Abstract
Objectives. Knee osteoarthritis (OA) is a major cause of disability, particularly in the elderly. The factors determining disability remain unclear. The aim of this study was to assess the impact of clinical and psychosocial variables on function in knee OA and to develop models to account for observed variance in self‐reported disability.
Methods. The subjects (n = 69) were hospital out‐patients. Self‐reported disability was measured by the Western Ontario and McMaster Universities (WOMAC) OA index. Pain was measured by the WOMAC and the McGill pain questionnaire. Depression, anxiety, helplessness, self‐efficacy, fatigue and quality of life were measured by standard instruments. A detailed knee examination, including pain threshold by dolorimetry, was performed. Radiographs were scored for individual features.
Results. Pain severity, obesity and helplessness were the most important determinants of disability: a model including these variables accounted for 59.9% variance in WOMAC disability. Anxiety remained associated with disability in some models. Disability was unrelated to radiographic change.
Conclusions. Function in symptomatic knee OA is determined more by pain and obesity than by structural change, at least as seen on plain X‐ray. Our study provides further support for interventions targeting anxiety and helplessness in knee OA.
A number of cross‐sectional community studies [
2,
7,
8] found a relationship between radiographic disease severity and physical function. Radiographic status at baseline has also been shown to predict future disability in longitudinal studies [
9–
11]. However, knee pain is a better predictor of disability than radiographic change [
2,
7,
9,
11] and, in multiple models, the association between disability and structural change may disappear altogether when pain is included [
8,
12]. Studies in patients presenting to medical care are fewer, but in general confirm the findings of community studies [
13–
16].
The aim of this study was to examine the determinants of self‐reported function in out‐patients with symptomatic knee OA. The factors studied included not only clinical measures but also psychosocial and personality traits which, if shown to be related to disability, might be amenable to multi‐disciplinary interventions.
Patients and methods
The subjects (
n = 69) were consecutive out‐patients with a clinical diagnosis of knee OA made by a rheumatologist. All subjects fulfilled American College of Rheumatology (ACR) clinical criteria for knee OA [
20] and all had ‘current knee pain’, as defined by a score of at least 2 on an 0–10 scale (where 0 = ‘no pain’, 10 = ‘pain as bad as I can imagine’) by preliminary telephone screening. Patients were excluded if they had a total knee replacement, significant hip or spinal arthritis or major concurrent illness. Inability to attend hospital was not an exclusion criteria as those subjects were interviewed at home. Local ethical committee approval and informed written consent were obtained.
Demographic data [age, gender, race, body mass index (BMI), duration of disease, years of formal education] were obtained. Self‐reported difficulty with tasks was measured by the function subscale of the Western Ontario and McMaster Universities (WOMAC) OA index [
21]. This is a well‐validated scale designed to reflect the problems experienced by individuals with lower limb OA. It asks subjects to rate their difficulty with tasks on a visual analogue scale from 0 to 100, with higher scores indicating greater difficulty. The 17 tasks are listed in Table 1
⇔. The mean of responses to these 17 questions is taken as the overall disability score.
The worse knee, as selected by the patient, was the ‘index’ knee. Pain severity was measured using two methods: the pain section of the WOMAC OA index and the McGill pain questionnaire (MPQ) [
22]. The following psychosocial variables were measured: depression [the Centre for Epidemiological Studies depression scale (CES‐D)] [
23], modified for use in arthritis patients [
24]; anxiety [trait section of the State‐Trait Anxiety Inventory (STAI)] [
25]; fatigue (Fatigue Severity Scale) [
26]; helplessness [helplessness subscale of the Rheumatology Attitudes Index (RAI)] [
27]; self‐efficacy for pain [pain subscale of the Arthritis Self‐Efficacy Scale (SELF)] [
28] and quality of life [Perceived Quality of Life Index (PQOL)] [
29].
Pain threshold was measured at six points using a hand‐held dolorimeter with a neoprene stopper footplate with a diameter of 1.1 cm (area 0.95 cm
2) (Pain, Diagnostics and Thermograph Inc, Great Neck, NY, USA). The footplate was placed over the test spot and stabilized with the examiner's non‐dominant hand to prevent slippage. Pressure was increased at approximately 1 kg/s until the subject indicated that the sensation had become painful. Threshold was assessed at three ‘fibromyalgia’ sites: the medial fat pad of the index knee; the ipsilateral lateral epicondyle and the contralateral trapezius; and three ‘control’ sites: ipsilateral lateral knee and lateral forearm and contralateral lateral thigh [
30].
A standard examination of the index knee was performed, including bony and soft tissue swelling (0–3, where 0 = none, 1 = mild, 2 = moderate, 3 = severe), pain on passive movement (0–1, where 0 = absent, 1 = present), joint line tenderness (0–1), tenderness on patellofemoral compression (0–1), crepitus (either patellofemoral or tibiofemoral, 0–1), fixed flexion deformity (0–1), collateral instability (0–1), cruciate instability (0–1) and range of movement (degrees).
Radiographs were not routinely obtained for this study but were viewed when available if taken less than 2 yr prior to this study. Standing anteroposterior radiographs of the tibiofemoral joints were available on 51/68; skyline or lateral patellofemoral joint views were available on 35/68. The radiographs were read by a single trained observer and graded for Kellgren and Lawrence (KL) grade where grade 0 = normal; 1 = doubtful osteophyte; 2 = definite mild; 3 = moderate; 4 = severe [
31]. The radiographs were also scored for individual features (osteophyte 0–3, joint space narrowing 0–3, sclerosis 0–1) using a standard atlas [
32] and the minimum tibiofemoral joint space width recorded (mm). The results are presented for the index knee. Individual feature scores from medial and lateral tibiofemoral joints and patellofemoral joints were totalled to give total osteophyte (0–15) and joint space narrowing scores (0–9).
Analysis of the data was performed using Statistical Analysis System version 6.06 (SAS Institute Inc, Cary, NC, USA). Unadjusted analyses were performed using unpaired t‐tests and ANOVAs for categorical variables and correlation coefficients for continuous variables. Multiple regression analyses were then performed using a backward stepwise elimination model. Independent variables were initially entered into the model, using WOMAC disability as the dependent variable. Linearity between variables was established by examining correlations and scatterplots. Variables were removed stepwise until only those significantly (P < 0.1) contributing to pain severity remained.
Results
The baseline characteristics of the group are shown in Table 2
⇔. The mean age was 65.8 (± 10.4) yr with a mean symptom duration of 8.2 (± 7.3) yr. Forty‐eight of 69 (69.6%) were women; 45/69 (65.2%) were Caucasian, the remainder being Afro‐American. A range of radiographic severity was observed; KL grades were as follows: grade 1: 10/62 (16.1%); grade 2: 21/62 (33.9%); grade 3: 21/62 (33.9%); grade 4: 10/62 (16.1%). Of those with definite OA (KL ≥ 2) in the index knee, 57% also had definite changes (KL ≥ 2) in the contralateral knee. As a group, the subjects were relatively highly educated, with a mean of 14.3 (± 3.3) yr of formal education; 22 (31.9%) were college graduates.
The mean disability score (0–100) was 42.1 ± 22.3 (range 1.0–90.9; median score 42.9, interquartile range 23.6–58.9). The frequency distribution of disability scores is shown in Fig. 1
⇔. The results for self‐reported difficulty with individual tasks are given in Table 1
⇔. Greatest difficulty was reported with: ‘heavy domestic duties’ (58.1 ± 29.2), ‘descending stairs’, ‘rising from sitting’, ‘going shopping’ and ‘ascending stairs’. Least difficulty was reported for ‘lying in bed’ (27.6 ± 25.5) and ‘sitting’ (27.5 ± 23.2).
Women tended to report more difficulty than men (44.5 ± 21.9 vs 36.5 ± 22.8; P = 0.17) and Black subjects more than Caucasian (48.0 ± 22.1 vs38.9 ± 22.0; P = 0.11) but in neither case was this statistically significant. There was no significant relationship between disability and radiographic change measured by KL grade (ANOVA P = 0.56). The mean score for subjects with grade 1 KL change (doubtful radiographic OA) was not different from grades 2, 3 and 4 combined (definite radiographic OA) (42.5 ± 27.2 vs 41.0 ± 21.3; P = 0.84). There was no correlation between difficulty with tasks and specific radiographic features including total osteophyte score (range 0–15; r = 0.13; P = 0.44), total joint space narrowing score (range 0–9; r = 0.10; P = 0.57) and minimum tibiofemoral joint space (r = 0.17; P = 0.24). Furthermore, in those subjects with lateral or skyline radiographs, the presence of patellofemoral changes was not associated with increased disability (definite patellofemoral joint narrowing vs no patellofemoral joint narrowing, WOMAC disability score 39.9 ± 20.0 vs 44.0 ± 19.1; P = 0.56).
Unadjusted correlations between disability score and continuous variables are shown in Table 2
⇔. Significant direct correlations were seen between disability and BMI, pain (by both measures), anxiety, fatigue and helplessness. Significant inverse correlations were seen with years of formal education, self‐efficacy and perceived quality of life.
Disability scores for all 17 tasks were higher with increasing BMI. The mean WOMAC function score for ‘obese’ subjects (BMI ≥ 30) was 50.2 ± 19.1; for ‘non‐obese’ (BMI < 30) the mean score was 34.2 ± 22.6 (P = 0.002). There was a trend for subjects to report less difficulty with function with increasing age: however, this was probably due to the confounding effect of BMI. BMI was greater in younger subjects (≤ 65 yr vs > 65 yr: 33.5 ± 7.4 vs 29.8 ± 5.7; P = 0.026). The effect of age on function disappeared after stratification on obesity (correlation age: function in non‐obese subjects r = 0.025; P = 0.88; obese r = 0.17; P = 0.34).
Regarding examination findings (see Table 3
⇔), in unadjusted analysis, disability was significantly higher in subjects with bony swelling than in those without (swelling
vs no swelling: 44.8 ± 21.7
vs 30.6 ± 22.8;
P = 0.046). Joint line tenderness and tenderness on patellofemoral compression were also associated with increased WOMAC scores. No significant relationship was seen with soft tissue swelling, pain on motion or joint instability. The range of motion in degrees correlated inversely with reported disability (
r = −0.25;
P = 0.042).
Threshold to pain was significantly inversely related to self‐reported disability: the higher the threshold, the less the disability. This applied whether measuring threshold at all six points (r = −0.28; P = 0.02), at the two points around the knee (r = −0.31; P = 0.01) or at the three ‘fibromyalgia’ points (r = −0.29; P = 0.01).
A backward stepwise elimination model was examined to establish the relative contributions of each variable to self‐reported disability, using the WOMAC disability score as the dependent variable. Independent variables were those correlating with disability in unadjusted analysis at P < 0.05, excluding joint line and patellofemoral tenderness as the reproducibility of these findings has not been established. Radiographic variables were forced into the model to test specific hypotheses. The adjusted model therefore contained seven variables: pain severity by WOMAC; BMI; helplessness; education; fatigue; anxiety; range of movement plus a forced radiographic variable (KL score). Inclusion of all variables accounted for 61.6% of variance in WOMAC disability. Variables were removed stepwise until only those significant at the P < 0.1 level remained. The final model included forced variable KL grade (F = 0.65; P = 0.42); other variables were pain severity (F = 31.4; P = 0.0001), BMI (F = 6.91; P = 0.011) and helplessness (F = 4.27; P = 0.04). This model accounted for 59.9% of variance in WOMAC disability.
Models were also examined forcing osteophyte score and joint space narrowing score instead of KL grade. Pain and BMI remained in both models but anxiety entered the model for both osteophyte (F = 6.37; P = 0.02) and joint space narrowing (F = 6.77; P = 0.01). Neither osteophyte nor joint space narrowing contributed significantly to disability.
Discussion
Disability is a major consequence of lower limb OA [
33]. Since knee OA is very common, the impact of such disability on the community in terms of health care utilization is high. It is clearly important to understand the factors which contribute to such disability.
Most previous studies of function in knee OA have utilized generic instruments such as the Health Assessment Questionnaire (HAQ) [
8,
18,
34,
35], NHAWES‐1 (National Health & Nutrition Examination Survey) interview questionnaire [
9], Sickness Impact Profile (SIP) [
7,
15] and the Arthritis Impact Measurement Scale (AIMS) [
13,
14]. Some of these (e.g. SIP, AIMS) reflect global health status rather than pure functional ability, whilst others (e.g. HAQ) measure upper as well as lower limb function. It is possible that indices which include functions unaffected by OA may ‘dilute’ the influence of knee OA on lower limb function [
2,
36]. For these reasons, we chose to use the WOMAC questionnaire, developed specifically for patients with lower limb OA. It is important to note that it reflects self‐reported disability rather than actual disability: we did not require patients to undergo any formal functional tests in this study.
Even before adjusting for pain, we were unable to find a relationship between radiographic change and self‐reported function. Reports of community subjects [
7,
8,
11,
12] and patients [
14,
16] have generally shown an increase in disability with increasing radiographic severity, although this effect may disappear when adjusted for pain and other confounders. Our study extends these observations by examining individual radiographic features: although these have been studied in relation to pain [
37–
39], to our knowledge only one previous report has considered disability. Van Baar
et al. [
16] reported unadjusted correlations of 0.23–0.26 between individual radiographic features (osteophyte, narrowing) and self‐reported disability in patients with knee OA. In contrast, we were unable to show a significant correlation between either osteophyte or narrowing and disability. Our population had radiographic disease of varying severity and all had pain. The Dutch population had milder radiographic disease and shorter duration of symptoms than ours (median 65 weeks
vs 312 weeks). It is possible that function bears a stronger relationship to radiographic change in patients with early disease (and community subjects). There is some evidence of this from our own study: the correlation between, for example, joint space narrowing and disability was much stronger in subjects with disease duration less than 5 yr (
r = 0.64;
P = 0.017) than in those with 5 or more years of symptoms (
r = 0.10;
P = 0.54). Other factors may become more important in determining disability as the disease becomes chronic. Perhaps, however, we should not be too surprised by a lack of correlation between radiographs and disability: radiographs show only bone whilst other structures—notably muscle and ligament—may be more important determinants of function. Other authors [
40,
41] have noted a discrepancy between radiographic and symptom change in longitudinal studies.
The risk of disability increases with presence of knee pain in the community [
2,
9–
12]. There is less literature on the relationship between severity of pain and disability. Jordan
et al. [
8] showed that severity of knee pain (as measured by a simple ‘mild‐moderate‐severe’ scale) correlated with self‐reported disability in the community. In hospital out‐patients with knee OA, Salaffi
et al. [
14] reported correlations of 0.61 (
P < 0.001) between pain severity (by MPQ) and function (by AIMS). van Baar
et al. [
16] reported correlations of 0.40 (
P < 0.01) between self‐reported disability (by AIMS) and pain (by visual analogue scale) in patients attending primary care with knee OA. We confirmed a strong correlation between pain severity and disability which, when assessed by the WOMAC pain scale or the MPQ, remained significant after adjustment for confounding variables. The mechanism by which pain contributes to disability is unclear. Pain may lead to avoidance of physical activity, resulting in a cycle of pain, inactivity and muscle wasting [
19]. Psychological factors such as anxiety may serve to amplify this loop by increasing the degree of avoidance: lack of knowledge or education may also contribute by fostering inappropriate beliefs about pain and activity.
Obesity was the other major determinant of functional loss in these patients. The most obvious explanation for this is that obesity places the individual at a mechanical disadvantage: greater muscle power is required to undertake activities, particularly lower limb tasks such as those which make up the WOMAC. Even in subjects who do not have knee OA, obesity has been shown to be associated with voluntary weakness of the quadriceps muscle [
42]. Weight reduction, combined with moderate exercise, has been shown to improve function in patients with knee OA [
43].
We did not find significant correlations between depression and disability: this is in contrast to other studies of patients with lower limb OA. Summers
et al. [
15] found correlations of 0.5 or higher between depression (by the Beck Depression Inventory) and function (by SIP); Salaffi
et al. [
14] using the AIMS and the Zung Depression Inventory reported correlations of 0.31 for physical function. Others [
16] have failed to find such an association. The lack of association in our study may reflect the low prevalence of depression in our population: only 15.4% subjects were ‘possibly depressed’ (CES‐D score 16 or above). We believe our measure of disability to be more appropriate for studies of knee OA. Although there may be disagreement on the role of depression in reported disability, all previous reports find an association with anxiety: we confirm this, even after adjustment for potential confounders, notably pain severity. This extends our observations from the Baltimore Longitudinal Study of Aging cohort that anxiety, but not depression, is related to pain reporting in the community [
44]. It is unclear from this cross‐sectional study whether anxiety is a risk factor for subsequent disability or whether disability is, itself, a reason why subjects become more anxious. Acute injury or pain triggers anxiety and avoidance of movement as a normal response, presumably with an evolutionary advantage in terms of a warning mechanism to protect the organ—in this case the knee—and reduce the chances of further injury. However, prolonged anxiety may, by resulting in persistent attempts to avoid knee pain, lead to loss of muscle bulk, generalized deconditioning and loss of confidence. Only prospective studies will allow clarification of this issue.
Helplessness was also related to disability. As with anxiety, subjects may feel helpless because they have poor function: alternatively, feelings of helplessness may directly result in a reduction in function, at least as measured by self‐report. Strategies designed to reduce helplessness such as education [
45] or provision of social or telephone support [
46–
47] have been shown to improve self‐reported function.
Certain findings on physical examination were associated with greater disability: for example bony swelling. Although reliability of clinical measurement in knee OA varies, bony swelling is one of the most reproducible measures [
48]. However, bony swelling correlates with pain severity [
49] and the effect of bony swelling on disability disappeared when adjusted for this variable. Joint line tenderness, and threshold to pain by dolorimetry were associated in unadjusted analysis but did not remain in an adjusted model. The range of joint motion was significantly related to disability, as previously reported [
16,
19]. Although disability may be secondary to restricted range of motion (due to capsular contractures, muscle contractures and muscle spasm) it is of interest that some tasks which do not appear to depend on knee movement (for example, ‘lying in bed’) showed the strongest correlation (
r = −0.35;
P = 0.003), whilst other tasks such as ‘getting in and out of a car’ or ‘putting on socks’ showed no correlation at all. It may be that diminished range of movement and disability share a common cause and reflect a more severe form of disease outcome.
Our study has certain limitations. We did not measure quadriceps strength although this has been shown by several authors to correlate with self‐reported disability [
12,
16,
19]. Our population may not be representative of all knee OA patients: our group, for example, contained a disproportionate number of highly educated individuals and overall prevalence of depression was less than that reported in other populations with OA. We report on a hospital‐based population of individuals who have elected to seek medical care: results may not be generalizable to the community. X‐rays were not available on all subjects. Finally, our assessment of anxiety and depression relied on self‐administered questionnaires which are, inevitably, a surrogate for formal psychiatric interview.
In conclusion, self‐reported disability in patients with symptomatic knee OA bears little relationship to radiographic severity, at least in subjects with over 5 yr disease duration. It is, however, strongly related to pain severity, BMI and anxiety, all of which are potentially treatable. Interventions focused on weight and pain reduction, together with strategies designed to reduce helplessness and anxiety, are logical approaches to tackling the burden of disability associated with knee OA [
50], although the impact of such interventions can only be assessed in formal prospective studies.
