Gradual tapering TNF inhibitors versus conventional synthetic DMARDs after achieving controlled disease in patients with rheumatoid arthritis: first-year results of the randomised controlled TARA study

Objectives The aim of this study is to evaluate the effectiveness of two tapering strategies after achieving controlled disease in patients with rheumatoid arthritis (RA), during 1 year of follow-up. Methods In this multicentre single-blinded (research nurses) randomised controlled trial, patients with RA were included who achieved controlled disease, defined as a Disease Activity Score (DAS) ≤ 2.4 and a Swollen Joint Count (SJC) ≤ 1, treated with both a conventional synthetic disease-modifying antirheumatic drugs (csDMARD) and a TNF inhibitor. Eligible patients were randomised into gradual tapering csDMARDs or TNF inhibitors. Medication was tapered if the RA was still under control, by cutting the dosage into half, a quarter and thereafter it was stopped. Primary outcome was proportion of patients with a disease flare, defined as DAS > 2.4 and/or SJC > 1. Secondary outcomes were DAS, European Quality of Life-5 Dimensions (EQ5D) and functional ability (Health Assessment Questionnaire Disability Index [HAQ-DI]) after 1 year and over time. Results A total of 189 patients were randomly assigned to tapering csDMARDs (n = 94) or tapering anti-TNF (n = 95). The cumulative flare rates in the csDMARD and anti-TNF tapering group were, respectively, 33 % (95% CI,24% to 43 %) and 43 % (95% CI, 33% to 53 % (p = 0.17). Mean DAS, HAQ-DI and EQ-5D did not differ between tapering groups after 1 year and over time. Conclusion Up to 9 months, flare rates of tapering csDMARDs or TNF inhibitors were similar. After 1 year, a non-significant difference was found of 10 % favouring csDMARD tapering. Tapering TNF inhibitors was, therefore, not superior to tapering csDMARDs. From a societal perspective, it would be sensible to taper the TNF inhibitor first, because of possible cost reductions and less long-term side effects. Trial registration number NTR2754

How might this impact on clinical practice or future developments? ► Tapering TNF inhibitors was not superior to tapering csDMARDs. We, therefore, advise to taper the TNF inhibitor first. This supports current EULAR guidelines.

AbsTRACT
Objectives The aim of this study is to evaluate the effectiveness of two tapering strategies after achieving controlled disease in patients with rheumatoid arthritis (Ra), during 1 year of follow-up. Methods in this multicentre single-blinded (research nurses) randomised controlled trial, patients with Ra were included who achieved controlled disease, defined as a Disease activity score (Das) ≤ 2.4 and a swollen Joint Count (sJC) ≤ 1, treated with both a conventional synthetic disease-modifying antirheumatic drugs (csDMaRD) and a TnF inhibitor. eligible patients were randomised into gradual tapering csDMaRDs or TnF inhibitors. Medication was tapered if the Ra was still under control, by cutting the dosage into half, a quarter and thereafter it was stopped. Primary outcome was proportion of patients with a disease flare, defined as Das > 2.4 and/or sJC > 1. secondary outcomes were Das, european Quality of life-5 Dimensions (eQ5D) and functional ability (Health assessment Questionnaire Disability index [HaQ-Di]) after 1 year and over time. Results a total of 189 patients were randomly assigned to tapering csDMaRDs (n = 94) or tapering anti-TnF (n = 95). The cumulative flare rates in the csDMaRD and anti-TnF tapering group were, respectively, 33 % (95% Ci,24% to 43 %) and 43 % (95% Ci, 33% to 53 % (p = 0.17). Mean Das, HaQ-Di and eQ-5D did not differ between tapering groups after 1 year and over time. Conclusion Up to 9 months, flare rates of tapering csDMaRDs or TnF inhibitors were similar. after 1 year, a non-significant difference was found of 10 % favouring csDMaRD tapering. Tapering TnF inhibitors was, therefore, not superior to tapering csDMaRDs. From a societal perspective, it would be sensible to taper the TnF inhibitor first, because of possible cost reductions and less long-term side effects.

Trial registration number nTR2754
InTROduCTIOn Treatment outcomes of rheumatoid arthritis (RA) have improved enormously during the past decades due to earlier detection of the disease, a treat-to-target approach and intensified treatment, especially combination therapy with conventional synthetic disease-modifying antirheumatic drugs (csDMARDs) and biological DMARDs (bDMARDs). As a result, 50%-60% of patients with early RA are able to reach low disease activity or even sustained remission. [1][2][3][4] Because of these improved outcomes, it is nowadays more common to taper medication in patients with RA, who are in sustained remission. This is in accordance with current treatment guidelines. 4 However, an optimal tapering approach, including in which order, still has to be unravelled. The benefits of tapering treatment are: (1) a decreased risk of long-term adverse events due to immunosuppression, that is, increased infection risk and possibility of malignancy development, (2) a reduction of healthcare costs, especially when biologicals are tapered and (3) a possibly improved compliance. 5 6 On the other hand, tapering treatment may lead to more transient or persistent disease flares with potential harmful consequences. 1 7 8 Previous studies have shown that it is possible to taper DMARDs in various ways, which has been extensively reviewed by several research groups. 7 9-15 bDMARDs are most frequently completely withdrawn. However, with this tapering strategy, the risk of disease flares in the first year of follow-up is very high. Other bDMARD-tapering studies used a dose-reduction approach, which resulted in less disease flares. However, to our knowledge, no randomised trials have been performed that investigate which DMARD should be tapered first.
Therefore, the aim of this study is to compare the effectiveness of two tapering strategies, namely gradually tapering csDMARDs or tumor necrosis factor (TNF) inhibitors, in patients with RA with controlled disease under a combination of csDMARDs and a TNF inhibitor.

PATIenTs And MeTHOds study design
Data were used from a clinical trial (NTR2754)-namely, TApering strategies in Rheumatoid Arthritis (TARA). TARA, a multicentre, single-blinded (research nurses) randomised trial, was carried out in 12 rheumatology centres in the Southwestern part of the Netherlands. Inclusion started in September 2011 and ended July 2016.

Patients
Adult patients with RA with controlled disease, defined as a Disease Activity Score (DAS) ≤2.4 and a Swollen Joint Count (SJC) ≤1 at two consecutive time points within a 3-month interval, with a combination of a csDMARD and TNF inhibitor, were included. Exclusion criteria were: (1) not being able to understand, speak and write in Dutch; (2) being diagnosed with a psychiatric or personality disorder and (3) tapering or stopping therapy due to other reasons.

Randomisation and blinding
Patients were randomised using minimisation randomisation stratified for centre. Trained research nurses, blinded to the allocated treatment arm throughout the study, examined patients and calculated the DAS.

Tapering schedule
Patients were randomised into gradual tapering their csDMARD or TNF inhibitor. csDMARD tapering was realised by cutting the dosage into half, a quarter and thereafter it was stopped. The TNF inhibitor was tapered by doubling the dose interval, followed by cutting the dosage into half, and thereafter it was stopped. The total tapering schedule took 6 months, with dose adjustments every 3 months as long as there was still a controlled disease. At the start of the study, patients were asked to refrain from glucocorticoids (GCs). There were no restrictions on the use of non-steroidal anti-inflammatory drugs (NSAIDs) or intra-articular GC injections.
If a disease flare occurred, defined as DAS >2.4 and/or SJC >1, tapering was stopped and the last effective treatment, when RA was under control, was restarted. In case of a flare, one intramuscular GC injection was allowed as bridging therapy. After a flare, no further attempts were taken to taper medication during the remainder of the first year of follow-up.

Outcomes
The primary outcome was the proportion of patients with a disease flare within 1 year. Secondary endpoints were disease activity, functional ability, quality of life, medication usage and radiographic progression.
Disease activity was measured with the DAS. Functional ability was measured with the Health Assessment Questionnaire Disability Index (HAQ-DI). 16 Higher HAQ-DI scores indicate poorer function. Quality of life was measured with the European Quality of Life-5 Dimensions (EQ-5D) and Short Form-36 (SF-36). [17][18][19] A higher EQ-5D index or SF-36 score indicates a better quality of life. Radiographic progression was measured with the modified total Sharp score (mTSS). 20 Radiographs were scored chronologically by two out of three qualified assessors, who were blinded for study allocation and the identity of the patients. 21 Median mTSS are reported. 22 The weighted overall κ was 0.75 with >99% agreement. The percentage of patients with radiographic progression, defined as a change in mTSS >0.5 and >0.9 (the smallest detectable change), are given. 22

Follow-up and assessments
Treatment strategies were tightly controlled, with patients being examined at baseline and every 3 months thereafter. At each time point, the DAS, medication usage, development of complications and self-reported questionnaires were collected, except for hand and foot radiographs, which were obtained at baseline and after 1 year of follow-up.

safety monitoring
Safety monitoring took place according to Dutch guidelines, and included laboratory tests every 3 months. [23][24][25] The medication was stopped or the dosage was lowered in case of adverse events related to medication use.

statistical analysis
The TARA study was a superiority trial, powered to detect a 20% difference in flare rates between both tapering strategies. Based on related prospective cohort studies from 2011 and before, following assumptions were made: (1) 40% of the patients tapering their TNF inhibitors to half will have controlled disease after 6 months and (2) 60% of the csDMARD tapering group will have controlled disease after 6 months. [26][27][28] Therefore, to detect this 20% difference using a significance level of α=0.05 and a power of 80%, 107 patients were needed in each treatment arm, also taking a 10% drop-out ratio into account.
Outcomes were calculated in an intention-to-treat analysis, using all available data. Differences in cumulative flare rates between groups were analysed with a logistic regression model. To account for stratified randomisation by centre, intercepts for each centre were included. Flare-free survival was visualised with Kaplan-Meier curves. Descriptive statistics were used to assess the proportion of patients with a controlled disease after 12 months of follow-up. A linear mixed model with maximum likelihood optimisation was used to compare DAS, HAQ-DI and EQ-5D over time. Random intercepts were included for both hospital and individual patients. Residual correlation was modelled by inclusion of an autoregressive order correlation structure. In the final model, the differences in evolution over time for the outcome DAS, HAQ-DI and EQ-5D between the two groups were assessed.    Statistical comparisons of the baseline characteristics and outcomes were made by Student's t-test, χ 2 test or Wilcoxon rank-sum test, when appropriate.

Rheumatoid arthritis
All data were analysed using STATA V.15. A p≤0.05 was considered statistically significant.

ResulTs Patients
A total of 330 patients were assessed for eligibility and 189 of those were randomly assigned to tapering their csDMARD (n=94) or tapering their TNF inhibitor (n=95). Most patients who were not eligible did not meet the inclusion criteria for remission or refused participation (figure 1). During the first year of follow-up, 14 patients withdrew from the study, mainly because of refraining from further participation (figure 1). Table 1 shows the baseline characteristics for both tapering strategies. Patients had an average symptom duration of 6.8 years and were predominantly female (66.1%) with an average age of 56.6 years. Baseline mean (SD) HAQ-DI was 0.52 (0.47) and 0.47 (0.53) and EQ-5D was 0.86 (0.12) and 0.87 (0.11) for, respectively, the csDMARD and TNF inhibitor tapering group.

Outcomes
After 1 year of follow-up, the cumulative flare rate was 33% (95% CI, 24% to 43%) in the csDMARD and 43% (95% CI, 33% to 53%) in the TNF inhibitor tapering group (figure 2). This means that 63/94 (67%) in the csDMARD tapering group and 54/95 (57%) in the TNF inhibitor tapering group still had a well-controlled RA (p=0.17). Of the patients who flared and restarted the last effective treatment strategy, 46% regained a DAS <2.4 within 3 months, which increased to 67% by 6 months. Two patients (1%) were unable to get back in remission within the first year.
No significant differences were seen in DAS (p=0.72), HAQ-DI (p=0.63) and EQ-5D (p=0.58) after 1 year between both tapering strategies (table 2). Also over time, the DAS (p=0.49) and EQ-5D (p=0.35) were not significantly different between both tapering strategies ( figure 3). Although the TNF inhibitor tapering group seems to have lower HAQ-DI scores over time, this was not significantly different (p=0.15) (figure 3).
Over time, the patients with a disease flare increased and thus the proportion of patients with a DAS <2.4 decreased in both tapering strategies. A similar trend was seen for the HAQ-DI and EQ-5D over time ( figure 3).

Treatment
After 12 months, 58 patients in the csDMARD-tapering group and 45 patients in the TNF inhibitor tapering group completely tapered their medication (p=0.09). On the other hand, 8 and 16 patients were using the same dosage as at start of the trial. The remaining patients were able to taper their medication partially ( figure 4C). The course of the tapering schedule is visualised in figure 4A,B. There was an overall significant difference in tapering status after 12 months of follow-up between the two tapering strategies (p=0.02). During the follow-up period, we found no significant differences in GC and NSAID usage between both tapering groups (figure 4D).

Adverse events
In the csDMARD-tapering group, 82 adverse events were self-reported versus 98 in the TNF inhibitor tapering group (online supplemental table S1). Serious adverse events (SAEs) were seen in 10 (12%) patients tapering csDMARDs and 5 (6%) patients tapering TNF inhibitors (p=0.3, online supplemental table S1). Reported SAEs were hospitalisation, herpes zoster infection, basal cell carcinoma, large cell lung carcinoma and a bruised rib. None of the SAEs were considered to be related to the trial treatment.

dIsCussIOn
Tapering csDMARDs resulted in a 33% (95% CI, 24% to 43%) flare rate (DAS44 >2.4 and/or an SJC>1), while tapering TNF inhibitors gave a 43% (95% CI, 33% to 53%) flare rate over a 1-year period in the randomised controlled TARA trial. At 12 months, 103 (59%) patients were able to stop either their TNF inhibitor or csDMARD, while 47 (37%) patients were using a lower dosage. Clinical and patient-reported outcomes were comparable in both tapering groups over time and after 1 year of follow-up. Also, no significant differences in adverse events or radiological progression were seen between both tapering strategies.
Nowadays, more patients with RA achieve a state of sustained remission, which makes them eligible for tapering treatment. This is reflected in current European league against rheumatism (EULAR) recommendations for the management for RA. The advice is to taper DMARD therapy in patients with RA who are in sustained remission in the following ordering: GCs, bDMARDS and csDMARDs. 4 Our results and the fact that TNF-blockers are more expensive than csDMARDs support aformentioned tapering order.
The majority of previous tapering trials focused on the withdrawal of TNF inhibitors alone. Flare rates for tapering TNF inhibitors varied between 51% and 77%. The Potential Optimalisation of Expediency and Effectiveness of TNF-blockers (POET) study, for example, reported a 51.2% flare rate (DAS28 >3.2 or ΔDAS28 >0.6) after stopping the TNF inhibitor. 7 The Spacing of TNF-blocker injections in Rheumatoid ArthritiS Study (STRASS) showed a 76.6% flare rate (DAS28 >2.6 or ΔDAS28 >0.6) when extending the dosage interval of the TNF inhibitor. 14 The Dose REduction Strategy of Subcutaneous TNF inhibitors study (DRESS) reported a 55% flare rate (ΔDAS28-CRP>0.6) after a dose reduction of the TNF inbitor. 13   With Active Rheumatoid Arthritis) reported a 57.4% flare rate (DAS28 >3.2) when the TNF inhibitor was stopped, and a 20.9% flare rate (DAS28 >3.2) when the TNF inhibitor dose was cut into half. 15 Only few randomised controlled trials investigated tapering of csDMARDs, but the majority looked at the combined tapering of csDMARDs and biologicals. Flare rates within these studies varied between 35% and 56%. 1 6 29-32 Although flare rates of aforementioned studies are similar to or higher than our findings, direct comparison is difficult, because of the differences in the study design. The most important study design differences are: (1) no common definition for relapse or flare, (2) no comparison between tapering of csDMARDs and TNF inhibitors, and (3) DMARD therapy could only be tapered or stopped once during follow-up. If we would use other criteria to define a flare in the TARA population, we would observe higher hypothetical flare rates. We would have encountered a 74.1% flare rate if using DAS28 >3.2 or ΔDAS28 >0.6, an 80.5% flare rate if using DAS28 >2.6 or ΔDAS28 >0.6, a 52.3% flare rate if we use ΔDAS28-CRP>0.6, and a 39.1% flare rate if using DAS28 >3.2. Mostly, these flare rates are higher than our reported flare rates, but are similar to previous mentioned trials. This indicates that our criteria were more strict than other studies, but that flare rates are comparible between the tapering studies.
Also, the flare duration was longer in the TARA trial compared with other trials, which could be due to the measurement intervals of 3 months. If patients did not have a controlled disease 3 months after flare, we assumed that the duration of flare was 6 months. That might be a reason that our results seem to have a long flare duration compared with the Dose Reduction or Discontinuation of Etanercept in Methotrexate-Treated Rheumatoid Arthritis Patients Who Have Achieved a Stable Low Disease Activity-State (DOSERA) study or DRESS study, in which they knew the exact duration of flare in weeks. 12 13 In this study, there are several strengths and limitations. Strengths of the study are that we performed a randomised controlled trial to asess tapering in patients with RA with a controlled disease. The TARA trial is one of the first trials which assesses the differences in tapering strategies, and elaborates on current viewpoints concerning tapering treatment, instead of only determining if tapering is feasible or not.

Rheumatoid arthritis
Some limitations should be noted as well. First of all, inclusion was terminated earlier due to difficulties with recruiting. This was due to the initial inclusion criteria being too strict (DAS ≤1.6), and the start of another trial (POET study) which used the same pool of eligible patients. 7 The study sample size was based on a 20% difference between both tapering strategies resulting in 96 patients per arm. We found, however, a 10% difference with 85 patients in the csDMARD and 89 patients in the TNF inhibitor arm. This resulted in a power of 70% instead of 80%. For this reason, we performed a worstcase scenario analysis to see if our results were valid. We used the following assumptions: (1) all extra included patients in the csDMARD tapering group had no flares and (2) all extra included patients in the TNF inhibitor tapering group flared. This analysis showed an 18% difference in flares, which is still below the 20% difference on which our power calculation was based. Therefore, we think our current results and conclusions are valid.
Second, rheumatologist could have only referred patients who achieved low disease activity quickly and had less severe disease and, therefore, creating selection bias. However, we think that our target population is the same as the one we would apply our results to, because those are the patients who are suitable for tapering and are willing to taper their medication. Furthermore, only research nurses, who did the DAS assessment, were blinded. Rheumatologists, therefore, knew the tapering strategy of their patients. This design was chosen to mimic daily practise as much as possible. However, it could be a possible source of bias, since rheumatologist might prefer one of the two tapering strategies and would possibly treat patients differently depending on the tapering strategy.
Third, the time frame of follow-up was only 1 year. Although the differences in flare rates were not significantly different between both tapering strategies, the largest difference was seen at 12 months. Data of the second year are needed to investigate if this difference will increase.
Last, we encountered 19% protocol violations, which could underestimate the effect of one of the two tapering strategies. We analysed the type of violations and we can conclude that most protocol violations were randomly distributed over the two treatment arms and were made due to a treat-to-target approach.
To ensure optimal rheumatic care in the future, efficient use of biological treatment is needed. 33 By tapering medication, costs can be reduced, especially when tapering bDMARDs. On the other hand, 38% of the patients in the TARA study flared within the first year, which may have a direct impact on patients' lives (i.e, worker productivity and unemployment). Therefore, it is important to know which tapering strategy is most cost-effective, which will be addressed in a follow-up analysis.
In conclusion, the TARA study showed that up to 9 months, flare rates of tapering csDMARDs or TNF inhibitors were similar. After 1 year, a non-significant difference in flare rates was found of 10% in favour of csDMARD tapering. Tapering TNF inhibitors was, therefore, not superior to tapering csDMARDs. From a societal perspective, it would be sensible to taper the TNF inhibitor first, because of possible cost reductions and less long-term side effects.