Methods

The study was a single-centre, open-label 6-week trial approved by the local ethics committee. Study subjects fulfilled the New York modified criteria for AS6 and were eligible for anti-TNF therapy according to the Delphi study criteria for axial or peripheral disease.7 All patients gave informed written consent. Inclusion criteria were evaluated during stable treatment; after inclusion, NSAID treatment was discontinued for a 1-week washout period before starting etoricoxib. Baseline clinical and laboratory measurements and scans were performed after the washout period (not receiving any NSAID treatment).

Exclusion criteria were (a) diagnosis of other inflammatory diseases; (b) contraindication for etoricoxib administration; (c) previous etoricoxib administration and (d) administration of disease-modifying antirheumatic drugs, bisphosphonates, or anti-TNF therapy in the 4 weeks before recruitment.

Etoricoxib was given at a dose of 90 mg daily. If patients could not tolerate this dose, it was reduced to 60 mg. NSAIDs other than the study drug, corticosteroids, bisphosphonates and disease-modifying antirheumatic drugs were not allowed during the study.

Clinical and laboratory measurements were made at the two study visits (baseline and week 6). Clinical measures included patient’s global assessment of disease activity (100 mm visual analogue scale (VAS)), spinal night pain (100 mm VAS) and day pain (100 mm VAS), a 66 swollen and 68 tender joint count, the Maastricht AS Enthesitis Score (MASES),8 Bath Ankylosing Spondylitis Functional Index (BASFI),9 Bath Ankylosing Spondylitis Disease Activity Index (BASDAI)10 and Ankylosing Spondylitis Quality of Life questionnaire (ASQoL).11 Laboratory variables included ESR and C-reactive protein (CRP). Left and right lumbar lateral flexion, modified Schober’s index and occiput-to-wall measurements were also performed.

MRI scans of the SIJ, lower thoracic (T8–T12) and lumbar spine were performed at baseline and at 6 weeks using a commercially available 1.5 T Gyroscan ACS NT (Philips Medical Systems, Best, The Netherlands). The following sequences were used: T1-weighted turbo spin echo and T2-weighted SPIR coronal oblique of the SIJ, and T2 SPIR sagittal sequence of the spine. MRI parameters were as previously described.3 Two experienced observers (DMcG and HM-O) blinded for the clinical findings and the timing of the scans scored paired scans using the Leeds scoring system. When the two observers disagreed, a consensus was reached. As previously described,3 each SIJ was divided into four quadrants and oedema was recorded as present or absent within each. In the spine, each vertebra constituted a unit divided into vertebral body and posterior elements including facetal joint, spinous process and paraspinal soft tissues. Within each area, the number of lesions was recorded. On paired scans, each lesion was reported as new, worse, stable, improved or resolved.

The primary outcome was the proportion of patients fulfilling the SpondyloArthritis international Society (ASAS) response criteria for biological therapies (ASASBIO).12 Secondary objectives included the changes in MRI lesions and the ASAS20 response criteria.13 We also sought to correlate MRI lesions with treatment response.

Intention-to-treat analysis was performed on clinical data. When data were unavailable for the 6-week visit the last observation was carried forward. Mann–Whitney U tests and Wilcoxon signed rank tests for paired data were used for the analysis.

Results

Twenty-two patients were recruited (20 men) with median age 42 years (range 28–72). All patients had established AS and had been refractory to at least two NSAIDs at maximal tolerated doses for 3 months—mean prior use of four NSAIDs (range 2–8). At inclusion, 20 patients were receiving daily NSAID treatment (six diclofenac, three indometacin, three naproxen, two rofecoxib, two celecoxib, two phenylbutazone, one piroxicam and one meloxicam). Median BASDAI at screening was 5.1 (range 2.4–8.6) with 16 patients having a BASDAI over 4.

Twenty patients completed the 6-week treatment period: one patient withdrew owing to lack of efficacy and another owing to side effects (mood disorder); this patient declined further follow-up.

MRI outcomes

Baseline MRI scans were available from 21 patients (one SIJ MRI was discarded owing to technical failure of fat suppression); 19 complete sets of scans were available (the two non-completers refused follow-up MRI).

At baseline, 71% of the patients (15/21) had a total of 63 MRI-detectable lesions; seven patients showed abnormalities at both the SIJ and the spine. Overall, 60 lesions belonged to patients with complete sets of MRI scans and 22% of lesions (13/60) either improved or resolved after etoricoxib treatment. Figure 1 shows an MRI scan of one of the study patients.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

(A) represents a STIR fat suppressed MRI of the lumbar spine of one of the study patients. The white arrow shows an active lesion on the posterior inferior aspect of the T11 vertebral body. (B) corresponds to the same patient after treatment with etoricoxib. As shown by the white arrows the pre-existing lesion remains unchanged and there is a new lesion on the posterior superior aspect of the T12 vertebral body.

Findings in the SIJ

Eleven patients had osteitis (subchondral oedema) with a total of 25 lesions at baseline; lesions were more common in the ilium (15 lesions). Follow-up was available for 22 lesions (nine patients); 36% resolved (n = 8 lesions in three patients)—with all three patients showing complete resolution (see fig 2)—while 64% (n = 14) remained unchanged.

Findings in the spine

Eleven patients had a total of 38 active lesions in the spine: vertebral body (n = 26), spinous process (n = 2), perifacetal (n = 2) and interspinous lesions (n = 8). Over half the spinal lesions (53%) were in the lower thoracic region. Three lesions resolved after treatment and two improved, with one patient becoming free of spinal lesions. Two new lesions developed, three worsened and the remaining 30 remained stable. Figure 1 shows the findings for one of the patients.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

A T2 fat suppressed coronal oblique scan of one of the study patients. The white arrows in (A) show bone oedema in the ilial and sacral aspects of both sacroiliac joints. (B) The corresponding image after treatment with etoricoxib showing resolution of the bone oedema.

Clinical outcomes

Eight patients achieved an ASASBIO response and 12 patients achieved an ASAS20 response. Most of the clinical assessments showed a significant improvement (table 1).

Correlation of clinical and MRI findings

At week 6, CRP was associated with the total number of MRI-detectable lesions (Spearman correlation (r_s) = 0.35, p = 0.169), even though no correlation was found at baseline. Of the four patients whose overall number of MRI lesions improved, three fulfilled ASASBIO response criteria and the other fulfilled ASAS20 criteria. Six patients had no baseline MRI lesions; their median baseline BASDAI was 6.4 (range 2.7–9.4) and at 6 weeks 5.0 (2.0–9.1). Patients with baseline MRI lesions (15 patients) had a median baseline BASDAI of 5.7 (2.5–9.5), and at 6 weeks of 2.6 (1.3–9.5). Patients with a higher number of baseline MRI lesions, tended to have a greater improvement in their BASDAI scores (r_s = −0.414, p = 0.078). Eighty-eight per cent of ASASBIO responders had MRI baseline lesions, while only 55% of non-responders had MRI lesions (NS).

Discussion

This 6-week open-label study suggests that even though etoricoxib can partially relieve clinical symptoms in patients with established AS, MRI lesions show only slight improvement.

All patients had well-established AS and most had moderate–high clinical activity. However, in keeping with some previous reports,4 only 71% of the patients show inflammation on MRI. The discordance between clinical and imaging assessment highlights the need for careful study design for the use of MRI findings as inclusion criteria. In our small cohort the presence of basal MRI lesions seemed to correlate with therapeutic response. In a recent analysis a high degree of spinal inflammation increased the likelihood of reaching a major clinical response to biological therapy.1

We and others have consistently noted that over 60% of axial osteitis lesions improve after anti-TNF therapy. The degree of improvement in this study was small and comparable with that previously seen with methotrexate.3 In other studies no improvement in bone oedema was reported in placebo groups. The duration of treatment was certainly short and the natural history of MRI bone lesions is not well documented, although early response to anti-TNF therapy has been demonstrated.13 This possibly indicates that etoricoxib might have a small effect on bone oedema changes but further studies are warranted with the inclusion of a placebo arm to test this formally.

A number of limitations need to be taken into account when interpreting the results of this paper. The lack of a control group makes difficult the interpretation of the clinical data as spontaneous fluctuations in disease activity are not uncommon in AS. The size of this cohort is clearly limited and even though we planned to recruit patients with a high disease activity—as evidenced by their need for biological therapy—a few of the patients had a moderate or even low disease activity as estimated by BASDAI. Study duration, while relatively short, was chosen in accord with published data, suggesting a similar proportion of NSAID responders at 6 weeks and at 1 year.14 The extent of the spinal imaging is also in keeping with current minimum recommendations as it includes both lower thoracic and lumbar spine.15

In conclusion, this study shows that treatment with etoricoxib provides clinical improvement; however, this is not associated with a significant reduction in MRI-determined axial skeleton bone oedema. As noted previously, a significant proportion of patients with AS with high clinical disease activity may present no lesions on MRI. Special care is required with decisions to start anti-TNF therapy and there is a need for further investigation into the predictive value of MRI lesions for disability, ankylosis and treatment response.