Thrust during ambulation and the progression of knee osteoarthritis




To determine whether the presence of varus thrust at baseline increases the risk of progression of medial tibiofemoral osteoarthritis (OA), whether knees with thrust have a greater adduction moment, whether thrust has any additional impact on top of static varus, and whether thrust is associated with poor physical function outcome.


Two hundred thirty-seven patients with knee OA (definite osteophytes and symptoms) underwent baseline gait observation to assess varus thrust and full-limb radiography to assess alignment. Sixty-four of these 237 patients also underwent quantitative gait analysis to determine the maximum knee adduction moment. Two hundred thirty patients (97%) returned for followup at 18 months. At baseline and 18 months, the 230 participants had semiflexed, fluoroscopically confirmed knee radiographs (with progression defined as worsening of medial joint space grade); self-reported and performance-based measures of function were also assessed. Logistic regression with generalized estimating equations was used to estimate odds ratios (ORs) for medial OA progression, after excluding knees that were not at risk for progression.


Varus thrust was present in 67 of 401 knees. Thrust increased 4-fold (age-, sex-, body mass index–, and pain-adjusted OR 3.96, 95% confidence interval [95% CI] 2.11–7.43) the odds of medial progression, with some reduction after further adjustment for varus alignment severity. In varus-aligned knees, thrust increased the odds of OA progression 3-fold (adjusted OR 3.17, 95% CI 1.60–6.31). In the gait substudy, the adduction moment was greater in knees with a thrust compared with knees without a thrust. Having a thrust in both knees versus neither knee was associated with a 2-fold increase in the OR for poor physical function outcome (P not significant).


Varus thrust is a potent risk factor, identifiable by simple gait observation, for disease progression in the medial compartment, the most common site of OA involvement at the knee. Varus thrust may also predict poor physical function outcome. Varus thrust increased the odds of progression among varus-aligned knees considered separately, suggesting that knees with a thrust are a subset of varus-aligned knees at particularly high risk for progression of OA.

Knee osteoarthritis (OA) is a major source of chronic disability (1). Current treatments may improve symptoms but do not delay disease progression. Knee OA that has progressed is the leading indication for joint replacement surgery. Factors that contribute to OA progression may represent targets for novel disease-modifying interventions.

Increased joint loading is theorized to play a key role in the progression of knee OA, but few specific mechanical factors have been identified (2). Varus alignment (hip–knee–ankle angle of >0° in the varus direction) is a static measurement assessed in the standing position using full-limb radiography. In contrast, varus thrust is the visualized dynamic bowing-out of the knee laterally, i.e., the abrupt first appearance of varus (or the abrupt worsening of existing varus) while the limb is bearing weight during ambulation, with return to a less varus alignment during the non–weight-bearing (swing) phase of gait (see Figure 1).

Figure 1.

Drawings representing the stance phase of gait for the right leg. The lateral knee motion represents varus thrust, the abrupt first appearance of varus alignment (bow-leggedness), or the abrupt worsening of existing varus alignment in the weight-bearing limb, with return to less varus alignment after push-off.

The impact of a thrust on the progression of knee OA has not previously been reported. There are several compelling reasons to look at the impact of thrust. With each step of walking, a varus thrust acutely increases load across the medial tibiofemoral compartment, the commonest site of OA disease at the knee. The knee of the weight-bearing limb sustains forces comparable to 3–5 times body weight during normal gait (3–5). Alignment influences how body weight forces are distributed between the medial and lateral compartments; alignment during ambulation is crucial. A thrust is a dynamic worsening of alignment in the phase of the gait cycle when the knee is most vulnerable to malalignment, i.e., when full body weight is on 1 leg. The impact of a thrust on function has also not been reported. In theory, the instability associated with a thrust may cause difficulty with knee-related physical tasks and may predict a decline in function.

It is likely that at least some of the effect of thrust on OA progression comes from static malalignment. However, a thrust reveals greater derangement than static malalignment: worsening of alignment at the most vulnerable phase of gait and dynamic instability. In an unstable knee, opposing joint surfaces are shifted so that fit is reduced, and both shear and compression forces are increased (6). If a varus thrust does harm to the knee beyond that attributable to the static varus condition, then the thrust should increase the likelihood of progression in varus knees considered separately. If the thrust effect on progression is fully explained by static varus, then the presence of thrust should matter little among varus-aligned knees.

Direct measurement of medial load requires invasive approaches that are not feasible in cohort studies. However, closely related to medial load is the moment (torque) that adducts the knee during the stance phase of gait (adduction moment). The adduction moment can be estimated using quantitative gait analysis (5, 7, 8). In theory, a varus thrust may influence the magnitude of the adduction moment: if a varus thrust increases medial load, then the adduction moment should be greater in knees with a varus thrust than in knees without a thrust.

We hypothesized that 1) the baseline presence of a varus thrust during ambulation increases the risk of medial knee OA progression over the ensuing 18 months, 2) this relationship is reduced after adjusting for static varus malalignment, which is likely to partially explain the thrust effect, 3) the presence of a varus thrust during ambulation increases the likelihood of medial progression among (static) varus-aligned knees considered separately, and 4) knees with (versus knees without) a thrust have a greater adduction moment during quantitative gait analysis. Finally, we explored the relationship between varus thrust at baseline and physical function outcome.



The Mechanical Factors in Arthritis of the Knee (MAK) study is a natural history study of knee OA, conducted at Northwestern University. As previously described (2), MAK participants were recruited from several community sources. Inclusion and exclusion criteria were based on those proposed at a National Institute of Arthritis and Musculoskeletal and Skin Diseases/National Institute on Aging (NIAMS/NIA)–sponsored multidisciplinary workshop for knee OA progression studies (9). Inclusion criteria were definite tibiofemoral osteophyte presence (Kellgren/Lawrence [K/L] radiographic grade of ≥2) in 1 or both knees and at least some difficulty (Likert category) with 2 or more items in the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) physical function scale. Exclusion criteria were corticosteroid injection within the previous 3 months or a history of avascular necrosis, inflammatory arthritis, periarticular fracture, Paget's disease, villonodular synovitis, joint infection, ochronosis, neuropathic arthropathy, acromegaly, hemochromatosis, Wilson's disease, osteochondromatosis, gout, pseudogout, osteopetrosis, bilateral total knee replacement, or a planned total knee replacement within the next year. Institutional review board approval was obtained, and all participants provided written informed consent.

Measurement of varus thrust, varus–valgus alignment, and the external adduction moment.

All measurements were obtained in both knees. For all participants in the cohort (n = 237), gait observation for thrust was performed at baseline. For 64 of these participants, an additional quantitative gait analysis was performed within 1 month, affording the chance to determine whether knees with a thrust had a greater adduction moment compared with knees without a thrust. The examiner/reader for each of these analyses (thrust, alignment, adduction moment) was blinded to data obtained by the other examiners.

Observation of gait for the presence of thrust was performed in a single unit and walkway, following a protocol that standardized the instructions given to the participant and the position and steps for the examiner. To assess intrarater reliability, it was not possible to use the live observation of gait, because the examiner may remember the presence or absence of thrust in specific individuals. Therefore, we videotaped 40 MAK participants with varying body habitus and severity of knee OA. Each subject's identity was concealed by videotaping subjects from the waist down and having all subjects wear identical shirts, shorts, and socks with no shoes. Each examiner viewed the videotapes during 2 separate sessions, at which the order of tapes had been altered, revealing very good intrarater reliability (κ = 0.81).

To assess alignment, a single anteroposterior full-limb radiograph was obtained, adhering to a previously described protocol (2). Alignment was measured by 1 reader as the angle formed by the intersection of the line connecting the centers of the femoral head and intercondylar notch with the line connecting centers of the ankle talus surface and tibial spine tips. Reliability for this approach is high, as previously reported (2).

Sixty-four of the 237 participants underwent quantitative gait analysis at Rush–Presbyterian–St. Luke's Medical Center. The Computerized Functional Testing Corporation (Chicago, IL) system was used, including 4 Qualisys (Atlanta, GA) optoelectronic cameras and a single Bertec (Columbus, OH) multicomponent force platform, with a sampling frequency of 120 Hz. Six passive markers, placed at the following bony landmarks, were used in the link segment model: lateral-most aspect of the superior iliac crest, greater trochanter, lateral joint axis line of the knee, lateral malleolus, lateral aspect of the calcaneous, and base of the fifth metatarsal joint. Inverse dynamics were used to calculate the external moments (in 3 planes) at the hip, knee, and ankle joint centers, using the 3-dimensional kinematic data acquired by the cameras along with the ground reaction forces and moments acquired by the force plate. The external moment was calculated by taking into account the moment about the joint center (created by the ground reaction force) and inertial forces. An external moment is equal and opposite to the internal moment created by muscles, soft tissues, and joint contact forces. The maximum of the adduction moment peaks was used in the analysis.

Measurement of knee pain intensity.

As previously described (10), knee pain was measured using 100-mm visual analog scales, with separate scales for the right and left knees. The scales were anchored at 0 (no pain) and 100 (pain as bad as it could be), and standardized instructions were given.

Assessment of OA progression.

Bilateral weight-bearing knee radiographs were obtained at baseline and 18 months, following the Buckland-Wright protocol (2, 11). This protocol meets criteria set by the NIAMS/NIA workshop (9) and the OA Research Society International (OARSI) (12). Knee position, beam alignment, magnification correction, and measurement landmarks were specified. The semiflexed position of this protocol superimposes anterior and posterior medial tibial margins. Tibial rim alignment and tibial spine centering in the notch were confirmed fluoroscopically before radiographs were obtained. Radiographs were obtained in a single unit by 2 trained technicians. Foot maps made at baseline were used at followup.

Medial tibiofemoral OA progression was defined as any worsening in the grade for radiographic medial joint space between baseline and 18 months. The grades described in the illustrated OARSI atlas (none, possibly, definitely, severely narrowed joint space) (13) were used by 1 reader (κ = 0.80–0.86). To describe the sample, the K/L grading system was used. In this system, 0 = normal, 1 = possible osteophytes, 2 = definite osteophytes, possible joint space narrowing, 3 = moderate osteophytes, definite narrowing, some sclerosis, possible attrition, and 4 = large osteophytes, marked narrowing, severe sclerosis, definite attrition. The reader of the knee radiographs was blinded to the gait observation, full-limb, and gait analysis data.

Measurement of physical function outcome.

Physical function was assessed at baseline and followup using chair-stand performance and the WOMAC physical function subscale. Chair-stand performance (i.e., rising from a chair and sitting down 5 times) depends heavily on knee function. The protocol described by Guralnik et al (14) was used. Among elderly individuals, better chair-stand performance at baseline was associated with a decreased frequency of disability in the activities of daily living and mobility-related disability 4 years later (14). The chair-stand rate (rate of chair stands per minute, based on the time required to complete 5 repetitions of rising from a chair and sitting down) was used in the analysis. The chair-stand rate was categorized using quintiles of the MAK cohort at baseline, ranging from those with the worst chair-stand performance to those with the best, as follows: first quintile (≤5.0), second quintile (>5.0 and ≤20.4), third quintile (>20.4 and ≤23.8), fourth quintile (>23.8 and ≤27.5), and fifth quintile (>27.5). The WOMAC is a disease-specific (for knee or hip OA), validated, self-report instrument (15, 16). The possible range of the WOMAC physician function scale is 0–68, with higher scores indicating worse function. The WOMAC scores were categorized using quintiles of the MAK cohort at baseline, ranging from individuals with the best WOMAC function scores to those with the worst function, as follows: first quintile (WOMAC 0–7), second quintile (WOMAC 8–14), third quintile (WOMAC 15–22), fourth quintile (WOMAC 23–33), and fifth quintile (WOMAC >33). For both rate of chair-stand performance and WOMAC function, poor outcome was defined as moving into a lower-functioning group or remaining within the 3 lowest-functioning groups between baseline and followup, as previously described (10).

Statistical analysis.

To test hypotheses dealing with structure outcome, knees with a joint space grade of 3 (severe narrowing) at baseline were excluded, because they could not progress further. Odds ratios (ORs), representing the likelihood that knees with a thrust at baseline will progress compared with knees without a thrust (referent group), and 95% confidence intervals (95% CIs) for the OR were calculated using logistic regression. Generalized estimating equations were used to validly include data from 1 or both knees of each participant (17). Analyses were adjusted for age (treated as a continuous variable), sex, body mass index (BMI; continuous variable), and pain. Because the study by Schnitzer et al dealt with adduction moment and not with varus thrust, we adjusted for knee pain intensity (continuous variable) to address potential confounding variables (18). Additional adjustment for varus alignment (continuous variable) was undertaken to explore the mechanism of the thrust effect to test our second hypothesis.

In the analysis of physical function outcome, participants were grouped as having thrust in neither knee, 1 knee only, or both knees. Adjusted ORs for poor functional outcome associated with bilateral or unilateral thrust versus no thrust (referent group), controlling for age, sex, BMI, intensity of knee pain, and varus malalignment, were estimated using multiple logistic regression.


In this study, 237 participants, of whom 7 (3%) did not return at 18 months (5 had died, and 2 could not be located), were evaluated. Of the 230 remaining subjects, 8 had severe medial joint space narrowing at baseline in both knees and were excluded, leaving 222 participants with at least 1 knee at risk for disease progression. Among these 222 participants, 163 (73%) were women, the mean (±SD) age was 68 years ± 10.7 years, the mean (±SD) height was 1.65 ± 0.09 meters, the mean (±SD) weight was 83.1 ± 17.6 kg, and the mean (±SD) BMI was 30.0 ± 6.0. The K/L score in both knees of each subject was determined, and the higher K/L score was 2 in 114 persons, 3 in 92 persons, and 4 in 16 persons. Forty-three individuals had advanced medial OA at baseline in 1 knee; exclusion of these knees left 401 knees as the sample for this study. These 401 knees included 169 knees (42%) with varus alignment (mean ± SD 3.6° ± 2.4°), 199 knees (50%) with valgus alignment (mean ± SD 3.8° ± 2.8°), and 33 knees (8%) with neutral alignment (0°).

Of the 401 knees, 67 (17%) had a varus thrust at baseline. Of these 67 knees, 54 were varus on the full-limb radiograph (i.e., >0° in the varus direction [mean ± SD 4.7° ± 2.5°]), 12 were valgus (i.e., >0° in the valgus direction [mean ± SD 2.7° ± 2.3°]), and 1 was neutral (0°). Knees with a thrust had a mean (±SD) alignment of 3.3° + 3.8° in the varus direction, while knees without a thrust had a mean alignment of 1.1° in the valgus direction (i.e., the difference in static alignment between knees with and those without a thrust was 4.4° [95% CI 3.3–5.6]). The SEM was 0.6° for limbs without thrust and 0.5° for limbs with thrust. The difference in alignment (4.4°) between knees with and those without thrust was substantially larger than the difference that could be attributed to measurement error.

Varus thrust and OA progression.

Of 401 knees, 76 (19%) had medial OA progression. Fifteen percent of knees without a thrust (49 of 334) versus 40% of knees with a thrust (27 of 67) had medial progression. As shown in Figure 2, results were comparable for knees with mild and moderate OA at baseline. Among all knees at risk for OA progression (including those with static varus, valgus, or neutral alignment), a varus thrust was associated with a 4-fold increase (age-, sex-, BMI-, and pain-adjusted OR 3.96, 95% CI 2.11–7.43) in the likelihood of medial OA progression in the subsequent 18 months. Further adjustment for severity of baseline varus malalignment led to an OR of 1.76 (95% CI 0.87–3.56). Results were not altered by adjusting for baseline disease severity or use of medication.

Figure 2.

Among knees with a Kellgren/Lawrence (K/L) score of 2 at baseline, 13% of the thrust-negative knees (25 of 192) had medial progression, compared with 39% of the thrust-positive knees (14 of 36) (P = 0.0006). In knees with a K/L score of 3 at baseline, 17% of thrust-negative knees (17 of 101) progressed, compared with 43% of thrust-positive knees (12 of 28) (P = 0.01).

Odds of progression associated with thrust in varus-aligned knees examined separately.

If the thrust effect were attributable only to static malalignment (and was not related to the dynamic worsening of malalignment during ambulation or to knee instability), then thrust should have little effect in varus-aligned knees examined separately. When only the varus knees (n = 169) were analyzed, thrust increased the odds of progression 3-fold (age-, sex-, BMI-, and pain-adjusted OR 3.17, 95% CI 1.60–6.31).

In a varus-aligned knee, a varus thrust adds stress to a medial compartment that is already stressed. We explored where, in the range of static varus malalignment, this impact was largest. An impact of a thrust was apparent in knees with varus alignment >0° and <5° (Table 1), but not in knees with varus alignment ≥5°.

Table 1. Subgroups of knees according to the presence of thrust and the severity of varus alignment*
Varus thrustVarus alignment >0° and <5°Varus alignment ≥5°
  • *

    Values are the number of knees with disease progression/total number of knees (%). The classification of static varus (for the examination of thrust impact in knee subgroups) was based on the clinical cutoff of ≥5°. Increasing the cutoff point consistently revealed that thrust appeared to have no impact in more statically malaligned knees.

Absent18/93 (19)13/22 (59)
Present10/25 (40)16/29 (55)

The adduction moment in thrust-positive and thrust-negative knees.

The quantitative gait analysis substudy included 64 participants. Twenty-three knees with advanced OA at baseline were excluded. Of the remaining 105 knees, 39 (37%) had a varus thrust. Knees with a thrust had a greater adduction moment (3.63 ± 0.66 percent body weight × height versus 2.60 ± 0.81 in knees without a thrust); this difference (1.03 [95% CI 0.64–1.42]) was significant (P < 0.0001).

The relationship of varus thrust to physical function.

Having a varus thrust in both knees versus neither knee at baseline doubled the odds of a poor outcome of chair-stand performance (as assessed from baseline to followup), although this difference did not achieve significance in analyses adjusted for age, sex, BMI, pain intensity, and varus alignment (adjusted OR 2.23, 95% CI 0.71–6.94). Varus thrust did not predict poor outcome on the WOMAC scale for physical function.


A varus thrust visualized during gait was associated with a 4-fold increased likelihood of progression of medial knee OA over the next 18 months. The mechanism of this effect relates at least in part to static malalignment. A thrust was associated with a 3-fold increased likelihood of OA progression in varus-aligned knees examined separately, suggesting that a thrust further increases the likelihood of progression over and above the risk conferred by static varus. Knees with a thrust had a greater peak adduction moment during gait.

A relationship between thrust and OA progression has not been previously reported. Disease progression occurred more frequently in knees with a thrust than in knees without a thrust, with similar proportions in mildly and moderately osteoarthritic knees examined separately. The persistence in the elevation in the odds of progression when varus knees were considered separately is consistent with the concept that elements unique to the thrust and not captured by full-limb standing alignment—such as dynamic malalignment and instability—add further to the likelihood of progression.

It is likely that the thrust effect is attributable to greater medial loading. Greater joint loading is believed to contribute to OA progression. This is difficult to substantiate within cohort studies, given the invasive methods required to directly assess load. A visible thrust suggests that efforts to stabilize the joint—either from lateral soft tissues or muscle activity—are not effectively counteracting the knee adduction moment (5); this is supported by the greater adduction moment in knees with versus those without a thrust. This finding provides some validation for gait observation for the presence of thrust. The fact that the presence of thrust at baseline significantly predicted disease progression is another validation of the method of gait observation that was applied.

As the definition of the primary outcome attests, our focus is on progression of disease in the medial compartment, the compartment stressed by a varus thrust. Of note, these results were detected despite not restricting analyses to knees with evidence of medial OA at baseline. The goal of examining the impact of thrust in knees with mild or moderate OA necessitated including knees that were osteoarthritic but that still had a normal joint space width. Until the joint space has narrowed, it is not possible to determine on radiography which knees have medial versus lateral tibiofemoral OA. The exclusion of knees with definite lateral joint space narrowing had minimal effect on our results. In any case, the admixture of (unidentifiable) knees in which lateral OA was destined to develop among the pool of knees with normal joint space width would tend to reduce the magnitude of the thrust-associated OR for progression, and the true result would be even stronger than that reported.

In varus-aligned knees, thrust was associated with a 3-fold increased likelihood of disease progression. A thrust effect was apparent in knees with varus alignment of <5° but not in those with varus alignment of ≥5°. In the neutral knee, more load is transmitted medially than laterally (19), and, with increasing varus alignment, medial transmission approaches 100% (3, 4, 20, 21). Our findings suggest that, in knees with varus alignment of <5°, medial load can still be increased further, and thrust dynamically increases the compartment load imbalance. In knees with varus alignment of ≥5°, it is likely that the vast majority of load is already being transmitted medially, and that thrust has no added impact. These results are consistent with prior results illustrating the central role of alignment in the effect of other factors, even factors that do not compound the compartment load imbalance, such as obesity (22) and strength (23).

Thrust in both knees versus neither knee was associated with a 2-fold increase in the odds of poor functional outcome as assessed by chair-stand performance; this change did not achieve significance, perhaps reflecting inadequate power for this question. There was no effect on WOMAC outcome, possibly reflecting lower sensitivity of a self-report measure versus that of a task performed under loaded conditions. We previously observed that static malalignment (≥5°) in both knees (versus more neutral alignment) was associated with a poor functional outcome as assessed by chair-stand performance (2).

Although this is the first report of the relationship between thrust and OA disease progression, prior reports have described the relationship between the adduction moment and worse outcome and are consistent with our results (24–26). Patients with a high adduction moment had a worse outcome of high tibial osteotomy, as determined by the physician's global assessment, compared with patients with a low adduction moment (24, 25). Miyazaki et al reported that a greater adduction moment at baseline was associated with a greater risk of disease progression in patients with knee OA (26).

Identifying risk factors for progression is strategically important. First, such risk factors may be a target for disease-modifying interventions. Second, they may modify the effect of potential disease-modifying drugs; addressing them may enhance the response to drugs. Third, their presence defines subsets of individuals at higher risk, which is useful at the individual level in terms of care and at the public health level in terms of the development of programs to prevent progression. Ease of assessing a risk factor is a critical practical advantage.

These results have several implications. Gait observation cannot replace quantitative gait analysis. Nevertheless, because gait observation requires no equipment or radiography, it identifies a subset of persons at high risk for OA progression, with a simplicity and potential for widespread clinical application that full-limb radiography (with its pelvic radiation exposure, unit and equipment requirement, and cost) and quantitative gait analysis (with its equipment and gait laboratory requirement, and cost) cannot approach. It would not be feasible to perform quantitative gait or motion analysis in the clinic given the laboratory equipment requirements and cost.

These results introduce the possibility that, for prognostic purposes, the presence of a thrust may influence the necessity of obtaining full-limb radiographs; this needs to be examined in future studies. Specific treatments that address both the medial-to-lateral compartment load imbalance and the dynamic instability of a thrust should be developed and tested, because they may delay knee OA progression. Until this occurs, referral of patients with a thrust to physical therapists and orthotists should be considered, for evaluation for specific exercise, stabilizing orthoses, and perhaps for a lateral wedged insole, which has been shown to reduce the force of the varus thrust (27).

In conclusion, varus thrust is a potent risk factor, identifiable by simple gait observation, for OA progression in the medial tibiofemoral compartment. Varus thrust may also predict poor physical function outcome. The mechanism of the thrust effect on disease progression relates at least in part to the severity of static varus. A varus thrust increased the odds of progression among varus-aligned knees considered separately, suggesting that knees with a thrust are a subset of varus-aligned knees at particularly high risk for OA progression.