Thecharacteristics of the ASSIST (6 cm, 49 cm/sec, 900cm/sec
) and the TUG (1.5 cm, 49 cm/sec, 900 cm/sec
)were determined through pilot data collected before sub- ject testing. The lateral translation was triggered exter-nally during the early phase of the APA when the sub- jects produced a vertical force equivalent to 55% of theirweight on the initial right swing leg as detected by the
 Patient characteristics
Patient Age (yr) Height (cm) Weight (kg) UPDRS Motor
 Hoehn & Yahr stagePD 1 69 194.6 71.35 33 2.5PD 2 78 180.5 88.12 27 2PD 3 58 166.4 88.25 19.5 1.5PD 4 62 176.6 99.25 20 1.5PD 5 67 171.5 75.45 52 3.0PD 6 62 181.8 89.04 18 2PD 7 62 179.2 87.89 10 1.5PD 8 60 169.0 65.78 21 2PD Mean 64.75 177.4 83.14 25.1 2PD - SD 6.43 8.90 11.15 12.77 0.53CS Mean 64 174.72 81.55 N/A N/ACS - SD 6 11.34 9.0UPDRS Motor subscale total maximum score 108, 108 being the most severe. Hoehn & Yahr scale of 0-5:1.5
unilateral disease plus axial involvement, 2
bilateral disease without impairment of balance, 2.5
mildbilateral disease with recover on pull test, 3
 mild to moderate bilateral; postural instability; physicallyindependent.UPDRS, Unified Parkinson’s Disease Rating Scale; PD, Parkinson’s diseases; CS, Control subjects; SD,standard deviation.
FIG. 1.
 Schematic representation of experimental setup for applying alateral postural assist during step initiation.
 Movement Disorders, Vol. 22, No. 1, 2007 
vertical loading force from the force platform. A safetyharness was worn by all subjects to prevent injury but didnot otherwise restrict movement.
Data Analysis
Customized interactive graphical analysis programs(MatLab 6.0, MathWorks, Inc., Natick, MA) were usedto compute the temporal and spatial characteristics of theAPAs and initial stepping responses (Fig. 2). As de-scribed in previous reports
and shown in Figure 2,the lateral APA is characterized by an initial displace-ment of the net center of foot pressure (CoP) in themediolateral (M-L) direction toward the first step sidethat propels the body center of mass laterally toward thesingle support limb side. At the same time, a backwardshift in the CoP propels the body forward in the directionof stepping. An unloading of the step limb and displace-ment of the foot represents the beginning of the stepcycle.The characteristics of the APA (onset, duration, andamplitude) were determined from the M-L displacementof the net CoP recorded beneath both feet. The APAonset was identified when the CoP began its M-L dis-placement toward the first step foot (when the first de-rivative of ML CoP is not 0). The APA duration was theperiod between the onset of the APA and the time whenthe M-L CoP reached maximum displacement toward thesingle stance side before step liftoff. The APA amplitudecorresponded to the maximum swing side M-L CoPdisplacement.The first step characteristics (onset, duration, length,and clearance) were derived from the displacement of theright ankle marker. The step onset time was determinedrelative to the instant of the initial M-L CoP change frombaseline, that is, APA onset. The beginning and end of the step were defined from the vertical velocity of themarker to identify the step duration. Step length wasassessed as the displacement of the ankle between steponset and end in the forward direction. Step clearancewas calculated as the maximum vertical displacement of the ankle marker.
Statistical Analysis
Mean values were calculated for each subject and allvariables. A mixed-model repeated measures analysis of variance (ANOVA) was used to assess the groups (PDand CS) as between-factors and the conditions (BASE,ASSIST, NONE, and TUG) as within-factors. A New-man–Keuls post hoc analysis with the level of signifi-cance adjusted for multiple comparisons, was used whenthe results of the ANOVA were significant. For all com-parisons, a significance level was set at
RESULTSAPA Characteristics
Across all conditions, the patients with PD exhibitedlonger APA durations than Controls (
0.01; Fig. 3).As expected, the total duration of the APA was signifi-cantly shorter in the ASSIST condition compared withthe other conditions (
 0.001; Fig. 3). The meandecrease in the APA duration for the ASSIST conditioncompared with the other three conditions in the PDsubjects ranged from 93 to 116 msec and for the CS from63 to 108 msec. The total duration of the APA did notdiffer significantly between BASE, NONE, and TUG
FIG. 3.
 The group mean values plus 1 SD for anticipatory posturaladjustment (APA) duration for Control subjects (CS, white bars) andsubjects with PD
 medication (OFF, black bars). The four experi-mental conditions are initial baseline trials without postural assistance(BASE), trials with lateral postural assistance (ASSIST), follow-uptrials without postural assistance (NONE), and trials with a mechanicaltug that provided no direct postural assistance (TUG).
Significantdifferences between groups at
 0.05. ***Significant differencesbetween conditions at
FIG. 2.
 Representative example of one trial during rapid step initiationin the baseline condition. The top record (solid line) is the net center of pressure displacement in the mediolateral direction showing the antic-ipatory postural adjustment (APA) for lateral weight transfer. Thevertical arrow (
) indicates the onset of the APA, and the horizontalarrow (
) denotes the APA duration. Negative values indicatedisplacement in the direction of the initial swing limb side. The bottomrecord (broken line) is the vertical displacement of the first step anklemarker. The vertical broken line indicates the onset of the first step footmovement.
 Movement Disorders, Vol. 22, No. 1, 2007 
conditions (0.70
0.80). There was no significantgroup by condition interaction.
Sequencing of Posture and Gait
Although PD subjects generally showed a tendency toinitiate the first step relative to APA onset later thancontrols, there was no significant group effect for the steponset time (
 0.26; Fig. 4). However, a significanteffect of condition was observed (
0.001). The meanstep onset time occurred 77 to 97 msec earlier in the PDsubjects and 41 to 97 msec earlier in the CS during theASSIST condition (
 0.001) compared to the otherconditions.
Step Characteristics
There were no significant group effects for either steplength or step clearance (Table 2). Step duration waslonger (
0.027) across all of the conditions for patientswith PD compared with Controls (Table 2). A significantgroup by condition interaction (
0.015) indicated thatstep duration became shorter for PD subjects but not forControls during the ASSIST condition (Fig. 5). No con-dition effects were found for step length or clearance.
The results of this study have shown that the introduc-tion of externally applied robotic assistance with APAsfor lateral weight transfer and stability during rapid,self-paced step initiation significantly reduced the dura-tion of the APA and triggered an earlier step onset timein both patients with mild PD and healthy control sub- jects. While the subjects with PD had generally longerAPA and step duration than control subjects, the use of the lateral assist improved performance such that PDsubjects’ APA duration resembled the baseline perfor-mance of the healthy controls, while their step durationwas reduced. Because the first step length was compa-rable between the groups for all conditions, the reductionin step duration for PD subjects during the ASSIST trialsindicated that their execution time actually became fasterduring the ASSIST trials. These findings indicated thatrapid step initiation could be acutely enhanced throughexternal assistance with postural accompaniments that
FIG. 4.
 The group mean values plus 1 SD for first step onset time forControl subjects (CS, white bar) and subjects with Parkinson’s disease
 medication (OFF, black bar). The four experimental conditions areinitial baseline trials without postural assistance (BASE), trials withlateral postural assistance (ASSIST), follow-up trials without posturalassistance (NONE), and trials with a mechanical tug that minimizeddirect postural assistance (TUG). ***Significant differences betweenconditions at
 Group mean (standard deviation) step characteristics
GroupDuration (msec) Length (cm) Clearance (cm)BASE ASSIST NONE TUG BASE ASSIST NONE TUG BASE ASSIST NONE TUGCS *530.5 547.5 527.0 525.3 66.9 66.0 67.5 68.5 5.5 5.6 5.8 5.466.2 52.5 67.2 69.3 18.8 13.7 18.5 19.2 2.1 2.1 2.5 2.2PD *608.2
575.2 602.4 602.6 67.6 65.6 69.7 68.7 7.7 7.4 8.3 7.852.0 39.3 42.2 54.2 17.5 20.9 16.1 16.8 2.7 2.8 2.6 2.9Asterisks indicate significant main effect of group (
0.027). Step duration was longer across all conditions for PD subjects versus CS. Daggersindicate significant group by condition interaction (
0.015), ASSIST
all other conditions for PD subjects.CS, Control subjects; PD, subjects with Parkinson’s disease.
FIG. 5.
 The group mean values for first step duration for Controlsubjects (filled diamonds) and subjects with Parkinson’s disease (PD)
 medication (filled squares). The four experimental conditions areinitial baseline trials without postural assistance (BASE), trials withlateral postural assistance (ASSIST), follow-up trials without posturalassistance (NONE), and trials with a mechanical tug that minimizeddirect postural assistance (TUG).
Significant differences betweengroups at
A group by condition interactions with the PDgroup showing a decrease in step duration with a lateral postural assist.
 Movement Disorders, Vol. 22, No. 1, 2007 
of 8