Patient-derived Joint Counts Are a Potential Alternative for Determining Disease Activity Score

MATERIALS AND METHODS

Patients

Data were derived from ETN-naive patients (n = 447) with active RA (≥ 18 yrs old) participating in an open-label study assessing the immunogenicity of ETN⁷. Patients were randomized to receive subcutaneous ETN 50 mg or ETN 50 mg + MTX once weekly for 24 weeks. All patients provided written informed consent, and all study procedures were approved by the appropriate institutional ethics committee.

Assessments

Joint evaluations by the patient and his or her physician were performed during the study visit. The patient’s evaluation was performed first, followed by the physician’s evaluation during the same study visit. The physicians were blinded to the results of the patients’ evaluations and were not involved in the treatment strategies of the clinical trial. Neither patient nor physician was aware of the other’s evaluation. Joint assessments included the 28-joint count for tenderness and swelling as established by the European League Against Rheumatism (EULAR)⁸. Assessments were performed at baseline, Week 12, and Week 24 or termination of the study.

To train patients in performing tender joint self-assessments, instructions were given to “Please put an ‘X’ on each joint you see on a mannequin that corresponds to the joints that hurt you. There may be pain when you press or move your joints.” For swollen joint counts, patients were instructed to “Please put an ‘X’ on each joint you see on a mannequin that corresponds to the joints you feel are swollen. The joints may feel swollen or look swollen”⁹. DAS28 and Pt-DAS28 were calculated from these joint counts using the formula: DAS28=0.56×√TEN28  (Physician or Patient)+0.28×√SW28(Physician or Patient)+0.36×ln (CRP+1)+0.014×PtGA+0.96

DAS28 and Pt-DAS28 scores range from 0 to 9.4. Disease activity can be categorized as low (≤ 3.2), moderate (> 3.2 to ≤ 5.1), or high (> 5.1) based on the DAS28 score¹⁰. Baseline and posttreatment DAS28 and Pt-DAS28 scores were further used to identify EULAR responders, that is, those with “good” or “moderate” responses to treatment (Table 1)¹¹.

Patients also assessed physical function at baseline, Week 12, and Week 24 using the Health Assessment Questionnaire Disability Index (HAQ-DI)^12,13. Two other composite scores, the Clinical Disease Activity Index (CDAI)¹⁴ and Simplified Disease Activity Index (SDAI)¹⁵, were each calculated using either physician joint counts or patient joint counts at baseline, Week 12, and Week 24.

Statistics

Patients who received at least 1 dose of ETN or ETN + MTX were included in the analyses. Convergent validity was assessed by Spearman correlation using rank scores and Pearson product-moment correlation. Intraclass correlation coefficients were used to measure agreement between the continuous variables of physician and patient tender and swollen joint counts. Kappa statistics were used to determine agreement between the noncontinuous variables of physician and patient individual joint assessments. Further, Bland-Altman plots were included to visually assess the agreement between physician and patient tender and swollen joint counts across different levels of joint involvement. Agreement between DAS28 and Pt-DAS28 in classifying patients’ disease as low, moderate, or high was measured by kappa statistics¹⁶. EULAR responses based on DAS28 scores were used as a standard to measure the sensitivity and specificity of the EULAR response based on Pt-DAS28 scores. Logistic regression analysis, adjusting for age, sex, and treatment, was used to determine differences in physician- and patient-derived joint counts between baseline, Week 12, and Week 24.

RESULTS

Patients

A total of 447 patients were enrolled; 218 received ETN alone and 229 received ETN + MTX. Baseline demographics were similar between treatment groups (Table 2). Most patients were women (79%), white (81%), and < 65 years of age (77%; mean 54.5 yrs old; range 19–88 yrs). At baseline, mean DAS28 (5.4 vs 5.0) and Pt-DAS28 (5.5 vs 5.1) were similar between the ETN and ETN + MTX arms. Mean RA disease duration was 8.5 years in ETN-treated patients compared with 7.2 years in ETN + MTX-treated patients.

Tender and swollen joint assessments

The results of physician-derived TJC and SJC were strongly correlated with the results of patient-derived TJC and SJC (Table 3). The proportion of patients with differences of ≤ 3 joints between physician and patient TJC was 52.6% at baseline, 67.6% at Week 12, and 65.4% at Week 24. Similarly, the proportions of patients with SJC differences of ≤ 3 were 46.6%, 58.6%, and 61.7% at baseline, Week 12, and Week 24, respectively. At Week 24, the intraclass correlation coefficients (r) between physician and patient TJC (r = 0.7771) were greater than those for SJC (r = 0.4341). There were no clinically relevant differences in correlations between tender or swollen joints when the cohort was stratified by ETN or ETN + MTX treatment (data not shown).

The correlation between CDAI scores calculated using either physician or patient joint counts was 0.73 or higher at baseline, Week 12, and Week 24 (Table 3). Similarly, the correlation between SDAI scores calculated using physician or patient joint counts was 0.75 or higher at baseline, Week 12, and Week 24. Thus, the DAS28, CDAI, and SDAI all show a high degree of correlation when calculated using either physician or patient joint counts over the 24-week study period.

To assess whether baseline differences resulted in differences at 12 and 24 weeks, logistic regression models were used to adjust for covariates. When physician and patient baseline assessments differed by ≥ 5 tender/swollen joints, patients were 4.7 (95% CI 2.87–7.60) and 2.7 (95% CI 1.71–4.21) times more likely to have such differences at 12 and 24 weeks, respectively, compared with patients without such differences at baseline.

To assess agreement between the physician and patient assessments across varying levels of joint involvement, Bland-Altman plots were used to illustrate differences between physician and patient joint counts versus their average at baseline and at Week 24 (Figure 1). At baseline, the mean absolute difference between physician and patient tender and swollen joints was 4.81 and 5.66, respectively (Figure 1A, 1C). By Week 24, the mean absolute difference between physician and patient tender and swollen joint counts decreased compared with baseline and was 3.86 and 4.37, respectively. In addition, the average number of joint counts decreased, indicating disease improvement due to treatment (Figure 1B, 1D). Generally, the largest variation between physician and patient assessment occurred at intermediate levels of joint involvement, or between 10 to 20 joints. More patients tended to view their disease condition to be worse than their physician, as evidenced by more data points clustering below the mean.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Bland-Altman plots of agreement between physician and patient tender and swollen joint counts over time. For each plot (+) represents the difference in joint counts assessed by the physician and the patient. Reference lines above and below zero represent the mean of the absolute difference between physician and patient joint counts for the analysis. Joint differences plotted at zero represent perfect agreement; positive joint count differences represent a higher physician joint count compared with the patient joint count; negative joint count differences indicate that patients counted more joints than the physician. When assessing a plot in total, a diamond-shaped geometry represents good correlation at low and high joint numbers with less agreement at an intermediate number of joints. A. Tender joint counts at baseline. The mean of the absolute difference between physician and patient joint counts was ± 4.81. B. Tender joint counts at Week 24. The mean of the absolute difference between physician and patient joint counts was ± 3.86. C. Swollen joint counts at baseline. The mean of the absolute difference between physician and patient joint counts was ± 5.66. D. Swollen joint counts at Week 24. The mean of the absolute difference between physician and patient joint counts was ± 4.37.

When individual joints were scored by physicians and patients for tenderness and swelling, the greatest agreement was observed for larger joints, particularly shoulders, knees, and elbows (Table 4). Agreement was less frequent when assessing swollen joints, with the exception of the knee (kappa = 0.40–0.55). Agreement was similar across time-points and treatment groups. Similarly, HAQ-DI moderately correlated with physician- and patient-derived TJC (0.54, 0.58 at Week 12 and 0.51, 0.54 at Week 24, respectively) and less with SJC (0.38, 0.39 at Week 12 and 0.31, 0.48 at Week 24, respectively; Table 3).

Agreement of DAS28 and Pt-DAS28

Agreement between DAS28 and Pt-DAS28 was high throughout the study (Figure 2). For DAS28, 8%, 35%, and 57% of patients could be classified as having low, moderate, or high disease activity at baseline, respectively. For Pt-DAS28, 7%, 36%, and 57% had low, moderate, and high disease activity at baseline. Of the 34 patients who had low disease activity at baseline by physician assessment, 14 (41%) were also categorized as having low disease activity by patient assessment. Of the 256 patients with high disease activity by physician assessment, 218 (85%) also had high disease activity by patient assessment. At Week 24, 48%, 39%, and 12% of patients were classified as having low, moderate, or high disease activity by DAS28 and 43%, 39%, and 18% by Pt-DAS28. Overall agreement across disease severity was 75% at baseline, 73% at Week 12, and 72% at Week 24. Kappa statistics for agreement between physician and patient assessments of disease severity were 0.54 (95% CI 0.48–0.61) at baseline, 0.57 (95% CI 0.50–0.63) at Week 12, and 0.55 (95% CI 0.48–0.62) at Week 24. Similar results were found when patients were assessed according to treatment.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Agreement between DAS28 and Pt-DAS28 at baseline (n = 447), Week 12 (n = 416 and n = 414) and Week 24 (n = 378 and n = 379), respectively. DAS and Pt-DAS expressed as mean ± SD. DAS28: 28-joint count Disease Activity Score; Pt-DAS28: patient DAS28.

Agreement between EULAR responses using DAS28 and Pt-DAS28

When DAS28 and Pt-DAS28 were used to classify patients according to EULAR criteria, 78% and 72% of patients, respectively, were classified as EULAR responders (good or moderate) at Week 24. Sensitivity (0.83) and specificity (0.67) were high, suggesting good agreement between EULAR response based on DAS28 and Pt-DAS28. At 24 weeks, 79% of patients showed agreement in EULAR response using DAS28 or Pt-DAS28 (kappa = 0.45).

DISCUSSION

This study demonstrated good agreement between physician- and patient-derived DAS28 scores over time in patients treated with ETN. Physician- and patient-derived TJC and SJC were well correlated, with better agreement noted for tender versus swollen joints. The HAQ-DI was similarly correlated with physician- and patient-derived tender and swollen joint counts. EULAR response rates for patients were similar when DAS28 and Pt-DAS28 were used to assess disease activity. Together, these data support the possible use of patient-derived tender and swollen joint counts to help determine disease activity in patients with RA.

Other studies have compared the level of agreement between physician and patient assessment of TJC and SJC, and, consistent with the data presented here, stronger correlations have been observed between physicians and patients for tender joints compared with swollen joints^9,17–19. One explanation for this finding is that physicians may rely more heavily on patient input when assessing tenderness, whereas physical measurements of synovitis may play a greater role when assessing swelling. In addition, swelling may be difficult for patients to assess compared with pain. Typically, there is a higher correlation between physician and patient assessments with pain compared with swelling¹⁸. The weaker correlation seen with swollen joints may be interpreted as a limitation of the DAS.

When assessments of tender and swollen joints were compared individually, better agreement was generally observed for large joints than small joints and for tender joints than swollen joints. Although this trend was consistent among most joints, the knee joint assessment showed good agreement for both tenderness and swelling. The reason for this difference is unclear, but agreement was consistent over the study duration. Potentially, swelling may be easier to determine in the knee compared with other large joints, such as the shoulder.

Agreement between physician and patient assessments of tender and swollen joints remained stable over the course of the study. Although there can be considerable variation in determining joint involvement in patients with RA, training for both medical staff and patients has been shown to significantly increase the accurate detection of both tender and swollen joints^18,20. However, even among trained medical professionals with RA experience, substantial variation has been shown to occur when different assessors examine the same patients²⁰. A potential means of reducing variation in joint counts would be to use data collected from the patient, which might be more consistent because joint assessment would be performed routinely by only 1 person. In addition, appropriate patient training would be expected to increase the accuracy of joint assessments over time and with experience¹⁸. Although patient self-assessment may be beneficial for many patients, the limitations of some patients in performing this assessment should be considered. In general, patients tended to have higher joint counts than physicians, and despite repeated assessments over time, some patients in this study with large differences between physician assessment and patient self-assessment at baseline had similarly large differences in subsequent measurements. Perhaps some patient self-assessments may be affected by musculoskeletal conditions unrelated to RA disease activity, while physicians may tend to focus more on joint pain and swelling secondary to RA.

Another possible limitation to our results was the considerable average difference between physician and patient SJC at baseline (5.7) and Week 24 (4.4). This is potentially important given that SJC is a stronger predictor of structural damage than TJC²¹. The good correlation of physician and patient DAS may be due in part to the stronger agreement with TJC between physicians and patients and heavier weighting of tender versus swollen joint counts in the DAS28. However, patient and physician correlations in scales that equally weigh tender and swollen joint counts such as the CDAI and SDAI appear to have correlations similar to the DAS28. Nonetheless, the overall results from this analysis support the use of patient self-assessment of tender and swollen joint counts during RA treatment. Patient-derived joint counts may be beneficial when physician time and resources are limited, but they are not meant to replace the physician assessment. Patient self-assessment is likely to result in more frequent monitoring, which may allow more rapid medication adjustments by the physician, thereby improving patient outcomes². Additionally, self-monitoring may increase patient involvement and prompt patients to seek changes in treatment earlier. Further studies are warranted to determine whether the Pt-DAS has predictive value for radiographic outcomes and functional status over time.