Results
Process of the Literature Review
Overall, 409 reports were identified in MedLine and 630 reports were identified in the Embase database, while no report was found in the Cochrane database. The search of the abstract meetings revealed 254 reports at EULAR and 202 reports at the ACR meeting. After exclusion of duplicates, 257 reports remained for validation, 64 reports were found to be dealing with efficacy and/or safety outcomes of patients and finally, 25 papers were found to have useful data for analysis (Fig. 1).
Efficacy
Calculation of ESs for Treatment Outcomes The comparison between anti-TNF treatment vs placebo showed superior outcome for the treatment effect in favour of the anti-TNF treatment. For the evaluation of the BASDAI, the ES (95% CI) varied between studies from 0.34 (0.08, 0.6) to 1.5 (0.45, 2.5) ( Table 1 ). For evaluation of the BASFI, the ES (95% CI) varied between 0.27 (−0.08, 0.63) and 2.5 (1.3, 3.7), whereas for evaluation of the BASMI, the ES (95% CI) was only available for golimumab [0.08 (−0.20, 0.31)] ( Table 1 ). Furthermore, data for different other outcomes such as occiput-to-wall measurements, chest expansion, physician's global and patient's global were only available in some of the studies. The treatment effect for patient's global assessment was 0.53 (0.17, 0.89), for physician's global assessment 1.3 (0.67, 1.9), for chest expansion zero (−0.35, 0.35), for occiput-to-wall it varied between −0.22 (−0.52, 0.09) and 0.01 (0.34, 0.37) and for modified Schober's test between 0.06 (−0.29, 0.42) and 0.28 (−0.03, 0.58).
The treatment effect for continuous vs on-demand anti-TNF treatment could only be calculated for infliximab, with an ES (95% CI) of 0.76 (0.44, 1.1) for BASDAI and 0.74 (0.42, 1.1) for BASFI, 0.53 (0.22, 0.84) for patient's global assessment, 0.03 (−0.28, 0.34) for the physical component of the short form 36 (SF-36) questionnaire and 0.19 (−0.12, 0.5) for the mental component of SF-36.
The Guyatt's ES could only be calculated for golimumab in AS and infliximab in non-radiographic SpA.
Calculation of Numbers Needed to Treat The calculation of the NNT for achieving all different treatment outcomes revealed only minor variations between the TNF blockers but superiority as compared with placebo, with NNTs of 2.3–2.7 for ASAS 20, 2.9–3.7 for ASAS 40, 2.4–2.8 for ASAS 5/6, 2.5 for BASDAI 50 and 4.7–5.9 for ASAS partial remission. Similar NNTs were found for patients with non-radiographic axial SpA, with 2.3 for ASAS 20, 1.6–2.4 for ASAS 40, 3.2 for ASAS 5/6 and 2.3–2.7 with ASAS partial remission ( Table 2 ).
In the comparison of continuous vs on-demand treatment with TNF blockers, the NNTs for ASAS 20 response were 4.2 vs 9.1 patients, for ASAS 40 response 6.7 vs 8.3 patients and for ASAS partial remission 5.9 vs 20.0 patients, respectively. For the differentiation between patients with vs without total spinal ankylosis, the NNTs varied between 2.4 for ASAS partial remission and 9.1 for ASAS 5/6 ( Table 3 ).
Efficacy of TNF Blockers on Extraspinal Manifestations of the Disease
One study from patients diagnosed as SpA according to the Amor criteria provided data on the efficacy of TNF blockers in peripheral manifestations of the disease. Patients with refractory disabling heel enthesitis were treated with etanercept or placebo. Patient's global assessment, heel pain and WOMAC improved significantly in the etanercept group as compared with placebo, already after 2 weeks of treatment.
Treatment With Biologics Other Than TNF Blockers
Overall, only small studies on biologics other than TNF blockers were available, and all of these studies included patients with advanced disease. None of the studies was placebo controlled. The compounds used were rituximab, anakinra or abatacept. All of the compounds showed only minor improvement in disease-related indices, and because there are no control groups, the level of improvement is difficult to interpret. For rituximab in anti-TNF naïve patients, there were significant within-group improvements in BASDAI (P = 0.047), pain as reported by the patient (P = 0.021) and improvement in CRP (P = 0.017). Further data published in the full paper of this abstract in 2010 showed a good improvement of all assessed parameters (50% in BASDAI50, 40% in ASAS 40) as compared with a poor response in those patients who had failed TNF blocker therapy before rituximab treatment (10% in ASAS40, none in BASDAI 50). For anakinra, the rate of patients showing sufficient ASAS response was reported as 25% for ASAS 20 and 20% for ASAS 40, while BASDAI improved from 5.8 to 4.6 and there was no change in CRP, as compared with baseline. The data of this study were included in abstract form in the first version of the recommendations, whereas the full paper is now available for the current report. For abatacept, there was only minor response of single patients.
TNF Blockers in Juvenile SpA
Only one small study published in abstract form including patients with juvenile SpA patients with established AS (n = 5 patients) and undifferentiated SpA (n = 19) treated with infliximab could be used for data analysis. In this study, the amount of active joints, tender entheses, pain, CRP and HAQ showed significant decrease after 1 year in all patients. The mean amount of active joints decreased from 4.7 (1.7) to 0, the mean amount of tender entheses from 11.9 (10.7) to 0, the mean CRP from 24.8 (10) to 1.3 (3.1), the pain (mean of NRS) from 7.2 (2.0) to 1.7 (2.7), while the mean score in the childhood HAQ did not show changes in the patients who were initially treated with infliximab and remained on this treatment.
Level of Evidence and Strength of Recommendations for Treatment With Biologics in AS
The overall research evidence for all TNF blockers is rated with 1b+ ( Table 4 ), including two studies with patients with non-radiographic axial SpA, which showed similar outcomes as compared with studies of patients with established AS.
Furthermore, the research evidence for the use of infliximab on demand and for the use of etanercept in a dose of 1 × 25 mg/week in patients with low disease activity was also rated with 1b+. There are no data on dose adjustment for adalimumab at the current time point. The strength of recommendation (SOR) for the use of all available TNF blockers in AS in the recommended dose is rated with A, with the exception of treatment with etanercept 1 × 25/week, where the SOR is rated with B ( Table 4 ).
The research evidence for the treatment of patients with DMARDs concomitant to TNF blockers as well as switching between TNF blockers is 3+, while the SOR was rated with C. Although the analyses for switching between anti-TNF compounds have been based on patients treated with infliximab after failure of treatment with etanercept, it is expected that other combinations among other TNF blockers would reveal similar outcomes.
For treatment with biologics other than TNF blockers, the available data showed a research evidence of 3 for anakinra based on the same study as already included in the previous review; however, this result remains to be confirmed by further studies. Data for abatacept and rituximab are scarce and did not allow for any conclusions, while no data for tocilizumab were available within the period of analysis in this update. The SOR for the use of anakinra in AS was rated with C.
For the use of biologics in patients with juvenile onset of SpA, only data on infliximab were available. The research evidence was 3, which can be translated to SOR rated with C ( Table 4 ).
Incidence of Concomitant Extra-articular Manifestations in AS During Treatment With TNF Blockers
TNF blockers showed beneficial effect on the treatment of extra-articular manifestations (EAMs) of AS as compared with treatment with placebo. Data were available for infliximab, etanercept and adalimumab, while data from studies with golimumab were not available at this time point. Two main concomitant EAMs were recognized: anterior uveitis (AU) and IBDs.
As suggested in a meta-analysis for the treatment of AU, which included only patients with infliximab and etanercept (adalimumab data were not available at this time point), the incident rates during anti-TNF treatment were 3.4 (range 1.1–8.0) per 100 patient-years (pys) as compared with 15.6 (7.8–27.9)/100 pys during placebo treatment (all P < 0.05). In a more recent paper, the incidence of AU flares under open-label adalimumab treatment was 7.4/100 pys and statistically significantly lower than the incidence rate of AU during the previously performed placebo-controlled period of the same trial with 15.0 AU flares/100 pys (P = 0.001). Another follow-up study with patients treated open label with etanercept showed similar superiority of etanercept (11 AU flares/100 pys), as compared with the numbers known from the placebo-controlled period of the same trial. Similar data were shown in a meta-analysis that was available in abstract form (, the full paper was published in 2010). A summary of the studies dealing with the occurrence of AU in patients with anti-TNF during the time period analysed in this update is shown in Table 5 .
For the incidence of IBD, other differences between the TNF blockers were found, with significantly lower incidence rates during treatment with infliximab, as compared with etanercept or adalimumab ( Table 6 ).
Safety
AEs The incidence of AEs between treatment with TNF blocker and placebo, between TNF blockers in different treatment doses or during treatment with TNF blockers with or without concomitant treatment with other compounds is shown in Table 7 . Overall, the incidence of AEs as reported in the present updated review is in line with those reported in the first version of the recommendations.
Infections In a meta-analysis comparing the risk difference between TNF blockers and placebo, the incidence rate of non-serious infections was 84.5 (58.4)/100 pys in patients treated with TNF blockers during the randomized control phases of the trials (RCTs) and reduced to 64.4 (56.7)/100 pys during the open-label phases. The latter was similar to the incidence of non-serious infections registered in the placebo arm of the RCTs, with an incidence of 63.6 (63.0) non-serious infections/100 pys.
In contrast, the analysis of serious infections showed an incidence of 2.3 (4.0) under TNF blockers during the RCTs and of 1.4 (2.8) during the open-label phases, as compared with an incidence of 1.4 (2.83) serious infections under placebo. An overview on the available data of the relative risk for infections in patients with AS is shown in Table 7 .
Formation of Antibodies against TNF blockers
Only a few studies were dealing with the issue of antibody formation during treatment with TNF blockers in AS. In one study with infliximab—patients who discontinued and re-started TNF blockade in the same treatment regimen—immunogenicity had no influence on the response to re-treatment or on safety outcomes in the long-term follow-up. While antibody formation due to immunogenicity was not detected during and after treatment with etanercept, antibody formation correlated well with undetectable serum trough levels, with inefficacy and infusion or injection reactions in patients treated with infliximab or adalimumab in two small studies.