Disease activity measurement of rheumatoid arthritis: Comparison of the simplified disease activity index (SDAI) and the disease activity score including 28 joints (DAS28) in daily routine

Authors


Abstract

Objective

To assess the reliability and congruency of the Simplified Disease Activity Index (SDAI) compared with the Disease Activity Score including 28 joints (DAS28) in daily practice.

Methods

In 399 consecutive rheumatoid arthritis patients (307 women, 92 men), the SDAI and the DAS28 were calculated. Additionally, 115 of them were observed for 1 year and changes of both values were recorded. Joint assessments were performed by 4 experienced physicians. DAS28 and SDAI values and the respective changes were compared by correlation and regression analyses. Reliability assessment and factor analyses were performed. Disease activity categorizing was compared by the Wilcoxon's rank sum test.

Results

The median ± SD scores were 3.42 ± 1.45 for the DAS28 and 11.50 ± 11.50 for the SDAI. Spearman's rho was 0.897 (P < 0.0001). Score changes were also significantly correlated. Reliability testing and factor analysis revealed that both scores can be regarded as monocomponent. Categorizing patients according to the European League Against Rheumatism response criteria (EULARC) or the SDAI revealed statistically significant differences between the 2 scales (P < 0.0001).

Conclusion

SDAI values are considerably shifted to the left compared with DAS28 levels. Internal consistency and reliability of both scores are comparable. For the differences in disease activity categorizing due to the SDAI compared with the EULARC, a major limitation of the application of this newly developed disease activity score is given, unless these incongruencies can be cleared.

INTRODUCTION

Giving the physician enough information for treatment decisions is the primary requirement, but also easy applicability and time-sparing documentation are prerequisites, of disease activity monitoring in rheumatoid arthritis (RA) patients in daily routine. In clinical trials, the American College of Rheumatology (ACR) response criteria or the Disease Activity Score including 28 joints (DAS28), and consequently the European League Against Rheumatism response criteria (EULARC), are most frequently applied to measure disease activity and therapeutic efficacy (1–3).

There is consensus that the ACR response criteria do not meet the needs of daily rheumatologic practice (4). The original DAS28 consists of 4 items, namely the number of swollen and tender joints (SJC and TJC) of a 28-joint count, the visual analog scale (VAS) of patient's general health (VAS-GH scored 0–100), and the erythrocyte sedimentation rate (ESR in the first hour). The value is the result of a complicated calculation (DAS28 = 0.56 × √[t28] + 0.28 × √[sw28] + 0.70 × Ln [ESR] + 0.014 × GH). The Simplified Disease Activity Index (SDAI), however, is derived by adding the following parameters: SJC and TJC of a 28-joint count, the C-reactive protein (CRP in mg/dl), the VAS-GH (scored 0–10), and the VAS of physician's health assessment (VAS-PH scored 0–10) (5). Both scores result in absolute numbers reflecting disease activity; however, by the way of calculation, different weightings of the single items are to be expected.

All parameters included in the SDAI and the DAS28 are of proven sensitivity to change and, moreover, the SDAI's sensitivity to change was demonstrated in the leflunomide study patient population (5). The SDAI and DAS28 values for the single patients were reported to be highly correlated and, therefore, the SDAI (because of its easy applicability) was proposed to provide advantages in daily practice compared with the DAS28 (5).

The DAS28 is commonly acknowledged as a tool for disease activity monitoring, although for its application an electronic device or computed assistance is necessary. The SDAI, although claiming an identical position, has to prove its accuracy, applicability, and comparability with the DAS28 in daily practice.

Because the patient population examined in developing the SDAI constituted a highly selected one (5), it was our goal to assess the correlation of both scores in unselected RA patients under various treatments and with unselected comorbidity, possibly influencing some components of the scores (e.g., VAS-GH). Moreover, the degree of congruence between the SDAI disease activity categories and the EULARC was of interest.

PATIENTS AND METHODS

The patient cohort investigated comprised 399 consecutive (December 2002 to April 2003) RA patients (307 women, 92 men; mean age 60.6 years, range 16–85 years), according to ACR (formerly American Rheumatism Association) criteria (6). Mean disease duration was 93.6 months (range 6–750 months) and mean functional stage was 1.84 (range 1–4) (7). SDAI and DAS28 values of these patients were analyzed cross sectionally.

Most of the patients (n = 320; 80%) were taking disease-modifying antirheumatic drugs (DMARDs): 168 patients were taking methotrexate (mean 15.1 mg/week, range 7.5–25); 80 chloroquine; 65 sulfasalazine; 26 leflunomide; 2 OM-389; and 29 tumor necrosis factor α blockers. Seventy-six patients received DMARD combination therapy. All patients were treated with nonsteroidal antiinflammatory drugs on an on-demand basis and 276 patients (69%) were taking low-dose corticosteroids (mean 4.1 mg/day prednisolone, range 2.5–25).

For a 1-year period, 115 of these patients (95 women, 20 men; mean age 60.6 years, range 17–86 years) were observed longitudinally with 4 simultaneous assessments of the DAS28 and the SDAI every 3 months, on average. In this cohort, the patients' individual therapies were adapted as deemed necessary with respect to the course of their disease.

Because the single items of both disease activity indices (SJC, TJC, ESR, CRP, VAS-GH, and VAS-PH) are part of the routine monitoring program for RA patients, both scores were calculated as described (1, 5). For the calculation of the DAS28, an electronic device was used. Joint assessments were performed by 4 experienced physicians. Consensus meetings concerning joint assessment were performed at regular intervals to avoid high interrater variabilities among the physicians.

Although a DAS28 including CRP (mg/liter) instead of the ESR had been available, it was decided to stick with the original score mainly for 2 reasons (8, 9). First, the original DAS28 was applied in the leflunomide studies and therefore for the initial comparison with the SDAI (5); second, this score is far more commonly used in daily routine because most of the devices provide only the original scale.

Besides the comparison of the entire patient population, patient categories according to their disease activity were established. With respect to the EULARC (3) and the SDAI disease activity categories, respectively, patients were divided into those at low (DAS28 <3.2, SDAI <20), moderate (DAS28 3.2–5.1, SDAI 20–40), and high disease activity (DAS28 >5.1, SDAI >40) (5).

Statistics.

Primarily the Kolmogorov-Smirnov test for normal distribution was applied. DAS28 and SDAI values were compared by calculating correlation coefficients (Spearman's rho) and by linear regression analysis, as were the respective changes in the longitudinally observed patients.

Internal consistency analysis of the population was performed by the half-split method, accounting the correlation between even and odd listed values. Additionally, item weighting was assessed by exploratory factor analysis (principal component analysis [PCA]) and, subsequently, Cronbach's alpha was calculated as a marker of reliability (10). Because disease activity categories were not normally distributed, disease activity categorizing was compared by the Wilcoxon's rank sum test.

RESULTS

Cross-sectional analysis.

SDAI values were not normally distributed in the 399 patients, contrasting the DAS28 values (Z = 2.412, P < 0.0001 versus Z = 1.187, P = 0.119; see Figure 1). Thus for better comparability, SDAI and DAS28 values are given as medians (95% confidence interval [95% CI]; slope). The median DAS28 score was 3.42 (95% CI 1.92–4.92; slope 0.36) for the 399 patients, indicating moderate disease activity (3). The median SDAI was 11.5 (95% CI 0–23; slope 1.19), reflecting low disease activity according to the respective categorizing (5) (Figure 1). The correlation between the 2 scores was statistically significant (r = 0.897, P < 0.0001), as it was revealed by linear regression analysis (corrected R2 = 0.805, P < 0.0001).

Figure 1.

Distribution of Disease Activity Score including 28 joints (DAS28) and Simplified Disease Activity Index (SDAI) values in the entire patient cohort (n = 399).

Using the half-split method to assess homogeneity, a Spearman-Brown coefficient of 0.6902 for the SDAI and of 0.6357 for the DAS28 could be calculated. Additionally, factor analysis by PCA revealed both scores to be monocomponent (see Tables 1–4). Cronbach's alpha, as a measure of reliability and internal consistency, was 0.720 for the SDAI and 0.654 for the DAS28, indicating reasonable reliability of both scores.

Table 1. Correlation matrix for SDAI based on 399 cross-sectionally evaluated patients*
 SJCVAS-PHCRPVAS-GHSDAI
  • *

    SDAI = Simplified Disease Activity Index; SJC = swollen joint count; VAS-PH = Visual analog scale of physican's health assessment; CRP = C-reactive protein; VAS-GH = visual analog scale of patient's general health; TJC = tender joint count.

TJC0.6020.3420.1660.4340.881
SJC 0.5640.2310.3540.822
VAS-PH  0.2030.5200.644
CRP   0.0800.367
VAS-GH    0.624
Table 2. Correlation matrix for DAS28 based on 399 cross-sectionally evaluated patients*
 SJCVAS-GHESRDAS28
  • *

    DAS28 = Disease Activity Score for 28 joints; ESR = erythrocyte sedimentation rate. For other abbreviations, see Table 1.

TJC0.6020.4340.1310.768
SJC 0.3540.2530.700
VAS-GH  0.1780.667
ESR   0.486
Table 3. Item loading for the SDAI based on 399 cross-sectionally evaluated patients*
 Component 1
  • *

    One principal component extracted; principal component analysis. For abbreviations, see Table 1.

TJC0.304
SJC0.331
VAS-GH0.284
VAS-PH0.313
CRP0.143
Table 4. Item loading for the DAS28 based on 399 cross-sectionally evaluated patients*
 Component 1
  • *

    One principal component extracted; principal component analysis. DAS28 = Disease Activity Score in 28 joints; TJC = tender joint count; SJC = swollen joint count; VAS-GH = visual analog scale of patient's general health; ESR = erythrocyte sedimentation rate.

TJC0.406
SJC0.405
VAS-GH0.344
ESR0.209

Applying PCA for the SDAI, almost the same weighting was found for TJC, SJC, and VAS-PH, whereas a considerably lower one was found for VAS-GH and CRP. The higher weighting of VAS-PH than VAS-GH and the CRP in the SDAI is of particular interest. For the DAS28, PCA revealed a similar impact of the TJC and SJC upon the total score, whereas VAS-GH and ESR exerted a lower influence. As expected, for the higher number of items in the SDAI, the single item's weighting was seen to be lower than in the DAS28.

Disease activity categorizing.

Categorizing patients according to the proposed SDAI disease activity scoring revealed 298 patients (74.4%) to be at low, 84 patients (21.1%) at moderate, and 17 patients (4.3%) at high RA activity. According to the EULARC, 171 patients (42.9%) were at low, 165 patients (41.4%) at moderate, and 63 patients (15.8%) at high RA activity. Disease activity categories according to both tools were not normally distributed (see Figure 2).

Figure 2.

Distribution of disease activity categories in 399 patients according to the European League Against Rheumatism criteria (EULARC) and the original Simplified Disease Activity Index activity categories (SDAIAC): 1 = low, 2 = moderate, 3 = high, including a normal distribution curve. P < 0.0001; Wilcoxon's rank sum test.

Using Wilcoxon's rank test and the Monte Carlo approximation, a highly significant difference between the disease activity ranking according to both scores was found (Z = −13.078, P < 0.0001). Calculating lambda (0.322, P < 0.0001) proved these incongruencies. Remarkably, not a single patient was found at a higher disease activity stage according to the SDAI classification than according to the EULARC, but 172 patients were found higher according to EULARC than SDAI.

The impression arising from this observation was to reduce the limit for low disease activity according to the SDAI from 20 to 10 and the limit of high disease activity from 40 to 25 to achieve congruence between the 2 scales. Rating disease activity according to those limits resulted in insignificant differences between the EULARC and the SDAI (P = 0.07, Wilcoxon's rank test).

Longitudinal analysis.

The median values for the SDAI and the DAS28 in the 115 longitudinally observed patients at the 4 assessments are given in Table 5. Mean changes of the SDAI and the DAS28 between the 4 observations were insignificant. For the fourth observation, the DAS28 using the CRP (8) was also calculated. No differences could be found between both DAS scales, nor could any be found with respect to the correlation with the SDAI.

Table 5. SDAI and DAS28 values in the longitudinally observed patient cohort (n = 115)*
 First obs.Second obs.Third obs.Fourth obs.
  • *

    SDAI = Simplified Disease Activity Index; DAS28 = Disease Activity Score in 28 joints; obs. = observation.

SDAI (median ± SD)14.3 ± 12.014.5 ± 12.013.1 ± 10.7310.9 ± 10.5
DAS28 (median ± SD)4.06 ± 1.434.05 ± 1.363.88 ± 1.373.69 ± 1.4

Correlating the corresponding SDAI and DAS28 changes revealed a highly significant relationship between the respective changes at a level of P < 0.0001 (r1 = 0.896; r2 = 0.850; r3 = 0.839). However, disease activity categorizing at each of the 4 observations was significantly different (P < 0.0001). The number and the extent of changes of disease activity categories within the 2 scoring systems at the 4 observations were not statistically significantly different (χ2 = 4.815, P = 0.439).

DISCUSSION

To our knowledge, this is the first report dealing with the application of the SDAI in daily routine. Our main objective was to assess this tool's applicability for monitoring RA activity in daily practice. In this particular respect, the DAS28 and the EULARC (3) can be regarded as the gold standards. Thus a new tool has to prove at least the same (if not an even better) quality with respect to reliability, internal consistency, and easy applicability. In case of disease activity scores in daily practice, the applicability of the tool is highly dependent on the method of calculation.

In this respect, the easy formula provides clear advantages for the SDAI, which can be calculated simply by summing up the parameters. However, many devices and computed assistance are available that enable one to calculate the DAS28 promptly and easily. Thus, the advantage of the SDAI's simple calculation can be regarded as a relative one.

The metrics of the SDAI show that it is not normally distributed, with a heavy weighting on the lower end of the scale and a large skewing rightward on the graphic display. This general behavior of the instrument affects all later analyses and accounts for the anomalies of poor agreement with the EULARC, as well as the poor correlation in the lower portion of the scale. A transformation of the data with a logarithmic function may allow better normalization, as occurred with the transformation of the basic DAS28 elements. However, this would defeat the “simplified” aspect of the SDAI.

Reliability and internal consistency testing of both scores gave comparable results as expressed by very similar Spearman-Brown coefficients. Cronbach's alpha was slightly higher for the SDAI, probably due to the fact that alpha usually increases with an increasing number of items.

Interestingly, factor analysis showed that the acute phase reactants were found to exert the lowest influence upon the total score for both indices. In the case of the SDAI, it is not surprising that the CRP contributed relatively little to the overall instrument value, given the relatively low value range relative to the scores on the VAS, and even the joint counts. After all, the approach of addition used in the SDAI will weight those values with larger scores more heavily in the final sum.

With respect to absolute values, the strong correlation between the SDAI and the DAS28 as described was clearly reproducible (5). Moreover, longitudinal changes of both scores were found to be highly correlated. Thus the conclusion can be drawn that both instruments have the same sensitivity to change to describe disease activity fluctuations.

However, the crucial question as to whether initial disease activity categorizing by the SDAI and the DAS28 would be congruent could not be answered positively. Several measurements of congruency resulted in highly significant differences. Not surprisingly, the 2 scores are highly correlated, as they comprise the same parameters, yet their ability to distinguish between patients at different levels of disease activity is quite discrepant. Due to the results of this observation, low disease activity according to the SDAI or the DAS28 does not necessarily mean the same thing. However, the degree of congruence between the 2 scores seems to increase with higher disease activity. The SDAI activity categorizing was developed by experts' opinions only, and not in comparison with the EULARC, which may constitute the main reason for the discrepancies.

The question, which of the 2 scores describes RA activity more accurately, may be answered bidirectionally. On the one hand, this investigation could suggest that the SDAI is not as useful a measure as the DAS28, in that the DAS28 has a normal distribution curve through the middle of the scoring range, rather than at the left side of the curve. Moreover, the incongruence in disease activity categorizing limits the application of the SDAI. On the other hand, however, the good correlation between the SDAI and the DAS28 values and also between the changes of the scores indicate that both tools have comparable capability of describing disease activity and its changes.

Disease activity categorizing by the SDAI should be reconsidered on the basis of the EULARC. A direct comparison of both scales, including such patient-orientated scores as the Rheumatoid Arthritis Disease Activity Index (11), and patient satisfaction could also help to achieve a solution as to whether these scores are of similar applicability, particularly with respect to disease activity categorizing. According to the results of this study, we would propose to decrease the upper limit of low disease activity from 20 to 10 and the limit of high disease activity from 40 to 25 to achieve a significant degree of congruence between the 2 scales.

In summary, after some adjustments the SDAI could constitute a practicable tool for RA disease activity measurement in daily routine. However, since disease activity categorizing due to the SDAI shows significant discrepancies to the EULARC, this newly developed disease activity score should be applied cautiously to avoid misunderstandings and misinterpretations of disease activity.

Ancillary