Magnetic resonance imaging of sacroiliitis in children: frequency of findings and interobserver reliability

Sacroiliitis affects 30% of children with the enthesitis-related arthritis subtype of juvenile idiopathic arthritis (JIA) [1]. Typically, patients with juvenile spondyloarthritis present with lower limb arthritis and enthesitis, with sacroiliitis occurring later [2, 3]. The diagnosis can be challenging. Inflammatory back pain is a late feature in children who can be asymptomatic and examination findings are unreliable [3, 4]. However, sacroiliitis causes reduced mobility and disability [5], with worse functional status in those presenting as children [3]. Axial disease responds poorly to conventional treatments [5‐10]. Anti-tumour necrosis factor drugs as first-line agents for sacroiliitis increase mobility, improve quality of life and reduce pain [3, 6‐10]. Increasingly, clinicians rely on imaging in JIA to identify those with axial disease who would benefit from biologics, highlighting the importance of quality imaging and reliable reporting to allow correct patient selection.

The complex structure and orientation of the sacroiliac joints pose challenges for imaging. Interpretation is more challenging in the immature skeleton. Radiographs and computed tomography (CT) are now obsolete for sacroiliac joint imaging [4, 5, 11, 12]. Magnetic resonance imaging (MRI) is now the modality of choice as it can demonstrate active inflammation, seen long before structural changes develop [12].

Acute and chronic manifestations of sacroiliitis can be seen on MRI. Acute or active findings include bone marrow oedema, effusions and synovitis [12]. Fluid within the joint (effusion) can be physiological in adults, but in children some report this as evidence of synovitis [2, 13]. High T2-weighted signal intensity or enhancement of the capsule, insertions of ligaments or tendons represents capsulitis or enthesitis, respectively [12]. Chronic/structural findings include erosions, sclerosis, fat deposition and ankylosis. Periarticular fat deposition in spondyloarthropathy represents areas of previous inflammation after biological treatment [12]. Despite clinicians’ reliance on MRI for diagnosing sacroiliitis, the evidence in children is limited [2, 14].

There are no reporting standards or diagnostic criteria for paediatric sacroiliac MRI. Most paediatric imagers adapt adult scoring systems [4, 5], but these are not validated in children [15]. Inter-reporter reliability is variable [4, 14, 16]. MRI is often used as the gold standard for diagnosis [4, 16]. This may be justified as MRI can be superior to clinical modes of assessment in children, but definitive histopathological correlation is rarely available. There is a mismatch between the reliance on MRI for diagnosis and the lack of evidence to support current imaging and reporting practice.

Our aim was to review the imaging findings of sacroiliac joint MRIs in children with suspected sacroiliitis to determine the presence and distribution of features of sacroiliitis and determine the reliability of each MRI finding. A further aim was to evaluate the additional value of diffusion-weighted imaging (DWI) and contrast-enhanced sequences with regard to the presence and distribution of disease and reliability of the detection.

One hundred and two MRI examinations had been performed and were eligible for the study. Three were excluded as incomplete studies. This resulted in the inclusion of 99 MRI examination performed in 88 patients. Eleven examinations were repeat studies in patients who had already been scanned within the study period. These patients were not excluded and were reported independently (not in comparison to the initial study) but were noted by the primary investigator so that their impact on final results could be considered. The mean patient age was 15 years (range: 6–20 years, median: 15 years). Forty-one patients were female (47%). Examples of findings are demonstrated in Figs. 1, 2, 3 and 4.

In this explorative study, 12 cases out of 99 (12%) were considered to have sacroiliitis based on the panel’s MRI findings, a lower percentage than the 20–31% reported elsewhere in the literature [2, 16]. This may be due to different imaging referral practices/demographics. One study reported a higher frequency (64%), possibly due to referral bias or MRI overdiagnosis (due to poor imaging technique, no diagnostic criteria or subjectivity of interpretation, as demonstrated in our study by reader 3’s findings (39% sacroiliitis) [4].

Bone marrow oedema (Figs. 1 and 2) was seen in 8 cases (8%), fewer than the 20% reported in other studies [2, 16]. This may be due to different referral practices, e.g., including those with a diagnosis [16], as opposed to those with back pain.

Use of contrast medium must be justified, especially given concerns about intracranial gadolinium deposition [17‐19]. Abnormal enhancement was seen in 5% of cases, but these all had other features of active sacroiliitis. Giving contrast did not identify any additional cases. Some authors say contrast is essential to identify synovial enhancement, some saying this can be the only evidence of synovitis, but they do not report effusions [4, 5]. Others, similar to this study, report that all cases with synovial enhancement also have effusions [2, 13]. Furthermore, it is unknown if isolated synovial enhancement is pathological. Those advocating contrast state that normal paediatric sacroiliac joints do not enhance [2]. However, the quoted paper on normal sacroiliac joints in children states that capsular enhancement is normal, with no specific comment on the synovial portion [20]. In adults, synovitis alone is not sufficient to diagnose sacroiliitis [12]. Some say synovitis can be a solitary finding in children, but this is not proven. We believe a thin rim of synovial enhancement may be normal in children (Fig. 4), but this is difficult to prove. Despite this, discounting those not reporting effusions, the literature shows that all cases with enhancement have other MRI signs of sacroiliitis, which refutes the use of contrast in children.

DWI is increasingly used. Inflammatory changes in sacroiliitis are high signal on DWI and demonstrate free diffusion. Only 5% of MRIs in this study had abnormal diffusion-weighted signal. Of note, three patients with oedema and four with effusion had no corresponding DWI abnormality. This could be due to lower resolution of images or artefact. One group found significantly higher ADC values in sacroiliitis and advocate diffusion to quantify severity of inflammation in enthesitis-related arthritis [21]. However, they also found that ADC values in skeletally immature controls overlap with sacroiliitis [22]. Others use ADC values to assess response to anti-tumour necrosis factor treatment [23]. In this study, all cases with abnormal diffusion-weighted signal had high signal on STIR. However, we found that DWI can differentiate periarticular high signal due to red marrow (can exhibit restricted or free diffusion and is iso- or hyperintense to muscle on T1-weighted images) from bone marrow oedema (exhibits free diffusion and is hypointense to muscle on T1-weighted images). For these reasons, and as DWI is safe, we believe it should be included in protocols.

The frequency of erosions (8%) (Figs. 1 and 2) and sclerosis (6%) (Fig. 3) is surprising as structural changes are uncommon in children [2]. One explanation is that the positive cohort was older (mean: 15 years, range: 9–19 years, median: 14 years) compared to the total group (mean: 15 years, range: 6–20 years, median: 15 years). However, the youngest patient with erosions and sclerosis was 9 years old. Some say structural changes are uncommon until late teenage years [2, 15]; others report a higher frequency of 56% [4]. In our study, aside from a case with unilateral sclerosis, all cases with structural changes also had oedema and/or effusion, supporting beliefs that children with structural change usually still have active disease [15]. However, there is a reported case of an erosion without coexisting oedema or effusion, highlighting the need for paediatric diagnostic criteria [16]. In adults, structural changes are not sufficient evidence of sacroiliitis [12]. The definition of a positive MRI diagnosis in children, mostly focused on acute signs, is disputed. Adding structural changes to diagnostic criteria may not alter the number of diagnoses and more research is needed.

Deducing sensitivity and specificity of MRI findings for sacroiliitis without a gold standard is problematic. One study reports that consensus agreement on the presence of sacroiliitis is most sensitive (55%), followed by synovial enhancement (52%), concluding a negative MRI cannot exclude sacroiliitis [2]. Using synovial enhancement as a sign of sacroiliitis is questionable. Most agree that structural changes are specific for sacroiliitis, but these often occur late. Oedema is nonspecific in adults [12, 24] but more specific in children [2], possibly due to a reduced likelihood of other pathology in children. Some suggest that oedema on a single image can represent sacroiliitis in children, in contrast to adults [15]. However, normal MRI appearances of the immature skeleton are not well known; foci of T2 hyperintensity can be seen elsewhere in the body [25], but the relevance of this phenomenon in the sacroiliac joints is uncertain.

The most reliable MRI findings in our study were the overall presence of sacroiliitis and the identification of erosions, both demonstrating moderate interobserver agreement. The remainder of findings demonstrated poor agreement, ranging from slight to fair (Table 3). The poor concordance of findings between Reader 1 and Reader 3 emphasizes that reporting these MRIs is difficult and interpretation is subjective. The literature reports variable inter-reader reliability. Some report moderate to substantial agreement [1, 16]. One group found poor agreement on composite scores and individual signs but substantial agreement on the presence of sacroiliitis [4], possibly reflecting their high frequency of positive scans.

Some use MRI as the gold standard for diagnosis [4, 16], but no papers can prove the MRI features represent true pathology. There are conflicting opinions on the reporting and the significance of findings. In addition, some discuss imaging findings but show images where pathology is not appreciated or with pitfalls presented as pathology. Finally, there is a tendency for paediatric studies to apply adult literature to children, but sacroiliac joint imaging is more difficult in children given their smaller size, and increased cartilage and red marrow [4]. Also, structural changes commonly seen in adults are not predominant in children. These factors, combined with no diagnostic criteria and variable inter-reporter agreement make interpreting the evidence difficult. Given the lack of a definitive means of diagnosis and that MRI is often used as the gold standard, more emphasis needs to be placed on the reliability of MRI findings.

There is limited discussion about the need for more standardised reporting. As these MRIs are frequently performed in patients with JIA with no symptoms of sacroiliitis, radiologists’ reports influence management decisions. We suspect that current practice is to overcall findings and overdiagnose sacroiliitis on MRI in children. Using a pictorial reporting pro forma, which categorizes MRI features of sacroiliitis by the appropriate sequence as used in this study may help improve consistency in reporting and would allow internal and external inter-reader agreement in greater numbers of reporters across hospitals. In the absence of a gold standard, radiologists must improve reliability based on agreed criteria. In addition, consensus and expert groups should create strict imaging definitions to improve agreement.

Bone marrow oedema and erosions were the most common findings of sacroiliitis in children. Reliability of MRI signs of sacroiliitis in children is at best moderate for overall impression and presence of erosions with lower agreement for the active signs of sacroiliitis. Contrast is not required in routine sacroiliac joint MRI in children with JIA because enhancement does not occur without other features. More standardised reporting and paediatric diagnostic criteria are required to improve the quality of evidence in the future, using a reporting checklist may help with this.