Article Text

Download PDFPDF

Extended report
Short-term changes in magnetic resonance imaging and disease activity in response to infliximab
  1. H M Bonel1,
  2. C Boller2,
  3. B Saar1,
  4. S Tanner2,
  5. S Srivastav3,
  6. P M Villiger2
  1. 1
    Institute for Diagnostic, Interventional and Pediatric Radiology, Inselspital, University of Bern, Bern, Switzerland
  2. 2
    Department of Rheumatology and Clinical Immunology/Allergology, Inselspital, University of Bern, Bern, Switzerland
  3. 3
    Department of Biostatistics, Tulane University, New Orleans, Los Angeles, California, USA
  1. Correspondence to Professor P M Villiger, Department of Rheumatology and Clinical Immunology/Allergology, Inselspital, University of Bern, 3010 Bern, Switzerland; peter.villiger{at}


Objectives: To characterise and quantify short-term changes in local inflammation using magnetic resonance imaging (MRI), and to correlate the findings with clinical disease activity in response to infliximab in patients with spondyloarthritis.

Methods: 28 consecutive patients with established spondyloarthritis under successful long-term treatment with infliximab underwent MRI immediately before and one week after re-administration of the TNF blocker. C-reactive protein and the Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) were assessed at both time points. The MRI protocol included coronal and sagittal turbo-short T1 inversion recovery (STIR) images as well as contrast-enhanced sagittal T1-weighted, fat-suppressed images. Images were assessed in independent sessions using the ASspiMRI-a score, the signal-difference-to-noise ratios (SDNR) and volumetry to assess oedematous and inflamed tissues.

Results: BASDAI values were expectedly low at study entry (3.3, SD 2.3). One week after administration of infliximab, 46% of patients reached a BASDAI 20, 39% a BASDAI 50. Kappa values for qualitative assessments and all measurements were excellent (range between 0.83 and 1.0) The ASspiMRI-a dropped most in the thoracic (3.3 points), less in the lumbar (1.21 points) and least in the cervical spine (0.38 points). The decrease of the ASspiMRI-a, the SDNR and the inflamed volumes in response to infliximab re-treatment was significant (p<0.01). The BASDAI showed a weak correlation with the ASspiMRI-a (r  =  0.41).

Conclusions: MRI proves to be a valid method to assess and quantify short-term effects of therapy in spondyloarthritis. Comparison between MRI and BASDAI changes show that the BASDAI may underestimate local inflammation. It suggests an explanation for the structural disease progression despite clinical remission.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Ankylosing spondylitis (AS) and related spondyloarthritides are inflammatory diseases predominantly affecting the axial skeleton of young men and presenting with inflammatory back pain.1 The Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) is a validated assessment tool for monitoring clinical disease activity.2 3 4 The magnetic resonance imaging (MRI) spine score for the assessment of acute changes in ankylosing spondylitis (ASspiMRI-a) has been developed recently.5 It scores acute changes, such as oedema and active erosions, with a point scoring system ranging from 1 to 6 for each of the 23 vertebral units. The score has been used for documentation of new treatments or to assess inflammatory changes after established treatment of several weeks.5 A recent study showed a correlation between the ASspiMRI and clinical findings over 54 weeks of therapy with infliximab.6

TNF blocking strategies proved to be highly effective in the treatment of AS.5 However, recent data indicate that they do not stop the progression of pathognomonic structural changes.7 This may suggest that the pathway leading to the characteristic new bone formation of AS is not (only) dependent on TNF bioactivity. Alternatively, one may argue that the current treatment protocols do not sufficiently suppress TNF bioactivity. Indeed, a recent study showed discordance between clinical and structural remission, supporting the second hypothesis.8

The goals of this study were: to analyse whether MRI qualifies to assess and quantify early changes in response to infliximab in a cohort of patients under successful long-term therapy; second, to compare clinical and radiological changes at the start of flare and in response to treatment.



Twenty-eight consecutive patients (23 men, five women; median 38.5 years of age (SD 12.43), range 24–62 years), who had AS according to the modified New York diagnostic criteria for AS9 or related spondyloarthritides10 and had continuously received infliximab for more than 19 months, were recruited for this study (for patient characteristics see table 1).

Table 1

Patient characteristics

Exclusion criteria were: patients younger than 18 years, contraindications for MRI (such as implants), renal insufficiency, known allergies to contrast media and claustrophobia. 57% of the patients were classified as AS, 14% as psoriasis spondyloarthritis, 25% as undifferentiated spondyloarthritis and 3.6% as spondyloarthritis associated with Crohn’s disease. The median disease duration was 39 months (range 24–62).

The intervals of infliximab infusions were individualised, ie, they were lengthened until patients reported the first symptoms of a relapse in the week before the next infusion. Therefore, intervals ranged between 4 and 12 weeks; see table 1. In order to analyse comparable disease conditions, intervals and drug doses had to be kept stable for each individual patient before study entry. The patients entered the study at the time of re-administration of infliximab. Clinical assessments, determination of C-reactive protein (CRP) and MRI were performed immediately before the administration of the TNFα inhibitor and exactly one week later.

The study was approved by the local ethical commission and performed according to GCP standards. All patients gave written informed consent.

MRI protocol

Patients were examined on an advanced 3 Tesla MR system (Magnetom Trio Matrix; Siemens, Erlangen, Germany) using a standard phased-array spine coil. The cervical, thoracic and lumbar spines as well as the sacroiliac joints were imaged in one session. The patients were centered in the gantry for the cervical spine and upper half of the thoracic spine, and in an additional step the lower thoracic and lumbar spine were centered and imaged.

Turbo-short T1 inversion recovery sequences (tSTIR: TR 6300 ms, TE 81 ms, TI 200 ms, echo train length 13), were applied in coronal and sagittal planes for the cervical, thoracic and lumbar spine. Contrast-enhanced T1-weighted turbo-spin echo images (TR 422–650 ms, TE 11 ms, echo train length 7) were used after intravenous contrast in sagittal planes. Slice thickness was 3 mm for the cervical and cranial thoracic spine and 3.5 mm for the caudal thoracic and lumbar spine. A rectangular field of view was used for all patients with a craniocaudal length of 480 mm and a corresponding acquisition matrix of 512 voxels to yield an in-plane resolution of 0.88 mm2.

In addition, a three-dimensional dynamic T1-weighted sequence (volumetric interpolated breath-hold examination: TR 3 ms, TE 1.1 ms, flip angle 11°) was done in sagittal planes with a slice thickness of 2.5 mm, a field of view of 320 mm and an in-plane resolution of 1.25 × 1.25 mm once per minute for 6 minutes. The dynamic sequence was applied on the site of most intense disease activity at the first examination. The purpose of this dynamic sequence is multiple: assessment of inflammatory lesions in the bone can be done with high reliability. Enthesitis can be differentiated with ease from vascular enhancement using three-dimensional maximum intensity projections. Subtractions assist in the identification of inflamed regions and segmentation of its volume.

Intravenous contrast (10 ml gadolinium BOPTA (Multihance); Bracco, Milan, Italy)11 was applied via a 22 G needle over a cubital vein using a standardised injection protocol (3 ml/s, 10 ml saline flush also with a flow of 3 ml/s).

Image evaluation protocol

Two experienced senior radiologists, a general magnetic resonance radiologist and a radiologist specialised in musculoskeletal imaging, blinded to clinical and imaging data and the time of the examination, evaluated all magnetic resonance examinations in independent sessions. Before the evaluation sessions reader training was performed on 15 routine cases that were not part of this study. All readings were performed on a standard PACS workstation with two 3 megapixel monitors suitable for reporting (Easyvision; Philips, Hamburg, Germany). The examinations were presented in random order and the readers were not aware whether they evaluated the baseline or follow-up examination. In the first of the two reading sessions either the baseline or the follow-up examination of a specific patient was presented, in the second reading session 6 weeks later the matching examination was presented in order to avoid a similar rating just by remembering. In a third reading session again 6 weeks later only the STIR sequences of the baseline examinations were evaluated to test the detection of actively inflamed segments by STIR sequences alone and thus the necessity to use of intravenous contrast.

The complete spine with all 23 intervertebral spaces was scored using the ASspiMRI-a scoring system and a summary score was calculated.5 In addition, signal-difference-to-noise ratios (SDNR)11 12 were measured for all active lesions in tSTIR and contrast-enhanced T1-weighted images with fat suppression. The inflamed areas in the vertebral bodies were also segmented and inflammatory volume assessed in cm3 by multiplication with slice thickness and with respect to the gap between the slices at both time points. Finally, the readers determined the most useful sequence for the evaluation of each patient.

Clinical assessment

The BASDAI was assessed at both time points by investigators who were aware of the study but were blinded to all patient information including medication, MRI findings and laboratory results.


All statistical analyses were performed using SAS statistical software, version 9.1, and all results were tested at a 5% level of significance. Data were tested for normal distribution, and the Student’s ttest (or Wilcoxon signed rank test) was used as a parametric (or non-parametric) technique for comparing two groups. The analysis of variance method was used to compare more than two groups for studied variables. Interobserver variability was calculated and analysed using the Cohen κ test. Interobserver agreement was scored as poor (κ  =  0.1–0.4), fair (κ  =  0.41–0.6), moderate (κ  =  0.61–0.8) or excellent (κ  =  0.81–1.0).13 Correlation was tested using Spearman coefficients.


Clinical evaluation

As outlined above, patients were treated within one week of relapse (ie, first clinical signs of emerging disease activity). Accordingly, BASDAI values were low at study entry (mean 3.3, SD 2.3). A decline of 47% could be recorded one week after the administration of infliximab (mean BASDAI of 1.95, SD 2.27, p<0.01). Thirty-nine per cent of patients reached a BASDAI 50 and 46% a BASDAI 20. Changes in CRP levels did not reach statistical significance (6.26 (SD 6.78) to 4.08 (SD 2.88), p = 0.08).

MRI evaluation

Using only STIR images, reader 1 detected 58 and reader 2 63 segments with signs of acute inflammation. With the addition of contrast-enhanced images both readers correctly assessed 83 segments with disease activity. Based on the observation that up to 30% of acutely inflamed segments would not have been correctly detected on STIR images, the readers decided to use the T1-weighted fat-suppressed contrast-enhanced sequence for MRI evaluation.

Kappa values were excellent and ranged between 0.91 and 1.0 for the baseline assessment of the ASspiMRI-a and between 0.83 and 1.0 for the follow-up examination. For the quantitative evaluation interreader correlation was also excellent (κ  =  0.85–1.0).

All patients demonstrated signs of acute inflammation at the time of infliximab infusion despite the long-term successful treatment of the disease with this TNF blocking antibody. Summary scores for the complete spine are given in fig 1. There was a small but significant decrease in the ASspiMRI-a as assessed by both readers (p<0.01).

Figure 1

Box and whiskers plot (median, third and first interquartile ranges, maximum and minimum observations) of magnetic resonance imaging spine score for the assessment of acute changes in ankylosing spondylitis (ASspiMRI-a) at baseline examination and at follow-up examination. The score summarises the assessment of the whole spine for both readers. The asterisk (*) indicates a significant decrease in the ASspiMRI-a.

The signs of acute inflammation as documented by the ASspiMRI-a were found significantly more frequently in the lower thoracic and lower lumbar spine before re-treatment. The ASspiMRI-a dropped most in the thoracic (3.3 points), less in the lumbar (1.21 points) and least in the cervical spine (0.38 points) (fig 2).

Figure 2

Diagram of magnetic resonance imaging spine scores for the assessment of acute changes in ankylosing spondylitis (ASspiMRI-a) indicating the localisation of acute changes found in the spine and their change within one week. The x-axis shows average scores per segment. CV, cervical vertebrae; LV, lumbar verterbrae; TV, thoracic vertebrae.

Figure 3 summarises all the quantitative data at baseline and follow-up examination. Volumes of inflamed bone and SDNR in both magnetic resonance sequences decreased significantly within one week (p<0.01). The reduction in altered bone marrow, however, is better seen in contrast-enhanced images compared with in tSTIR imaging (fig 4). The high noise of the tSTIR sequence causes a relatively low SDNR in comparison with T1-weighted contrast-enhanced images; however, in both sequences there was a significant drop in the SDNR.

Figure 3

Box and whiskers plot (median, third and first interquartile ranges, maximum and minimum observations) of the quantitative evaluation of signal-difference-to-noise ratios (SDNR) (A) and volumes (B) for active lesions in the 23 intervertebral segments. The asterisk (*) indicates a significant change of p<0.005. Short T1 inversion recovery (STIR) is displayed on the left y-axis, contrast-enhanced T1 with fat suppression on the right y-axis.

Figure 4

Thoracic spine in a 22-year-old man (Th6–Th8). The oedema in the annulus fibrosus and the adjacent bone marrow can be well appreciated in short T1 inversion recovery (STIR) sequences (A, B), but enhancement is obvious in the fat-suppressed T1-weighted images (C, D) (arrows). One week after the administration of infliximab (B, D), the inflammation has decreased significantly.

Assessment of oedematous and inflamed volumes is much more difficult in tSTIR images, because delineation of altered bone marrow regions is supported by a steep contrast and impeded by flaw depiction. The range of volumes in the follow-up of tSTIR images is therefore much broader compared with T1-weighted fat-suppressed images (fig 3), as the assessment of altered bone marrow regions is more difficult because of reduced image contrast in tSTIR images.

Correlation of parameters

Correlation of the ASspiMRI-a with volume measurements using contrast-enhanced T1-weighted sequences was best (r  =  0.47), correlation with the BASDAI and volume measurements based on STIR imaging was similarly high (r  =  0.41).

Table 2 informs about the correlation of magnetic resonance findings. Generally, volume measurements correlate better than SDNR measurements and T1-weighted contrast-enhanced image data correlate better than tSTIR.

Table 2

Correlation of quantitative magnetic resonance findings at baseline (upper right) and follow-up (lower left)


The TNFα blocker infliximab has been shown to be safe and effective in the treatment of AS and related diseases. Patients frequently report a rapid improvement in signs and symptoms, often within one week after the administration of infliximab. Despite clinical benefit, recent data have challenged the hypothesis that the effect on disease activity is paralleled by an effect on disease evolution, and a limited correlation between persistent local inflammation measured in MRI and clinical disease activity parameters was found.7 8 Regarding pathogenesis, it has been proposed that osteoanabolic changes are caused by low levels of dickkopf protein, which are not directly influenced by the TNF pathway.14 15 However, it is also conceivable that persisting subclinical disease activity may promote disease progression. Recent data support the hypothesis that spinal inflammation increases the likelihood of new developing syndesmophytes.16

The effect of TNF inhibition on local inflammation has been analysed by MRI in several studies.5 6 7 8 17 18 19 20 21 22 23 However, in most cohorts, MRI was performed at the end of regular treatment intervals, and the dose of TNF blocker administered was chosen according to the initial study protocols. As a consequence, patient cohorts were rather heterogeneous regarding clinical disease activity at study entry and regarding their degree of response. As we were interested to analyse the performance of MRI to detect early signs of flare, but also to study early changes in response to infliximab re-administration, we decided to choose a cohort of spondyloarthritis patients on long-term and effective treatment with infliximab. Disease activity at study entry was kept low by individualisation of treatment intervals. This study design should allow the assessment and quantification of MRI signals, to test the performance of the ASspiMRI-a at low levels of inflammation and to compare the dynamic of clinical versus local (ie, spinal) response.

Despite low to moderate disease activity at study entry, a BASDAI 50 was reached by almost 40% of patients one week after the re-administration of infliximab and several patients recorded complete remission. In parallel, the decrease in MRI signals (SDNR, inflammatory volume and the ASspiMRI-a) demonstrated a significant reduction in inflammation in the spine. Remarkably, the statistical correlation between the ASspiMRI-a and the BASDAI was rather low and the decrease in MRI signals was far less impressive than the clinical response (25% and 29% decrease in SDNR for tSTIR and contrast-enhanced sequences, respectively). Taken together, the results strongly support the earlier notion that clinical assessments such as the BASDAI will underestimate biological disease activity. It is conceivable that persistent or fluctuating local disease activity is sufficient to promote bony ankylosis. Further prospective studies should address the question of whether MRI signals precede clinical signs of relapse and whether MRI might help to predict the dose of TNF blocking agents required to achieve complete and lasting remission.

Acute changes as detected by the ASspiMRI-a were dominant in the lower thoracic spine and the lower lumbar spine. In these regions, the ASspiMRI-a also decreased most considerably. In the original paper introducing the scoring system of the ASspiMRI-a, the average score of untreated patients was 112.5, and a treatment of 3 months resulted in an average score of 68.5 In comparison with these data, the score of our patients at study entry corresponds to a low disease activity. The fact that the ASspiMRI-a showed a significant decrease within one week after the administration of infliximab demonstrates the potential of the score to pick up even small changes in intervertebral segments. This is the first study to exclude a ceiling effect of the ASspiMRI-a at low levels of disease activity.

As a result of additional costs and of the risk of side effects the use of contrast agents is frequently questioned. Turbo-STIR is frequently given the preference to show inflammatory bone oedema reliably. Turbo-STIR imaging, however, also carries the disadvantages of high noise, long imaging times and soft contrast between normal and altered bone marrow. In addition, oedematous regions that are depicted with very high sensitivity may be difficult to differentiate from haematopoietic bone marrow. The soft image contrast and the poor discrimination from physiological bone marrow changes were the main problem in the volumetry of inflamed bone marrow. We tried to overcome the problem of poor magnet homogeneity, which is a major limitation at the caudal and cranial end of the field of view, by reducing the field of view to 48 cm instead of 50 cm and doing a two-step examination with optimal centering of the patient. In these fat-saturated contrast-enhanced T1-weighted images, inflammatory bone marrow changes can be differentiated with ease from chronic changes caused by AS or normal and haematopoietic bone marrow. Enhancement is much more pronounced in inflamed regions, and inflammatory tissues are depicted with a steeper contrast and higher SDNR (fig 3). This resulted in an increased number of acutely inflamed segments of up to 30%! The problem of the poor delineation of inflamed regions in tSTIR images becomes obvious in fig 3. In the follow-up examination, assessment of inflamed volumes results in a much larger statistical range of volumes compared with baseline examinations and contrast-enhanced imaging. The better delineation of enhancing tissues compared with the presentation of oedema results in a demarcation that allows easier segmentation of volumes in this setting. The preference for contrast-enhanced sequences in our study seems to contradict previous findings. However, previous studies were performed using a lower main field strength of the MRI system, and disease activity was higher compared with our patient cohort. In addition, interreader correlation benefits from contrast-enhanced sequences. Finally, subtraction images were available to support the readers; these supported the use of contrast-enhanced images by using a dynamic aspect of the examination and a smaller slice thickness. Regarding biological response to TNF inhibition, adjacent residual bone oedema might decrease slower than inflammatory tissue; this way contrast-enhanced imaging would contribute more precision. In agreement, correlation between contrast-enhanced fat-suppressed imaging and the ASspiMRI-a was better than with to tSTIR imaging. In addition, in the quantitative evaluation, correlations were better using contrast-enhanced sequences and focusing on volumes than on relying on signal intensities. In conclusion, our data strongly suggest that T1-weighted fat-suppressed contrast-enhanced sequences show multiple advantages over tSTIR imaging. We are convinced that these advantages outweigh additional costs and the low risk of side effects. In clinical routine, however, the absolute number and size of lesions might not have the same importance as in a clinical trial.

It is remarkable that we found only a moderate correlation between absolute inflamed bone volume and the ASspiMRI-a. It is currently unknown whether an assessment of inflamed bone volume bears information of clinical interest. It is possible that the dose of TNF blocking agents necessary to suppress spinal inflammation completely is correlated to one or the other variable. Prospective studies will have to address this issue. We therefore suggest that the size of the inflamed tissue and number of affected segments should be assessed separately to yield an optimal score for short-term controls.

In summary, the data presented show that at the time of the start of a clinical flare, MRI detects inflammatory signals in the axial skeleton of all patients analysed, with a predominance of the thoracic and lumbar spine. MRI proved to be a reliable tool to quantify short-term changes of disease activity, eg, the reduction in local inflammation one week after re-administration of the TNF blocker infliximab. Furthermore, the data exclude a ceiling effect of the ASspiMRI-a at the lower end of disease activity. Comparison of MRI data with clinical response showed a persistent subclinical disease activity in several patients. Smoldering disease may contribute to the recently described progression of ankylosis in patients with clinically successful TNF inhibition. Regarding MRI technology, intravenous contrast agents significantly improved the assessment of inflammatory signals. Its use is recommended especially for clinical studies. Volume measurements correlate better than a mere assessment of signal intensities and therefore should be part of any score evaluating disease activity of spondyloarthritis.



  • Competing interests None.

  • Ethics approval The study was approved by the local ethical commission and was performed according to good clinical practice standards.

  • Patient consent Obtained.

  • Provenance and Peer review Not commissioned; externally peer reviewed.