Infliximab versus alpha interferon in the treatment of Behçet's disease: the BIO BEHÇET'S RCT

Article history

The research reported in this issue of the journal was funded by the EME programme as award number 12/205/46. The contractual start date was in January 2015. The draft manuscript began editorial review in May 2022 and was accepted for publication in August 2023. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The EME editors and production house have tried to ensure the accuracy of the authors’ manuscript and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this article.

Copyright statement

Copyright © 2024 Moots et al. This work was produced by Moots et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This is an Open Access publication distributed under the terms of the Creative Commons Attribution CC BY 4.0 licence, which permits unrestricted use, distribution, reproduction and adaptation in any medium and for any purpose provided that it is properly attributed. See: https://creativecommons.org/licenses/by/4.0/. For attribution the title, original author(s), the publication source – NIHR Journals Library, and the DOI of the publication must be cited.

2024 Moots et al.

Background

Behçet syndrome (BS; also called Behçet’s disease or, simply, Behçet’s) is a systemic inflammatory vasculitis of unknown aetiology, characterised by recurrent episodes of acute inflammation in a variety of organs, typically including mucus ulceration in the mouth and genitally, but also manifestations in other organs from the skin to the eyes, where it can cause blindness. 1,2 These clinical features, the absence of specific autoantibodies and its spontaneously relapsing and remitting nature have led to its characterisation as a polygenic autoinflammatory disorder. It is considered to be a very rare disease in the UK. The early estimates of prevalence in the UK ranged from 0.27 to 0.64 per 100,000, but a recent unvalidated population-based estimate utilising a large primary care database suggested a higher range of between 12 and 14/100,000, with the rider that this may be an overestimate as a proportion of such cases will not have been validated by an expert multidisciplinary team (MDT). Behçet’s disease is up to 10-fold more prevalent in the Middle East and Far East along the ‘silk route’ to Japan and increases towards Southern Europe.

There is considerable variation in its clinical presentation between and within individuals, and there appear to be differences in disease manifestations and response to therapy in patients in the UK compared to those in southern European or Far East countries. While little is known about the underlying pathophysiological processes, recent years have witnessed the successful application of biologic therapies, with good outcomes in patients who had not previously responded to standard therapy with steroids and/or immunosuppressants such as azathioprine. 3–5 Much of the evidence base for biologic therapy has arisen from case series or other uncontrolled trials performed in countries outside the UK – where the phenotype appears to differ. For example, more severe ocular disease is reported in patients in Turkey and Japan compared to the UK and Western populations, where mucocutaneous and gastrointestinal (GI) manifestations are more prevalent. Accordingly, most trials have reported effects on ocular disease. Very few data are available to inform useful systematic reviews, and most guidelines therefore stem from expert-based consensus. 6

The establishment of three National Centres of Excellence for Behçet’s disease in England7 led to the creation of a national cohort of patients with agreed pathways for assessment and treatment by MDTs, comprising the specialist and support staff needed to cover the wide spectrum of organ involvement and the ability to fund biologic therapy when indicated. At the time of the trial, it was generally considered that only two biologics, infliximab and alpha interferon (Roferon), had sufficient evidence to support their use as first-line biologics in refractory disease. Other biologics, such as alemtuzumab, with a less favourable adverse event (AE) profile, were reserved for patients who were refractory to or intolerant of infliximab and alpha interferon.

Infliximab is a mouse/human chimeric monoclonal antibody originally developed for use in rheumatoid arthritis that works by neutralising tumour necrosis factor (TNF) alpha. Its long-term safety record is well established in rheumatoid arthritis,8 and its utility in Behçet’s disease reported, typically in uncontrolled studies and in countries outside the UK. 3 In 2011, Arida et al. conducted an extensive PubMed/MEDLINE search on the published experience of 375 patients treated with a TNF-inhibitor for Behçet’s disease9 and with a variety of organ systems involved. Of these, the vast majority (325) were treated with infliximab, and all had been inadequately controlled with, or were intolerant to, other immunosuppressive regimens such as glucocorticoids, azathioprine and ciclosporin. Sustained organ-specific clinical responses were evident in 90%, 89%, 100% and 91% of patients with resistant mucocutaneous, ocular, GI and central nervous system involvement, respectively. None of those trials were randomised or placebo controlled.

At the time of the trial, Roferon was used in several inflammatory and rheumatic disorders, including Behçet’s disease, and was well summarised by Kotter et al. in 2010. 10 Much of the reported experience of alpha interferon in Behçet’s disease originated from uncontrolled studies in patients with ocular disease. Kotter et al. reported in 2003 an open, non-randomised, uncontrolled prospective study using Roferon in 50 patients with inflammatory eye disease due to Behçet’s disease. 11 An overall ocular response rate of 92% was reported. A rapid response: with the posterior uveitis score of the affected eye falling by 46% per week and full remission achieved by week 24. A retrospective single-centre uncontrolled series reported by Bodaghi et al. 12 of ocular Behçet’s disease also reported a high response rate of 82.6%, with other groups reporting similar findings. Reports of controlled trials of alpha interferon and its efficacy in extraocular disease are limited. Alpsoy et al. 13 published a randomised placebo-controlled study of 50 patients with mucocutaneous Behçet’s disease randomised to alpha interferon or placebo, reporting that alpha interferon was effective in the management of mucocutaneous lesions, with a trend towards improvement of joint symptoms. The formulation of and dosing regimens for alpha interferon have varied between these studies, most utilising the preparation Roferon, with short half-life and more frequent administration. The subsequent development of pegylated alpha interferon (Viraferon Peg)14 allowed dosing once a week. However, a UK prospective trial evaluating Viraferon Peg in Behçet’s disease reported only modest benefit, with responses far less than that for Roferon. 15

Assessment of disease activity: As a multisystem disease with lack of laboratory biomarkers to determine activity and limited disease-specific outcome measures, devising and undertaking a clinical trial of active medicinal products in Behçet’s disease is challenging, and international standards for this have not yet been adopted. Mumcu et al. 16 summarised the various methods used to measure overall clinical activity. The International Scientific Committee on Behçet’s disease produced the ‘Behçet’s Disease Current Activity Form’ (BDCAF), with investigators from five countries participating. 17 Thirty dichotomous questions were reduced to a Behçet’s disease activity index (BDAI) lying between 0 and 12 but then transformed to a 0–20 scale for international comparisons. In Iran, the Iranian Behçet’s Disease Dynamic Measure is used. 18 The Behçet’s Syndrome Activity Scale was developed as a patient-reported outcome measure, which correlates with the BDCAF. 19 A Behçet’s Disease-specific quality-of-life measure (BD-QoL) was derived by the Psychometric Group in the Leeds Institute of Rheumatic and Musculoskeletal Medicine, Leeds University; it consists of 30 easily answered dichotomised questions. 20 Other disease activity measurements have been proposed for specific organ symptoms. We chose to use a slightly modified form of the BDAI as the primary outcome for the proposed clinical trial, as this is a verified measure of the overall severity of the disease and is routinely used in the three UK Centres of Excellence.

As in other complex chronic diseases, there is a need in BS to not only better understand the phenotype of patients in clinical trials but also to explore the potential to identify biomarkers that may help target therapies and minimise AEs. No such biomarkers are currently available. We, therefore, built into the study the exploratory analysis of two potential genetic and metabolomic biomarkers that focused on potential responses to alpha interferon and infliximab, respectively. Three genome-wide association studies in patients with hepatitis C virus genotype 1 infection implicated single nucleotide polymorphisms (SNPs) in the vicinity of the interferon lambda 3 (IFNL3) [interleukin 28B (IL28B)] gene on chromosome 19q13.13 with response to alpha interferon therapy. 21–23 Patients with the CC genotype at rs12979860 had higher response rates to alpha interferon. 24 A recent parallel sequencing study25 was able to show that rs4803221 and rs7248668 predicted failure to respond better than rs12979860. IFNL3 encodes a lambda type of infliximab, while the SNP at rs12979860 affects interferon-stimulated gene production as part of the innate immune response, but the actual mechanism is unclear. 26 Despite this, treatment algorithms incorporating IFNL3 genotyping are now used in many clinics for the treatment of hepatitis C. 27 A recent study28 showed that rs12979860 is in linkage disequilibrium with a frameshift variant, ss469415590[ΔG], which also creates a new gene, interferon lambda 4 (IFNL4), reduced expression of which may be associated with reduced responsiveness of cells to alpha interferon 8. Whether the same SNPs affect response to alpha interferon in other diseases is unclear, but given the role of the innate immune system in the pathogenesis of BS,29 it was biologically plausible that a similar effect to that seen in hepatitis C with alpha interferon may be operating in BS. We intended to test this hypothesis in this trial.

To our knowledge, no convincing genetic predictors had been identified through genome-wide association studies as determinants of response to infliximab. We therefore chose initially to pursue a metabolomic route to explore this. Nuclear magnetic resonance (NMR)-based metabolomics allows the examination of the changes in hundreds or thousands of low-molecular-weight metabolites in an intact tissue or biofluid and offers several distinct advantages in a clinical setting since it can be carried out on standard preparations of blood cells, serum, plasma or urine. Pattern recognition techniques are applied to the NMR spectra of samples taken from individuals. Metabolomic analysis was able to distinguish between patients with rheumatoid arthritis who responded to anti-TNF therapy compared to those who did not, with a sensitivity of 88.7% and a specificity of 85.9%. 30 We have previously shown that metabolomic analysis of vitreous humour could separate with high sensitivity and specificity samples from patients with two inflammatory conditions, lens-induced uveitis and idiopathic chronic uveitis with urea and oxaloacetate levels associated with the different conditions. 31 There is therefore a sound rationale for exploring similar methodology in the current study of patients with BS.

Assessing biomarkers as part of this trial also promised the potential to not only help in identifying determinants of response but also provide insights into the potential mechanisms of action of these biologics in BS patients.

Rationale

Behçet syndrome is associated with significant morbidity and mortality in the UK and abroad. In the UK, it can take up to 12 years to diagnose, leads to blindness and stroke, often does not respond to simple immunosuppressive therapy and, as we have previously reported, has a major impact on quality of life (QoL). 32 Although the biologic drugs infliximab and alpha interferon have been reported to be effective in refractory BS, they are expensive [at the time of starting the trial, the cost of infliximab was < £20,000/year and alpha interferon (Roferon) was £4000/year] and have not been subjected to rigorously undertaken randomised controlled trials (RCTs) compared directly against each other for efficacy and safety. Evidence for their efficacy arose from uncontrolled studies in other countries, where the disease phenotype may exhibit subtle differences from that in the UK. 6 Funding for biologic drugs for BS in England is held by the three National Centres of Excellence, allocated from highly specialised NHS commissioning. 33 Anecdotal experiences from the three centres of excellence also indicated that response rates to the individual biologics used in this study might differ when used within a UK population compared to results reported from other international cohorts. While these biologic drugs were generally considered effective for patients with BS in the UK, the national centres anecdotally observed efficacy rates to differ from those reported in other countries, with variable and unpredictable responses, and therefore there was a poor evidence base on which to inform clinical decisions. The identification of novel biomarker(s) predicting response to biologic therapy was deemed necessary to allow a more precision-based approach to treatment. A polymorphism in the IFNL3 (IL28B) was predictive of reduction in viral load in response to alpha interferon in hepatitis B or C infections. 21,34,35 As similar alpha interferon-mediated pathways of innate immunity are involved in BS, the potential effect of IFNL3 SNP on response to therapy with alpha interferon may well be relevant in BS. Similarly, metabolic analyses of urine samples from patients afforded the potential to provide biomarkers for treatment response to infliximab and/or alpha interferon.

Objectives

The aim of the study was to create the evidence base to underpin clinically effective prescribing of the biological drugs infliximab and alpha interferon for BS.

The objectives of the study were to:

1. Undertake a RCT to compare infliximab versus alpha interferon in patients with BS who were unresponsive to standard oral therapy.

2. Examine whether IFNL3 and IFNL4 SNPs can predict responses to alpha interferon and/or infliximab in BS.

3. Examine the potential for urine metabolomics to act as biomarkers for drug responses to infliximab and/or alpha interferon in BS.

Trial design

Bio Behçet’s was a randomised, two-arm, parallel design comparing the efficacies of infliximab versus alpha interferon. The population was patients with refractory disease eligible for the first biologic drug. Patients were recruited from the National Behçet’s Centres and supporting clinics in England and randomised to the two arms of the trial with stratification by centre. A total of 80 patients were to be randomised on a 1 : 1 ratio for arms. Recruitment was scheduled to take approximately 36 months. Assessments were made at baseline, 12, 24 and 36 weeks. The end of study was when the final patient completed their 6-month follow-up assessment.

Study setting and study population

Trial interventions

Schedule of trial procedures

The schedule of trial procedures are listed in Table 2.

Data capture

Trial data were captured using electronic Case Report Forms (eCRFs) and transcribed to a MACRO database. This database was designed and maintained by the LCTC, which was also responsible for the randomisation. The eCRF was the primary data collection instrument for the study.

All eCRFs were entered directly into a MACRO database and accessed via a secure web page by research site staff and the clinical trial co-ordinator at LCTC. The client application was secured with a unique username/password combination allocated to each delegated member of the research team. When data were entered into an eCRF, it was electronically stamped with the date, time and the person who entered it. If data were changed on an eCRF, it was electronically stamped with the change and would be accompanied with the date, time, person and a reason for making the change or correction. The previous value was recorded in an audit trail for each data item.

Each eCRF contained specific validation checks on the data being entered. If any values were outside what was expected, or data missing, this was flagged up and raised as a discrepancy on the main database system. Regular reports were generated to identify discrepancies in the data and allow for follow-up. Comprehensive guidelines for eCRF data entry were provided to all staff who have been delegated the responsibility for data collection. Where the site was unable to upload data using the eCRF, a backup paper CRF was available to use and accessed from the LCTC portal. In such cases, the site research staff would enter the data onto the trial MACRO database following the assessment.

Electronic and paper screening logs were kept in clinics to record the number of patients declining participation and, when volunteered, the reason given. All data were kept in a secure, locked location on NHS premises. All routine eCRFs were to be completed within 14 days of the study visit occurring.

Paper versions of the CRFs were available for download from the LCTC website, www.LCTC.org.uk and used as an aid to research staff. Quality control (QC) processes, including onsite source data verification for primary and secondary end points, were put in place in line with the eCRF platform.

Biological samples

Outcome measures

Sample size

The primary outcome was a modified version of the BDAI after 3 months of therapy, which will range from 0 to 30 for a patient.

If a traditional frequentist equivalence design were to be used, then based on equivalence being defined as the difference in means being < 20% (i.e. 20% of mean BDAI of 10 = 2), then for significance level, 0.20 and power 90%, a sample size of 176 (88 per arm) is required. (Here we have assumed standard deviation of 4 for BDAI at 3 months, a difference in means of 0.5, in accordance with the opinions of the international experts recruited for the Bayesian design. Also, baseline measurements have not been taken into account which would be expected to reduce the sample size to some extent.) As the recruitment of this number of patients is not feasible, the Bayesian design is adopted.

Bayesian design: Analysis of the data obtained from the small survey of experts both from the UK and regions of high prevalence of BS internationally, described in Chapter 4 (Research Design), gives a prior distribution of the difference in mean values of BDAI as N (0.52, 1.062) and < 24% difference in means to define equivalence. The mode for the latter was 20%, and this value is used in the sample size calculation as it fits better with FDA guidelines.

If the difference in means, D, of BDAI at 3 months is considered without the use of baseline BDAI, then (assuming the above) prior for D and a normal distribution, N (10, 42) for the distribution of the 3-month BDAI scores. For a sample size of 45 per arm, the Bayesian power based on an equi-tailed 80% credible interval for testing for equivalence is 0.71.

To be more accurate, a simulation exercise was carried out using R and WinBUGS to establish the sample size for the analysis of covariance (ANCOVA) model. For one arm, random baseline and 3-month BDAI scores were generated from a bivariate normal distribution with mean vector (12, 10), variances 4.0 for both and correlation r. For the other arm, random baseline and 3-month BDAI scores were generated from a bivariate normal distribution with mean vector (12, 10+ m), variances 4.0 for both and correlation r. For each simulation, r was randomly chosen from a uniform distribution (0.05, 0.5) and m from a N (0.52, 1.062) distribution. For a sample size of 45 per arm, testing for equivalence using an 80% equi-tailed credible interval calculated from the posterior distribution, Bayesian power of 91% was obtained. When a 90% credible interval was used, the Bayesian power dropped to 73%.

Using this design with the 80% credible interval, a sample size of 45 patients per arm was deemed suitable, which, allowing for 10% dropout, requires 100 patients to be recruited.

Following recommendations to reduce the overall length of the study, the study was amended to recruit a total of 80 patients. Including a 10% dropout rate, estimates of study power are based on 72 evaluable patients (36 on each arm). Here, when an 80% equi-tailed credibility interval is used, a power of 88% is obtained; when a 90% credible interval is used, the power drops to 71%.

Randomisation

Stratified block randomisation, based on randomly permuted blocks with random block sizes of 2 and 4, was employed. The randomisation code list was generated by the LCTC trial statistician with the software package Stata^® using the ‘ralloc’ statement. The trial was open-labelled. The stratification factor included in the design is Centre. Data from the eCRFs will be entered onto a MACRO4 database with extensive data validation checks alerting all missing data to be queried. Missing data were monitored, and strategies were developed to minimise its occurrence. Central statistical data monitoring will summarise missing or inconsistent data periodically.

Analysis

Primary outcome

Alpha interferon. A Bayesian ANCOVA model will be used:

y = β0 + β1 × x + α × tr + ε,

where y = BDAI at 3 months, x = baseline BDAI, tr = 0 if a patient is in the infliximab group and tr = 1 if a patient is in the alpha interferon group. β0, β1 and α are the parameters to be estimated, and ε is the error term with variance σ² (to be estimated). The parameter of particular interest is α, as it measures the difference between the two treatment groups. Prior distributions will be placed on β0, β1, α and σ² WinBUGS will be used to fit the model.

Prior information

Vague priors for β0 and β1 were set as following a normal distribution with mean zero and a larger variance [i.e. ~ N (0.0, 100,000)]. The prior distribution for σ was set as a uniform distribution with limits of 0 and 3, respectively [i.e. ~ U (0, 3)].

The prior distribution for alpha is based on data obtained from a small group of international BS experts using the question:

On the assumption that the average BDAI score for patients treated with infliximab is 10, share out 100 points on the following scale on how better or worse alpha interferon is compared to infliximab at relieving/controlling BS symptoms.

Alpha interferon better	Alpha interferon worse
Scale: 21% + 16–20% 11–15% 6–10% 0–5%|	0–5% 6–10% 11–15% 16–20% 21%+
For example 0 0 10 20 3 20 10 10 0 0

The experts’ answers revealed a mean of 0.053 and a variance equal to 0.0126. The results are also shown in Table 3 and Figure 1.

TABLE 3 - Results from experts’ survey

Difference	Expert 1	Expert 2	Expert 3	Expert 4	Expert 5	Expert 6
–25	0	0	0	0	0	0
–18	0	0	5	0	0	0
–13	10	0	10	5	5	0
–8	10	0	15	20	10	5
–3	15	0	20	20	10	10
3	30	5	20	5	10	20
8	15	15	15	25	10	30
13	15	20	10	25	10	20
18	5	30	5	0	10	10
25	0	30	0	0	35	5

FIGURE 1.

Prior distribution.

A second question assessed where a point of equivalence between the two drugs was reached. The question posed was:

Suppose you are generally prescribing just one of the two drugs (infliximab or alpha interferon) at the moment. If you were told that the efficacies of infliximab and alpha interferon are exactly the same, then presumably you would not change to prescribing the other drug. However, if you were told that the other drug is 40% more efficacious than the one you are currently prescribing, then you would presumably change. Somewhere between 0% and 40% you would probably change from prescribing the current drug to the other. What % would this be? (Ignore any other factors such as cost of drug. Percentages are based on a mean of 10 for the BDAI.)

The responses to the second question informed the boundaries as to where infliximab and alpha interferon can be considered to be equivalent and also where one is superior to the other.

Let the equivalence boundary be given by γ. Then the equi-tailed, 80% Bayesian credible region (α_L, α_U) obtained from the posterior distribution for α will be used to describe the difference in efficacy for the two treatments, guided by the following:

• if (αL, αU) lies between (–γ, γ), then equivalent. if αU < –γ, then infliximab is superior

• if α_U < –γ and α_U < γ, then infliximab could be equivalent or superior. if α_U < –and α_U > γ, then equipoise

• if α_U > –γ and α_U > γ, then alpha interferon could be equivalent or superior. if α_L > γ, then alpha interferon is superior.

Trial recruitment and disposition

Patients were recruited between June 2016 and February 2020. Recruitment by centre is summarised in Table 23 (see Appendix 1) and occurred at a linear rate that was slower than anticipated (Figure 2), resulting in the premature termination of the trial and fewer participants than originally planned. One hundred and sixty-one patients were screened, and seventy-nine patients randomised. The subsequent disposition of the randomised patients is summarised in Figure 3.

FIGURE 2.

Trial recruitment over time.

FIGURE 3.

Patient disposition. αIFN, alpha interferon; IFX, infliximab.

For the reasons outlined, the ITT analysis was restricted to 37 subjects allocated to infliximab and 37 to Roferon.

Assessment of data quality

Withdrawals from the study protocol within each treatment arm are summarised, together with their reasons and losses to follow-up, in Table 4.

A summary of study protocol deviations is provided in Appendix 1 of Table 24: Summary of study protocol deviations. Patients with a major protocol deviation were removed from analyses performed on the per-protocol data set.

Description of baseline subject characteristics

Baseline characteristics of the study population are summarised in Table 4 for the 74 trial participants. Mean age [interquartile range (IQR)] was 39.1(31.6–47.2) and did not differ significantly between the two treatment arms. There were 50 (68%) female and 24 male (32%) participants with similar proportions in each treatment arm by sex. Ethnic profile and baseline disease characteristics did not differ between treatment arms. Steroid use was also similar in each treatment arm (IFX 49%, alpha interferon 51%). These are detailed in Table 5.

Exposure to treatment and compliance

Summary information is provided in Appendix 1: describing patients’ exposure to treatment with infliximab (mean dose per patient and the percentage of patients continuing to receive treatment over time) and interferon (dose changes over time and mean number of missed doses recorded at weeks 12, 24 and 36 reviews).

Analysis of primary outcome measures

The primary outcome for the trial was defined as the change in mBDAI between baseline and 3 months (with 6 months as a secondary outcome). The statistical analysis examined the difference in mean mBDAI scores between the two treatment arms at 3 and 6 months, with clinically significant response defined as a difference in 20% or more between the two treatment arms (assessed using Bayesian ANCOVA for change in mean from baseline adjusted for baseline score). Analysis based on planned ITT therefore included 37 patients allocated to infliximab and 37 to Roferon.

Table 6 presents the results from the Bayesian linear regression model to estimate the impact of treatment group on mBDAI. The linear model includes as an adjusting covariate the baseline mBDAI, and so there are three parameters presented:

• β0: model intercept

• β1: covariate associated with baseline mBDAI

• α (alpha interferon vs. infliximab): impact of treatment group.

Results are presented in terms of model estimates [standard error (SE)] and the 80% credibility interval, which is consistent with the study design.

The results show that 3-month BDAI is reliant on the baseline BDAI, which is demonstrated by an 80% credibility interval which does not include zero for the β1 parameter. The impact of treatment is not statistically significant based on the 80% credibility interval [Est (SE) = 0.13 (0.25), 80% CI = (–0.19 to 0.46)]. So the two treatment arms did not differ significantly with respect to change in BDAI at 3 or 6 months. These results show that, after adjustment for baseline BDAI, patients who were treated with alpha interferon had a BDAI score of 0.13 points higher than those on infliximab. The study set a margin of equivalence of 20% change between the two treatment arms. Here, for example, a patient with a baseline BDAI score of 8 points could expect a follow-up BDAI score of approximately 4.45 on infliximab and 4.58 on alpha interferon, representing an approximate 3% change between the two treatment groups.

Figure 4 demonstrates this graphically, as the probability density for the prior does not differ significantly from the posterior.

FIGURE 4.

Density of prior distribution, likelihood and posterior distribution for mBDAI.

However, Figure 5 also shows that both treatments appear to be associated with significant improvements in BDAI at 3 and 6 months. Defining response categories as 20%, 50% and 70% improvement with respect to baseline revealed that for infliximab, 17, 13 and 8 patients met the 20%, 50% and 70% response definitions. For Roferon, figures were 22, 12 and 7, respectively.

FIGURE 5.

Boxplot of mBDAI at baseline, 3 and 6 months across treatment groups.

The improvement in mBDAI within both treatment arms is illustrated further in Figures 6 and 7 which show the BDAI score measured at baseline and 6 months for infliximab and Roferon.

FIGURE 6.

Change in BDAI score – infliximab.

FIGURE 7.

Change in BDAI score – Roferon.

If there were no impact of treatment, we would expect each point to lie close to the diagonal red dashed line. However, most points lie above the line, showing a higher BDAI at baseline than at 6 months. This illustrates the reduction in BDAI associated with both treatments. The median (IQR) change for infliximab is –1.5 (–4 to 0) BDAI, and regressing the 6-month BDAI data against the baseline BDAI reveals a significant difference and estimated that on average, the 6-month BDAI score is 70% [0.7 (0.109); p < 0.001] that of the baseline score (i.e. a 30% reduction). The median (IQR) change for Roferon is –3 (–5.25 to –1) BDAI and regressing the 6-month BDAI data against the baseline BDAI also indicates a significant difference and estimates that, on average, the 6-month BDAI score is 61% [0.61 (0.069); p < 0.001] that of the baseline score (i.e. a 39% reduction).

Results of the sensitivity analyses demonstrate that the protocol variations and switches did not influence the conclusion that the two treatments did not differ overall in relation to response and provide confidence that the estimate of the difference between the two treatments is reliable.

A significantly higher proportion of patients randomised to alpha interferon swapped away from their randomised treatment compared to those randomised to infliximab treatment (Roferon 11 of 37, infliximab 3 of 37; Table 7, p = 0.0104) (Table 6).

Reasons for switching are summarised in Table 8. It should be noted that for four of the patients switching away from alpha interferon, the reason for the switch was not recorded.

Analyses of secondary outcome measures

Improvements in organ systems after 3 and 6 months compared to baseline were evaluated as secondary outcome measures.

Original Behçet’s disease activity index

The box plot results for the original BDAI at baseline and after 3 and 6 months by treatment group are presented in Figure 5 and Table 27 (see Appendix 1) and, as expected, did not differ from those presented for the mBDAI. There were no statistically significant differences demonstrated between groups for the Bayesian linear model comparing the original BDAI at baseline and 3 months and between baseline and 6 months. Statistics for within-group changes over time are not presented, but they show that both treatments appear to be associated with significant improvements in BDAI at 3 and 6 months.

Ophthalmological assessments included assessment for intraocular inflammation (vitreous haze) and visual function (BCVA; numbers of letters read) at baseline, 3 months and 6 months. For vitreous haze at baseline, 11 patients (infliximab) and 16 patients (Roferon) were evaluated. For BCVA, 18 patients underwent baseline assessments. Not all patients had follow-up assessments. As ocular examination was symptom-directed, the number of patients with these measurements was small. Tables 28–31 (see Appendix 1) summarise the results for vitreous haze and BCVA, respectively, comparing measurements at baseline with results at 3 and 6 months for each eye. For each of the outcome measures, there were no notable differences between treatment groups.

Oral ulcer activity score

Most patients experienced a reduction in their oral ulcer activity score between baseline and 3 months (Table 9) and between baseline and 6 months (Table 7), with a median reduction (for both treatment arms) of 50% at 3 months and 53.8% at 6 months.

Using caution, due to small sample sizes, a slightly higher proportion of patients randomised to alpha interferon experienced a clinically significant (at least 20%) reduction compared to patients randomised to infliximab (64% and 58%, respectively) at both 3 and 6 months; however, this small difference was not statistically significant at the 5% level (Fisher’s exact test p = 0.7600).

Genital ulcer activity

A similar pattern emerged for genital ulceration. Most patients experienced a reduction in their genital ulcer activity between baseline and 3 months (Table 10) and between baseline and 6 months (Table 11), with a median percentage reduction (for both treatment arms) of 100% at 3 months and 100% at 6 months.

Noting the small sample sizes, there was no statistically significant difference (at the 5% level) in the proportion of patients in each treatment arm who experienced a clinically significant (at least 20%) reduction in their genital ulcer activity at 3 months (Fisher’s exact test p = 1.0000) and 6 months (Fisher’s exact test p = 0.7600).

Likert pain score

Though sample sizes were small, there was no statistically significant difference (at 5% level) in the proportion of patients in each treatment arm who experienced a clinically significant (at least 20%) improvement in their Likert pain score at 3 months (Fisher’s exact test p = 1.0000) and 6 months (Fisher’s exact test p = 0.1142). Summary data (see Appendix Tables 32 and 33) indicate modest early improvements for both infliximab and Roferon which were lost by 6 months for the infliximab group.

Prednisolone usage

Prednisolone use was reported as a binary (yes/no) variable longitudinally throughout the study. Table 12 details the number (percentage) of patients receiving prednisolone (or another steroid). The two groups were well matched a baseline with similar proportions receiving steroids [20/39 (51.3%)] and there appeared a modest steroid-sparing effect observed in each group. For infliximab at baseline, 15 patients were using steroids, reducing to 12 at 24-week follow-up. This results in 20% of patients on steroids ceasing their use. The overall rate decreased from 15/29 (52%) to 12/29 (41%). For Roferon at baseline, 16 patients were using steroids which reduced to 9 at 24-week follow-up. This results in 44% of patients on steroids ceasing their use. Two patients on Roferon began taking steroids; therefore, the overall rate decreased from 16/32 (50%) to 11/32 (34%).

The results of a logistic regression model are presented in Table 13 to explore the impacts of time points and treatment on the use of steroids. Treatment and time are included as an interaction to detect either a consistent overall difference due to treatment or a difference between treatments that appears over the course of the study. No significant differences are observed, showing that there is no evidence of any difference in prednisolone (or other steroid) dose reduction between the treatment groups. Further analysis using actual steroid doses will be explored.

Clinician’s overall perception (Physician’s Global Assessment disease activity)

The physician’s overall perception of disease activity (a 7-point Likert scale) was completed as part of (but assessed independently of) the BDAI at baseline, 3 and 6 months. A change of 2 points in the score was considered a clinically meaningful change. The clinician’s overall perception of disease activity indicated a reduction in disease activity for most patients between baseline and 3 months and between baseline and 6 months, with a median reduction of –2.0 (infliximab) and –1.0 (Roferon) at 3 months and –3.0 (infliximab) and –2.0 (Roferon) at 6 months. There was a statistically significant difference (at the 5% level) between treatment arms in the change in clinician’s overall perception of disease activity at 3 months from baseline (Wilcoxon test p = 0.0421) and at 6 months from baseline (Wilcoxon test p = 0.0420) in favour of infliximab providing a greater reduction in the clinician’s overall perception of disease activity.

At 3 months, physician’s overall perception of disease activity was higher for the alpha interferon arm compared to the IFX arm at the 5% level (p = 0.002) and remained significantly higher at 6 months at the 5% level (p = 0.001).

These results are summarised in Tables 14 and 15.

Quality-of-life measures

Quality-of-life measures included the number of patients with reported problems (levels 2, 3, 4, 5) in the EuroQol-5 Dimensions, five-level version (EQ-5D-5L) domains of Mobility, Self-Care, Usual Activities, Pain/Discomfort and Anxiety/Depression at baseline, 3 and 6 months for infliximab and Roferon. There were no differences between groups for these important subdimensions of the EQ-5D-5L score evaluating the number of patients who reported problems.

There were also no significant differences between the two treatment groups for EQ-visual analogue score assessment comparing the difference between baseline versus 3 months and baseline versus 6 months comparing infliximab with alpha interferon (3 months Wilcoxon test, p = 0.8318; 6 months Wilcoxon test, p = 0.8600).

For BD-QoL, only minor differences emerged in favour of infliximab compared to Roferon when comparing differences in scores for QoL at 3 months versus baseline between the two treatment groups (Wilcoxon test, p = 0.0274), but this was no longer present at 6 months (Wilcoxon test, p = 0.3029).

Quality-of-life measures results are summarised in Tables 35–39 (see Appendix 1).

Analysis of safety and tolerability

Table 40 evaluates the difference in AE severity across the two treatment arms for all reported AEs. In total, 46 patients reported 270 events. A proportion test shows that there were a greater number of AEs observed on alpha interferon (p < 0.001). A Fisher’s test to evaluate any differences in the distribution of AEs across treatment arms was not significant (p = 0.224). The overview of AEs by severity can be found in Table 16.

In total, eight serious adverse events (SAEs) from five patients were reported across the study. One patient on the infliximab arm reported four SAEs [hypertension (×2), bacterial urinary tract infection and blood pressure inadequately controlled]. In total, three patients (six events) were reported on the infliximab arm and two patients (two events) were reported on the alpha interferon arm. There were no suspected drug interactions and no suspected unexpected serious adverse reactions (SUSARS) reported in the study.

Summaries of all study-related SAEs together with an aggregated summary of AEs are listed in Table 40 (see Appendix 1).

Genetics analysis

Four SNPs within the IFNL4 gene locus were selected owing to a priori knowledge of effects on gene/protein function or clinical association.

Genotyping was undertaken, and a summary of genotype counts and minor allele frequencies can be found in Table 17. Hardy–Weinberg p-values were within the tolerable threshold (> 0.0001), indicating that genotype distributions are not significantly different to those that might be expected, and therefore there were no quality control issues associated with the genotyping assays.

Genotypes were obtained for a total of 62 individuals (30 in Arm A and 32 in Arm B) for all SNPs except for rs7248668 where a genotype for one individual (Arm B) could not be obtained despite repeated attempts.

The data suggest that there is high linkage disequilibrium between rs12979860 and rs368234815, and rs4803221 and rs7248668, summarised in Table 18.

Genotype association with a binary response outcome based on either 20%, 50% or 70% response was undertaken using a Pearson’s chi-squared test for all SNPs in all individuals plus stratifying for Arm A or Arm B only (Table 19).

These analyses suggest the only statistically significant associations are for the rs4803221 and rs7248668 SNPs in Arm B and only when applying the 70% response binary phenotype (p = 0.021 and 0.025, respectively). However, after correction for multiple testing [false discovery rate (FDR)], these associations are no longer significant (p > 0.05).

Subsequent analysis determined genetic association with four continuous variable outcome measures: BDAI at baseline, 3 months, 6 months and a baseline-adjusted BDAI area under the curve (AUC). This used an ANOVA with Bonferroni correction (Table 20).

A notionally statistically significant association was observed for baseline-adjusted BDAI AUC for both rs12979860 and rs368234815 in Arm A only (infliximab) (p = 0.055). However, this was no longer significant after correction for multiple testing (FDR) (p > 0.05) (Table 20).

Metabolomics

Urine samples from the patients with BS were analysed by NMR spectroscopy and principal component analysis. Initial results showed no significant metabolite differences at baseline between patients allocated to drug A (Infliximab) or drug B (Roferon) (Figure 8). NMR data were analysed as bins which contain multiple metabolites. Three bins showed significant differences between patient groups A and B (Figure 9). However, when corrected for multiple comparisons of all bins, none remained significant (Figure 10).

FIGURE 8.

PCA analysis of baseline urine samples prior to randomised drug treatment. PCA, principle component analysis.

FIGURE 9.

PCA analysis of baseline urine samples showing specific bins significantly different between groups.

FIGURE 10.

PCA analysis of baseline urine samples prior to randomised drug treatment; unpaired t-test for multiple comparisons.

To test whether drug treatments altered metabolite profiles, baseline samples from patients on infliximab (see Figures 8–10) or Roferon (Figures 11–13) were compared to samples from the same patient taken at week 24. PCA analysis, Figures 8 and 11, showed no major difference in metabolite profiles between the two samples. Specific bins, Figures 9 and 12, did show significance, though this was lost when correction for multiple comparisons was made (see Figures 10 and 13).

FIGURE 11.

Analysis of NMR analysis of urine samples from patients on infliximab at baseline vs. 24 weeks. PCA analysis.

FIGURE 12.

Analysis of NMR analysis of urine samples from patients on infliximab at baseline vs. 24-week significant bins.

FIGURE 13.

Unpaired t-test for multiple comparisons.

To determine whether metabolite profiles were associated with responses to the individual drugs, urine samples at week 24 from responders and non-responders were compared. Samples from patients on infliximab showed similar clustering by PCA analysis (Figure 14). Specific bins once again showed significant differences (Figure 15), with one bin remaining significant after correction for multiple comparisons (Figure 16).

FIGURE 14.

Analysis of NMR analysis of urine samples from patients on Roferon at baseline vs. 24 weeks. PCA analysis.

FIGURE 15.

Analysis of NMR analysis of urine samples from patients on Roferon at baseline vs. 24 weeks. PCA analysis – significant bins.

FIGURE 16.

Analysis of NMR analysis of urine samples from patients on Roferon at baseline vs. 24 weeks. PCA analysis – unpaired t-test for multiple comparisons.

For responders versus non-responders to Roferon, there was weak separation between the groups by PCA analysis, particularly clustering of responder samples (Figure 17). This pattern may have been due to specific bins which were significantly different between the groups (Figure 18), although such significance was lost when corrected for multiple comparisons (Figure 19).

FIGURE 17.

Nuclear magnetic resonance metabolite analysis of urine samples from responders to infliximab compared to non-responders. PCA analysis.

FIGURE 18.

Nuclear magnetic resonance metabolite analysis of urine samples from responders to infliximab compared to non-responders. PCA analysis showing significantly different specific bins.

FIGURE 19.

Nuclear magnetic resonance metabolite analysis of urine samples from responders to infliximab compared to non-responders. PCA analysis – unpaired t-tests for multiple comparisons.

Metabolomic analysis showed no significant differences between the patients at baseline before randomised allocation to either of the study drugs (Figure 9). This confirms that there were no major confounding factors that may have influenced response to a particular treatment. Analysis between individual patient’s urine metabolite profile at baseline compared to 24 weeks indicated no major differences, suggesting that the drugs were not inducing wide-ranging changes to the patients’ metabolic processes, but rather that any effects would be due to specific changes to each drug’s target pathways. This was supported by comparison of 24-week urine samples from responders and non-responders to the same drug. For infliximab, one bin remained significant after multiple corrections (Figure 18), and PCA clustering was weaker for patients on Roferon.

While of interest, it should be noted that comparisons between responders and non-responders split each group, leading to a smaller number of samples for analysis. However, the results for infliximab seem to indicate that within one of the bins we have detected potential marker(s) of a metabolic response to treatment which is worthy of further study to identify the individual metabolites and associated metabolic pathways responsible. The signal was weaker for Roferon and may not necessarily be due to the same metabolites and pathways (Figures 20–23). The specific bins that vary between each comparison will therefore be further analysed to determine the specific metabolites responsible for this difference, which may identify the particular pathways being influenced by each treatment.

FIGURE 20.

Nuclear magnetic resonance metabolite analysis of urine samples from responders to Roferon compared to non-responders. PCA analysis.

FIGURE 21.

Nuclear magnetic resonance metabolite analysis of urine samples from responders to Roferon compared to non-responders. PCA analysis, showing significantly different specific bins.

FIGURE 22.

Nuclear magnetic resonance metabolite analysis of urine samples from responders to Roferon compared to non-responders. PCA analysis results using unpaired t-tests for multiple comparisons.

Patient and public involvement

This has been embedded into the Bio Behçet’s study from its inception. This trial was planned with involvement of patients with Behçet’s. The National UK patient charity (Behçet’s UK: Chair Mr Tony Thornburn) was a co-applicant/coinvestigator on the grant. The patient charity helped provide awareness of the study and will be fully involved in dissemination of results. Patients have been involved in data analysis, and a representative from Behçet’s UK was part of the TSC.

Equality, diversity and inclusion

Our aim has been to be inclusive and reflective of the diverse population of patients with Behçet’s in the UK in the recruitment and running of this study. We have recruited through the National Centres for Behçet’s in the UK – to understand response to drugs in the UK Behçet’s population, as set out in the grant application. As part of this, we have recruited patients from a broad spectrum of ethnicities found in UK Behçet’s patients, but we are aware that the spectrum of ethnicities in the UK may differ from that in regions of greater preference for Behçet’s, such as in Silk Route countries.

Contributions of authors

Robert J Moots (https://orcid.org/0000-0001-7019-6211) (Professor of Rheumatology) led the design of the study, recruited patients, contributed to analysis of results and preparation of the manuscript.

Farida Fortune (https://orcid.org/0000-0001-5930-1973) (Professor of Oral Medicine) contributed to study design, patient recruitment, analysis of results and preparation of the manuscript.

Richard Jackson (https://orcid.org/0000-0002-7814-5088) (Lecturer in Statistics) contributed to analysis of results and preparation of the manuscript.

Tony Thornburn (https://orcid.org/0000-0001-7566-7596) (Chair, Behçet’s UK) contributed to analysis of results and preparation of the manuscript.

Ann W Morgan (https://orcid.org/0000-0003-1109-624X) (Professor of Rheumatology) contributed to study design, analysis of results and preparation of the manuscript.

Dan Carr (https://orcid.org/0000-0001-8028-4282) (Senior Lecturer in Clinical Pharmacology) contributed to study design, analysis of results and preparation of the manuscript.

Philip Ian Murray (https://orcid.org/0000-0001-8491-3795) (Professor of Ophthalmology) contributed to study design, analysis of results and preparation of the manuscript.

Graham Robert Wallace (https://orcid.org/0000-0003-2054-0509) (Senior Lecturer in Rheumatology) contributed to study design, analysis of results and preparation of the manuscript.

Deva Situnayake (https://orcid.org/0000-0002-0643-0760) (Consultant Rheumatologist) contributed to study design, recruitment of patients, analysis of results and preparation of the manuscript.

Acknowledgements

We are also most grateful for the help of:

Professor David Jayne (Chair, Trial Steering Committee)

Professor Dorian Haskard (Chair, Data Monitoring Committee)

Professor Hasan Yaziki (Trial Steering Committee)

Professor Thomas Jaki (Independent Statistician)

Professor Cristoph Deuter (Independent Expert)

Trustees and Members of Behçet’s UK

Ms Helen Frankland, Ayren Mediani, Elisabeth Ssendi, Nardos Wakjira, Debbie Mitton and Jo Kennedy (Research and Specialist Nurses)

Dr Andy Cross (Laboratory Manager, Liverpool Behçet’s Centre)

Dr Trevor Cox (Statistician)

Dr Elaine Howarth (Statistician)

Dr Karen Wilding, University of Liverpool

Diane Carlton, Chris Harvey, Claire Taylor and other staff at the Liverpool Clinical Trials Consortium

Data-sharing statement

All available data can be obtained from the corresponding author via the Liverpool Clinical Trials Consortium.

Ethics statement

This study was assessed and approved by the NRES Committee North West – Liverpool Central. Date of approval: 17 March 2015. Reference number: 15/NW/0008.

Information governance statement

The University of Liverpool is committed to handling all personal information in line with the UK Data Protection Act (2018) and the General Data Protection Regulation (EU GDPR) 2016/679. Under Data Protection legislation the University of Liverpool is the Data Processor and the Data Controller, and we process personal data in accordance with their instructions. You can find out more about how we handle personal data, including how to exercise your individual rights and the contact details for the University of Liverpool Data Protection Officer https://www.liverpool.ac.uk/legal/data_protection/.

Disclosure of interests

Full disclosure of interests: Completed ICMJE forms for all authors, including all related interests, are available in the toolkit on the NIHR Journals Library report publication page at https://doi.org/10.3310/HTFC6304.

Primary conflicts of interest: Ann W Morgan reports Investigator Initiator grant from Roche for GCA; consultancy and honoraria from Roche and Chugai; consultancy from AstraZeneca, Sanofi, Regeneron and Vifor; travel subsidence from Roche and MRC DPFS panel membership. Philip Ian Murray reports payments from Oxford University Press and payments made to the author’s institution from Scope Eyecare. Graham Robert Wallace received travel subsidence from Cerrahpasa University Hospital to attend the Cerrahpasa Behçet’s Disease Symposium. All other authors have no interests to disclose.

Disclaimers

This article presents independent research. The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, the MRC, the EME programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, the EME programme or the Department of Health and Social Care.