SEARCH

SEARCH BY CITATION

Keywords:

  • Knee pain;
  • Osteoarthritis;
  • Radiographs;
  • Sensitivity and specificity;
  • Older adults;
  • Clinical assessment study;
  • Primary care

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES
  10. Supporting Information

Objective

To determine whether clinical information can practically rule in or rule out the presence of radiographic osteoarthritis in older adults with knee pain.

Methods

We conducted a cross-sectional diagnostic study involving 695 adults ages ≥50 years reporting knee pain within the last year identified by postal survey and attending a research clinic. Potential indicators of radiographic osteoarthritis were gathered by self-complete questionnaires, clinical interview, and physical examination. Participants underwent plain radiography (posteroanterior, skyline, and lateral views). Radiographic osteoarthritis was defined as the presence of definite osteophytes in at least 1 joint compartment of the index knee.

Results

Independent predictors of radiographic osteoarthritis were age, sex, body mass index, absence of whole leg pain, traumatic onset, difficulty descending stairs, palpable effusion, fixed-flexion deformity, restricted-flexion range of motion, and crepitus. Using this model, 245 participants had a predicted probability ≥80% (practical rule in), of whom 231 (94%) actually had radiographic osteoarthritis (specificity 93%). Twenty-one participants had a predicted probability <20% (practical rule out), of whom only 2 (10%) had radiographic osteoarthritis (sensitivity 99.6%). The predicted probability of radiographic osteoarthritis for the remaining 429 participants fell into an intermediate category (20–79%).

Conclusion

Simple clinical information can be used to estimate the probability of radiographic osteoarthritis in individual patients. However, for the majority of community-dwelling older adults with knee pain this method enables the presence of radiographic osteoarthritis to be neither confidently ruled in nor ruled out. Prospective validation and updating of these findings in an independent sample is required.


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES
  10. Supporting Information

In the UK, an estimated 2 million adults consult their general practitioner with joint symptoms diagnosed as osteoarthritis each year (1). However, attributing joint pain in older adults to osteoarthritis with confidence is difficult (2), and the frequency of osteoarthritis diagnosis varies considerably between general practitioners (3).

Of all the painful conditions presented by older adults to doctors, knee pain is the most common (4). Although osteoarthritis is the most likely diagnosis, general practitioners often request plain radiographs in an attempt to help clinical decision making (5, 6). However, this practice runs contrary to current guidelines (7) that recognize that changes on radiograph are common in the asymptomatic population (8) and, in many cases, do not guide the choice of treatment or alter its outcome.

Relatively simple educational reminders can reduce the number of radiograph requests by general practitioners (9, 10). However, if ascertaining disease status remains a consideration in some consultations, is there an alternative to sending the patient for a radiograph? Can information available at the time of consultation effectively rule in or rule out the presence of radiographic knee osteoarthritis?

In this study, we quantified the simple risk factors and clinical characteristics that can be used, in combination, to estimate the probability of radiographic knee osteoarthritis in older adults with knee pain not associated with specific inflammatory arthropathy. We also investigated the extent to which these simple indicators can effectively rule in and rule out the presence of radiographic knee osteoarthritis. These steps equate to the derivation of a clinical prediction rule before requiring external validation in a separate sample (11).

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES
  10. Supporting Information

Study population.

Participants were recruited from a 2-stage cross-sectional postal survey of all adults ages ≥50 years registered with 3 general practices in North Staffordshire (irrespective of actual consulting patterns). Respondents reporting knee pain within the previous 12 months were invited to attend a research clinic at a local National Health Service Hospital Trust.

The study protocol was approved by North Staffordshire Local Research Ethics Committee (project number 1430) and details have been published elsewhere (12, 13). All participants provided written informed consent to undergo clinical and radiographic assessment. In addition, they were asked for consent to medical record review to assist in excluding preexisting inflammatory disease.

The inclusion criteria for the current analysis were as follows: age ≥50 years, registered with one of the participating general practices at the time of study, responded to both postal questionnaires, consented to further contact, and attended the research clinic. Participants were excluded if they had not experienced knee pain within the 6 months prior to clinic attendance, had a preexisting diagnosis of inflammatory arthropathy in their medical records, or had a total knee replacement in their most affected knee. Thus, the target population for this study were adults with current or recent knee pain of an age range at which knee osteoarthritis would be one of the main diagnoses considered.

Data collection.

All data were planned and gathered prospectively. Participants underwent a standardized clinical interview and physical examination, which were abbreviated versions of assessments developed in an earlier stage of this research (14, 15). Inter- and intrarater reliability have been reported elsewhere (16, 17). Assessments were conducted by 1 of 6 research therapists blinded to the findings from radiography, postal questionnaires, and medical records. Training of the assessors took place prior to the study and was updated after every 100 participants recruited. Training included comparisons against rheumatologists, open and blinded comparisons against each other using “expert patients,” and peer observation. Assessors were issued a manual of detailed protocols for assessing each sign and symptom.

Plain knee radiographs were obtained on the day of clinic attendance in the Trust's radiology department. The films were obtained by a team of 6 radiographers who had all undergone training to standardize the radiographs and met for regular quality control sessions.

Three views were taken of each knee: a weight-bearing semiflexed posteroanterior (PA) view according to the protocol developed by Buckland-Wright et al (18), and lateral and skyline views, both in a supine position with the knee flexed to 45°. Participants filled in a brief self-complete questionnaire about their knee symptoms on the day of their clinic attendance.

Presence or absence of radiographic knee osteoarthritis.

The reference standard was definite osteoarthritis from radiographs of the participants' most affected knee (identified by participants at interview). Our definition included osteoarthritis in the tibiofemoral and/or patellofemoral joint. The latter has often been omitted from previous studies but has been shown to be a common form of knee osteoarthritis (19). A single reader (RD), blinded to all other information on participants, scored all films. Films were scored for individual radiographic features, including osteophytes, joint space width, sclerosis, subluxation, and chondrocalcinosis. The atlas and scoring system developed by Altman et al (20, 21) were used for the PA and skyline views and the atlas developed by Burnett et al (22) was used for the lateral view. Additionally, PA and skyline views were assigned a Kellgren and Lawrence (K/L) grade (23). In a subsample of 50 participants, both intra- and interreader reliability for PA K/L score, skyline K/L score, and lateral osteophytes were very good (κ = 0.81–0.98 and 0.49–0.76, respectively). Definite radiographic osteoarthritis was defined as K/L grade 2 or higher for the PA view, K/L grade 2 or higher for the skyline view, the presence of superior or inferior patella osteophytes (lateral view), and/or posterior tibial osteophytes (lateral view) (24). Our definition of radiographic osteoarthritis essentially applied the same principle as earlier studies (i.e., definite osteophyte [25]) but extended this to all 3 views of the knee.

Potential indicators of the presence or absence of radiographic knee osteoarthritis.

Prior to analysis, potential indicators were identified from information in the 2 postal questionnaires, the clinical assessment, or the brief self-complete questionnaire. Potential indicators were chosen if they were known risk indicators for radiographic knee osteoarthritis (e.g., age, sex, body mass index [26, 27]), clinical signs and symptoms with a known or putative link to the occurrence of radiographic knee osteoarthritis (e.g., morning stiffness, crepitus, difficulty climbing stairs [28–37]), or clinical manifestations of differential diagnoses (e.g., whole leg pain suggesting widespread pain or referred pain from the lumbar spine, restricted hip range of movement suggesting referred pain from hip osteoarthritis [8, 38–40]). All indicators had to be practicable for assessment within a routine primary care consultation.

Statistical analysis.

We quantified the relationship between each potential indicator and the presence of radiographic knee osteoarthritis by univariable logistic regression. For continuous indicator variables or ordinal variables with several levels, cutoffs were chosen prior to univariable analysis on the basis of previous empirical evidence or, in the absence of this evidence, frequency in the current sample. Variables with a P value less than 0.10 were carried forward to multivariable logistic regression. We used a backward stepwise procedure to fit the regression function to the data (Pr = 0.05; Pe = 0.01). Patients with missing data on any of the univariably associated indicators were excluded from this stage of the analysis. Beta coefficients and odds ratios (ORs) with their respective 95% confidence intervals were used to summarize the marginal contribution of each indicator in the regression function. The resulting multivariable regression function was then fitted to all patients with complete data on indicators included in the multivariable function. The goodness-of-fit of the regression function was assessed by splitting the sample into deciles and plotting the estimated probability against the observed proportion with radiographic osteoarthritis (41).

We used the area under the receiver operating characteristic curve (ROC) as a summary measure of how well the regression function discriminated between patients with and without radiographic knee osteoarthritis. To allow for overoptimism in the regression function, we performed bootstrapping techniques as a form of internal validation (42). The resulting multivariable function was applied to 200 bootstrap samples, with replacement, to determine the level of bias in each of the observed beta coefficients.

Using the bias-corrected regression function, we described the proportion of the sample falling into the practical rule-in category (≥80% probability of radiographic osteoarthritis) and practical rule-out category (<20% probability). Finally, using the bias-corrected beta coefficients, we derived a simple scoring method (43) that would enable the probability of radiographic osteoarthritis to be calculated for an individual patient with knee pain. All analysis was carried out in Stata 7.0 (StataCorp, College Station, TX).

RESULTS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES
  10. Supporting Information

Between August 2002 and September 2003, a total of 819 individuals ages ≥50 years who had knee pain in the past 12 months attended the research clinics. Compared with all 3,106 survey respondents reporting knee pain in the previous 12 months, those attending the research clinics tended to be younger, male, married, have more education, and be less likely to come from routine occupational backgrounds or to be anxious or depressed (Table 1). The number of patients who were eligible and had complete data from the 819 clinic attendees is shown in Figure 1.

Table 1. Descriptive characteristics of the observed target population and the subgroup attending the research clinic*
CharacteristicReported knee pain in last 12 months (n = 3,106)Attended research clinic (n = 819)
  • *

    Values are the number (percentage) unless otherwise indicated. Individual items may not add to totals due to missing data. SF-12 = Medical Outcomes Study Short Form 12 (44); HAD = Hospital Anxiety and Depression scale (45); WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index (46).

  • Participants were selected from registers of these 3 general practices.

Practice  
 A1,227 (40)302 (37)
 B1,251 (40)330 (40)
 C628 (20)187 (23)
Age, years  
 50–59898 (29)236 (29)
 60–69964 (31)312 (38)
 70–79822 (26)222 (27)
 ≥80422 (14)49 (6)
Female sex1,832 (59)440 (54)
Married/cohabiting1,985 (65)599 (74)
Higher education327 (11)117 (15)
Currently employed668 (22)167 (21)
Occupational class: manual1,904 (68)420 (54)
Baseline SF-12 score, mean ± SD  
 Physical (0–100)37.0 ± 12.237.6 ± 11.9
 Mental (0–100)47.9 ± 11.650.6 ± 10.9
Baseline HAD anxiety  
 None (0–7)1,656 (55)500 (62)
 Possible (8–11)842 (28)217 (27)
 Probable (12–21)508 (17)85 (11)
Baseline HAD depression  
 None (0–7)2,198 (73)663 (83)
 Possible (8–11)593 (20)102 (13)
 Probable (12–21)221 (7)37 (4)
Baseline WOMAC score, mean ± SD  
 Pain (0–20)6.4 ± 4.4
 Stiffness (0–8)2.7 ± 1.9
 Physical function (0–68)20.9 ± 15.2
thumbnail image

Figure 1. Flow diagram. CAS(K) = Clinical Assessment Study (Knee); TKR = total knee replacement; PA = posteroanterior; OA = osteoarthritis.

Download figure to PowerPoint

Of the 650 eligible participants with complete data on all indicators, 452 (70%) had radiographic knee osteoarthritis. A total of 257 patients were classified as having moderate or severe radiographic osteoarthritis (K/L grade 3 or 4 in either the tibiofemoral or patellofemoral joint, or grade 3 posterior or lateral osteophytes).

Of the 57 candidate indicators, 17 were not associated with the presence of definite radiographic knee osteoarthritis in univariable analysis. A further 11 indicators were excluded from the multivariable stage due to high collinearity (see Appendix A, available at the Arthritis Care & Research Web site at http://www.interscience.wiley.com/jpages/0004-3591:1/suppmat/index.html).

Of the remaining 29 indicators with a univariable P value less than 0.10, 10 were independently associated with the presence of radiographic knee osteoarthritis in the multivariable logistic regression function. This model was then refitted to all 695 participants with complete data for these 10 indicators. Older age (adjusted OR 2.0 for ages 60–69 years and 2.8 for ages ≥70 years), male sex (OR 2.1), overweight (OR 2.2) or obese (OR 3.3), history of the knee problem starting following an accident or injury in the past (OR 1.8), difficulty descending stairs due to the knee problem (OR 1.3 for mild, 2.8 for moderate, and 3.0 for severe/extreme), presence of palpable effusion on examination (OR 2.2 for mild and 1.8 for moderate/gross), fixed-flexion deformity (OR 7.1), restricted knee flexion range of motion (OR 2.3), and the presence of palpable coarse crepitus (OR 1.4 for possible and 2.4 for definite) all increased the likelihood of radiographic knee osteoarthritis being present. Pain in the whole leg was associated with a lower likelihood of radiographic knee osteoarthritis (OR 0.4) (Tables 2 and 3).

Table 2. Uncorrected multivariable logistic regression function for radiographic knee osteoarthritis (ROA)*
IndicatorTotal, no.ROA, no.Uncorrected function
β(95% CI)OR(95% CI)P
  • *

    β = multiple logistic regression coefficient; 95% CI = 95% confidence interval; OR = odds ratio (adjusted for all other listed variables); ROM = range of motion.

Age, years       
 50–59 (reference)207109    
 60–692782030.71(0.28, 1.14)2.0(1.3, 3.1)0.001
 ≥702101701.05(0.54, 1.55)2.8(1.7, 4.7)< 0.001
Sex       
 Female (reference)381237     
 Male3142450.74(0.35, 1.13)2.1(1.4, 3.1)< 0.001
Body mass index, kg/m2       
 <25 (reference)11055     
 25–29.93042100.77(0.24, 1.30)2.2(1.3, 3.7)0.004
 ≥302812171.19(0.63, 1.75)3.3(1.9, 5.7)< 0.001
Started after accident/injury       
 No (reference)593401     
 Yes102810.60(0.03, 1.18)1.8(1.0, 3.3)0.04
Pain in whole leg       
 No (reference)610435     
 Yes8547−0.99(−1.56, −0.43)0.4(0.2, 0.6)0.001
Difficulty descending stairs       
 None (reference)17999     
 Mild2161380.29(−0.17, 0.76)1.3(0.8, 2.1)0.21
 Moderate1891521.04(0.50, 1.57)2.8(1.7, 4.8)< 0.001
 Severe/extreme111931.11(0.41, 1.81)3.0(1.5, 6.1)0.002
Palpable effusion       
 None (reference)443275     
 Mild84730.81(0.06, 1.56)2.2(1.1, 4.7)0.04
 Moderate/gross1681340.56(0.09, 1.53)1.8(1.1, 2.8)0.02
Fixed-flexion deformity       
 No (reference)610400     
 Yes85821.96(0.74, 3.18)7.1(2.1, 24.1)0.002
Knee flexion ROM, degrees       
 ≥120 (reference)599397     
 <12096850.83(0.07, 1.60)2.3(1.1, 4.9)0.03
Coarse crepitus       
 None (reference)412254     
 Possible130980.34(−0.17, 0.85)1.4(0.8, 2.3)0.20
 Definite1531300.89(0.34, 1.44)2.4(1.4, 4.2)0.001
Constant  −5.04(−6.89, −3.18)  < 0.001
Table 3. Bias-corrected multivariable logistic regression function for radiographic knee osteoarthritis (ROA)*
IndicatorTotal, no.ROA, no.Bias-corrected function (after bootstrapping)
β(95% CI)OR(95% CI)
  • *

    See Table 2 for definitions.

Age, years      
 50–59 (reference)207109    
 60–692782030.69(0.26, 1.19)2.0(1.3, 3.3)
 ≥702101701.00(0.63, 1.53)2.7(1.9, 4.6)
Sex      
 Female (reference)381237    
 Male3142450.74(0.30, 1.15)2.1(1.4, 3.2)
Body mass index, kg/m2      
 <25 (reference)11055    
 25–29.93042100.73(0.11, 1.21)2.1(1.1, 3.3)
 ≥302812171.14(0.56, 1.64)3.1(1.7, 5.2)
Started after accident/injury      
 No (reference)593401    
 Yes102810.61(0.03, 1.27)1.8(1.0, 3.5)
Pain in whole leg      
 No (reference)610435    
 Yes8547−0.98(−1.65, −0.55)0.4(0.2, 0.6)
Difficulty descending stairs      
 None (reference)17999    
 Mild2161380.29(−0.12, 0.80)1.3(0.9, 2.2)
 Moderate1891521.02(0.50, 1.66)2.8(1.7, 5.3)
 Severe/extreme111931.10(0.38, 1.94)3.0(1.5, 7.0)
Palpable effusion      
 None (reference)443275    
 Mild84730.72(−0.22, 1.70)2.1(1.0, 5.5)
 Moderate/gross1681340.52(0.11, 1.10)1.7(1.1, 3.0)
Fixed-flexion deformity      
 No (reference)610400    
 Yes85821.87(1.04, 3.19)6.5(2.8, 24.4)
Knee flexion ROM, degrees      
 ≥120 (reference)599397    
 <12096850.76(0.07, 1.52)2.1(1.1, 4.6)
Coarse crepitus      
 None (reference)412254    
 Possible130980.32(−0.09, 0.95)1.4(0.9, 2.5)
 Definite1531300.84(0.27, 1.42)2.3(1.3, 4.1)
Constant  −4.94(−6.74, −2.91)  

The formula of the multiple logistic regression function was used to estimate the probability of radiographic knee osteoarthritis for each individual based on their clinical profile. The individual probabilities ranged from 0.07 to 0.99 (where 0 represents certainly absent and 1 represents certainly present). The function was a good fit for the data (Pearson's χ2 = 430.89, P = 0.88) (calibration plot not shown). The area under the ROC curve was 0.80, suggesting reasonable discrimination between patients with and those without radiographic knee osteoarthritis (41). Bootstrapping revealed only a very minor degree of overfitting in the model, with the degree of shrinkage of coefficients for each indicator in the model well below the level indicative of serious bias (25% of the standard error of the coefficient [47]) (data not shown).

The effect of applying the bias-corrected function to individuals in the sample is summarized in Table 4. A total of 245 individuals fell into the practical rule-in category. Of these, 231 (94%) did have radiographic knee osteoarthritis on radiograph. Twenty-one individuals fell into the practical rule-out category, of whom only 2 (10%) had radiographic knee osteoarthritis. The remaining 429 individuals fell into an indeterminate category (estimated probability 20–79%). Of these, 249 (58%) had radiographic knee osteoarthritis. Adopting a threshold of predicted probability of ≥50% identified 266 (96%) of 277 individuals with moderate or severe radiographic knee osteoarthritis.

Table 4. Severity of radiographic knee osteoarthritis (ROA) by range of estimated probability of ROA from a bias-corrected logistic regression function*
Estimated probability of definite ROA, %Severity of ROA
Moderate/severeMildDoubtfulNoneTotal
  • *

    Values are the number (percentage).

  • Severity defined as follows: moderate/severe = posteroanterior Kellgren and Lawrence (K/L) grade 3 or higher, skyline K/L grade 3 or higher, posterior osteophytes grade 3, or lateral osteophytes grade 3; mild = posteroanterior K/L grade 2, skyline K/L grade 2, posterior osteophytes grade 1 or 2, or lateral osteophytes grade 1 or 2; doubtful = posteroanterior K/L grade 1 or skyline K/L grade 1; none = posteroanterior K/L grade 0, skyline K/L grade 0, posterior osteophytes grade 0, and lateral osteophytes grade 0.

  • Definite ROA = mild or moderate/severe.

90–100110 (40)26 (13)2 (4)2 (1)140 (20)
80–8964 (23)31 (15)1 (2)9 (6)105 (15)
70–7943 (16)34 (17)12 (21)17 (11)106 (15)
60–6928 (10)33 (16)12 (21)27 (17)100 (14)
50–5921 (8)26 (13)6 (11)19 (12)72 (10)
40–495 (2)24 (12)6 (11)16 (10)51 (7)
30–394 (1)22 (11)8 (14)24 (15)58 (8)
20–292 (1)7 (3)6 (11)27 (17)42 (6)
10–190 (0)2 (1)3 (5)15 (10)20 (3)
0–90 (0)0 (0)1 (2)0 (0)1 (<1)
Any277 (100)205 (100)57 (100)156 (100)695 (100)

Points were assigned to each indicator by dividing the bias-corrected coefficient by the value of the smallest coefficient and rounding to the nearest 0.5 points (see Appendix B, available at the Arthritis Care & Research Web site at http://www.interscience.wiley.com/jpages/0004-3591:1/suppmat/index.html). Adding the relevant points for an individual patient provides a total score that is then converted into an estimated probability of radiographic knee osteoarthritis. For example, a 67-year-old (2.0 points) overweight (2.5 points) man (2.5 points) who reports localized knee pain (3.5 points) that first began after a previous knee injury (2.0 points), who now reports moderate difficulty going downstairs due to the knee (3.5 points), and who on examination has a mild effusion (2.5 points) and definite crepitus (3.0 points) but no fixed-flexion deformity or limited knee flexion would have a total score of 21.5. The estimated probability of that individual having radiographic knee osteoarthritis is >90%.

DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES
  10. Supporting Information

Our purpose was to establish whether information potentially available at the time of a primary care consultation could be used to effectively rule in or rule out the presence of radiographic osteoarthritis in older adults presenting with knee pain. Our findings demonstrate that this can be done with partial success.

Estimating the probability of radiographic knee osteoarthritis among symptomatic individuals requires a combination of several pieces of information: no single indicator adequately discriminates between patients with and those without structural disease. Diagnosis in clinical practice rarely relies on a single test, and diagnostic research therefore relies on multivariable modeling (48–50). Well-recognized risk indicators (e.g., age, body mass index) increase the probability that the patient has radiographic knee osteoarthritis, as does information from the history (e.g., difficulty descending stairs) and physical examination (e.g., restricted movement at the knee, coarse crepitus).

But despite a relatively well-calibrated multivariable model, when we used this model to estimate the probability of radiographic knee osteoarthritis for individual patients, only a minority of our sample fell into a practical rule-in category (≥80% probability) and very few fell into a practical rule-out category (<20% probability). In short, a fairly comprehensive assessment of risk indicators and clinical signs and symptoms still leaves us unable to confidently rule in or rule out radiographic knee osteoarthritis in 3 out of 5 older adults experiencing knee pain in the community. We do not advocate ordering radiographs to resolve this uncertainty. Instead, given the difficulty in discriminating patients with radiographic osteoarthritis from those without, and given that structural change on radiograph may be a relatively late indicator of disease, the responsibility is perhaps more on justifying why knee pain in this age group (after excluding relatively rare specific differential diagnoses) should not be managed as if it were osteoarthritis.

Many previous studies have described the association between risk factors, clinical variables, and radiographic knee osteoarthritis, but often only in univariable analyses. There are few comparable studies in which multivariable analysis has been conducted with the express purpose of individual risk prediction. LaValley et al found that brief questionnaire-based screening instruments failed to adequately discriminate between patients with and those without symptomatic radiographic knee osteoarthritis (symptoms on most days and radiographic evidence of knee osteoarthritis) in the general population (33). Our focus was rather different. We have argued that, from a clinical perspective, the starting point is undifferentiated knee pain. Ascertaining symptom status poses little challenge: we ask the patient. Identifying the presence of radiographic knee osteoarthritis in those with knee pain is the challenging part. Like us, Claessens et al concluded that “no single clinical finding can accurately predict radiographic OA” in symptomatic individuals (29). The content and performance of their multivariable model were not fully reported. In contrast, we found that clinical variables do add to the prediction of radiographic knee osteoarthritis, as they have done in similar studies for hip osteoarthritis (40, 51). Variables used in the American College of Rheumatology clinical classification criteria for knee osteoarthritis (duration of morning stiffness, bony enlargement, and crepitus) (28) were associated with radiographic knee osteoarthritis in the current study, but only crepitus was retained in the multivariable model.

We defined our target population as adults ages 50 years and older (an age at which osteoarthritis becomes a common diagnosis in primary care) with current or recent knee pain who did not have a preexisting medical diagnosis of specific inflammatory arthropathy. This definition may be criticized for being too broad. It has been argued that “using less frequent knee pain together with radiographic osteoarthritis … corresponds to an entity which is less consistently reflective of osteoarthritis” (33). We would argue that osteoarthritis symptoms can be less frequent and such individuals do present to general practice in significant numbers where their symptoms may be attributed to osteoarthritis (52). A similar argument concerns the inclusion of individuals who reported knee pain as part of a more diffuse pain problem. Pain persistence (pain days in last 6 months), widespread pain, and whole leg pain were included in our list of candidate indicators.

Our reference standard included tibiofemoral and patellofemoral radiographic osteoarthritis. Risk factors may differ between the two (53). However, the combined pattern of osteoarthritis was the most common in our sample, and has had the strongest associations with risk indicators in several previous studies (54–56). Rather than looking separately at tibiofemoral and patellofemoral radiographic osteoarthritis, we thought that discriminating between any or no radiographic osteoarthritis was an appropriate choice and probably better reflects current clinical distinctions. Distinguishing between mild and severe radiographic osteoarthritis might also be important. Using our regression function, the estimated probability of patients with moderate or severe radiographic disease was significantly higher than those with milder forms of disease. For example, only 11 of 172 individuals with an estimated probability <50% had moderate or severe radiographic osteoarthritis. It may be that selecting a more severe threshold of radiographic disease would permit greater discrimination.

We considered many candidate indicators of radiographic knee osteoarthritis, each justified on the basis of prior evidence or clinical knowledge. Occupation and sports participation have been highlighted as risk indicators for knee osteoarthritis (27) but information about these was not gathered in the current study. Participating practices were located partly within areas of traditional mining and farming communities, which might explain the higher prevalence of radiographic knee osteoarthritis among men in our sample. The inclusion of these additional indicators may modify our model but would also increase its complexity with potentially only a small incremental benefit to discrimination (11, 57). Ninety-five participants had 1 or more missing values on either the outcome or diagnostic indicators. These individuals were more likely to be women, older, and have milder symptoms. Nevertheless, when 45 of these individuals were reintroduced for the final multivariable model, there was little noticeable change in the regression coefficients and age, sex, and a marker of clinical severity (difficulty descending stairs) were included in the multivariable regression function. We therefore do not believe that missing data pose a significant threat of bias in our results.

Our model was derived within a single sample of participants drawn from one geographic area. Bootstrap procedures were used to guard against overfitting our logistic regression function. However, prospective validation and updating in a separate sample, particularly one more ethnically diverse than the present sample (individuals from ethnic minorities constituted <1% of the current sample), would be beneficial.

The majority of older persons with knee pain have some evidence of radiographic osteoarthritis in their most affected knee. The prior probability of radiographic osteoarthritis among symptomatic older adults is high. Simple clinical information can help estimate the probability of radiographic osteoarthritis in individual patients, although in many cases this will not allow radiographic osteoarthritis to be ruled in or ruled out with a high degree of confidence.

AUTHOR CONTRIBUTIONS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES
  10. Supporting Information

Dr. Peat 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 design. Peat, Thomas, Duncan, Wood, Wilkie, Hill, Hay, Croft.

Acquisition of data. Peat, Thomas, Duncan, Wood, Wilkie, Hill.

Analysis and interpretation of data. Peat, Thomas, Duncan, Wilkie, Hay, Croft.

Manuscript preparation. Peat, Thomas, Duncan, Wood, Wilkie, Hill, Hay, Croft.

Statistical analysis. Thomas.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES
  10. Supporting Information

The authors thank the administrative and health informatics staff at Keele University's Primary Care Sciences Research Centre, and staff of the participating general practices and Haywood Hospital. We also gratefully acknowledge the assistance of Professor C. Buckland-Wright for advice and training for the radiograph protocols. Dr. K. Dziedzic, Ms H. Myers, Ms J. Handy, and Ms C. Clements contributed to aspects of the design of the Knee Clinical Assessment Study (CAS[K]) and the collection of clinical data. Drs. G. Carpenter, L. Coar, V. Cooper, P. Dawes, J. Greig, A. Hassell, M. Porcheret, M. Shadforth, and A. Rees contributed to aspects of the conception and design of CAS(K). Dr. J. Saklatvala, Ms C. Jackson, Ms J. Myatt, Ms J. Wisher, Ms S. Stoker, Ms S. Yates, and Ms K. Hickson from the Department of Radiography, Haywood Hospital conducted all study radiographs.

REFERENCES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES
  10. Supporting Information
  • 1
    Arthritis Research Campaign. Arthritis: the big picture. Chesterfield (UK): ARC; 2004. URL: http://www.arc.org.uk/about_arth/bigpic.htm.
  • 2
    World Health Organization. The burden of musculoskeletal conditions at the start of the new millennium: report of a WHO Scientific Group, Technical Report Series No 919. Geneva: WHO; 2003. URL: http://whqlibdoc.who.int/trs/WHO_TRS_919.pdf.
  • 3
    Bierma-Zeinstra SM, Lipschart S, Njoo KH, Bernsen R, Verhaar J, Prins A, et al. How do general practitioners manage hip problems in adults? Scand J Prim Health Care 2000; 18: 15964.
  • 4
    Mantyselka P, Kumposalo E, Ahonen R, Kumposalo A, Kauhanen J, Viinimaki H, et al. Pain as a reason to visit the doctor: a study in Finnish primary health care. Pain 2001; 89: 17580.
  • 5
    Morgan B, Mullick S, Harper WM, Finlay DB. An audit of knee radiographs performed for general practitioners. Br J Radiol 1997; 70: 25660.
  • 6
    Linsell L, Dawson J, Zondervan K, Rose P, Carr A, Randall T, et al. Population survey comparing older adults with hip versus knee pain in primary care. Br J Gen Pract 2005; 55: 1928.
  • 7
    The Royal College of Radiologists. Making the best use of a department of radiography: guidelines for doctors. 5th ed. London: The Royal College of Radiologists; 2003.
  • 8
    Creamer P. Current perspectives on the clinical presentation of joint pain in human OA. In: ChadwickDJ, GoodeJ, editors. Osteoarthritic joint pain. Chichester (UK): John Wiley & Sons; 2004. p. 6478.
  • 9
    Eccles M, Steen N, Grimshaw J, Thomas L, McNamee P, Soutter J, et al. Effect of audit and feedback, and reminder messages on primary-care radiology referrals: a randomised trial. Lancet 2001; 357: 14069.
  • 10
    Ramsay CR, Eccles M, Grimshaw JM, Steen N. Assessing the long-term effect of educational reminder messages on primary care radiology referrals. Clin Radiol 2003; 58: 31921.
  • 11
    Laupacis A, Sekar N, Stiell IG. Clinical prediction rules: a review and suggested modifications to methodological standards. JAMA 1997; 277: 48894.
  • 12
    Thomas E, Wilkie R, Peat G, Hill S, Dziedzic K, Croft P. The North Staffordshire Osteoarthritis Project: NorStOP. Prospective, 3-year study of the epidemiology and management of clinical osteoarthritis in a general population of older adults. BMC Musculoskelet Dis URL: http://www.biomedcentral.com/1471-2474/5/2.
  • 13
    Peat G, Thomas E, Handy J, Wood L, Dziedzic K, Myers H, et al. The Knee Clinical Assessment Study, CAS(K): a prospective study of knee pain and knee osteoarthritis in the general population. BMC Musculoskelet Disord URL: http://www.biomedcentral.com/1471-2474/5/4.
  • 14
    Peat G, Lawton H, Hay E, Greig J, Thomas E, for the KNE-SCI Study Group. Development of the Knee Standardized Clinical Interview: a research tool for studying the primary care clinical epidemiology of knee problems in older adults. Rheumatology (Oxford) 2002; 41: 11018.
  • 15
    Wood LR. The physical examination of older people with knee pain: a study of reliability and relationship to self-reported function in primary care [dissertation]. Keele (UK): Keele University; 2004.
  • 16
    Peat G, Wood L, Wilkie R, Thomas E, for the KNE-SCI Study Group. How reliable is structured clinical history-taking in older adults with knee problems? Inter- and intraobserver variability of the KNE-SCI. J Clin Epidemiol 2003; 56: 10307.
  • 17
    Wood L, Peat G, Wilkie R, Hay E, Thomas E, Sim J. Reliability of physical examination of the knee in older adults with knee problems. J Clin Epidemiol 2006; 59: 51220.
  • 18
    Buckland-Wright JC, Wolfe F, Ward RJ, Flowers N, Hayne C. Substantial superiority of semiflexed (MTP) views in knee osteoarthritis: a comparative radiographic study, without fluoroscopy, of standing extended, semiflexed (MTP), and Schuss views. J Rheumatol 1999; 26: 266474.
  • 19
    McAlindon TE, Snow S, Cooper C, Dieppe PA. Radiographic patterns of osteoarthritis of the knee joint in the community: the importance of the patellofemoral joint. Ann Rheum Dis 1992; 51: 8449.
  • 20
    Altman RD, Hochberg M, Murphy WA Jr, Wolfe F, Lesquesne M. Atlas of individual radiographic features in osteoarthritis. Osteoarthritis Cartilage 1995; 3 Suppl A: 370.
  • 21
    Altman RD, Fries JF, Bloch DA, Carstens J, Cooke TD, Genant H, et al. Radiographic assessment of progression in osteoarthritis. Arthritis Rheum 1987; 30: 121425.
  • 22
    Burnett S, Hart D, Cooper C, Spector T. A radiographic atlas of osteoarthritis. London: Springer-Verlag; 1994.
  • 23
    Lawrence JS. Rheumatism in populations. London: WM Heinemann Medical Books; 1977.
  • 24
    Duncan RC, Hay EM, Saklatvala J, Croft PR. Prevalence of radiographic osteoarthritis: it all depends on your point of view. Rheumatology (Oxford) 2006; 45: 75760.
  • 25
    Spector TD, Hart DJ, Byrne J, Harris PA, Dacre JE, Doyle DV. Definition of osteoarthritis of the knee for epidemiological studies. Ann Rheum Dis 1993; 52: 7904.
  • 26
    Felson DT, Lawrence RC, Dieppe PA, Hirsch R, Helmick CG, Jordan JM, et al. Osteoarthritis: new insights. Part 1: the disease and its risk factors. Ann Intern Med 2000; 133: 63546.
  • 27
    Doherty M. Risk factors for progression of knee osteoarthritis [letter]. Lancet 2001; 358: 7756.
  • 28
    Altman R, Asch E, Bloch D, Bole G, Borenstein D, Brandt K, et al. Special article: development of criteria for the classification and reporting of osteoarthritis: classification of osteoarthritis of the knee. Arthritis Rheum 1986; 29: 103949.
  • 29
    Claessens AA, Schouten JS, van den Ouweland FA, Valkenburg HA. Do clinical findings associate with radiographic osteoarthritis of the knee? Ann Rheum Dis 1990; 49: 7714.
  • 30
    Bagge E, Bjelle A, Eden S, Svanborg A. Osteoarthritis in the elderly: clinical and radiological findings in 79 and 85 year olds. Ann Rheum Dis 1991; 50: 5359.
  • 31
    Bijl D, Dekker J, van Baar ME, Oostendorp RA, Lemmens AM, Bijlsma JW, et al. Validity of Cyriax's concept capsular pattern for the diagnosis of osteoarthritis of hip and/or knee. Scand J Rheumatol 1998; 27: 34751.
  • 32
    Klippel JH, Dieppe PA, Ferri FF. Primary care rheumatology. London: Mosby; 1999.
  • 33
    LaValley M, McAlindon TE, Evans S, Chaisson CE, Felson DT. Problems in the development and validation of questionnaire-based screening instruments for ascertaining cases with symptomatic knee osteoarthritis: the Framingham Study. Arthritis Rheum 2001; 44: 110513.
  • 34
    WONCA International Classification Committee. ICPC-2 International Classification of Primary Care. 2nd ed. Oxford: Oxford University Press; 1998.
  • 35
    Primary Care Rheumatology Society. Knee osteoarthritis: management options. Northallerton (UK): Primary Care Rheumatology Society; 1999.
  • 36
    Ersoz M, Ergun S. Relationship between knee range of motion and Kellgren-Lawrence radiographic scores in knee osteoarthritis. Am J Phys Med Rehabil 2003; 82: 1105.
  • 37
    Angst F, Ewert T, Lehmann S, Aeschlimann A, Stucki G. The factor subdimensions of the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) help to specify hip and knee osteoarthritis: a prospective evaluation and validation study. J Rheumatol 2005; 32: 132430.
  • 38
    Creamer P, Hochberg MC. Why does osteoarthritis of the knee hurt: sometimes? Br J Rheumatol 1997; 36: 7268.
  • 39
    Brandt KD, Heilman DK, Slemenda C, Katz BP, Mazzuca S, Braunstein EM, et al. A comparison of lower extremity muscle strength, obesity, and depression scores in elderly subjects with knee pain with and without radiographic evidence of knee osteoarthritis. J Rheumatol 2000; 27: 193746.
  • 40
    Birrell F, Croft P, Cooper C, Hosie G, Macfarlane G, Silman A, and the PCR Hip Study Group. Predicting radiographic hip osteoarthritis from range of movement. Rheumatology (Oxford) 2001; 40: 50612.
  • 41
    Hosmer DW, Lemeshow S. Applied logistic regression. 2nd ed. New York: John Wiley & Sons; 2000.
  • 42
    Steyerberg EW, Harrell FE Jr, Borsboom GJ, Eijkemans MJ, Vergouwe Y, Habbema JD. Internal validation of predictive models: efficiency of some procedures for logistic regression analysis. J Clin Epidemiol 2001; 54: 77481.
  • 43
    Moons KG, Harrell FE, Steyerberg EW. Should scoring rules be based on odds ratios or regression coefficients? [letter]. J Clin Epidemiol 2002; 55: 10545.
  • 44
    Ware J Jr, Kosinski M, Keller SD. A 12-item Short-Form Health Survey: construction of scales and preliminary tests of reliability and validity. Med Care 1996; 34: 22033.
  • 45
    Zigmond AS, Snaith RP. The hospital anxiety and depression scale. Acta Psychiatr Scand 1983; 67: 36170.
  • 46
    Bellamy N. WOMAC osteoarthritis index. A user's guide. Ontario: London Health Services Centre, McMaster University; 1996.
  • 47
    Efron B. The jacknife, the bootstrap and other resampling plans. Philadelphia: Society for Industrial and Applied Mathematics; 1982.
  • 48
    Moons KG, Biesheuvel CJ, Grobbee DE. Test research versus diagnostic research. Clin Chem 2004; 50: 4736.
  • 49
    Miettinen OS. Theoretical epidemiology: principles of occurrence research in medicine. New York: John Wiley & Sons; 1985.
  • 50
    Moons KG, Grobbee DE. Diagnostic studies as multivariable, prediction research [letter]. J Epidemiol Community Health 2002; 56: 3378.
  • 51
    Bierma-Zeinstra SM, Oster JD, Bernsen RM, Verhaar JA, Ginai AZ, Bohnen AM. Joint space narrowing and relationship with symptoms and signs in adults consulting for hip pain in primary care. J Rheumatol 2002; 29: 17138.
  • 52
    Jinks C, Jordan K, Ong BN, Croft P. A brief screening tool for knee pain in primary care (KNEST). 2. Results from a survey in the general population aged 50 and over. Rheumatology (Oxford) 2004; 43: 5561.
  • 53
    Cooper C, McAlindon T, Snow S, Vines K, Young P, Kirwan J, et al. Mechanical and constitutional risk factors for symptomatic knee osteoarthritis: differences between medial tibiofemoral and patellofemoral disease. J Rheumatol 1994; 21: 30713.
  • 54
    McAlindon T, Zhang Y, Hannan M, Naimark A, Weissman B, Castelli W, et al. Are risk factors for patellofemoral and tibiofemoral knee osteoarthritis different? J Rheumatol 1996; 23: 3327.
  • 55
    Cicuttini FM, Spector T, Baker J. Risk factors for osteoarthritis in the tibiofemoral and patellofemoral joints of the knee. J Rheumatol 1997; 24: 11647.
  • 56
    Baker KR, Xu L, Zhang Y, Nevitt M, Niu J, Aliabadi P, et al. Quadriceps weakness and its relationship to tibiofemoral and patellofemoral knee osteoarthritis in Chinese: the Beijing Osteoarthritis Study. Arthritis Rheum 2004; 50: 181521.
  • 57
    Kannel WB, D'Agostino RB, Sullivan L, Wilson PW. Concept and usefulness of cardiovascular risk profile. Am Heart J 2004; 148: 1626.
  • 58
    Lacey RJ, Lewis M, Jordan K, Jinks C, Sim J. Interrater reliability of scoring of pain drawings in a self-report health survey. Spine 2005; 30: E4558.
  • 59
    Von Korff M, Ormel J, Keefe FJ, Dworkin SF. Grading the severity of chronic pain. Pain 1992; 50: 13349.

Supporting Information

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. PATIENTS AND METHODS
  5. RESULTS
  6. DISCUSSION
  7. AUTHOR CONTRIBUTIONS
  8. Acknowledgements
  9. REFERENCES
  10. Supporting Information
FilenameFormatSizeDescription
22785.pdf31KSupporting Information file 22785.pdf

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.