Introduction

Ankylosing spondylitis (AS), a disease characterized by inflammation of the axial skeleton, peripheral joints, and entheses can cause considerable disability and pain. After approval of the first tumor necrosis factor (TNF) inhibitor infliximab, for the treatment of patients with AS, the treatment of this chronic inflammatory disease has changed remarkably. Recent meta-analyses of randomized controlled trials (RCTs) [1, 2] demonstrate that the efficacy of TNF inhibitors show no detectable difference in AS. So, their safety profile is likely to be an important determinant for decision making in AS care. Since a meta-analysis published on this subject described the safety of anti-TNF agents, this study shows the safety outcomes and withdrawals did not indicate statistically significant differences between treatment and control groups, but the search of studies was completed in 2012 [3]. The most recent meta-analysis [4] shows that serious adverse events and all-cause withdrawals did not indicate statistically significant differences between treatment and control groups, but the meta-analysis did not report the risk of malignancy and the risk of serious infection events. Recently, more and more RCTs [5,6,7] focused on malignancy risk and serious infection events in patients with AS not exposed and exposed to TNF inhibitors. Continuous updates can aggregate data from new studies, as well as provide more robust information for physicians to determine the most appropriate therapies. So we aim to comparatively update the key relevant safety profiles (i.e., overall serious adverse events, malignancy, serious infection, and discontinuation due to adverse events) of TNF inhibitors (i.e., adalimumab, certolizumab pegol, etanercept, golimumab, or infliximab) in AS patients.

Methods

Data Sources and Searches

Study selection, assessment of eligibility criteria, data extraction, and statistical analysis were performed based on a pre-defined, peer-reviewed protocol according to the Cochrane Collaboration guidelines (http://www.cochrane.org/resources/handbook/index.html). We undertook a systematic literature search including RCTs that selected adult patients with AS. We searched studies using MEDLINE via OVID and PubMed, the Cochrane databases, Google Scholar, Clinical trials.gov, and manual searches of reference lists from systematic reviews and original publications and identified studies published in English from 1990 to May 31, 2016. PubMed Auto Alerts was set up to provide weekly updates of new literature until May 31, 2016. The search terms included ankylosing spondylitis; etanercept; infliximab; adalimumab; certolizumab pegol; golimumab; randomized controlled trial; adverse effects; infection; malignancy; all adult.

Study Selection

We pre-specified the target population, interventions, comparators, outcome measures of interest, timing, and settings (PICOTS) following the PICOTS framework [8]. To be eligible, RCTs had to (1) compare the safety of any of the TNF inhibitor is against placebo and/or traditional disease-modifying anti-rheumatic drugs (DMARDs); (2) include only patients with AS; (3) report a minimum of 12 weeks of the study duration. We excluded studies that were open-labeled. We defined the target population as 18 years of age or older, with adult patients diagnosed with active AS, as defined by the modified NewYork criteria [9], were included in this analysis. Eligible interventions included all five currently available TNF inhibitors (adalimumab, certolizumab pegol, etanercept, golimumab, and infliximab). Eligible comparators included placebo or traditional DMARDs. Eligible outcomes included (1) serious adverse events, which was any adverse event that resulted in death, was life threatening, resulted in hospitalization or prolongation of hospitalization, or caused persistent or substantial disability [10];( 2) serious infection, defined as an infection that requires antimicrobial therapy or hospitalization [11]; (3) malignancies, defined as each RCT specified [12]; (4) treatment discontinuation due to adverse events. Two investigators (Li-qiong Hou and Ga-xue Jiang) independently determined the eligibility of the studies and discrepancy was resolved by consensus.

Data Extraction and Study Quality Assessment

Two investigators (Lei Meng and Miao Xue) independently extracted baseline patient characteristics, drug doses and treatment duration, the number of subjects experiencing an event by outcomes in randomized groups, and the number of randomized patients for intention-to-treat analysis. Discrepancy was resolved by consensus. We assessed the quality of evidence using the Grading of Recommendations Assessment, Development and Evaluation (GRADE) system for grading evidence [13,14,15]. The ranks of the quality of evidence were based on the type of study design, the risk of bias in the body of evidence, the consistency of the results, and the precision of the overall estimate. For each outcome, the strength of the evidence was rated as high, moderate, low, or very low. We examined risk of bias in individual studies using the criteria from the Cochrane risk-of-bias tool [16]. To appraise the risk of bias, we used pre-defined criteria, which included random sequence generation, allocation concealment, masking of the treatment status, masking of outcome assessment, selective outcome reporting, and intention-to-treat principles. Consistency was obtained based on the pooled statistical heterogeneity at the significant level α = 0.1.

Data Synthesis and Statistical Analysis

We compared each TNF inhibitors with placebo alone or in combination with traditional DMARDs. In addition, both odds ratios (ORs) and their 95% confidence intervals (CIs) were calculated. It is recommended that studies with zero events in both arms be excluded from meta-analyses of ORs [17]. To address the potential dose impact, we also compared the event rates according to TNF inhibitors dose (i.e., high dose vs normal dose), similar to a recent systematic review by Aaltonen et al. [18]. For this analysis, high dose referred to a higher than normal TNF inhibitors dose as per package insert of each TNF inhibitors as follows: infliximab 3 mg/kg/every 8 weeks, etanercept 50 mg/week (or 25 mg/twice per week), adalimumab 40 mg/every other week (or 20 mg/week), certolizumab pegol 400 mg/every 4 weeks, and golimumab 50 mg/every 4 weeks. Finally, to explore the impact of the study duration, a meta-analysis was performed on two subgroups of studies (risk difference by Mantel-Haenszel’s method), according to the study duration (< vs ≥24 weeks). We explored heterogeneity between the trials using the chi-square test for heterogeneity and a 10% level of significance. In addition, the I 2 statistic was calculated from the results of the meta-analysis as I 2 = 100% × (Q − df)/Q, where Q is Cochran’s heterogeneity statistic (chi-square) and df is the degrees of freedom, to quantify inconsistencies across studies, and results were complied with the recommendations put forward in the Cochrane Handbook [19]. A value of 0% indicates no observed heterogeneity, and larger values show increasing heterogeneity [20]. The study was considered heterogeneous when a p value of chi-square statistics was less than 0.1 or the value of I 2 was over 50% [21]. All statistical tests and creation of forest plots were conducted with STATA 12 and Meta-Analyst software [22].

Results

Search Results

The database search results are summarized in Fig. 1. Among the 28 full-text articles that were assessed for eligibility, 14 RCTs [5,6,7, 23,24,25,26,27,28,29,30,31,32,33] that fulfilled all the inclusion criteria and none of the exclusion criteria. The main reasons for the exclusion were no drug of interest, patients without AS, lack of randomization or a control group, and the study duration of <12 weeks. We excluded the study of Braundt J [34] because the study with zero events in both arms, the study Braundt J [35] did not compare the safety of the TNF inhibitor is against control group, the studies Dijkmans B [36] and Inman RD [37] are open-labeled studies (Fig. 1).

Fig. 1
figure 1

Flow chart of search results

Full size image

Study Characteristics

Of the 14 studies, 10 reported total serious adverse events, 3 reported malignancies, 8 reported serious infections, and 10 reported discontinuation due to adverse events. The mean age ranged from 29 to 48 years, 63 to 95% of patients were male, and the disease duration ranged from 3.2 to 16.4 years for patients with AS included in the analysis. From the 14 RCTs included in the systematic review, 2 used adalimumab, 7 etanercept, 2 golimumab, and 3 infliximab for intervention. The included trials have 3062 patients, of which 2032 and 1030 were in the treatment and control groups (Table 1).

Table 1 Characteristic of randomized controlled trials included in the analysis
Full size table

Overall Serious Adverse Events

Nine trials and 11 comparisons provided data about the overall serious adverse events. There were 1992 patients in the TNF inhibitor group. There were 1118 patients in the control group. The risk of overall serious adverse events in patients treated with TNF inhibitors as a group was not significantly different from those treated with placebo and/or traditional DMARDs in the control arm (OR, 1.34; 95% CI, 0.87–2.05) (Fig. 2).

Fig. 2
figure 2

Effect of TNF inhibitors vs control therapy on the occurrence of overall serious adverse events in patients with ankylosing spondylitis

Full size image

Serious Infection Events

Eight trials provided data about the serious infection events. There were 1501 patients in the TNF inhibitor group. There were 707 patients in the control group. The risk of serious infection events in patients treated with TNF inhibitors as a group was not significantly different from those treated with placebo or traditional DMARDs in the control group (OR, 1.59; 95% CI, 0.63–4.01) (Fig. 3).

Fig. 3
figure 3

Effect of TNF inhibitors vs control therapy on the occurrence of overall serious infection events in patients with ankylosing spondylitis

Full size image

Discontinuation Due to Adverse Events

Ten trials and 12 comparisons provided data about the discontinuation due to adverse events. There were 1919 patients in the TNF inhibitor group. There were 1065 patients in the control group. The risk of discontinuation due to adverse events in the TNF inhibitor group was not significantly different from the control group (OR, 1.55; 95% CI, 0.95–2.54) (Fig. 4).

Fig. 4
figure 4

Effect of TNF inhibitors vs control therapy on the occurrence of discontinuation due to adverse events in patients with ankylosing spondylitis

Full size image

Malignancy Events

Three trials and four comparisons provided data about the malignancy events. There were 425 patients in the TNF inhibitor group. There were 302 patients in the control group. The risk of malignancy in the TNF inhibitor group was not significantly different from the control group (OR, 0.98; 95% CI, 0.25–3.85) (Fig. 5).

Fig. 5
figure 5

Effect of TNF inhibitors vs control therapy on the occurrence of malignancy events in patients with ankylosing spondylitis

Full size image

Additional Analyses

To address the potential dose impact, we compared the safety event rates according to TNF inhibitors dose (i.e., high dose vs normal dose). High-dose TNF inhibitors was not significantly associated with an increased risk of overall serious adverse events (OR, 1.43; 95% CI, 0.61–3.38), serious infection (OR, 1.26; 95% CI, 0.28–5.63), or discontinuation due to adverse events (OR, 2.70; 95% CI, 0.70–10.47) as compared with patients treated with a normal dose. Similarly, our subgroup analysis for these safety events according to duration of the trial did not appear to differ substantially (summary ORs, 1.72 vs 1.10 for overall serious adverse events and 4.27 vs 0.33 for serious infection events, respectively, and 1.32 vs 0.96 for discontinuation due to adverse events) (Table 2).

Table 2 The results of the subgroup analyses
Full size table

Quality of Evidence for Our Analyses

The study qualities of the selected trials were diverse: nine trials were classified as high quality (Jadad score ≥ 4) and five trials were classified as moderate quality (Jadad score = 3) (Table 3).

Table 3 Quality of included trials
Full size table

Discussion

Our objective was to systematically update major safety profiles reported in the RCTs of all approved TNF inhibitors to date to inform the field.

Our study shows that the risk of overall serious adverse events in patients treated with TNF inhibitors was not significantly different from those treated with placebo and/or traditional DMARDs in the control arm. This null conclusion has been consistently reported in previous meta-analysis [3], but our meta-analysis included the recently published RCTs [7, 32]. The most recent meta-analysis [3] also shows the same results; the difference is that we excluded the open-labeled study. Open-labeled studies usually lack a control group, consequently establishing causality between a treatment and an event is impossible. Our study shows that there was no significant difference in the risk of serious infection events between TNF inhibitors and placebo. The lack of significant increase in risk of serious infections with TNF inhibitors may be explained by a lack of power. Indeed, more infections were detected in the TNF inhibitors group; it would, however, necessitate a RCT of more than 5000 patients to reach statistical significance with the difference in risks observed. The few serious infections reported in the articles reviewed here were very heterogeneous, both viral and bacterial infections. In 2010, a meta-analysis of nine trials suggested that the serious infections with TNF inhibitors compared with placebo were not significant [38]. Different from the previous study, the studies with zero events in both arms be excluded from our meta-analyses of OR, and we included the latest RCTs. Furthermore, no significant risk difference was found among increased doses of TNF inhibitors-treated patients and patients treated with normal doses in the included trials. A prospective cohort study indicated a high incidence of serious infections with infliximab treatment in AS [39]. However, they had a small population and no comparison with placebo. As for tumor incidence, AS has not been linked with malignancy, unlike rheumatoid arthritis (RA), which is associated with an increased incidence of lymphoma. Two studies assessing solid malignancy rates with TNF inhibitors have failed to show an increased risk compared with the general population [40]. Our findings confirm no increased risk of malignancy associated with TNF inhibitors either as a group or individual in AS. A meta-analysis public in 2009 analysis included 19, 041 patients exposed to adalimumab in 36 global clinical trials in RA, psoriatic arthritis (PsA), AS, Crohn’s disease (CD), psoriasis, and juvenile idiopathic arthritis (JIA) has the same conclusion [41]. A single-center series and systematic review of RCTs of malignancies in patients with RA, PsA, and AS receiving TNF inhibitors have the same conclusion too. Furthermore, findings from observational studies as well as from a large long-term extension study of a RCT have reported no increased risk of malignancy across all TNF inhibitors [42]. However, there were few RCTs report the risk of malignancy event with TNF inhibitors treatment in AS, addressing this issue in future studies would be valuable. Our study shows that the risk of overall serious adverse events in patients treated with TNF inhibitors was not significantly different from those treated with placebo and/or traditional DMARDs in the control arm. This conclusion has been consistently reported in the previous meta-analysis [3]. A recently meta-analyses report the risk of all cause withdrawals in patients treated with TNF inhibitors was not significantly different from those treated with placebo in the control arm [4]. It should be noted that the discontinuation due to adverse events rate in our meta-analysis is different from all cause withdrawals rates, which include a lack of efficacy.

The present systematic review has several advantages. This systematic review was performed using all available literature sources and included all published data to date. Furthermore, the meta-analysis was performed in accordance with the recommendations of the Cochrane Collaboration. No evidence for significant publication bias was apparent on a funnel plot (data not shown) and no statistical heterogeneity was detected. Furthermore, several methods of meta-analysis were applied, assessing the risk difference vs the OR. Both methods confirmed the robustness of the results. Several sensitivity analyses were also performed and confirmed the results. Therefore, we consider these results to be valid.

Although we believe that the current meta-analysis provides useful information, some potential limitations should be addressed. First, some safety data are not available in all finally selected articles and thus could not be included in our analysis. Disease severity was not taken into account in our study, partly due to the difficulties associated with the different severity measures across the trials. Third, the study durations were short (from 12 to 30 week). So more trials are needed to investigate the safety of TNF inhibitors in patients treated for AS.

In conclusion, our meta-analysis of RCTs to date indicates TNF inhibitors generally safe for treatment of AS. Our findings should be relevant to clinical comparative effectiveness guidelines on treatment for AS patients.