Patients and methods

Patients

We analysed a total of 692 patients from three different AS cohorts. AS was classified according to the modified New York criteria.21 From all patients information about the presence or absence of peripheral manifestations was available. Baseline data are shown in table 1. Populations 1, 2 and 3 were used to assess truth aspects including correlation of the mini-BASDAI with other disease parameters. Populations 1 and 2 were also used to assess sensitivity to change.

Population 1: AS observational cohort

A total of 234 AS patients was seen at baseline (cross-sectional evaluation) in 13 rheumatological centres in the German Spondyloarthritis Inception Cohort (GESPIC). Only patients with a symptom duration of less than 10 years were included. The details of this cohort have been reported recently in more detail.22 From these 234 AS patients the BASDAI could be calculated in 228 patients (full data of six patients were missing).

Population 2: Randomised controlled trials with TNFα inhibitors in AS (TNFα trial population)

Sixty-nine AS patients treated with infliximab (5 mg/kg) and 30 AS patients treated with etanercept (25 mg twice a week) were available at baseline. Details of these two studies have been reported previously.23 24 A BASDAI score of 4 or more and failure of standard treatment were required for inclusion. Of these 99 AS patients, only those for whom the full set of BASDAI and patient global values were available at baseline, week 6 and week 12 were included in our analysis. These three visits were the visits both studies had in common. Finally, the data of 91 AS patients could be pooled for this analysis, eight could not be analysed because of missing data. Because data on different time points during treatment were available sensitivity to change could also be assessed.

Population 3: Randomised controlled trial in AS patients comparing celecoxib versus diclofenac (NSAID trial population)

This cohort comprised a total of 373 AS patients. AS patients either received celecoxib 200 mg once a day (33.2%), celecoxib 200 mg twice a day (32.3%) or diclofenac 75 mg twice a day (33.6%). The study duration was 12 weeks. Inclusion criterion was the presence of an acute episode of moderate to severe pain (visual analogue scale score ⩾40 mm) with an increase in pain score of 30% or greater in comparison to the screening visit after having stopped non-steroidal anti-inflammatory drug (NSAID) treatment (flare design). Importantly, only patients without present peripheral arthritis were included in the study. This study was reported in more detail elsewhere.25

Statistical analysis

BASDAI, the Bath Ankylosing Spondylitis Functional Index (BASFI), patient global and physician global were assessed by using a numeric rating scale with scores ranging from 0 (no disease activity) to 10 (worst disease activity). After calculation of the original BASDAI, we calculated a BASDAI by omitting questions 3 (Q3_arthritis) and 4 (Q4_enthesitis), here termed the mini-BASDAI, as follows: [(Q1_fatigue + Q2_{spinal pain}) + mean of (Q5_{strength morning stiffness} + Q6_{duration morning stiffness})] divided by 3 (see equation 1).

Peripheral arthritis was defined as the presence of at least one swollen joint according to the rheumatologist’s opinion, and enthesitis was defined as the presence of at least one enthesitic region according to the rheumatologist’s opinion. In the GESPIC as well as in the TNFα trial populations 12 enthesitic points were assessed according to the rheumatologist’s opinion, including the iliac crest, the greater trochanter of the femur, the medical and lateral condyles of the femur, the insertion of the Achilles tendon and the plantar fascia to the calcaneus, as has been reported before.22 23 24 26

For the GESPIC and the TNFα trial population we defined three groups of patients: (1) the total group of AS patients; (2) the subgroup without peripheral involvement (with neither arthritis nor enthesitis), here called the “P−” group; and (3) the subgroup with peripheral manifestations with at least arthritis or enthesitis being present, here called the “P+” group. A discrimination into “P+” and “P−” subgroups was not possible in the NSAID trial population as peripheral arthritis was an exclusion criterion for the study.

To compare the different groups with regard to scores on the original BASDAI and mini-BASDAI, on the separate questions of the BASDAI, on patient global, on physician global and on CRP a statistical analysis was performed by using the independent t test or the Mann–Whitney test, depending on whether data were normally distributed or not. A two-sided p value of less than 0.05 was considered statistically significant. To assess the correlation of the original BASDAI and mini-BASDAI with other disease parameters we calculated Spearman correlation coefficients. In the two trial populations effect sizes were calculated to assess sensitivity to change. The formula for effect size was the mean change (mean value at baseline minus mean value at week 12) divided by standard deviation at baseline.27 28

Results

Distribution of patients with and without peripheral joint involvement

Demographic data and the frequency of peripheral involvement for the three AS cohorts are shown in table 1.

The percentage of patients with pure axial disease manifestation without peripheral involvement (“P−” group) was 70.6% in the GESPIC at baseline and 39.6% in the TNFα trial population at baseline. By definition, patients in the NSAID trial population did not have peripheral arthritis.

Disease activity in subgroups with and without peripheral involvemesnt

In the GESPIC and TNFα trial population in which “P−” groups and “P+” groups could be compared those patients in the “P+” group clearly had higher disease activity than those in the “P−” group, as shown by higher values in all parameters that were assessed, including the single questions of the BASDAI.

Statistically significant differences were found in both populations for the original BASDAI, for BASFI (table 2) and for BASDAI Q3_arthritis, which showed mean values of 3.85 (“P+” group) versus 1.85 (“P−” group) in the GESPIC and 5.96 (“P+” group) versus 4.36 (“P−” group) in the TNFα trial population. Furthermore, statistically significant differences were found in the GESPIC for patient global (table 2) and for BASDAI Q4_enthesitis, with mean values of 4.13 (“P+” group) versus 2.73 (“P−” group).

Clear differences between the “P−” groups and “P+” groups, although not statistically significant, were also found for BASDAI Q1_fatigue, with mean values of 5.50 (“P+” group) versus 4.52 (“P−” group) in the GESPIC and 6.87 (“P+” group) versus 6.44 (“P−” group) in the TNFα trial population as well as for BASDAI Q2_{axial pain}, with mean values of 5.50 (“P+” group) versus 5.12 (“P−” group) in the GESPIC and 8.05 (“P+” group) versus 7.47 (“P−” group) in the TNFα trial population. Similarly, clear differences were found for BASDAI Q5_{strength morning stiffness} (mean values of 5.22 vs 4.52 in the GESPIC and 7.15 vs 6.72 in the TNFα trial population), for BASDAI Q6_{duration morning stiffness} (mean values of 3.55 vs 3.52 and 5.98 vs 5.66), and surprisingly also for BASDAI Q4_enthesitis in the TNFα trial population, with mean values of 6.35 (“P+” group) versus 5.94 (“P−” group).

The mini-BASDAI was always higher than the original BASDAI in all subgroups at both time points at baseline and week 12 (tables 2 and 3). In the “P−” groups the differences between the mini-BASDAI and the original BASDAI were higher than in the “P+” groups. So, deleting Q3_arthritis and Q4_enthesitis led to a stronger increase of the mini-BASDAI in the “P−” groups (tables 2 and 3).

However, when we used other disease activity parameters such as patient global and BASFI for comparison, the mini-BASDAI did not reflect disease activity in the “P−” group better, as shown by similar differences between the “P−” group and the “P+ group” for all these parameters.

Correlation analysis

In order to assess further whether the mini-BASDAI correlates better with other disease parameters, we assessed the correlation of the mini-BASDAI compared with the original BASDAI with patient global, BASFI, physician global and CRP in all three AS cohorts at baseline. Correlation analysis was also performed for the “P−” group and the “P+” group as well as at week 12 for the two trial populations.

As shown in table 4 there was a similarly good correlation of the mini-BASDAI with patient global, BASFI, CRP and physician global if compared with the correlation of the original BASDAI with the disease parameters, including the “P−” group.

We further calculated the correlation between the change scores of the original BASDAI (difference of the original BASDAI at baseline minus at week 12) and the mini-BASDAI (difference of the mini-BASDAI at baseline minus at week 12). In the TNFα trial population this correlation was 0.94 for the total group and 0.94 for the “P−” group compared with 0.93 for the “P+” group. In the NSAID cohort this correlation was 0.93.

In addition, we assessed the correlation of BASDAI Q3_arthritis and BASDAI Q4_enthesitis with the joint and enthesitis score (data not shown). There was only a weak correlation between BASDAI Q3_arthritis and the joint count in the “P+” group, with correlation coefficients ranging between 0.17 and 0.36 in all three AS cohorts. The correlation between BASDAI Q4_enthesitis and the enthesitis score was similarly weak, with correlation coefficients ranging between 0.03 and 0.54. These findings indicate that there might be patients with a small number of affected joints and enthesitic sites but a high activity or vice versa.

Interestingly, the correlation of the original BASDAI and mini-BASDAI with patient global and BASFI was worse when disease activity was elevated, as shown by low correlation coefficients of approximately r  =  0.5 at baseline in the clinical trial populations. In contrast, with lower disease activity this correlation improved, as can be seen by high correlation coefficients of approximately r  =  0.7–0.8 at week 12 in the two clinical trial populations and at baseline in the AS non-interventional cohort.

Sensitivity to change and assessment of response

Sensitivity to change of the mini-BASDAI and original BASDAI was assessed by calculating absolute and relative differences between baseline and week 12 and by calculating effect sizes in the two trial populations. As shown in table 3 the absolute and relative change between baseline and week 12 was similar for the mini-BASDAI, original BASDAI, patient global and BASFI, including again the “P−” subgroups. We also did not find clear differences between the original BASDAI, mini-BASDAI and patient global if effect sizes were calculated, although effect sizes were slightly higher for the mini-BASDAI if compared with the original BASDAI in the “P−” group of the TNFα trial population as well as in the NSAID trial population. Figure 1 also illustrates that in the NSAID trial population the mini-BASDAI was not more sensitive to change than the original BASDAI.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Original Bath Ankylosing Spondylitis Disease Activity Index (BASDAI), mini-BASDAI (BASDAI without Q3_arthritis and Q4_enthesitis) and patient global in a non-steroidal anti-inflammatory drug trial over 12 weeks: 373 ankylosing spondylitis patients were treated with celecoxib 200 mg once a day, 200 mg twice a day or diclofenac 75 mg twice a day for 12 weeks in a double-blind randomised controlled trial.25 The mini-BASDAI was not more sensitive to change than the original BASDAI.

Also when composite scores were used there was no significant difference, as shown by 81.3% patients in the total TNFα trial population who reached BASDAI 20 response, 58.2% who reached BASDAI 50 response and 33.0% who reached BASDAI 70 response when the original BASDAI was applied, similar to 80.2%, 52.7% and 29.7%, respectively, when the mini-BASDAI was applied. In the “P−” group and the “P+” group there were also no clear differences with regard to (mini-)BASDAI 20, 50 and 70 response (data not shown).

In the NSAID cohort the percentage of patients reaching a BASDAI 20, 50 and 70 response was 60.1%, 33.2% and 14.5% when the original BASDAI was used, similar to 60.1%, 31.4% and 14.2% when the mini-BASDAI was used.

Subgroup without peripheral involvement

In order to investigate the question of whether there is a small but relevant subgroup of patients with no peripheral involvement but with high axial disease activity that would not be visible on the group level, we tried to identify such a subgroup in the GESPIC. For the definition of this subgroup we looked for patients who had scored on the BASDAI Q3_arthritis and Q4_enthesitis each 1 or less, on the original BASDAI less than 4, but on the mini-BASDAI more than 5. We identified 12 out of 228 patients (5.3%) who fulfilled these criteria and showed mini-BASDAI values of 5.59 (SD 0.38), which were very similar to patient global with a mean value of 5.75 (SD 2.01). Disease activity was also further illustrated by high scores for questions 1, 2, 5 and 6 of the BASDAI, which were 5.33 (SD 2.54), 6.58 (SD 1.56), 5.0 (SD 1.86) and 4.69 (SD 2.27), respectively. Questions 3 and 4 were each scored with 0.25 (SD 0.45). The original BASDAI was 3.45 (SD 0.24).

Discussion

In the observational AS cohort (GESPIC) approximately 70% of patients belonged to the “P−” group showing pure axial disease activity (back pain) at baseline without peripheral involvement, which is very similar to the percentage in the OASIS cohort.16 These findings are also similar to the TNFα trial population at week 12 (after active treatment) when approximately 80% showed no peripheral involvement. In the TNFα trial population the percentage of patients without peripheral involvement was only approximately 39% at baseline. These data show that there is indeed a substantial proportion of AS patients without peripheral involvement, which has also been shown in many other studies.7 8 9 10 11 12 13 14 15 16 17

The “P+” group was more active compared with the “P−” group in all scores, as shown by higher values of the original BASDAI, the mini-BASDAI, patient global and BASFI. This finding is compatible with previous studies, which also found that patients with peripheral arthritis generally have more active disease compared with patients without peripheral arthritis.29 30 31 Based on our analysis we do not know at the moment whether this indicates more active spinal disease in the “P+” group or whether patients with peripheral manifestations just score axial symptoms higher.

The mini-BASDAI was always higher than the original BASDAI in all three populations at all visits. This is due to generally lower mean values of Q3_arthritis and Q4_enthesitis of the BASDAI compared with the higher mean values of questions Q1_fatigue, Q2_{spinal pain}, Q5_{strength morning stiffness} and Q6_{duration morning stiffness}. Interestingly, this was also the case in the “P+” group. Therefore, even in the “P+” group the axial aspect of disease activity was more pronounced than the peripheral aspect of disease activity, which underlines the fact that AS is primarily a disease of the axial skeleton.

Although the mini-BASDAI was higher than the original BASDAI, it did not correlate better compared with the original BASDAI with parameters of global disease activity such as patient global, CRP, physician global and BASFI. Furthermore, the sensitivity to change was similar for the original BASDAI and mini-BASDAI in AS patients treated with TNFα inhibitors and with NSAID. The mini-BASDAI was thus not superior compared with the original BASDAI following aspects of the OMERACT criteria with regard to truth and sensitivity to change. Feasibility was not assessed separately, but leaving out two out of six original BASDAI questions is likely to improve feasibility.

The question whether in AS patients without peripheral involvement (“P−” group) the application of the original BASDAI may lead to a false low estimation of disease activity has already been assessed in a similar study by Heuft-Dorenbosch et al.16 They also found that the mini-BASDAI does not offer an advantage to assess disease activity in patients without peripheral manifestations. However, whereas in that study only one observational cohort of AS patients was investigated, we had the chance to investigate three AS cohorts, including two treatment trials with a total of 692 patients. Patients from therapy studies normally have a higher disease activity at baseline, as was the case here, and this offers the possibility to assess sensitivity to change upon treatment.

Finally, we addressed the question of whether there is a relevant subgroup of AS patients with very low scoring of questions Q3_arthritis and Q4_enthesitis of the BASDAI, but very high scoring of the other questions of the BASDAI. We identified approximately 5% of patients with a mean mini-BASDAI value of 5.6. All these patients scored questions Q3_arthritis and Q4_enthesitis 1 or less and had a similarly high patient global value of 5.9. In these patients the original BASDAI was only 3.5. In this small but relevant subgroup, the original BASDAI indeed has some limitations in correctly assessing disease activity. However, as the percentage is quite small this might be acceptable.

Mainly because of the already discussed poor correlation of the BASDAI with other markers of disease activity, including CRP, erythrocyte sedimentation rate and magnetic resonance imaging findings as well as radiographic progression, currently the Assessment of SpondyloArthritis international Society (ASAS) has developed a new disease score called the ASAS-endorsed disease activity score.32 It has to be seen whether this score catches the small 5% group better in whom disease activity would be underestimated by using the original BASDAI.

Overall, the original BASDAI, including questions Q3_arthritis and Q4_enthesitis, is a valid instrument to assess disease activity also in patients without peripheral involvement.