Original Research

Plantar Pressures and Ground Reaction ForcesDuring Walking of Individuals With UnilateralTransfemoral AmputationMarcelo Peduzzi de Castro, PhD, MSc, PT, Denise Soares, PhD, Emília Mendes, MSc,Leandro Machado, PhD

M.P.C. Center of Research, Education, Inno-vation and Intervention in Sport, Faculty ofSport, University of Porto, Porto, Portugal; PortoBiomechanics Laboratory, University of Porto,Rua Dr. Plácido Costa, 91 - 4200.450, Porto,Portugal. Address correspondence to: M.PdC.;e-mail: marcelocastro_fisio@hotmail.comDisclosure: nothing to disclose

D.S. Center of Research, Education, Innova-tion and Intervention in Sport, Faculty of Sport,University of Porto, Porto, Portugal; PortoBiomechanics Laboratory, University of Porto,Porto, PortugalDisclosure related to this publication: receiveda PhD scholarship supported by Foundationfor Science and Technology (FCT) from

Objective: To describe and compare the plantar pressures, temporal foot roll-over, andground reaction forces (GRFs) between both limbs of subjects with unilateral transfemoralamputation and with those of able-bodied participants during walking. We also verify therelevance of a force plate and a pressure plate to discriminate changes in gait parameters ofsubjects with limb loss.Design: Cross-sectional study.Setting: Biomechanics laboratory.Subjects: A total of 14 subjects with unilateral transfemoral amputation and 21 able-bodied participants.Methods: We used a force plate and a pressure plate to assess biomechanical gait pa-rameters while the participants were walking at their self-selected gait speed.Main Outcome Measurements: We measured plantar pressure peaks in 6 footregions and the instant of their occurrence (temporal foot roll-over); and GRF peaks andimpulses of anterior-posterior (braking and propulsive phases), medial-lateral, and vertical(load acceptance and thrust phases) components.Results: The thrust, braking, and propulsive peaks, and the braking and propulsive im-pulses, were statistically significantly lower in the amputated limb than in the sound limb(P < .05) and in able-bodied participants (P < .05). In the amputated limb, we observedhigher pressure peaks in the lateral rearfoot and medial and lateral midfoot, and lower valuesin the forefoot regions compared to those in the other groups (P < .05). The temporal footroll-over showed statistically significant differences among the groups (P < .05).Conclusions: The plantar pressures, temporal foot roll-over, and GRFs in subjects withunilateral transfemoral amputation showed an asymmetric gait pattern, and different valueswere observed in both of their lower limbs as compared with those of able-bodied subjectsduring walking. The force plate and pressure plate were able to determine differencesbetween participants in gait pattern, suggesting that both plantar pressure and GRF ana-lyses are useful tools for gait assessment in individuals with unilateral transfemoralamputation. Because of the convenience of pressure plates, their use in the clinical contextfor prosthetic management appears relevant to guide the rehabilitation of subjects withlower limb amputation.

PM R 2014;-:1-11

Portugal (The author derives no financialbenefit from this publication.)

E.M. Department of Bioengineering, Universityof Strathclyde, Scotland, UK; Center of Pro-fessional Rehabilitation of Gaia (CRPG),Arcozelo, PortugalDisclosure: nothing to disclose

L.M. Center of Research, Education, Innova-tion and Intervention in Sport, Faculty of Sport,University of Porto, Porto, Portugal; PortoBiomechanics Laboratory, University of Porto,Porto, PortugalDisclosure: nothing to disclose

Submitted for publication May 29, 2013;accepted January 26, 2014.

INTRODUCTION

As a consequence of the absence of a natural limb, subjects with unilateral transfemoral(TF) amputation often complain about pain in the sound limb (SL) [1]. They show a higherprevalence of osteoarthritis [2,3], scoliosis [4], and lumbar pain [5], as well as lower bonemineral density [4] and reduction in the ability to perform all desired tasks [6] compared toable-bodied (AB) subjects. Previous studies showed a longer stance phase and higher ankle,knee, and hip joint moments [7] and vertical ground reaction forces (GRFs) [7-9] in the SLcompared with the amputated limb (AL) and AB subjects. These alterations in load dis-tribution between lower limbs increase the risk of injuries such as anterior cruciate ligament

PM&R1934-1482/14/$36.00

Printed in U.S.A.

ª 2014 by the American Academy of Physical Medicine and RehabilitationVol. -, 1-11, - 2014

http://dx.doi.org/10.1016/j.pmrj.2014.01.0191

2 Castro et al WALKING WITH UNILATERAL TRANSFEMORAL AMPUTATION

tears [10] and knee joint osteoarthritis [3,9] in response tooverload. Higher values for the step width [11] anddisplacement of the center of pressure in both limbs of theindividuals with lower limb amputation compared to ABsubjects have also been reported [12,13].

The prosthesis is limited in its mechanical functionalitywhen the dynamic movement changes, resulting in movement-specific compensatory mechanisms in the residual jointsand intact limb [14]. The alterations in gait features pro-moted by the adaptation to a prosthetic limb may causeor reinforce physical impairments. During the alignmentprocess of the prosthesis and gait training in subjects withlower extremity amputation, a prosthetic foot roll-over asclose as possible to the physiological foot [15] and bilateralsymmetry [16] are pursued. This process is highly sub-jective and variable [17], leading to the need for in-struments able to easily and reliably provide quantitativemeasurements of the gait of individuals with limb loss tohelp and improve the rehabilitation process [17].

The analysis of plantar pressures is considered clinicallyuseful in identifying anatomical deformities on the foot, inguiding the diagnosis and treatment of gait disorders, and inpreventing pressure ulcers [18,19]. The plantar pressure peakis the main parameter used in plantar pressure analysis, andreflects the highest pressure to occur in a specific region of thefoot during the stance phase. The instant of the pressure peaksalso can be calculated, allowing recognition of the sequence ofrecruitment of different regions of the foot (loosely referred toin the present study as temporal foot roll-over). However, thevalidity of the plantar pressure analysis is not yet well estab-lished, and its clinical applicability in the case of prostheticfeet is minimal. Geil and Lay [17] investigated the capacity of aplantar pressure analysis system to identify alterations intranstibial amputees’ prosthetic alignment for clinical pur-poses. These authors observed that angular changes in theprosthetic alignment in the frontal plane produce predictableshifts in the plantar pressures between lateral and medial footregions, concluding that plantar foot pressure analysis is asensitive and feasible tool to help clinicians quantify gait pa-rameters and the way in which these parameters are influ-enced by the prosthetic alignment [17].

Regarding the SL, although some studies have evidencedits overload during gait [1,7-9], to the best of our knowledge,the identification of the plantar pressure distribution pattern,temporal foot roll-over features, and specific overloaded re-gions were not yet addressed. This information may helpclinicians to prevent plantar foot injuries such as blisters,callosity, or skin ulcerations by avoiding the development ofregions with high pressure peaks, and may also guide gaittraining programs. The light-weight plantar pressure platescould be a feasible and practical way to record these data.

Information about mechanical stress can be obtainedfrom vertical force analysis [20]. The vertical GRF providesthe global aspect of the vertical forces, whereas the plantarpressure analysis informs about the distribution of this force

along the plantar surface of the foot [21]. The vertical GRF isrelated to joint contact forces and can therefore provide in-sights into the development of some pathological conditions,such as back pain and osteoarthritis [20]. Increased verticalforces represent a decreased capacity of the musculoskeletalsystem in absorbing the body loading during gait [22] and,as a consequence, express an increment on the likelihood ofdeveloping overuse injuries [23]. The anterior-posterior GRFinforms about the friction between the sole and floor, rele-vant for assessing foot-related injuries (eg, blister or ulcera-tion) and tendency to slip [24]. The medial-lateral GRF hasbeen suggested to provide information about gait balance[23]. The GRF are influenced by gait speed, in which theincrease of speed promotes linear increases in the GRF peaks[25] and decreases the GRF impulses [26]. In the presentstudy, the GRF peaks and GRF impulses were calculated.Although the peak variables are widely used and informabout the highest forces experienced by the body, the im-pulse variables provide complementary information relatedto the amount of force received by the body during a givenactivity. Based on the viscoelastic properties of the muscu-loskeletal tissues, the peak variables appear to be moreinsightful in showing aggressive loads to the body. However,in terms of gait pattern and long-term adaptations, the im-pulse variables might also provide valuable informationabout the kinetic features of walking.

The GRFs provide global information about the verticaland shear stress forces, whereas the plantar pressure analysisidentifies the distribution of the vertical GRF over the plantarfoot surface [21]. The combination of both analyses providesmore detailed and complementary information about specificfeatures of forces acting on the prosthetic and the soundlimbs. Such information might be useful in verifying the ef-fects of adjustments in prosthesis components and differenttherapeutic approaches to gait performance. Hence, the aimof this study was to describe and compare plantar pressures,temporal foot roll-over, and GRF parameters between bothlimbs of subjects with unilateral TF amputation and with ABsubjects during walking. The plantar pressure and GRF datawere also analyzed to verify their relevance in determiningdifferences between participants in gait parameters of in-dividuals with unilateral TF amputation walking.

METHODS

This was a cross-sectional study with a convenience sample.This project was approved by the ethical board from theProfessional Rehabilitation Center of Gaia (Arcozelo, Portugal),and all participants freely signed an informed consent basedon the Declaration of Helsinki.

Subjects

Two groups of participants were analyzed. For the experi-mental group, patients with unilateral TF amputation were

Figure 1. Distribution of the plantar pressures, as acquired byFootScan 7 gait 2nd generation, in 1 participant from 1 group.

PM&R Vol. -, Iss. -, 2014 3

selected from the rehabilitation center’s database. Patientswho had been living with unilateral TF amputation for at least2 years were included, whereas those who had a prosthesiswith electronically controlled knee or prosthetic ankle withenergy storing system, and/or who presented with pain orlacked independent walking ability (without the help of somekind of aid device) were excluded. The AB participants wereselected from the project of physical activity for elderly in-dividuals developed at the university engaged in this study.They were excluded if they presented with any musculo-skeletal impairment, limitation, or pain during walking.

Fourteen subjects with TF amputation, 13 male and 1 fe-male, with a mean age of 56.7 � 11.7 years and a mean bodymass of 71.4 � 11.7 kg were enrolled in this study. Allparticipants had undergone the amputation more than 9 yearsbefore the experiment. Twelve amputations were traumatic,1 was from osteomyelitis complications, and 1 was fromvascular disease; in these non-traumatic cases, the individualsdid not show any sign of co-morbidity or difficulty in gaitrelated to the pathology that caused the amputation. Twelveof the participants used a modular prosthesis of endoskeletaltype, and 2 participants used the exoskeletal type. Twelveprostheses had a prosthetic foot with articulated ankle and 2with fixed ankle; all of them had a friction-controlled kneejoint, where 12 were uniaxial and 2 polyaxial. All prostheticsockets were of total contact type without the use of a pelvicsuspension belt. All individuals had concluded the adaptationand prosthesis alignment process, a process completed by thesame 2 technicians in all cases. Participants used their ownshoes (standard classic shoes), which were provided by thesame manufacturer, and they were all of similar type (notorthopedic shoes). Furthermore, the prosthesis was individ-ually aligned for these shoes. They had no special insole. TheAB group was composed of 6 men and 15 women with amean age of 68.3 � 9.4 years and a mean body mass of66.0 � 9.0 kg. They also wore their own shoes, which weretheir preferred ones and were of the same type as the entireAB group (athletic shoes).

As a measure of the capacity and physical independenceof the participants, the questionnaire SF-36 v.2 was appliedand the physical function domain was analyzed [27].

Instruments

We used a pressure plate (RsScan, Olen, Belgium) measuring0.5 � 0.4 m, with a spatial resolution of 2.7 sensors/cm2 oper-ating at 300Hz to capture pressure data. To collect theGRFs, weused a piezoelectric force plate (Kistler Instruments,Winterthur,Switzerland) operating at 1000 Hz. Both plates were calibratedbefore the study, and they were synchronized by an externaltrigger that started them simultaneously.

Data Acquisition

We used the FootScan 7 gait 2nd generation software(RsScan, Olen, Belgium) to acquire the pressure plate data

(Figure 1), and used SIMI Motion System software (SIMIReality Motion Systems, Unterschleissheim, Germany) fordata acquisition of the force plate.

Procedures

The pressure plate was positioned over the force plate, andthey were placed in the middle of an 8-m walkway. Anti-slipmats were placed along the trajectory to avoid unlevelground caused by the pressure plate. The participants werefamiliarized with the environment by freely walking overthe walkway. During data collection, the participants walkedat their self-selected speed. They performed at least 7 steps(3 before and 3 after stepping on the plate). The participantswith amputation performed 6 valid trials, 3 of them re-cording data from the AL and 3 from the SL. The AB par-ticipants performed 3 valid trials with their right leg. Thetrial was considered valid if the participants hit the plate withthe entire foot over it and did not alter their gait pattern.Alterations in the gait pattern such as step length or pacewere assessed subjectively by visual inspection comparingthe gait performed during the familiarization time and thetrials.

Data Analysis and Outcome Measures

Data from the pressure plate (values of each sensor at eachframe) and force plate (3 components of the GRFs) wereexported to Matlab 7.0 (Mathworks, Natick, MA), and aprogram for data processing and variable calculations wasdeveloped. The shoe imprint (hereafter loosely referredto as foot) was divided into 6 regions: in the longitudinal(anterior/posterior) direction, the boundary between rearfootand midfoot was located at 73% of foot length, measuredfrom toes to heel, and that between midfoot and forefoot was

4 Castro et al WALKING WITH UNILATERAL TRANSFEMORAL AMPUTATION

located at 45% along this length [28]. Each region wasfurther divided into 2 parts, 50% medial and 50% lateral.

We calculated 2 plantar pressure and 10 global GRF-dependent variables. The plantar pressure variables werecalculated for all 6 foot regions and were as follows: thepressure peak (unit, N/cm2), which was defined as the peak ofthe average of all active sensors in the region; and temporalfoot roll-over, the instant of pressure peak (unit, percentage ofthe stance phase), which was defined as the moment ofoccurrence of the pressure peak. The GRF variables were asfollows (Figure 2):

� Stance phase duration—time from the first contact to toeoff;

� Load acceptance peak (VLA-Peak) —first peak (highestvalue at the first half of the curve) from the vertical GRF;

� Thrust peak (VT-Peak)—second peak (highest value at thesecond half of the curve) from the vertical GRF;

� Braking peak (APB-Peak)—lowest value (negative peak)from the anterior-posterior GRF;

� Propulsive peak (APP-Peak)—highest value (positive peak)from the anterior-posterior GRF;

� Medial-lateral peak (MLPeak)—the highest value from themedial-lateral GRF;

� Load acceptance impulse (VLA-Imp)—time integral fromthe first foot contact until midstance (minimum betweenVLA-Peak and VT-Peak) from the vertical GRF;

Figure 2. Ground reaction force dependent variables.VLA-Peak ¼ load acceptance peak; VT-Peak ¼ thrust peak;APB-Peak ¼ braking peak; APP-Peak ¼ propulsive peak; MLPeak ¼medial-lateral peak; VLA-Imp ¼ load acceptance impulse;VT-Imp ¼ thrust impulse; APB-Imp ¼ braking impulse; APP-Imp ¼propulsive impulse.

� Thrust impulse (VT-Imp)—time integral from midstance totoe-off from the vertical GRF;

� Braking impulse (APB-Imp)—time integral from the firstcontact to middle 0 from the anterior-posterior GRF; and

� Propulsive impulse (APP-Imp)—time integral from themiddle zero to toe-off from the anterior-posterior GRF.

The variables were calculated for the 3 groups understudy. The GRFs were normalized by body weight (ie, theweight of the prosthesis was added to the body weight). Inthe present study, an asymmetric pattern was consideredpresent when statistically significant differences were foundbetween AL and SL in any variable [29].

Statistical Analysis

The mean of the 3 repetitions of each group (AL, SL, andAB groups) was computed and statistical procedures wereperformed with these values. Data normality was verified bythe Shapiro-Wilk test and homogeneity of variances by theLevene’s test. The between-trial repeatability of all variableswas verified by the intraclass correlation coefficient (ICC).To compare the variables among the groups, paired Studentt tests (AL versus SL) and independent Student t tests (SLversus AB and AL versus AB) were used. As there was only1 woman in the experimental group, and because in thecontrol group there were more females than males, we alsoused independent Student t tests to compare male andfemale participants in the control group. Statistical pro-cedures were performed using SPSS software (SPSS Inc,Chicago, IL) with an a value set at 0.05.

RESULTS

All data showed normality of distribution and homogeneityof variance. The ICC values were above 0.90 for all variables,indicating excellent intertrial repeatability. The AL showedshorter stance time compared to SL (AL ¼ 0.88 � 0.12seconds versus SL ¼ 1.02 � 0.12 seconds; P ¼ .003) andhigher stance time than AB (0.73 � 0.10 seconds; P ¼ .001),and the AB participants also showed shorter stance time thanthe SL (P < .001). The values of the questionnaire SF-36were 62.8 � 24.9 points for the subjects with lower limbamputation and 82.3 � 18 points for the AB participants.There was no statistically significant difference betweenmales and females in the control group for any parameter(P < .05) (Appendix 1).

The pressure peaks showed statistically significant dif-ferences among the 3 groups (AL, SL, and AB). The ALshowed higher pressure peaks on the lateral rearfoot andlateral midfoot compared with the AB group, and higherpressure peaks on the medial midfoot compared with the ABand SL groups. No differences among groups were found onthe medial rearfoot. The AL presented with lower pressurepeaks than the AB on the medial forefoot and also lowervalues than AB and SL on the lateral forefoot (Figure 3A).

PM&R Vol. -, Iss. -, 2014 5

The temporal foot roll-over also showed statistically sig-nificant differences among the 3 groups (Figure 3B). In bothrearfoot regions, the instants of pressure peaks occurredearlier in both AL and SL compared to the AB. In the medialmidfoot, the instants of pressure peaks occurred statisticallysignificantly later in the AL compared to the SL and AB, andin the lateral midfoot they occurred earlier in the SL com-pared to the AL and AB. In the AL group, the instants ofpressure peaks occurred earlier in the lateral and medialforefoot compared to the AB.

The 3 GRF components were statistically significantlydifferent among the groups. The AL showed lower VT-Peak

and VLA-imp, and lower braking (APB-Peak and APB-Imp) andpropulsive (APP-Peak and APP-Imp) forces than the SL and ABgroups. For the MLPeak, the AL showed higher values thanthe SL and AB groups. The SL presented higher VT-Imp andAPB-Imp and lower VT-Peak, APB-Peak, and VLA-imp than did theAB participants (Table 1 and Figure 4).

Figure 3. (A) Plantar pressure peaks and (B) Temporal foot roll-over—instant of the plantar pressure peaks for the 3 groups.Equal symbols in the groups represent statistically significantdifferences between *Amputated limb and Sound limb,yAmputated limb and Able-bodied, and between. #Sound limband Able-bodied subjects (P < .05).

DISCUSSION

This study was conducted to describe and compare plantarpressure and GRF parameters between both limbs of in-dividuals with TF amputation (AL and SL) and with AB sub-jects during gait. As an indicator of physical function, theSF-36 Health Survey questionnaire showed that the subjectswith lower limb amputation have lower values than their able-bodied counterparts. Nevertheless, these data indicate thatthese subjects experienced few limitations to performing ac-tivities of daily living involving physical function, and thattherefore the gaits of both groups of participants appearssuitable for comparison. We observed an asymmetrical patternin pressure peaks, temporal foot roll-over, and GRF variableswhen the participants with TF amputation walked. Further-more, statistically significant differences in both SL and ALcompared to the AB participants were also identified. Bothplantar pressure and GRF analyses were able to discriminatedifferences in gait pattern among the groups. To the best of ourknowledge, no studies presenting values of plantar pressureson subjects with TF amputation are available in the literature tocompare with our results. Nevertheless, other studies analyzingthe center of pressure [12,13], GRFs [7,8,30], lower limb ki-nematics [30], and joint moments [7] also found an asym-metrical pattern when subjects with TF amputation walked.However, it is not clear whether these asymmetries arenecessary (in both magnitude and pattern) to promote safetyduring walking and to compensate for the difficulties inherentto walking with a prosthetic limb [31], or whether they shouldbe reduced to promote a kinetic gait pattern as symmetric aspossible to decrease the injury risks caused by overloading orunloading [9].

During shod walking at a medium speed, healthy olderadults showed plantar pressure peaks ranging from 10 N/cm2 (medial midfoot) to 23 N/cm2 (medial forefoot) and 25N/cm2 (central forefoot) [32]. In the current study, theplantar pressure peaks in the AB group ranged from 8 N/cm2

(medial midfoot) to 18 N/cm2 (lateral rearfoot), in the ALfrom 10 N/cm2 (lateral forefoot) to 25 N/cm2 (lateralrearfoot), and in the SL from 13 N/cm2 (medial and lateralmidfoot) to 18 N/cm2 (lateral rearfoot). Therefore, evenwith different plantar pressure systems—in-shoe [32]versus pressure plate (present study)—similar magnitudesof plantar pressures were observed. This increases theexternal reliability of our findings.

A set of studied parameters provided informationregarding the beginning of the stance phase: pressure plate(medial and lateral rearfoot pressure peaks and their instantsof occurrence) and force plate (VLA-Peak, VLA-Imp, APB-Peak,and APB-Imp). The present findings suggest that both in-struments were able to identify differences between thesubjects with amputation and AB participants while theywalked. Subjects with TF amputation did not show a higherload acceptance peak (VLA-Peak) in the SL compared to theprosthetic one, but they did present with higher vertical GRF

Table 1. Between-group comparison of ground reaction force variables

Variable Group Mean SD Comparison P Lower CL Upper CL

VLA-Peak (%BW) AL 101.6 5.7 AL vs AB .994 �5.2 5.2SL 103.7 4.8 SL vs AB .396 �2.9 7.0AB 101.6 7.1 AL vs SL .239 �5.7 1.6

VT-Peak (%BW) AL 97.9 5.3 AL vs AB* <.001 �19.2 �7.1SL 104.4 7.2 SL vs AB* .042 �13.3 �0.3AB 111.1 9.0 AL vs SL* .013 �11.2 �1.7

APB-Peak (%BW) AL �7.12 2.3 AL vs AB* <.001 6.0 11.1SL �12.3 3.4 SL vs AB* .023 0.5 6.2AB �15.6 3.8 AL vs SL* .003 2.1 8.2

APP-Peak (%BW) AL 7.4 2.8 AL vs AB* <.001 �12.1 �6.3SL 16.4 1.8 SL vs AB .881 �3.1 2.7AB 16.6 4.1 AL vs SL* <.001 �12.0 �6.5

MLPeak (%BW) AL 8.17 2.1 AL vs AB* <.001 1.6 4.1SL 7.5 2.2 SL vs AB* .002 0.9 3.4AB 5.3 1.0 AL vs SL .283 �0.7 2.1

VLA-Imp (%BW $ s) AL 27.3 3.7 AL vs AB* <.001 �11.8 -6.6SL 32.4 2.1 SL vs AB* <.001 �6.2 �2.0AB 36.5 3.0 AL vs SL* <.001 �7.0 �3.2

VT-Imp (%BW $ s) AL 33.8 7.3 AL vs AB* .011 4.4 16.3SL 47.9 10.1 SL vs AB* <.001 17.6 31.4AB 23.4 7.8 AL vs SL* <.001 �19.8 �8.5

APB-Imp (%BW $ s) AL 2.02 1.1 AL vs AB* .016 �1.6 �0.2SL 4.4 1.0 SL vs AB* <.001 0.8 2.2AB 2.9 0.7 AL vs SL* <.001 �3.2 �1.5

APP-Imp (%BW $ s) AL 1.5 0.6 AL vs AB* <.001 �2.1 �1.2SL 3.1 1.1 SL vs AB* .002 �2.6 �0.7AB 3.1 0.6 AL vs SL .973 �0.7 0.7

VLA-Peak ¼ load acceptance peak; VT-Peak ¼ thrust peak; APB-Peak ¼ braking peak; APP-Peak ¼ propulsive peak; MLPeak ¼ medial-lateral peak; VLA-Imp ¼ loadacceptance impulse; VT-Imp ¼ thrust impulse; APB-Imp ¼ braking impulse; APP-Imp ¼ propulsive impulse; BW ¼ body weight; AL ¼ amputated limb; SL ¼ soundlimb; AB ¼ able-bodied; SD ¼ standard deviation; lower CL ¼ lower bound of 95% confidence limit; upper CL ¼ upper bound of 95% confidence limit.*Significant difference between groups with P < .0.5.

6 Castro et al WALKING WITH UNILATERAL TRANSFEMORAL AMPUTATION

impulse (VLA-Imp). Thus, as the impulse variables are influ-enced by the magnitude and duration of force application,the longer time spent on the sound limb (compared with theprosthetic limb) is likely to be the main reason for thisbehavior, which might play an important role in the higherlevels of knee pain and knee osteoarthritis observed in the SLsubjects [1-3].

Changes in the temporal foot roll-over in both lower limbsof subjects with unilateral TF amputation compared to subjectswithout any amputation and an altered plantar pressure dis-tribution between the medial and lateral rearfoot region in theAL were identified by plantar pressure analysis. Such eventscould not be identified by the GRF analysis, as it assesses theoverall GRFs and ground reaction moments, and thereforedoes not provide any information about how the forces arebeing distributed among different plantar foot regions.

The asymmetrical plantar pressure distribution and thechanges in the temporal foot roll-over in the AL might be aconsequence of the absence of physiological knee and anklejoints, which causes altered motor control of the prostheticlimb [33]. Michaud et al [34] found an increased pelvicobliquity during the prosthetic swing in individuals withunilateral lower limb amputation, suggesting hip hiking. As aresult, earlier instants of pressure peaks at the beginning ofthe stance phase might occur because of the prosthetic foot’s

abrupt landing. Hip hiking was also evidenced by the lowerAPB-Peak and APB-Imp observed in the AL compared to the SLand AB. It is possible that the increased motion in the frontalplane (hip hiking) resulted in a more vertical landing withthe prosthetic foot, reflected in lower anterior-posteriorforces (braking forces). This also indicates difficulties indecelerating the prosthetic limb at the beginning of thestance phase. The possible slower gait speed of the partici-pants with TF amputation may also have contributed tothese lower braking peaks in the prosthetic limb. Sapin et al[30], corroborating our results, observed a lower capacity forbraking in subjects with TF amputation who used a single-axis prosthetic knee (which coordinates ankle and kneeflexion), and in those using other knee joints without a knee-ankle link. Considering the SL, we observed an earlier oc-currence of the pressure peaks in the rearfoot and midfootof the SL compared to the AB. This may reflect a strategyto compensate for the lower stability in the late stance withthe prosthetic limb. Thus the stance with the SL was antic-ipated, and the maximal weight bearing in the rearfoot andmidfoot also occurred earlier.

At the midstance phase, in which the medial and lateralmidfoot pressure peaks occur, higher values of pressurepeaks and later instants of their occurrence were found forboth regions in the AL. These higher magnitudes of pressure

Figure 4. Ground reaction forces (GRF) of all tests and all participants. (A) Anterior-posterior GRF; (B) medial-lateral GRF; and (C)vertical GRF. Equal symbols in the groups represent statistically significant difference between *Amputated limb and Sound limb,yAmputated limb and Able-bodied and between #Sound limb and Able-bodied subjects (P < .05). As all the MLPeak were in the samedirection (positive values), and only some participants showed negative values at the edge of the curve (first or last 20% of the stancephase), the figure graphically presents only absolute positive values of the medial-lateral GRF.

PM&R Vol. -, Iss. -, 2014 7

might be caused by the need for the subjects with TFamputation to block their prosthetic knee in extension [35].Moreover, at the midstance phase, the prosthetic knee tendsto be in extension as the line of action of the GRFs passesover or anterior to the center of the knee joint [31]. Pressurepeaks for the AL were postponed in the midfoot regions,possibly to spend more time in this phase, as subjectsconsider it the most stable phase of stance.

Considering the MLPeak, both AL and SL showed highervalues compared to AB participants. As subjects with TFamputation showed a higher vertical displacement of thecenter of mass to block the knee in extension [8,35], onemay expect a larger body movement on the frontal plane[34] to smooth out this greater center of mass verticaldisplacement, resulting in higher values of MLPeak, as foundin the present study. The significantly increased load in the

SL [8,35], the hip-hiking pattern, and the lower control ofthe AL for lowering the pelvis on the SL side before the initialfoot contact might also contribute to these higher magni-tudes of MLPeak [36], which might reflect gait instability.

We found, at the late stance, the most different behaviorsamong groups in pressure peaks (medial and lateral forefootpeaks), temporal foot roll-over, and GRFs (VT-Peak, VT-Imp,APP-Peak and APP-Imp). The AL displayed lower values inalmost all of these variables, and earlier instants of pressurepeaks, compared to the SL and AB groups. Similar resultswere found in patients using a prosthetic limb, with orwithout a knee-ankle link [30]. The plantar flexor musclesare responsible for plantar flexion during the gait propulsionphase, and their absence implies the recruitment of the hipflexors to lift the foot, anticipating the AL swing phase [33].Therefore, the lower values that we observed might be

8 Castro et al WALKING WITH UNILATERAL TRANSFEMORAL AMPUTATION

caused by the absence of these muscles [37]. Moreover, allparticipants used a Solid-Ankle Cushioned Heel (SACH)foot. This foot includes a solid ankle and a rigid keel thatruns along the length of the prosthetic foot until the end ofthe midfoot, with most of the forefoot composed mainly offlexible, rubber-type material that does not provide anysupport or energy return.

The higher GRF impulses, pressure peaks in the medialforefoot, and the longer stance phase duration in the subjectswith TF amputation suggest a significantly greater load in theSL during walking. Nolan et al [8] analyzed 4 subjects withTF amputation and found higher vertical GRF impulses inthe SL compared to the AL and AB groups, whereas a longerstance and a shorter swing phase in the SL were observed inprevious studies [8,13,35]. The subjects with TF amputationraise their body by excessive plantar flexion of the soundfoot in a well-known compensatory mechanism, that is,vaulting. This behavior could occur as an adaptive mecha-nism to increase the foot clearance of the prosthetic foot andto protect the residual limb (AL) by charging it with load forless time [8]. Because mechanical forces are related to jointdamage and dysfunctions such as those found in osteoar-thritis [3], this “protective pattern” in subjects with unilateralTF amputation might be 1 of the causes of the high inci-dence of dysfunction in their SL [3].

Pressure plates allow the collection of information relatedto the interface between shoe sole and ground, whereas in-shoe pressure systems capture data related to the plantarfoot surface interface and the internal surface of the shoe.We chose a pressure plate in this study for its convenienceand ease of use for clinical purposes. Moreover, the effectpromoted by the deformable interface of the shoe couldexert major influence on pressure insole measurements [38].The insole sensors could be affected by the internal envi-ronment of the shoe, such as temperature, contour, andhumidity [39]. Furthermore, the presence of cables attachedto the individuals, necessary for in-shoe pressure systems,could prove to be another conditioning factor affecting thegait pattern.

The present study showed that both plantar pressures andGRFs are sensitive to detecting alterations in gait patternswhen subjects with TF amputation walked. However, usingboth instruments in a clinical context might be complex andexpensive. As using a pressure plate does not require anelaborate laboratory set up, it is less expensive and morepractical than using a force plate (ie, pressure plates aresmaller and lighter than force plates), its use for analyzingdifferent aspects of gait training and prosthesis alignment issuggested. Moreover, data regarding AB subjects can be usedto guide the interpretation of data from patients with lowerlimb amputation, providing some indications about theplantar foot regions that should be more or less used by ABsubjects while walking. However, this interpretation shouldbe made with caution, and data from AB subjects are notsuggested to be a template that should be matched.

Angular changes between the pylon and the socket in thefrontal plane during prosthetic alignment shift the plantarpressure in a lateral/medial direction [17]. This informationcould be used during prosthesis alignment to identify animbalance of pressures between medial and lateral plantarfoot regions. After that step, alterations in prosthesis align-ment could be tested and the clinician could use data fromAB subjects, gait symmetry, or any other criteria in agree-ment with the aim of intervention to improve the gaitpattern. Future studies on the influence of specific musclestrengthening, special insoles or shoes, gait training, andprosthesis components on pressure distribution wouldprovide a vast amount of quantitative information for reha-bilitation purposes. Such findings could be used to verify themost appropriate approach to diminish gait asymmetry to anoptimum level.

Study Limitations

The adopted walking speed in this study was the one atwhich the subjects felt most comfortable (self-selected).Thus, an identical gait speed for the subjects with TFamputation and AB participants was not warranted. Basedon the differences in duration of the stance phase betweenthe experimental and control groups, it is plausible to as-sume the participants with TF amputation walked moreslowly than their able-bodied counterparts. A previous study[40] assessing AB subjects observed similar plantar pressurepeaks between slow and normal self-selected gait speeds. Inthe mentioned study [40], the duration of the stance phasefor slow gait speed was 0.82 � 0.11 seconds and for normalgait speed was 0.70 � 0.05 seconds. The differences in theduration of stance phase between groups in our study wereonly slightly higher (AB ¼ 0.73 � 0.10, AL ¼ 0.88 � 0.12and SL 1.02 � 0.12) than in the aforementioned study [40].Therefore, we believe that the possible differences in gaitspeed between groups had only small effects on the plantarpressure parameters. On the other hand, the GRF peaksincrease linearly with increasing gait speed [25], and theGRF impulses decrease with increasing gait speed [26].Thus, the comparisons between the GRF parameters fromthe AL and SL with those from the AB should be made withcaution. Our subjective analysis of gait speed during datacollection was that the participants with TF amputationwalked slightly more slowly than the AB participants. Weadopted the self-selected gait speed to prevent disturbancesin gait patterns, caused by restricting speed with treadmillsor using a metronome, and to ensure normal walking.

Other limitations should also be considered in the presentstudy. The distribution between men and women among theparticipants was not homogenous. However, we did notobserve any difference between able-bodied male and femalesubjects for any parameter. Putti et al [41] also reported thatthere were no differences in plantar pressure parametersobserved between genders. Only the right lower limbs of the

PM&R Vol. -, Iss. -, 2014 9

AB participants were assessed; however, similar GRFs andplantar pressures have been shown between limbs [42,43].The shoe types were different between groups (TF amputeesand AB participants). Using new shoes or ones different fromthose usually worn could alter the prosthesis alignment andgait pattern. To preserve a situation as real as possible whilecausing the least interference in gait patterns, the participantswere asked to use their own shoes, which were of similartype inside the groups. Finally, the clinical usefulness of theanalyses used in the present study—GRFs and, mainly,plantar pressures—as tools to guide and follow the reha-bilitation process must be interpreted with caution, as therewas no follow-up to verify the effects of intervention (ie, gaittraining or changes in prosthesis alignment) using theseinstruments.

CONCLUSION

In this study, the participants with unilateral TF amputationshowed an asymmetrical plantar pressure distribution andGRF patterns. In terms of plantar pressure distribution,temporal foot roll-over, and GRFs, both lower limbs fromthe participants with TF amputation (AL and SL) weredifferent from those of the AB subjects. The GRFs suggest asignificantly increased load in the SL, whereas the plantarpressure analysis suggests a higher recruitment of the lateralrearfoot and medial and lateral midfoot regions, and a lowerrecruitment of the medial and lateral forefoot on the ALcompared to the SL and AB groups. Both systems seemed tobe able to discriminate between the lower limbs of in-dividuals with unilateral TF amputation. The pressure plateseems to be a useful and sensitive instrument for gait eval-uation in subjects with limb loss.

REFERENCES1. Lemaire ED, Fisher FR. Osteoarthritis and elderly amputee gait. Arch

Phys Med Rehabil 1994;75:1094-1099.2. Pieter AS, Caroline MvH, Minou WH, Rob JS. The prevalence of

osteoarthritis of the intact hip and knee among traumatic leg amputees.Arch Phys Med Rehabil 2009;90:440-446.

3. Morgenroth DC, Gellhorn AC, Suri P. Osteoarthritis in the disabledpopulation: A mechanical perspective. PM R 2012;4:S20-S27.

4. Burke MJ, Roman V, Wright V. Bone and joint changes in lower limbamputees. Ann Rheum Dis 1978;37:252-254.

5. Skinner H, Effeney D. Gait analysis in amputees. Am J Phys Med 1985;64:82-89.

6. Christensen B, Ellegaard B, Bretler U, Ostrup EL. The effect of pros-thetic rehabilitation in lower limb amputees. Prosthet Orthot Int 1995;19:46-52.

7. Nolan L, Lees A. The functional demands on the intact limb duringwalking for active trans-femoral and trans-tibial amputees. ProsthetOrthot Int 2000;24:117-125.

8. Nolan L, Wit A, Dudzinski K, Lees A, Lake M, Wychowanski M. Ad-justments in gait symmetry with walking speed in trans-femoral andtrans-tibial amputees. Gait Posture 2003;17:142-151.

9. Royer T, Koenig M. Joint loading and bone mineral density in per-sons with unilateral, trans-tibial amputation. Clin Biomech 2005;20:1119-1125.

10. Griffin LY, Agel J, Albohm MJ, et al. Noncontact anterior cruciate lig-ament injuries: Risk factors and prevention strategies. J Am AcadOrthop Surg 2000;8:141-150.

11. Mensch G, Ellis PE. Running patterns of transfemoral amputees: Aclinical analysis. Prosthet Orthot Int 1986;10:129-134.

12. Castro MP, Soares D, Mendes E, Machado L. Center of pressure analysisduring gait of elderly adult transfemoral amputees. J Prosthet Orthot2013;25:68-75.

13. Schmid M, Beltrami G, Zambarbieri D, Verni G. Centre of pressuredisplacements in trans-femoral amputees during gait. Gait Posture2005;21:255-262.

14. Schoeman M, Diss CE, Strike SC. Kinetic and kinematic compensationsin amputee vertical jumping. J Appl Biomech 2012;28:438-447.

15. Hansen AH, Childress DS, Knox EH. Prosthetic foot roll-over shapeswith implications for alignment of trans-tibial prostheses. ProsthetOrthot Int 2000;24:205-215.

16. Petersen AO, Comins J, Alkjær T. Assessment of gait symmetry intransfemoral amputees using C-Leg compared with 3R60 prostheticknees. J Prosthet Orthot 2010;22:106-112.

17. Geil MD, Lay A. Plantar foot pressure responses to changes duringdynamic trans-tibial prosthetic alignment in a clinical setting. ProsthetOrthot Int 2004;28:105-114.

18. Rodgers M. Dynamic foot biomechanics. J Orthop Sports Phys Ther1995;21:306-316.

19. Hessert MJ, Vyas M, Leach J, Hu K, Lipsitz LA, Novak V. Foot pressuredistribution during walking in young and old adults. BMC Geriatr2005;5:8.

20. Piscoya JL, Fermor B, Kraus VB, Stabler TV, Guilak F. The influence ofmechanical compression on the induction of osteoarthritis-relatedbiomarkers in articular cartilage explants. Osteoarthritis Cartilage 2005;13:1092-1099.

21. Castro M, Abreu S, Sousa H, Machado L, Santos R, Vilas-Boas JP.Ground reaction forces and plantar pressure distribution during oc-casional loaded gait. Appl Ergon 2013;44:503-509.

22. Simpson KM, Munro BJ, Steele JR. Effects of prolonged load car-riage on ground reaction forces, lower limb kinematics and spatio-temporal parameters in female recreational hikers. Ergonomics2012;55:316-326.

23. Birrell SA, Hooper RH, Haslam RA. The effect of military load carriageon ground reaction forces. Gait Posture 2007;26:611-614.

24. Chang W-R, Chang C-C, Matz S. The effect of transverse shear force onthe required coefficient of friction for level walking. Hum Factors 2011;53:461-473.

25. Perry J. Stride analysis. In: Perry J, ed. Gait Analysis: Normal andPathological Function. Thorofare, NJ: SLACK; 1992, 431-441.

26. Jordan K, John HC, Karl MN. Walking speed influences on gait cyclevariability. Gait Posture 2007;26:128-134.

27. Ware JE, Gandek B. Overview of the SF-36 Health Survey and theInternational Quality of Life Assessment (IQOLA) project. J Clin Epi-demiol 1998;51:903-912.

28. Gurney JK, Kersting UG, Rosenbaum D. Between-day reliability ofrepeated plantar pressure distribution measurements in a normalpopulation. Gait Posture 2008;27:706-709.

29. Sadeghi H, Allard P, Prince F, Labelle H. Symmetry and limb domi-nance in able-bodied gait: A review. Gait Posture 2000;12:34-45.

30. Sapin E, Goujon H, de Almeida F, Fodé P, Lavaste F. Functional gaitanalysis of trans-femoral amputees using two different single-axisprosthetic knees with hydraulic swing-phase control: Kinematic andkinetic comparison of two prosthetic knees. Prosthet Orthot Int 2008;32:201-218.

31. Murray M, Sepic S, Gardner G, Mollinger L. Gait patterns of above-knee amputees using constant-friction knee components. Bull ProsthetRes 1980;17:25-45.

32. Burnfield JM, Few CD, Mohamed OS, Perry J. The influence of walkingspeed and footwear on plantar pressures in older adults. Clin Biomech2004;19:78-84.

10 Castro et al WALKING WITH UNILATERAL TRANSFEMORAL AMPUTATION

33. Rietman JS, Postema K, Geertzen JHB. Gait analysis in prosthetics:Opinions, ideas and conclusions. Prosthet Orthot Int 2002;26:50-57.

34. Michaud SB, Gard SA, Childress DS. A preliminary investigationof pelvic obliquity patterns during gait in persons with trans-tibial and transfemoral amputation. J Rehabil Res Dev 2000;37:1-10.

35. Segal AD, Orendurff MS, Klute GK, et al. Kinematic and kineticcomparisons of transfemoral amputee gait using C-Leg and Mauch SNSprosthetic knees. J Rehabil Res Dev 2006;43:857-870.

36. Goujon-Pillet H, Sapin E, Fode P, Lavaste F. Three-dimensional mo-tions of trunk and pelvis during transfemoral amputee gait. Arch PhysMed Rehabil 2008;89:87-94.

37. Winter DA. Human balance and posture control during standing andwalking. Gait Posture 1995;3:193-214.

38. Chesnin KJ, Selby-Silverstein L, Besser MP. Comparison of an in-shoepressure measurement device to a force plate: Concurrent validity ofcenter of pressure measurements. Gait Posture 2000;12:128-133.

39. Cavanagh PR, Hewitt FG Jr, Perry JE. In-shoe plantar pressure mea-surement: A review. Foot 1992;2:185-194.

40. Taylor AJ, Menz HB, Keenan A-M. The influence of walking speed onplantar pressure measurements using the two-step gait initiation pro-tocol. Foot 2004;14:49-55.

41. Putti AB, Arnold GP, Abboud RJ. Foot pressure differences in men andwomen. J Foot Ankle Surg 2010;16:21-24.

42. Seeley MK, Umberger BR, Shapiro R. A test of the functional asymmetryhypothesis in walking. Gait Posture 2008;28:24-28.

43. VanZant RS, McPoil TG, Cornwall MW. Symmetry of plantar pressuresand vertical forces in healthy subjects during walking. J Am PodiatrMed Assoc 2001;91:337-342.

PM&R Vol. -, Iss. -, 2014 10.e1

APPENDIX 1

Table A. Comparisons between able-bodied male and female subjects for all studied parameters

Variable (unit)

Male Subjects Female Subjects

t Value df P valueMean SD n Mean SD n

Ground Reaction Forces (Peaks - %BW;Impulse - % BW.s)

VLA-Peak 101.654 4.134 6 104.132 11.781 15 �0.404 19 .692VT-Peak 108.474 8.212 6 111.975 9.424 15 �0.661 19 .519APB-Peak 18.708 4.017 6 15.861 4.097 15 1.208 19 .247APP-Peak �17.971 3.123 6 �14.842 3.777 15 �1.486 19 .159MLPeak 5.427 0.219 6 5.269 1.168 15 0.263 19 .796VLA-Imp 35.501 1.272 6 36.868 3.338 15 �0.785 19 .446VT-Imp 27.101 4.379 6 22.196 8.377 15 1.104 19 .288APB-Imp 3.409 0.516 6 2.745 0.745 15 1.638 19 .124APP-Imp 3.618 0.535 6 2.916 0.591 15 2.097 19 .055

Pressure Peak (N/cm2)FF_med 14.483 2.829 6 12.865 2.188 14 1.391 18 .181FF_lat 13.296 3.871 6 13.967 3.600 15 �0.378 19 .709MF_med 9.851 3.231 6 7.512 2.922 14 1.592 18 .129MF_lat 12.831 5.935 6 11.067 4.417 15 0.751 19 .462RF_med 18.089 3.854 6 16.831 4.592 15 0.591 19 .562RF_lat 21.619 5.042 5 18.073 5.708 15 1.234 18 .233

Instant of Pressure Peak (% of stance phase)FF_med 81.065 2.457 6 79.649 2.575 14 1.883 18 .065FF_lat 77.618 4.702 6 78.389 4.429 15 �0.591 19 .557MF_med 49.699 19.819 6 53.865 13.649 14 �0.917 18 .363MF_lat 47.547 15.453 6 40.800 12.899 15 1.702 19 .094RF_med 19.933 5.554 6 19.253 7.038 15 0.352 19 .726RF_lat 20.449 8.851 5 19.424 8.626 15 0.408 18 .685

VLA-Peak ¼ load acceptance peak; VT-Peak ¼ thrust peak; APB-Peak ¼ braking peak; APP-Peak ¼ propulsive peak; MLPeak ¼ medial-lateral peak; VLA-Imp ¼ loadacceptance impulse; VT-Imp ¼ thrust impulse; APB-Imp ¼ braking impulse; APP-Imp ¼ propulsive impulse; FF ¼ forefoot; MF ¼ midfoot; RF ¼ rearfoot; med ¼ medial;lat ¼ lateral; BW ¼ body weight; SD ¼ standard deviation.