Validity of the disease activity score in undifferentiated arthritis

Authors


Abstract

Objective

To study whether the Disease Activity Score (DAS) is a valid measure of disease activity in undifferentiated arthritis (UA).

Methods

Data from a randomized, double-blind, placebo-controlled trial of methotrexate (MTX) and placebo involving 110 patients with UA were used. Data included baseline and 3, 6, 9, and 12 months, as well as diagnosis at 18 months. Validity of the DAS was analyzed using factor analysis, correlations with disease activity variables, correlations with changes in disability and joint damage, differences in DAS between diagnoses, and detecting the difference between placebo and MTX.

Results

Three disease activity factors were retrieved from the disease activity variables: patient reported outcomes, tender and swollen joints, and acute phase reactants. The DAS had its highest correlations (r > 0.77) with tender joint counts, followed by swollen joint counts (r > 0.63) and patient reported outcomes (r > 0.30), but the DAS correlated less with C-reactive protein levels (r = 0.32). Over time, the DAS was related to the Health Assessment Questionnaire response with an odds ratio of 4.1 (95% confidence interval 2.1–8.0), but not with change in joint damage. At 18 months, the mean DAS was 2.6 for rheumatoid arthritis patients, 2.2 for UA patients, and 1.9 for patients in remission (P = 0.001). The DAS discriminated better than all single variables between MTX and placebo, with a Guyatt's effect size of 0.89.

Conclusion

The DAS appears to be a reasonably valid measure of disease activity for use in UA clinical trials.

INTRODUCTION

According to the American College of Rheumatology (ACR; formerly the American Rheumatism Association) criteria for rheumatoid arthritis (RA) (1, 2), the disease process develops before the diagnosis can be made. Indeed, depending on the study population, 6–55% of patients with undifferentiated arthritis (UA) may develop RA within 1 year (3). UA is defined as an inflammatory arthritis in which no definitive diagnosis can (yet) be made. UA may be successfully treated before it develops into erosive arthritis, such as RA (4). Until now, there is no available valid measure of disease activity in UA. The Disease Activity Score (DAS) is a valid measure of RA disease activity (5) and may also be useful and valid in UA clinical trials, but has not been formally tested in that setting.

The DAS was developed in RA patients, using data from the first 3 years after diagnosis according to the ACR criteria (2, 5). Although the variables included and the weights applied in the DAS were derived from early RA patients, later validation proved that the DAS is also valid in patients with longer disease duration (6). UA covers a spectrum of disease before the ACR classification criteria are fulfilled, and not all patients with UA will eventually develop RA. Moreover, in UA, generally fewer joints are involved, and the pattern of involvement may differ from RA. Therefore, extended joint counts may be more appropriate than reduced joint counts in UA.

Our objective was to study whether the DAS is a valid measure of disease activity in early UA, using the data from a randomized placebo-controlled trial. The modified DAS in 28 joints (DAS28), which includes reduced joint counts, was also studied. The alternative DAS28 modifications (Clinical Disease Activity Index or Simple Disease Activity Index) could not be calculated due to missing components.

PATIENTS AND METHODS

Trial data.

We used data from the PROMPT (Probable Rheumatoid Arthritis: Methotrexate Versus Placebo Treatment Trial), which has been described previously (4). In summary, this study was a multicenter, randomized, double-blind, placebo-controlled trial involving 110 patients with early UA. Treatment started with methotrexate (MTX; 15 mg/week) or placebo tablets, and every 3 months the dose was increased if the DAS was >2.4. After 12 months, the study medication was tapered and discontinued in patients not fulfilling the ACR criteria for RA. When a patient fulfilled the ACR criteria for RA (primary end point), the study medication was changed to MTX. The diagnosis at the end of the study and the progression of joint damage were prespecified primary outcomes of this trial to test whether MTX indeed modifies prognosis in UA (4).

Patients and assessments.

Eligible patients attended the rheumatology outpatient clinics of the participating hospitals and had symptoms of arthritis for ≤2 years, were ≥18 years of age, and were diagnosed as having UA according to the ACR 1958 criteria for probable RA (7).

At baseline and at 3, 6, 9, and 12 months, the erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) levels, the 68 tender joint count (TJC68), the Ritchie Articular Index (RAI), the 66 swollen joint count (SJC66), the disability measurement (Health Assessment Questionnaire [HAQ]) and the visual analog scales (VAS; range 0–100 mm) for assessing pain, global disease activity, severity of morning stiffness, fatigue, and general health (GH) were obtained. Every 6 months, radiographs of the hands and feet were obtained and scored according to the Sharp/van der Heijde (SHS) scoring method. At 18 months, the diagnosis was recorded by the patient's rheumatologist as RA, UA, or remission. Remission was defined as no clinical symptoms of arthritis and no use of disease-modifying antirheumatic drugs during the preceding year. The DAS was calculated according to the original formula:

equation image

where SJC is a 44 swollen joint count (SJC44) and GH is a patient-assessed VAS of 0–100 mm (5). The RAI contains 53 joints that are rated for tenderness on a scale where 0 = no tenderness, 1 = pain on pressure, 2 = pain and winced, and 3 = winced and withdrew (8). For the SJC44, joints included in the RAI are scored for swelling on a 0–1 scale, excluding the temporomandibular joints, neck, hips, and subtarsal and tarsal joints. Levels of DAS <2.4 are equated with “low” disease activity, and levels of DAS >3.7 are equated with “high” disease activity (9). Similarly, the DAS28 was calculated using the formula on the swollen 28-joint count and tender 28-joint counts (SJC28 and TJC28, respectively) (6).

Statistical analysis.

Cumulative frequency plots of the TJC68, the 53-joint count of the RAI without grading (TJC53), the TJC28, and the SJC66, SJC44, and SJC28 counts were produced to analyze the degree of misclassification when using reduced joint counts. Differences between SJC66 and SJC44, SJC44 and SJC28 at baseline, and at 12 months were tested using the signed rank test; the same was used for TJC68 and TJC53, and TJC53 and TJC28.

Analysis of the validity of the DAS was performed in 5 steps which are outlined below. For these analyses, data were used from baseline and at 3, 6, 9, and 12 months, and at the final diagnosis obtained at 18 months. For comparison, the DAS28 was analyzed in the same way.

Factor analysis.

The DAS had been devised to draw information from 3 underlying factors of disease activity in RA: physical examination, laboratory values, and patient reported outcomes (5). To reveal the factors underlying the assessment of disease activity in UA, factor analysis with varimax rotation was used with a minimum eigenvalue >1 to obtain factors using baseline data.

Concurrent validity.

To inform how strong changes in the DAS within a patient are correlated with changes in other variables assessing disease activity, within-subject correlations were calculated using linear regression with correction for repeated observations, using the 5 visits of the first 12 months (10).

Construct validity.

It was studied whether the DAS was related to changes in disability and to progression of joint damage over the first 12 months. Ordinal logistic regression was used to relate the time-averaged DAS (0–12 months) to the HAQ response at 12 months. The minimal clinically important difference of 0.22 in the HAQ score was used to classify patients as “worsened,” “unchanged,” or “improved” (11). The proportional odds assumption was checked by dichotomizing the HAQ response into worsened versus unchanged or improved, and worsened or unchanged versus improved, and by applying 2 separate logistic regression models. It was checked graphically as to whether the logits of the dichotomized HAQ responses were linearly related to the time-averaged DAS (12). The Hosmer-Lemeshow test was used to indicate goodness-of-fit for both logistic regression models (P > 0.05), and a chi-square test score was used to test the proportional odds assumption in the ordinal logistic regression model (P > 0.05) (12). Time-averaged DAS, time-averaged CRP levels, and the change in modified SHS over 0–12 months were analyzed using logistic regression with progression >3 points (yes/no) in modified SHS as a dependent variable, as well as using linear regression with change in modified SHS as a dependent variable.

Criterion validity.

Using analysis of variance and ordinal logistic regression, it was studied whether the mean DAS over the 5 visits in the first 12 months was different for patients who had erosive arthritis, UA, or who were in remission at visit 6 (18 months). The underlying assumption is that disease activity and mean DAS are highest in patients developing erosive arthritis and lowest in patients entering remission. The proportional odds assumptions for the ordinal logistic regression were tested in a similar way as described above for HAQ responses.

Responsiveness.

Responsiveness, or sensitivity to change, was calculated over the visits at baseline and 12 months, using the standardized response mean (SRM; mean change/SD of change), and Guyatt's responsiveness statistic (mean changeactive treated group/SDplacebo group). Moreover, the t value as a measure of discrimination between the placebo and MTX trial arms was calculated.

RESULTS

The values of the disease activity parameters at study baseline are shown in Table 1. The majority of the patients were women with an average age of ∼50 years; 25% (27 of 110) were positive for anti–cyclic citrullinated peptide (anti-CCP) antibodies, and 39% (39 of 110) were positive for rheumatoid factor (RF). According to the distributions of the disease activity parameters in Table 1, the majority of patients were in moderate-to-low states of disease activity. At baseline, 36% (n = 20) of the patients in the MTX group and 45% (n = 25) of the patients in the placebo group had a DAS <2.4. There were no statistically significant differences between the trial arms at baseline in any of the parameters. The DAS and the DAS28 had a Gaussian distribution (data not shown).

Table 1. Disease activity parameters at baseline*
 MTX group (n = 55)Placebo group (n = 55)
  • *

    Values are the median (interquartile range) unless otherwise indicated. MTX = methotrexate; RF = rheumatoid factor; anti-CCP = anti–cyclic citrullinated peptide; ESR = erythrocyte sedimentation rate; CRP = C-reactive protein; SJC28 = swollen joint count in 28 joints; TJC28 = tender joint count in 28 joints; RAI = Ritchie Articular Index; VAS = visual analog scale; HAQ = Health Assessment Questionnaire (disability index); DAS = Disease Activity Score; DAS28 = Disease Activity Score in 28 joints.

Women, no. (%)35 (64)38 (69)
Age, mean ± SD years53 ± 1450 ± 14
Symptom duration at first visit, days312 (195–507)263 (169–432)
RF positive, no. (%)20 (36)19 (35)
Anti-CCP positive, no. (%)12 (22)15 (27)
ESR, mm/hour12 (5–24)11 (5–25)
CRP, mg/dl5 (3–11)5 (3–9)
SJC282 (1–5)2 (1–3)
SJC443 (2–5)2 (2–4)
SJC663 (2–5)3 (2–5)
TJC286 (2–9)4 (2–7)
TJC5311 (4–16)7 (4–14)
RAI, range 0–787 (3–10)5 (2–8)
TJC6811 (4–17)7 (4–14)
Pain, VAS44 (18–61)48 (28–54)
Disease activity, VAS49 (15–60)52 (29–64)
Morning stiffness severity, VAS45 (20–61)47 (23–62)
Fatigue, VAS55 (22–76)46 (9–72)
General health, VAS40 (19–52)36 (16–55)
HAQ, range 0–30.75 (0.38–1.13)0.75 (0.25–1.13)
DAS, mean ± SD2.72 ± 0.782.52 ± 0.76
DAS28, mean ± SD3.94 ± 1.043.72 ± 1.06

Cumulative frequency plots.

In Figure 1A, a cumulative frequency plot of the SJC28, SJC44, and SJC66 is shown at baseline. Every individual patient is represented in the curves as a single symbol with patients being ranked by the number of joints involved. The different types of symbols denote the different joint counts and the patients do not necessarily have the same rank in the 3 curves. All differences between reduced and extended joint counts were statistically significant (P ≤ 0.002) for the tender joint counts (Figure 1C), as well as for the swollen joint counts (Figure 1A) at baseline and at 12 months (not shown).

Figure 1.

Cumulative frequency plots of joint counts at baseline: average differences (A and B)and individual differences (C and D). Every patient is represented in the plots as a single symbol with patients being ranked by the number of joints involved. In panel A and B, the patients do not necessarily have the same rank in the 3 curves. The position of the median scores is where the cumulative frequency is 50%. In panel C and D, every individual patient is represented in the plots as a single symbol, patients being ranked by their scores on the extended joint count. Only the median of the extended joint count is at a frequency of 50%. SJC = swollen joint count; TJC = tender joint count.

It can be seen in Figure 1A that the SJC66 detects most swollen joints, with the SJC44 being similar. The SJC28 scores are somewhat lower than the other scores, and the SJC28 therefore slightly underestimates the number of swollen joints. The number of patients with no swollen joints (according to the SJC28) and those with ≥1 swollen joint (according to the extended joint counts) does not exceed 10% at baseline and at 12 months (not shown). Figure 1C similarly shows the cumulative frequency plots of the TJC28, TJC53, and TJC68 at baseline. It can be seen that the number of tender joints involved is larger than the number of swollen joints. In Figure 1C, the scores of the TJC follow the same trend, but the TJC28 is the lowest. Misclassification (no tender joints according to the TJC28, but at least 1 tender joint according to the extended joint counts) occurred in 6% at baseline (Figure 1C) and 11% at 12 months.

Cumulative frequency plots of the SJC28 and SJC44 (Figure 1B) and of the TJC28 and TJC53 are shown (Figure 1D). Now the patients are ordered by their scores on the extended joint count only. It can be seen in Figure 1B that patients with no or few swollen joints according to the SJC28 may have ≥1 swollen joint according to the SJC44. The same is seen for the TJC28 and TJC53 in Figure 1D.

Analysis of factors.

Three factors were retrieved by factor analysis from the disease activity parameters at baseline (Table 2). The first factor consisted of all the items on patient reported outcomes, the second factor consisted of the examinations of tender and swollen joints, and the third included the acute phase reactants ESR and CRP levels. The HAQ loaded on both the first and second factors. The ordering of factors in factor analysis and the amount of variance explained also reflect the number of items included.

Table 2. Exploratory factor analysis at baseline*
 Factor
Patient-reported outcomesJoint countsAcute phase reactants
  • *

    Factors derived using exploratory factor analysis with varimax rotation and a minimum eigenvalue of >1 to obtain factors using baseline data. See Table 1 for abbreviations.

  • Indicates the highest loading of a variable to a factor.

ESR, mm/hour−0.050.020.88
CRP, mg/dl−0.02−0.090.88
SJC44−0.150.60−0.03
TJC530.180.93−0.06
RAI, range 0–780.280.85−0.03
Pain, VAS0.860.060.10
Disease activity, VAS0.82−0.080.08
Morning stiffness, VAS0.650.29−0.14
Fatigue, VAS0.720.29−0.12
General health, VAS0.85−0.01−0.13
HAQ, range 0–30.530.570.06
Variance explained, %463222

Within-subject correlations.

In Table 3, the within-subject correlations of the DAS with other disease activity variables are shown, using the visits of the first 12 months. The DAS had its highest correlations with the RAI and the tender joint counts; it correlated well with the swollen joint counts and most of the patient assessments and correlated less with CRP. Correlations of the DAS28 generally were similar.

Table 3. Within-subject correlations with the DAS and DAS28 over visits 1–5*
VariablesDASDAS28
r95% CIr95% CI
  • *

    Within-subject correlations inform how far a change in DAS (or DAS28) is paralleled by a change in the variables. 95% CI = 95% confidence interval. See Table 1 for additional abbreviations.

ESR, mm/hour0.470.31–0.610.500.34–0.63
CRP, mg/dl0.320.14–0.480.360.19–0.52
SJC280.620.49–0.720.630.50–0.73
SJC440.630.50–0.730.520.37–0.65
SJC660.620.49–0.720.500.35–0.63
TJC280.620.49–0.720.690.58–0.78
TJC530.710.60–0.790.630.51–0.73
RAI, range 0–780.770.68–0.840.600.47–0.71
TJC680.640.51–0.740.570.43–0.69
Pain, VAS0.430.26–0.570.490.33–0.62
Disease activity, VAS0.430.27–0.570.480.32–0.61
Morning stiffness, VAS0.390.22–0.540.420.25–0.56
Fatigue, VAS0.300.12–0.460.350.17–0.50
General health, VAS0.460.30–0.600.540.39–0.66
HAQ, range 0–30.350.17–0.500.350.21–0.53

Changes in disability and progression of joint damage.

It was studied whether the time-averaged DAS over 12 months was associated with a change in HAQ over the same period of time, irrespective of treatment allocation (Figure 2). HAQ scores were available for 102 patients. Over 12 months, there were 16 patients with a worsening in the HAQ score by >0.22 points, 40 patients with an improvement in the HAQ score of >0.22 points, and 46 patients with smaller changes in either direction who were regarded as “unchanged” (Figure 2). In ordinal logistic regression, the time-averaged DAS was significantly associated with a minimal clinically important change (>0.22 points) in HAQ, with an odds ratio (OR) of 4.1 (95% confidence interval [95% CI] 2.1–8.0), corrected for baseline HAQ, anti-CCP, age, and sex as confounders. The assumptions for ordinal logistic regression, as outlined in the Methods section, were met. Furthermore, it was studied whether time-averaged disease activity and joint damage progression were correlated. However, there was no relation between time-averaged CRP or time-averaged DAS and change in SHS, irrespective of RF positivity, anti-CCP positivity, treatment, or final disease outcome (results not shown).

Figure 2.

Response in HAQ and DAS scores averaged over time. Box plots of the patient-averaged DAS over months 0–12, by HAQ response at month 12. HAQ response defined as worsening (change in HAQ score by >−0.22), no change (change in HAQ score between −0.22 and 0.22), and improvement (change in HAQ score >0.22) according to Wells et al (11). HAQ = Health Assessment Questionnaire; DAS = Disease Activity Score.

Differences between criteria.

The level of disease activity is supposedly different for patients who develop RA, have UA, or reach remission. The mean ± SD DAS for patients with RA (n = 49) was 2.6 ± 0.90, for patients with UA (n = 21) the mean ± SD DAS was 2.2 ± 0.72, and for patients reaching remission (n = 29) the mean ± SD DAS was 1.9 ± 0.64 (P = 0.001 for the differences between diagnostic groups). The OR for the differences between diagnostic groups in ordinal logistic regression was 2.5 (95% CI 1.5–4.2) per time-averaged DAS point. A similar difference was observed for the DAS28. The assumptions for ordinal logistic regression were met for both analyses.

The mean ± SD DAS28 for patients with RA was 3.7 ± 1.09, for patients with UA the mean ± SD DAS28 was 3.2 ± 1.91, and for patients reaching remission the mean ± SD DAS28 over time was 2.8 ± 0.79 (P = 0.0007 for the differences between diagnostic groups). At 18 months, 13 patients fulfilled the ACR criteria for remission. At that time, there were 32 patients with a DAS <1.6, and 13 (40%) of those patients fulfilled the ACR criteria for remission. The number of patients with a DAS28 <2.6 was 37, with 13 (35%) of those patients fulfilling the ACR criteria.

Disease activity.

The changes in disease activity parameters from baseline to 12 months are shown in Table 4. Disease activity improved in both treatment arms, but changes were generally larger in the MTX group. When adopting P = 0.05 as the critical value for determining statistical significance, the critical value of t with this sample size is 2.0. In Table 4 it can be seen that in the MTX group the DAS had a higher responsiveness (SRM) than the single parameters. The tender joint counts and the DAS and DAS28 showed the best performance in discriminating MTX from placebo according to Guyatt's responsiveness statistic and the t value, with increasing discrimination as the number of joints assessed increased. The other variables did not exceed the critical t value of 2.0.

Table 4. Responsiveness of disease activity parameters*
 MTX group (n = 55)Placebo group (n = 55)
Mean ± SDSRMMean ± SDSRMGuyatt's ESt value
  • *

    Changes over 0–12 months are shown. SRM = standardized response mean; ES = effect size. See Table 1 for additional abbreviations.

DAS−0.8 ± 0.81.02−0.4 ± 0.90.400.892.40
DAS28−1.1 ± 1.11.05−0.6 ± 1.30.470.852.15
ESR, mm/hour−5.6 ± 11.80.47−1.8 ± 15.20.120.371.40
CRP, mg/dl−3.6 ± 15.00.24−0.7 ± 17.00.040.210.87
SJC28−1.7 ± 2.60.62−1.2 ± 2.90.420.570.77
SJC44−1.9 ± 3.60.53−1.0 ± 4.70.200.551.17
SJC66−2.3 ± 4.50.52−1.4 ± 4.70.290.551.08
TJC28−2.7 ± 3.90.68−0.5 ± 5.20.100.522.44
TJC53−3.8 ± 6.40.59−0.31 ± 6.80.040.572.71
RAI, range 0–78−3.3 ± 5.00.65−0.7 ± 4.40.160.742.77
TJC68−4.8 ± 7.80.620.17 ± 9.00.020.543.04
Pain, VAS−11 ± 250.45−15 ± 280.520.410.59
Disease activity, VAS−10 ± 240.41−17 ± 280.590.341.39
Morning stiffness, VAS−13 ± 250.53−15 ± 280.540.460.49
Fatigue, VAS−7 ± 250.27−7 ± 300.240.220.10
General health, VAS−9 ± 210.44−6 ± 310.200.290.54
HAQ, range 0–3−0.19 ± 0.450.42−0.13 ± 0.470.290.400.59

DISCUSSION

In this post hoc analysis of the data from the PROMPT study of MTX versus placebo in patients with early UA, it was questioned whether the DAS is a valid measure of disease activity in early UA. According to the results of this study, the DAS indeed appears to be a reasonably valid measure of disease activity for use in UA clinical trials. However, the validity of the DAS to measure arthritic disease activity in UA should also be tested in other UA samples in different settings.

In the absence of an unambiguous “gold standard” measure, validation took 5 steps: factor analysis, concurrent validity, construct validity, criterion validity, and responsiveness. The primary interest of this study in UA was in the DAS, because it includes extended joint counts. The DAS28 with its reduced (28) joint counts was included in order to determine whether or not there were large differences between the 2 measures in their performance.

According to the comparison of reduced with extended joint counts in cumulative frequency plots, it appeared that up to 10% of patients were misclassified as having no swollen or tender joints by the 28-joint count. In contrast, these patients had 1–10 joints involved according to the extended joint counts. The largest difference between 28-joint and extended joint counts is that the former does not include the joints in the feet, which may typically be involved in early RA (13). Since the DAS includes extended joint counts, it may be preferred over the DAS28 in clinical trials in early RA or UA. It could be shown in the frequency plots that the reduced joints are underestimating the involvement of the number of joints. However, the lesser performance of the reduced joint counts was not substantiated in the correlation of the DAS28 with other variables and its responsiveness.

By the factor analysis, similar factors of disease activity were found in this study as previously described in RA (5). The 3 factors were patient reported outcomes, tender joints and swollen joints, and acute phase reactants. Indeed, as seen in its formula, the DAS taps information from each of these constructs. Results of factor analysis depend on the variables that are included. Contributing to the retrieval of patient-reported outcomes as the first factor is the fact that a large proportion of the disease activity parameters, which are naturally closely correlated, were patient-assessed.

In this study, the DAS correlated as moderate-to-good with all other disease activity parameters, including patient assessments, except for the acute phase reactants. The contribution of the GH item to the DAS is relatively low (maximally 0.72 if GH is scored on a scale of 0–100 mm), which may be criticized as neglecting the patient perspective. However, despite the weighting, the DAS is reasonably well-associated with the other frequently used patient assessments. The low within-subject correlations of the DAS with ESR and CRP levels will be caused by the large proportion of patients who had acute phase reactants in the normal range, so changes in the acute phase reactants were not highly related to changes in the DAS or the DAS28. On the other hand, it is known that the DAS28 is sensitive for changes in the low (even normal) range of ESR levels (14). Over time the DAS was indeed related to a change in HAQ score, as expected in advance. However, both DAS over time and CRP levels over time were not associated with the progression of joint damage in this study. This could not be explained, not even approximately, by differences in RF positivity, anti-CCP positivity, treatment, or final disease outcome. However, the small number of patients experiencing joint damage progression made it difficult to analyze these relations.

It was supposed in advance that the course of disease activity would be different for patients who turned out to have RA, who still had UA, or who were in remission at the end of the study. Indeed, the mean DAS levels were clearly different for these 3 diagnostic groups. Both the DAS and the DAS28 did not perform well in predicting the ACR criteria for clinical remission; all patients with ACR clinical remission had a DAS <1.6 and a DAS28 <2.6, but only 40% of patients with a DAS <1.6 and 35% of patients with a DAS28 <2.6 fulfilled the ACR remission criteria.

The original PROMPT trial was powered for the primary outcome, which was the diagnosis reached at 18 months (4). When analyzing the responsiveness of the several disease activity parameters included as secondary outcomes, the DAS discriminated well between MTX and placebo. Neither of the other parameters, except the tender joint counts, reached statistical significance. Therefore, as in RA clinical trials, the pooling of information in the DAS may be advantageous for trials in UA.

An advantage of the DAS for assessment of disease activity is that it includes extended joint counts. Although the 28-joint counts were introduced to facilitate a quick performance of joint counts, their use may lead to an underestimation of disease activity if only a few joints are involved that are not among the 28 counted, such as the feet. This may play a role in UA, as well as in early RA. In UA, disease activity is generally low, and by using extended joint counts, including the feet, an underestimation of the number of joints involved may be prevented. For that reason, for clinical trials of early interventions in UA or RA that aim at clinical remission, the DAS may be preferred over the DAS28 (13). On the other hand, in this validation study in UA, the DAS28 performed closely to the DAS regarding all parameters of validity and responsiveness. It was the assessment of clinical remission that was impeded by the use of reduced joint counts.

In summary, it can be concluded that the DAS, which is a valid measure of RA disease activity, appears to be a reasonably valid measure of disease activity for use in UA clinical trials. Similar factors of disease activity were found in this study as have been found previously in RA: the DAS correlated as moderate-to-good with all other disease activity parameters including patient assessments; the DAS over time was related to changes in HAQ scores; the DAS was different for patients with RA, UA, and patients in remission; and the DAS discriminated well between MTX and placebo. It may be regarded that the PROMPT trial we used was relatively small (n = 110) for validation purposes and, therefore, studying the validity of the DAS in UA in other samples is most welcome. The DAS may be used as a primary or secondary outcome measure for clinical trials in early UA. Instead of the DAS, the DAS28 may also be used, but the use of reduced joint counts underestimates the involvement of inflamed joints.

AUTHOR CONTRIBUTIONS

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. Dr. van der Heijde 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. Fransen, Huizinga, van Riel, van der Heijde.

Acquisition of data. Visser, Dongen.

Analysis and interpretation of data. Fransen, Visser, Dongen, van der Heijde.

Ancillary