Gait modification strategies for altering medial knee joint load: A systematic review




To evaluate the effect of gait modification strategies on the external knee adduction moment (KAM), a marker of medial knee joint load; determine potentially adverse effects; assess the methodologic quality; and identify areas of future research.


Five electronic databases were searched. Studies evaluating the effects of gait modifications on the KAM in either healthy individuals or those with knee osteoarthritis (OA) were included. Methodologic quality was evaluated by 2 reviewers using the Downs and Black checklist.


Twenty-four studies met the inclusion criteria, exploring 14 different gait modifications of varying sample sizes, age groups, and OA classifications. Contralateral cane use, increased step width, medial knee thrust, increased hip internal rotation, weight transfer to the medial foot, and increased lateral trunk lean demonstrated KAM reductions. Tai Chi gait, ipsilateral cane use, Nordic walking poles, and increased knee flexion exhibited increases in the KAM, demonstrating a potential detriment to their use. The effects of reduced stride length, as well as increases and reductions in either toe-out or gait speed, were inconsistent across the studies and gait cycle.


This review demonstrates that some gait modifications have the ability to alter knee load. Future research is required to determine the magnitude of modification required to maximize beneficial effects, the best method of training, long-term patient adherence, and if these biomechanical changes can translate into clinically relevant changes in symptoms or disease progression risk.


Osteoarthritis (OA) is a common chronic joint disease (1) frequently occurring at the medial tibiofemoral joint (2). As there is no cure for OA (3), management essentially involves alleviating the symptoms. Clinical guidelines stress the importance of conservative nonpharmacologic management (4), as drug therapies are often associated with adverse side effects (5) and only a limited number of surgical interventions are available prior to end-stage disease. Interventions that aim to slow OA progression are required to ease disease burden (6). Since increased dynamic knee joint load, particularly in walking, is a contributing factor to knee OA progression (7), there has been recent interest in identifying load-modifying interventions for this population. However, few interventions have been found to reduce knee load. Gait modification is a frequently used conservative strategy in the clinic that offers promise in managing knee OA. Teaching a patient with medial knee OA to modify their walking may be beneficial in reducing dynamic medial knee load, although it is presently unclear which gait modifications are most likely to be successful.

Medial compartment knee load is commonly evaluated indirectly using 3-dimensional gait analysis, as direct measurement is invasive. The parameter of most relevance to medial knee OA is the external knee adduction moment (KAM). This moment, which acts to force the tibia into varus, has been validated as a reliable indicator of medial knee load (8, 9). Typically, there are two peaks in the KAM during the stance phase of gait. The initial and generally largest peak occurs during the load acceptance phase of early stance, whereas the second peak occurs during the propulsion phase of late stance. Studies evaluating gait have demonstrated that an elevated KAM typically occurs in people with knee OA (10, 11). Hence, the peak KAM is a strong predictor of medial compartment OA presence, radiographic disease severity (12), rate of progression (7), and the presence of OA symptoms (13). Although the late-stance peak KAM has also been shown to be higher in knee OA (14), the clinical significance of this finding remains unknown at present.

Gait modification represents a simple and inexpensive treatment strategy that may be employed by a range of health professionals to reduce medial knee load. As the KAM magnitude is largely determined by the product of the ground reaction force and frontal plane lever arm length (perpendicular distance from the knee joint center to the ground reaction force) (15), gait modifications that alter either of these factors may influence the KAM. Some of the gait modifications reported to benefit knee OA include increases in toe-out angle and lateral trunk lean, reductions in walking speed, and the use of gait aids. Cross-sectional studies lend credence to the use of these strategies. The KAM has been related to several gait biomechanical variables, including toe-out angle (the angle between the foot-long axis and the line of progression) (16, 17), the mediolateral trunk position (18), body segment acceleration (19), and body mass (20). Therefore, modification of the abovementioned variables has the potential to alter the KAM magnitude. Although a range of modifications offer prospection for reducing medial knee load, the efficacy of gait modification for knee OA remains unknown.

In order to evaluate the efficacy of gait modifications for reducing the KAM, we conducted a systematic review of all of the biomechanical studies evaluating gait modification in people with knee OA and healthy asymptomatic individuals. The purposes of this systematic review were to: 1) determine the effect of different gait modifications on the early- and late-stance peak KAM, 2) determine adverse effects and biomechanical effects at other lower extremity joints with implementation of gait modifications, 3) assess the methodologic quality of studies, and 4) identify areas of future research.


Inclusion and exclusion criteria.

As few randomized controlled trials (RCTs) were anticipated, any study design that evaluated the effect of any gait modification on the KAM was eligible for inclusion. Between- and within-subject studies were eligible. Gait modifications were defined as any researcher-initiated alteration of natural gait or the provision of a gait aid. Studies were included if they measured the KAM using 3-dimensional motion analysis during both natural and modified gait conditions. Due to limited research expected to meet the selection criteria, studies involving healthy asymptomatic participants and/or people with knee OA were included. Interventions relating to orthotic management (such as braces, shoes, and orthotics) were excluded because they have previously been subjected to systematic review (21, 22). Dissertations, conference proceedings, studies reported only in abstract form, and studies in non-English languages were excluded.

Search strategy.

We searched the following electronic databases: Medline (from 1950), CINAHL (1981), ISI Web of Science (1900), Expanded Academic ASAP (1980), and Cochrane Central Register of Controlled Trials (1966). Databases were searched until August 2010, week 3. The search strategy used with database-specific truncation terms was as follows: (train* OR retrain* OR interven* OR modif* OR strateg* OR patter* OR rehab*) OR (aid* OR pole* OR cane* OR stick* OR crutch* OR frame* OR walker* OR rollator*) AND (gait OR walk* OR ambulat* OR locomot*) AND (knee* OR tibiofemoral*) AND (moment* OR load* OR force*). The search was limited to human research and to studies reported in English. Bibliographies of potentially eligible studies were searched recursively until no other potentially eligible studies were identified.

Inclusion determination.

Eligibility was assessed by 2 independent reviewers (MS and RSH), with disagreements resolved by consensus. In the case of continued disagreement, a third reviewer was available for arbitration (MAH). First, all titles and abstracts from the initial yield were assessed. Then, potentially eligible studies and studies whose titles and abstracts provided insufficient information were obtained in full text for further assessment.

Methodologic quality evaluation.

Study quality was evaluated by 2 reviewers (MS and MAH), with disagreements resolved by consensus. In the case of continued disagreement, a third reviewer was available for arbitration (RSH). Quality was assessed using the Downs and Black checklist (23), a valid and reliable tool commonly used in medical research (24). The tool contains 27 items evaluating 5 subscales: reporting, external validity, bias (intervention and outcome measurement), confounding (cohort selection bias), and power. As recommended in the literature (25), the power subscale (question 27) was not used in this study due to item ambiguity. Therefore, a 26-item scale was used to evaluate study quality. Each question is scored as 0 (poor quality) or 1 (good quality), with the exception of question 5 (“clear description of principal confounders”), which is scored as 0 (not satisfying), 1 (partially satisfying), or 2 (fully satisfying). Therefore, the score may range from a minimum of 0 to a maximum of 27 points.

Data extraction.

A single reviewer (MS) extracted the following information from each study: study design, sample size, participant characteristics, gait modification, modification training method, method for ensuring skill accuracy, magnitude of modification implemented, KAM parameters evaluated, calculation of joint moment data, followup duration, main study findings, adverse effects, and biomechanical effects at lower extremity joints (other than the KAM). Quantitative data pertaining to KAM outcomes were extracted. Where quantitative point estimates were not reported, authors were contacted and asked to provide data, if possible. If data were not provided, data were extracted from graphs, if available. The percentage change in KAM observed with the modification was calculated for each study by evaluating the difference between KAM measures relative to natural gait values. To evaluate modification effects throughout the gait cycle, a graphical representation of change in KAM with gait modifications was completed separately for the early- and late-stance phases of gait. Studies reporting only the overall peak KAM were considered to occur during early stance, as this is commonly the case (15). A meta-analysis of the studies was not performed due to expected high heterogeneity and the anticipated small number of studies.


Studies included in the review.

We retrieved 1,604 articles (Figure 1), of which 47 potential studies remained for full-text review. Twenty-four studies satisfied the eligibility criteria. Table 1 describes the individual studies included, whereas Table 2 summarizes the major results. All of the studies utilized a within-subject design and evaluated the immediate within-session effect of gait modification on the KAM, with the exception of 2 studies, where reevaluation occurred after 1 and 9 months, respectively (26, 27). The methodologic quality of the included studies was predominantly fair to moderate. No RCTs were identified. Sample sizes varied from 1–88 participants. Fourteen studies involved only healthy participants of diverse ages, 6 included only participants with knee OA of varying severities, and 4 evaluated healthy and knee OA participants concurrently.

Figure 1.

Flow chart of study selection process. OA = osteoarthritis; KAM = knee adduction moment.

Table 1. Description of the studies included*
ModificationAuthor, year (ref.)SampleOA diagnostic criteriaAge, mean or mean ± SD yearsNGait modificationDescription and training method
  • *

    OA = osteoarthritis; NR = not reported; K/L = Kellgren/Lawrence; N/A = not applicable; JSN = joint space narrowing; ACR = American College of Rheumatology.

Toe-outFregly et al, 2008 (29)Adult with knee OANR401Increased toe-outNot specified
 Guo et al, 2007 (30)Pain-free medial knee OAK/L 1–364 ± 8.09Increased toe-outParticipants were instructed by the researcher to increase toe-out by 15° with the use of visual cues consisting of lines on the force plate representing the 15° higher than their natural toe-out value
 Lin et al, 2001 (31)Healthy childrenN/ANR (range 11–13)44Increased toe-outParticipants were instructed to walk in 2 different postures: intentionally in-toeing and intentionally out-toeing
      Reduced toe-out 
 Lynn et al, 2008 (33)Healthy young adultsN/A22.9 ± 1.811Increased toe-outParticipants were instructed by the researcher to perform the following strategies: increase toe-out by 30° from natural angle and reduce toe-out by 30° from natural angle. Visual cues with 2 parallel lines on the floor at the specific angle were used
      Reduced toe-out 
 Lynn and Costigan, 2008 (32)Healthy older adultsN/A68.7 ± 8.412Increased toe-outInstructed by the researcher to externally rotate the foot as much as possible (toe-out) and to internally rotate the foot as much as possible (reduce toe-out). Participants were instructed to perform the magnitude of the modification to which point it felt abnormal but not uncomfortable
  Knee OAScott OA score67.4 ± 10.012Reduced toe-out 
 Reinbolt et al, 2008 (34)Adult with knee OANR411Increased toe-outNot specified
 Schache et al, 2008 (35)Young healthy adultN/A261Increased toe-outNot specified
Gait speedKirtley et al, 1985 (36)Healthy adultsN/A3710Increased gait speedInstructed by the researcher to walk at: 2 speeds above and 2 speeds below their natural gait speed. Magnitude of the change in speed was self-selected
      Reduced gait speed 
 Landry et al, 2007 (37)Healthy older adultsN/A50.7 ± 10.243Increased gait speedInstructed to increase gait speed to 150% of natural pace; training not specified
  Knee OAK/L 1–358.2 ± 8.341  
 McClelland et al, 2010 (38)Healthy older adultsN/A69.6 ± 8.340Increased gait speedParticipants were asked to walk as fast as they could without running
 Mundermann et al, 2004 (39)Healthy older adultsN/A63.3 ± 10.744Increased gait speedInstructed by the researcher to walk at self-selected slow and fast gait speeds
  Adults with medial knee OAK/L >2; medial JSN65.4 ± 10.044Reduced gait speed 
 Robbins and Maly, 2009 (40)Healthy younger adultsN/A32 ± 832Increased gait speedParticipants were required to walk 15% slower and 15% faster for the 2 gait modifications. The researcher monitored speed after each trial and provided feedback to participants regarding the speed achieved
      Reduced gait speed 
 Shultz et al, 2009 (41)Healthy normal weight childrenN/A10.4 ± 1.610Increased cadence (accompanied with increased gait speed)Participants were instructed to walk at 130% of their self-selected walking cadence. A metronome was used to guide the timing of steps
  Healthy overweight childrenN/A10.4 ± 1.810  
 Zeni and Higginson, 2009 (42)Healthy older adultsK/L 059 ± 1122Control speed; self-selected speed; fastest tolerable speedParticipants had 4 minutes of familiarization time with the treadmill. Participants walked on a treadmill for 3 speed conditions: 1) speed was controlled by the researcher to 1.0 meter/second, 2) self-selected walking speed, determined by a 10-meter timed walk, and 3) fastest tolerable speed, determined at which point the subject felt comfortable without the use of handrails
  Adults with moderate knee OAK/L 2–363 ± 9.321  
  Adults with severe knee OAK/L 459 ± 9.813  
Gait aid useChan et al, 2005 (43)Knee OAACR clinical diagnosis61.8 ± 9.414Gait aid: contralateral cane useParticipants were taught by a researcher to perform a 2-point gait pattern with the cane on either the contralateral or ipsilateral side of the affected extremity, practiced until comfortable. Participants were encouraged to practice at home once a day for 1 month. Poor adherence with the training program was reported
       During the assessment session a further 15 minutes of practice time was allocated. Testing of the contralateral and ipsilateral gait patterns was randomized
      Gait aid: ipsilateral cane use 
 Kemp et al, 2008 (44)Knee OAACR clinical and radiographic criteria65 ± 10.220Gait aid: contralateral cane useBriefly trained by a physiotherapist to place the cane on the ground during the stance phase
 Stief et al, 2008 (45)Healthy experienced Nordic walkersN/A31.0 ± 4.6415Gait aid: Nordic walking polesParticipants exercised a reciprocal gait pattern with the Nordic walking poles using a “diagonal technique” (not described in the study). A fast gait speed was chosen (2.0 meters/second) to accentuate biomechanical changes
Medial knee thrustFregly et al, 2007 (27)Adult with medial knee OAK/L 2371Medial knee thrustMedial knee thrust gait incorporated a combination of increased pelvic axial rotation and increased lower extremity flexion, resulting in a “medialization” of the knee, as predicted from computer optimization modeling (without changing foot and trunk positioning) using the participant's natural gait data. The participant studied plots of optimized kinematics and kinetics obtained from modeling. The gait modification was self-taught in a manner that was “natural and easy to achieve” over a period of 9 months
 Schache et al, 2008 (35)Young healthy adultN/A261Medial knee thrustInstructed to walk “rubbing the insides of knees together”
Step widthFregly et al, 2008 (29)Adult with knee OANR401Increased step widthNot specified
 Reinbolt et al, 2008 (34)Adult with knee OANR411Increased step widthNot specified
OtherBarrios et al, 2010 (26)Healthy younger adultsN/A21.4 ± 1.68Hip internal rotation and adductionTraining included 8 treadmill sessions with real-time biofeedback presented via a screen providing frontal knee angle information. Participants were instructed to reduce their angle in relation to a threshold representing 1 SD of a normative database by “bringing thighs inward” and “walking with knees closer together.” Feedback time was gradually reduced with each session. The timeline of sessions is not reported
 Dowling et al, 2010 (46)Healthy younger adultsN/A25.2 ± 5.29Increased weight transfer to medial side of footParticipants were evaluated using 2 different methods of feedback: verbal instruction to minimally load the lateral side of the foot; and active feedback system using a shoe pressure sensor provided vibratory feedback when lateral foot pressure exceeded the resting threshold
 Riskowski, 2010 (47)Healthy younger womenN/A26.8 ± 2.515Increased knee flexion and reduced vertical acceleration at initial contactA biofeedback-based monitoring knee brace was fitted with audible sensors activated with reduced knee flexion or high vertical acceleration at initial contact. Participants wore the brace for 30 minutes and independently modified their gait in order to minimize the feedback signals. Gait assessment was conducted without the brace, 1 hour posttraining
 Russell et al, 2010 (48)Healthy-weight younger womenN/A25.1 ± 3.810Reduced stride length (increased cadence)Participants were instructed to walk with a 15% reduction in their preferred stride length. Reduced stride length was performed with the speed monitored to match the preferred condition. Foot placement was guided by tape marks on the floor and a metronome
  Obese younger womenN/A25.3 ± 9.810  
 Mundermann et al, 2008 (49)Healthy younger adultsN/A22.8 ± 3.119Increased trunk leanInstructed by the researcher to “move their trunk more” from side to side (increased mediolateral trunk sway) to a self-selected amount. Practice was undertaken until participants were comfortable and walking at natural gait speed
 Wu and Millon, 2008 (50)Healthy younger adultsN/A28 ± 66Tai Chi gaitInstructed to perform Tai Chi gait. The training technique is not specified. Tai Chi gait incorporated deep knee flexion, large step width, 7–9 times slower speed, and “gentle foot-floor contact” (59)
  Healthy older adultsN/A72 ± 86  
Table 2. Major results of included studies*
ModificationAuthor, year (ref.)Gait modificationNatural gait: mean value of target parameterModified gait: mean value of target parameterKAM outcome reportedKAM unit of measureNatural gait: mean ± SD KAMModified gait: mean ± SD KAMKAM change, %Main findings
  • *

    KAM = knee adduction moment; NR = not reported; KAM1 = early-stance peak KAM; BW = body weight; Ht = height; KAM2 = late-stance peak KAM; OA = osteoarthritis; N/A = not applicable; IR = internal rotation.

  • Data obtained from personal communication with the authors.

  • Findings are statistically significant.

  • §

    Data obtained were extracted from graphical format.

Toe-outFregly et al, 2008 (29)Increased toe-outNR15° increase in toe-outKAM1%BW × Ht2.50 ± NR2.50 ± NR0Increased toe-out only altered the KAM2
     KAM2%BW × Ht3.20 ± NR2.00 ± NR−37.5 
 Guo et al, 2007 (30)Increased toe-out2.0° toe-out18.6° toe-outKAM1%BW × Ht2.81 ± 0.492.84 ± 0.44+1.10Large reductions demonstrated in late-stance KAM, with no change observed during early stance
     KAM2%BW × Ht2.27 ± 0.631.37 ± 0.53−39.7 
 Lin et al, 2001 (31)Increased toe-out10.0° toe-out30° toe-outKAM1§Nm/kg0.105 ± NR0.047 ± NR−55.2Increased toe-out reduced the peak KAM and created a large increase during late stance. However, the late-stance near-zero natural KAM must be considered when interpreting these results
     KAM2§Nm/kg0.003 ± NR0.024 ± NR+700.0 
  Reduced toe-out10.0° toe-out−15° toe-outKAM1§Nm/kg0.105 ± NR0.126 ± NR+20.0Reduced toe-out demonstrated a large increase in the KAM waveform. However, the late-stance near-zero natural KAM must be considered when interpreting these results
     KAM2§Nm/kg0.003 ± NR0.190 ± NR+6,233.3 
 Lynn et al, 2008 (33)Increased toe-out18.5° toe-out40.2° toe-outKAM1Nm/kg0.31 ± 0.160.35 ± 0.18+12.9Large reductions demonstrated in KAM2 with no significant change during early stance
     KAM2Nm/kg0.25 ± 0.160.02 ± 0.16−92.6 
  Reduced toe-out18.5° toe-out−9.1° toe-outKAM1Nm/kg0.31 ± 0.160.28 ± 0.16−9.7Significant increase in late-stance KAM with reduced toe-out
     KAM2Nm/kg0.25 ± 0.160.41 ± 0.14+64.0 
 Lynn and Costigan, 2008 (32)Increased toe-out healthy sample11.5° toe-out22.5° toe-outKAM1Nm/kg0.37 ± 0.110.36 ± 0.13−2.7Increased toe-out significantly reduced late-stance KAM, with no change during early stance
     KAM2Nm/kg0.27 ± 0.120.19 ± 0.14−57.8 
  Increased toe-out OA sample7.5° toe-out17.1° toe-outKAM1Nm/kg0.45 ± 0.150.46 ± 0.13+2.2Increased toe-out significantly reduced late-stance KAM, with no change during early stance
     KAM2Nm/kg0.40 ± 0.140.31 ± 0.13−31.1 
  Reduced toe-out healthy sample11.5° toe-out2.5° toe-outKAM1Nm/kg0.37 ± 0.110.32 ± 0.12−13.5Reduced toe-out (in-toeing) demonstrated a significant change in early-stance KAM in the healthy sample only. No significant change during late stance
     KAM2Nm/kg0.27 ± 0.120.26 ± 0.09−3.7 
  Reduced toe-out OA sample7.5° toe-out4.4° toe-inKAM1Nm/kg0.45 ± 0.150.43 ± 0.15−2.5Reduced toe-out (in-toeing) demonstrated no significant change in KAM values
     KAM2Nm/kg0.40 ± 0.140.39 ± 0.14−4.4 
 Reinbolt et al, 2008 (34)Increased toe-outNRNRKAM1§%BW × Ht2.36 ± NR2.50 ± NR+6.4Increased toe-out produced a small increase in the KAM during early stance and a reduction during late stance
     KAM2§%BW × Ht2.94 ± NR2.02 ± NR−31.3 
 Schache et al, 2008 (35)Increased toe-outNRIncreased toe-out by 11°KAM1Nm/kg (distal tibia reference)0.56 ± 0.030.56 ± 0.070No change during early stance and small reduction in late stance. Results were reference-frame dependent, with early-stance KAM increasing up to 29.4% in the laboratory frame, the late-stance KAM reducing up to 34.0% in the femoral reference frame, and KAM impulse showing increases of 10.0% in the proximal tibial frame
     KAM2Nm/kg (distal tibia reference)0.52 ± 0.050.40 ± 0.02−22.9 
     KAM impulseNm/kg (distal tibia reference)0.21 ± 0.010.18 ± 0.01−14.3 
Gait speedKirtley et al, 1985 (36)Increased gait speed1.4 meters/second speedNRPeak KAMNm25.3 ± 10.0NRN/AGait speed and peak KAM were weakly correlated (r = 0.41, P < .001)
  Reduced gait speed1.4 meters/second speedNRPeak KAMNm NRN/A 
 Landry et al, 2007 (37)Increased gait speed; healthy sample1.4 meters/second speed1.8 meters/second speedKAM1§Nm/kg0.52 ± NR0.61 ± NR+17.3Waveform data analysis showed no change in the overall KAM magnitude. Increasing gait speed demonstrated reductions in late-stance KAM values in relation to early stance. This was evident in healthy and OA participants
     KAM2§Nm/kg0.31 ± NR0.27 ± NR−12.9 
  Increased gait speed; OA sample1.3 meters/second speed1.8 meters/second speedKAM1§Nm/kg0.51 ± NR0.63 ± NR+23.5 
     KAM2§Nm/kg0.36 ± NR0.33 ± NR−8.3 
 McClelland et al, 2010 (38)Increased gait speed1.3 meters/second speed1.8 meters/second speedPeak KAM%BW × Ht3.59 ± 0.684.84 ± 1.17+34.8Increased gait speed resulted in a larger peak KAM
 Mundermann et al, 2004 (39)Reduced gait speed; healthy sample1.2 meters/second speed0.8 meter/second speedKAM1%BW × Ht3.19 ± 0.803.12 ± 0.68−2.2Gait speed did not alter the KAM values in healthy participants
     KAM2%BW × Ht2.49 ± 0.862.57 ± 0.71+3.2 
  Reduced gait speed; OA sample1.2 meters/second speed0.8 meter/second speedKAM1%BW × Ht3.24 ± 0.753.04 ± 0.71−6.2A positive linear correlation was identified between gait speed and early-stance KAM, more pronounced in mild knee OA (regression line slope 0.90)
     KAM2%BW × Ht2.20 ± 0.732.47 ± 0.77+12.3 
  Increased gait speed; healthy sample1.2 meters/second speedNRKAM1%BW × Ht3.19 ± 0.80NR Gait speed did not alter the KAM values in healthy participants
     KAM2%BW × Ht2.49 ± 0.86NR  
  Increased gait speed; OA sample1.2 meters/second speedNRKAM1%BW × Ht3.24 ± 0.75NR NR
     KAM2%BW × Ht2.20 ± 0.73NR  
 Robbins and Maly, 2009 (40)Increased gait speed1.4 meters/second speed1.6 meters/second speedPeak KAMNm30.05 ± 10.9732.28 ± 13.18+7.4Increased gait speed did not result in significant differences in the peak KAM or KAM impulse
     KAM impulseNm8.97 ± 3.888.69 ± 3.93−3.1 
  Reduced gait speed1.4 meters/second speed1.2 meters/second speedPeak KAMNm30.05 ± 10.9727.63 ± 10.08−8.1Reduced gait speed had no significant effect on peak KAM, but a significant increase was demonstrated in the KAM impulse
     KAM impulseNm8.97 ± 3.8810.07 ± 4.69+12.3 
 Shultz et al, 2009 (41)Increased cadence (increased speed); overweight childrenNRNRPeak KAMN26.24 ± 14.1528.66 ± 10.13+9.2No significant differences were identified in the KAM with increased cadence in either group
  Increased cadence (increased speed); normal- weight childrenNRNRPeak KAMN14.11 ± 4.3517.50 ± 5.04+24.0 
 Zeni and Higginson, 2009 (42)Reduced speed        
   Healthy1.22 meters/second1.0 meter/secondPeak KAMNm/kg0.37 ± 0.110.36 ± 0.11−4.6No significant difference
   Moderate OA1.13 meters/second1.0 meter/secondPeak KAMNm/kg0.40 ± 0.200.42 ± 0.19+5.5No significant difference
   Severe OA1.03 meters/second1.0 meter/secondPeak KAMNm/kg0.47 ± 0.180.45 ± 0.18−4.5Small statistically significant reduction with controlled gait speed
   Increased speed        
   Healthy1.22 meters/second1.75 meters/secondPeak KAMNm/kg0.38 ± 0.110.44 ± 0.14+19.0Small statistically significant increase in peak KAM
   Moderate OA1.13 meters/second1.50 meters/secondPeak KAMNm/kg0.40 ± 0.200.45 ± 0.25+13.3Small statistically significant increase in peak KAM
   Severe OA1.03 meters/second1.37 meters/secondPeak KAMNm/kg0.47 ± 0.180.45 ± 0.16−3.4No significant difference
Gait aidChan et al, 2005 (43)Gait aid: contralateral cane useNo aidContralateral cane usePeak KAMNm/kg0.55 ± 0.200.51 ± 0.14−7.3Contralateral cane use had no significant effect on peak KAM
  Gait aid: ipsilateral cane useNo aidIpsilateral cane usePeak KAMNm/kg 0.77 ± 0.46+40.0Ipsilateral cane use significantly increased peak KAM; however, this effect was seen in only 35% of the participants.
 Kemp et al, 2008 (44)Gait aid: contralateral cane useNo aidContralateral cane useKAM1%BW × Ht3.76 ± 0.953.38 ± 0.68−10.1Significant reductions were achieved in both early- and late-stance KAM peaks with contralateral cane use
     KAM2%BW × Ht2.92 ± 0.902.47 ± 0.95−15.4 
 Stief et al, 2008 (45)Gait aid: Nordic walking polesNo aidNordic walking pole use on both sidesKAM1Nm66.13 ± 13.6457.59 ± 14.37+14.8Significant increase demonstrated in early-stance KAM, with no change during late stance
     KAM2Nm40.08 ± 7.6341.12 ± 8.68+2.6 
Medial knee thrustFregly et al, 2007 (27)Medial knee thrustNo specific quantifiable parameterNo specific quantifiable parameterKAM1%BW × Ht3.80 ± NR1.90 ± NR−50.0Medial knee thrust gait reduced the KAM1 in this subject
     KAM2%BW × Ht4.60 ± NR2.10 ± NR−55.0 
 Schache et al, 2008 (35)Medial knee thrustNo specific quantifiable parameterNo specific quantifiable parameterKAM1Nm/kg (distal tibia reference)0.56 ± 0.030.32 ± 0.03−43.8Reduction in both KAM peaks and impulse occurred with medial knee thrust. Results were reference-frame dependent, with the smallest early-stance KAM changes reducing 2.4% in the proximal tibia frame, the late-stance KAM reductions starting from 7.8% in the femoral reference frame, and KAM impulse showing increases of up to 10.0% in the proximal tibial frame
     KAM2Nm/kg (distal tibia reference)0.52 ± 0.050.43 ± 0.03−17.3 
     KAM impulseNm/kg (distal tibia reference)0.21 ± 0.010.15 ± 0.01−29.7 
Step widthFregly et al, 2008 (29)Increased step widthNR20-cm increase in step widthKAM1%BW × Ht2.60 ± NR2.20 ± NR−15.4Increased step width altered KAM peaks during early and late stance
     KAM2%BW × Ht4.20 ± NR3.80 ± NR−9.5 
 Reinbolt et al, 2008 (34)Increased step widthNRNRKAM1§%BW × Ht2.36 ± NR2.14 ± NR−9.3Increased step width reduced the KAM throughout stance
Other modificationsBarrios et al, 2010 (26)Increased hip IR and adduction5.3° hip IR; 9.2° hip adduction13.5° hip IR; 12.3° hip adductionPeak KAMNm/kg0.43 ± 0.070.34 ± 0.07−20.9Increased hip IR and adduction significantly reduced the peak KAM. Participants were able to reproduce the gait pattern 1 month posttraining
 Dowling et al, 2010 (46)Increased weight transfer to medial side of foot; verbal feedbackNRNRKAM1%BW × Ht2.48 ± 0.402.29 ± 0.55−7.7Increased weight transfer toward the medial side of the foot reduced the KAM1, with statistical significance achieved using the shoe active feedback system only
  Active feedback  KAM1%BW × Ht2.54 ± 0.562.18 ± 0.57−14.2 
 Riskowski, 2010 (47)Increased knee flexion and reduced vertical acceleration1.2° knee flexion; −5.87 meters/second2 vertical acceleration3.8° knee flexion; −4.89 meters/second2 vertical accelerationPeak KAM%BW × Ht0.51 ± 0.070.57 ± 0.07+11.8The biofeedback knee brace used for training produced a significant increase in the peak KAM 1 hour following the training period
 Russell et al, 2010 (48)Reduced stride length; obese sample1.49 meters stride lengthNRKAM1Nm40.17 ± 13.7940.69 ± 12.44+1.3Reduced stride length did not significantly alter the peak KAM. Significant reductions in the KAM impulse were achieved with a reduced stride length
     KAM impulseNm2,712 ± 1,8802,508 ± 1,489−7.5 
  Reduced stride length; healthy- weight sample1.60 meters stride lengthNRKAM1Nm31.94 ± 7.6829.32 ± 8.22−8.2 
     KAM impulseNm2,025 ± 6981,696 ± 724−16.3 
 Mundermann et al, 2008 (49)Increased trunk leanNR10° trunk leanKAM1%BW × Ht2.02 ± 0.500.73 ± 0.56−63.9Significant reduction demonstrated in early-stance KAM. No change observed during late stance
     KAM2%BW × Ht1.59 ± 0.561.51 ± 0.79−5.0 
 Wu and Millon, 2008 (50)Tai Chi gait: results include younger and older participantsNR; speed: 52 cm/secondNR; speed: 8 cm/secondPeak KAM§%BW × Ht3.46 ± 0.477.00 ± 1.27+102.3Significant increases were found in the peak KAM with Tai Chi gait in both the younger and older participants

In studies evaluating knee OA participants, there was considerable variation in the diagnostic criteria used. Most commonly the Kellgren/Lawrence radiographic grading scale was used (5 studies), and the American College of Rheumatology OA criteria (2 studies) (28) and the Scott OA score (1 study) were also used. Diagnostic criteria were not specified for 3 studies.

Gait modification strategies.

Most studies investigated one or more gait modifications. As seen in Table 1, 14 different modifications were evaluated. Seven studies (29–35) evaluated the effect of increasing toe-out, 3 of which also evaluated reducing toe-out (31–33). Alteration of gait speed was also investigated, with 7 studies evaluating increasing speed (36–42) and 4 studies investigating reduced speed (36, 39, 40, 42). Gait aids were investigated in 3 studies, 2 of which evaluated contralateral cane use (43, 44), 1 investigated ipsilateral cane use (43), and 1 examined Nordic walking poles (45). A medial knee thrust gait pattern was implemented in 2 single-subject studies (27, 35), with another study implementing elements of a medial knee thrust pattern (increased hip internal rotation and adduction) (26). Two single-subject studies evaluated the effect of increased step width, achieved by increasing the frontal plane distance between feet during consecutive steps (29, 34). Other gait modifications included single studies evaluating increased weight transfer to the medial foot (46), increased knee flexion and reduced vertical acceleration at initial contact (47), reduced stride length (48), increased mediolateral trunk lean (49), and Tai Chi gait (50). Gait modifications were inadequately described in most studies, with little detail provided regarding training methods, instructor qualifications, or accuracy of participant skill acquisition (Table 1).

Outcome assessment.

Most studies evaluated the overall or early-stance peak KAM (Table 2). Twelve studies reported early-stance peak KAM values (27, 30–33, 35, 43–46, 48, 49), with late-stance peak KAM values reported in 6 studies (27, 30, 32, 33, 35, 45). Point estimates and measures of variability were not reported for 1 study and could not be extracted from graphical data (36).

Effect of gait modifications on the KAM.

Gait modifications resulted in KAM changes during the early- (Figure 2A and Table 2) and late-stance phases of gait (Figure 2B and Table 2), with the size of the effect differing according to the modification used, level to which it was implemented, sample recruited, and KAM reference frame. Increased toe-out resulted in wide-ranging effects on early-stance KAM, where changes ranged from a reduction of 55.2% to an increase of 12.9%; however, most changes were nonsignificant. In contrast, late-stance KAM reductions were observed with this strategy (22.9% to 92.6%) with the exception of 1 study, whose baseline late-stance KAM value was minimal (0.003 Nm/kg) (31) and consequently found a 700.0% increase in this parameter. Generally, reduced toe-out resulted in small reductions in early-stance KAM (4.4% to 13.5%) apart from 1 study (31), which demonstrated a 20.0% increase. Inconsistent effects of reduced toe-out were observed during late stance, with half of the studies demonstrating small reductions and the rest demonstrating large increases. Importantly, the largest increase in late-stance KAM (31) should be interpreted with caution due to inconsistencies in the reporting of these data and the near-zero value of the natural late-stance KAM value.

Figure 2.

Effects of gait modifications on medial knee joint load during the early- (A) and late-stance (B) phases of gait, as measured by the external knee adduction moment (KAM). The bars correspond to a KAM percentage change with implementation of the gait modification in comparison to natural gait. Reductions in the KAM (bars to the left of the vertical axis) signify reductions in medial knee compartment loading. H = healthy sample; OA = osteoarthritis sample; OA-S = severe OA sample; OS = overweight sample; OA-M = moderate OA sample; NS = normal weight sample; VF = visual feedback training; AF = active feedback training; * = study results exceeding the graph scale.

Although increased gait speed increased the early-stance KAM in most studies (by up to 34.8%), small reductions (8.3% to 12.9%) were noted during the late-stance phase (in both healthy and OA participants) in the single study reporting these data (38). Effects of reduced gait speed were variable and inconsistent with respect to early-stance KAM (8.0% reduction to 5.5% increase), and the single study reporting late-stance KAM exhibited small increases (3.2% to 12.3%) in both healthy and OA participants (39).

Effects of gait aids varied according to the aid and technique implemented. Contralateral cane use reduced early-stance KAM in both studies evaluating this strategy (7.3% to 10.1%), as well as during late stance (15.4%). Conversely, ipsilateral cane use resulted in a 40% increase in KAM during early stance. Nordic pole walking increased the early-stance KAM (14.8%), with little change during late stance.

The effect of increased step width resulted in small reductions in both early- (4.7% to 15.4%) and late-stance KAM (9.5% to 11.7%). The medial knee thrust gait pattern exhibited reductions in the KAM during both early- (43.8% to 50.0%) and late-stance phases (17.3% to 55%). Increased hip internal rotation and adduction demonstrated a significant reduction in early-stance KAM (20.9%). Weight transfer toward the medial foot achieved a significant reduction in early-stance KAM (14.2%) when implemented using an active biofeedback device. A biofeedback brace aimed at increasing knee flexion and reducing acceleration at initial contact resulted in an increased peak KAM (11.8%). A reduction in stride length without alteration of speed did not alter the peak KAM, and exhibited small reductions in the KAM impulse (7.5% to 16.3%). Increased lateral trunk lean demonstrated a large reduction in early-stance KAM (63.9%), with little change during late stance. Conversely, Tai Chi gait demonstrated a large increase in early-stance KAM of 102.3%.

Adverse effects and effects at other joints.

Only 1 study (49) explicitly defined and evaluated adverse biomechanical effects, which they characterized as increases in axial loading at the ankle, knee, or hip. The results of this study showed no changes in axial loading rates with increased trunk lean. Despite limited reporting of adverse effects, other studies did report some lower extremity kinematic and/or kinetic effects of gait modifications (other than changes in the KAM), and these are summarized in Table 3. Only 1 study evaluated participant-subjective views of the gait modification (26), specifically the strategy naturalness and perceived effort. No studies evaluated participant symptoms with the implementation of modifications.

Table 3. Kinematic, kinetic, and/or spatiotemporal effects (other than change in knee adduction moment) at the lower extremity joints reported with implementation of gait modifications*
  • *

    IR = internal rotation; OA = osteoarthritis.

Increased toe-out
  ↑ power absorption in early stance (extension) (31)
  ↑ frontal plane power generation in midstance (31)
  ↑ peak extension moment in early stance (31)
  ↑ flexion moment (35)
  ↑ peak IR moment (31)
  ↓ lateral–medial shear force during early stance (OA sample) (32)
  ↓ lateral–medial shear force during late stance (32, 33)
  ↓ sagittal plane power generation in midstance (31)
  ↑ sagittal plane power generation in late stance (31)
Reduced toe-out
  ↑ frontal plane power generation in late stance (31)
  ↑ peak flexion moment in late stance (31)
  ↑ peak IR moment (31)
  ↓ sagittal plane power generation in midstance (31)
  ↑ sagittal plane power generation in late stance (31)
  ↑ lateral–medial shear force during late stance (32, 33)
Increased gait speed
  ↑ flexion range throughout stance (41)
  ↓ abduction during stance (41)
  ↑ flexion moment (42)
  ↑ flexion range throughout gait (37)
  ↑ flexion moment (37, 38, 42)
  ↑ external rotation moment during early stance (37)
  ↑ IR moment during late stance (37)
  ↓ ankle plantar flexion during stance (41)
  ↓ foot inversion during stance (41)
  ↑ plantar flexion moment (42)
  ↑ vertical ground reaction force (38)
Reduced gait speed
  ↓ loading rate (moderate OA sample) (42)
Contralateral cane use
  ↓ flexion moment (43, 44)
  ↓ cadence (43, 44)
  ↓ speed (43, 44)
  ↑ step length (44)
  ↑ vertical ground reaction force (44)
Ipsilateral cane use
  ↑ adduction moment (43)
  ↓ cadence (43)
  ↓ speed (43)
Nordic pole walking
  ↓ flexion early stance (45)
  ↓ maximal flexion (45)
  ↓ flexion moment (45)
  ↓ IR moment (45)
  ↓ plantar flexion during stance (45)
  ↓ maximal eversion during stance (45)
  ↓ eversion moment (45)
  ↓ loading rate during early stance (45)
  ↓ horizontal ground reaction force peaks (45)
Medial knee thrust
  ↓ adduction moment (27)
  ↑ flexion during stance (35)
  ↑ extension moment (27)
  ↑ flexion moment (35)
  ↑ plantar flexion (27)
  ↑ inversion moment (27)
Increased knee flexion and reduced vertical acceleration at initial contact
  ↑ flexion moment (47)
  ↓ rate of loading (47)
  Increased trunk lean
  ↑ abduction moment (49)
  ↓ adduction moment during early stance (49)
  ↑ flexion angle at heel strike (49)
  ↑ abduction moment (49)
  No change in axial loading rates: hip, knee, and ankle (49)
Tai Chi gait
  ↓ joint compressive force (50)
  ↑ peak flexion, IR, and adduction moments (50)
  ↑ joint compressive force (50)
  ↑ peak shear forces (50)
  ↑ peak flexion and IR moments (50)
  ↓ plantar flexion moment (50)

Methodologic quality.

The methodologic quality of studies ranged from 7 to 17 points of 27 (mean ± SD score 12.3 ± 2.8). While many studies performed strongly with regard to reporting (Table 4), the most common flaws were failure to provide detailed explanations of either the gait modification implemented, instructions given to participants, or training methods used. Several studies also poorly described study findings and failed to provide estimates of random variability. All of the studies scored poorly regarding external validity and only slightly better for internal validity bias and confounding variables. A major shortfall of these studies was a sample selection bias, and thus reduced the ability to generalize study findings across the population.

Table 4. Methodologic quality of included studies as evaluated using the Downs and Black checklist (23)
Author, year (ref.)Reporting (n = 11)External validity (n = 3)Internal validity: bias (n = 7)Internal validity: confounding (n = 6)Total quality score, no. (%) (n = 27)
Barrios et al, 2010 (26)1004216 (59)
Chan et al, 2005 (43)903113 (48)
Dowling et al, 2010 (44)704213 (48)
Fregly et al, 2007 (27)603211 (41)
Fregly et al, 2008 (29)503210 (37)
Guo et al, 2007 (30)905115 (56)
Kemp et al, 2008 (44)914216 (59)
Kirtley et al, 1985 (36)502411 (41)
Landry et al, 2007 (37)904417 (63)
Lin et al, 2001 (31)50218 (30)
Lynn et al, 2008 (33)904215 (56)
Lynn and Costigan, 2008 (32)904114 (52)
McClelland et al, 2010 (38)805215 (56)
Mundermann et al, 2004 (39)802212 (44)
Mundermann et al, 2008 (49)903113 (48)
Reinbolt et al, 2008 (34)40217 (26)
Riskowski, 2010 (47)704213 (48)
Robbins and Maly, 2009 (40)905216 (59)
Russell et al, 2010 (48)904215 (56)
Schache et al, 2008 (35)703111 (41)
Shultz et al, 2009 (41)1004115 (56)
Stief et al, 2008 (45)603413 (48)
Wu and Millon, 2008 (50)804315 (56)
Zeni and Higginson, 2009 (42)804214 (52)


This systematic review evaluated the effect of different gait modifications on medial knee load as measured indirectly via the KAM. Results showed that different modifications exert different effects on dynamic knee load at varying points throughout the gait cycle. Of 14 gait modifications identified, strategies that most consistently reduced the early-stance KAM (the parameter known to predict radiographic disease progression) are contralateral cane use and increased trunk lean. It must be noted that in some cases, results are conflicting and/or based on very few/single studies. Medial knee thrust gait and increased trunk lean demonstrated the greatest reductions in early-stance KAM. The adverse effects of these strategies have not been systematically evaluated, and while some studies evaluated kinetic and kinematic changes at other lower extremity joints, the clinical implications of these changes remain largely unknown. Due to limited high-quality research in this area, particularly the absence of RCTs, and the lack of studies evaluating the effectiveness of gait modification over the longer term, it is impossible to conclude which gait modification is superior for reducing medial knee load.

Increased toe-out consistently reduced the late-stance KAM. This is likely due to a lateral shift in the center of pressure (16) consequently reducing the KAM lever arm, which would be most pronounced after midstance (17) as the body progresses over the midfoot. Although late-stance KAM reductions were substantial and achievable with only modest changes in toe-out angle (10–20°), the clinical implications of this parameter remain unknown at present. This is particularly important in light of the tendency of increased toe-out to elevate the early-stance peak KAM. In contrast, the reduced toe-out modification resulted in small nonsignificant reductions in early-stance KAM, potentially due to a lateral shift in the center of pressure across the foot (51), thereby reducing the KAM lever arm. Whether these reductions would translate into clinically meaningful improvements in patient status or ultimately slow disease progression is unknown. Nonetheless, changing toe-out during gait is likely to be relatively easy to implement in the clinical setting, despite the generally poor description of training methods implemented. However, the ability of patients to maintain the gait modification in the longer term remains unknown.

Altering gait speed resulted in inconsistent early- and late-stance KAM changes, with effects varying across study samples. Speed may affect the KAM by altering the vertical and frontal plane center of mass acceleration, thus influencing the ground reaction force magnitude (52). It was therefore hypothesized that reductions in speed would proportionally reduce the KAM, and vice versa. However, study results do not consistently support this hypothesis. Findings regarding increased gait speed were generally more consistent, resulting in an increased early-stance peak but lower late-stance peak KAM, and thus rendering this gait strategy of questionable value for knee load reduction in OA. Furthermore, the desirability of training slower or faster than normal gait speeds must be considered in light of potential adverse implications on overall physical function, daily tasks, and safety.

Gait aids are widely recommended for managing OA (4). A Dutch survey of people with OA found that 44% possessed a cane, most commonly those experiencing pain (53). Cane use is postulated to influence the KAM by offloading the weight borne through the affected lower extremity via the upper extremity (20), thereby reducing the ground reaction force. As hypothesized, contralateral cane use demonstrated small reductions in both early- and late-stance KAM. Conversely, ipsilateral cane use increased the early-stance KAM, which is consistent with effects observed on hip joint moments (54). These results highlight the potential long-term dangers of inappropriate (ipsilateral) cane use. This is particularly important, given that many people self-prescribe the use of gait aids (53) and may receive little, if any, education regarding appropriate use. Even when used appropriately (cane positioned in the contralateral hand to the symptomatic extremity), there is the possibility of adverse kinetic effects on the ipsilateral knee (less symptomatic/unaffected extremity), given that there is a small portion of the gait cycle where the cane and ipsilateral extremity contact the ground simultaneously. Compared to canes, Nordic walking poles exhibited different biomechanical effects, possibly due to bilateral use with greater shoulder and elbow flexion and the further positioning from the body. Nordic poles significantly increased early-stance KAM (45), suggesting that they are an unfavorable choice for knee load reduction over the longer term. Nevertheless, Nordic walking poles may be useful as an intermittent strategy to improve balance and facilitate enhanced participation in regular walking. A limitation of all gait aid studies is the failure to measure load borne through the gait aid. Therefore, conclusions regarding the participants' abilities to offload, or optimal levels of body weight support needed to reduce knee load, cannot be drawn.

Increasing step width during gait produced small but consistent reductions in both KAM peaks (29, 34). Increasing step width may alter the KAM by lateralizing the center of pressure, sequentially reducing the KAM lever arm. Importantly, studies evaluating this strategy involved single subjects only limiting the generalizability of the current evidence.

The “medial knee thrust” gait pattern, implemented in 2 single-subject studies (27, 35), attempted to alter dynamic knee alignment (and thus KAM lever arm) during gait by medially directing the knee. This pattern achieved reductions in KAM throughout the gait cycle, with the size of the effect showing some variability during late stance. Despite this, training of this modification may pose several challenges given the complexity of the movement, and may require use of special training and biofeedback mechanisms. Several elements of this modification were implemented in another study, which evaluated increased hip internal rotation and adduction and demonstrated similar effects to medial thrust gait during early stance (26). In this study, a real-time biofeedback system was used to aid training, with favorable results shown in participant subjective measures such as perceived effort and naturalness. A possible adverse effect of gait modifications promoting increased hip internal rotation during gait could be patellofemoral pain due to alterations in patellar tracking (55), although this has not been investigated to date.

Increased weight transfer toward the medial side of the foot demonstrated a small reduction in the early-stance peak KAM (46). This modification may be altering the KAM by a change in the ground reaction force origin, thus reducing the moment's lever arm. The study demonstrated that a shoe biofeedback mechanism was required for a significant change in knee load with this gait modification, which may limit the clinical usefulness of such a strategy.

A biofeedback training brace used to increase knee flexion at initial contact and reduce vertical acceleration resulted in a significant peak KAM increase (47). Although this modification may not be appropriate for participants for medial knee OA, it may be more applicable for people post–anterior cruciate ligament injury as the rate of loading was reduced.

Reduced stride length without alteration of gait speed resulted in nonsignificant peak KAM changes and small reductions in the KAM impulse (48). The change in KAM impulse was most likely achieved through a reduced stance phase time. It must be noted that a reduced stride length would be accompanied with a higher cadence; consequently, a reduced KAM impulse may not be reflected in the cumulative loading of the knee.

Increased lateral trunk lean demonstrated the largest reduction in early-stance KAM (49). However, this modification has only been evaluated in one study of healthy participants. Trunk lean may influence the KAM by altering the ground reaction force direction in response to the changed frontal plane center of mass location, consequently reducing the KAM lever arm (18). Although findings are promising, the clinical feasibility is unclear and training may present several challenges. Patients with OA may have difficulty coordinating increases in trunk lean during the stance phase of gait and in achieving adequate lean to reduce knee load. Currently, it is unknown how much lean is required to reduce knee load, and future research should aim to determine this.

Tai Chi gait (50) generated a large increase in early-stance KAM in the single study evaluating this strategy, and may therefore be detrimental for people with knee OA. Although described by authors as a gait, this modification may best be viewed as an intermittent exercise for balance and stability rather than a long-term gait modification. Irrespective of its effects on knee load, Tai Chi gait has questionable feasibility as a gait modification, given the involvement of extremely slow speeds and high amounts of lower extremity flexion.

Despite the limited research available, the reduction in knee load evident with some gait modifications may be clinically important. For example, based on a longitudinal study evaluating disease progression (7), early-stance peak KAM reduction achieved by increased trunk lean (1.3 Nm/body weight × height%) (49) could equate to a more than 6-fold decreased risk of radiographic OA progression. Although large reductions in late-stance KAM were observed with increased toe-out and medial knee thrust gait, the clinical relevance of a late-stance KAM reduction remains unknown. Future research is needed to determine whether this parameter is important in disease pathogenesis. However, indirect support is present from a longitudinal study demonstrating a reduced rate of OA progression in people who naturally ambulated with higher toe-out angles (16).

The methodologic quality of included studies could be considered fair to moderate. No RCTs were identified, and with the exception of 2 studies (26, 27), none implemented modifications over the longer term. Studies lacked an adequate description of procedures, the modification implemented, the training methods, and the evaluation of skill acquisition. Other methodologic issues included the failure to thoroughly control extraneous variables such as speed and step length, inadequate standardization of gait modification magnitudes, and small sample sizes. Also, several clinically important measures were not evaluated, such as biomechanical or symptomatic adverse effects and participant satisfaction with the modification.

Several limitations should be considered when interpreting the results of this review. The KAM was chosen as the primary outcome, given it is the most commonly used measure of medial joint loading in knee OA and its important relationship to disease progression (7). However, alternate measures of knee load are available, including medial compartment compressive force measured via instrumented joint replacement (9, 56, 57) or by mathematical musculoskeletal modeling (29, 34, 58). Despite this, only limited research is available using these measures due to their invasive nature, potential expense, and complicated and/or time-consuming nature. Importantly, strong correlations between internal medial knee joint contact force and the KAM were found (9), supporting the KAM use as an appropriate noninvasive measurement of medial knee load.

The body of current research is an important initial step in the development of noninvasive and potentially structure-modifying treatment modalities for medial knee OA. Future research should aim to evaluate gait modifications in large samples of people with medial knee OA and using the rigor of a randomized controlled study design. Studies are also required to determine the magnitude of gait modification required to maximize knee load reduction and the optimal method of training. Evaluation of the effects of gait modification on OA symptoms and at lower extremity joints other than the knee, as well as long-term patient adherence, is also required. Finally, it remains unknown whether immediate reductions in the KAM with gait modification translate into longer-term clinically relevant changes in risk of disease progression, and this should be evaluated in the future.

In conclusion, this systematic review provides limited evidence for the benefits of some gait modifications in reducing medial knee joint load. Findings from this review are hypothesis generating rather than conclusive due to the limited research available and the methodologic limitations of currently available literature. The gait modifications that show the most promise in reducing early-stance KAM are increased lateral trunk lean and contralateral cane use. Although increased toe-out appears to be the optimal modification for reducing late-stance KAM, the clinical importance of this effect remains unknown.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Ms Simic had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Simic, Hinman, Wrigley, Bennell, Hunt.

Acquisition of data. Simic.

Analysis and interpretation of data. Simic, Hinman, Bennell, Hunt.