Examination of exercise effects on knee osteoarthritis outcomes: Why should the local mechanical environment be considered?



Osteoarthritis (OA) is the most common form of human arthritis. Almost every older adult has some evidence of radiographic OA: 905 of 1,040 participants aged 55–65 years in the Rotterdam study had definite radiographic OA in the hands, knees, hips, or spine (1). Although less prevalent than radiographic OA, symptomatic OA is a leading cause of chronic disability. It is estimated that 12% of Americans between ages 25 and 75 years have clinical signs and symptoms of OA (2).

Disability due to OA is largely a result of knee or hip involvement. The risk of disability attributable to knee OA alone is as great as that due to cardiovascular disease and greater than that due to any other medical condition in elderly persons (3). Beyond its direct effect, knee OA synergistically increases the risk of disability originating from other medical conditions (4).

Exercise is widely used in the management of knee OA, based on the results of predominantly short-term trials that demonstrate a clear beneficial effect on both symptoms and physical function. Less is known about the effect of exercise on structural outcome—i.e., new OA development and progression of established OA—or on risk of chronic disability.

Buckwalter and Lane propose that exercise (especially exercise that involves repetitive impact or torsion) may have a more deleterious effect in joints that have a local anatomic or physiologic impairment (5). From the perspective of the local environment, OA knees are heterogeneous. Local mechanical factors may be thought of as factors that have a clinically important biomechanical effect, may be linked to joint anatomy or physiology, and are specific to joint site. In published investigations, exercise effects have commonly been examined in samples narrowed to include subjects with OA predominantly at the joint site of interest, e.g., the knee. To date, minimal attention has been devoted to the examination of whether exercise effects differ between subsets of OA knees based upon local mechanical factors.

Statement of the Problem

In studies of arthritis that deal with the effect of specific exercise programs, investigators have endeavored to examine homogeneous samples of subjects. It is generally believed that homogeneity has been achieved when the sample includes subjects with a single type of arthritis, at a single joint site of predominant involvement. These criteria for homogeneity are essential but may not be sufficient: a single type of arthritis (i.e., OA) even at a single joint site (the knee) is a heterogeneous condition in terms of the local joint-organ environment.

The same exercise program may have a dissimilar effect on different knee subsets based on local mechanical environment. Such variation in effect would dilute the detected impact of the intervention on study outcomes in therapeutic trials that consider the gamut of OA knees together, and may explain in part the relatively modest impact of exercise interventions on physical function in many trials. Ultimately, exercise interventions for those with knee OA may be more effective if they are tailored to knee subset.

Review of the Literature

There is increasing awareness that the profiles of determinants of symptoms, physical function, and structural disease outcomes are not identical in OA; the effect of any intervention on each of these outcomes needs to be specifically examined. In the exercise literature dealing with knee OA, several studies demonstrate a beneficial effect of exercise on symptoms and on patient-centered physical function. In terms of structural outcome, there is minimal information on either risk of incident or progressive OA. Nonoccupational physical activity in general (with the exception of certain elite athletic activities) was not associated with an increase in the risk of newly occurring, knee OA in most studies in which this has been examined (6).

The lack of attention to the effect of exercise on disease progression is multifactorial. Even with x-ray acquisition and measurement protocols that increase the ability to detect change, the study duration required to gauge the effect of an exercise intervention on radiographic disease progression may be prolonged and the necessary sample size large. Magnetic resonance imaging (MRI)-based outcome measures may be superior in terms of reducing study duration and sample size but are expensive. Ensuring compliance with any intervention in a therapeutic trial over a long followup is challenging; this challenge can be even greater with exercise interventions.

Despite these challenges, the effect of exercise on incident OA, OA disease course, and disability each warrant further study. In theory, specific exercise programs might have some disease-modifying effect (i.e., delay progression or worsening of OA) via beneficial actions on joint tissues and the potential to enhance joint-protective mechanisms. Manipulating the local mechanical environment is relatively untapped as a strategy to delay disease progression. In addition, specific exercise programs may differ in their impact on physical function and on disability risk.

It is unlikely that a single exercise program will serve the heterogeneous process labeled “knee OA.” In the evaluation of exercise intervention on structure and disability outcomes, the local mechanical environment needs to be considered for at least 2 reasons: To stratify response to broadly applied exercise approaches according to presence/absence or severity of local impairment; and to develop exercise interventions that are tailored to knee OA subset.

The biomechanics literature and clinical OA literature offer insight into which local factors to consider in the derivation of subsets of OA knees. Malalignment and laxity are local impairments that are common in knee OA though of variable severity, influence load distribution at the knee, and can be measured relatively easily in clinical settings. This is not equivalent to subsetting according to OA disease stage or severity. In many individuals, abnormalities in alignment and/or laxity may be present prior to or at early stages of OA. Also, specific features of OA disease may have opposing effects on a given local impairment (e.g., as described for laxity below).

It is likely that the effect of an exercise intervention will differ between knees that are malaligned versus closer to neutral in alignment, between lax versus stable knees. These are subsets that differ in structural vulnerability to the forces that develop during exercise or physical activity.

Examining exercise interventions according to knee alignment.

Alignment (i.e., the hip-knee-ankle angle) is a key determinant of load distribution at the knee. In theory, any shift from a neutral or collinear alignment of the hip, knee, and ankle affects load distribution at the knee (7). In a varus knee, the load-bearing axis passes medial to the knee, and a moment arm is created that increases force across the medial compartment. In a valgus knee, the load-bearing axis passes lateral to the knee, and the resulting moment arm increases force across the lateral compartment. Biomechanical studies support that varus and valgus alignment increase medial and lateral load, respectively (7–9). Also, severity of varus alignment correlates with the ratio of medial to lateral bone mineral density in patients with OA, i.e., greater density in the higher load-bearing region (10).

During gait, disproportionate transmission of load to the medial compartment results from a stance-phase adduction moment (11). The adduction moment reflects the magnitude of intrinsic compressive load on the medial compartment (12). Varus-valgus alignment is a key determinant of this moment. Varus alignment further increases medial load during gait (13). Valgus alignment is associated with an increase in lateral compartment peak pressures (9); however, more load is still borne medially until more severe valgus is present (14, 15).

In theory, varus and valgus alignment may each be both a cause and result of progressive knee OA. Varus or valgus alignment that predates knee OA may be due to genetic, developmental, or posttraumatic factors. Animal model data support a link between preexisting varus or valgus alignment and OA development (7). Knee alignment that results from knee OA may be due to loss of cartilage and bone height.

A large body of literature gives evidence that preoperative alignment is a determinant of the outcome of surgical procedures involving the knee (e.g., arthroplasty, osteotomy, complete or partial meniscectomy, meniscal debridement). In the operated knee, the development or progression of OA is linked to several factors not at play in the examination of natural progression (e.g., nature of surgery, stage of OA at time of surgery, complications). Considerably less attention has been paid to the role of knee alignment in the nonsurgical, natural evolution of knee OA. Few longitudinal studies have dealt with alignment and the natural history of OA.

Schouten et al found that subject recollection of bow legs or knock knees in childhood was associated with a 5-fold increase in the risk of OA progression (odds ratio [OR] 5.13, 95% confidence interval [95% CI] 1.14–23.1) over a 12-year period, after adjusting for age, sex, and body mass index (BMI) (16). In another study involving patients from a hospital practice who had not undergone surgery, and in whom alignment was considered at the end of followup, 50% of 35 varus knees had progressive joint space narrowing (17).

We recently reported on the effect of varus and valgus alignment measured at baseline on subsequent progression of medial and lateral tibiofemoral OA, respectively, in 240 community-recruited subjects with knee OA (with definite osteophyte presence and at least a little difficulty with physical function) (18). Alignment was measured as the angle made by the intersection of the femoral and tibial mechanical axes from a full-limb radiograph. Knees with grade 3 (most severe) joint space narrowing at baseline were excluded.

First, the relationship between baseline varus alignment (varus in degrees as a positive value, neutral 0, and valgus in degrees as a negative value) and magnitude of decrease in medial joint space width (from baseline to 18 months on semiflexed, fluoro-confirmed knee radiographs corrected for magnification error), each as a continuous variable, was examined in the dominant knee of each subject using linear regression. Severity of varus alignment correlated with the magnitude of loss of medial joint space width (R = 0.52, P < 0.0001). Similarly, the relationship between baseline valgus alignment and magnitude of decrease in lateral joint space width was examined. Severity of valgus alignment correlated with the magnitude of loss in lateral joint space width (R = 0.35, P < 0.0001). These relationships persisted in analyses adjusted for age, sex, and BMI.

Second, the relationship between baseline alignment and compartment-specific progression, defined as a 1 grade increase in grade of severity of joint space narrowing was examined. Odds ratios were calculated using logistic regression and generalized estimating equations methodology to include data from one or both knees of each subject.

As shown in Figure 1, varus versus nonvarus at baseline was associated with a 4-fold increase in the odds of medial progression after adjusting for age, sex, and BMI. In calculating risk in varus versus nonvarus knees, we recognized that medial OA may be associated with varus, valgus, or neutral alignment. Therefore, the risk associated with varus alignment was compared with the risk conferred by any other possible alignment for a given knee. To determine the progression risk associated with varus alignment when the comparison group had neutral or nearly neutral knees, we repeated the analysis with a referent group consisting of neutral (0°) or mildly valgus (≤2°) knees. Varus alignment was still associated with a 3-fold increase in risk of medial progression in adjusted analyses (Figure 1).

Figure 1.

The odds of medial progression conferred by varus alignment are presented for 2 reference groups: nonvarus knees and neutral/mild valgus knees (i.e., neutral or ≤ 2° valgus). The unadjusted odds ratio (OR) for varus versus nonvarus (referent) was 5.00, 95% confidence interval (95% CI) 2.77–9.02; and for varus versus neutral/mild valgus (referent) 3.54, 95% CI 1.85–6.77. The adjusted (for age, sex, and body mass index [BMI]) OR for varus versus nonvarus was 4.09, 95% CI 2.20–7.62, and for varus versus neutral/mild valgus 2.98, 95% CI 1.51–5.89. Knees with grade 3 (most severe grade) joint space narrowing in either the medial or lateral compartment at baseline were excluded. The number for analysis involving the first reference group was 381 knees. The number for analysis involving the second reference group was 281 knees. —•— = varus versus nonvarus; −−•−− = varus versus neutral/mild valgus.

Valgus versus nonvalgus (referent) alignment at baseline was associated with an almost 4-fold increase in the odds of lateral progression during the subsequent 18 months (Figure 2). This relationship persisted after adjustment for age, sex, and BMI. When the referent group had neutral or nearly neutral (≤2° varus) knees, valgus alignment was associated with a more than 3-fold increase in the odds of subsequent lateral OA progression (Figure 2). It is likely that malalignment and OA progression are in a vicious cycle. The results of this study support the concept that, whether a given alignment precedes or results from OA disease, malalignment may contribute to subsequent progression.

Figure 2.

The odds of lateral progression conferred by valgus alignment are presented for 2 reference groups: nonvalgus knees and neutral/mild varus knees (i.e., neutral or ≤ 2° varus). The unadjusted odds ratio (OR) for valgus versus nonvalgus (referent) was 3.88, 95% confidence interval (95% CI) 1.82–8.24; and for valgus versus neutral/mild varus (referent) 3.23, 95% CI 1.30–8.05. The adjusted (for age, sex, and body mass index [BMI]) OR for valgus versus nonvalgus was 4.89, 95% CI 2.13–11.20; and for valgus versus neutral/mild varus 3.42, 95% CI 1.31–8.96. Knees with grade 3 (most severe grade) joint space narrowing in either the medial or lateral compartment at baseline were excluded. The number for analysis involving the first reference group was 381 knees. The number for analysis involving the second reference group was 278 knees. —•— = valgus versus nonvalgus; −−•−− = valgus versus neutral/mild varus.

The burden of malalignment at baseline also predicted deterioration in physical function between baseline and 18 months (18). Subjects were classified into 1 of 3 groups, i.e., subjects who had no knees with alignment >5°; 1 knee with alignment >5°; both knees with alignment >5°. Physical functional outcome was analyzed as a continuous variable, i.e., baseline to 18 month change in chair stand rate (time required to complete 5 chair stands, converted to a rate or number of stands per minute). As shown in Table 1, change did not differ between the first 2 groups. However, a significantly greater deterioration in chair stand performance was found in subjects having both knees versus no knees with alignment >5°. The difference between these groups persisted after adjusting for age, sex, BMI, and pain.

Table 1. Alignment group differences in the baseline to 18-month change in chair stand performance*
Groups based on malalignment burdenDifference between groups (95% CI)
UnadjustedAdjusted for age, sex, and BMIAdjusted for age, sex, BMI, and pain
  • *

    95% confidence interval (95% CI) that excludes 0 represents a significant difference between groups. Subjects with 2 knees with alignment >5° had significantly greater deterioration in chair stand rate between baseline and 18 months versus subjects with 0 knees with alignment >5°, in unadjusted and adjusted analyses. BMI = body mass index.

One knee >5° versus no knee >5°0.48 (−1.40, 2.36)0.43 (−1.44, 2.31)0.17 (−1.66, 2.01)
Two knees >5° versus no knee >5°2.88 (0.75, 5.01)2.73 (0.52, 4.94)2.23 (0.05, 4.41)

We also explored the relationship between burden of malalignment and functional decline, designating decline as ≥20% worsening in rate of chair stand performance. Thirty-four subjects (16% of the 215 subjects able to perform the test at baseline) experienced functional decline by this definition. The proportion of subjects experiencing decline steadily rose as the burden of malalignment increased from none to 1 to 2 knees. The odds of functional decline were doubled (OR 2.33, 95% CI 0.97–5.62) by having one knee with alignment >5° versus no knee with alignment >5°, and were tripled by having both knees with alignment >5° versus no knee with alignment >5° (OR 3.22, 95% CI 1.28–8.12).

In summary, there is biomechanical evidence that alignment influences load distribution and longitudinal evidence that this biomechanical effect is clinically relevant in knee OA structural and functional outcome.

Examining exercise interventions according to knee laxity.

Stability is an important component of the mechanical environment of any joint. Knee laxity may be broadly defined as abnormal displacement or rotation of the tibia with respect to the femur (19). In the unloaded state, knee stability is provided by the ligaments, capsule, and other soft tissues, and in the loaded state by interactions between these tissues, condylar geometry, and contact forces generated by muscle activity and gravitational forces (19). Dynamic stability also depends upon proprioceptive input and reflex and centrally driven muscle activity (20). During normal motion, ligament stiffness is low. In the setting of large forces, soft tissue stiffness increases to limit displacement between the femur and tibia, protecting cartilage and other tissues from injury (21).

As clinically assessed, joint laxity represents an impairment for which muscle activity may or may not be able to compensate. Laxity results in more abrupt joint motion with larger displacements. Deleterious effects of laxity include alteration of the congruence and regions of contact of the opposing articular surfaces, and an increase in shear and compression forces on some regions of the articular cartilage (22). Bruns et al demonstrated that ligament division in cadaver knees resulted in further increases in peak articular pressure, even in the presence of severe malalignment (9). Such alterations in pressure may lead to cartilage damage, lessening the subsequent ability of cartilage to withstand stress (21).

In individuals without arthritis, frontal plane or varus-valgus laxity may reflect primary capsuloligamentous laxity (related to genetic factors or aging-related soft tissue changes) or prior injury. In knees with moderate to severe OA, laxity may be due to loss of cartilage and/or bone height, chronic capsuloligamentous stretch, or combinations of ligamentous, meniscal, muscular, and capsular pathology. The ligaments and the menisci of the OA knee develop fraying and cracking similar to what is seen in articular cartilage (23).

The paucity of clinical information on varus-valgus laxity in knee OA relates in part to the absence of measurement systems. In clinical settings, varus-valgus laxity is most commonly assessed by physical exam, an unreliable approach (24, 25). Sources of variation during the physical exam test have been identified as inadequate immobilization of the thigh and ankle, incomplete muscle relaxation, variation of the knee flexion angle, variation of load applied, and imprecise measures of rotation with load application (24–26). Devices to measure anterior-posterior (AP) laxity are commercially available.

Aging is associated with alterations in ligament properties. Ligament stiffness and ultimate load decreased substantially with specimen age in a study of human femur-anterior cruciate ligament (ACL)-tibia complex involving 3 groups, i.e., younger (22–35 years), middle (40–50 years), and older (60–97 years) (28). Ultimate load was more than 300% higher in the younger group than in the older group. In subjects without clinical or radiographic evidence of OA, a modest correlation between varus-valgus laxity and age has been described (27). Such age-related changes may be intensified by anatomic factors, by patterns of use, and by comorbid conditions.

The knee injury literature provides some evidence of the clinical importance of laxity. OA develops in a canine model by inducing an unstable joint via complete ACL transection. Lundberg and Messner found that while the majority of those with isolated medial collateral ligament (MCL) injuries (grades I–II) did not develop OA by 10 years, combined injury to the MCL and the ACL led to OA in close to 50% of patients (29). Kannus reported that 50% of those with a grade III sprain of the lateral collateral ligament developed OA within 8 years of injury, and that 63% of those with grade III sprains of the MCL developed OA within 9 years (30, 31). Attention to concomitant injury to tissues other than ligaments varies between injury studies.

There is evidence that a portion of the varus-valgus laxity present in OA knees predates the development of full-blown disease. In support of this concept, we found that varus-valgus laxity was greater in subjects with knee OA—even in their uninvolved knee or their mildly involved knee—than in older subjects without any clinical or radiographic evidence of knee OA (27). These differences persisted after adjusting for age and sex. Although Brage et al did not statistically compare the knees with mild OA with the knees of older control subjects, mildly arthritic knees in their study appear to be more lax than the knees of the control subjects (32).

Although some portion of the varus-valgus laxity of idiopathic OA appears to predate the development of full-blown disease, specific aspects of the disease itself exacerbate the problem. Varus-valgus laxity increased as joint space decreased, and was greater in knees with evidence of bony attrition (27). This is presumably related to the points of ligamentous attachment to the femur and tibia moving closer together as a result of loss of bone and cartilage height.

It is likely that osteophytes prevent laxity to some extent, as demonstrated by Pottenger et al, who measured varus-valgus laxity before and after intraoperative osteophyte removal in patients with advanced knee OA (33). Given their findings, it is possible that, at earlier stages of OA, osteophytes make some contribution to varus-valgus stability. With progressive disease, loss of cartilage and bone height appear to override this stabilizing effect. At advanced stages, although osteophytes continue to have some stabilizing activity, they cannot prevent further increases in varus-valgus laxity. The opposing effects of specific features of OA on varus-valgus laxity may have contributed to the mixed results seen when studies have relied on global radiographic assessment of OA status.

Using a KT1000 arthrometer, no relationship between AP laxity and age or sex in subjects without OA was detected (27). The AP laxity did not differ between subjects with OA and controls, and was not associated with specific features or global grade of OA severity. In studies using the Genucom computerized measurement system, AP laxity declined with increasing severity of OA. In one study including arthroscopic examination, the decline in AP translation was noted in spite of the fact that among those with severe OA, ACLs were absent or torn in the majority of subjects (34). The ACL type did not predict AP translation. Joint stiffness due to capsular changes or osteophytic growth may override the cruciate ligament insufficiency that can occur in progressive knee OA.

We have found that varus-valgus laxity, i.e., the sum of right and left knees, was associated with physical function in cross-sectional analyses (35). Physical function was worse in subjects with high laxity (Western Ontario and McMaster Universities Osteoarthritis Index physical function score 26.5 ± 13.3) than low laxity (20.8 ± 13.8; P = 0.008). There is evidence that varus-valgus laxity may mediate in the relationship between muscle strength and physical functioning in patients with knee OA (35). Laxity necessitates that greater muscular work be directed toward joint stabilization. We found that greater laxity was consistently associated with a weaker relationship between strength (quadriceps or hamstring) and physical function (self reported or observed). These results raise the possibility that muscle strengthening may have less impact on physical function in high laxity than low laxity knees, and that addressing varus-valgus laxity may improve the outcome of strengthening intervention.

In summary, there is biomechanical evidence and cross-sectional clinical evidence that varus-valgus laxity is a key local factor in the course of knee OA.


Knees with OA are heterogeneous in terms of the local mechanical environment. Differences in the local environment explain, in part, interindividual variation in the rate of OA disease progression. The same exercise program may have a dissimilar effect on different knee subsets based upon local environment. Malalignment and laxity are local factors to consider in the derivation of subsets of OA knees. These factors are suggested as a first step; the list of relevant mechanical factors to apply toward developing subsets will extend beyond the factors described here.

There is evidence that alignment influences load distribution at the knee, is a determinant of surgical outcomes, increases the risk of natural progression of knee OA in the expected compartment-specific fashion, and increases the risk of functional decline in persons with knee OA. There is evidence that varus-valgus laxity has deleterious biomechanical effects that may lead to cartilage damage, increases with age, is associated with a greater risk of OA in the setting of ligament injury, is present in OA patients to some extent before full-blown disease, is made worse by specific aspects of OA disease, is associated with worse physical function, and alters the strength/function relationship.

The effect of exercise or physical activity may differ between knees that are malaligned versus closer to neutral in alignment, between lax versus more stable knees. Malalignment and laxity may reduce the impact of generic exercise programs on functional outcome, and may change the effect of exercise on structural outcome.

These findings point to areas for future research. A profile of key local mechanical factors for each joint site should be identified. In exercise trials of knee OA, investigators should consider stratifying results according to alignment (if feasible considering both direction and severity of malalignment) and varus-valgus laxity, as well as other key mechanical factors. Certain knee subsets (e.g., malaligned knees or lax knees) are likely to benefit from a more tailored approach; exercise interventions should be developed for these subsets. Exercise trials should strive to include structure and disability outcomes. Interventions (e.g., orthotics) directed towards laxity- and malalignment-associated stresses on the knee should be further developed and studied. Methods to assess laxity and alignment in clinical settings should be developed.