The 'Switch’ study protocol: a randomised-controlled trial of switching to an alternative tumour-necrosis factor (TNF)-inhibitor drug or abatacept or rituximab in patients with rheumatoid arthritis who have failed an initial TNF-inhibitor drug

Cochrane reviews provide clear evidence that the licensed TNFi drugs (etanercept, infliximab, adalimumab, certolizumab and golimumab) produce better outcomes in RA compared with placebo or treatment with conventional DMARDs[16‐19]. All these are in the same class of drug i.e. TNFi, but differ in several respects:

Despite the extensive benefits of TNF-directed biologic therapies, a significant proportion of RA patients fail to achieve sufficient response[23]. Two broad approaches can be employed to manage initial TNFi non-response; switching to an alternative TNFi therapy or use of another mechanism agent. Of the latter, rituximab, a B-cell depleting therapy, abatacept, and more recently, tocilizumab, have been licensed, although only rituximab is currently approved by the National Institute for Health and Care Excellence (NICE) at the TNFi-failure stage[24].

Current NICE guidance does not permit switching to an alternative TNFi as a second-line biologic therapy choice unless rituximab +/- methotrexate is contraindicated. Several early phase, uncontrolled studies and an initial, small, randomised study suggested benefit in switching between TNFi agents[25‐35]. A report of high ACR20 responses on an alternative TNFi agent in specific sub-group of patients[27] also indicates the potential value of and the need to explore this approach further. The rationale and argument for switching between different TNFi drugs was strengthened by a large, randomised industry-led efficacy study comparing golimumab with placebo. This phase III study of 461 patients who had previously received and either failed or were intolerant to one or more TNFi were randomised to placebo, subcutaneous golimumab 50 mg or 100 mg 4-weekly. Significantly higher ACR20 response rates at week 14 were observed in the 50 mg and 100 mg golimumab groups compared to placebo group (35% and 38% versus 18% respectively)[36].

A key benefit of the TNFi is their suitability in both seropositive and seronegative disease [to rheumatoid factor (RF) +/- anti-citrullinated peptide antibody (ACPA)]. This is in contrast with data implying the influence of antibody status and response rates in patients treated with rituximab (particularly at the TNFi-failure stage, see below) due to its distinct target and rationale for use (rituximab depletes the autoantibody producing B-cells)[37, 38]. It is therefore important not to prematurely discount an alternative TNFi drug as an effective therapeutic option, particularly in the context of such resistant and aggressive disease cohorts. In addition, patients with RA may have a co-existing immune-mediated inflammatory disease (for example, inflammatory bowel disease, psoriasis) that would also be amenable to treatment with a TNFi (with rituximab not as suitable and potentially toxic)[36, 39‐42]. Having the option of using a second TNFi in this scenario would be clinically more appropriate than having to potentially consider two different classes of bDMARDs.

Industry-led efficacy studies have demonstrated benefits of rituximab, abatacept and tocilizumab after TNFi failure[43‐47] although only rituximab is NICE-approved (and neither abatacept nor a TNFi has been compared to rituximab). Rituximab, however, is not appropriate for certain patients and may even lead to unpredictable responses or toxicity[48], or failure to respond (up to a third of patients). Furthermore, meta-analyses of rituximab suggests seronegative antibody status, seen in up to 25-30% of patients, appears to be associated with poorer response, particularly in the TNFi-failure trial; although this has not been formally tested[37, 44, 45, 49, 50]. Recent data on abatacept from an observational registry also suggests seropositive status may confer greater benefits to abatacept therapy[51].

A Swiss observational study[52] comprised 116 patients who had failed at least one TNFi agent and were switched to either an alternative TNFi therapy or to one cycle of rituximab with the results suggesting that rituximab was the more favourable treatment option. Aside from the small sample size, this retrospective study had several other design limitations with outcome taken from differing time-points and inclusion of all types of initial TNFi failure; in addition it was neither controlled nor randomised to treatment type. The observational studies MIRAR and SWITCH-RA have also reported the use of rituximab as a better strategy compared to an alternative TNFi drug following insufficient response to a first TNFi[53, 54]. The collaborative CERRERA registry[55, 56] has also suggested utility of rituximab but in contrast to the Swiss study, following TNFi toxicity as opposed to lack of efficacy[57]. Other observational studies comparing alternative TNFi with other bDMARDs, such as abatacept and tocilizumab as well as rituximab, also favour these therapies over the use of alternative mechanism TNFi as second line treatment[58‐60]. These results have also been consolidated by recent RCTs (preliminary data)[61, 62] and meta-analyses, which have failed to demonstrate superiority of one therapy over another[63], with European recommendations also confirming all currently licensed therapies as appropriate options[64].

It therefore remains unclear how best to utilise the alternative bDMARDs described above following initial TNFi failure. It is apparent that no universally effective treatment exists with the present approach, and clinicians treating patients in the absence of sufficiently strong data is unsatisfactory. The current reality of second-line bDMARD restricted to a single option (rituximab) in the UK seriously impedes effective management. This is particularly pertinent to patient sub-groups where alternative licensed therapies may seem more appropriate (e.g. in seronegative RA, concomitant immune mediated inflammatory diseases). This poses a significant problem to the NHS and is in conflict with the patient agenda. Despite several treatment options now being available, no good quality head-to-head comparisons investigating the efficacy of sequential biologic treatments have been conducted to date.

The SWITCH trial aims to provide clear guidance to clinicians. The results of this study will enable the development of a rational treatment algorithm and should enable more judicious and cost-effective management; in particular it will potentially allow personalised treatment regimens as opposed to switching all patients to only one available (and potentially unsuccessful) therapy, potentially leading to long-term net-benefits and improved patient outcomes.

Whilst more recent technology appraisal permits the use of abatacept, a T-cell co-stimulation blockade agent, and tocilizumab, an interleukin-6 receptor monoclonal antibody, as first-line biologic together with TNFi[65], TNFi, remains the predominant first-line bDMARD currently prescribed in the UK[66].

The general aim of the trial is to compare alternative-mechanism TNFi to rituximab, and abatacept to rituximab in terms of disease response, quality of life, cost-effectiveness, toxicity and safety over a 12-month (48 weeks) period. Each of the two comparisons (TNFi vs. rituximab and abatacept vs. rituximab) is considered to be of interest independently, and the trial aims to establish non-inferiority; therefore no adjustments for multiple comparisons have been planned.

To establish whether an alternative-mechanism-TNFi or abatacept are non-inferior to rituximab in terms of disease response at six months (24 weeks) post randomisation.

SWITCH is a pragmatic, phase IV, multi-centre, parallel-group, open-label, RCT comparing alternative mechanism TNFi with rituximab, and abatacept with rituximab in a total of 477 patients with rheumatoid arthritis who have failed an initial TNFi drug. Participants will be randomised to receive one of the following for a maximum of 48 weeks (interventional phase):

The British Society of Rheumatology (BSR) provides guidelines on the use of TNFi[67]. These guidelines include important exclusion criteria that are adhered to in clinical practice. These will also be applied in this study. The inclusion and exclusion criteria for randomisation into this study are detailed in Table 1.

Participants will be recruited from multiple research sites within the United Kingdom; some of the collaborating research centres have been initially selected under the guidance of the Arthritis Research UK’s Adult Inflammatory Arthritis Clinical Study Group (AIA CSG). In addition, potentially eligible patients may also be identified via Participant Identification Centres (PICs). The identified clinicians at these PICs will refer potential participants to the research team based in one of the participating research sites for assessment and possible recruitment to the trial.

Patients will be approached during standard clinic visits for management of their disease. Alternatively, patients identified by other means (such as waiting lists, registries, review of case records) may be sent the personalised Switch Invitation letter inviting them to take part. Patients will be provided with verbal and written details about the trial (Participant Information Sheet and Informed Consent Document). Patients will have as long as they need to consider participation. Assenting patients will be invited to provide informed, written consent before being registered into the trial and formally assessed for eligibility.

Patients who are eligible to take part in the trial will also be eligible to have a number of biological samples (blood and urine) taken for the Switch Trial BioBank. Participation will be discussed with patients at the same time as discussing their participation in the main trial. Patients who agree to have biological samples taken for the Switch Trial BioBank will be asked to sign an additional consent form.

Following written informed consent and prior to any trial related invasive or non-invasive procedures, patients will be registered into the study. All patients will undergo a screening assessment within 4 weeks prior to the baseline assessments to determine eligibility for the study.

Following registration, confirmation of eligibility and completion of baseline assessments and questionnaires, participants will be randomised in a 1:1:1 ratio to receive alternative mechanism TNFi, abatacept or rituximab. Treatment group allocation will use minimisation incorporating a random element, via a computer-generated programme, to ensure treatment groups are well-balanced for: centre; disease duration (<5 years or ≥ 5 years); non-response (primary or secondary); rheumatoid factor status (RF seropositive or ACPA positive) or (RF seronegative and ACPA negative)). Both registration and randomisation will be performed centrally using an automated 24-hour telephone system based at the Leeds Clinical Trials Research Unit (CTRU).

Participating research sites will be required to complete a log of all patients over the age of 18 with RA who have failed an initial TNFi agent and have been considered for the trial, but not registered for screening or randomised, either because they are ineligible or because they decline participation.

Treatment will be administered in the three arms as detailed in Table 2. Participants will receive the randomised treatment for a minimum of 24 weeks, after which, 24-week responders will continue treatment to 48 weeks. After week 48, the randomised treatment may be continued if response is maintained and local practice/guidelines permits on-going use of a non-NICE approved treatment if relevant. The observational phase constitutes the follow up period from week 49 to the end of the trial, with a maximum follow-up duration to week 96. The duration of the observational phase will therefore vary amongst the participants. After the observational phase participants will return to NHS routine care.

Participants who, in the investigator’s opinion, demonstrate an unacceptably high level of disease activity prior to week 24 may discontinue treatment if clinically indicated. These participants will be followed up as part of the observational phase of the trial. The DAS28 score[68] obtained at week 24 will be used for the primary endpoint.

All protocol-required assessments will be recorded on paper case report forms at each site.The trial visits are structured as detailed below (see also Figures 2,3 &4):

The primary outcome is the change in Disease Activity Score 28 (DAS28)[68] at 6 months (24 weeks). Secondary outcomes at weeks 12, 24, 36 and 48 are: the DAS28 score and the proportion of participants who achieve a reduction in DAS28 score of greater than 1.2 from baseline, Low Disease Activity Score (LDAS) rate[69] and remission rate[70], EULAR (European League Against Rheumatism) and ACR (American College of Rheumatology) response scores[69, 71], changes in scores and proportion of participants in each category of the Clinical Disease Activity Index (CDAI)[72] and Simplified Disease Activity Index (SDAI)[73], the proportion of participants that achieve ACR/EULAR Boolean remission rate[70] at each time-point. The outcomes relating to quality of life at weeks 12, 24, 36 and 48 are: the RA Quality of Life (RAQoL)[74], the Hospital Anxiety and Depression Scale (HADS)[75], and the Health Assessment Questionnaire Disability Index (HAQ-DI)[76]; the HAQ-DI will also be evaluated at weeks 60, 72, 84 and 96. The outcomes required for the cost-effectiveness analysis and collected at weeks 12, 24, 36 and 48 are: EQ-5D[77], Health Utilities Index (HUI)[78], Health and Social Care Use & Expenditure due to Rheumatoid Arthritis[79]. EQ-5D and Health Utilities Index will also be evaluated at weeks 60, 72, 84 and 96. Further outcomes correspond to safety (adverse events and reactions) and toxicity (requiring cessation of treatment) reported throughout the duration of the trial treatment (up to week 48). In addition, outcomes related to radiographic measures at week 48 will be: changes in Genant-Sharp scores[80] of hands and feet, and bone densitometry T-scores of neck of femur and lumbar spine.

A total of 477 participants will be recruited to this study. A total of 429 evaluable participants are required to have 80% power for demonstrating non- inferiority of either abatacept or alternative mechanism TNFi to rituximab at the 5% significance level. A total of 143 evaluable participants in each treatment group will ensure that the lower limit of the two-sided 95% confidence interval for the true difference in DAS28 (abatacept/alternative mechanism TNFi – rituximab) lies above -0.6 units, assuming no difference between treatment groups and a between-participant standard deviation of 1.8 units[44]. Allowing for a loss to follow-up of 10%, a total of 477 participants will be recruited.

All analyses will be conducted on the Intention-To-Treat (ITT) population, where patients will be included according to the treatment to which they were allocated at randomisation. A Per-Protocol (PP) population will also be defined for the non-inferiority analyses, which will exclude participants who violate the protocol or fail to comply with the required treatment regime. Non-inferiority will need to be demonstrated in both ITT and PP populations in order to infer non-inferiority. All formal analyses will be carried out at a 2-sided 5% level of significance.

An interim analysis will be conducted after 239 participants have completed 24 weeks of follow-up to allow for early stopping of a treatment arm; specifically if either abatacept or alternative mechanism TNFi is shown to be inferior to rituximab, which will be based on the confidence interval excluding the value zero.

Multiple-variable linear regression will be used to compare the alternative mechanism TNFi and abatacept to rituximab with the dependent variable, the change in DAS28 at 6 months (24 weeks), and the minimisation factors (centre, disease duration (<5 years, ≥5 years), rheumatoid factor status, primary/secondary non-response) and the baseline value of DAS28 included as independent variables. The mean treatment differences, 95% CIs and p-values from this analysis will provide the main comparisons for each treatment group with rituximab.

Alternative mechanism TNFi and abatacept will be compared to rituximab at 12, 24, 36 and 48 weeks using the following methods:

To determine if there is a differential response according to TNFi type (monoclonal antibody or fusion protein) initially failed, a linear regression model will be fitted to DAS28 at 24 weeks including baseline DAS28, type of TNFi initially received and minimisation factors as independent variables. To determine if there is a differential treatment effect according to primary and secondary failure to initial TNFi received, a linear regression model will be fitted to DAS28 at 24 weeks on baseline DAS28, treatment, primary/secondary failure, remaining minimisation factors, and an interaction term between treatment and type of failure. Finally a linear regression model will be fitted to DAS28 at 24 weeks including baseline DAS28, treatment, rheumatoid factor status, remaining minimisation factors, and an interaction term between treatment and rheumatoid factor status in order to assess if there is a differential treatment response between seropositive and seronegative patients.

The economic evaluation aims to assess overall cost-effectiveness from the perspective of the health system (NHS) and patients. It will consist of a within-trial cost effectiveness analysis and a decision analytic cost effectiveness model. The within-trial analysis will evaluate the costs and outcomes of the patients recruited to SWITCH for the follow-up of the trial. As with the primary analysis, the economic evaluation will be an intention to treat analysis. The outcome used in the primary analysis will be the Quality Adjusted Life Year. Using the NICE cost effectiveness threshold of 20,000 per QALY[81], we will convert costs and outcomes for each patient on to the Net Benefit scale and use linear regression analysis to estimate the expected Net Benefit of the trial interventions compared to current practice[82]. Analysis of uncertainty will be undertaken using the non-parametric bootstrap, to characterise the uncertainty in the estimates of Net Benefit.

A second analysis will synthesise the data from the SWITCH trial with existing evidence to estimate the lifetime expected net benefit of the trial interventions compared to current standard care. The perspective for this analysis will be the same as for the within trial analysis. In order to capture the switching nature of the treatment pathways, we will construct patient level simulation model, rather than the Markov cohort model, which is frequently used for decision analytic cost effectiveness analyses.

The primary modelled analyses will adopt the perspective of the NHS and Public Social Services. Secondary analyses will adopt a broader perspective incorporating carer quality of life and cost impacts, and productivity costs.

Resource utilisation will be captured at each follow-up visit. Personal expenditures related to the management of RA, and time spent away from work by the patient and carers will be collected using the Cost Diary, a validated questionnaire[79]. Unit costs will be taken from routine national databases such as the British National Formulary, the NHS Reference Costs and the Personal Social Services Research Unit (PSSRU) costs of health and social care[83]. Health-related quality of life will be captured using the EQ-5D supplemented by the Health Utilities Index. These data will be collected at baseline and at each clinical follow-up. Parameter uncertainty will be addressed through probabilistic sensitivity analysis. For the within trial analysis this will be done using the non-parametric bootstrap; for the decision analytic cost effectiveness model this will be done using Monte Carlo simulation.

Outputs from the analyses will be presented as Expected Incremental Cost Effectiveness Ratios; Cost Effectiveness Acceptability Frontiers, Expected Net Benefit[84] and Net Benefit Probability Maps[85]. In addition to the primary analyses, secondary analyses adopting different perspectives, different utility measures and different approaches to dealing with missing data will be presented. The final set of analyses will present estimates of the global and partial value of perfect information, to inform future research.

The first patient was enrolled into SWITCH on the 31st July 2012 and recruitment is due to end in December 2016. The study is being conducted in multiple NHS sites across the UK, with a planned total of up to 50 sites. We expect to report the main trial results in Autumn 2018. Ethical and governance approval for this trial has been obtained from the Leeds West Ethics Committee (ref 11/H1307/6) and the Leeds Teaching Hospitals NHS Trust respectively. The trial progress is monitored by an independent Data Monitoring and Ethics Committee (DMEC) and Trial Steering Committee (TSC).

Since opening, the trial has undergone a major trial re-design. Our original target sample size was 870 patients, to have 80% power for determining whether abatacept or alternative TNFi were non-inferior to rituximab at 24 weeks post randomisation in terms of achieving a DAS28 reduction of greater than 1.2 points without toxicity. The corresponding non-inferiority margin was set at 12% and assumed a response rate of 65% in the rituximab arm. The original trial design was also powered for a definitive sub-group analysis to determine if there is a differential treatment response between seropositive and seronegative patients Following challenges in recruiting patients and securing site participation, as well as re-discussion of meaningful endpoints, a decision was made to re-design the trial by modifying the primary outcome measure from a binary to a continuous outcome (which was also deemed clinically relevant). This allowed a reduction in sample size to 477 patients whilst still ensuring a trial of clinical relevance. The previous planned definitive sub-group analysis is now an exploratory analysis. The trial re-design was unanimously supported by the DMEC and the TSC, approved by the funder and has received favourable ethical opinion.