Article Text

Extended report
Long-term outcomes after disease activity-guided dose reduction of TNF inhibition in rheumatoid arthritis: 3-year data of the DRESS study - a randomised controlled pragmatic non-inferiority strategy trial
1. Chantal AM Bouman1,
2. Noortje van Herwaarden1,
3. Frank HJ van den Hoogen1,2,
4. Jaap Fransen2,
5. Ronald F van Vollenhoven3,4,
6. Johannes WJ Bijlsma5,
7. Aatke van der Maas1,
8. Alfons A den Broeder1,2
1. 1Department of Rheumatology, Sint Maartenskliniek, Nijmegen, The Netherlands
2. 2Department of Rheumatology, Radboud University Medical Centre, Nijmegen, The Netherlands
3. 3Department of Clinical Immunology and Rheumatology, Academic Medical Center, Amsterdam, The Netherlands
4. 4Department of Rheumatology, VU University Medical Centre, Amsterdam, The Netherlands
5. 5Department of Rheumatology & Clinical Immunology, Utrecht University Medical Centre, Utrecht, The Netherlands
1. Correspondence to Chantal AM Bouman, Department of Rheumatology, Sint Maartenskliniek, PO box 9011, Nijmegen 6500 GM, The Netherlands; C.Bouman{at}maartenskliniek.nl

## Abstract

Objective Tumour necrosis factor inhibitors (TNFi) are effective in rheumatoid arthritis (RA), but disadvantages include adverse events (AEs) and high costs. This can be improved by disease activity-guided dose reduction (DR). We aimed to assess long-term outcomes of TNFi DR in RA by using 3-year data from the DRESS study (Dose REduction Strategy of Subcutaneous TNF inhibitors study).

Methods In the intervention phase (month 0–18) of the DRESS study (Dutch trial register, NTR 3216), patients were randomised to DR or usual care (UC). In the extension phase (month 18–36), treatment strategies in both groups converged to continuation of protocolised tight control and allowed dose optimisation. Intention-to-treat analyses were done on flare, disease activity (28 joint count-based disease activity score with C reactive protein (DAS28-CRP)), functioning (health assessment questionnaire-disability index (HAQ-DI)), quality of life (Euroqol 5 dimensions 5 levels questionnaire (EQ5D–5L)), medication use, radiographic progression (Sharp van der Heijde score (SvdH)) and AE.

Results 172/180 patients included in the DRESS study were included in the extension phase. Cumulative incidences of major flare were 10% and 12% (−2%, 95% CI −8 to 15) in DR and UC groups in the extension phase, and 17% and 14% (3%, 95% CI −9 to 13) from 0 to 36 months. Cumulative incidences of short-lived flares were 43% (33 to 52%)%) and 35% (23 to 49%)%) in DR and UC groups in the extension phase, and 83% (75 to 90%)%) and 44% (31 to 58%)%) from 0 to 36 months. Mean DAS28-CRP, HAQ-DI, EQ5D-5L and SvdH remained stable and not significantly different between groups. TNFi use remained low in the DR group and decreased in the UC group. Cumulative incidences of AE were not significantly different between groups.

Conclusions Safety and efficacy of disease activity guided TNFi DR in RA are maintained up to 3 years, with a large reduction in TNFi use, but no other benefits. Implementation of DR would vastly improve the cost-effective use of TNFi.

• rheumatoid arthritis
• TNF inhibitor
• treatment optimization
• flare

## Introduction

The treatment of rheumatoid arthritis (RA) has improved in the last two decades, due to, among other things, the introduction of the first widely used class of anticytokine drugs: tumour necrosis factor inhibitors (TNFi). These drugs are effective and safe in the treatment of RA, providing benefits for symptoms, functioning, quality of life and inhibition of joint damage.1 2

However, TNFis are not without their drawbacks. These include (dose dependent) increased risk of infection,3 skin cancer4 and idiosyncratic adverse reactions like induction of multiple sclerosis, lupus or heart failure.5–7 Furthermore, the need for regular self-injection poses a burden for patients. Lastly, the costs of these drugs are significant, both per patient per year (Europe: $17 000; USA:$26 000), as well as on a macroeconomic scale.8 9

These disadvantages might be ameliorated by dose reduction (DR) or discontinuation of TNFi after disease control has been achieved, and this indeed has been shown to be possible in a relevant proportion of patients.10–12 A disease activity-guided TNFi DR strategy has been tested previously in the DRESS study (Dose REduction Strategy of Subcutaneous TNF inhibitors) and has been shown to be feasible and non-inferior to usual care (UC) with regard to prolonged flaring.13 The strategy also did not result in differences in disease activity, functioning, quality of life or relevant radiographic progression after 18 months. However, short-lived flares and minimal radiographic progression occurred more frequently in the DR arm, probably due to the temporary effects of unsuccessful DR attempts on disease activity.14 Although no benefits were seen with regard to side effects, the cost effectiveness was very high, reaching \$440 000 saved per lost quality-adjusted life year.15

Some important questions remain, especially with regard to long-term risks and benefits of this strategy. Does the occurrence of major flare remain comparable between groups after longer follow-up? Is the small difference in radiographic joint damage between groups only due to a temporary difference in disease activity, or should we expect more damage in subsequent years? And finally, can a lower risk of adverse events (AEs) (eg, infections) be demonstrated?

In an attempt to answer these questions, we performed an extension study of the original DRESS study, exploring long-term effects of this DR strategy on disease activity, functioning, quality of life, radiographic progression and (serious) adverse events ((S)AE).

## Methods

### Study design and participants

This is a long-term extension study of the DRESS study, an 18-month, pragmatic, open label, randomised controlled, non-inferiority strategy trial in patients with RA, in which a disease activity-guided DR strategy of adalimumab or etanercept was compared with UC. For inclusion criteria, we refer to van Herwaarden et al.13 Disease activity was measured using DAS28-CRP (28 joint count-based disease activity score with C reactive protein (CRP)).

The DRESS study was registered at the Dutch trial register (www.trialregister.nl, NTR 3216), and its design and results have been reported previously.13–17 The extension study was performed from May 2014 to January 2016 in the Sint Maartenskliniek Nijmegen and Woerden, The Netherlands, and was approved by the local ethics committee (CMO region Arnhem-Nijmegen, NL37704.091.11).

In the intervention phase (month 0–18) of the DRESS study, patients were randomised to the DR or UC group in a ratio of 2:1, stratified for adalimumab and etanercept. In the extension phase (month 18–36), the original group allocation was maintained. Both the intervention and extension phase were non-blinded.

### Procedures

In the intervention phase, patients allocated to UC continued a standardised tight control treatment protocol (maintaining DAS28-CRP <3.2).

In the DR group, patients received identical care, with addition of a specific DR advice given to the treating rheumatologist for that particular patient. The DR strategy consisted of 3-monthly stepwise increase of injection time interval until flare or discontinuation. For details we refer to van Herwaarden et al.13 If the flare persisted after 4 weeks despite bridging with intramuscular or intra-articular steroids or non-steroidal anti-inflammatory drugs, TNFi was increased stepwise, if needed, to the shortest registered interval. If a flare persisted thereafter, treatment was switched. Only one DR attempt was advised.

As flare criterion, a DAS28-CRP increase >1.2, or a DAS28-CRP increase >0.6 compared with baseline and current DAS28-CRP ≥3.2 was used (short-lived flares).18 Major flare was defined as a flare persisting >12 weeks.

In the extension phase, the treatment strategies in both groups converged to the same strategy: treatment choices were left to the discretion of the treating rheumatologist and were based on local treatment protocols that included (1) disease activity measurement every 3–6 months and using treat-to-target to achieve at least low disease activity, (2) a preferential order of biological Disease Modifying Anti-Rheumatic Drugs (bDMARDS) (see online supplementary appendix 1) and (3) a bDMARD dose optimisation protocol (see  online supplementary appendix 2). Patients originally allocated to UC were therefore also able to initiate TNFi DR (see online supplementary appendix figure 1) but without specific DR advices. After March 2015, DAS28-CRP cut-off levels for low disease activity and remission were slightly lowered to 2.9 and 2.4, as it was shown that DAS28-CRP thresholds should be slightly lower in comparison with DAS28-ESR.19

### Supplementary Material

Supplementary material

### Supplementary Material

Supplementary appendix

### Outcomes

For the extension phase, the same endpoints were used as in the original DRESS study, although in an explorative, non-hypothesis testing manner. The primary endpoint was the difference in cumulative incidence of major flare between DR and UC group, during the extension phase and during the entire study (0–36 months).

Secondary endpoints were cumulative incidence of patients with short-lived flares, difference in mean time-weighted (MTW) DAS28-CRP score, MTW functioning, measured with the health assessment questionnaire-disability index (HAQ-DI) and quality of life at 36 months measured with EQ5D-5L (EUROQOL 5 dimensions 5 levels questionnaire), proportions of patients in whom DR or discontinuation was successful, bDMARD use, mean change (Δ) in Sharp van der Heijde score (SvdH) and cumulative incidence and incidence density (ID) of (S)AE.

Successful DR was defined as being on a lower dose than at baseline with concomitant low disease activity, measured both at 18 and 36 months. Successful discontinuation was defined as complete withdrawal of adalimumab or etanercept with concomitant low disease activity, measured both at 18 and 36 months.

In the extension phase, DAS28-CRP and HAQ-DI were measured at least every 6 months, and an EQ5D-5L was repeated at 36 months. For bDMARD use, the normalised proportion of the defined daily dose (DDD) was calculated with IQR with 1.0 being the full dose equivalent.

Radiographs of hands and feet were obtained at 36 months and assessed using the modified SvdH score, by the same two blinded, trained readers that assessed the original DRESS radiographs. Scoring was done pairwise with radiographs from months 18 and 36 in known sequential order, but without rescoring baseline and 18 months, for efficiency reasons, as suggested for long-term studies.20 Mean Δ in SvdH and proportion of patients with ΔSvdH exceeding three different cut-off levels were compared between groups: (1) minimal clinical important change (MCIC) of eight points in 18 months,21 (2) smallest detectable change (SDC)22 23 and (3) 0.5 SvdH units for minimal radiographic progression.

### Statistical analysis

Stata IC V.13.1 was used. In the DRESS study, per protocol (PP) analyses were used for the primary outcomes because of the non-inferiority nature of the analyses. Because of (1) the more exploratory analyses in this extension phase, (2) difficulty defining ‘per protocol’ when treatment decisions are left to the treating physician and (3) minor differences in PP and intention-to-treat (ITT) analyses in the original study, an ITT approach was chosen. Patients who did not give informed consent or were lost to follow-up before 24 months were excluded. All analyses were done for the extension phase and when appropriate for the entire study. Since previously no differences between TNFi (adalimumab or etanercept) were found, stratified analyses were deemed unnecessary for the extension phase.

For the primary analysis, we kept the original non-inferiority margin of 20% difference in major flare between DR and UC groups, based on the same reasoning as mentioned before.16 Point estimates with CI (95% CI) of the difference in cumulative incidence of major flare between groups were calculated and the upper limits of the CI were compared with the non-inferiority margin. A t-test compared mean/median medication use (MTW) DAS28-CRP, HAQ-DI and EQ5D-5L. Differences in cumulative incidence of flares and levels of disease activity at 24, 30 and 36 months were compared by χ2 testing. Different time points were tested separately, with no repeated measure analyses performed.

### Role of the funding source

This study was funded by the department of rheumatology at the Sint Maartenskliniek Nijmegen, The Netherlands.

## Results

Of 180 patients included in the DRESS study, 172 patients were enrolled in the extension phase: 115 patients in the DR group and 57 in the UC group (figure 1). Baseline characteristics at start of the extension phase were similar between groups, except for higher prevalence of conventional synthetic DMARD (csDMARD) comedication in the UC group (table 1). The percentage of missing data was low: 2% of planned visits and 2%–8% missing per variable, therefore multiple imputation was deemed unnecessary and simple imputation using last observation carried forward in case the last observation was missing or mean of previous and next were calculated for in-between missings.

Table 1

Patient characteristics at DRESS baseline and at start of DRESS extension phase

Figure 1

Flow chart. DRESS, Dose REduction Strategy of Subcutaneous TNF inhibitors.

### Flaring

The cumulative incidences of major flare during the extension phase were 12/115 (10%) in the DR and 7/57 (12%) in the UC group (difference −2%, 95% CI −8% to 15%). The upper limit of the 95% CI around the difference was <20%, compatible with non-inferiority of DR to UC group. The cumulative incidence from month 0–36 was 20/115 (17%) in the DR and 8/57 (14%) in the UC group (difference 3%, −10 to 15). There was no significant difference in cumulative incidence of short-lived flares during the extension phase: 49/115 (43%, 33% to 52%) in the DR and 20/57 (35%, 23% to 49%) in the UC group. From month 0–36, the cumulative incidence of flare remained different between groups: 96/115 (83%, 75% to 90%) in the DR and 25/57 (44%, 31% to 58%) in the UC groups. Additional analyses within the DR and UC groups on the occurrence of major and short-lived flares comparing adalimumab with etanercept, showed no significant differences in both the extension phase (18–36 months) as well as the whole study duration (0–36 months).

### Disease activity, function and quality of life

In the extension phase, MTW-DAS28-CRP was 2.2 (SD 0.7) in the DR group and 2.1 (SD 0.7) in the UC group (difference 0.08, –0.15 to 0.30). MTW–DAS28-CRP from 0 to 36 months was 2.3 (SD 0.6) in the DR group and 2.1 (SD 0.7) in the UC group (difference 0.16, –0.03 to 0.35). DAS28-CRP, HAQ-DI and EQ5D-5L remained stable during the extension phase and complete follow-up and were not significantly different between groups at any time point (figure 2). Disease activity states were not significantly different between groups at any time point in the extension phase (see online supplementary appendix table 1).

Figure 2

Mean (A) disease activity (measured with DAS28-CRP), (B) functioning (measured with HAQ-DI) and (C) quality of life (measured with EQ5D-5L). DAS28-CRP, 28 joint count-based disease activity score with C reactive protein; HAQ-DI, health assessment questionnaire-disability index. EQ-5D-5L, Euroqol 5 dimensions 5 levels questionnaire.

### TNFi tapering and medication use

In the intervention phase, 23/115 (20%, 13% to 28%) patients in the DR group had successfully discontinued their bDMARD, 52/115 (45%, 36% to 55%) successfully reduced their bDMARD and in 40/115 (35%, 26% to 44%) no DR was possible. Nineteen out of 115 (17%, 10% to 25%) patients in the DR group persisted being biological free with maintenance of low disease activity from the intervention phase until 36 months, and 33/115 (29%, 21% to 38%) of patients in the DR group persisted being successfully dose reduced from the intervention phase until 36 months.

During the intervention phase, 49/57 (86%, 74% to 94%) patients in the UC group did not attempt DR (eight patients tapered or discontinued their bDMARD, mostly due to AEs). In the extension phase, in 32/49 (65%, 50% to 78%) patients, a DR attempt was made. Of these, 19/49 (39%, 25% to 54%) were successfully dose reduced and 7/49 (14%, 1% to 27%) were successfully discontinued at 36 months. In 12/32 (38%, 21% to 56%) patients in the UC group in whom a DR attempt was made, a short-lived flare occurred. In patients in the UC group in whom a DR attempt was made, four experienced a major flare. Two of these patients reached low disease activity at the next visit after re-escalation or reinstallment. The remaining two patients had a major flare that was due to a high VAS score and high tender joint count. Re-escalation was thus deemed unnecessary by both the treating rheumatologist as well as the patient.

At study end, between group differences in numbers of patients successfully tapered or stopped were smaller but still existent (figure 3).

Figure 3

Dose optimisation in dose reduction and usual care group (percentages) at 18 and 36 months.

During the intervention phase, the proportion of the DDD of TNFi use was 0.50 (IQR 0.48 to 0.51) in the DR group and 0.92 (0.90 to 0.94) in the UC group (difference −0.42, –0.45 to −0.39). During the extension phase, this difference decreased but remained significant: 0.54 (0.51 to 0.58) in the DR and 0.67 (0.64 to 0.71) in the UC group (difference −0.13, –0.18 to 0.08). From 0 to 36 months, DDD was 0.53 (0.51 to 0.54) in the DR and 0.80 (0.78 to 0.82) in the UC group (difference −0.27, –0.30 to −0.25).

In the extension phase, no significant between-group differences in csDMARD use were observed. At 36 months, <10% of patients in both groups used oral steroids (difference −1%, −10% to 8%). During the extension phase, intramuscular or intra-articular glucocorticoid injections were given to 48/115 (42%, 33% to 51%) in the DR and 21/57 (37%, 24% to 51%) in the UC group (difference 5%, −11% to 21%).

One hundred and fifty-six patients (101 DR; 55 UC) had radiographs available at 18 and 36 months. No significant difference in mean progression score between groups was observed for the extension phase (table 2). Two out of 101 (2%) patients in the DR group and no patients in the UC group exceeded the MCIC. No significant between-group differences were seen for the SDC (calculated as 5.1 points) or minimal radiographic progression as cut-off values.

Table 2

### Safety

The cumulative incidence of AEs during the extension phase was equal in both groups: 39/115 (34%, 25% to 43%) in the DR and 22/57 (39%, 26% to 52%) in the UC group (difference −5%, −11 to 21), and the number of patients with SAEs was also not different between groups (difference 3%, −11 to 15). From month 0–36, 103/115 (90%, 82% to 94%) patients had an AE in the DR group and 54/57 (95%, 85% to 99%) in the UC group (difference −5%, −14% to 4%) and the number of patients with SAEs was different (difference 17%, 8% to 31%), caused by a higher incidence of elective surgery in the DR group. Overall, low IDs per 100 patient-years were observed for other SAE categories with no significant between-group differences (see online supplementary appendix table 2).

## Discussion

To our knowledge, this is the first study investigating long-term effects of disease activity-guided DR of adalimumab and etanercept in patients with RA. Results show that the initial efficacy and safety of this strategy are maintained. No relevant difference in the number of major flares could be demonstrated between DR and UC group, and disease activity, functioning and quality of life were also very similar. Furthermore, no significant difference in radiographic progression was found, although this might be caused by less contrast between groups, due to the converging treatment strategies. However, other benefits of tapering, including less AEs (eg, infections), were not observed.

There are some factors to take into account when interpreting our data. The design choices that were made for the extension study have advantages but also some drawbacks. Considering the latter, the convergence of strategies between groups and subsequent loss of contrast should lead to caution when interpreting the lack of differences between groups. The similarity in outcomes may be caused by the former DR group doing well, but also in part by the original UC group doing worse than before. However, the latter seems less likely, considering the very stable 3-year course in disease activity, functioning and quality of life. The outcomes are also highly comparable between the DR group from 18 to 36 months and the UC group from 0 to 18 months. Furthermore, in the extension phase of this study, flare criteria were altered since it was shown that cut-off values of DAS28-CRP for low disease activity and remission should be slightly lower than the validated flare criterion cut-offs using DAS28-ESR.19 It is unclear, however, how this would have altered our results. Tight control would have become even more tight, but this would have occurred in both DR and UC groups. Future studies should investigate how using different flare criteria influences treatment strategy outcomes.

Nevertheless, the design of the study did allow to assess—in the former UC group—what level of TNFi DR can be achieved when no specific DR advice is given. Interestingly, in the majority of those patients an attempt to dose reduce was observed, and subsequent results were also comparable with those in the original DR group. This further supports generalisability of the results to clinical practice.

Although there was some drop-out during the extension phase, for the primary outcome, our study seems well powered. A type II error might be present for analyses we did not power for. In the design of the original DRESS study, a sample size calculation showed that to be able to reject the null hypothesis in this study (ie, the intervention is inferior compared with the control arm by more than the non-inferiority margin) with a power of 80%, and accounting for a 10% drop-out, 180 patients in total were included. At the end of the long-term extension phase (month 36), 113 and 57 patients were still included. Thus, this is only slightly below the numbers as calculated above and total drop-out is still below 10%.

Comparison of our findings with other long-term studies on disease activity-guided DR or discontinuation is difficult, due to paucity and heterogeneity of these studies, with different DR strategies (eg, interval lengthening vs dose tapering, gradual vs fixed DR, or using different tapering schemes) and different definitions for relapse or flare.10 A recently published paper of Raffeiner et al is the only other study to show long-term (median follow-up 3.6 years) data from a prospective semirandomised tapering trial.24 However, only etanercept was studied, and fixed dose halving was used instead of disease activity-guided DR versus continuation of etanercept. Reassuringly, the outcomes of this study were very similar to ours, with no significant differences in clinical outcomes and radiographic progression, although with much higher absolute baseline radiographic damage.

Two points are of interest with regard to AEs, SAEs being more frequent in the DR group and no observed benefits of TNFi tapering on risk for infections. First, the higher incidence of SAEs seems an artefact, because it is almost exclusively caused by more elective surgery and SAEs that arose from study-related PET/CT scanning, that was only done in the DR group, resulting in information bias. Second, we were not able to demonstrate lower infection risks in the DR group, which is in contrast to Raffeiner et al and to what might be pharmacologically expected.3 This difference in outcome might be caused by lower patient numbers and follow-up time and less contrast between the treatment arms in the extension phase. Furthermore, duration of TNFi use before study start was much longer in our study and as patients susceptible to infections would have been more likely to have discontinued their bDMARD before inclusion, this could have led to selection bias (healthy survivor bias).

In conclusion, a disease activity-guided DR strategy of TNFi in patients with RA doing well seems a reasonable long-term approach in RA treatment. Further optimisation of this strategy could consist of identification of predictors for successful DR or discontinuation, as this might prevent short-lived flaring.

## Acknowledgments

The study received no external funding. We would like to thank all the rheumatologists in the Sint Maartenskliniek Nijmegen and Woerden, all the specialised nurses and Kimberly Bouman for participating in data collection.

View Abstract

## Footnotes

• Contributors CAMB, NvH, FHJvdH, JF, RFvV, HB, AvdM and AAdB were involved in the study design. CAMB, NvH, AvdM, FHJvdH and AAdB were involved in the data collection. CAMB, NvH, JF, AvdM and AAdB performed the data analyses. All authors were involved in writing and revision of the manuscript. CAMB declares she had full access to all the data in this study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

• Competing interests HB received grants and personal fees from Pfizer and AbbVie during the conduct of the study; grants and personal fees from Roche, BMS, MSD, UCB, all outside the submitted work. RFvV received grants from AbbVie, Amgen, BMS, GSK, Pfizer, Roche, UCB and personal fees from AbbVie, Biotest, BMS, Celgene, Crescendo, GSK, Janssen, Lilly, Merck, Novartis, Pfizer, Roche, UCB, Vertex, outside the submitted work. JF received a research grant from BMS. AAdB received congress invitations from Roche and Abvie and an expert witness fee from Amgen. The other authors declare that they have no conflicts of interest.

• Ethics approval CMO region Arnhem Nijmegen.

• Provenance and peer review Not commissioned; externally peer reviewed.

• Data sharing statement The authors commit to making the relevant anonymized patient level data available on reasonable request.

## Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.