Gratacós et al. Arthritis Research & Therapy (2019) 21:11 https://doi.org/10.1186/s13075-018-1772-z RESEARCH Open Access Non-inferiority of dose reduction versus standard dosing of TNF-inhibitors in axial spondyloarthritis Jordi Gratacós1, Caridad Pontes2* , Xavier Juanola3, Jesús Sanz4, Ferran Torres5, Cristina Avendaño6, Antoni Vallano7, Gonzalo Calvo8, Eugenio de Miguel9, Raimon Sanmartí10 and The REDES-TNF investigators Abstract Objective: The objective was to determine if dose reduction is non-inferior to full-dose TNFi to maintain low disease activity (LDA) in patients already in remission with TNFi, in axial spondyloarthritis. Methods: Randomized, parallel, non-inferiority, open-label multicentre clinical trial. Patients were eligible if they had axial spondyloarthritis and had been in clinical remission for ≥ 6 months with any available TNFi (adalimumab, etanercept, infliximab, golimumab) at the dose recommended by product labelling. Patients were randomized by automated central allocation to continue the same TNFi dose schedule, or to reduce the dose by roughly half according to the protocol. The main outcome was the proportion of subjects with LDA after 1 year. Serious adverse reactions or infections were recorded. Results: The trial stopped due to end of the funding period, after 126 patients were randomized; 113 patients (84.1% male, mean age (SD) 45.6 (13.0) years) were included in the main per-protocol subset. Non-inferiority was concluded for LDA at 1 year (47/55 (83.8%) patients in the full-dose and 48/58 (81.3%) patients in the reduced-dose arm, adjusted difference (95% CI) − 2.5% (− 16.6% to 11.7%)). Serious adverse reactions or infections were reported in 7/62 patients (11.3%) assigned to full dose and 2/61 patients (3.3%) assigned to reduced dose (p value = 0.164). Conclusion: In patients with ankylosing spondylitis in clinical remission for at least 6 months, dose reduction is non-inferior to full TNF inhibitor doses to maintain LDA after 1 year. Serious adverse events may be less frequent with reduced doses. Trial registration: EU Clinical Trials Registry, EudraCT 2011–005871-18 and ClinicalTrials.gov, NCT01604629. Keywords: Spondyloarthritis, Dose-tapering, Non-inferiority, TNF inhibitors Introduction achieve and maintain clinical remission or low disease Spondyloarthritis (SpA), a group of rheumatic diseases activity (LDA) [5]. The first step is non-steroidal that share immunogenic, clinical and radiological char- anti-inflammatory drugs (NSAIDs) and physical therapy, acteristics, are classified as axial (axSpA) or peripheral although 60% of patients obtain no clinical benefit [6]. by the Assessment of Spondyloarthritis International So- Disease-modifying anti-rheumatic drugs (DMARDs) are ciety (ASAS). Ankylosing spondylitis (AS) is the main not deemed efficacious, but tumour necrosis factor in- disease within SpA, with a prevalence of 0.4–0.9% in hibitors (TNFi) can induce and sustain long-term clin- Caucasians, and marked differences by race, prevalence ical remission in patients not responding to NSAIDs [7]. of HLA B27, and geographical area [1, 2]. Despite treat- However, the optimal duration and dose for long-term ment, many Patients with SpA have impaired quality of treatment remain unclear, and the recommended TNFi life and work disability [3, 4]. SpA management aims to doses are the same for treatment of acute and chronic disease. Although the risk-benefit is favourable, pro- * Correspondence: caridad.pontes@uab.cat longed treatment with TNFi chronically suppresses the 2 Clinical Pharmacology Unit, Hospital de Sabadell, Institut Universitari Parc immune response, increasing the risk of serious infection Taulí, Universitat Autònoma de Barcelona, c/Taulí n°1, 08208 Sabadell, dose-dependently, especially in frail patients. It is un- Barcelona, Spain Full list of author information is available at the end of the article clear whether chronic use may increase malignancies © The Author(s). 2019 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Gratacós et al. Arthritis Research & Therapy (2019) 21:11 Page 2 of 10 and neurological demyelinating diseases [8–10]. Finally, Randomization and treatments TNFi are expensive and a cost burden. Eligible patients were randomized to receive either the Discontinuation of TNFi after achieving clinical remis- full recommended TNFi dose or a reduced dose accord- sion is followed by early relapse in most cases [11–13]. ing to an agreed protocol supported by clinical practice However, lower doses of TNFi may be sufficient in pa- (Table 1). Patients were screened, and information on tients with LDA and low levels of inflammatory media- previous TNFi treatment, clinical activity and other eligi- tors, according to uncontrolled or observational studies bility criteria entered in the electronic case report form [14, 15]. Some clinical guidelines recommend empirical (eCRF). Patients were automatically randomized by cen- dose reductions despite a lack of robust supportive evi- tral allocation according to a list generated using SAS dence [16]. Likewise, although drug-level monitoring PROC PLAN v9.2 (SAS Institute Inc., Cary, NC, USA) may help physicians optimize and prevent overtreatment with a 1:1 assignment ratio between arms, stratifying by in patients with rheumatoid arthritis treated with TNFi prior TNFi treatment in blocks of four elements. The [17], there are no conclusive data on useful biomarkers randomization list was loaded to a separate module of for the monitoring of TNFi in AS [18]. the eCRF so that the module automatically assigned the The objective of this study was to determine the suit- lowest sequential number available within the ability of dose reduction for long-term treatment with randomization stratum. An auditable registry of the date, TNFi in axSpA and potential predictors of poor clinical time and other variables related to stratification and response. The main study hypothesis was that in patients treatment assignment was kept. Patients and the study with axSpA in persistent clinical remission with TNFi, investigators were not blinded to treatment once reduced TNFi doses would be not inferior to full doses, assigned. as assessed by the proportion of patients not reaching LDA criteria for changing treatments. In addition, we Clinical outcomes analysed pre and post withdrawal drug levels, and The main clinical endpoint was the percentage of pa- markers of inflammation and bone markers in a subset. tients with LDA (BASDAI score < 4, plus physician glo- bal assessment < 4, patient global assessment < 4 and Patients and methods nocturnal axial pain < 4 as assessed on a 0–10 visual A prospective multicentre, parallel, controlled and ran- analogue scale (VAS)) at one year. The key secondary domized open-label clinical trial was conducted at 22 endpoint was the proportion of patients who remained Spanish centres. Patients were eligible if they met criteria in clinical remission (BASDAI <= 2, physician global as- for axSpA according to the ASAS classification criteria sessment <= 2, and patient global assessment <= 2 with [19], on treatment with the recommended doses of com- serum CRP below the upper limit of normality) at 1 year. mercially available TNFi (infliximab, adalimumab, eta- Other secondary objectives were the usual axSpA clinical nercept or golimumab) at the time of the study, and in outcomes (Ankylosing Spondylitis Disease Activity Score sustained clinical remission defined as Bath Ankylosing (ASDAS)-CRP, ASAS response criteria, BASDAI, func- Spondylitis Disease Activity Index (BASDAI) ≤ 2, no tional assessment as measured by Bath Anklylosing clinically active arthritis or enthesitis and C-reactive pro- Spondylitis Functional Index (BASFI) and quality of life tein (CRP) equal to or higher than the upper limit of as measured by Ankylosing Spondylitis Quality of Life normality for ≥ 6 months. Exclusion criteria were sec- (ASQoL)). Relapses defined according to ASAS group, ondary SpA or predominantly peripheral arthritis, co- based on pain, BASDAI and ASDAS-CRP [20] were cal- morbidity interfering with the clinical assessment, culated a posteriori on recorded data. Adverse events, pregnancy and breast feeding. Before any study-related serious adverse reactions requiring hospitalization and/ procedure, patients were given comprehensive informa- or treatment withdrawal, and serious infections requir- tion on the study objectives, procedures and risks, and ing systemic antibiotic treatment and/or hospitalization written informed consent was obtained from all patients. were recorded. Table 1 Study treatments Anti-TNF drug Posology according to SPC Route Full-dose group Reduced-dose group Adalimumab 40 mg every 2 weeks SC 40 mg every 2 weeks 40 mg every 3 weeks Etanercept 25–50 mg every 3–7 days SC 25 mg every 3 days or 50 mg every 10 days 50 mg every 7 days Golimumab 50 mg every month SC 50 mg every 4 weeks 50 mg every 6 weeks Infliximab 5 mg/Kg every 6–8 weeks IV 5 mg/Kg every 6–8 weeks 3 mg/Kg every 8 weeks SPC Summary of Product Characteristics, SC subcutaneous, IV intravenous Gratacós et al. Arthritis Research & Therapy (2019) 21:11 Page 3 of 10 Blood samples for the measurement of TNFi, strategy of 85 patients per arm [24, 25]. The anti-drug antibodies and inflammatory mediators were non-inferiority (delta margin) of 17% was prospectively collected within the 24 h before TNFi injection in 18 set based on the consensus on clinical relevance reached participating centres. The number of samples requested by the rheumatologists involved, who decided that a pro- were ≥ 2 per patient, separated by ≥ 90 days when feas- portion of patients with acceptable control < 70% after 1 ible. Samples were frozen and stored at − 80 °C until year would severely discourage the use of dose reduction shipment at the end of the study. [22]. To ensure the reliability of the study, the lower Inflammatory mediators were assayed at Laboratori de confidence interval (CI) in the full-dose arm had to be > Recerca – I3PT (Sabadell, Spain). Plasma levels of TNF 60% to confirm non-inferiority, in order to ensure that alpha, IL-6, sclerostin (SOST) and Dickkopf-related protein the control treatment had been reasonably effective [22]. 1 (DKK-1) were determined using a Luminex multiplex im- A per-protocol set (PPS) for analysis was the munoassay (Merck-Millipore, Darmstadt) (accuracy = 82– pre-defined primary population for this non-inferiority 97%, inter-assay and intra-assay precision = < 15% and < study, following a conservative approach where any po- 10%, respectively); high-sensitivity CRP (hs-CRP) using a tential differences can be maximized. The principal end- Luminex immunoassay (Merck-Millipore, Darmstadt) (ac- point and the key secondary endpoints were also tested curacy = 93%, inter-assay and intra-assay precision = < 15% in the full analysis set (FAS) population for sensitivity. and < 10%, respectively); plasma calprotectin levels using The principal and key secondary end-points were Calprolab ELISA test HRP (Calpro, Lysaker) (inter-assay assessed by estimating the between-treatment risk differ- and intra-assay precision < 7% and < 6%, respectively). ences after 1 year of randomization and checking these Drug and anti-drug antibody (ADA) levels were deter- against the pre-defined non-inferiority margin (delta (δ)) mined by assay at the Immunology Department – Hos- of 17%. Rates and risk differences were estimated using pital La Paz (Madrid, Spain). TNFi was measured by log-binomial regression including treatment and the fac- commercially available capture ELISA kits, (Promonitor, tors used to stratify assignment. The effects of baseline Progenika, Derio, Vizcaya), with cutoff values established clinical and inflammatory parameters on the results were using serum from 150 healthy blood donors and 100 pa- evaluated, and the heterogeneity of the main results tients with RA who were naïve to biological drugs. The checked for stratification factors. A detailed statistical positivity threshold was mean optical density + 6 stand- analysis plan was issued and approved on 25th Septem- ard deviations: positive levels were infliximab > 0.035 μg/ ber 2014. The statistical analysis was run in the SAS Sys- ml, adalimumab > 0.024 μg/ml, golimumab > 0.022 μg/ tem (SAS Institute Inc., Cary, NC, USA) v9.2. The ml and etanercept > 0.035 μg/ml. ADAs were assayed detailed statistical methods have been reported else- only in samples with no detectable TNFi, using a where [22]. home-made, two-site (bridging) ELISA. Results Visit schedule From 6th July 2012 to 15th May 2014, 157 patients were Visits were scheduled according to routine clinical prac- screened. Due to recruitment delays and time restric- tice, although recommended intervals between visits tions on funding, recruitment was closed after 126 pa- were 8–16 weeks [21]; a mandatory visit was required 1 tients had been randomized; the last patient visit was on year after randomization for the main assessment. Fol- 10th June 2015. Figure 1 shows the number of patients low up continued until the end of study, change in bio- assigned to each treatment arm (FAS), lost to follow up logical treatment, withdrawal of informed consent or during the study (total of three patients) and the reasons loss to follow up. NSAID treatment and dose modifica- for exclusion from the PPS main analysis set. Baseline tion was permitted throughout the study and was re- participant characteristics were similar between groups corded. The study protocol has been reported in full (Table 2). Although the protocol allowed inclusion of pa- elsewhere [22]. tients with axSpA, all included patients fulfilled criteria for AS. Sample size and statistical methods The main study result fulfilled the objective of The estimated proportion of patients at LDA after 1 year non-inferiority, with a proportion (95% CI) of patients of treatment with full-dose TNFi was 87% in patients with LDA at one year of randomization of 83.8% (64.8– who were in stable remission for 4 months [23]. A 102.7%) in the full-dose arm and 81.3% (62.8–99.8%) in non-inferiority margin of 17% was established, and pro- the reduced-dose arm, and absolute difference in LDA at tection against type I error = 2.5% (unilateral) and 1 year between treatment groups of − 2.5% (95% CI of − against type II error = 20%, being the estimated sample 16.6 to 11.7%) in the pre-defined PPS population (Fig. 2a size to test the non-inferiority hypothesis of the and c). The lower bound of CI for the estimate in the reduced-dose strategy with respect to the full-dose control arm was > 60%, fulfilling the pre-specified Gratacós et al. Arthritis Research & Therapy (2019) 21:11 Page 4 of 10 Fig. 1 Disposition of patients. FAS, full analysis; PPS, per-protocol main analysis set; BASDAI, Bath Ankylosing Spondylitis Disease Activity Index; CRP, C-reactive protein. *Not meeting inclusion criteria for remission (BASDAI ≤ 2 and CRP ≤ upper limit of normality) validation criteria. The percentage (95% CI) of patients TNFi on the outcome is independent of the assigned dose in clinical remission (the key secondary endpoint) at one schedule (full dose or reduced dose) (Additional file 1: year of randomization was 83.7% (64.7–102.7%) in the Table S2). Considering inflammatory biomarkers and drug full-dose arm and 78.2% (59.7–96.8%) in the plasma levels, only hs-CRP levels were higher in subjects reduced-dose arm, with absolute differences between without LDA at 1 year (Additional file 1: Table S3). Uni- treatment groups of − 5.5% (− 20.6 to 9.7%) (p = 0.480) variate models including both clinical and analytical bio- (Fig. 2a and c). The sensitivity analysis in the FAS popu- markers had poor predictive value, with the lower CI for lation showed similar estimates for the differences be- the ROC AUC always < 0.62; multivariate methods were tween the two arms for the primary and key secondary unfeasible as no combination of factors improved the uni- end-points (Fig. 2b and d). No significant differences variate models (Additional file 1: Table S3). were observed between treatment arms for any of the Of the 126 randomized patients, serious and related secondary variables (Table 3). adverse events or infections were reported in 7/62 pa- At least one blood sample was available from each of tients (11.3%) in the full-dose group and 2/61 (3.3%) pa- 55 subjects (29 in the full-dose group and 26 in the tients in the reduced-dose group (p value = 0.164) reduced-dose group) for assessment of plasma levels of (Table 3). A total of 34 infections were reported by 26 inflammatory biomarkers and drugs (Additional file 1: subjects, among whom 15 infections were reported by Table S1). All samples tested for TNFi levels were posi- 11/62 subjects (17.7%) in the full-dose group and 4 in- tive, and thus no ADA levels were determined. fections were reported by 3/61 subjects (4.9%) in the Amongst clinical variables at baseline, BASFI, ASQoL reduced-dose group as being at least possibly related to and the type of TNFi were all significantly associated with treatment (Additional file 1: Table S4). Treatment was the probability of LDA after 1 year in univariate models, discontinued in 6/62 (9.7%) and 13/61 patients (21.3%) but only the type of TNFi and BASFI remained significant in the full-dose and reduced-dose groups, respectively (p in multivariate models. The treatment effect (dose group value = 0.086), among whom 3/62 (4.8%) in the full-dose assigned) showed no significant heterogeneity according and 8/61 (13.1%) in the reduced-dose group had treat- to the BASFI (p = 0.5244) or type of TNFi (p = 0.887), thus ment discontinued due to poor disease control, accord- suggesting that the predictive value of BASFI and type of ing to physician judgment (p value = 0.204). Gratacós et al. Arthritis Research & Therapy (2019) 21:11 Page 5 of 10 Table 2 Baseline characteristics of patients Baseline characteristics Full analysis set Per-protocol set Full dose Dose reduction Total Full dose Dose reduction Total (n = 60) (n = 60) (n = 120) (n = 55) (n = 58) (n = 113) Gender (male), n (%) 53 (88.3) 49 (81.7) 102 (85.0) 48 (87.3) 47 (81.0) 95 (84.1) Age (years), mean (SD) 46.2 (13.7) 43.7 (12.4) 44.9 (13.1) 47.2 (13.6) 43.6 (12.4) 45.6 (13.0) BMI, mean (SD) 25.9 (3.4) 25.8 (3.8) 25.9 (3.6) 25.8 (3.4) 25.9 (3.8) 25.9 (3.6) Years from diagnosis, median (P25, P75) 10.4 (7.1, 20.8) 9.3 (5.2, 17.6) 10.0 (5.9, 19.8) 10.4 (7.1, 22.6) 9.3 (5.0, 19.0) 10.0 (5.9, 20.3) ASAS criteria for sacroilitis, n (%) 58 (96.7) 59 (98.3) 117 (97.5) 53 (96.4) 57 (98.3) 110 (97.3) ASDAS-CRP, median (P25, P75) 0.7 (0.5, 1.1) 0.7 (0.5, 1.1) 0.7 (0.5, 1.1) 1.1 (0.8, 3.5) 1.0 (0.7, 1.9) 1.1 (0.7, 2.0) BASDAI, median (P25, P75) 1.0 (0.6, 1.7) 1.0 (0.2, 1.4) 1.0 (0.4, 1.6) 1.0 (0.6, 1.7) 1.0 (0.2, 1.4) 1.0 (0.4, 1.6) VAS nocturnal axial pain, mean (SD) 0.85 (1.0) 1.03 (1.16) 0.94 (1.09) 0.84 (1.01) 1.02 (1.15) 0.93 (1.08) IGA, median (P25, P75) 1.0 (0.0, 2.0) 1.0 (0.0, 1.0) 1.0 (0.0, 1.0) 1.0 (0.0, 2.0) 1.0 (0.0, 1.0) 1.0 (0.0, 1.0) PGA, median (P25, P75) 1.0 (0.0, 2.0) 1.0 (0.0, 2.0) 1.0 (0.0, 2.0) 1.0 (0.0, 2.0) 1.0 (0.0, 1.0) 1.0 (0.0, 1.0) Current TNFi, n (%) Adalimumab 24 (40.0) 23 (38.3) 47 (39.2) 22 (40.0) 22 (37.9) 44 (38.9) Etanercept 21 (35.0) 20 (33.3) 41 (34.2) 19 (34.5) 19 (32.8) 38 (33.6) Golimumab 4 (6.7) 5 (8.3) 9 (7.5) 4 (7.3) 5 (8.6) 9 (8.0) Infliximab 11 (18.3) 12 (20.0) 23 (19.2) 10 (18.2) 12 (20.7) 22 (19.5) N of previous TNFi, n (%) None 50 (83.3) 44 (73.3) 94 (78.3) 46 (83.6) 42 (72.41) 88 (77.9) One 10 (16.7) 11 (18.3) 21 (17.5) 9 (16.4) 11 (19.0) 20 (17.7) Two 0 (0.0) 5 (8.3) 5 (4.2) 0 (0.0) 5 (8.6) 5 (4.4) NSAID use, n (%) 16 (26.7) 14 (23.3) 28 (24.8) 15 (27.3) 13 (22.4) 28 (24.8) n number, SD standard deviation, P25 percentile 25, P75 percentile 75, BMI body mass index, ASAS Assessment of Spondyloarthritis International Society, ASDAS- CRP Ankylosing Spondylitis Disease Activity Score including C-reactive protein, BASDAI Bath Ankylosing Spondylitis Disease Activity Index; VAS patient’s rating of nocturnal axial pain by visual analogue scale ranging from 0 (none) to 10 (worst), PGA Patient Global Assessment of disease activity rated from 0 (best) to 10 (worst), IGA Investigator’s Global Assessment of disease activity rated from 0 (best) to 10 (worst), TNFi TNF inhibitor, NSAID non-steroidal anti-inflammatory drug Discussion reported in patients with AS between BASDAI, BASFI This trial has shown that a dose-reduction strategy may or ASDAS-CRP and disease activity, function and be non-inferior to full-dose treatment in patients with cost-effectiveness of treatments [28–30]. Our study has AS in persistent clinical remission with TNFi, in terms shown similar results in both randomized arms for these of maintenance of LDA after 1 year. In the reduced-dose variables. No differences were observed between the group, less serious and related adverse events or infec- groups in the frequency of flares, and no substantial tions were observed, but also more withdrawals due to changes in the proportion of NSAID treatments were insufficient disease control, although not statistically observed during follow up in either TNFi regimen. How- significant. ever, despite the overall small number of early study Observational studies have reported that complete withdrawals due to poor disease control, these were nu- TNFi discontinuation in patients with AS is not feasible merically higher in the reduced-dose group. This obser- [16] despite 60–80% of patients have LDA at 12–21 vation has to be interpreted in the context of an months after TNFi dose reduction [19, 26, 27]. However, open-label study, and considering which the clinical op- these studies were uncontrolled or retrospective and tions were in the case of observing worse clinical scores: sometimes had a small sample size, wide inclusion cri- in the reduced-dose arm, the options included returning teria and many dose schedules, leading to substantial to full-dose schedule, while in the full-dose group the heterogeneity that limited the generalizability of any options were more aggressive, either dose intensification dose-reduction recommendation. or switching. Our results show that more than 80% and 77% of pa- Since there are no universally accepted criteria for re- tients with AS treated with TNFi had LDA or remission mission in AS, we used strict criteria at baseline, and ran- at 1 year after randomization to standard or reduced- domized only patients with persistent and maintained dose schedules, respectively, in a pragmatic setting, with clinical remission during the last 6 months while on TNFi high external validity. Clear correlation has been [31]. In the absence of an alternative internationally agreed Gratacós et al. Arthritis Research & Therapy (2019) 21:11 Page 6 of 10 Fig. 2 Proportion of patients with low disease activity and clinical remission at 12 months. a Proportion of subjects with low disease activity and clinical remission at 12 months, per-protocol subset (main analysis). b Proportion of subjects with low disease activity and clinical remission at 12 months, intention-to-treat subset. c Adjusted differences between groups and non-inferiority testing for low disease activity and clinical remission, per-protocol subset (main analysis). d Adjusted differences between groups and non-inferiority testing for low disease activity and clinical remission, intention-to-treat subset. Low disease activity was defined by Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) < 4 and Physician Global Assessment < 4, Patient Global Assessment < 4 and axial pain at night < 4. Clinical remission was defined by BASDAI ≤ 2, Physician Global Assessment ≤ 2 and Patient Global Assessment ≤ 2 consensus on thresholds for response and control at the Most patients were on etanercept and adalimumab, with time of the study design, we used the Spanish Society of < 20% of patients on infliximab, consistent with previous Rheumatology definition for low disease activity [16]. The reports [12]. Randomization was stratified by drug to en- definition was reflective of a consensus on the clinical sure comparability of groups; subgroup analysis based therapeutic goal of treatment requiring action if not met. on stratification factors suggested that infliximab was as- The definition was well-known, widely accepted and ap- sociated with a higher risk of failure, although no inter- plied in clinical practice by rheumatologists in Spain. action with the randomized strategy was observed, and The standardized dose-reduction protocol included thus the risk was similar regardless of whether doses lengthening the dosing interval by 50% for all subcutane- were reduced or not. Considering that patients were not ous TNFi and reducing the infliximab dose from 5 to 3 randomized to a given TNFi, but to the dosing strategy, mg/8 week, according to reported data [15]. This this observation is of limited value for causality assess- allowed a reduction in the heterogeneity of drug re- ment, and indication bias cannot be discarded. Similarly, gimes, a limitation reported by other studies [27], al- although observational data suggest that dose reduction though it did not allow evaluation of the potential may be easier with etanercept [14], we found no differ- benefit of additional patient-tailored dose reductions. ences for this stratum. Gratacós et al. Arthritis Research & Therapy (2019) 21:11 Page 7 of 10 Table 3 Secondary endpoints Efficacy secondary variables at 1 year Variables Full dose (n = 55) Dose reduction (n = 58) Differences between groups P value a ASDAS-CRP < 1.3 61.4% [47.3%; 75.5%] 53.5% [39.7%; 67.3%] 7.8% [−10.0%; 25.8%] 0.389 ASDAS-CRP relapsea 6.6% [−5.4%; 18.7%] 12.7% [1.4%; 24.0%] −6.1% [−25.2%; 12.9%] 0.529 Relapse BASDAI-VASa 15.9% [−3.1%; 34.9%] 10.4% [−9.1%; 29.8%] 5.5% [−12.3%; 23.4%] 0.545 Relapse SERa 6.4% [−13.0%; 25.8%] 10.1% [−7.9%; 28.1%] − 3.7% [− 19.2%; 11.8%] 0.638 NSAIDs usea 18.3% [− 1.2%; 37.7%] 20.6% [1.5%; 39.8%] −2.4% [− 18.9%; 14.2%] 0.779 ASDAS-CRP scoreb 1.1 (0.1) [0.9;1.3] 1.1 (0.1) [0.9;1.3] 0.0 (0.2) [−0.3;0.3] 0.783 BASDAIb 1.4 (0.2) [1.1;1.7] 1.4 (0.2) [1.1;1.7] − 0.0 (0.2) [− 0.5;0.4] 0.890 VAS night axial painb 1.4 (0.2) [1.0;1.8] 1.1 (0.2) [0.7;1.6] 0.3 (0.3) [−0.3;0.9] 0.337 b PGA 1.6 (0.2) [1.2;2.0] 1.6 (0.2) [1.1;2.0] 0.0 (0.3) [−0.6;0.6] 0.962 IGAb 1.1 (0.2) [0.8;1.4] 1.1 (0.2) [0.8;1.4] −0.0 (0.2) [− 0.4;0.40] 0.923 BASFIb 1.7 (0.2) [1.2;2.1] 1.8 (0.2) [1.3;2.3] −0.2 (0.3) [− 0.8;0.5] 0.616 ASQoLb 2.3 (0.5) [1.2;3.3] 2.2 (0.6) [1.0;3.3] 0.1 (0.7) [−1.3;1.5] 0.858 Safety secondary variables Adverse event or infections Full dose (n = 62) Dose reduction (n = 61) Differences between groups P value c Any 22 (35.5%) [23.6%;47.4%] 17 (27.9%) [16.6%;39.1%] (7.6%) [−8.8%;24.0%] 0.439 Relatedc 8 (12.9%) [4.6%;21.2%] 5 (8.2%) [1.3%;15.1%] (4.7%) [−6.1%;15.5%] 0.559 c Severe 14 (22.6%) [12.2%;33.0%] 11 (18.0%) [8.4%;27.7%] (4.6%) [−9.6%;18.7%] 0.655 Severe and relatedc 7 (11.3%) [3.4%;19.2%] 2 (3.3%) [−1.2%;7.7%] (8%) [− 1.1%;17.1%] 0.164 Ankylosing Spondylitis Disease Activity Score including C-reactive protein (ASDAS-CRP) relapse was defined by increase ≥ 1.1. Bath Ankylosing Spondylitis Disease Activity Index-visual analogue scale (BASDAI-VAS) was defined by increases of 20% or a 2-unit increase in the 0–10 scale. SER relapse was defined by BASDAI ≥ 4, global clinical assessment by physician ≥ 4 and at least one of three following criteria: patient assessment ≥ 4, axial nocturnal pain (VAS) ≥ 4, and increase in acute phase reactants (reactive °C protein (PCR) and/or erythrocyte sedimentation rate (ESR). Ankylosing Spondylitis Disease Activity Score-C reactive protein (ASDAS-C), which is calculated as (0.12 x back pain) + (0.06 x duration of morning stiffness) + (0.11 x patient GA) + (0.07 x peripheral pain/swelling) + (0.58 x Ln(CRP + 1)); if CRP is not available but ESR is available, the last term is changed by (0.29 x √(ESR)). BASDAI is calculated as {A + B + C + D+ [(E + F)/2]}/5 where A to E are 6 VAS, rated 0 (best) to 10 (worst) assessing (A) fatigue, (B) axial skeletal pain, (C) peripheral joint pain, (D) pain on contact or pressure, (E) intensity of morning stiffness and (F) duration of morning stiffness; VAS nocturnal axial pain is the patient’s rating of nocturnal pain by VAS ranging from 0 (none) to 10 (worst). Patient Global Assessment (PGA) of disease activity was rated from 0 (best) to 10 (worst). Investigator’s Global Assessment (IGA) of disease activity was rated from 0 (best) to 10 (worst). The Bath Ankylosing Spondylitis Functional Index (BASFI) is calculated as the average value of answers to 10 questions rated from 0 (best) to 10 (worst). Ankylosing Spondylitis Quality of Life (ASQoL - Spanish validated version) scores from 0 to 18, where lower scores indicate better health-related quality of life. Adverse event (AE) or infection presented as proportion (95% CI) of patients with at least one reported adverse event and/or infection; related presented as proportion (95% CI) of patients with at least one reported AE and/or infection assessed by the investigator as at least possibly related to TNF inhibitor (TNFi) treatment; severe presented as proportion (95% CI) of patients with at least one severe reported AE and/or infection according to standard definitions (fatal or life-threatening, required or prolonged the patient’s hospitalization, caused significant or persistent disability, caused congenital anomaly/birth defect or required immediate medical intervention to avoid any of the previous outcomes); severe and related presented as proportion (95% CI) of patients with at least one reported severe and related AE and/or infection. SER: Spanish Society of Rheumatology (Sociedad Española de Reumatología) a Adjusted percentage estimates [95% CI], binomial regression b Adjusted least square means (standard error of the mean) [95% CI], mixed model for repeated measurements cNumber (%) [95CI]), analyzed by Fisher’s exact test One goal of personalized medicine is to identify probable are not unexpected, since it is a more sensitive tech- beneficiaries of an intervention through clinical or bio- nique. Reports have suggested a potential association logical markers [18]. We attempted to identify clinical and with TNFi levels in rheumatoid arthritis, but data on AS biological markers of a successful TNFi dose-reduction are controversial [31–33]. We found no association be- strategy. Baseline values of BASFI, ASQoL and the TNFi tween serum TNFi concentrations and the persistence of drug were associated with the odds of LDA after 1 year in LDA at 1 year, although we had only few samples and univariate models; but in multivariate models only TNFi these were for heterogeneous drugs. All samples had drug and BASFI remained significant. Similarly, hs-CRP quantifiable TNFi levels, so none of the samples was was higher in subjects without LDA at 1 year. CRP best re- tested for ADA. Other biomarkers, including calprotec- flects objective inflammation and serum CRP is predictive tin, did not significantly predict the likelihood of LDA at of TNFi response in patients with AS [30]. 1 year in our study. Recently, calprotectin levels have Although reports have suggested no advantages of been suggested as a predictor of the clinical response hs-CRP as compared with CRP in predicting the clinical [18], although in acute clinical response, and not in pa- response to TNFi [18], the results observed for hs-CRP tients in remission. Gratacós et al. Arthritis Research & Therapy (2019) 21:11 Page 8 of 10 There were no interactions between dose regimen and reduced-dose TNFi doses compared with full doses in pa- TNFi or BASFI, indicating that their predictive value was tients with SpA. The results were consistent across many independent of the actual dose regimen assigned, and variables and TNFi drugs, and suggest a better safety pro- multivariate models including biological markers were of file with reduced doses, although the safety data must be low predictive value. Thus, we could not propose any interpreted with caution due to the small sample size. marker to identify patients with higher chances of main- Overall, the study results support TNFi dose-reduction in taining a low disease activity after reduction of TNFi dose. patients with AS in stable clinical remission. Serious adverse events and infections were nominally less than one third in the reduced-dose group as com- Additional file pared to full dose, although not significant. This suggests low TNFi doses might have a better safety profile, but Additional file 1: Table S1. Baseline characteristics, subset with confirmation is needed, since the size and duration of inflammatory biomarkers. Table S2. Baseline predictors of loss of low our study were insufficient to draw conclusions. How- disease activity at 1 year in the logistic regression analysis. Table S3. Univariate methods for loss of low disease activity at 1 year– subset with ever, the differences might be relevant, especially since inflammatory biomarkers. Table S4. Listing of infectious adverse events. these are long-term treatments for often-fragile patients. (DOCX 38 kb) Our study has several limitations. First, due to limited funding and recruitment delays, the final sample size Abbreviations was smaller than anticipated. Although the sample was AS: Ankylosing spondylitis; ASAS: Assessment of Spondyloarthritis sufficient to conclude on the pre-determined primary International Society; ASDAS: Ankylosing Spondylitis Disease Activity Score; objective, the limited sample size hindered the imple- ASQoL: Ankylosing Spondylitis Quality of Life; axSpA: Axial spondyloarthritis; BASFI: Bath Anklylosing Spondylitis Functional Index; CI: Confidence interval; mentation of secondary objectives. Differences in ad- CRP: C-reactive protein; DMARD: Disease-modifying anti-rheumatic drug; verse events were not significant and biological samples eCRF: Electronic case report form; ELISA: Enzyme-linked immunosorbent were obtained from only half the study population, likely assay; FAS: Full analysis set; hs-CRP: High-sensitivity C-reactive protein; LDA: Low disease activity; NSAID: Non-steroidal anti-inflammatory drug; reducing the ability to build a predictive model to iden- PPS: Per-protocol set; SD: Standard deviation; SpA: Spondyloarthritis; tify patients with poorer chances of maintaining LDA TNFi: TNF inhibitors after dose-halving. In addition, although the study was designed to include axSpA, only patients with AS were Acknowledgements finally included. In addition, the study was not designed REDES-TNF Investigators to determine which TNFi is more effective, nor powered Miriam Almirall (IMIM – Hospital del Mar, Barcelona); Maria Aparicio (Hospital Universitari de Bellvitge, Barcelona); Agustí Sellas (Hospital Vall d’Hebron, to determine whether a specific TNFi can be reduced Barcelona); Roser Vives, Nestor Albiñana; Mireia Moreno (Hospital Universitari more successfully than another. ParcTaulí. I3PT. UAB. Sabadell (Barcelona); Teresa Clavaguera (Hospital A conservative TNFi reduction schedule was applied Comarcal Palamós, Girona); Juan Carlos Torre-Alonso (Hospital Monte Naranco, Oviedo); Raúl Veroz (Hospital de Mérida, Mérida); Carlos Rodríguez- in this study (about 50% of the recommended dose), and Lozano (Hospital Universitario Dr. Negrín, Las Palmas de Gran Canaria); Luís neither the potential benefit of an extra dose reduction Francisco Linares (Hospital Virgen de la Arrixaca, Murcia); Ana Urruticoechea or the effects of treatment withdrawal were studied. In (Hospital Can Misses, Eivissa, Baleares); Eduardo Collantes (Hospital Reina Sofía, Córdoba); Rosa María Morlà (Hospital de Sant Pau i Santa Tecla, clinical practice, infliximab can be reduced more than Tarragona); Dèlia Reina (Hospital de Moises Broggi, Esplugues de Llobregat other TNFi, i.e., dose and interval; however, given the (Barcelona)); Eduardo Cuende (Hospital Príncipe de Asturias, Madrid); Pedro limited statistical power and the fact that this was an un- Zarco (Hospital Fundación Alcorcón, Madrid); Maria Cruz Fernández-Espartero (Hospital Universitario de Móstoles, Madrid); Rosario García-Vicuña (Hospital planned analysis, we cannot venture to show results of Universitario de La Princesa, Madrid); Carlos Alberto Montilla (Hospital Clínico subgroup analysis by dose. Likewise, the 12-month fol- de Salamanca, Salamanca); Alejandro Villalba, Dora Pascual (Hospital La Paz, low up was insufficient to determine the potential effects Madrid); Cristina Campos (Hospital General de Valencia; Valencia); Antonio Juan (Hospital Son Llatzer; Mallorca); Rafael Ariza (Hospital Virgen Macarena; of long-term TNFi reduction, especially on structural Sevilla); Consuelo Díaz-Miguel (Hospital Ramón y Cajal; Madrid); Manuel damage, which requires 2–4 years to detect relevant Maqueda (Hospital de Llerena, Llerena, Badajoz); Maria Pilar Fernández- changes [31, 34]. A suggestion that TNFi dose tapering Dapica (Hospital Doce de Octubre, Madrid); Manuel Fernández-Prada (Hospital Universitario de Guadalajara, Guadalajara); Enrique Batlle (Hospital may be associated with more rapid radiographic progres- Clinico de Sant Joan d’Alacant, Alacant); Carlos González-Fernández (Hospital sion than full doses regardless of LDA maintenance [31], Universitario Gregorio Marañón, Madrid); Rubén Queiro (Hospital especially in patients with baseline syndesmophytes, Universitario Central de Asturias; Oviedo). must be confirmed by prospective studies. Strict moni- toring of radiographic progression of patients with AS is Funding The study was fully funded by grants from the Spanish Ministry of Health recommended if dose reduction is considered [31]. within the programme “Ayudas para el fomento de la investigación clínica independiente del Ministerio de Salud, Política Social e Igualdad - Orden SPI/ Conclusion 2885/2011, de 20 de octubre” (project ID: EC11–229) and from the Ministerio de Economia y Consumo (MINECO)-Instituto de Salud Carlos III-Subdirección In spite of a number of limitations, this is the first General de Evaluación and the European Fund for Regional Development randomized trial to support the non-inferiority of (Project ID: PI13/02680). Gratacós et al. Arthritis Research & Therapy (2019) 21:11 Page 9 of 10 Availability of data and materials Received: 1 August 2018 Accepted: 19 November 2018 The datasets supporting the conclusions of this article are available from the corresponding author on reasonable request. References Authors’ contributions 1. Braun J, Bollow M, Remlinger G, Eggens U, Rudwaleit M, Distler A, Sieper J. All authors contributed to the design and the manuscript and FT performed Prevalence of spondylarthropathies in HLA-B27 positive and negative blood the analysis. All authors read and approved the final manuscript. donors. Arthritis Rheum. 1998;41(1):58–67. 2. Gran JT, Husby G. The epidemiology of ankylosing spondylitis. Semin Arthritis Rheum. 1993;22(5):319–34. Ethics approval and consent to participate 3. Braun J, Pincus T. Mortality, course of disease and prognosis of patients with The trial was performed according to the study protocol and protocol ankylosing spondylitis. Clin Exp Rheumatol. 2002;20(6 Suppl 28):S16–22. amendments, which were approved by the Ethics Committee for Clinical 4. Castillo-Ortiz JD, Ramiro S, Landewe R, van der Heijde D, Dougados M, van Research of the participating sites and by the Spanish Agency for Medicines den Bosch F, Boonen A. Work outcome in patients with ankylosing and Health Products, and complied with the ethical principles of biomedical spondylitis: results from a 12-year followup of an international study. research, Good Clinical Practice [35] and applicable Spanish legislation. Two Arthritis Care Res (Hoboken). 2016;68(4):544–52. amendments to the protocol reflected in the clinical trial registration sites 5. Smolen JS, Braun J, Dougados M, Emery P, Fitzgerald O, Helliwell P, were implemented after recruitment was initiated, which referred to Kavanaugh A, Kvien TK, Landewe R, Luger T, et al. Treating inclusion of new investigators and recruiting sites, and to allow collection of spondyloarthritis, including ankylosing spondylitis and psoriatic arthritis, blood samples for measurement of inflammatory mediators, drug plasma to target: recommendations of an international task force. Ann Rheum measurements and ADA. Dis. 2014;73(1):6–16. 6. Zochling J, Bohl-Buhler MH, Baraliakos X, Feldtkeller E, Braun J. Nonsteroidal anti-inflammatory drug use in ankylosing spondylitis–a population-based Consent for publication survey. Clin Rheumatol. 2006;25(6):794–800. All authors have read and approved the final version and consent to 7. Coates LC, Cawkwell LS, Ng NW, Bennett AN, Bryer DJ, Fraser AD, Emery P, publication. Marzo-Ortega H. Real life experience confirms sustained response to long- term biologics and switching in ankylosing spondylitis. Rheumatology (Oxford). 2008;47(6):897–900. Competing interests 8. Bongartz T, Sutton AJ, Sweeting MJ, Buchan I, Matteson EL, Montori V. Anti- No author has received any private funding, nor has any been influenced in TNF antibody therapy in rheumatoid arthritis and the risk of serious the preparation of the manuscript. The following authors declare that they infections and malignancies: systematic review and meta-analysis of rare have no competing interests: CP, FT, CAS, AV and GC. In the past 2 years, JG harmful effects in randomized controlled trials. JAMA. 2006;295(19):2275–85. has received grants to attend congresses and educational courses, has 9. Busquets N, Carmona L, Suris X. Systematic review: safety and efficacy of obtained fees for scientific consultancies and has received speaking fees and anti-TNF in elderly patients. Reumatol Clin. 2011;7(2):104–12. fees for participation in educational programmes from the following 10. Dreyer L, Magyari M, Laursen B, Cordtz R, Sellebjerg F, Locht H. Risk of companies: Roche, MSD, Pfizer, AbbVie, Janssen Cilag, UCB Pharma, Novartis multiple sclerosis during tumour necrosis factor inhibitor treatment for and Celgene. EdM has received grant/research support from AbbVie and arthritis: a population-based study from DANBIO and the Danish Multiple Pfizer, honorarium as a speaker from Abbvie, MSD and Novartis and Sclerosis Registry. Ann Rheum Dis. 2016;75(4):785–6. consultant fees from Abbvie and Novartis. JS has obtained grants to attend 11. Baraliakos X, Listing J, Brandt J, Zink A, Alten R, Burmester G, Gromnica-Ihle congresses, speaking fees and fees for participation in educational E, Kellner H, Schneider M, Sorensen H, et al. Clinical response to programmes from Abbvie, Pfizer, MSD, Roche, UCB and Menarini. XJ has discontinuation of anti-TNF therapy in patients with ankylosing spondylitis attended advisory board meetings of Celgene, has received investigational after 3 years of continuous treatment with infliximab. Arthritis Res Ther. grants from Abbvie and Pfizer and has received fees for participation in 2005;7(3):R439–44. educational programmes from MSD, Pfizer and Abbvie. RS has received 12. Torrente V, Gratacos J, Juanola X, Sanmarti R, Suarez D, Moreno M. speaking fees and fees for attending advisory boards from Abbott/Abbvie, Infliximab withdrawal in patients with spondyloarthritis who presented Bristol-Myers Squibb, MSD, Roche, UCB and Pfizer, investigational grants from criteria of clinical disease remission. an open study of clinical practise Bristol-Myers Squibb, MSD, Roche, UCB, Pfizer, FER and SCR and funding for (REMINEA). Arthritis Rheum. 2009;60(Suppl 10):S1785. educational projects from MSD, Bristol-Myers Squibb and Abbvie. 13. Brandt J, Listing J, Haibel H, Sorensen H, Schwebig A, Rudwaleit M, Sieper J, Braun J. Long-term efficacy and safety of etanercept after readministration in patients with active ankylosing spondylitis. Rheumatology (Oxford). 2005; Publisher’s Note 44(3):342–8. Springer Nature remains neutral with regard to jurisdictional claims in 14. Navarro-Compan V, Moreira V, Ariza-Ariza R, Hernandez-Cruz B, Vargas- published maps and institutional affiliations. Lebron C, Navarro-Sarabia F. Low doses of etanercept can be effective in ankylosing spondylitis patients who achieve remission of the disease. Clin Author details Rheumatol. 2011;30(7):993–6. 1 Rheumatology Department, Hospital de Sabadell, Institut Universitari Parc 15. Keeling S, Oswald A, Russell AS, Maksymowych WP. Prospective Taulí, Universitat Autònoma de Barcelona, Sabadell, Barcelona, Spain. 2Clinical observational analysis of the efficacy and safety of low-dose (3 mg/kg) Pharmacology Unit, Hospital de Sabadell, Institut Universitari Parc Taulí, infliximab in ankylosing spondylitis: 4-year followup. J Rheumatol. 2006; Universitat Autònoma de Barcelona, c/Taulí n°1, 08208 Sabadell, Barcelona, 33(3):558–61. Spain. 3Rheumatology Department, Hospital Universitario de Bellvitge, 16. Gonzalez-Alvaro I, Martinez-Fernandez C, Dorantes-Calderon B, Garcia-Vicuna Universitat de Barcelona, Bellvitge, Barcelona, Spain. 4Rheumatology R, Hernandez-Cruz B, Herrero-Ambrosio A, Ibarra-Barrueta O, Martin-Mola E, Department, Hospital Universitario Puerta de Hierro- Majadahonda, Madrid, Monte-Boquet E, Morell-Baladron A, et al. Spanish Rheumatology Society Spain. 5Medical Statistics core facility, IDIBAPS, Hospital Clínic, Biostatistics and Hospital Pharmacy Society Consensus on recommendations for Unit, School of Medicine, Universitat Autònoma de Barcelona, Barcelona, biologics optimization in patients with rheumatoid arthritis, ankylosing Spain. 6Clinical Pharmacology Department, Hospital Puerta de Hierro spondylitis and psoriatic arthritis. Rheumatology (Oxford). 2015;54(7):1200–9. Majadahonda, Madrid, Spain. 7Clinical Pharmacology Department, Hospital 17. Chen DY, Chen YM, Hsieh TY, Hung WT, Hsieh CW, Chen HH, Tang KT, Lan Universitario de Bellvitge - Universitat de Barcelona, Barcelona, Spain. JL. Drug trough levels predict therapeutic responses to dose reduction of 8 Clinical Pharmacology Department, Hospital Clínic de Barcelona - Universitat adalimumab for rheumatoid arthritis patients during 24 weeks of follow-up. de Barcelona, Barcelona, Spain. 9Rheumatology Department, Hospital Rheumatology (Oxford). 2016;55(1):143–8. Universitario La Paz, Madrid, Spain. 10Rheumatology Department, Hospital 18. Maksymowych WP. Biomarkers in axial spondyloarthritis. Curr Opin Clínic de Barcelona - Universitat de Barcelona, Barcelona, Spain. Rheumatol. 2015;27(4):343–8. Gratacós et al. Arthritis Research & Therapy (2019) 21:11 Page 10 of 10 19. Rudwaleit M, van der Heijde D, Landewe R, Listing J, Akkoc N, Brandt J, Braun J, Chou CT, Collantes-Estevez E, Dougados M, et al. The development of Assessment of SpondyloArthritis international Society classification criteria for axial spondyloarthritis (part II): validation and final selection. Ann Rheum Dis. 2009;68(6):777–83. 20. Gossec L, Portier A, Landewe R, Etcheto A, Navarro-Compan V, Kroon F, van der Heijde D, Dougados M. Preliminary definitions of ‘flare’ in axial spondyloarthritis, based on pain, BASDAI and ASDAS-CRP: an ASAS initiative. Ann Rheum Dis. 2016;75(6):991–6. 21. Juanola Roura X, Zarco Montejo P, Sanz Sanz J, Munoz Fernandez S, Mulero Mendoza J, Linares Ferrando LF, Gratacos Masmitja J, Garcia de Vicuna R, Fernandez Carballido C, Collantes Estevez E, et al. Consensus Statement of the Spanish Society of Rheumatology on the management of biologic therapies in spondyloarthritis except for psoriatic arthritis. Reumatol Clin. 2011;7(2):113–23. 22. Pontes C, Gratacos J, Torres F, Avendano C, Sanz J, Vallano A, Juanola X, de Miguel E, Sanmarti R, Calvo G. Evaluation of dose reduction versus standard dosing for maintenance of remission in patients with spondyloarthritis and clinical remission with anti-TNF (REDES-TNF): study protocol for a randomized controlled trial. Trials. 2015;16:370. 23. Conti F, Ceccarelli F, Marocchi E, Magrini L, Spinelli FR, Spadaro A, Scrivo R, Valesini G. Switching tumour necrosis factor alpha antagonists in patients with ankylosing spondylitis and psoriatic arthritis: an observational study over a 5-year period. Ann Rheum Dis. 2007;66(10):1393–7. 24. Elashoff JD. nQuery Version 7.0 Advisor User’s Guide. Statistical Solutions Ltd, Los Angeles, CA; 2007. 25. Newcombe RG. Interval estimation for the difference between independent proportions: comparison of eleven methods. Stat Med. 1998;17(8):873–90. 26. Lee J, Noh JW, Hwang JW, Oh JM, Kim H, Ahn JK, Lee YS, Cha HS, Koh EM. Extended dosing of etanercept 25 mg can be effective in patients with ankylosing spondylitis: a retrospective analysis. Clin Rheumatol. 2010;29(10): 1149–54. 27. Zavada J, Uher M, Sisol K, Forejtova S, Jarosova K, Mann H, Vencovsky J, Pavelka K. A tailored approach to reduce dose of anti-TNF drugs may be equally effective, but substantially less costly than standard dosing in patients with ankylosing spondylitis over 1 year: a propensity score- matched cohort study. Ann Rheum Dis. 2016;75(1):96–102. 28. Pavy S, Brophy S, Calin A. Establishment of the minimum clinically important difference for the bath ankylosing spondylitis indices: a prospective study. J Rheumatol. 2005;32(1):80–5. 29. Kobelt G, Andlin-Sobocki P, Maksymowych WP. The cost-effectiveness of infliximab (Remicade) in the treatment of ankylosing spondylitis in Canada. J Rheumatol. 2006;33(4):732–40. 30. Ramiro S, van der Heijde D, van Tubergen A, Stolwijk C, Dougados M, van den Bosch F, Landewe R. Higher disease activity leads to more structural damage in the spine in ankylosing spondylitis: 12-year longitudinal data from the OASIS cohort. Ann Rheum Dis. 2014;73(8):1455–61. 31. Park JW, Kwon HM, Park JK, Choi JY, Lee EB, Song YW, Lee EY. Impact of dose tapering of tumor necrosis factor inhibitor on radiographic progression in ankylosing spondylitis. PLoS One. 2016;11(12):e0168958. 32. Bouman C, van Herwaarden N, van den Hoogen F, van der Maas A, van den Bemt B, den Broeder AA. Prediction of successful dose reduction or discontinuation of adalimumab, etanercept, or infliximab in rheumatoid arthritis patients using serum drug levels and antidrug antibody measurement. Expert Opin Drug Metab Toxicol. 2017;13(6):597–604. 33. Marsman AF, Kneepkens EL, Ruwaard J, Wei JC, Nurmohamed MT, van Denderen C, van der Horst-Bruinsma IE, Rispens T, Wolbink G. Search for a concentration-effect curve of adalimumab in ankylosing spondylitis patients. Scand J Rheumatol. 2016;45(4):331–4. 34. Poddubnyy D, Sieper J. Radiographic progression in ankylosing spondylitis/ axial spondyloarthritis: how fast and how clinically meaningful? Curr Opin Rheumatol. 2012;24(4):363–9. 35. International Conference on Harmonisation. ICH: E 6 (R1): Guideline for good clinical practice - Step 5. [https://www.ich.org/fileadmin/Public_Web_ Site/ICH_Products/Guidelines/Efficacy/E6/E6_R1_Guideline.pdf]. Last access: 30th November 2018.