Introduction

Ankylosing spondylitis (AS) is a chronic inflammatory disease, which is accompanied by enthesis and inflammation of the spinal and sacroiliac joints, and eventually induces bone and joint erosion. Additionally, it progressively contributes to new bone growth, syndesmophytes, and ankylosis, thereby resulting in increased structural damage, fatigue, and decreased quality of life (QOL) [1,2,3]. Non-steroidal anti-inflammatory drugs (NSAIDs) are prescribed as first-line pharmacological treatment for AS. Biological disease-modifying antirheumatic drugs (DMARDs), i.e., anti-tumor necrosis factor (TNF) therapies and secukinumab, an interleukin-17A (IL-17A) inhibitor, are prescribed for treating AS patients who do not respond well to NSAIDs [4]. However, the absence or failure of response to current treatments remains a major concern in some patients. In about 40% of patients with AS, anti-TNF treatment either fails to achieve sufficient disease management or induces undesirable side effects [5]. There is also an unmet requirement for elucidating alternate mechanisms of action to effectively manage AS treatment.

Several cytokines, including those involved in the IL-23/IL-17 axis, signal through the Janus kinase (JAK) family of tyrosine kinases. The JAK pathway is thus a potential therapeutic target in AS. Tofacitinib, an oral inhibitor of JAK [6], selectively inhibits JAK1, JAK2, and JAK3, and exhibits a specificity for JAK1 and JAK3 over JAK2. Upadacitinib was developed to confer higher selectivity for JAK1 than for JAK2, JAK3, and Tyk2 [7]. Likewise, filgotinib, an inhibitor of JAK1, was developed to confer higher selectivity for JAK1 than for others.

Several clinical studies have been conducted to determine the effectiveness and safety of JAK inhibitors in active AS patients exhibiting insufficient response or intolerance toward two or more NSAIDs [8,9,10]. In this study, we aimed to improve the precision and accuracy of the effectiveness and safety estimates of JAK inhibitors in active AS patients exhibiting an inadequate response or intolerance to two or more NSAIDs, through meta-analysis [11,12,13] of the results of randomized clinical trials (RCTs). The findings of such a strategy would encourage the regular evaluation of the publicly available data.

Materials and methods

Identification of eligible studies and data extraction

We conducted an extensive search for RCTs investigating the use of JAK inhibitors for the treatment of AS. We initially performed a literature search using MEDLINE, EMBASE, and the Cochrane Controlled Trials Registry to identify eligible publications (until October 2020) to be included in this study. To perform the searches, the following search terms were used: “ankylosing spondylitis” and “JAK inhibitor.” All references in the selected research articles were further verified to find relevant studies that were not included in the online repositories. The following criteria were used for selecting the RCTs: (1) comparison of JAK inhibitors with placebo for the treatment of active AS patients exhibiting inadequate response or intolerance to two or more NSAIDs, and (2) reporting of the clinical effectiveness and safety endpoints of JAK inhibitors with placebo at 12–14 weeks. The following criteria were used for exclusion: (1) duplicate data and (2) lack of data needed for inclusion. The efficacy endpoints included the following: Assessment of SpondyloArthritis International Society 20% improvement (ASAS20) response rate; ASAS40 response rate; ASAS5/6; ASAS partial remission; Ankylosing Spondylitis Disease Activity Score (ASDAS) with C‑reactive protein (CRP) major response (improvement ≥2.0) and clinical improvement (≥1.1); proportion of patients with clinically relevant improvement (decrease of ASDAS from baseline ≥1 · 1), major improvement (decrease of ASDAS from baseline ≥2 · 0), or inactive disease (ASDAS <1 · 3), ASDAS low disease activity (defined as less than 2 · 1); BASDAI50; Bath AS Functional Index (BASFI); change from baseline in Spondyloarthritis Research Consortium of Canada (SPARCC) score of SI joints and spine (six most severely affected discovertebral units). The safety outcomes were analyzed based on the following parameters: the number of patients experiencing adverse events (AEs), the number of patients experiencing serious adverse events (SAEs), and the number of withdrawn patients owing to adverse events. The following details were collected from each report: first author, year of publication, dosage of JAK inhibitor, number of patients treated with JAK inhibitor with placebo, and safety and efficacy results at 12–14 weeks post JAK inhibitor administration. We assessed the methodological quality of the selected studies using the Jadad score [14]. The Jadad score measures random assignment, blinding, and patient withdrawal and dropout rates, and it varies from zero to five. Quality was categorized as high (a score of 3–5) or low (a score of 0–2). We performed a meta-analysis in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines [15].

Evaluation of statistical associations

The effect size of the study outcomes was represented as odds ratio (OR) for dichotomous data or standardized mean difference (SMD) for continuous data and the corresponding 95% confidence intervals (95% CIs). We tested the differences and heterogeneities within and between the samples using Cochran’s Q‑statistics [16]. The heterogeneity method was applied to measure the null hypothesis, which stated that both experiments measured the same effect. If relevant Q‑statistics (p < 0.10) indicated variance among the analyses, the random impact model was used to perform the meta-analysis; otherwise, the fixed impact model was used. The fixed impact model suggests that all experiments estimate the same underlying effect and recognize only differences within the sample. We quantified the effect of heterogeneity using the following equation: I2 = 100% × (Q − df) /Q [17], where I2 assesses the level of inconsistency between the studies and determines whether the percentage of the total variation across the studies is due to the heterogeneity rather than by chance. I2 varies from 1% to 100%; however, I2 values of 25%, 50%, and 75% are referred to as low, moderate, and high values, respectively. Statistical analyses were performed using the Applied Meta-Analysis Software System (Biosta, Englewood, NJ, USA).

Evaluation of publication bias

Funnel plots are usually generated to detect bias in publications. However, as funnel plots require a large number of studies with varying sizes and individual decisions, we assessed publication bias using Egger’s linear regression test [18], which tests funnel plot asymmetry using a normal logarithm OR scale.

Results

Studies selected to perform the meta-analysis

Initially, 370 articles were identified using online and manual searches, of which 10 were chosen for full-text review based on their title and abstract. However, 7 of the 10 articles were excluded as they contained redundant data, non-RCT data, or no outcome data. Therefore, only three RCTs met the inclusion criteria for performing the meta-analysis [8,9,10]. These three selected studies included a total of 406 patients (203 experimental subjects and 203 controls), and included the investigation of upadacitinib, filgotinib, and tofacitinib. The JAK inhibitors were used at the following dosages: upadacitinib 15 mg once daily, filgotinib 200 mg once daily, and tofacitinib 5 mg twice daily. The Jadad score across all the studies ranged from 3 to 4, thereby indicating high quality. All patients underwent standard therapy, and the related aspects of the trials used in the meta-analysis are listed in Table 1.

Table 1 Characteristics of the individual studies included in the meta-analysis
Full size table

Meta-analysis to assess the efficacy of JAK inhibitor for treating AS

The ASAS20 response was significantly higher in the JAK inhibitor group than in the placebo group (OR = 3.762, 95% CI 2.474–5.721, p < 0.001; Table 2; Fig. 1). ASAS40 and ASAS5/6 responses were also substantially higher in the JAK inhibitor group than in the placebo group (Table 2; Fig. 1). Other efficacy outcomes in the JAK inhibitor group, such as ASAS partial remission, BASDAI50, ASDAS, SPARCC MRI scores, and BASFI, were also significantly higher in the placebo group (Table 2; Fig. 1).

Table 2 Meta-analysis of randomized controlled trials of JAK inhibitors in AS
Full size table
Fig. 1
figure 1

Meta-analysis of the a ASAS20, b ASAS40, and c ASDAS clinically important improvement response rates for JAK inhibitors versus placebo treatments in patients with AS. ASAS20 Assessment of SpondyloArthritis International Society 20% improvement, ASAS40 Assessment of SpondyloArthritis International Society 40% improvement, ASDAS Ankylosing Spondylitis Disease Activity Score, JAK Janus kinase

Full size image

Meta-analysis to assess the safety of JAK inhibitors for treating AS

The incidence of AEs did not differ between the JAK inhibitor and placebo groups (Table 2; Fig. 2). Additionally, no differences were identified in SAEs and withdrawal owing to the difference in AEs between the JAK inhibitor and placebo groups (Table 2; Fig. 2). There was a non-serious deep vein thrombosis in the calf of a man aged 53 years who had a heterozygous factor V Leiden mutation, diagnosed 3 days after the patient’s last dose of filgotinib. However, there was no further thromboembolic event in upadacitinib and tofacitinib RCTs. Herpes zoster infection was not reported in any of the three RCTs.

Fig. 2
figure 2

Meta-analysis of a adverse events (AEs), b serious adverse events (SAEs), and c withdrawal attributed to AEs for JAK inhibitors versus placebo treatments in patients with AS

Full size image

Heterogeneity and publication bias

Heterogeneity between the studies was not found in most of the meta-analyses assessing the safety and efficacy of JAK inhibitors, excluding the ASDAS major response, SPARCC spine score, SPARCC SI joint score, and BASFI. The cause of heterogeneity was identified to be the difference in the magnitude of the effect size, and not its direction. It was difficult to compare the funnel map, which is typically used to identify the reporting bias, as the number of studies included was very limited. However, no evidence of publication bias was detected (Egger’s regression test p-values >0.1).

Discussion

In this meta-analysis, we extracted data from three RCTs that compared JAK inhibitors to placebo in AS. All efficacy results, including ASAS20, ASAS40, and ASAS5/6 responses, were substantially higher in the JAK inhibitor group than in the placebo group. In comparison, there was little disparity between the frequency of AEs and SAEs and the AE withdrawal rate between the JAK inhibitor and the placebo groups. JAK inhibitors reached the endpoints demonstrating substantial improvements in disease activity (ASAS partial remission, BASDAI50, ASDAS), work (BASFI), and MRI (SPARCC MRI spine) outcomes. JAK inhibitors decreased disease activity and symptoms and signs more significantly than placebo in patients with active AS who did not respond to NSAIDs, and were well tolerated. Two studies are phase II studies compared to one phase II/III study. Tofacitinib and filgotinib were evaluated in phase II studies, while upadacitinib was assessed in a phase II/III study, which is ongoing to evaluate long-term safety and efficacy of upadacitinib treatment in AS. These data with JAK inhibitors in AS are very promising and indicate that JAK inhibitors could be used as a potential therapeutic alternative for AS.

Currently, the therapeutic regimes for patients with AS who do not respond well to NSAIDs are restricted to TNF inhibitors and secukinumab [4]. AS patients with allergic responses to IL-17A and/or TNF inhibitor cannot undergo these treatments [19] and the use of IL-17 antagonists is not recommended in patients with active inflammatory bowel disease [20]. Considering these unmet needs, our data suggest that JAK inhibitors can serve as an effective and safe treatment regimen for patients with active AS who do not respond well to NSAIDs; these findings also encourage further study on the use of JAK inhibitors for managing AS.

The JAK-STAT cascade regulates the proliferation and cytokine network associated with different T‑cell subpopulations, such as Th17 cells and the IL-23/IL-17 cytokine axis [21]. As the IL-23/IL-17 cytokine axis plays a vital role in the pathogenesis of AS, JAK inhibitors are predicted to have therapeutic potential for AS. Inhibition of TNF‑α has been shown to be an efficient strategy in the management of AS. However, the IL-23/IL-17 cytokine is not specifically inhibited by JAK inhibitors, but it has been shown that the selective inhibition of JAK can result in secondary inhibition of additional pathways that do not rely on JAK1 signaling [22].

There are certain limitations to be considered in the present meta-analysis. First, the number of experiments involved was limited and the possibilities for sampling mistakes and publishing prejudices cannot be excluded. While we did not find publication bias, it should be noted that publishing bias is difficult to eliminate with full confidence, particularly when the number of studies involved is limited, and only three RCTs were included in our analysis. Second, long-term findings have not been included in this meta-analysis. The follow-up time of the included trials was 14 weeks, and therefore, follow-up studies conducted for a longer duration are expected to be included in the future analyses. Third, variation in clinical characteristics, such as ethnicity, sex, age, and AS severity, complicates the results of meta-analysis and could have influenced our analysis.

Conversely, this meta-analysis study also has several strengths, as it is up to date and included all the existing information. The number of AS patients in each sample varied from 103 to 187, but the combined analysis comprised 406 patients. Compared to the individual research papers, we provided more reliable evidence by increasing the statistical power and resolution through combining the findings of independent analyzes [23]. This is—to the best of our knowledge—the first meta-analysis assessing the efficacy and safety of JAK inhibitors in treating patients with active AS despite failure of NSAID treatment. Our meta-analysis has provided detailed reliable evidence across all accessible RCTs regarding the use of JAK inhibitors in AS with statistical significance and addresses the variability of the outcomes of independent analyzes across current literature [24,25,26]. Therefore, these might be the best available data in this field [24].

Conclusively, upon performing this meta-analysis, we found that treatment with JAK inhibitor was successful in patients with active AS who demonstrated an insufficient response or intolerance toward two or more NSAIDs and JAK inhibitor therapy. While the long-term efficacy and safety of JAK inhibitors needs to be assessed, they were found to exhibit potential therapeutic effects against AS. Further long-term research is required in this field to better assess the safety and efficacy of JAK inhibitors for treating AS.