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

There is broad evidence on the favorable risk/benefit of anti-TNF agents for the treatment of axial spondyloarthritis in patients who do not respond to NSAIDs [14]. It seems plausible to believe that low maintenance doses of anti-TNF might suffice to control the disease in patients in remission or with very low inflammatory activity [18‐20]. Based on this biological plausibility, empirically based anti-TNF dose reduction during maintenance treatment is a common off-label practice in routine patient care, sometimes formally supported by recommendations from scientific societies. It is not based, however, on solid clinical evidence from randomized trials. The REDES-TNF study aims to confirm whether the dose-reduction strategy is as effective and safe as full-dose maintenance strategies, while reducing the healthcare burden and costs.

The study has several strengths, including the large number of sites involved and the centrally randomized design requiring entry of patient data in the eCRF before randomization, thus minimizing the risk of errors in eligibility and of biases in the randomization process. The eligibility requirements include axial involvement and remission after induction for at least 8 weeks (2 routine controls), and applies strict remission criteria - according not only to clinical symptoms but also to biologic parameters (CRP), in order to ensure both disease stability and sample homogeneity, thus maximizing the sensitivity of the study in detecting differences between treatments.

The high-level comparison of interventions, which includes all anti-TNF agents available in Spain during the study period, is intended to provide testing of the general concept behind dose-reduction in a low-intensity inflammation setting. The selection of the dose regimens to be tested in the dose-reduction arm was made according to existing recommendations [27] and has been refined by systematizing the clinical experience of the participating rheumatologists; thus, it can be considered as representative of current off-label clinical practice. In general, the change to dosage reduction has been made by increasing the dosing interval by 50 %, except in the case of infliximab, where - based on previous reports [24, 27] and clinical experience - a reduction in the dose per kg while maintaining the dosing interval of 8 weeks was considered more suitable.

The design of the study is pragmatic, with the intervention limited to random assignation to one of two dosing options but, otherwise, patient follow-up and assessments are those currently used in routine patient care. The study treatment is prescribed and dispensed following routine outpatient procedures. For this reason, the study results are expected to have high external validity. The choice of a noninferiority design is suitable to the clinical situation, where a new strategy is now being applied empirically, and a “saving” therapeutic strategy is promoted by healthcare payers despite the lack of evidence and of regulatory endorsement in the summary of product characteristics.

The noninferiority (delta margin) of 17 % was set based on the consensus on clinical relevance reached by the rheumatologists involved, who decided that a proportion of patients with acceptable control <70 % after 1 year would severely discourage the use of dose-reduction. The main variable is based on the acceptable therapeutic goal instead of the optimal therapeutic goal. While complete resolution of signs and symptoms is a desirable goal, the definition of acceptable therapeutic goal represents the degree of residual signs and symptoms that, clinically, may not merit a change in the therapeutic strategy; thus, with respect to medical decision making, the acceptable therapeutic goal is a more useful primary variable than the optimal therapeutic goal. A thorough assessment of treatment safety during the study may help to establish a comparative risk/benefit profile of the two options, which is a key point in the indication of long-term treatment. Finally, the collection of biological samples during treatment and when there is disease relapse may serve as a basis to predict the clinical or biological features of patients who may be at higher risk of failure, which could be for future guidance.

Statistically, the choice of the FAS as the primary analysis set is generally not regarded as conservative [33] because it is considered that the PP conditions most closely reflect the scientific model and more reliably replicate the conditions in which active control is likely to be effective, thus ensuring the sensitivity of the assay. However, ideally both analyses should have equal importance and should lead to similar conclusions [45]. Therefore, although analysis of the PP set will be the predefined primary analysis, the principal end-point and the key secondary end-points will also be tested in the RS and FAS populations. Losses to follow up should be minimized in order to avoid bias by ensuring that patients are followed further even if the assigned treatment strategy is modified. However, continuous efficacy variables with repeated measurements will be analyzed using MMRM. This approach is robust to the presence of data missing at random, and makes the analysis in all subjects despite missing values. With this method, estimates are calculated based on the variance-covariance structure but without any formal imputations.

The primary endpoint was initially intended for analysis by the Mantel-Haenzel method [46]. However, since the noninferiority margin was predefined and justified on clinical grounds using a risk difference scale, we prospectively amended the statistical method to directly address testing against the margin on the same scale. Therefore, noninferiority will be checked by testing the estimated treatment of the one-sided 97.5 % CI of the risk difference against the noninferiority margin of 17 % using a log-binomial regression model as detailed in the statistical section. The main outcome will also be analyzed using the Mantel-Haenzel method predefined in the protocol for sensitivity purposes.

Considering the study operations, the strategies compared, full and reduced dose, are currently considered as equally appropriate for the treatment of patients with spondyloarthritis and are widely used. This may represent a challenge to recruitment because, due to financial pressures in healthcare, many patients will have already undergone dose reduction. In fact, one of the main challenges of the study is to meet the recruitment goal on time in order to ensure that all patients are assessed for the main efficacy outcome before study funding ends.

The study has a number of potential limitations. First, the study is not blinded, since it was thought that the complexity, costing and risk of medication errors associated with the use of double dummy (different doses in iv injections for infliximab, placebo subcutaneous injections matching the alternative posology for adalimumab, etanercept and golimumab) did not outweigh the benefits of blinded assessment of outcomes, especially for the subcutaneous treatments, which would have required an increase in the number of injections over a long time period. Likewise, blinded assessors will not be used because the decision to modify treatment in the event of clinical flare will mainly be based on the patient’s reporting of disease signs and symptoms (BASDAI score and axial night pain). Therefore, it is expected that if the patient’s reports are biased due to knowing the identity of treatment, they will tend to consider the low-dose treatment as less efficacious than the full dose treatment. Thus, the bias would likely be conservative, contrary to the main study hypothesis, and reflective of what may be expected after dose reduction in routine clinical practice. Second, the pragmatic approach means the use of NSAIDs and DMARDs will not be standardized and may be varied by the investigators as required during the study to control symptoms. While this may be regarded as a potential source of confusion, the use of these additional treatments is quantified during the trial and will be analyzed to determine whether either of the arms is associated with an increase in the use of concurrent anti-inflammatory drugs or DMARDs. Third, patients will be stratified only according to the anti-TNF drug, in order to maintain a reasonable number of strata. Additional factors with potential prognostic implications (for example, time since first diagnosis, use of DMARDs, duration of clinical remission at randomization) were not considered. Although these factors have been reported to predict a clinical response to anti-TNF therapy [47, 48], until now there is no data on their value in predicting the clinical result of anti-TNF dose reduction. Fourth, the duration of follow-up until the main efficacy assessment is limited to one year. Although this period may be considered a reasonable timeframe in which to test the clinical acceptability of each therapeutic option, the study will not be able to provide information to guide clinical decisions after this time. Additionally, it is unlikely the study will be able to detect relevant differences in structural end-points by spine imaging because changes are generally slower in axial spondyloarthritis. It is also anticipated that the study will not be able to detect differences in efficacy or safety between the dosages of each anti-TNF studied.

In summary, the REDES-TNF study is a pragmatic clinical trial sponsored by two medical scientific societies that aims to answer the need for evidence to support medical decisions now taken empirically. The results of the trial may be useful in guiding the management of patients with spondyloarthropathies requiring anti-TNF treatment.