Article Text
PDF
Abstract
Objectives To determine the 5-year outcomes of early remission induction therapy followed by targeted treatment aimed at drug-free remission (DFR) in patients with early arthritis.
Methods In 12 hospitals, 610 patients with early (<2 years) rheumatoid arthritis (RA) or undifferentiated arthritis (UA) started on methotrexate (MTX) 25 mg/week and prednisone (60 mg/day tapered to 7.5 mg/day). Patients not in early remission (Disease Activity Score <1.6 after 4 months) were randomised (single blind) to arm 1, adding hydroxychloroquine 400 mg/day and sulfasalazine 2000 mg/day, or arm 2, switching to MTX plus adalimumab 40 mg/2 weeks. Treatment adjustments over time aimed at DFR. Outcomes were remission percentages, functional ability, toxicity and radiological damage progression after 5 years.
Results After 4 months, 387 patients were in early remission, 83 were randomised to arm 1 and 78 to arm 2. After 5 years, 295/610 (48%) patients were in remission, 26% in sustained DFR (SDFR) (≥1 year) (220/387 (57%) remission and 135/387 (35%) SDFR in the early remission group, 50% remission, 11% SDFR in the randomisation arms without differences between the arms). More patients with UA (37% vs 23% RA, p=0.001) and more anticitrullinated protein antibody (ACPA)-negative patients (37% vs 18% ACPA-positive, p<0.001) achieved SDFR.
Overall, mean Health Assessment Questionnaire was 0.6 (0.5), and median (IQR) damage progression was 0.5 (0–2.7) Sharp/van der Heijde points, with only five patients showing progression >25 points in 5 years.
Conclusions Five years of DFR-steered treatment in patients with early RA resulted in almost normal functional ability without clinically relevant joint damage across treatment groups. Patients who achieved early remission had the best clinical outcomes. There were no differences between the randomisation arms. SDFR is a realistic treatment goal.
- early rheumatoid arthritis
- treatment
- disease activity remission
- DMARDs (biologic)
- DMARDs (synthetic)
Statistics from Altmetric.com
Introduction
Rheumatoid arthritis (RA) is characterised by inflammation of synovial joints,1 which can lead to destruction of joints, even before the classification criteria for RA are met (undifferentiated arthritis (UA)).1–3 Current recommendations are to start disease-modifying antirheumatic drugs (DMARDs) as soon as possible and to intensify or change medication until remission or at least low disease activity has been achieved.4–17
Previous studies have shown that some patients even achieve early drug-free remission (DFR).4 8 There is evidence that sustained DFR, approximating cure, is more likely if treatment is started very early, possibly before disease characteristics meet classification criteria.12
In IMPROVED (Induction therapy with MTX and Prednisone in Rheumatoid Or Very Early arthritic Disease) study, we aimed at early DFR in patients with early RA and UA. Here we report 5-year clinical and radiological outcomes.
Methods
Study design
IMPROVED study is a multicentre study designed by Dutch rheumatologists participating in the Foundation for Applied Rheumatology Research, in patients with RA or UA. It combines a treat-to-target observational cohort (treatment target DFR) with a randomised clinical trial comparing two treatment strategies in the subset of patients who after 4 months of induction therapy did not achieve early remission (see below). In all patients, the general aim was to achieve drug-free clinical remission as early as possible. Every 4 months, disease activity was monitored using the Disease Activity Score (DAS).11 If DAS was <1.6 (remission),9 patients tapered medication until DFR was achieved, but treatment was intensified or changed if DAS was ≥1.6. The induction therapy was with methotrexate (MTX) and prednisone, followed, for patients not in remission at 4 months, by either immediately switching to a biologic DMARD (randomisation arm 2) or first trying additional synthetic DMARDs (arm 1). The study was conducted in 12 hospitals in the western part of the Netherlands. The study protocol was approved by the medical ethics committee of each participating centre.
Patients
Eligible patients were ≥18 years, with early RA fulfilling the 2010 American College of Rheumatology (ACR)/European League Against Rheumatism (EULAR) classification criteria5 with symptom duration ≤2 years or UA suspected to be early RA according to the rheumatologist, regardless of symptom duration, with DAS ≥1.6, who had not been treated with prednisone and/or DMARDs. Exclusion criteria were pregnancy/pregnancy wish during the study, malignancy within the last 5 years, bone marrow hypoplasia, aspartate transaminase and/or alanine transaminase >3 times normal value, serum creatinine level >150 µmol/L or estimated creatinine clearance <75%, uncontrolled diabetes mellitus, uncontrolled hypertension, heart failure (New York Heart Association class III/IV), alcohol or drug abuse, serious infections in the previous 3 months or chronic infectious disease, active or latent hepatitis B infection, known HIV infection, lymphoproliferative disease and multiple sclerosis.7 13 Patients with active tuberculosis (TB) and patients with UA with latent TB were excluded. Patients with RA with latent TB could be enrolled if they started adequate antituberculous therapy prior to initiation of high-dose prednisone, according to local recommendations. All patients gave written informed consent.
Study outlay
Treatment decisions during the study were based on the DAS (based on 53/44 joints), which was assessed every 4 months.11 DAS <1.6 was considered to denote remission.9 For all patients over 5 years, if DAS was ≥1.6, dose intensification, change or restart of the last discontinued medication was required, and medication was tapered to 0 as soon and as long as DAS was <1.6, until DFR was achieved.
All patients started on MTX 7.5 mg/week increased to 25 mg/week in 5 weeks (or highest tolerated dose, oral or subcutaneous at the discretion of the rheumatologist) and prednisone tapered in 7 weeks from 60 to 7.5 mg/day. Patients who were in remission after 4 months (early remission) tapered and after 3 weeks stopped prednisone, then, if remission continued at 8 months, over 10 weeks tapered and stopped MTX, thus achieving DFR at year 1 (online supplementary figure S1). If, at 8 months, DAS was ≥1.6, prednisone could be restarted once, at 7.5 mg/day. With regained remission, this could be tapered and stopped again.
Supplementary file 1
For patients not in early remission, the study investigated if early start of biologic DMARD adalimumab (arm 2) resulted in more patients achieving DFR compared with first trying a combination of conventional DMARDs (arm 1), MTX (25 mg/week or highest tolerated dose), sulfasalazine 2000 mg/day and hydroxychloroquine 400 mg/day with prednisone 7.5 mg/day, to be changed to MTX and adalimumab 40 mg/2 weeks if remission was not achieved. In both treatment arms, adalimumab was intensified to 40 mg/week if remission was not achieved on the standard dose of 40 mg/2 weeks. If remission was still not achieved, subsequent treatment steps were left to shared decision making by rheumatologist and patient (online supplementary figure S1).
Primary and secondary outcomes
Primary outcomes after 5 years were percentages of remission and DFR based on a DAS <1.6 or on the ACR/EULAR 2011 Boolean remission definition.6 ‘Sustained drug-free remission’ was defined by DFR during ≥1 year, starting at any time point, and called ‘early sustained drug-free remission’ if present from the earliest possibility, at t=12 months in early remission patients.
Secondary outcomes were DAS, functional ability assessed by the Dutch version of the Health Assessment Questionnaire (HAQ),10 damage progression on annual radiographs of the joints in hands and feet, scored in chronological order by two independent readers (GA and SAB) using the Sharp/van der Heijde Score (SHS)18 and toxicity as recorded by the research nurses at each 4 monthly visit in the first 2 years and afterwards annually and/or by the treating physician. (Serious) Adverse events (AEs) were reported per 100 patient years.
Statistical analysis
The target sample size was calculated with a power calculation to detect differences of ≥50% in remission rates and 0.2 points in HAQ between the early or delayed adalimumab randomisation arms, with 80% power. Based on an estimated 30% of patients achieving early remission, we would need 535 patients to randomise 100 patients in each arm. During the study, more patients achieved early remission and the target sample size was recalculated and increased to 610 patients. Comparisons in outcomes were made between randomisation arms. In addition, outcomes were compared across the whole cohort in relation to DFR-steered treatment, for baseline characteristics such as disease activity, autoantibody status and symptom duration.
Outcomes were compared using Student’s t-tests, Mann-Whitney U tests and χ2 tests. DAS and HAQ over time (dependent variables) were studied using linear mixed models (LMMs), with treatment strategy (arms 1 and 2), time (study visit) and interaction term (time×treatment strategy) as fixed effects. An additional LMM was done with multiple imputation for missing values for DAS and HAQ (due to missing values for variables to calculate DAS and HAQ), using chained equations (10 cycles), with age, gender, symptom duration, body mass index (BMI), smoking status, diagnosis, autoantibody status (rheumatoid factor (RF), anticitrullinated protein antibodies (ACPAs) and anticarbamylated protein antibodies (anti-CarP)), baseline DAS and HAQ, as well as allocated treatment strategy and annual total SHS score added in the imputation model. We performed intention-to-treat analyses. Patients were included until the time of study discontinuation (see flow chart in figure 1). Patients with missing data were not excluded from the analysis. In patients with available baseline and 5-year radiographs, the progression score over 5 years was calculated. Missing values for annual erosion and narrowing scores of hands and feet were imputed by multiple imputation using chained equations (10 imputed datasets), after first log-transformation because of skewed data, with age, gender, symptom duration, BMI, smoking status, diagnosis, autoantibody status (RF, ACPAs and anti-CarP), baseline DAS and HAQ, as well as allocated treatment strategy, annual DAS and HAQ and log-transformed annual erosion and joint space narrowing scores of hands and feet added in the imputation model. All statistical analyses were conducted with SPSS V.23.0. The study is registered with the ISRCTN Register (number 11916566) and EudraCT (number 2006-006186-16).
Trial profile IMPROVED study. First DAS evaluation was at 4 months. Fifty patients who were not in remission at 4 months but were not randomised according to the protocol were treated outside of protocol (OP group). Of those, 19 discontinued treatment before 5 years and 31 patients were included in the intention-to-treat analysis. DAS, Disease Activity Score; IMPROVED, Induction therapy with MTX and Prednisone in Rheumatoid Or Very Early arthritic Disease.
Results
Between 13 March 2007 and 24 September 2010, 610/730 assessed patients were eligible and included in the study (figure 1), 479 (79%) with classifiable RA and 131 (21%) with UA. During 5 years, 152/610 (25%) patients (112 with RA, 40 with UA) were lost to follow-up: 17 patients died, 13 left the study due to comorbidity, 12 had a revised diagnosis and 110 discontinued their consent. Twelve patients left the study before the first assessment at 4 months. Of 610 patients, 387 (63%) achieved early remission. One-hundred and sixty-one patients (26% of 610) were randomised: 83 patients to delayed adalimumab (arm 1) and 78 to early adalimumab (arm 2); 50 others were not in early remission but were erroneously not randomised (OP group). Online supplementary table S1 shows the number of patients still in the study at each time point for the different treatment groups. Baseline characteristics were well balanced between randomisation arms.
Supplementary file 2
General outcomes of the whole group
Baseline characteristics of all patients are shown in table 1. All but six patients achieved remission at least once during 5 years of follow-up. At 5 years, 295/610 patients (48%) were in remission and 137/610 (22%) were in ACR/EULAR Boolean remission. During 5 years, on average, 54% of all patients were in DFR at least once (figure 2). Of the 295 in remission at 5 years, 159 (26% of 610) had achieved sustained (≥1 year) DFR, 58 from the earliest possible time point at year 1. Of the 58 in early sustained DFR, 24 (4% of 610) were still in DFR at 5 years, while 26 patients had lost remission and restarted medication, and 8 had left the study early while still in DFR. Online supplementary table S2 shows percentages missing per patient still participating in the study at each time point for figures 2 and 3. At 5 years, the mean difference in HAQ was −0.4 (95% CI −0.5 to −0.3) between patients in and not in remission.
Treatment during 5 years in (A) ER group, (B) arm 1, (C) arm 2, in percentage of participating patients at each time point per treatment group. (D) Percentages in remission and percentages in DFR per treatment group. Lines are approximations of the proportions of patients discontinuing medications (according to tapering strategies or due to side effects) or starting medications according to remission-steered escalation strategies, across various treatment steps per arm, during 5 years. Percentages are calculated for total number of participating patients at each time point. The category ‘Other’ includes medications that were prescribed per protocol in the ‘treatment according to rheumatologist’ step after failure on methotrexate plus adalimumab, as well as medications prescribed outside of the protocol but still maintaining a remission-targeted strategy. Shaded areas denote patient proportions in remission during 5 years. ADA, adalimumab; DAS, Disease Activity Score; DFR, drug-free remission; ER, early remission; HCQ, hydroxychloroquine; mono, monotherapy; MTX, methotrexate; pred: prednisone; SSZ, sulfasalazine.
![[SP1.jpg]](https://ard.bmj.com/content/annrheumdis/77/1/111/DC1/embed/inline-supplementary-material-1.jpg?download=true){kind=link}
(A) Mean HAQ over time. (B) mean DAS over time. Means were calculated per time point for patients that had a value at that time point. (C) Probability plot SHS progression in completers with available X-rays at baseline and 5 years. (D) Estimated mean total SHS over time after imputation. DAS, Disease Activity Score; HAQ, Health Assessment Questionnaire; SHS, Sharp/van der Heijde Score.
Baseline characteristics of IMPROVED study population
Radiographs at baseline and 5 years were available in 362/458 study completers. SHS progression ≥0.5 points was seen in 180 (39%) completers, with a median SHS progression (IQR) of 0 (0–3) points. 58/458 (13%) had progression ≥5. The mean yearly progression rate in completers was 0.43 points/year. A probability plot with SHS progression after 5 years in completers (patients with available X-rays at baseline and 5 years) is shown in figure 3C.
During 5 years follow-up, 555 (91%) patients had in total 2897 AEs (21.4 AEs per 100 patient years; online supplementary table S3). The most common AEs were upper airway infections, increased liver enzymes and skin rash. One hundred and forty-eight (24%) patients reported 242 serious AEs (SAEs) (5.7 SAEs per 100 patient years).
Comparisons between patients in and not in early remission
Patients who achieved early remission had, at baseline, lower DAS and slightly shorter symptom duration; fewer were female; fewer fulfilled the RA classification criteria, although more were positive for autoantibodies; and more had erosions on radiographs at baseline, compared with patients who did not achieve early remission (online supplementary table S4). In general, patients who achieved early remission had better outcomes over time than patients in the randomisation arms or out of protocol group. Sustained DFR was achieved by 135/387 (35%) in the early remission group, compared with 11% of the patients in randomisation arms 1 (9/83) and 2 (10/78) and the OP group (5/50). At 5 years, 220/387 (57%) in the early remission group patients were in remission, and 111/387 (29%) in ACR/EULAR Boolean remission, compared with 75/211 (36%) and 26/211 (12%), respectively, in the other patients (online supplementary table S4 and figure 2D). HAQ improvement over time was −0.6 (0.7) in the early remission group and −0.5 (0.8) in other patients (figure 3A), resulting in mean HAQ over time of 0.4 (0.4) and 0.9 (0.5), respectively. After imputation, total radiological damage progression was similar in the early remission group and the other patients (figure 3C,D), but more patients in the early remission group had erosion progression.
In the early remission group, use of medication initially decreased, with a steep decline in use of prednisone and to a lesser degree also of MTX, then remained stable over time (figure 2A). During 5 years, 98/387 (25%) patients who initially achieved early remission, because of later increases in DAS, received treatment with adalimumab and/or other DMARD.
Comparisons between the randomisation arms
For patients who did not achieve early remission and were randomised to arms 1 (first DMARD combination therapy) and 2 (immediately adalimumab), there were no significant differences at the time of randomisation or in clinical and radiological outcomes or use of DMARD over 5 years of DFR-steered treatment (table 2). Over time, there was no difference in DAS (LMM 0.05 (95% CI −0.30 to 0.40), p=0.768) and HAQ (LMM 0.05 (95% CI −0.20 to 0.29), p=0.705) between the randomisation arms.
Outcomes at time of randomisation and after 5 years in the randomisation arms
Comparisons between patients with RA and UA
At 5 years, 234/479 patients with RA (49%) and 61/131 (47%, p=0.366) patients with UA were in remission. More patients with UA (41/131 (31%)) than patients with RA (93/479 (19%), p<0.001) were in DFR at 5 years. Over 5 years, sustained DFR was achieved by more patients with UA (49/131 (37%)) than patients with RA (110/479 (23%), p=0.001). Despite these differences and although mean DAS over time was significantly lower in patients with UA (1.5 (0.6)) compared with patients with RA (1.7 (0.7), p=0.003), mean HAQ over time was similar in patients with RA and UA. More patients with RA than patients with UA had SHS progression ≥0.5 points (observed) (152/479 (32%) vs 28/131 (21%), p=0.118) with a higher median (3 (1–6.9) vs 2.3 (1–3.5), p=0.199), and 54/479 (11%) vs 4/131 (3%) had SHS progression ≥5 (p=0.011), with a higher median progression score (7.8 (6–12.3 vs 6 (5.3–19.5), p=0.288).
Comparison between antibody-positive and antibody-negative patients
Absence of autoantibodies (RF and ACPAs) was associated with more DFR (81/262 (31%) vs 50/332 (15%), p<0.001), but not more remission (120/262 (46%) vs 172/332 (52%), p=0.887) at 5 years. Over 5 years, sustained DFR was achieved by 96/262 (37%) of ACPA-negative compared with 60/332 (18%) ACPA-positive patients (p<0.001). Similarly, over 5 years, sustained DFR was achieved by 79/245 (32%) of RF-negative and 74/339 (22%) RF-positive patients (p=0.005). Mean DAS over time was comparable in ACPA-negative (1.7 (0.7)) and ACPA-positive patients (1.7 (0.7), p=0.882). Also, mean HAQ was similar in ACPA-negative (0.54 (0.6)) and ACPA-positive patients (0.52 (0.5), p=0.712). More ACPA-positive patients compared with ACPA-negative patients had SHS progression ≥0.5 after 5 years (117/333 (35%) vs 60/255 (24%), p=0.016), and 41/333 (12%) vs 16/255 (6%) had SHS progression ≥5 (p=0.041<0.001), with a higher median progression score (0.5 (0–3) vs 0 (0–1.5), p<0.001). Also, erosive disease was seen in more ACPA-positive patients (62/333 (19%; was 15% at baseline)) than in ACPA-negative patients (23/255 (9%; was 9% at baseline), p=0.001, for comparison at 5 years).
Discussion
This study in patients with early RA and UA shows for the first time that a treatment strategy aiming at sustained DFR is successful, although most patients periodically still need a DMARD. Remission defined by DAS<1.6, on or off DMARD, was achieved in 48% of patients after 5 years and DFR in 54% at least once. Functional ability approached normality in these patients, and radiological damage progression was well suppressed and clinically negligible.19 Patients with UA, in our study largely overlapping with autoantibody-negative patients, achieved more DFR than patients with RA and autoantibody-positive patients but overall showed similar disease activity and functional ability over time and similarly little radiological damage progression. In general, the 63% of patients who were in remission after 4 months of treatment had better outcomes than patients who were not. For patients not in early remission, an immediate switch to biologic DMARD adalimumab or first trying a combination of conventional DMARDs resulted in similar outcomes over time. Patients who achieved early remission already had lower DAS and HAQ at baseline, placing the DAS target within closer reach. Our study suggests that, rather than starting with relatively ‘mild’ treatment, remission induction therapy for these patients can be especially successful. Also, our results show that regardless of presence of autoantibodies, all patients benefit from early remission induction therapy followed by treatment adjustments as long as remission is not achieved. Previous subanalyses of the current study showed that ACPA positivity was associated with early remission,13 but not with early DFR.7 From year 1, ACPA positivity was associated with loss of DFR,20 a trend that is now confirmed with finding fewer ACPA-positive patients than ACPA-negative patients achieving sustained DFR. ACPA-positive patients also showed statistically significantly more progression in damage scores. Total joint damage progression after 5 years was similarly low in patients who were or were not in early remission, but we saw slightly more erosive joint damage progression in patients in the early remission group. More patients in the early remission group already had erosions at baseline, which is associated with more erosion progression.21 It is also possible that, due to more drug tapering to DFR and less use of anti-TNF therapy compared with the other treatment groups, there has been subclinical inflammation.
Experts considered a yearly progression score of ≥5 points to be clinically relevant.22 Only 5/610 patients in the current study had >25 points progression after 5 years. In general, progression scores were remarkably low, even though the radiographs were scored in chronological order, which is associated with higher sensitivity.23 Previous studies aiming at low DAS (≤2.4)24 25 or even stricter remission definitions26 27 reported more radiological damage progression (scored in random order) despite reporting similar or higher remission rates. This may be due to the inclusion of more advanced and/or severe RA types.
This study is the first to aim for relatively rapid tapering of medication aiming at sustained DFR, which we felt is the outcome closest approaching cure. Therefore, we aimed to include and treat patients in an early phase of (suspected, if classification criteria were not met) RA, as it appears that earlier treatment may result in better and long-lasting suppression of inflammatory processes, which at that time may be reversible. During this so-called ‘window of opportunity’, chronicity of inflammation may be prevented and prolonged remission may be induced.28–30 It is estimated to encompass around 12 weeks from symptom onset.12 However, we found few differences in remission rates and only a trend for more DFR in patients with symptom duration <12 weeks compared with ≥12 weeks. It may be that the initial combination with fast acting prednisone makes up for possible downsides of a longer symptom duration.
Due to drug tapering strategies, the use of prednisone over time was low. Most patients remained on MTX monotherapy or were in DFR (figure 2). In the randomisation arms, use of adalimumab stabilised at twice the level of use as in the early remission group and also use of other biologic DMARDs was higher, which will have impacted treatment costs. Toxicity in the early remission group and the randomisation arms was comparable.
There are several limitations to the study design. First, a comparator arm being absent, we cannot claim that the good clinical and radiological outcomes are the result of the initial treatment, subsequent medications or the remission-steered treatment adjustments. We may have temporarily overtreated patients who would have achieved spontaneous remission. This was partly why we tried to taper and discontinue medication early. We chose the MTX and prednisone doses for induction therapy based on the results of the COBRA31 and BeSt studies.24 More recent studies have shown that lower dosages of prednisone may be equally effective.32 33 The 4 monthly evaluation time points may not have provided sufficiently tight control in combination with targeted treatment. This and rapid tapering may have resulted in fewer patients achieving sustained (drug-free) remission than we hoped. The remission target may be insufficiently stringent, even though we used the original DAS, based on more joints than the DAS28. Had we treated aiming at a more strict remission definition, more patients might have achieved DFR, but it would also have risked higher use of costly medications. All definitions of remission may be influenced by non-inflammatory pain.34 Finally, early study termination in the various patient groups may have influenced the results, and missing data may not have been completely at random and can potentially bias data analysis. We have performed multiple imputations and repeated the analysis (data not shown), which showed results that did not differ from the complete case analysis.
In conclusion, in this study in patients with early RA and UA, after 5 years of treatment aiming at DFR, HAQ results indicate almost normal functional ability over time and radiological damage progression was generally well suppressed. Forty-eight per cent of patients were in remission after 5 years, and 26% had achieved DFR for at least 1 year, but few achieved continuous DFR. Aiming at DFR appears to be an effective way to prevent damage and deterioration, provided that disease activity is vigilantly monitored, weighing the benefits of discontinuation of medication against the risk of a disease relapse.
Acknowledgments
We would like to thank all patients as well as the following rheumatologists (other than the authors) who participated in the IMPROVED study group (all locations are in The Netherlands): C Bijkerk (Reinier de Graaf gasthuis, Delft), M H W de Bois (Medical Center Haaglanden), P D M de Buck (Medical Center Haaglanden, Leidschendam); G Collée (Medical Center Haaglanden, The Hague); J A P M Ewals (Haga Hospital, The Hague); Y P M Goekoop-Ruiterman (Haga Hospital, The Hague), J B Harbers (Franciscus Hospital, Roosendaal); E T Molenaar (Groene Hart Hospital, Gouda); N Riyazi (Haga Hospital, The Hague); H K Ronday (Haga Hospital, The Hague); A A (Haga Hospital, The Hague); I Speyer (Bronovo Hospital, The Hague); M L Westedt (Bronovo Hospital, The Hague). We would also like to thank all other rheumatologists and trainee rheumatologists who enrolled patients in this study and all research nurses for their contributions.
References
Footnotes
Handling editor Tore K Kvien
Contributors GA performed the statistical analysis, interpreted the data and drafted the manuscript. LH, SAB, RJG, MvO, JHLMvG, AJP, GMS-B, LRL, PBJdS and BAMG contributed in the acquisition of the data and revised the manuscript. TWJH participated in the study design, contributed in the acquisition of the data and was involved in revising of the manuscript. CFA participated in the study design, contributed in the acquisition of the data and was involved in analysing and interpreting the data and helped to draft the manuscript. All authors read and approved the final version of the manuscript.
Funding The study was designed by the investigators and financially supported during the first year of follow-up by AbbVie.
Disclaimer Trial management, data collection, patient recruitment, data analysis, interpretation and preparation of the manuscript were performed by the authors. We have not been paid to write this manuscript by a pharmaceutical company or other agency. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.
Competing interests TWJH reports grants from AbbVie during the conduct of the study and grants from UCB, Bristol Myers Squibb, Pfizer, Roche, Sanofi-Aventis and Boeringher from outside the submitted work. The other authors declare no competing interests.
Patient consent Obtained.
Ethics approval The study protocol was approved by the medical ethical committees of all participating hospitals.
Provenance and peer review Not commissioned; externally peer reviewed.