SEARCH

SEARCH BY CITATION

Abstract

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

Objective

To compare the reliability of 3 different simplified joint counts with the gold standard 66 swollen/68 tender joint count (JC66/68) for assessing clinical response in patients with polyarticular psoriatic arthritis (PsA).

Methods

The 28–joint count (JC28), in the same way that it is used in rheumatoid arthritis, and 2 measures including distal interphalangeal (DIP) joints (the 32–joint count [JC32], including all finger joints as well as wrists and knees, and 36–joint count [JC36], which additionally included elbows and ankles), were compared with the JC66/68 in 182 patients using data from the Infliximab Multinational Psoriatic Arthritis Controlled Trial 2 trial database. Pearson's correlation coefficients were calculated to compare the swollen and tender JC28, JC32, and JC36 with the corresponding results of the total JC66/68. American College of Rheumatology (ACR) responses based on the individual measures were compared, and their ability in predicting a clinical response of ACR 20% improvement (ACR20) based on the JC66/68 was assessed by calculating the area under the receiver operating characteristic curve via logistic regression and the maximum Youden indices at weeks 14 and 24.

Results

All simplified joint counts were highly correlated to the standard JC66/68 both for tenderness and swelling at each individual visit (Pearson's correlation coefficients consistently >0.8, n = 182–200; P < 0.0001). Logistic regression for ACR20 response showed that area under the curve was constantly >0.91, with comparable results for Youden indices of the simplified joint counts.

Conclusion

All simplified joint counts considered seemed sufficiently sensitive and specific to measure clinical response in trial patients with polyarticular PsA when compared with the JC66/68, no matter whether DIP joints were included (the JC36 and JC32) or excluded (the JC28). Further research will be needed to clarify this issue.


INTRODUCTION

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

A considerable number of patients with psoriasis show musculoskeletal symptoms, and ∼5–10% of them (1, 2), and in recent investigations even 20–30% of them (3–5), are experiencing psoriatic arthritis (PsA). Like other destructive forms of arthritis, PsA responds well to treatment with conventional disease-modifying antirheumatic drugs such as methotrexate (MTX) (6) or leflunomide (7), as well as biologic agents such as tumor necrosis factor α (TNFα) inhibitors, to reduce pain and inflammation and ideally achieve protection from structural damage. During the last decade, considerable progress has been made in the diagnosis, monitoring, and treatment of PsA in the course of multiple clinical trials, which have particularly studied the efficacy and safety of TNFα inhibitors in PsA (8–11).

These achievements markedly improved the management of PsA in daily clinical practice and led to suggestions in assessing the disease-related activity (12–16). Although treatment strategies in PsA and rheumatoid arthritis (RA) are similar with regard to MTX and TNFα inhibitors, there are crucial differences in the clinical picture of both diseases. With the possible occurrence of dactylitis and enthesitis aside from the frequent appearance of nail or skin involvement, PsA shows a more widespread spectrum of symptoms than RA, representing a challenge for simple and global assessment of disease activity. One key challenge related to this topic is an appropriate pattern of assessment to capture joint involvement. Whereas RA primarily affects the wrist, metacarpophalangeal (MCP), and proximal interphalangeal (PIP) joints of both hands and normally spares the distal interphalangeal (DIP) joints, the pattern of joint involvement in PsA is usually asymmetric and frequently involves the DIP joints (17–22).

So far, joint assessment in PsA has been accomplished using the American College of Rheumatology (ACR) 66/68–joint count (JC66/68) (23), which has been established in most clinical trials and monitors the tenderness and swelling of virtually all peripheral joints. This method allows a global analysis of inflammatory involvement in PsA by capturing a maximum number of joints and minimizing chances to miss affected ones. In routine practice, the JC66/68 is, for practical reasons, rarely used, and rheumatologists often use a scheme including 28 joints. However, it remains to be seen whether a 28–joint count (JC28) neglecting the DIP joints and originally designed for patients with RA is as useful in patients with PsA as compared with simplified joint counts including the DIP joints.

Fransen and colleagues have previously shown that calculation of disease activity based on a JC28 (the Disease Activity Score in 28 joints [DAS28]) is sensitive to change and may reliably distinguish patients with PsA treated with placebo from those treated with TNFα inhibitors in clinical trials (11). These data indicate that a simplified joint count such as the JC28, which excludes the DIP joints, may be regarded as a useful tool in the assessment of polyarticular PsA. Considering these observations, the DAS28 has been studied in PsA and compared with RA (24). However, in contrast to the unidimensional meaning of the DAS28 in RA, principal component analysis showed that the meaning of the DAS28 in PsA is rather bidimensional, which is most likely due to skin involvement (24). Thus, some components of the DAS28 appear to have additional meaning in PsA, which reflects the diversity of the symptoms reported. Before a set of variables forming a simplified disease activity score such as the DAS28 can be validated in PsA, the most suitable composition of a simplified joint count needs to be defined, owing to the fact that the distribution of joint involvement in PsA is somewhat different from that in RA.

This study aimed to compare the performance of the complete but laborious JC66/68 with 3 different, easier to perform scoring methods that are based on limited numbers of frequently involved joints (28, 32, and 36 joints) in patients with polyarticular PsA. In particular, the study aimed to address the impact of assessment of DIP joints in polyarticular PsA and to deal with the question of whether simplified joint counts based on counting DIP joints perform any better than those excluding them.

PATIENTS AND METHODS

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

Patients.

For comparing the results of total and simplified joint counts, we considered the swollen joint counts (SJCs) and tender joint counts (TJCs) derived from the Infliximab Multinational Psoriatic Arthritis Controlled Trial 2 (IMPACT 2) database (8), which appraised the efficacy of infliximab therapy in PsA. Following this approach, data from 182 patients included from December 20, 2002 (first patient, first visit) to January 22, 2004 (last patient, last visit) were analyzed with regard to the results of joint assessments at week 0 (baseline), week 14, and week 24 after initiation of therapy. At the start of the IMPACT 2 trial, 200 patients were included at baseline with a 1:1 ratio in the placebo arm and the infliximab treatment arm. Eighteen patients discontinued treatment during the treatment period, and thus only 182 patients completed all 3 visits, including the corresponding joint assessments.

Joint counts.

The following compositions of the joint count schemes were taken into account. First, the JC66/68, which is considered the gold standard and is widely used in clinical trials to evaluate 66 swollen joints and 68 tender joints. The JC66/68 comprises the following joints: 2 temporomandibular, 2 sternoclavicular, 2 acromioclavicular, 2 shoulder, 2 elbow, 2 wrist, 8 DIP (fingers only), 20 PIP (fingers and toes), 10 MCP, 10 metatarsophalangeal, 2 knee, 2 tarsal, and 2 ankle joints (n = 66 joints). For tenderness only, the 2 hip joints are assessed additionally (n = 68 joints). The simplified joint counts, including 28, 32, or 36 joints, have been derived from the JC66/68. Second, the 36–joint count (JC36), which is a simplified joint count including DIP joints and is composed of 8 DIP, 10 PIP, 10 MCP, 2 wrist, 2 elbow, 2 knee, and 2 ankle joints. Third, the 32–joint count (JC32), which is a simplified joint count including DIP joints and is limited to hand, finger, and knee joints and includes 8 DIP, 10 PIP, 10 MCP, 2 wrist, and 2 knee joints. Fourth, the standard JC28 without the DIP joints used in RA, which includes 10 PIP, 10 MCP, 2 wrist, 2 elbow, 2 shoulder, and 2 knee joints (Figure 1).

thumbnail image

Figure 1. Illustration of the 36–joint count (JC36), 32–joint count (JC32), and 28–joint count (JC28).

Download figure to PowerPoint

Statistical analysis.

SJCs and TJCs based on the JC66/68, JC36, JC32, and JC28 schemes were compared by Pearson's correlation analysis at each individual time point (baseline, week 14, and week 24). Because the ACR 20% improvement (ACR20) response criterion was the primary end point of the IMPACT 2 trial, the total number of responders and nonresponders was calculated based on the various joint counts in order to compare changes from baseline measured by using the JC66/68, JC36, JC32, and JC28. To focus on the reliability of simplified joint assessments in comparison with the JC66/68, the outcome after initiation of infliximab therapy in view of the ACR20 response for each of the joint counts was calculated and compared with each other by receiver operating characteristic (ROC) technique. Therefore, in this context reliability was determined as the extent to which simplified joint counts were able to correctly predict response (sensitivity) or nonresponse (specificity) with regard to the underlying classification of the gold standard, the JC66/68. For numeric comparison of ROC curves, the cumulative predicted probabilities, i.e., the amount of area under the curve (AUC), was calculated. In a subsequent step, Youden indices (i.e., sensitivity + specificity − 1) were taken into account to determine which simplified joint count offered the best tradeoff between specificity and sensitivity. Finally, we looked at the number of patients that would have been excluded from the IMPACT 2 trial if simplified joint counts had been used for evaluating patient eligibility for the trial instead of the JC66/68. Data analysis was done using SAS software, version 9.1.3 (SAS Institute).

RESULTS

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

Pearson's correlation analysis of TJC and SJC.

With regard to tender joints, bivariate correlation coefficients between JC66/68 and the simplified joint counts consistently showed highly significant results (P < 0.0001) for baseline, week 14, and week 24 assessments, with the highest average correlation (r = 0.93) between the 68-TJC and the 36-TJC for all 3 assessments, followed closely by the results of the 32-TJC and the 28-TJC (for each, r = 0.92). Correlation between total and simplified TJCs slightly improved over time (Table 1). This tendency could not be observed for the correlations between total and simplified SJCs. For swelling, all correlations between total SJC and simplified SJCs were also significant (P < 0.0001), with the highest average correlation (r = 0.93) again between the 66-SJC and the 36-SJC. In all, these findings were as we had expected.

Table 1. Pearson's correlation analysis and AUC characteristics of simplified joint counts*
 36–joint count32–joint count28–joint count
  • *

    Values are the r (number of patients) unless otherwise stated. AUC = area under the curve; TJC = tender joint count; SJC = swollen joint count.

  • P < 0.0001 for all correlation coefficients.

  • Calculated via Fisher's z-transformation (for unequal sample sizes).

Pearson's correlation between 68-TJC and simplified TJCs   
 Baseline0.91 (200)0.88 (199)0.88 (200)
 Week 140.95 (195)0.93 (194)0.93 (195)
 Week 240.95 (182)0.94 (182)0.94 (182)
 Average correlation0.930.920.92
 Correlation of the percentage of improvement from baseline   
  At week 140.83 (195)0.80 (195)0.84 (195)
  At week 240.88 (182)0.68 (182)0.83 (182)
  Average correlation0.860.750.84
Correlation between 66-SJC and simplified SJCs   
 Baseline0.89 (200)0.87 (199)0.81 (200)
 Week 140.95 (195)0.93 (194)0.92 (195)
 Week 240.94 (182)0.93 (182)0.90 (182)
 Average correlation0.930.920.89
 Correlation of the percentage of improvement from baseline   
  At week 140.80 (195)0.75 (195)0.73 (195)
  At week 240.81 (182)0.78 (182)0.75 (182)
  Average correlation0.810.770.74
AUC   
 Week 140.950.940.94
 Week 240.940.920.94

Pearson's correlation analysis of percentage of improvement from baseline.

The percentage of improvement from baseline was defined as concordance between the relative improvement of joint status in total and simplified joint counts at week 14 and week 24 in comparison with baseline. To address this issue, we computed Pearson's correlation coefficients for the percentage of improvement of TJCs and SJCs at week 14 as well as at week 24. All correlations were highly significant (P < 0.0001), with coefficients ranging from r = 0.68 to r = 0.88 for tenderness and from r = 0.73 to r = 0.81 for swelling. Again, the average correlation between change in the JC66/68 and change in the JC36 (r = 0.86 for tenderness and r = 0.81 for swelling) was slightly higher than the coefficients obtained for changes in the JC66/68 versus the JC32 and for the JC66/68 versus the JC28 (Table 1).

ROC curve analysis and logistic regression for the ACR20 response criterion.

Based on these results, we went another step further and conducted an ROC curve calculation followed by a logistic regressions analysis to confirm the expectations derived from the correlation analysis, i.e., the JC36 being the best way of simplifying the JC66/68. As shown in Figure 2, all 3 simplified joint counts showed similar ROC curves, whereas the JC32 seemed to have minor disadvantages in identifying patients as truly positive and truly negative in fulfilling the ACR20 response criterion with respect to the JC66/68. A closer look at the results of the corresponding logistic regression models, with the ACR20 response being the dependent variable and each simplified joint count being the independent variable, allowed a more differentiated view on this topic, highlighting that the JC36 showed the best predictive results when considering the AUC (AUCweek14 = 0.95, AUCweek24 = 0.94), whereas differences between simplified joint counts in general were rather small. Surprisingly, the JC28 (which did not include the DIP joints) showed results (AUCweek14 = 0.94, AUCweek24 = 0.94) that were nearly as good as those obtained by using the JC36, whereas the performance of the JC32 (AUCweek14 = 0.94, AUCweek24 = 0.92) was slightly lower. The results of the ROC curve analysis seemed sensible when investigating the number and percentages of patients correctly classified as true ACR20 responders and true ACR20 nonresponders according to their involvement of joints by the JC36 and JC28 (Table 2). Finally, we determined the average maximum of Youden indices for all simplified joint counts and found results that slightly argued for the JC28 (meanYouden = 0.83) in direct comparison with the JC36 (meanYouden = 0.81), being another distinctive mark of a high sensitivity and specificity, especially of these 2 joint counts.

thumbnail image

Figure 2. A, Sensitivity versus (1 − specificity) for numeric American College of Rheumatology (ACR) variables, based on partial joint counts at week 14 of the Infliximab Multinational Psoriatic Arthritis Controlled Trial 2 (IMPACT 2). B, Sensitivity versus (1 − specificity) for numeric ACR variables, based on partial joint counts at week 24 of the IMPACT 2 trial. acrn36 = the numeric variable of the receiver operating characteristic curve of the 36–joint count.

Download figure to PowerPoint

Table 2. ACR20 response based on total versus simplified joint counts with calculation of corresponding sensitivity, specificity, and Youden index*
ACR20 responses, listed by total joint count responses at each weekSimplified joint counts
JC36JC32JC28
  • *

    Values are the number (percentage) unless otherwise stated. ACR20 = American College of Rheumatology 20% improvement; JC36 = 36–joint count; JC32 = 32–joint count; JC28 = 28–joint count.

Week 14   
 No, 125 patients   
  No, true-negative118 (94.40)117 (93.60)117 (93.60)
  Yes, false-positive7 (5.60)8 (6.40)8 (6.40)
 Yes, 70 patients   
  No, false-negative9 (12.86)12 (17.14)9 (12.86)
  Yes, true-positive61 (87.14)58 (82.86)61 (87.14)
 Maximum Youden index0.810.760.83
Week 24   
 No, 86 patients   
  No, true-negative76 (88.37)76 (88.37)79 (91.86)
  Yes, false-positive10 (11.63)10 (11.63)7 (8.14)
 Yes, 96 patients   
  No, false-negative8 (8.33)11 (11.46)10 (10.42)
  Yes, true-positive88 (91.67)85 (88.54)86 (89.58)
 Maximum Youden index0.810.780.82
Average sensitivity (meansens)0.890.860.88
Average specificity (meanspec)0.910.910.93
Average maximum of Youden index (meanYouden)0.810.770.83

Redistribution of patients according to use of simplified joint counts.

As can be seen in Table 3, a maximum of 13 (6.5%) of 200 patients would have been excluded from study participation due to not fulfilling the inclusion criteria with respect to joint involvement by using one of the simplified joint counts. Thus, the sensitivity of these joint count methods ranges from 0.96 for the JC36 to 0.94 for both the JC32 and the JC28. Even at the end point of the trial at week 24, the number of false-negative patients, meaning the number of those patients who had a total joint count of more than zero but a simplified joint count of zero (or missing results), was low (e.g., 17 patients with a JC28 [9.3%]) (Table 4).

Table 3. Patients with a partial joint count <2 or <1 at baseline (n = 200)*
Number of joints involvedJC36JC32JC28
Both SJC and TJCEither SJC or TJCBoth SJC and TJCEither SJC or TJCBoth SJC and TJCEither SJC or TJC
  • *

    Values are the number (percentage) of patients. JC36 = 36–joint count; JC32 = 32–joint count; JC28 = 28–joint count; SJC = swollen joint count; TJC = tender joint count.

  • Both SJC and TJC <2 or <1.

  • Any of SJC and TJC <2 or <1.

<21 (0.50)8 (4.00)1 (0.50)13 (6.50)2 (1.00)13 (6.50)
<10 (0.00)2 (1.00)0 (0.00)3 (1.50)0 (0.00)3 (1.50)
Table 4. Patients who had a total joint count greater than zero but a partial joint count equal to zero or missing at week 14 or week 24*
Visit, jointJC66/68JC36JC32JC28
  • *

    JC66/68 = 66/68–joint count. See Table 3 for additional definitions.

Week 14 (n = 195)    
 SJC, no. (%)173 (88.72)4 (2.05)8 (4.10)7 (3.59)
 Total SJC, mean ± SD12.84 ± 12.353.00 ± 2.832.63 ± 2.072.43 ± 2.15
 TJC, no. (%)178 (91.28)7 (3.59)11 (5.64)11 (5.64)
 Total TJC, mean ± SD22.47 ± 17.394.29 ± 2.984.82 ± 3.124.27 ± 2.61
Week 24 (n = 182)    
 SJC, no. (%)146 (80.22)11 (6.04)16 (8.79)17 (9.34)
 Total SJC, mean ± SD8.53 ± 9.211.64 ± 1.032.63 ± 2.252.53 ± 2.27
 TJC, no. (%)153 (84.07)11 (6.04)14 (7.69)16 (8.79)
 Total TJC, mean ± SD16.01 ± 15.432.64 ± 2.663.50 ± 2.953.13 ± 2.80

DISCUSSION

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

In this study, we compared global and simplified joint counts for assessing ACR20-related response to therapy and joint involvement in polyarticular PsA. We were able to show that differences between the simplified joint counts and the JC66/68 are rather small in view of AUC-related sensitivity and specificity, whereas these 2 parameters remained stable on a high level. Additionally, all 3 simplified joint counts showed highly significant correlations with the JC66/68, and furthermore with the percentage of improvement from baseline. Surprisingly, the ACR20-related sensitivity and specificity of the JC36 and JC32 (which include the DIP joints) were not superior to those of the JC28 (which does not include the DIP joints), although Pearson's correlation coefficients between the JC66/68 and the JC36 were slightly higher compared with those between the JC32 and the JC28. Finally, even the maximum Youden indices show findings suggesting a slight superiority of the JC28. Being that these results revealed a very similar performance of all 3 simplified joint counts in general, it appears that inclusion of DIP joints in a simplified joint count is not mandatory for appropriate monitoring of joint involvement and the corresponding response to therapy in polyarticular PsA. A reason for that might be found in difficulties in assessing swollen DIP joints compared with MCP or PIP joints, leading to common false-negative evaluation. This point is supported by the data of affected joint patterns, which generally show more positive ratings of swelling for MCP and PIP joints of the fingers than for DIP joints throughout all assessments. However, because data for patient-derived tenderness show the same frequency distribution, it remains questionable whether there truly is a negative rating bias, or just less common joint involvement of the DIP joints. Either way, our results strengthen the inclusion of MCP joints and PIP joints of both hands, knees, and wrists for a reliable simplification of joint counts for polyarticular PsA.

One of the previously-mentioned concerns is that DIP joints are not part of the ACR improvement criteria in RA (25), which are in part based on the DAS28 (26). In PsA, however, DIP joints play a major role in the clinical picture of disease (5, 27), not so much with regard to disease severity but especially during the process of diagnosis, e.g., while identifying the underlying subset of PsA. Our data correspond to prior findings (5, 28) in which the involvement of the DIP joints in polyarticular PsA appeared in >40% of the patients. Being that joint patterns in polyarticular PsA and RA seem to be somewhat similar and that the DAS28 has proved to be a valid tool for the assessment of the latter (29, 30), it might be allowable to conclude that leaving out the DIP joints still leads to a therapeutically reliable assessment of joint involvement in polyarticular PsA. This is also supported by the findings of other studies that have addressed the question of identifying adequate outcome and response parameters in medical treatment of PsA (31, 32).

However, a limitation of the present study is the fact that the conclusions drawn only account for a polyarticular, more severe pattern of PsA that was monitored in a clinical trial with predefined inclusion and exclusion criteria and consisted of patients with a high burden of disease. Therefore, simplified joint counts might not be appropriate means for monitoring mono- or oligoarticular cases of PsA, which are also part of the clinical routine. Thus, a mere direct transfer of our findings to all different kinds of patterns of PsA is difficult and associated with the limitation of missing affected joints that might be important for a correct evaluation and overall monitoring of disease activity.

Nonetheless, one should keep in mind that reduced joint counts are only one part of the possible approaches to facilitating measurement of disease activity in PsA. The findings of this study show that assessment of joint involvement using a JC28 (or another simplified scoring method that includes the DIP joints) led to considerably sensitive and specific decision making with regard to therapy response, supporting the point of view that composite indices for disease activity that are based on simplified joint counts might be useful tools in the future assessment of polyarticular PsA. This point is supported by both prior findings (11) as well as by the high sensitivity that the various simplified joint counts would have had if they had been used as entry criteria instead of the JC66/68. Surprisingly, our findings are not congruent to prior calculations of the Outcome Measures in Rheumatology Clinical Trials (OMERACT) study, in which ∼20% of the patients would not have fulfilled the entry criteria of PsA trials if patient eligibility had been evaluated based on a JC28 (33). The main reason for the low number of false-negative results, which accounts for less than one-third of the number suggested by OMERACT, might be based on different patient populations or the fact that there were nearly twice as many patients in the phase 3 IMPACT 2 trial. To clarify whether a JC28 is a reliable tool for evaluating initial eligibility into clinical trials for polyarticular PsA, further comparative analysis on joint counts will be needed. Moreover, the question of to which extent simplified joint counts might also be useful in clinical practice, e.g., in a broad range of patients, including those experiencing oligoarticular PsA, will have to be discussed in the future.

Assessment of overall disease activity in PsA remains tricky, being that it has to include symptoms related to skin or nail disease, dactylitis, enthesitis, and last but not least, involvement of the spine. Thus, a disease activity score based on the JC28, similar to the DAS28 in RA, does not suffice to describe the entire clinical disease pattern of PsA nor to monitor global disease activity (24, 34). However, a sensitive and specific reduction of the number of joints might be a helpful tool on the way to pursue that objective.

AUTHOR CONTRIBUTIONS

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

All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be submitted for publication. Mr. Englbrecht had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Englbrecht, Wang, Ronneberger, Manger, Vastesaeger, Veale, Schett.

Acquisition of data. Englbrecht, Wang, Ronneberger.

Analysis and interpretation of data. Englbrecht, Wang, Ronneberger, Vastesaeger.

REFERENCES

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