The effects of impaired joint position sense on the development and progression of pain and structural damage in knee osteoarthritis




Although cross-sectional studies have reported impaired proprioceptive acuity in people with osteoarthritis (OA), there have been no longitudinal studies to evaluate whether those with such impairments increase the risk of OA or its worsening.


We studied subjects from the Multicenter Osteoarthritis Study study, a longitudinal study of people with or at high risk of knee OA. At baseline, we quantified acuity as the amount of a subject's error when attempting to reproduce a test knee flexion angle (a measure of joint position sense). We tested proprioception 10 times in the right leg and used a person's worst score as their proprioceptive acuity. At baseline and the 30-month followup, we assessed the presence of frequent pain, obtained Western Ontario and McMasters Universities OA Index (WOMAC) scores, and acquired posteroanterior and lateral weight-bearing knee radiographs read for Kellgren/Lawrence grade and individual radiographic features. We examined the relation of baseline proprioceptive acuity in quartiles with baseline knee pain (frequent pain yes/no), WOMAC pain score, self-reported physical function, and radiographic OA, and with changes from baseline in pain, physical function, and radiographic OA adjusted for age, sex, body mass index, and quadriceps strength.


At baseline, proprioceptive acuity was associated with the presence and severity of knee pain but not with the presence of radiographic OA. However, among the 2,243 subjects with baseline acuity assessments and 30-month followup, there were no strong associations between proprioceptive acuity and development of adverse OA outcomes. Acuity was not significantly associated with the new onset of frequent knee pain. Those with the worst acuity at baseline had slightly greater worsening of WOMAC pain scores (0.47 on a 20-point scale) and physical function scores (by 1.5 points on a 0–68-point scale) compared with those with the best proprioceptive acuity, whose pain and physical function score deteriorated less (for pain P = 0.05; for physical function P = 0.02). Radiographic worsening was not significantly associated with proprioceptive acuity.


Proprioceptive acuity as assessed by the accuracy of reproduction of the angle of knee flexion had modest effects on the trajectory of pain and physical functional limitation in knee OA.


Proprioception is the conscious perception of body position, loading, and movement (1). The physiologic systems that contribute to proprioceptive acuity include visual and vestibular systems and articular, cutaneous, and muscle mechanoreceptors. Input from these systems are integrated with other information and used to plan and effect finely controlled muscle contraction enabling smooth, coordinated movement.

There is some evidence that impaired proprioceptive acuity may be involved in the pathogenesis of osteoarthritis (OA). In animals subjected to joint trauma, impaired sensory input increases the risk of developing OA (2). In humans, proprioceptive deficits are greater in people with knee OA compared with people of a similar age without disease (1, 3–6) These deficits are also seen in the contralateral legs of people with unilateral OA who have a high risk of developing bilateral OA (7), and in people with Charcot's arthropathy, severe impairment of position sense accelerates joint degeneration after minor joint injury. Proprioception deficits may result in poorly controlled, excess loading to the knee during gait, initiating or accelerating joint degeneration (6, 8, 9).

Unfortunately, the cross-sectional design of all of these studies limits our ability to investigate whether impaired proprioceptive acuity, which we shall examine as joint position sense, precedes the occurrence of OA or accelerates disease progression. In addition, all previous studies have compared people with clinical OA with age-matched healthy controls. Because clinical OA encompasses structural damage and symptoms (pain and physical functioning), these studies cannot discriminate the association between impaired proprioception and pain, disability, or structural damage.

Elucidating the role of proprioceptive deficits in the etiology and progression of OA may be important because proprioceptive acuity may be a modifiable factor (3), and because interventions that improve proprioceptive acuity might prevent or ameliorate some of the effects of OA. To investigate the role of proprioceptive deficit in the development and progression of knee OA, we undertook a prospective, longitudinal assessment of proprioceptive acuity and its relation to the trajectory of knee OA, examining radiographic change, symptoms, and disability in a large cohort of people.


Study design and subjects.

The Multicenter Osteoarthritis Study (MOST) is a large, prospective, epidemiologic study of people with OA or at high risk of developing OA (10). The goal of the MOST study is to identify risk factors for incident and progressive knee OA in a sample of older adults who either have the disease or who are at high risk of developing it. Those considered at high risk included people who were overweight or obese, had knee pain on most of the last 30 days, had a history of knee injury that made it difficult to walk for ≥1 week, or reported previous knee surgery. High risk due to being overweight was defined as present when a person weighed more than the median weight of people of the same age and sex from the Framingham OA Study (11, 12). For example, the weight cutoff for women age 50–59 years was 70 kg; for women age 60–69 years was 68 kg, and for women age 70–79 years was 67 kg. Weight cutoffs for men also varied by age. For men age 50–59 years the cutoff was 88 kg, for age 60–69 years it was 84 kg, and for age 70–79 years it was 83 kg.

Subjects age 50–79 years were recruited from 2 US communities, Birmingham, Alabama and Iowa City, Iowa, through mass mailings of letters and study brochures supplemented by media and community outreach campaigns. Each center also recruited ethnic minorities reflecting their local population. Subjects were excluded if they had rheumatoid arthritis (13), ankylosing spondylitis, psoriatic arthritis, chronic reactive arthritis, kidney disease necessitating hemodialysis or peritoneal dialysis, a history of cancer (except for nonmelanoma skin cancer), bilateral knee replacement surgery, inability to walk without the help of another person or a walker, or if they were planning to move out of the area in the next 3 years.

The study protocol was approved by the Institutional Review Boards at the University of Iowa, the University of Alabama, Birmingham, the University of California, San Francisco, and the Boston University Medical Campus.

Proprioceptive acuity.

Proprioceptive acuity relies on joint position sense, which we assessed using a protocol developed by Hurley (14). The protocol instructs a subject to attempt to reproduce a randomly chosen angle of knee flexion. Each subject is blindfolded and instructed to sit quietly with their legs dangling off the end of a raised chair. A table just above the thighs further obstructs any view of their legs. We attached an electrogoniometer to the outside of the right leg. A SG150 twin-axis electrogoniometer (Biometrics, Ladysmith, VA) was attached to the lateral aspect of their right lower leg using adhesive tape, the upper block just below and in line with the head of the fibula and the lower block just below and in line with the lateral malleolus. The goniometer was connected by leads to a handheld display unit that gave a continuous, real-time digital reading of knee flexion angle.

From this start position with their legs dangling, the subject was asked to very slowly extend their leg to a predetermined position, which they held for 5 seconds. They were then asked to relax and let their leg slowly return to the start position. After 3 seconds they were then asked to reproduce the test position, and the knee angle displayed on the display unit at this reproduced position was recorded. Because the movement was voluntary, no predefined speed could be imposed, but as an approximate idea of the speed of movement it took subjects ∼10 seconds to move their lower leg through an arc of ∼90 degrees. We repeated this process 10 times, each time at a slightly different predefined flexion angle so as to test a wide range of knee flexion, obtaining 10 measures of proprioceptive acuity from each subject using the right leg. For each trial, proprioceptive acuity was taken as the difference between the knee angles recorded from the display unit at the test position and the reproduced position.

Because our primary interest was in determining whether or not knee joints might be injured by poor proprioceptive acuity and thereby cause pain and structural damage, we chose to focus mainly on each subject's worst measured proprioceptive acuity of the 10 trials. We chose to look at this in preference to the average inaccuracy in joint position sense across all trials because a subject's worst acuity was more likely to provide an accurate characterization of their exposure to subclinical injuries. Variability in proprioception test results (15) was expected because this reflects a wide variation in human performance. Indeed, in predicting walking speed, one group of investigators has shown that the variability around the mean joint position sense predicted walking speed, whereas the mean joint position sense did not (16). We also examined the mean inaccuracy across all trials as a secondary measure of proprioceptive acuity.


At baseline and the 30-month clinic visit followup assessment, all subjects were asked the following question about frequent knee pain: “During the past 30 days, have you had pain, aching, or stiffness in your knee on most days?” We characterized a knee with no pain at baseline but with pain at 30 months as new knee pain.

At the baseline and 30-month assessment, subjects completed the Western Ontario and McMaster Universities OA Index (WOMAC) questionnaire and filled out surveys on medication use and were weighed and measured. Body mass index (BMI) was computed as weight/height2.

Structural damage on radiograph.

At baseline and at the 30-month followup, all subjects underwent weight-bearing posteroanterior (PA) fixed flexion knee radiographs using the protocol by Peterfy et al (17). Body weight was equally distributed between the 2 legs, and the great toes of the feet and the front of thighs were placed in contact with the front plate of a plexiglass frame. The external rotation of the feet was fixed at 10 degrees using a V-shaped foot angulation support on the frame. The central radiograph beam was directed to the midpoint between the back of the knees at ∼10 degrees caudal angle to allow the anterior and posterior lips of the medial tibial plateau to be optimally superimposed (film focus distance of 183 cm). Lateral weight-bearing films of each knee were also obtained at baseline and at 30 months using the Framingham OA Study protocol (18).

An experienced musculoskeletal radiologist and a rheumatologist, both blinded to clinical data, graded all of the PA films according to the scheme of Kellgren/Lawrence (K/L) and scored individual radiographic features, including joint space narrowing (JSN), on a scale of 0–3 using the Osteoarthritis Research Society International atlas (19). Lateral films were read using the protocol described by LaValley et al (18), in which joint space width is also scored from 0–3 in medial and lateral tibiofemoral and patellofemoral compartments. Atlases and protocols for anteroposterior and lateral view readings are targeted toward cross-sectional evaluations of joint space width. In preliminary readings, we found that knees often showed JSN longitudinally but did not show enough narrowing to move from one grade to the next (e.g., from 1 to 2). When the joint space was narrowed but narrowing did not reach a full grade on the 0–3 scale, readers were instructed to use half grades (20).

For knees with radiographic OA (K/L ≥2 or patellofemoral OA by radiograph), we defined radiographic progression as any increase in K/L score or joint space width score in either the tibiofemoral or patellofemoral compartment. For knees without radiographic OA, we defined incident OA as new onset of K/L grade ≥2 or patellofemoral OA (21). We counted knees that had a knee replacement at followup and that did not have a K/L grade of 4 or bone on bone JSN at baseline as having experienced progression.

Radiographs were read paired. Readers were blinded to the proprioceptive status of the participant. Each person's knee films were read by both readers working independently. If readers disagreed on the presence of progression (for knees with OA at baseline) or incidence (for knees without radiographic OA at baseline), the film readings were adjudicated by a panel of 3 readers to decide whether incidence or progression had occurred.

Statistical analyses.

As noted above, a subject's worst absolute error between the test and reproduced positions of the 10 trials of the right knee was used as his or her measure of proprioception. For example, if of the 10 attempts a subject's largest error in reproducing the referent flexion angle was 8 degrees, then their proprioceptive acuity was considered to be 8 degrees. Based on their proprioception score, each subject was placed into a quartile.

Cross-sectional and longitudinal analyses focused on the relation of proprioceptive acuity to frequent knee pain (yes/no), WOMAC pain score, and WOMAC physical function score. For 80 knees that underwent knee replacement between baseline and followup, we assumed WOMAC pain and physical function scores at followup based on mean WOMAC scores in US subjects undergoing knee replacements (22) and adjusted these scores based on the effects of age, sex, and baseline BMI on WOMAC scores in all MOST subjects at baseline. In effect, the difference between the age/sex/BMI category for each subject on the WOMAC pain/physical function was added (or subtracted) from the mean pre-knee replacement WOMAC score for those undergoing knee replacement. The relation of proprioceptive acuity to radiologic damage was tested in relation to baseline radiographic OA (yes/no), incident OA (restricted to knees without OA at baseline), and progressive OA (knees with OA at baseline). All data are concerned with the right knee only. For dichotomous outcomes, logistic regression was used, and for continuous ones (e.g., WOMAC score), multiple linear regression was used. Covariates included age, sex, BMI, isokinetic quadriceps strength (tested using the same leg and with a Cybex dynamometer [Cybex International, Medway, MA]), and others as noted. For the quartile analysis, we used the best acuity as the referent group and present odds ratios for disease (with 95% confidence intervals [95% CIs]) for each other quartile. In addition, we tested an ordinal variable (quartiles 1–4) and computed a test for trend as the significance associated with that variable.


There were no important differences in age, sex, prevalence of OA or pain, or proprioceptive acuity in subjects seen at baseline and those seen at both baseline and the 30-month followup (Table 1).

Table 1. Baseline characteristics in the whole MOST cohort who had measured proprioception and in those with full followup of proprioception*
Risk factorsTotal cohort (n = 2,440)Cohort with followup at 30 months (n = 2,243)
  • *

    Values are the percentage of participants unless otherwise indicated. MOST = Multicenter Osteoarthritis Study; BMI = body mass index; OA = osteoarthritis; TF = tibiofemoral joint; PF = patellofemoral joint; WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index.

 Age, mean ± SD years62.4 ± 8.162.4 ± 8.0
 BMI, mean ± SD kg/m230.4 ± 5.730.3 ± 5.6
 Radiographic OA ≥1 knee TF or PF51.251.5
 Frequent knee pain40.939.9
 Quadriceps strength, mean ± SD newtons92.0 ± 43.992.6 ± 44.1
 Radiographic OA39.638.2
 Frequent knee pain28.527.9
 WOMAC pain, mean ± SD score3.3 ± 3.63.2 ± 3.5
 Mean proprioception score, median degrees across subjects3.93.9
 Worst proprioception score, median degrees across subjects8.08.0

At baseline, proprioceptive acuity was associated with both pain and functional limitations (Table 2). WOMAC pain and physical functioning scores were higher (i.e., worse) in subjects who had worse proprioceptive acuity. The prevalence of knee pain at baseline increased from 25.3% among those with the best proprioceptive acuity to 31.5% among those in the worst acuity group (P = 0.03 for trend). The prevalence of radiographic knee OA was not related to proprioceptive acuity (Table 2).

Table 2. Baseline proprioceptive acuity, pain, and physical functioning*
 Quartile 1 (worst)Quartile 2Quartile 3Quartile 4 (best)P for trend
  • *

    WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index; OR = odds ratio; 95% CI = 95% confidence interval; OA = osteoarthritis.

  • Knee-specific and related to the limb tested for proprioception.

  • For all subjects, scores are for the knee tested only and are the least squares means adjusted for age, sex, body mass index (BMI), and isokinetic quadriceps strength.

  • §

    Adjusted for age, sex, BMI, and isokinetic quadriceps strength.

WOMAC scores     
 Pain3.< 0.001
Knee pain     
 Yes/no (%)156/495 (31.5)216/717 (30.1)167/607 (27.5)157/621 (25.3) 
 Adjusted OR (95% CI)§1.33 (0.99–1.78)1.32 (1.01–1.73)1.18 (0.89–1.56)1 (referent)0.03
Radiographic OA     
 Yes/no (%)191/495 (38.6)271/717 (37.8)237/607 (39.0)267/621 (43.0) 
 Adjusted OR (95% CI)§0.87 (0.67–1.12)0.79 (0.63–1.00)0.79 (0.62–1.01)1 (referent)0.22

Longitudinally, proprioceptive acuity was related modestly to changes in pain and function (Table 3). Change in WOMAC pain score over 30 months was modestly and significantly related to proprioceptive acuity at baseline (P = 0.05 for trend), although the difference in worsening of WOMAC pain scores in people with the worst acuity at baseline was modest: WOMAC pain scores increased by 0.47 on a 20-point scale in those with the worst acuity, whereas they increased by only 0.15 for those with the best acuity. Poor proprioceptive acuity was associated with worse physical function at followup. The change in physical function score worsened by 1.50 points in subjects in the worst proprioceptive acuity group at baseline compared with those with the best proprioceptive acuity at baseline, whose physical functioning worsened by 0.16 points (P = 0.03 for trend). Among subjects with no knee pain at baseline, there was no significant relation between baseline proprioceptive acuity and the risk of development of new knee pain. Subjects in the worst proprioceptive acuity quartile were only slightly (14.9%) more likely to develop new knee pain than subjects with the best proprioceptive acuity (12.0%).

Table 3. Proprioceptive acuity at baseline and development of new pain and change in WOMAC pain and physical functioning scores at the 30-month followup assessment*
Proprioceptive acuityQuartile 1 (worst)Quartile 2Quartile 3Quartile 4 (best)P for trend
  • *

    See Table 2 for definitions.

  • Knee-specific and related to the limb tested for proprioception.

  • For all subjects, scores are for the knee tested only and are the least squares means adjusted for age, sex, body mass index (BMI), and isokinetic quadriceps strength.

  • §

    Adjusted for age, sex, BMI, and isokinetic quadriceps strength.

Change in WOMAC scores at followup     
 Pain0.47 (worse)0.380.190.150.05
 Functioning1.50 (worse)1.080.670.160.02
New knee pain at followup     
 Yes/no (%)33/222 (14.9)51/358 (14.3)43/296 (14.5)38/317 (12.0) 
 Adjusted OR (95% CI)§1.33 (0.78–2.26)1.22 (0.76–1.95)1.24 (0.75–2.04)1 (referent)0.31

There was no association of proprioceptive acuity at baseline with worsening radiographic OA over time (Table 4). For subjects without knee OA at baseline, those in the worst proprioception grouping had a 1.12-fold increase in the odds of developing OA at 30 months (95% CI 0.62–2.02). For subjects with knee OA at baseline, the odds for progression were not increased.

Table 4. Proprioceptive acuity with incident or progressive radiographic OA*
Proprioceptive acuityQuartile 1 (worst)Quartile 2Quartile 3Quartile 4 (best)
  • *

    See Table 2 for definitions.

Knees without OA at baseline    
 Incident radiographic OA, yes/no (%)21/225 (9.3)25/386 (6.5)20/320 (6.3)31/347 (8.9)
 Adjusted OR (95% CI)1.17 (0.64–2.14)0.58 (0.32–1.05)0.70 (0.38–1.28)1 (referent)
Knees with OA at baseline    
 Radiographic progression, yes/no (%)122/213 (57.3)114/214 (53.3)128/219 (58.5)203/358 (56.7)
 Adjusted OR (95% CI)0.99 (0.68–1.43)0.93 (0.65–1.33)1.05 (0.73–1.51)1 (referent)

We focused on the worst acuity displayed by a person during their testing because we felt this was more valid than using the mean score. We carried out additional analyses using the mean score and findings were similar. The mean inaccuracy in joint position sense was significantly related to WOMAC pain and physical function scores at baseline after adjustment for age, sex, BMI, and quadriceps strength. This mean inaccuracy was also modestly related to later change in WOMAC pain score (P = 0.055 for trend), with the worst category experiencing an increase in WOMAC pain score of 0.63 versus an increase of 0.16 among those in the best category. For WOMAC physical function scores, mean inaccuracy was associated with worse later function (P = 0.07 for trend); those in the worst category had an increase of 1.96 in score versus an increase of only 0.40 in those in the best category. The differences between the worst and best acuity groups (unlike the test for trend) showed significant differences for both WOMAC pain and physical function scores. Similar to our measure of worst acuity, mean acuity was not related to radiographic OA.


To our knowledge, this is the first longitudinal study of proprioceptive acuity as a risk factor for the initiation and progression of pain and structural damage in knee OA. Our baseline findings confirm previous cross-sectional studies reporting a relationship between poor proprioceptive acuity and worse pain and physical functioning. Although our longitudinal findings showed that people with greater proprioceptive deficits were at an increased risk of deteriorating physical function and more severe pain, these effects were modest and their clinical relevance was unclear. We did not find that proprioceptive deficits affected the risk of new OA or new knee pain. Our data therefore suggest that proprioceptive deficits, at least as assessed by a person's ability to reproduce a knee flexion angle, are not a major risk factor for OA.

The major strengths of this study are its size and longitudinal design. Our failure to find that proprioceptive acuity had a major effect on knee OA outcomes is unlikely to be a consequence of inadequate power because this study was a large study, the largest yet undertaken, and 95% CIs were narrow, especially for baseline analyses and even for radiographic progression. When we looked at the average inaccuracy in joint position sense across all 10 trials in each individual, we found similar inaccuracy (3.9 degrees) as has been reported in other studies (23, 24). This strengthens our confidence in the generalizability of our inferences.

However, several limitations of the current study may mask any effect of proprioceptive deficit. The most important limitations are the difficulty in measuring proprioception and the insensitivity of the tools used to measure the outcomes. Accurate proprioception is essential for planning, executing, and monitoring safe, efficient movement. To ensure that the body has an accurate, real-time picture of what is happening, it collects, collates, and assimilates sensory information from several overlapping physiologic systems. If one physiologic system is compromised, others compensate to minimize sensory deficits. This makes measuring proprioception problematic (14, 15, 25) and could have prevented us from detecting associations between proprioceptive deficits.

There are several ways of measuring proprioceptive acuity. One often used is the threshold detection of passive movement, but passive movements do not reflect real life movement or function. We estimated people's ability to replicate limb position using active movement because this maximizes sensory input to the central processing systems and replicates normal movement that is almost always active, and as such is a closer measure of real life proprioceptive acuity. The technique does require concentration and cognitive skills by the subjects and if these skills are compromised, this will interfere with the accurate estimation of proprioception. Other investigators have suggested that there is a poor correlation between different approaches to proprioception assessment (26).

Similarly, radiographs are an insensitive measure of structural joint damage, so the poor correlation between proprioception and joint damage might be partially due to insensitive outcome measures. Because radiograph damage is weakly related to function loss and pain in OA, one might argue that any factor related to pain and function in OA might not be expected to correlate with radiographic OA. In addition, because OA is a slowly progressive condition, despite this study involving a relatively long followup it may still have been too short to see changes in structural damage and proprioception.

Many factors may combine to worsen a patient's course, and proprioceptive deficits may be only one of them, and one that may not have a unique independent effect. People may accommodate for (sub)conscious proprioceptive decline by adapting their behavior. Therefore, impaired proprioception may explain why people with symptomatic OA walk more slowly and with longer double limb stance to avoid risk of joint injury and prevent worsening of the disease (4, 27). These compensatory mechanisms could also explain the lack of association between poor proprioception and progression of OA. The modest association of joint position sense with function loss and pain worsening may not relate to OA progression at all, but rather to poor motor control and muscle function in those with impaired proprioception, and the contribution of these with pain and function.

Subjective assessment of pain and physical function, as measured by the WOMAC, is influenced by many psychological factors, traits, and emotions. Therefore, a weak relationship between structural articular damage, pain, and impaired proprioception is not surprising, and reflects the well-described lack of association between structural damage, pain, and disability. Including an objective measure of physical function might have provided a better comparison to assess the effect of proprioception on physical function.

The severity of the condition may affect the association between proprioception and outcomes measures (3). The baseline pain score (mean WOMAC score of ∼3 on a 0–20 scale where 0 = best and 20 = worst), physical functioning score (WOMAC score of ∼14 on a 0–64 scale where 0 = best and 64 = worst), and percentage of people with radiologic damage suggest that the cohort recruited into the MOST study from the community had mild disease. In early stages of disease when overlapping protective strategies function well, proprioception deficits may not emerge as identifiable risk factors for disease. Perhaps it is only when disease is further advanced and all protective strategies are impaired that proprioception deficits play a critical role.

In summary, poor proprioceptive acuity as assessed by joint position sense is related cross-sectionally with the presence of knee pain and its severity and worse physical functional limitations. However, it is not strongly related to risk of later outcomes: we found no significant relations with new knee pain or radiographic worsening. We did find that, compared with those with good proprioceptive acuity, people with poor acuity had more deterioration in physical function and worse pain over time, but the associations were modest.


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. Dr. Felson 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. Felson, Gross, Nevitt, Torner, Hurley.

Acquisition of data. Nevitt, Yang, Torner, Lewis.

Analysis and interpretation of data. Felson, Nevitt, Yang, Lane, Hurley.