Introduction

The term spondyloarthritis (SpA) refers to a group of inflammatory rheumatic disorders that can be broadly classified into axial SpA (axSpA), which primarily involves the spine and the sacroiliac joint (SIJ), and peripheral SpA, which primarily affects the extremities [1]. A recent revision of the classification criteria for axSpA by the Assessment of Spondyloarthritis International Society (ASAS) advocated for the further subdivision of axSpA into radiographic axSpA (r-axSpA; i.e., ankylosing spondylitis [AS]) and non-radiographic axSpA (nr-axSpA) [2]. Diagnosis of r-axSpA is based on the presence of definite sacroiliitis on X-ray imaging, in accordance with the modified New York criteria for AS [3]. Diagnosis of nr-axSpA is based on the presence of sacroiliitis on magnetic resonance imaging (MRI) or positivity for the human leukocyte antigen B27 (HLA-B27) [4], in addition to clinical and laboratory features associated with SpA [5].

MRI-evident sacroiliitis can precede the detection of sacroiliitis on radiographs by nearly a decade [6]; hence, early detection of axSpA may enable timely implementation of appropriate disease management strategies. Active inflammation in the SIJ and spine, as evidenced by inflammatory changes (bone marrow edema) that are followed by structural lesions (joint erosion, fat metaplasia) seen on MR images [7, 8], leads to bone repair and secondary bone formation, thus exacerbating disease progression in both early and established axSpA [9,10,11,12].

The management of patients with axSpA should be personalized according to their current disease state (e.g., axial, peripheral, and extra-articular manifestations), and any decision to initiate treatment with biological disease-modifying antirheumatic drugs (bDMARDs) should take into consideration C-reactive protein levels and MRI or radiographic findings [13]. Biological inhibitors of the proinflammatory cytokine tumor necrosis factor alpha (TNFα) have been shown to be an effective treatment in reducing SpA disease activity and improving patient function [14]. To date, four TNFα inhibitors—adalimumab, certolizumab, etanercept, and golimumab—have been approved by the European Medicines Agency for treatment of nr-axSpA. A fifth TNFα inhibitor, infliximab, has not been approved yet for use in patients with nr-axSpA, but its safety and efficacy in this patient population have been investigated in randomized controlled trials (RCTs). Considering the progressive nature of axSpA, it is important to determine whether TNFα inhibitors have an effect on early inflammatory and structural lesions detected on MR images.

The purpose of this literature review was to identify RCTs that evaluated the impact of TNFα-inhibitor therapy on inflammatory and structural lesions (particularly fatty lesions) in early axSpA, as assessed using MRI, and to summarize those findings.

Methods

Literature Search Strategy

A search of the MEDLINE® and PubMed Central® databases was conducted in February 2018 and repeated in December 2018, using the PubMed® platform and the following search string: “(axial spondyloarthritis OR non-radiographic axial spondyloarthritis OR nonradiographic axSpA OR nr-axSpA OR non-radiographic axSpA) AND (MRI OR magnetic) AND (adalimumab OR certolizumab OR etanercept OR golimumab OR infliximab).” Individual searches were conducted for each TNFα inhibitor, in combination with the disease subtype and imaging modality search terms. The search was limited to full-text, English-language articles published since January 2007. Titles and abstracts of retrieved articles were screened manually to identify RCTs of TNFα inhibitors that assessed inflammatory or fatty lesion progression in the spine or SIJ using MRI in a population that included at least 40% of patients with early axSpA, defined as nr-axSpA (unless data analysis was stratified by radiographic vs. nr-axSpA). In this context, the adjective “early” refers to the extent of damage the SIJ and spine have sustained (and detectability of that damage using radiography), and not necessarily to the duration of symptoms.

Compliance with Ethics Guidelines

This article is based on previously conducted studies and does not contain any studies with human participants or animals performed by any of the authors. All procedures performed in studies involving human participants that were cited in this review were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards, as reported in the primary reports.

Results

Literature Search

A total of 106 articles were retrieved from the literature searches. Of these, 33 reported results from RCTs involving TNFα-inhibitor therapy and its effect on structural progression in early axSpA, with the majority of RCTs presenting data from trials on etanercept (n = 20). A detailed appraisal of these 33 papers identified the 19 which were ultimately included in this review (Table 1). Articles were excluded from further review if they did not report the outcome measure of interest, they included < 40% of patients with early axSpA, and they did not stratify results by radiographic versus nr-axSpA, or were reviews or available in abstract form only (e.g., conference proceedings).

Table 1 Studies identified in the literature review and selection of patient baseline characteristics
Full size table

Patient Populations

The articles identified in this analysis reported either on trials that enrolled patients with nr-axSpA only (9 of 19 articles) or presented subgroup data for patients with nr-axSpA (10 of 19 articles) (Table 1). Key baseline characteristics of the patients included in the studies are shown in Table 1. Consistent with a high proportion of patients with early axSpA, participants in the selected studies had a mean age range of 28–39 years, disease duration of 1–10 years, and Bath Ankylosing Spondylitis Functional Index (BASFI) scores of 3.6–5.5 (Table 1).

Effect of TNFα Inhibitors on Inflammatory Lesions

The majority of selected articles included information on the effect of TNFα-inhibitor therapy on MRI-evident inflammatory lesions in the SIJ (15/19 articles) or spine (12/19) (Tables 2–6). All four TNFα inhibitors approved for treatment of nr-axSpA demonstrated significant improvements versus comparator therapies in MRI scores for inflammation. Infliximab, although not indicated for nr-axSpA, also showed a positive effect on MRI-assessed inflammation in this patient population.

Adalimumab

Evidence for an effect of adalimumab on inflammatory lesions in the SIJ and spine comes primarily from the ABILITY-1 trial, a randomized, placebo-controlled, phase III study in which patients with active nr-axSpA (n = 185) received adalimumab 40 mg or placebo every 2 weeks (Q2W) for 12 weeks, followed by an open-label extension to week 114 [15]. At week 12, adalimumab therapy was associated with significant reductions in MRI scores for inflammation compared with placebo in both the SIJ (mean change from baseline: − 3.2 vs. − 0.6, P  = 0.003) and spine (− 1.8 vs. − 0.2, P  = 0.001) (Table 2) [15].

Table 2 Adalimumab studies: measures of inflammatory lesions by magnetic resonance imaging
Full size table

A post hoc analysis of data from two phase II/III clinical trials of adalimumab (D2E7-Early AS, conducted in patients with nr-axSpA only, n = 46) and etanercept (ESTHER, conducted in patients with nr-axSpA or r-axSpA, n = 76) revealed that both TNFα inhibitors were associated with improvements in SIJ inflammation and that adalimumab was particularly effective in patients with early disease (mean improvement score change from baseline: 7.0 [symptom duration < 4 years] vs. 2.7 [symptom duration ≥ 4 years];  P  = 0.04) (Table 2) [16].

Certolizumab

In the randomized, placebo-controlled, phase III RAPID-axSpA study, patients with axSpA (imaging set, n = 163; n = 68 with nr-axSpA) received certolizumab (200 mg Q2W or 400 mg Q4W) or placebo for 24 weeks [17]; active treatment continued as dose-blinded to week 48 and as open-label to week 204 [18].

A pre-specified subanalysis of pooled-dose MRI data over 96 weeks demonstrated that patients treated with certolizumab achieved greater mean reductions in MRI inflammation scores from baseline to week 12 than did placebo-treated patients in both the SIJ (− 4.4 vs. 1.2; P  < 0.001) and spine (− 2.0 vs. 0.3; P  = 0.006) (Table 3) [18]. These improvements in inflammation were maintained through weeks 48, 96, and 204 [19] for all patients who received certolizumab, including those originally randomized to placebo (Table 3).

Table 3 Certolizumab study: measures of inflammatory lesions by magnetic resonance imaging
Full size table

Etanercept

Two major clinical trials of etanercept—the phase II ESTHER [20] and phase III EMBARK [21] trials—were conducted in patients with axSpA. ESTHER included both patients with r-axSpA (51% [39/76]) and those with nr-axSpA (49% [37/76]) [20], while EMBARK was conducted only in patients with nr-axSpA (n = 215) [21].

In ESTHER [20], patients with active axSpA refractory to non-steroidal anti-inflammatory drug (NSAID) treatment, symptom duration < 5 years, and MRI evidence of inflammatory lesions were randomized to twice-weekly etanercept 25 mg or daily sulfasalazine 2–3 g, both for 48 weeks, followed by a long-term, open-label treatment. The primary analysis of inflammatory lesions in the SIJ and spine demonstrated that etanercept treatment resulted in significant reductions in MRI inflammation scores at 24 and 48 weeks compared with sulfasalazine treatment (Table 4) [20]. For the primary endpoint at 48 weeks, the mean change in MRI inflammation scores in the SIJ was − 5.4 (etanercept) and − 1.9 (sulfasalazine) (P  = 0.02); the mean changes for spinal inflammation were − 1.3 and − 0.1 (P  = 0.01) (Table 4) [20]. Similar reductions in MRI inflammation scores in the SIJ and spine were observed over 3 years (156 weeks) of continuous etanercept treatment, in both a last observation carried forward (LOCF) analysis of patients with nr-axSpA (n = 30) [22] and an additional analysis of patients with axSpA (n = 41) (Table 4) [23].

Table 4 Etanercept studies: measures of inflammatory and structural lesions by magnetic resonance imaging
Full size table

In EMBARK, patients with active nr-axSpA who demonstrated an inadequate response to NSAID therapy and had symptom duration of between 3 months and 5 years were randomized to receive weekly doses of etanercept 50 mg or placebo on a background of NSAID treatment for 12 weeks, followed by a 92-week open-label period of etanercept therapy [21]. Etanercept treatment was associated with significant reductions in MRI-evident inflammation in the axial skeleton: mean changes in inflammation scores were − 3.8 (etanercept) and − 0.8 (placebo) (P  < 0.001) for the SIJ and − 2.1 and − 1.2, respectively (P  = 0.041) for the spine (Table 4) [21]. Notable improvements in inflammation scores in patients randomized to etanercept were sustained during open-label treatment to 48 and 104 weeks (Table 4) [24, 25]. An analysis of a subset of patients from Latin America, Central Europe, and Asia (n = 117) also found that etanercept therapy was associated with a significant improvement in the inflammation score versus placebo in the SIJ (− 3.2 vs. − 0.4; P  = 0.001), but not in the spine, despite a numerical difference favoring active treatment (− 2.2 vs. − 1.4; P  = 0.223) (Table 4) [26].

Golimumab

The effects of golimumab on MRI-evident inflammation in the SIJ of patients with nr-axSpA were assessed in a single phase III, randomized, placebo-controlled GO-AHEAD trial (n = 198); measures of spinal inflammation were not reported [27]. Treatment with golimumab 50 mg Q4W over 16 weeks was associated with significant reductions in SIJ inflammation versus placebo: − 5.3 vs. − 1.0, respectively (P  = 0.001) (Table 5) [27]. The overall improvement in SIJ scores was largely driven by patients with evidence of sacroiliitis on MRI and/or an elevated CRP level at baseline [27].

Table 5 Golimumab study: measures of inflammatory lesions by magnetic resonance imaging
Full size table

Infliximab

Infliximab is not indicated for the treatment of nr-axSpA, but the available data show that it has a positive effect on MRI-evident inflammatory lesions, particularly in the SIJ (Table 6).

Table 6 Infliximab studies: measures of inflammatory and structural lesions by magnetic resonance imaging
Full size table

In a study by Barkham et al., in which patients with early sacroiliitis (n = 40; ~ 88% with nr-axSpA) were randomized to receive infliximab 5 mg/kg body weight or placebo over 16 weeks, infliximab-treated patients had a median change from baseline in the SIJ MRI score of − 2.00, compared with no change in the placebo group (P  = 0.033) (Table 6) [28]. Moreover, significantly more lesions were resolved in patients who received infliximab (P  < 0.001), whereas significantly more new lesions developed in placebo-treated patients (P  = 0.004) [28].

The INFAST study was a randomized, double-blind, placebo-controlled trial of patients with MRI-evident r-axSpA (60%) or nr-axSpA (40%) and disease duration of ≤ 3 years [29]. Patients received intravenously administered infliximab 5 mg/kg body weight + naproxen 1000 mg/day (n = 106) or intravenously administered placebo + naproxen 1000 mg/day (n = 52) over 28 weeks; a total of 156 patients with available MRI data from at least one time point were included in the analysis (Table 6). Significant improvements in MRI inflammation scores were observed in the SIJ and spine in both treatment groups, but these were more notable in patients treated with infliximab (SIJ: − 4.3 vs. − 3.9, P  = 0.003; spine: − 2.9 vs. − 2.0, P  < 0.001) (Table 6) [30]. A post hoc analysis of INFAST data [31], with patients stratified on the basis of fulfilment of the modified New York criteria for AS [3], found that the effect of adding infliximab to NSAID therapy on MRI inflammation scores was greater in patients with AS than in those with nr-axSpA (data not shown). However, the latter also experienced reduction of active inflammation, most notably in the SIJ (Table 6). The apparent lack of treatment effect in the spine of patients with nr-axSpA was possibly due to low baseline levels of spinal inflammation in this subgroup.

Effect of TNFα Inhibitors on Structural Lesions

Several articles identified in this analysis included information on the effect of TNFα-inhibitor therapy on MRI-evident structural lesions in the SIJ (6/19 articles) or spine (3/19 articles) (Tables 4, 6). These studies have provided mixed results regarding the benefit of TNFα-inhibitor therapy on structural progression in axSpA. A small (n = 56), single-center retrospective study suggests that long-term treatment with a TNFα inhibitor may slow progression of structural lesions in patients with AS [32].

Etanercept

In the ESTHER trial, treatment with etanercept was associated with a significantly higher increase in MRI fatty lesion scores compared with sulfasalazine therapy in both the SIJ and spine at 24 and 48 weeks (Table 4) [33]. Increases in fatty lesion scores at 48 weeks were 0.8 and 0.2, respectively, in the SIJ (P  = 0.001), and 0.8 and 0.1 in the spine (P  = 0.020) (Table 4). After 1 year, active suppression of inflammation was strongly associated with the appearance of fatty lesions, which may be the first sign of chronic damage in the bone after prior inflammation [33]. Analysis of the long-term ESTHER data found a small increase in fatty lesion scores from baseline to month 24 (which was significant only for the spine; P  = 0.025), but no further increases in fatty lesion scores were observed during the third year of etanercept therapy (Table 4) [34]. New fatty lesion formation was primarily observed in those areas where active inflammation was present at baseline [34]. Notably, no changes in erosion or ankylosis scores—indicative of more chronic structural changes—were observed during the entire follow-up period [33, 34].

Similar observations of an increase in fatty lesion formation were noted in the EMBARK trial, but not before week 48 [24]. At week 12, there were significant differences between etanercept- and placebo-treated patients in the reduction of erosion (− 0.57 vs. − 0.08, respectively; P  = 0.017) and increase in backfill (0.36 vs. 0.06; P  = 0.022) at the SIJ, but not in the changes in fat metaplasia (0.06 vs. 0.05) (Table 4) [35]. In addition, changes in fat metaplasia at week 12 did not correlate significantly with the changes in SIJ inflammation (data not shown). At week 48, the mean SIJ structural lesion scores for fat metaplasia and backfill increased by 0.46 and 0.89, respectively (Table 4); the ankylosis score increased by 0.04, and the erosion score decreased by − 1.29 [24].

Finally, radiographic changes on the SIJ after 104 weeks of etanercept treatment in the EMBARK trial were compared with those from participants in a contemporary control cohort (DESIR) [36] who met the ASAS criteria for axSpA and who did not receive any biologic treatment for the first 2 years of follow-up [37]. At week 104, patients from EMBARK (n = 154) had an adjusted least-squares mean total SIJ score improvement of − 0.14, while their DESIR counterparts (n = 182) experienced an overall worsening of 0.08 (P  = 0.008) (Table 4). (The total SIJ score was calculated by adding up structural damage scores for both SIJs, using the modified New York grading system [5]). In addition, the net difference in the proportion of patients who experienced improvement versus worsening significantly favored etanercept-treated (EMBARK) over biologic-naïve (DESIR) patients on two out of three radiographic assessment criteria (Table 4) [37].

Infliximab

The effect of infliximab therapy on fatty lesion formation was also investigated in the INFAST trial [30]. As observed with short-term etanercept therapy, increases in fatty lesion MRI scores in the SIJ and spine were observed in both the infliximab + NSAID group and the placebo + NSAID group after 28 weeks, with no significant difference in treatment effect at either site (Table 6).

Discussion and Conclusions

A number of RCTs have shown that TNFα inhibitors reduce MRI-evident inflammatory lesions in the SIJ and spine of patients with early axSpA. Although studies reported improvements in MRI-evident inflammation primarily over the short to medium term (12 weeks to 1 year), reductions in inflammation were maintained for up to 4 years (204 weeks) with certolizumab therapy and for up to 3 years (156 weeks) with etanercept therapy. Little data are available on the effect of TNFα inhibitors on structural lesions, but the EMBARK etanercept trial indicates an improvement with up to 2 years of treatment, compared with a no-treatment cohort from another trial. In addition, increased fatty lesion formation following the resolution of inflammatory lesions with etanercept therapy appears to be transient in nature, with no associated change in joint erosion or ankylosis over the longer term.

Effective anti-inflammatory treatment of axSpA may be associated with an apparent increase in fatty lesion scores, irrespective of the presence or absence of concomitant therapy with a TNFα inhibitor. Fatty lesion formation may represent an important long-term parameter for assessing the effect of early suppression of joint inflammation on more chronic, structural bone changes, such as erosion and ankylosis. However, the specificity of fatty lesions in patients with axSpA needs to be investigated further. A 2012 ASAS/OMERACT (Outcome Measures in Rheumatology) consensus statement based on a systematic literature review suggested that the presence of several corner fatty lesions may indicate axSpA, especially in younger patients, but the authors cautioned that prospective studies in patients aged < 45 years would be needed to strengthen the evidence [8]. In one such study, conducted at two clinical centers, the presence of ≥ 3 corner inflammatory lesions and ≥ 6 corner fatty lesions did not help distinguish between patients with axSpA and those with nonspecific back pain, despite the mean age of all cohorts being < 40 years [38]. In conclusion, additional studies are required to determine the exact role of fatty lesions in axSpA progression and whether TNFα-inhibitor therapy can limit or delay radiological progression in patients with early axSpA.