doi: 10.2522/ptj.20110362Originally published online July 26, 2012

2012; 92:1452-1460.PHYS THER. and Melvyn RoerdinkO. Wiggerts, Wojtek Polomski, Thomas W.J. Janssen Han Houdijk, Mariëlle W. van Ooijen, Jos J. Kraal, HenriWith a Projected Visual ContextUnilateral Amputation on an Instrumented Treadmill Assessing Gait Adaptability in People With a

http://ptjournal.apta.org/content/92/11/1452found online at: The online version of this article, along with updated information and services, can be

Online-Only Material

10362.DC2.htmlhttp://ptjournal.apta.org/content/suppl/2012/10/29/ptj.201

10362.DC1.htmlhttp://ptjournal.apta.org/content/suppl/2012/10/29/ptj.201

Collections

Tests and Measurements     Outcomes Measurement    

Injuries and Conditions: Lower Extremity     Injuries and Conditions: Hip    

Gait Disorders     Amputation    

in the following collection(s): This article, along with others on similar topics, appears

e-Letters

"Responses" in the online version of this article. "Submit a response" in the right-hand menu under

or click onhere To submit an e-Letter on this article, click

E-mail alerts to receive free e-mail alerts hereSign up

at Vrije Universiteit, Library on November 20, 2012http://ptjournal.apta.org/Downloaded from

Assessing Gait Adaptability in PeopleWith a Unilateral Amputation on anInstrumented Treadmill With aProjected Visual ContextHan Houdijk, Marielle W. van Ooijen, Jos J. Kraal, Henri O. Wiggerts,Wojtek Polomski, Thomas W.J. Janssen, Melvyn Roerdink

Background. Gait adaptability, including the ability to avoid obstacles and to takevisually guided steps, is essential for safe movement through a cluttered world. Thisaspect of walking ability is important for regaining independent mobility but isdifficult to assess in clinical practice.

Objective. The objective of this study was to investigate the validity of an instru-mented treadmill with obstacles and stepping targets projected on the belt’s surfacefor assessing prosthetic gait adaptability.

Design. This was an observational study.

Methods. A control group of people who were able bodied (n�12) and groupsof people with transtibial (n�12) and transfemoral (n�12) amputations participated.Participants walked at a self-selected speed on an instrumented treadmill with pro-jected visual obstacles and stepping targets. Gait adaptability was evaluated in termsof anticipatory and reactive obstacle avoidance performance (for obstacles presented4 steps and 1 step ahead, respectively) and accuracy of stepping on regular andirregular patterns of stepping targets. In addition, several clinical tests were admin-istered, including timed walking tests and reports of incidence of falls and fear offalling.

Results. Obstacle avoidance performance and stepping accuracy were signifi-cantly lower in the groups with amputations than in the control group. Anticipatoryobstacle avoidance performance was moderately correlated with timed walking testscores. Reactive obstacle avoidance performance and stepping accuracy performancewere not related to timed walking tests. Gait adaptability scores did not differ ingroups stratified by incidence of falls or fear of falling.

Limitations. Because gait adaptability was affected by walking speed, differencesin self-selected walking speed may have diminished differences in gait adaptabilitybetween groups.

Conclusions. Gait adaptability can be validly assessed by use of an instrumentedtreadmill with a projected visual context. When walking speed is taken into account,this assessment provides unique, quantitative information about walking ability inpeople with a lower-limb amputation.

H. Houdijk, PhD, MOVE ResearchInstitute Amsterdam, Faculty ofHuman Movement Sciences, VUUniversity Amsterdam, van derBoechorststraat 9, 1081 BT,Amsterdam, the Netherlands, andHeliomare Rehabilitation,Research and Development, Wijkaan Zee, the Netherlands. Addressall correspondence to Dr Houdijkat: h.houdijk@vu.nl.

M.W. van Ooijen, MSc, MOVEResearch Institute Amsterdam,Faculty of Human Movement Sci-ences, VU University Amsterdam.

J.J. Kraal, MSc, MOVE ResearchInstitute Amsterdam, Faculty ofHuman Movement Sciences, VUUniversity Amsterdam.

H.O. Wiggerts, MD, AmsterdamRehabilitation Research Center–Reade, Amsterdam, theNetherlands.

W. Polomski, MD, HeliomareRehabilitation, Wijk aan Zee, theNetherlands.

T.W.J. Janssen, PhD, MOVEResearch Institute Amsterdam,Faculty of Human Movement Sci-ences, VU University Amsterdam,and Amsterdam RehabilitationResearch Center–Reade.

M. Roerdink, PhD, MOVEResearch Institute Amsterdam,Faculty of Human Movement Sci-ences, VU University Amsterdam.

[Houdijk H, van Ooijen MW, KraalJJ, et al. Assessing gait adaptabilityin people with a unilateral ampu-tation on an instrumented tread-mill with a projected visual con-text. Phys Ther. 2012;92:1452–1460.]

© 2012 American Physical TherapyAssociation

Published Ahead of Print:July 26, 2012

Accepted: July 23, 2012Submitted: October 25, 2011

Research Report

Post a Rapid Response tothis article at:ptjournal.apta.org

1452 f Physical Therapy Volume 92 Number 11 November 2012 at Vrije Universiteit, Library on November 20, 2012http://ptjournal.apta.org/Downloaded from

Walking is a context-specificactivity that demands gaitadjustments based on envi-

ronmental circumstances. For exam-ple, gait adjustments are essentialduring walking on uneven or clut-tered terrains to secure adequatefoot placement in relation to localenvironmental features, such asobstacles and stepping targets.1,2

Gait adaptability, defined as the abil-ity to adjust gait to such environmen-tal circumstances, is an importantdeterminant of the risk of falls3

because most falls result from a trip,a slip, or a misplaced step.4 How-ever, gait adaptability is difficult toassess in clinical practice.

This study was conducted to assessthe gait adaptability performance ofa heterogeneous group of peoplewho have a unilateral lower-limbamputation and who walk with aprosthesis. People with a lower-limbamputation are known to have anelevated risk of falls and a high inci-dence of falls (ie, 52% experienced afall in the previous year).5 Significantproportions of such people also havean increased fear of falling andreduced balance confidence (49%and 65%, respectively),5,6 whichlimit their participation.6 Therefore,regaining the ability to walk safelywithin the constraints of daily life isof great concern to people with alower-limb amputation.7,8 To helpguide targeted interventions, a com-prehensive assessment of walkingability is required; this assessmentshould include an evaluation of theability to adjust gait to environmentalcircumstances, given the aforemen-tioned relationship between gaitadaptability and risk of falls.

So far, prosthetic gait adaptability inrelation to environmental constraintshas been assessed in only a few stud-ies. In these studies, gait adaptabilitywas examined in terms of obstaclenegotiation during overground walk-ing9–11 or treadmill walking.12,13

Although people with a lower-limbamputation crossed overgroundobstacles in a different, more cau-tious manner than people in controlgroups of participants who wereable bodied,9,10 they were all able tosuccessfully avoid the obstacleswithout tripping or falling.9–11 Incontrast, for obstacle avoidance on atreadmill, when obstacles suddenlyfell in front of the prosthetic or non-prosthetic foot, people with a trau-matic transtibial amputation showedlower obstacle avoidance successrates than people in a control groupwho were able bodied.12,13 Thisresult was predominantly seen whenobstacles were presented under hightime pressure, because of delayedand smaller-amplitude muscularresponses.13 These findings signifiedthat people with a lower-limb ampu-tation had an impaired ability tomake online gait adjustments (eg,obstacle avoidance under time pres-sure)—but less so when theyadjusted gait in response to obstaclesthat were visible earlier, for whichrelevant visual information is typi-cally gathered 2 steps ahead of theobstacle and implemented in theavoidance response in an anticipa-tory manner.14,15 In addition, thesefindings indicated that a treadmill-based protocol, in which multipleobstacles can be presented consecu-tively under controlled responsetimes to modulate the time pressuredemands of obstacle avoidance, mayoffer a valid and clinically applicableassessment of prosthetic gaitadaptability.

The aim of this study was to test theface validity and construct validity ofa treadmill-based protocol for objec-tively assessing the ability of peoplewith a lower-limb amputation toadjust their gait in an online andanticipatory manner to various envi-ronmental circumstances. We usedan instrumented treadmill with avisual context (ie, visual obstaclesand stepping targets) projected on

the treadmill surface by means of aprojector.16–18 The visual contextwas used in 2 conditions. In condi-tion 1, visual obstacles were pre-sented either 1 or 4 steps ahead, andgait adaptability was evaluated interms of success rates for either reac-tive or anticipatory obstacle avoid-ance, respectively. In condition 2,regular and irregular patterns ofvisual stepping targets were pre-sented on the treadmill, and gaitadaptability was evaluated in termsof the accuracy of visually guidedstepping. Face validity was deter-mined by examining whether gaitadaptability scores (ie, success ratesand stepping accuracy) could dis-criminate among 3 groups of partic-ipants: participants with a transfemo-ral amputation (TF group),participants with a transtibial ampu-tation (TT group), and participantswho were able bodied and served ascontrols (CO group). Because walk-ing ability has been shown todepend on the level of amputa-tion,6,8 we expected inferior gaitadaptability scores for the TF group,intermediate scores for the TTgroup, and superior scores for theCO group. Construct validity wasdetermined by examining the rela-tionships between gait adaptabilityscores and various clinical testscores, including timed walking testsand self-reported incidence of fallsand fear of falling. Here, weexpected only moderate associa-tions, because gait adaptabilityscores were deemed to reflect aunique, complementary aspect of

Available WithThis Article atptjournal.apta.org

• eAppendix: Walking SpeedAffects Gait AdaptibilityPerformance

• Demonstration Video ofInstrumented Treadmill WithProjected Visual Context

Assessing Gait Adaptability in People With a Unilateral Amputation

November 2012 Volume 92 Number 11 Physical Therapy f 1453 at Vrije Universiteit, Library on November 20, 2012http://ptjournal.apta.org/Downloaded from

walking ability that is related, but notequal, to the walking ability con-structs of the aforementioned clini-cal tests.

MethodParticipantsA convenience sample of peoplewith a unilateral prosthesis aftertransfemoral amputation (n�12) ortranstibial amputation (n�12) andpeople who were able bodied andserved as controls (n�12) partici-pated in this study. Only people whowere 18 to 70 years of age and whowere able to walk without walkingaids for at least 4 minutes wereincluded. People with a lower-limbprosthesis additionally had to meetthe following criteria: be equippedwith a well-fitted prosthesis, be freeof stump problems, and have SpecialInterest Group in Amputee Medicine(SIGAM) mobility scale grades of C toF (ie, being able to walk indepen-dently with or without an assistivedevice).19 All participants were freeof any cardiorespiratory, neurologi-cal, and orthopedic pathologies (ie,other than related to the amputa-tion) that could affect walking abil-ity. All participants gave informedconsent to participate in this study.

InstrumentationAn instrumented treadmill with alarge embedded force platform(ForceLink BV, Culemborg, theNetherlands), which allowed for the

online detection of gait events (eg,heel-strike and toe-off) and gait char-acteristics (eg, cadence and steplength), was used to assess gaitadaptability. The procedure used toderive gait events from center-of-pressure (COP) profiles on a singleforce platform embedded in a tread-mill was previously outlined and val-idated by Roerdink et al.20 The tread-mill was connected to a projector toproject visual objects (movingtoward the participant at belt speed)on the treadmill belt in response todetected gait events.16 Specifically,visual obstacles could be presentedso that a person would step on themif he or she did not adjust gait (Figs.1A and 1B). Likewise, a sequence ofstepping targets could be presented(Fig. 1C) to correspond to a person’sstep length and cadence determinedonline (see video, available atptjournal.apta.org).17,18

Protocol and Data CollectionBefore the treadmill-based gait adapt-ability tests, the following standardtimed walking tests were adminis-tered to people in the TT and TFgroups: the 10-m walk test at com-fortable (10MWTcomf) and maximal(10MWTmax) walking speeds,21 theTimed “Up & Go” Test (TUG),22 andthe sub-item obstacle avoidance ofthe Emory Functional AmbulationProfile (Emory-FAP).23 In addition,the Activities-specific Balance Confi-dence Scale (ABC) was adminis-

tered,24 and the self-reported inci-dence of falls during the past 12months5 and fear of falling6 werescored.

Subsequently, all participants prac-ticed treadmill walking for a mini-mum of 5 minutes at various beltspeeds. Participants were encour-aged to walk unsupported, but hand-rail support was allowed if neces-sary. All participants walked withtheir own prosthesis and comfort-able walking shoes. They wore asafety harness that was suspendedover the treadmill (no weight bear-ing). The self-selected comfortabletreadmill walking speed was deter-mined by gradually increasing thebelt speed until a comfortable walk-ing speed was reported. Belt speedwas then increased beyond thejust-reported comfortable walkingspeed and subsequently graduallydecreased until a comfortable walk-ing speed was reported. Belt speedwas set to the average of the 2reported comfortable treadmill walk-ing speeds (hereafter referred to as“self-selected walking speed”) forthe execution of the following 2 con-ditions (offered in sequential order).

Condition 1: anticipatory andreactive obstacle avoidance. Thiscondition comprised two 4-minutetrials during which participants wereconfronted with a series of 12 visualobstacles each. Obstacles were pro-jected onto the treadmill belt as awhite patch of light moving towardthe participant at a speed corre-sponding to the belt speed. Theobstacle dimensions were setaccording to the width of the beltand the size of the participant’s shoe.The starting position of the obstaclewas such that it would arrive exactlybeneath either the left foot or theright foot (6 each, in random order)if the participant continued walkingin the way he or she had been walk-ing, as indicated by the foot positiondetermined online20 and with the

Figure 1.Instrumented treadmill with a projected visual context (ie, visual obstacles and steppingtargets) for assessing gait adaptability. (A) Anticipatory obstacle avoidance. (B) Reactiveobstacle avoidance. (C) Visually guided stepping on patterns of stepping targets.

Assessing Gait Adaptability in People With a Unilateral Amputation

1454 f Physical Therapy Volume 92 Number 11 November 2012 at Vrije Universiteit, Library on November 20, 2012http://ptjournal.apta.org/Downloaded from

average length of the 8 precedingsteps.

In the anticipatory obstacle avoid-ance trial, the 12 obstacles were pre-sented at the moment of foot con-tact, 4 steps in advance of theanticipated contact (Fig. 1A), allow-ing for an anticipatory gait adjust-ment to avoid the obstacle. In thereactive obstacle avoidance trial, the12 obstacles were presented undertime pressure, at the moment of foot-off, 1 step in advance of the antici-pated contact (Fig. 1B), requiring areactive gait adjustment during theswing phase of the ongoing step toavoid the obstacle. Sufficient timewas available between obstacles forparticipants to reach a steady-statewalking pattern again. Anticipatoryand reactive obstacle avoidance trialswere presented in counterbalancedorder, each followed by a 5-minuteseated rest period.

Success rates (as percentages) inboth obstacle avoidance trials werescored manually by one of the exper-imenters through visual inspection.To be classified as successful avoid-ance, both feet had to be positionedoutside the area of the obstacle (ie,no overlap of shoe and obstacle).This visual inspection was deemedsuperior to using available COP databecause COP data do not reveal thereal margins of the foot and hencecannot reveal whether part of thefoot area was within the area of theobstacle. Sagittal video recordingswere made for verification of uncer-tain obstacle crossings.

Condition 2: walking on regularor irregular patterns of steppingtargets. In this condition, rectan-gular stepping targets (length: theparticipant’s shoe length; width: halfthe belt width) were projected ontothe treadmill belt for both the leftfoot and the right foot. Participantswere instructed to walk on the pre-sented step pattern, aligning their

feet as much as possible in the ante-rior and posterior directions on thestepping targets. Approximately 4 or5 stepping targets were visible ateach moment in time (Fig. 1C). Thiscondition consisted of 3 trials withdifferent levels of regularity in thepatterns of stepping targets (ie, 0%,20%, and 30% variations). The 0%trial consisted of a regular pattern ofstepping targets corresponding to aparticipant’s mean gait pattern, asregistered in the preceding 20 sec-onds of unconstrained treadmillwalking.17,18 In the irregular trials, arandom variation of 20% or 30% of aparticipant’s mean stride length wasadded to the aforementioned regularpattern of stepping targets. Each trialconsisted of a sequence of 80 step-ping targets. The 3 trials wereoffered in a counterbalanced orderand were followed by a period ofseated rest.

Stepping accuracy (in millimeters)was quantified in terms of the varia-tion in foot placement relative to thestepping targets. The anterior-posterior COP position at mid-stance(ie, the COP position halfwaybetween the subsequent instances offoot-off and initial contact of the con-tralateral foot) were compared withthe anterior-posterior position of thecenter of the stepping targets; theformer position was used as a refer-ence for the foot placement locationon the treadmill. Note that althoughthe COP position at mid-stance isclose to the center of the foot, it isnot necessarily perfectly so; in thepilot study on people who were ablebodied (n�12; mean age�28.6years, SD�4.4), we found a system-atic error between the true center ofthe foot and the COP position at mid-stance of �1 cm (95% confidenceinterval), depending on the mannerin which people walked. As a conse-quence, a systematic offset betweenthe center of the stepping target andthe COP position at mid-stance maybe observed, even when the foot is

perfectly aligned on the stepping tar-get. To circumvent such bias, weused the standard deviations of thedistances between the COP positionat mid-stance and the center of thestepping target (instead of the mean)as a measure of the accuracy of footplacement. The smaller this standarddeviation, the more accurate the footplacement positions are relative tothe stepping targets.

Supplementary experiment: effectof walking speed on gait adapt-ability scores. Walking speed mayaffect gait adaptability performance.In a supplementary experiment, wetherefore assessed the effect of walk-ing speed per se on the gait adapt-ability scores of participants in theCO group. These participants per-formed both conditions at their self-selected walking speed and at aslower speed (expected to representthe lower range of walking speedsobserved in participants with anamputation), in randomized order, asfurther described in the eAppendix(available at ptjournal.apta.org).

Data AnalysisGeneral group characteristics andself-selected treadmill walkingspeed. The 3 groups were com-pared in terms of age, body weight,height, and self-selected treadmillwalking speed by use of separate1-way analyses of variance (ANO-VAs) (factor: group; levels: CO, TT,and TF) followed by post hoc inde-pendent t tests for significant groupeffects. The TT group and the TFgroup were compared in terms ofthe side and the cause of amputationby use of chi-square tests and interms of the time since amputationand the time using the current pros-thesis by use of independent t tests.

Condition 1. Nonparametric sta-tistics were used to examine groupeffects on obstacle avoidance suc-cess rates. First, Wilcoxon signedrank tests were performed to com-

Assessing Gait Adaptability in People With a Unilateral Amputation

November 2012 Volume 92 Number 11 Physical Therapy f 1455 at Vrije Universiteit, Library on November 20, 2012http://ptjournal.apta.org/Downloaded from

pare anticipatory and reactive obsta-cle avoidance success rates sepa-rately for the 3 groups. Differencesin success rates between groupswere evaluated with Kruskal-Wallistests; Mann-Whitney U tests wereused for post hoc follow-up in case ofsignificant group effects.

Condition 2. Differences in step-ping accuracy between groups weretested by use of a repeated-measuresANOVA with the between-subjectfactor group (3 levels: TF, TT, andCO) and the within-subject factorpattern regularity (3 levels: 0%, 20%,and 30%). Post hoc comparisonswere performed with independent

t tests (group effects) or paired-sample t tests (pattern regularityeffects).

Gait adaptability performance inrelation to clinical test scores.Gait adaptability scores were relatedto 10MWTcomf, 10MWTmax, TUG,Emory-FAP, and ABC scores by use ofSpearman (success rates) or Pearson(stepping accuracy) correlation coef-ficients. To examine the relation-ships between gait adaptabilityscores and self-reported incidence offalls and fear of falling, we stratifiedthe groups into subgroups (“fallers”/“nonfallers” and fearful/nonfearful)and then used Mann-Whitney U tests

to compare obstacle avoidance suc-cess rates for these subgroups. Like-wise, to compare stepping accura-cies for the stratified subgroups, weused repeated-measures ANOVAs forsubgroup (2 levels: fallers versusnonfallers and fearful versus nonfear-ful) � pattern regularity (3 levels:0%, 20%, and 30%).

All statistical tests were performedwith SPSS version 17.0 (SPSS Inc,Chicago, Illinois). Level of signifi-cance (�) was set at .05 for testingmain and interaction effects. For posthoc tests, � was corrected for multi-ple comparisons by dividing .05 (�)by the number of post hoc compari-sons performed. In general, for posthoc tests among the 3 subgroups, �was .017.

Role of the Funding SourceThis study was supported by a grantfrom the OIM Foundation, Assen, theNetherlands. The contribution of DrRoerdink was supported by VeniGrant 451-09-024 from the Nether-lands Organization for ScientificResearch (NWO).

ResultsAll participants were able to com-plete conditions 1 and 2 successfullyand were highly motivated to com-plete the tasks. Seven of 12 partici-pants in the TF group required hand-rail support for confidence.

General Group CharacteristicsThe 3 groups did not differ in termsof age, body weight, and height (allF2,35�1.5, P�.24). The TT groupand the TF group did not differ interms of the side and the cause ofamputation (all �2

1�2.27, P�.13) orin the time since amputation and thetime using the current prosthesis (allt22�1.0, P�.32). Table 1 shows thegroup characteristics. All partici-pants in the TT and TF groups woretheir own custom-designed prosthe-sis. The types of prosthetic compo-nents used varied among the partic-

Table 1.General Group Characteristics and Clinical Test Scoresa

Characteristic TF Group TT Group CO Group

Age, y 51 (10) 43 (14) 44 (12)

Weight, kg 74 (13) 79 (13) 76 (13)

Height, m 1.75 (0.13) 1.80 (0.07) 1.78 (0.09)

Sex (no. of men/women) 7/5 10/2 6/6

SIGAM (grades D/E/F) 6/2/4 0/1/11

Time since amputation (mo)

X (SD) 213 (214) 208 (218)

Range 6–648 14–540

Time using current prosthesis (mo)

X (SD) 15 (15) 22 (17)

Range 2–48 2–60

Cause of amputation (trauma/vascular condition),no. of participants

8/4 11/1

Side of amputation (right/left), no. of participants 5/7 6/6

Clinical test scores

10MWTcomf, s 9.2 (1.2) 7.5 (0.8)

10MWTmax, s 7.7 (1.2) 5.9 (0.8)

TUG, s 10.8 (2.5) 8.4 (1.8)

Emory-FAP, s 13.1 (2.3) 9.5 (0.8)

ABC, % 88.1 (8.1) 95.2 (4.5)

Incidence of falls (yes/no), no. of participants 7/5 2/10

Fear of falling (yes/no), no. of participants 4/8 2/10

a Data are presented as mean (standard deviation) unless stated otherwise. TF�participants with atransfemoral amputation, TT�participants with a transtibial amputation, CO�participants who wereable bodied (controls), SIGAM�Special Interest Group in Amputee Medicine, 10MWTcomf�10-m walktest at a comfortable walking speed, 10MWTmax�10-m walk test at the maximal walking speed,TUG�Timed “Up & Go” Test, Emory-FAP�Emory Functional Ambulation Profile, ABC�Activities-specific Balance Confidence Scale.

Assessing Gait Adaptability in People With a Unilateral Amputation

1456 f Physical Therapy Volume 92 Number 11 November 2012 at Vrije Universiteit, Library on November 20, 2012http://ptjournal.apta.org/Downloaded from

ipants, but all participants in the TTgroup used a type of dynamic foot.Within the TF group, 6 of 12 partic-ipants used a processor-controlledknee (C-leg; Otto Bock, Duderstadt,Germany), 5 participants used amechanically controlled knee, and 1participant used a rigid knee.

Self-selected Treadmill WalkingSpeedSelf-selected comfortable treadmillwalking speeds differed significantlyamong the groups (F2,35�6.35,P�.001), with participants in the TFgroup walking significantly moreslowly (2.3 km/h) than participantsin both the TT group (3.3 km/h;t22�2.90, P�.008) and the COgroup (3.8 km/h; t22�6.12, P�.001).The speed difference between theTT group and the CO group didnot reach significance (t22�1.98,P�.060).

Condition 1: Anticipatory andReactive Obstacle AvoidancePerformanceSuccess rates for anticipatory obsta-cle avoidance were significantlyhigher than those for reactive obsta-cle avoidance for all 3 groups (Fig. 2)(TF group: z�2.786, P�.005; TTgroup: z�2.764, P�.006; CO group:z�2.375, P�.018). Furthermore, sig-nificant group effects were observedfor both anticipatory obstacle avoid-ance (H2�13.4, P�.001) and reac-tive obstacle avoidance (H2�8.5,P�.015) (Fig. 2). Post hoc compari-sons revealed that success rates weresignificantly lower for the TF groupthan for the CO group for both antic-ipatory obstacle avoidance (U�12.0,P�.001) and reactive obstacle avoid-ance (U�23.5, P�.005). The TTgroup showed intermediate successrates for both anticipatory obstacleavoidance and reactive obstacleavoidance. However, post hoc com-parisons revealed that differences insuccess rates between the TT groupand the CO group (anticipatory:U�45.0, P�.062; reactive: U�35.0,

P�.030) as well as between the TTgroup and the TF group (anticipa-tory: U�42.5, P�.078; reactive:U�69.0, P�.862) did not reach sig-nificance (bearing in mind the cor-rected � of .017).

Condition 2: Walking on Regularor Irregular Patterns of SteppingTargetsData from 1 person in the TF groupwere lost because of measurementerror. The effect of group tendedtoward significance (F2,32�2.55,P�.094), with seemingly inferior

stepping accuracy for the TF group(43 mm), intermediate steppingaccuracy for the TT group (36 mm),and superior stepping accuracy forthe CO group (31 mm). Further-more, a significant effect of patternregularity (F2,64�23.2, P�.001) wasobserved, with higher stepping accu-racy for the regular sequence of step-ping targets (31 mm) than for bothirregular stepping target sequences(20%: 38 mm, P�.001; 30%: 41 mm,P�.001). As indicated in Figure 3, allgroups showed lower stepping accu-racy for more irregular patterns of

Figure 2.Success rates for anticipatory and reactive obstacle avoidance assessments for partici-pants who were able bodied (controls) (CO), participants with a transtibial amputation(TT), and participants with a transfemoral amputation (TF).

Figure 3.Stepping accuracy for participants who were able bodied (controls) (CO), participantswith a transtibial amputation (TT), and participants with a transfemoral amputation(TF). Accuracy was determined from visually guided stepping on regular and irregular(20% and 30%) patterns of stepping targets.

Assessing Gait Adaptability in People With a Unilateral Amputation

November 2012 Volume 92 Number 11 Physical Therapy f 1457 at Vrije Universiteit, Library on November 20, 2012http://ptjournal.apta.org/Downloaded from

stepping targets; the interaction ofgroup and pattern regularity was notsignificant (F4,64�0.32, P�.86).

Gait Adaptability Performance inRelation to Clinical Test ScoresAnticipatory obstacle avoidance suc-cess rates were negatively correlatedwith 10MWTcomf, 10MWTmax, TUG,and Emory-FAP scores (Tab. 2);higher success rates were associatedwith faster test completion. All sig-nificant correlations were of inter-mediate strength. No significant cor-relations were found for reactiveobstacle avoidance. No difference inanticipatory and reactive obstacleavoidance success rates was foundfor subgroups stratified into nonfall-ers and fallers (U�61.5 [P�.71] andU�50.0 [P�.294], respectively) orfor subgroups stratified by fear offalling (U�33.0 [P�.15] andU�35.5 [P�.22], respectively).

Stepping accuracy was not signifi-cantly correlated with the outcomesof the clinical tests (Tab. 2), nor didstepping accuracy differ for sub-groups stratified into fallers(F1,21�0.11, P�.74) or fear of falling(F1,21�0.17, P�.68).

DiscussionIn the present study, an instru-mented treadmill with projected

visual obstacles and stepping targetswas successfully used to assess gaitadaptability in 3 groups of partici-pants: 2 groups with amputation (TFgroup and TT group) and a controlgroup (CO group). We expected andfound significant differences in gaitadaptability performance among thegroups, with inferior gait adaptabil-ity scores for the TF group, interme-diate scores for the TT group, andsuperior scores for the CO group.Besides these between-group differ-ences, large intragroup variations ingait adaptability scores wereobserved. Although these variationsreduced the likelihood of finding sig-nificant differences among thegroups, they reflected the variationsin walking ability that can beexpected in a convenience sample ofpeople with a lower-limb amputa-tion; such variations can be due tothe cause of amputation, type ofprosthesis, prosthetic fit, comorbidi-ties, and age.8,25–27 Consequently,the large intragroup variations testifyto the sensitivity of the selected gaitadaptability assessments for identify-ing individual within-group differ-ences, thereby promoting their facevalidity. The sensitivity of the gaitadaptability assessments was furtherunderscored by 2 strong within-subject effects that held for all

groups; obstacle avoidance successrates decreased when fewer stepswere available for obstacle avoid-ance (Fig. 2), and stepping accuracydecreased with increased patternirregularity (Fig. 3).

Significant correlations between gaitadaptability scores and timed walk-ing test outcomes were found onlyfor anticipatory obstacle avoidancesuccess rates (Tab. 2). In particular, asignificant correlation was expectedand found between anticipatoryobstacle avoidance success rates andthe sub-item obstacle avoidance ofthe Emory-FAP.23 This test bestresembled the construct of anticipa-tory obstacle avoidance because itassesses obstacle negotiation perfor-mance on an overground obstaclecourse, and visual information aboutthe obstacle is available several stepsin advance. In contrast, the correla-tion with the Emory-FAP was not sig-nificant for reactive obstacle avoid-ance success rates (Tab. 2). Becausereactive obstacle avoidance requiresgait adjustments under time pres-sure,13 this condition apparently rep-resents a different construct. Simi-larly, we can draw a more generalconclusion that the observed moder-ate correlations (eg, for anticipatoryobstacle avoidance success rates) incombination with several nonsignifi-cant correlations (for reactive obsta-cle avoidance success rates and step-ping accuracy) indicated that ourgait adaptability scores quantified aunique, complementary aspect ofwalking ability. This unique aspect,putatively related to the ability tomake online gait adjustments undertime pressure, was not captured byany of the timed walking tests inisolation.

In addition to timed walking tests,we related gait adaptability scores tomarkers related to the risk of falls,including self-reported incidence offalls, fear of falling, and ABC scores.Gait adaptability and risk of falls, fear

Table 2.Correlations Between Clinical Test Scores and Gait Adaptability Performancea

Condition

Correlation With Scores on:

10MWTcomf 10MWTmax TUG Emory-FAP ABC

Obstacle avoidance

Anticipatory �.608 �.557 �.425 �.589 .257

Reactive �.376 �.288 �.207 �.249 .345

Pattern walking

0% variation .105 .067 .221 .344 �.221

20% variation .052 .070 .127 .265 �.147

30% variation �.180 �.093 �.079 .076 �.056

a In terms of success rates for anticipatory and reactive obstacle avoidance and accuracy of stepping onregular and irregular patterns of stepping targets. Bold type indicates significant correlations.10MWTcomf�10-m walk test at a comfortable walking speed, 10MWTmax�10-m walk test at themaximal walking speed, TUG�Timed “Up & Go” Test, Emory-FAP�Emory Functional AmbulationProfile, ABC�Activities-specific Balance Confidence Scale.

Assessing Gait Adaptability in People With a Unilateral Amputation

1458 f Physical Therapy Volume 92 Number 11 November 2012 at Vrije Universiteit, Library on November 20, 2012http://ptjournal.apta.org/Downloaded from

of falling, and balance confidenceare related but different con-structs,24,28 as evidenced by theabsence of significant correlationsbetween gait adaptability scores andABC scores as well as the absence ofsystematic differences in gait adapt-ability scores for subgroups stratifiedby fear of falling and incidence offalls. Gait adaptability scores objec-tively reflect a person’s ability toadjust gait during obstacle negotia-tion and visually guided stepping,whereas ABC scores and reportsabout fear of falling represent a per-son’s confidence with regard tothose activities and potentially thewillingness to engage in suchevents.24 Obviously, the incidence offalls is determined by the combina-tion of (1) gait adaptability and (2)balance confidence and fear of fall-ing. For example, the incidence offalls can differ between people withequally low gait adaptability scoresbecause of different activity intensi-ties or differences in risk-taking pro-pensities in relation to personal vari-ations in fear of falling or balanceconfidence. Notwithstanding themediating effect of fear of falling onthe incidence of falls, reduced gaitadaptability performance itselfseems to be an important, indepen-dent determinant of the risk offalls.3,12 The present study paves theway for a meaningful assessment of aperson’s capacity to adjust gait to theenvironmental context (such as step-ping targets and obstacles) withestablished face validity and con-struct validity.

LimitationsBefore our conclusions can beaccepted, the effect of walkingspeed as a potential confounding fac-tor for gait adaptability performancemust be considered for 2 reasons.First, the group showing inferior gaitadaptability scores (TF group) alsoperformed the tests at the slowestself-selected treadmill speed. Sec-ond, the supplementary experiment

revealed that walking speed directlyaffected gait adaptability scores forstepping accuracy and reactiveobstacle avoidance performance.Both scores improved with walkingat a speed that was slower than theself-selected speed (eAppendix). Onthe basis of these results, it could behypothesized that people with areduced capacity to adjust gait adopta slower walking speed to be betterable to cope with environmental cir-cumstances that demand gait adjust-ments. As a consequence, reactiveobstacle avoidance and guided step-ping performance of groups or indi-viduals walking at a slower self-selected walking speed may havebeen overestimated in the presentstudy relative to the performance ofgroups or individuals walking at afaster self-selected speed. Therefore,a comprehensive analysis of gaitadaptability should not only examinegait adaptability performance butalso evaluate this performance withrespect to possible changes in self-selected walking speed.

In addition to having a slower walk-ing speed, 7 of 12 participants in theTF group used handrail support forconfidence during the treadmilltests. This handrail support probablyfacilitated their performance onobstacle avoidance and steppingtasks, thereby diminishing group dif-ferences due to an overestimation ofgait adaptability scores in the TFgroup. Like the effect of walkingspeed on gait adaptability perfor-mance (as discussed above), this fac-tor must be considered in the inter-pretation of between-group effects(or the absence thereof).

The present study was a first attemptto explore the use of an instru-mented treadmill with projectedvisual obstacles and stepping targetsfor assessing prosthetic gait adapt-ability. We observed that all partici-pants, representing a heterogeneoussample of people with diverse char-

acteristics (eg, level and cause ofamputation, age, and time sinceamputation) (Tab. 1), were well ableto complete and enjoy obstacleavoidance and guided steppingtasks. In addition, we established theface validity and construct validity ofsuch assessments. However, to fullydetermine the usefulness of aninstrumented treadmill with a pro-jected visual context (ie, visual obsta-cles and stepping targets) for routineclinical evaluation, future researchmust next establish the reproducibil-ity of the selected gait adaptabilitytests and sensitivity to change afterclinical interventions.

ConclusionWe successfully assessed and com-pared gait adaptability in 3 groups ofparticipants using obstacle avoid-ance and visually guided steppingtasks on an instrumented treadmillwith a projected visual context (ie,visual obstacles and stepping tar-gets). The gait adaptability perfor-mance scores so obtained indicatedadequate face validity, as evidencedby significant between-group differ-ences in gait adaptability. The assess-ment of construct validity revealedthat gait adaptability performanceadds unique information to com-monly used timed walking tests andself-reported incidence of falls andfear of falling. This added informa-tion reflects a person’s ability tomake step adjustments under timepressure. We, therefore, proposethat a gait adaptability assessmentshould be considered an integral partof a walking ability evaluation, espe-cially when this construct is the maingoal of intervention, such as withnovel microprocessor-controlledprosthetic components or functionalgait training. Because walking speedcan affect gait adaptability perfor-mance, it should be taken intoaccount in the interpretation of gaitadaptability scores for different indi-viduals or over time.

Assessing Gait Adaptability in People With a Unilateral Amputation

November 2012 Volume 92 Number 11 Physical Therapy f 1459 at Vrije Universiteit, Library on November 20, 2012http://ptjournal.apta.org/Downloaded from

All authors provided concept/idea/researchdesign, writing, and consultation (includingreview of manuscript before submission).Mrs van Ooijen and Mr Kraal provided datacollection. Mrs van Ooijen, Mr Kraal, DrHoudijk, and Dr Roerdink provided dataanalysis. Dr Houdijk and Dr Roerdink pro-vided project management and fund pro-curement. Mr Wiggerts and Mr Polomskiprovided study participants and institutionalliaisons. Dr Houdijk and Dr Janssen providedfacilities/equipment. The contributions ofIrene Derksen, Marcia Teunissen, and OtelieBosboom in selecting and including partici-pants are gratefully acknowledged. Theauthors thank Bert Coolen for developingand implementing the software formovement-dependent presentation ofobstacles and regular and irregular patternsof stepping targets.

This study was approved by the Medical Eth-ical Review Board, VU University MedicalCentre, Amsterdam, the Netherlands.

Parts of the article were presented at theWorld Confederation for Physical TherapyCongress; June 20–23, 2011; Amsterdam,the Netherlands; and at the 13th World Con-gress of the International Society of Prosthet-ics and Orthotics; May 10–15, 2010;Leipzig, Germany.

This study was supported by a grant fromthe OIM Foundation, Assen, the Nether-lands. The contribution of Dr Roerdink wassupported by Veni Grant 451-09-024 fromthe Netherlands Organization for ScientificResearch (NWO).

DOI: 10.2522/ptj.20110362

References1 Hollands MA, Marple-Horvat DE. Coordina-

tion of eye and leg movements during visu-ally guided stepping. J Mot Behav. 2001;33:205–216.

2 Patla AE. Understanding the roles of visionin the control of human locomotion. GaitPosture. 1997;5:54–69.

3 Weerdesteyn V, Rijken H, Geurts ACH,et al. A five-week exercise program canreduce falls and improve obstacle avoid-ance in the elderly. Gerontology. 2006;51:131–141.

4 Berg WP, Alessio HM, Mills EM, Tong C.Circumstances and consequences of fallsin independent community-dwelling olderadults. Age Ageing. 1997;26:261–268.

5 Miller WC, Deathe AB, Speechley M, KovalJ. The influence of falling, fear of falling,and balance confidence on prostheticmobility and social activity among individ-uals with a lower extremity amputation.Arch Phys Med Rehabil. 2001;82:1238–1244.

6 Miller WC, Speechley M, Deathe B. Theprevalence and risk factors of falling andfear of falling among lower extremityamputees. Arch Phys Med Rehabil. 2001;82:1031–1037.

7 Sansam K, Neumann V, O’Connor R,Bhakta B. Predicting walking ability fol-lowing lower limb amputation: a system-atic review of the literature. J RehabilMed. 2009;41:593–603.

8 van Velzen JM, van Bennekom CAM,Polomski W, et al. Physical capacity andwalking ability after lower limb amputa-tion: a systematic review. Clin Rehabil.2006;20:999–1016.

9 Hill SW, Patla AE, Ishac MG, et al. Kine-matic patterns of participants with abelow-knee prosthesis stepping overobstacles of various heights in locomo-tion. Gait Posture. 1997;6:186–192.

10 Hill SW, Patla AE, Ishac MG, et al. Alteredkinetic strategy for the control of swinglimb elevation over obstacles in unilateralbelow-knee amputee gait. J Biomech.1999;32:545–549.

11 Vrieling AH, van Keeken HG, Schoppen T,et al. Obstacle crossing in lower limbamputees. Gait Posture. 2007;26:587–594.

12 Hofstad CJ, van der Linde H, Nienhuis B,et al. High failure rates when avoidingobstacles during treadmill walking inpatients with a transtibial amputation.Arch Phys Med Rehabil. 2006;87:1115–1122.

13 Hofstad CJ, Weerdesteyn V, van der LindeH, et al. Evidence for bilaterally delayedand decreased obstacle avoidanceresponses while walking with a lowerlimb prosthesis. Clin Neurophysiol. 2009;120:1009–1015.

14 Berg WP, Murdock LA. Age-related differ-ences in locomotor targeting performanceunder structural interference. Age Ageing.2011;40:324–329.

15 Patla AE, Vickers JN. How far ahead do welook when required to step on specificlocations in the travel path during locomo-tion? Exp Brain Res. 2003;148:133–138.

16 Roerdink M, Beek PJ, inventors; ForceLinkBV, assignee. Device for displaying targetindications for foot movements to personswith a walking disorder. US patent2009246746-A1 (October 1, 2009), Euro-pean patent 2106779-A1 (October 7,2009), Japanese patent 2009240775-A(October 22, 2009), and Dutch patent1035236-C2 (October 1, 2009).

17 Bank PJ, Roerdink M, Peper CE. Compar-ing the efficacy of metronome beeps andstepping stones to adjust gait: steps to fol-low! Exp Brain Res. 2011;209:159–169.

18 Wezenberg D, de Haan A, van BennekomCA, Houdijk H. Mind your step: metabolicenergy cost while walking an enforcedgait pattern. Gait Posture. 2011;33:544–549.

19 Ryall NH, Eyres SB, Neumann VC, et al.The SIGAM mobility grades: a newpopulation-specific measure for lowerlimb amputees. Disabil Rehabil. 2003;25:833–844.

20 Roerdink M, Coolen BH, Clairbois BH,et al. Online gait event detection using alarge force platform embedded in a tread-mill. J Biomech. 2008;41:2628–2632.

21 van Hedel HJ, Wirz M, Dietz V. Assessingwalking ability in subjects with spinal cordinjury: validity and reliability of 3 walkingtests. Arch Phys Med Rehabil. 2005;86:190–196.

22 Schoppen T, Boonstra A, Groothoff JW,et al. The Timed “up and go” test: reliabil-ity and validity in persons with unilaterallower limb amputation. Arch Phys MedRehabil. 1999;80:825–828.

23 Wolf SL, Catlin PA, Gage K, et al. Establish-ing the reliability and validity of measure-ments of walking time using the EmoryFunctional Ambulation Profile. Phys Ther.1999;79:1122–1133.

24 Powell LE, Myers AM. The Activities-specific Balance Confidence (ABC) Scale.J Gerontol A Biol Sci Med Sci. 1995;50:M28–M34.

25 Lamoth CJ, Ainsworth E, Polomski W,Houdijk H. Variability and stability analysisof walking of transfemoral amputees. MedEng Phys. 2010;32:1009–1014.

26 MacKenzie EJ, Bosse MJ, Castillo RC, et al.Functional outcomes following trauma-related lower-extremity amputation.J Bone Joint Surg Am. 2004;86:1636–1645.

27 Schmalz T, Blumentritt S, Jarasch R.Energy expenditure and biomechanicalcharacteristics of lower limb amputee gait:the influence of prosthetic alignment anddifferent prosthetic components. GaitPosture. 2002;16:255–263.

28 Tinetti ME, Richman D, Powell L. Fallsefficacy as a measure of fear of falling.J Gerontol. 1990;45:P239–P243.

Assessing Gait Adaptability in People With a Unilateral Amputation

1460 f Physical Therapy Volume 92 Number 11 November 2012 at Vrije Universiteit, Library on November 20, 2012http://ptjournal.apta.org/Downloaded from

doi: 10.2522/ptj.20110362Originally published online July 26, 2012

2012; 92:1452-1460.PHYS THER. and Melvyn RoerdinkO. Wiggerts, Wojtek Polomski, Thomas W.J. Janssen Han Houdijk, Mariëlle W. van Ooijen, Jos J. Kraal, HenriWith a Projected Visual ContextUnilateral Amputation on an Instrumented Treadmill Assessing Gait Adaptability in People With a

References

http://ptjournal.apta.org/content/92/11/1452#BIBLfor free at: This article cites 27 articles, 5 of which you can access

Information Subscription http://ptjournal.apta.org/subscriptions/

Permissions and Reprints http://ptjournal.apta.org/site/misc/terms.xhtml

Information for Authors http://ptjournal.apta.org/site/misc/ifora.xhtml

at Vrije Universiteit, Library on November 20, 2012http://ptjournal.apta.org/Downloaded from