Impact of shoulder, elbow, and knee joint involvement on assessment of rheumatoid arthritis using the American College of Rheumatology Core Data Set

Authors


Abstract

Objective

To determine the most sensitive scoring method for assessment of rheumatoid arthritis (RA) disease activity using the American College of Rheumatology Core Data Set.

Methods

The subjects were 4,530 patients with RA (mean age 57.9 years, mean disease duration 12.7 years) who participated in a large observational cohort study of RA patients. The 68 joints assessed were classified into 15 joint areas, and each joint variable was categorized based on the presence or absence of swelling or pain in these areas. Multiple linear regression and analysis of variance were used to evaluate the significance of effects of these 15 joint areas on variables for assessment of RA disease activity such as patient's assessment of pain on a visual analog scale (VAS), patient's and physician's global assessment of disease activity on a VAS, HAQ (Health Assessment Questionnaire), and Japanese HAQ.

Results

Although the 3 most frequently affected joints were the wrist, metacarpophalangeal joints, and proximal interphalangeal joints, the 5 joints with the largest contributions to all of the variables assessed for disease activity were the shoulder, elbow, and knee joints, followed by the wrist and ankle joints. The combination of shoulder, elbow, and knee joints accounted for approximately 70% of the contribution to all the variables, while addition of the wrist and ankle joints increased this value to approximately 90%.

Conclusion

Scoring for assessment of RA disease activity would be more sensitive if separate joints such as the shoulder, elbow, knee, wrist, and ankle joints were weighted differently.

INTRODUCTION

Rheumatoid arthritis (RA) is a systemic disease characterized by the involvement of multiple joints. Variables such as tender joint count, swollen joint count, grip strength, articular indices, patient- and physician-oriented visual analog scales (VAS), functional disability indexes such as the HAQ (Health Assessment Questionnaire), quality of life assessment, and laboratory data have been used to evaluate disease activity.

In recent clinical trials, response criteria based on the core sets of the American College of Rheumatology (ACR), the European League Against Rheumatism (EULAR), and the World Health Organization/International League Against Rheumatism (WHO/ILAR) have commonly been used (1–6). The ACR core set consists of tender joint count, swollen joint count, patient's assessment of pain, patient's and physician's global assessments of disease activity, patient's assessment of physical function, and laboratory evaluation of acute-phase reactants (1). Development of the WHO/ILAR response criteria was also based on the ACR core set (4, 5). In contrast to the WHO/ILAR and ACR criteria, the EULAR response criteria are based on the Disease Activity Score (DAS) (7, 8), which consists of the Ritchie index (9), swollen joint count, erythrocyte sedimentation rate (ESR), and patient's global assessment, and has later been modified as the DAS28 (10). In these 3 sets, the impact of subjective variables assessed by patients and physicians such as the patient's assessment of pain, patient's and physician's global assessments of disease activity, patient's assessment of physical function, and disability index are high, and these variables are indispensable for assessment of RA. However, there have been few reports of objective analysis of the contributions of different joints to these parameters.

Various methods have been used to score joint tenderness and swelling in patients with RA. These methods differ in the number of joints counted, use of weighted or graded joint counts, percentage contribution of each joint or joint area to the total score, use of combined or separate measures of tender and swollen joints, and whether joints are scored individually or as joint units. In the Lansbury Articular Index (11), joint score is weighted by joint size. In the Ritchie Articular Index (9), 50 joints are evaluated using a 4-point tenderness scale. Although the initial recommendation included 68 joints for the ACR core set (1), a simplified 28-joint scoring method has been developed. The score obtained with the simplified method has been confirmed to be a valid and reliable measure for joint assessment (10, 12–14). Therefore both the ACR Committee on Outcome Measures in Rheumatoid Arthritis Clinical Trials (15) and the EULAR Standing Committee on International Clinical Studies Including Therapeutic Trials (16) have adopted reduced joint counts for clinical trials. Neither the 68-joint index (ACR core set) nor the 28-joint index (DAS28) are weighted joint scoring methods. Determining the most appropriate weighting score system for joint evaluation thus remains controversial (17).

It is important to evaluate to what extent the involvement of each joint affects the subjective variables in the ACR core set using reliable statistical techniques. Based on the results of such studies, it might be possible to determine which joints should be used in the assessment of RA. In 2000, a database for a large observational cohort of RA patients including physician's assessments, patient's assessments, and laboratory data was established at Tokyo Women's Medical University. Using this database, we examined which joint involvements have the strongest impact on variables for the assessment of RA disease activity such as patient's assessment of pain on a VAS, patient's and physician's global assessment of disease activity on a VAS, the HAQ, and the Japanese version of the HAQ (J-HAQ). Matsuda et al modified the original HAQ to measure functional status in Japanese-speaking patients with RA (18). The J-HAQ has demonstrated excellent reliability and validity.

PATIENTS AND METHODS

Cohort database and RA patients in this study.

Since April 2000, all RA patients who have visited our clinic have been asked to answer questionnaires every 6 months assessing variables such as disability level, assessment of pain, global assessment of disease activity, use of medication, adverse events, health care utilization, and satisfaction with care. A combined database was prepared including patient-reported data, physician's assessment, and laboratory data. A total of 4,998 patients with RA participated in this cohort study in October 2002. Those patients with missing data in patient's or physician's assessments, and those whose diagnosis of RA was eventually changed were excluded. A cross-sectional analysis was performed on the remaining 4,530 patients (mean age 57.9 years, mean disease duration 12.7 years) who had received a diagnosis of RA from one of 43 specialized rheumatologists in our institution according to the 1987 classification criteria for RA of the ACR (formerly, the American Rheumatism Association) (19).

Assessments using the ACR Core Data Set.

For assessment of RA disease activity and functional disability, the patient's assessment of pain (patient's pain VAS), patient's global assessment of disease activity (patient's global VAS), physician's global assessment of disease activity (physician's global VAS), HAQ, and J-HAQ scores were used. The patient's pain VAS, patient's global VAS, and physician's global VAS were evaluated using a 100-mm double-anchored VAS where 0 = very well, and 100 = very poor. The 68 joints (68 joints for tenderness and 66 for swelling) included in the ACR core set were assessed.

Statistical analysis.

The 68 joints were classified into 15 joint areas: jaw, shoulder, sternoclavicular joint, acromioclavicular joint, elbow, wrist, metacarpophalangeal (MCP) joints, proximal interphalangeal (PIP) joints, distal interphalangeal (DIP) joints, hip, knee, ankle, forefoot, metatarsophalangeal (MTP) joints, and interphalangeal joints. The forefoot joint area included the transverse tarsal joint (Chopart joint) and the tarsometatarsal joint (Lisfranc joint). Right and left joints were regarded as one variable. Each of the 15 joint variables (independent variables) could have one of the following 3 values: level 2 (presence of swelling or pain in both joints), level 1 (presence of swelling or pain in 1 joint), and level 0 (presence of swelling or pain in neither joint).

The stepwise multiple linear regression method was applied to the data by setting each of the ACR core set measures (patient's general VAS, patient's pain VAS, physician's general VAS, HAQ, and J-HAQ) as a dependent variable. Joint variables were designated as independent variables. Variables such as laboratory data were not included as dependent variables. After a set of significant independent variables was selected, the relative contribution of each independent variable to the contribution of total independent variables was estimated by the analysis of variance method. Relative contribution was expressed as the variance of the dependent variable due to the variation of each joint variable divided by the summed variances due to the variations of all the joint variables.

RESULTS

Clinical features of the patients with RA.

The clinical features of the 4,530 patients with RA are shown in Table 1. Overall disease activities in these patients were relatively low, as reflected in mean tender joint count, swollen joint count, C-reactive protein titer, and ESR. The overall disability index was also relatively low, as demonstrated by the mean J-HAQ score. The mean patient's pain and general VAS scores were higher than the mean physician's general VAS (patient's pain VAS versus patient's general VAS r = 0.85, patient's pain VAS versus physician's general VAS r = 0.46, physician's general VAS versus patient's general VAS r = 0.42). Percentage use of disease-modifying antirheumatic drugs was 86.0% and the average dosages of methotrexate and corticosteroids were relatively low for these patients.

Table 1. Baseline features of 4,530 patients with rheumatoid arthritis*
  • *

    Values are mean ± SD unless otherwise indicated. RA = rheumatoid arthritis; VAS = visual analog scale; HAQ = Health Assessment Questionnaire; J-HAQ = Japanese HAQ; DAS = Disease Activity Score; NSAID = nonsteroidal antiinflammatory drug; DMARD = disease-modifying antirheumatic drug; MTX = methotrexate.

Age, years57.9 ± 12.7
Female, %83.0
RA disease duration, years11.0 ± 8.8
Tender joint count (68 joints)2.9 ± 5.0
Swollen joint count (66 joints)2.8 ± 3.9
C-reactive protein (mg/100 ml)1.3 ± 1.7
Erythrocyte sedimentation rate (mm/hour)34.7 ± 23.8
Patient's pain VAS (mm)30.8 ± 26.2
Patient's general VAS (mm)32.3 ± 25.4
Physician's general VAS (mm)16.7 ± 16.0
HAQ score0.75 ± 0.73
J-HAQ score0.79 ± 0.76
DAS283.6 ± 1.2
Positive rheumatoid factor, %75.2
NSAID use, %74.2
DMARD use, %87.2
MTX use, %44.9
Dosage of MTX per week (mg)2.7 ± 3.6
Corticosteroid use, %52.1
Corticosteroid dosage (mg/day)2.3 ± 3.0

Involvement of each joint.

The 2 most frequently affected joints were the wrist (45.5%) and MCP joints (34.4%), followed by the PIP joints (28.8%), knee (27.3%), ankle (23.3%), elbow (20.7%), shoulder (11.5%), and MTP joints (10.8%). In contrast, the frequencies of DIP joint and hip joint involvements were small (Figure 1).

Figure 1.

Percentage involvement of each joint in rheumatoid arthritis patients. MCP = metacarpophalangeal joints; PIP = proximal interphalangeal joints; DIP = distal interphalangeal joints; MTP = metatarsophalangeal joints; IP = interphalangeal joints. Dark bars indicate bilateral affected joints. Striped bars indicate unilateral affected joints.

Impact of each joint on variables used for assessment of disease activity.

Because the contribution of each independent variable was expressed as a relative value (i.e., the variance due to the variation of each independent variable divided by the summed variances due to the variations of all independent variables not including the residuals), the summed contributions become 100%. This does not necessarily mean that there are no contributions of the variables other than the joints.

Analysis of variance showed that the 3 joints with the greatest contributions to all assessment variables were the shoulder, elbow, and knee joints, followed by the wrist and ankle joints (Figure 2).

Figure 2.

Relative contribution of various joints to each disease-activity assessment of rheumatoid arthritis in 4,530 patients. The combination of the shoulder, elbow, and knee joints accounted for approximately 70% of the contribution to these variables, while the addition of the wrist and ankle joints increased contribution to approximately 90%. VAS = visual analog scale; HAQ = Health Assessment Questionnaire; J-HAQ = Japanese HAQ; MCP = metacarpophalangeal joints; PIP = proximal interphalangeal joints.

The shoulder, knee, and elbow joints were the 3 joints with the greatest contribution to patient's pain VAS score, followed by the wrist, ankle, and MCP joints. The combination of the shoulder, elbow, and knee joints accounted for 67.8% of the contribution to patient's pain VAS, while adding the wrist and ankle joints increased the contribution to 87.2% (Table 2).

Table 2. Relative contribution of each joint to patient's pain visual analog scale in 4,530 patients with rheumatoid arthritis*
Joint variable% ContributionCoefficients
Level 1Level 2
  • *

    Level 1 = presence of swelling or pain in 1 joint; Level 2 = presence of swelling or pain in both joints; MCP = metacarpophalangeal joints; PIP = proximal interphalangeal joints.

Shoulder24.78.210.1
Knee22.59.615.3
Elbow20.67.79.4
Wrist13.04.36.7
Ankle6.46.18.6
MCP5.03.93.6
Hip2.814.26.8
Jaw2.71.41.7
Acromioclavicular1.14.35.1
PIP0.41.11.0

The shoulder, knee, and elbow joints were the 3 joints with the greatest contribution to patient's general VAS, followed by the wrist, ankle, and MCP joints. The first 3 joints together accounted for 67.3% of the contribution, while addition of the wrist and ankle joints increased contribution to 86.5% (Table 3). The wrist, knee, and elbow joints were the 3 joints with the greatest contributions to physician's general VAS, followed by the shoulder, ankle, and MCP joints. The combination of these 5 joints accounted for 84.0% of the entire contribution (Table 4).

Table 3. Relative contribution of each joint to patient's general visual analog scale in 4,530 patients with rheumatoid arthritis*
Joint variable% ContributionCoefficients
Level 1Level 2
  • *

    See Table 2 for definitions.

Shoulder25.27.39.9
Knee25.18.814.8
Elbow17.05.86.8
Wrist12.83.15.8
Ankle6.45.37.9
MCP5.93.82.9
Hip2.411.26.5
Jaw2.0−2.5−0.5
PIP1.32.12.1
Acromioclavicular0.62.60.5
Sternoclavicular0.3−7.5−1.1
Table 4. Relative contribution of each joint to physician's general visual analog scale in 4,530 patients with rheumatoid arthritis*
Joint variable% ContributionCoefficients
Level 1Level 2
  • *

    MTP = metatarsophalangeal joints; DIP = distal interphalangeal joints. See Table 2 for additional definitions.

Wrist23.54.37.9
Knee23.26.613.3
Elbow16.53.95.9
Shoulder14.21.93.7
Ankle6.64.96.9
MCP6.32.43.2
Jaw2.34.6−5.9
MTP1.91.76.1
Acromioclavicular1.74.85.4
Hip1.67.85.3
PIP1.41.51.8
DIP0.42.13.7

The shoulder, knee, and elbow joints were also the 3 joints with the greatest contributions to HAQ scores, followed by the wrist, ankle, and hip joints. The combination of the former 3 joints accounted for 71.9% of the entire contribution, while addition of the wrist and ankle joints increased contribution to 88.3% (Table 5). The combination of the first 3 joints accounted for 72.5% of the contribution, while the addition of the wrist and ankle joints increased contribution to 88.4% (Table 6).

Table 5. Relative contribution of each joint to Health Assessment Questionnaire in 4,530 patients with rheumatoid arthritis*
Joint variable% ContributionCoefficients
Level 1Level 2
  • *

    Level 1 = presence of swelling or pain in 1 joint; Level 2 = presence of swelling or pain in both joints; MCP = metacarpophalangeal joints; IP = interphalangeal joints; PIP = proximal interphalangeal joints; DIP = distal interphalangeal joints.

Shoulder28.30.210.40
Knee23.20.220.45
Elbow20.40.220.26
Wrist11.80.070.19
Ankle4.60.150.20
Hip3.40.460.10
MCP3.30.090.12
Jaw2.40.070.02
Acromioclavicular0.90.21−0.11
IP0.4−0.150.00
PIP0.4−0.02−0.10
DIP0.40.14−0.20
Table 6. Relative contribution of each joint to Japanese Health Assessment Questionnaire in 4,530 patients with rheumatoid arthritis*
Joint variable% ContributionCoefficients
Level 1Level 2
  • *

    See Table 5 for definitions.

Shoulder27.40.210.41
Knee23.40.230.48
Elbow21.70.230.30
Wrist11.20.070.19
Ankle4.70.160.21
MCP3.60.100.13
Hip3.40.480.11
Jaw2.30.050.03
Acromioclavicular0.80.21−0.12
IP0.5−0.160.00
DIP0.30.13−0.22
PIP0.3−0.02−0.10

A summary of the relative contributions of various joints to each disease assessment variable is shown in Figure 2. The 3 joints with the largest contributions to all of the variables were the shoulder, elbow, and knee joints, followed by the wrist and ankle joints. The combination of the shoulder, elbow, and knee joints accounted for approximately 70% of the contribution to these variables, while the addition of the wrist and ankle joints increased contribution to approximately 90%. On the other hand, the relative contributions of the finger, toe, forefoot, hip, sternoclavicular, acromioclavicular, and jaw joints were small.

DISCUSSION

The combinations of the shoulder, elbow, and knee joints accounted for ∼70% of the entire contribution to each variable. This figure increased to ∼90% when the wrist and ankle joints were added to the calculation (Figure 2). These results indicate that the score for each variable can be predicted with 90% probability if only 5 joints (i.e., the shoulder, elbow, knee, wrist, and ankle joints) are evaluated bilaterally. The contributions of the finger, toe, and hip joints to each variable were small. One might have expected involvement of the hip joint to have a greater impact on the variables; however, the low frequency of involvement of the hip joint may have prevented it from being extracted as a joint making major contributions to the variables.

Determining the number of joints to be evaluated when assessing RA disease activity has been a subject of controversy (20). Recent reports have described the 28-joint count scoring method as simple to perform, reliable, and valid for joint assessment (10, 12–14). However, Johnson reported that it is important to evaluate the joints in the feet and that these joints should not be ignored in clinical practice (21).

The percentage contributions of shoulder, elbow, wrist, knee, and ankle joints, when affected, to the total joint indices of various outcome measurements of RA were quite different. Specifically, they were all 2.9% in the 68-joint index of the ACR core set (1), and all 7.7% in the Ritchie Articular Index (9). In the 28-joint index of the DAS28 (10) and in the Fuchs index (22) the total joint indices were 7.1%, 7.1%, 7.1%, 7.1%, and 0%; in the Lansbury Articular Index (11) they were 8.3%, 8.3%, 6.2%, 17.2%, and 6.2%; in the Egger Articular Index (23) they were 0%, 5.3%, 5.3%, 5.3%, and 5.3%; and in both the Thompson-Kirwan Index (24) and Stewart Articular Index (25), they were 0%, 18.0%, 12.0%, 35.6%, and 12.0%, respectively, for the shoulder, elbow, wrist, knee, and ankle joints. Prevoo et al found that weighted joint indices did not differ substantially from nonweighted joint indices in the evaluation of disease activity of RA, and recommended use of the 28-joint index based on its simplicity, validity, and reliability (10). Fuchs et al also suggested that weighted scoring system did not actually add substantially to the information derived from a joint count (22). However, the results obtained with weighted scoring systems of the Thompson-Kirwan and Stewart Articular Index appear to support our results, with the exception of the shoulder joint. These scoring systems, which weigh separate joints such as the shoulder, elbow, and knee joints differently, could be more sensitive to disease activities of RA, as the results of our study suggested. It is therefore difficult to conclude how many joints should be measured or which weighting score system should be used for joint evaluation. However, inclusion of the shoulder, elbow, wrist, knee, and ankle joints is strongly recommended and there is a need for an emphasis on joints that are not in the hands.

Because the average number of tender and swollen joints was low in our study, the patients were considered to have low RA disease activity. Our conclusions should not be applied to patients with high RA disease activity, who are often enrolled in clinical trials, but to patients who have well-controlled disease activity. We performed a subset analysis of patients with ≥6 tender and ≥6 swollen joints, and those with <6 tender and swollen joints (Tables 7 and 8). The results from this subset analysis were similar to the results in the 4,530 patients in the main study. However, patients in the subset analysis ranked jaw and hip joint involvement higher than when the total number of patients were included in the analysis (Table 8). This is probably because the proportion of the patients in whom the 2 joints were affected was very low among all the patients but was higher in the patients with >5 joints involved. Because involvement of the hip joints occurs rarely, the contribution of the joints was not very high in this study. However, P values were low (<0.05) for only the top 2 to 3-ranked joints in this subset analysis, perhaps due to the small sample size (∼360) of patients with ≥6 tender and swollen joints, and this conclusion should be interpreted with care.

Table 7. Patient characteristics of subset analysis*
CharacteristicAll patients (n = 4,530)<6 tender and swollen joints (n = 3,427)≥6 tender and swollen joints (n = 360)
  • *

    Values are the mean ± SD. RA = rheumatoid arthritis; HAQ = Health Assessment Questionnaire; J-HAQ = Japanese HAQ; DAS28 = Disease Activity Score.

Age, years57.9 ± 12.758.2 ± 12.657.1 ± 12.3
RA disease duration, years11.0 ± 8.810.7 ± 8.511.9 ± 9.0
Tender joint counts (68 joints)2.9 ± 5.01.3 ± 1.413.6 ± 8.2
Swollen joint counts (66 joints)2.8 ± 3.91.3 ± 1.410.6 ± 5.5
C-reactive protein (mg/100 ml)1.3 ± 1.71.1 ± 1.52.4 ± 2.3
HAQ score0.75 ± 0.730.64 ± 0.691.24 ± 0.73
J-HAQ score0.79 ± 0.760.68 ± 0.711.32 ± 0.75
DAS283.6 ± 1.23.2 ± 1.05.7 ± 1.0
Table 8. Relative contribution of each joint to each dependent variable in the analysis of the subset of patients with rheumatoid arthritis*
Relative contributionAll patients<6 tender and swollen joints≥6 tender and swollen joints
Joint variableContributionJoint variableContributionJoint variableContribution
  • *

    Values are the percentage contribution. VAS = visual analog scale; HAQ = Health Assessment Questionnaire; PIP = proximal interphalangeal joints; J-HAQ = Japanese HAQ.

Patient's pain VASShoulder24.7Knee34.8Shoulder27.3
 Knee22.5Elbow16.5Jaw13.3
 Elbow20.6Ankle15.0Knee11.1
 Wrist13.0Wrist13.0Hip10.4
 Ankle6.4Shoulder10.6Ankle10.4
Patient's general VASShoulder25.2Knee37.7Shoulder33.2
 Knee25.1Ankle14.2Knee16.1
 Elbow17.0Elbow13.6Jaw13.2
 Wrist12.8Wrist11.8Ankle11.5
 Ankle6.4Shoulder11.7Hip6.4
Physician's general VASWrist23.5Knee36.0Knee30.7
 Knee23.2Wrist25.4Jaw20.0
 Elbow16.5Elbow14.5Shoulder11.8
 Shoulder14.2Ankle10.4Ankle7.8
 Ankle6.6Shoulder3.9Acromioclavicular6.2
HAQShoulder28.3Knee33.7Shoulder31.3
 Knee23.2Elbow21.4Jaw12.8
 Elbow20.4Shoulder12.2Knee10.6
 Wrist11.8Wrist10.8PIP10.1
 Ankle4.6Ankle9.8Hip8.5
J-HAQShoulder27.4Knee33.4Shoulder29.0
 Knee23.4Elbow22.4Knee12.9
 Elbow21.7Shoulder12.3Jaw11.4
 Wrist11.2Wrist9.9Hip9.0
 Ankle4.7Ankle9.9PIP7.7

Response criteria using the ACR core set and EULAR core set are known to be effective for use in comparatively short-term clinical trials. However, the long-term functional prognosis of the joints, as well as that of disability indexes like the HAQ score, is more important for RA patients. In recent clinical trials of treatment of patients with RA, either the ACR core set or EULAR core set was used. In those core sets, although the numbers of joints evaluated differed, all joints were treated equally in scoring. Our results suggest that certain joints, especially the shoulder, elbow, and knee joints, should be treated separately from the other joints because their impact on patient-oriented outcome measures are markedly stronger than those of the other joints. Although weighted-joint scoring methods are effective for evaluating patient-oriented disease activity, we have not evaluated their effectiveness in determining response to treatment. In conclusion, shoulder, elbow, and knee joint involvement have a significant impact on variables of RA disease activity, specifically with regard to patient-oriented assessment.

Ancillary