Effect of Magnetic Resonance Imaging vs Conventional Treat-to-Target Strategies on Disease Activity Remission and Radiographic Progression in Rheumatoid Arthritis

Key Points

Question

Does a magnetic resonance imaging (MRI)–guided treat-to-target strategy aiming for imaging remission lead to an increased rate of disease activity remission (disease activity score in 28 joints–C-reactive protein [DAS28-CRP]  <2.6) rate and less radiographic progression in patients with rheumatoid arthritis in clinical remission?

Findings

In this randomized clinical trial that included 200 patients with rheumatoid arthritis with DAS28-CRP scores less than 3.2 and no swollen joints, an MRI-guided strategy compared with a conventional treat-to-target strategy resulted in DAS28-CRP remission rates of 85% vs 88%, respectively, and no radiographic progression (66% vs 62%, respectively). Neither comparison was statistically significant.

Meaning

Using MRI for treatment guidance in patients with rheumatoid arthritis did not improve the rate of disease activity remission or radiographic progression compared with a conventional treat-to-target strategy.

Abstract

Importance

Whether using magnetic resonance imaging (MRI) to guide treatment in patients with rheumatoid arthritis (RA) improves disease activity and slows joint damage progression is unknown.

Objective

To determine whether an MRI-guided treat-to-target strategy vs a conventional clinical treat-to-target strategy improves outcomes in patients with RA in clinical remission.

Design, Setting, and Participants

Two-year, randomized, multicenter trial conducted at 9 hospitals in Denmark. Two hundred patients with RA in clinical remission (disease activity score in 28 joints–C-reactive protein [DAS28-CRP] <3.2 and no swollen joints) were enrolled between April 2012 and June 2015. The final follow-up visit was April 2017.

Interventions

Patients were randomly allocated (1:1) to an MRI-guided vs a conventional treat-to-target strategy. In the MRI-guided group, the treatment goal was absence of MRI bone marrow edema combined with clinical remission, defined as DAS28-CRP of 3.2 or less and no swollen joints. In the conventional group, the treatment goal was clinical remission.

Main Outcomes and Measures

Co-primary outcomes were proportions of patients achieving DAS28-CRP remission (DAS28-CRP <2.6) and with no radiographic progression (no increase in total van der Heijde–modified Sharp score) at 24 months. Significance testing for the primary outcome was based on 1-sided testing. Secondary outcomes were clinical and MRI measures of disease activity, physical function, and quality of life.

Results

Of 200 patients randomized (133 women [67%]; mean [SD] age, 61.6 [10.5] years; median baseline DAS28-CRP, 1.9 [interquartile range, 1.7-2.2]; van der Heijde–modified Sharp score, 18.0 [interquartile range, 7.0-42.5]), 76 patients (76%) in the MRI-guided group and 95 (95%) in the conventional group completed the study. Of these, 64 (85%) vs 83 (88%), respectively, reached the primary clinical end point (risk difference, −4.8% [1-sided 95% CI, −13.6% to + ∞; 1-sided P = .19]) and 49 (66%) vs 58 (62%), respectively, reached the primary radiographic end point (risk difference, 4.7% [1-sided 95% CI, −7.0% to + ∞; 1-sided P = .25). Of 10 key secondary end points, 8 were null and 2 showed statistically significant benefit for the MRI treat-to-target group. Seventeen patients (17%) in the MRI-guided treat-to-target group and 6 patients (6%) in the conventional treat-to-target group experienced serious adverse events.

Conclusions and Relevance

Among patients with RA in clinical remission, an MRI-guided treat-to-target strategy compared with a conventional treat-to-target strategy did not result in improved disease activity remission rates or reduce radiographic progression. These findings do not support the use of an MRI-guided strategy for treating patients with RA.

Trial Registration

ClinicalTrials.gov Identifier: NCT01656278

This randomized clinical trial compares the effect of a magnetic resonance imaging (MRI)–guided vs a conventional clinical treat-to-remission-target strategy on improving disease activity scores and radiographic progression in patients with rheumatoid arthritis (RA) in clinical remission.

Introduction

Rheumatoid arthritis (RA) is a severe inflammatory joint disease associated with increased morbidity and mortality.¹ Patients with RA typically have joint pain, functional impairment, and reduced quality of life and are at risk of progressive joint destruction. The treatment target for RA is clinical remission, defined as no clinical signs or symptoms of disease activity. Early treatment, “treat-to-target” strategies, and biological therapy in patients with severe disease have made clinical remission an achievable goal.^2,3,4 However, structural joint damage develops in approximately 20% to 30% of patients fulfilling remission criteria,⁵ indicating that current remission criteria are not optimal.

The disease activity score in 28 joints–C-reactive protein (DAS28-CRP) is a composite disease activity score that is based on swelling and tenderness of 28 joints, the patient’s assessment of disease activity, and C-reactive protein and is commonly used to monitor RA disease activity.⁶ Structural joint damage progression is monitored using conventional radiography of hands and feet.

Magnetic resonance imaging (MRI) is a more sensitive measure of joint inflammation and damage than clinical examination or radiography.^7,8 MRI-detected synovitis and in particular bone marrow edema, which histologically represents osteitis,^9,10 are associated with subsequent bone damage progression in patients with active RA and in patients with RA who are in remission.^11,12,13,14 The high sensitivity of MRI for joint inflammation and damage, the presence of subclinical inflammation measured by MRI, and the association of these MRI measures with subsequent increased erosive progression have been acknowledged in European League Against Rheumatism (EULAR) recommendations.¹⁵ However, it remains unclear whether using MRI measures to guide treatment improves outcomes in RA compared with current clinical treatment strategies.

The IMAGINE-RA study was designed to assess whether a treat-to-target strategy that used systematic MRI assessments targeting absence of bone marrow edema improved clinical and radiographic outcomes, compared with a conventional disease activity–guided treat-to-target strategy in patients with RA in clinical remission.

Methods

The study was a 2-year, investigator-initiated, randomized multicenter trial conducted in Denmark. The trial protocol (Supplement 1) was approved by the Regional Committees on Health Research Ethics and the Danish Data Protection Agency and was carried out in compliance with the Declaration of Helsinki and the International Conference on Harmonization Guidelines for Good Clinical Practice. All patients gave written informed consent. End points were analyzed according to the intention-to-treat (ITT) principle (wherein patients who stopped treatment were invited for follow-up) as stated in the statistical analysis plan (Supplement 1).

Setting and Patients

Patients were recruited from outpatient clinics at 9 Danish rheumatology departments. Inclusion criteria were (1) presence of RA according to the American College of Rheumatology (ACR)/EULAR 2010 criteria¹⁶; (2) age 18 years and older; (3) clinical remission, defined as DAS28-CRP less than 3.2 and no swollen joints based on investigator assessment; (4) anticyclic citrullinated peptide (CCP) value greater than the upper limit of normal; (5) presence of erosive disease (bone erosion on conventional radiography described by the local radiologist); and (6) receipt of conventional synthetic disease-modifying antirheumatic drugs (csDMARD) in stable dosage of monotherapy or combination therapy more than 6 weeks prior to inclusion. Based on a search in the national quality database (DANBIO),¹⁷ as of September 7, 2018, 11 954 (49%) of a total of 24 382 patients with RA fulfilled these criteria (ie, were patients not previously treated with biologics, with DAS28-CRP ≤3.2 and no swollen joints), representing a substantial proportion of the RA population.

Exclusion criteria were previous treatment with biologics; intolerance to methotrexate (defined as tolerating <7.5 mg/wk); intramuscular, intra-articular, or intravenous glucocorticoid 6 weeks or less prior to inclusion; oral glucocorticoid of more than 5 mg/d; changes in oral glucocorticoid dose less than 3 months prior to inclusion; liver enzymes more than 2 times the upper limit of normal at screening; contraindications for tumor necrosis factor inhibitors; and contraindications for MRI. Ultrasonography and use of MRI outside the study protocol to visualize joint inflammation were not allowed at screening or during the study.

Twenty-two of 200 patients were unintentionally included before final trial registration was approved (and publicly available) in ClinicalTrials.gov. No data were analyzed before the entire study was completed and the database was locked.

Randomization

Patients were randomized 1:1 to either an MRI-guided treat-to-target treatment strategy or a conventional DAS28-CRP–guided treat-to-target (conventional treat-to-target) treatment strategy. To ensure that patients and investigators were unaware of the treatment assignment before study entry, a computer-generated allocation concealment process was developed before enrolling patients. Randomization was not performed before eligibility criteria were electronically confirmed in the web-based case report form. Randomization was performed centrally, without stratification, and the sequence was generated by an independent statistician using a random number generator with a 1:1 allocation using random block sizes of 2, 4, and 6.

Assessments

Patients were assessed every 4 months during the 2-year follow-up period. DAS28-CRP was measured at every visit. Contrast-enhanced MRI of the dominant hand, including the second to fifth metacarpophalangeal joints and wrist, was performed in agreement with the Outcome Measures in Rheumatology RA MRI scoring system (OMERACT RAMRIS).¹⁸

In the MRI-guided treat-to-target group, MRI scans were conducted before every study visit and assessed for bone marrow edema by a single experienced evaluator unaware of the group assignment (B.E.). The result, “presence” vs “absence” of bone marrow edema, was available to the investigator at each visit. In the conventional treat-to-target group, MRI scans were performed at baseline, 12 months, and 24 months. However, MRI results were not reported to the investigator. Different MRI scanners (1.0T-3.0T) were used at different sites and were not changed during the study period (detailed descriptions of MRI sequences and scanners used are in eTable 1 and eTable 2 in Supplement 2). Radiographs of hands (posteroanterior view) and feet (anteroposterior view) were performed at baseline, 12 months, and 24 months. Investigators were blinded to these results. Quality assessment of all imaging was carried out by a single experienced radiographer (J.M.). MRI/radiographs were repeated if quality requirements were not fulfilled.

Treatment Strategy and Interventions

The treatment target in the MRI-guided treat-to-target group was remission based on the MRI, defined as absence of bone marrow edema, combined with clinical remission defined as DAS28-CRP of 3.2 or less and no swollen joints. The treatment target for the conventional treat-to-target group was clinical remission only. In both groups, treatment was escalated stepwise according to a predefined algorithm if the target was not reached (Box). Intra-articular glucocorticoids were administered to every clinically swollen joint.

Box. Treatment Algorithm for Escalation of Conventional Synthetic Disease-Modifying Antirheumatic Drugs (csDMARDs) and Biologic Treatment^a.

csDMARD

• Step 1

  • csDMARD monotherapy < maximum dose

    • csDMARD monotherapy at less than maximum tolerated dose (maximum dose: up to methotrexate, 25 mg/wk; sulfasalazine, 3 g/d; hydroxychloroquine, 400 mg/d; or leflunomide, 20 mg/d. Minimum dose: methotrexate, 7.5 mg/wk; sulfasalazine, 1 g/d; hydroxychloroquine, 200 mg/d; leflunomide, 10 mg/d)

• Step 2

  • csDMARD monotherapy maximum dose or csDMARD combination therapy < maximum dose

    • csDMARD monotherapy at maximum tolerated dose (maximum dose: up to methotrexate, 25 mg/wk; sulfasalazine, 3 g/d; hydroxychloroquine, 400 mg/d; or leflunomide, 20 mg/d) or csDMARD combination therapy at less than maximum tolerated dose (maximum dose up to: methotrexate, 25 mg/wk; sulfasalazine, 2 g/d; hydroxychloroquine, 200 mg/d)

• Step 3

  • csDMARD combination therapy maximum dose

    • csDMARD combination therapy at maximum tolerated dose (maximum dose: up to methotrexate, 25 mg/wk; sulfasalazine, 3 g/d; hydroxychloroquine, 400 mg/d).

    • In case of adverse reactions or intolerance to 2 of the following: methotrexate, sulfasalazine, hydroxychloroquine in triple therapy, patients were instead receiving monotherapy leflunomide, 20 mg/d, without prior loading dose.

Biologic Treatment

• Step 4

  • Adalimumab and csDMARD monotherapy

    • Adalimumab (40 mg subcutaneously every other wk) and csDMARD monotherapy (methotrexate or leflunomide or sulfasalazine, prioritized sequence) at the maximum tolerated dose (maximum dose up to: methotrexate, 25 mg/wk; sulfasalazine, 2 g/d; hydroxychloroquine, 200 mg/d).

• Step 5

  • Another TNF inhibitor and csDMARD monotherapy

    • Other TNF inhibitor in combination with csDMARD monotherapy (methotrexate or leflunomide or sulfasalazine, prioritized sequence) at the maximum tolerated dose (maximum dose up to: methotrexate 25 mg/wk, sulfasalazine 2g/d, hydroxychloroquine 200 mg/d).

• Step 6

  • Biologic DMARD 3 and csDMARD monotherapy

    • Biologic DMARD treatment, decided by the investigator (preferably a non-TNF inhibitor) in combination with csDMARD monotherapy (methotrexate or leflunomide or sulfasalazine, prioritized sequence) at the maximum tolerated dose (maximum dose up to: methotrexate, 25 mg/wk; sulfasalazine, 2 g/d; hydroxychloroquine, 200 mg/d).

• Step 7

  • Biologic DMARD 4 and csDMARD monotherapy

    • Biologic DMARD treatment, decided by the investigator (preferably a non-TNF inhibitor) in combination with csDMARD monotherapy (methotrexate or leflunomide or sulfasalazine, prioritized sequence) at the maximum tolerated dose (maximum dose up to: methotrexate, 25 mg/wk; sulfasalazine, 2 g/d; hydroxychloroquine, 200 mg/d).

• Step 8

  • Biologic DMARD 5 and csDMARD monotherapy

    • Biologic DMARD treatment, decided by the investigator (preferably a non-TNF inhibitor) in combination with csDMARD monotherapy (methotrexate or leflunomide or sulfasalazine, prioritized sequence) at the maximum tolerated dose (maximum dose up to: methotrexate, 25 mg/wk; sulfasalazine, 2 g/d; hydroxychloroquine, 200 mg/d).

Primary Outcomes

Two predefined co-primary end points were assessed at 24 months. The primary clinical end point was the proportion of patients achieving DAS28-CRP remission, ie, DAS28-CRP of less than 2.6; the primary radiographic end point was the proportion of patients with no radiographic progression, ie, no increase (>0) in total van der Heijde–modified Sharp score (vdHSS; range, 0-448, with higher scores indicating more damage) from baseline to 24 months. Radiographs were scored on one occasion by an experienced, trained reader (L.M.Ø.) who was aware of the chronological order of the 3 sets of images but was unaware of all other data.^19,20,21,22

Secondary Outcomes

Predefined key secondary end points were ACR/EULAR remission using the Boolean-based definition (eTable 3 in Supplement 2) and change between baseline and 24-month follow-up in the following outcomes: physical function assessed by Health Assessment Questionnaire (HAQ; range, 0 [best] to 3 [worst]); quality of life (using the 36-item Short-Form Health Survey [SF-36] and EuroQol–5 dimensions [EQ-5D] score; range, 1 to −0.59, with 1 representing best possible health); total vdHSS and MRI synovitis, osteitis, and erosion; and progression in MRI erosions (ie, change in RAMRIS erosion score ≤0). MRI scans obtained at baseline, 12 months, and 24 months were scored by an experienced trained reader (D.G.) aware of the chronologic order of the 3 sets of images but blinded to all other data.^23,24,25

Additional secondary outcomes were ACR/EULAR remission using the Boolean-based definition; DAS28-CRP; DAS28-CRP remission; MRI synovitis and osteitis at 12 months; change between baseline and 12-month follow-up in physical function assessed by HAQ, quality of life (using SF-36 and EQ-5D questionnaires); change between baseline and 12-month and 12 to 24 month follow-up in total vdHSS and MRI erosion; and proportion of patients with no radiographic progression, ie, no increase (>0) in total vdHSS and proportion of patients with no progression in MRI erosion (ie, no increase [>0] in RAMRIS erosion score (presented in eTable 4 in Supplement 2).

Exploratory End Points

All exploratory end points were defined before analyses were performed. The OMERACT RAMRIS scoring systems for tenosynovitis and joint space narrowing (JSN)^23,24,26,27 were developed after the protocol was written and the trial registered in ClinicalTrials.gov but were incorporated into the statistical analysis plan before data analyses were performed. Changes in JSN and tenosynovitis from baseline to 24 months, combined scores of inflammation (sum of scores of synovitis, osteitis, and tenosynovitis), and damage (sum of scores of erosion and JSN) were analyzed.

Other exploratory end points were the composite indices Simple Disease Activity Index remission and Clinical Disease Activity Index remission (eTable 3 in Supplement 2), ACR core set assessment of tender joint count, swollen joint count, pain, patient and investigator global assessment of disease activity, and fatigue.

Adverse events were evaluated during the study period, and serious adverse events (SAEs) were registered.

Statistical Analyses

The trial’s primary objective was met if a statistically significant beneficial effect could be demonstrated on both end points (P < .05; 1-sided tests). The sample size was determined based on the assumption that 80% vs 60% of participants randomized to MRI-guided vs conventional treat-to-target strategies, respectively (assumed treatment difference of 20%), would reach the primary clinical end point (DAS28-CRP <2.6), and that 90% vs 75%, respectively, would reach the primary radiographic end point (no radiographic progression) at month 24. These assumptions were based on anti-CCP–positive patients with erosive RA in low disease activity taking csDMARDs in a Danish cohort study²⁸ and previous tumor necrosis factor inhibitor clinical trials.²⁹

Statistical power of 80% for the primary clinical and radiographic end points required 64 and 79 patients, respectively, in each treatment group. With a total sample size of 200 patients, the trial would have good statistical power even with 20% attrition rates: Using Pearson χ² statistic, with a 1-sided significance level of .05, a total sample size of 200 patients corresponded to an approximate power of 87.9%. One-sided significance testing was selected because investigators anticipated no possibility of an inferior result with the MRI treatment strategy because the strategy was more intensive. Furthermore, 2-sided testing would have required a trial of approximately twice the size, which was not considered feasible for funding reasons and the inability to perform approximately twice as many MRI scans due to capacity problems.

The co-primary end points and adverse events were analyzed on the ITT principle (full analysis set: all patients randomized and exposed to at least the baseline assessment independent of adherence to the protocol). The ITT principle dictates that the patients were evaluated on the basis of the randomized treatment regimen irrespective of the actual treatment given, ie, participants allocated to a treatment group were followed up, assessed, and analyzed as members of the group irrespective of their adherence to the planned course of treatment (statistical analysis plan in Supplement 1). Estimates of treatment effect for primary and secondary dichotomous outcomes were expressed as adjusted risk differences with 90% CIs (1-sided) and 95% CIs (2-sided) using multiple imputation based on the Markov chain Monte Carlo method assuming missing data were missing at random.

Secondary continuous outcomes were analyzed using repeated-measure linear mixed models, also valid under the assumption of data missing at random. Robustness of the analyses was assessed with separate sensitivity analyses conducted on the full analysis set, the per-protocol population, and the 178 patients included after final registration in ClinicalTrials.gov. Sensitivity analyses included “worst-case” and “best-case” imputation and tipping point analyses to take into account the possibility that data were not missing at random (eTable 9 and eAppendix in Supplement 2). The smallest detectable change for total vdHSS was calculated.³⁰ The trial would be interpreted as positive only if a statistically significant beneficial effect (P < .05) was demonstrated on both end points. No adjustments for multiple testing were applied, and therefore inferences drawn from the reported P values and 95% CIs from secondary outcomes should be considered exploratory. Analyses were performed using R version 3.3.3 (lme4 and mitml package; R Project for Statistical Computing).

Results

Disposition and Baseline Characteristics of Patients

Between April 2012 and June 2015, 228 patients were screened and 200 were randomized (100 in each group), included in the primary analyses, and constituted the ITT population. The last patient visit took place in April 2017. Seventy-six patients in the MRI-guided treat-to-target group and 95 patients in the conventional treat-to-target group completed the study (Figure 1). Participants in the MRI-guided treat-to-target group had a lower rate of DAS28-CRP remission at baseline (DAS28-CRP <2.6) (86% vs 96%) and higher HAQ and patient visual analogue scale global, pain, and fatigue scores (Table 1).

Figure 1. Patient Flow Throughout the IMAGINE-RA Randomized Clinical Trial.

Table 1. Baseline Characteristics.

Characteristic	Treat-to-Target Strategy, Median (Interquartile Range)a
	Magnetic Resonance Imaging (n = 100)	Conventional (n = 100)
Demographic characteristics
Sex
Women, No. (%)	69 (69)	64 (64)
Men, No. (%)	31 (31)	36 (36)
Age, mean (SD), y	62.7 (9.6)	60.5 (11.1)
Disease duration, y	9.0 (5.0-16.0)	11.0 (4.0-17.0)
Height, mean (SD), cm	169.0 (8.2)	169.2 (9.3)
Weight, mean (SD), kg	74.9 (14.6)	75.4 (17.3)
Clinical measures
IgM rheumatoid factor positive, No. (%)	86 (86)	88 (88)
CRP, mg/L	7.0 (4.0-9.0)	6.0 (3.0-8.3)
DAS28-CRPb	2.0 (1.7-2.3)	1.9 (1.6-2.1)
DAS28-CRP <2.6, No. (%)	86 (86)	96 (96)
American College of Rheumatology-European League Against Rheumatism Boolean remission, No (%)c	34 (34)	43 (43)
Simplified Disease Activity Indexd	2.8 (1.6-4.3)	2.3 (1.2-3.5)
Clinical Disease Activity Indexe	2.1 (0.9-3.5)	1.4 (0.6-2.7)
Tender joint count (0-28)	0 (0-0)	0 (0-0)
Swollen joint count (0-28)	0 (0-0)	0 (0-0)
Patient’s global assessment of disease activity on a visual analogue scale (0-100)	15.5 (6.8-29.0)	9.0 (3.0-20.3)
Patient’s assessment of pain, visual analogue scale (0-100)	15.0 (5.0-26.8)	9.0 (2.8-20.0)
Patient’s assessment of fatigue, visual analogue scale (0-100)	19.5 (6.0-32.0)	12.0 (4.0-27.3)
Investigator’s global assessment of disease activity, visual analogue scale (0-100)	2.0 (0.0-4.0)	2.0 (1.0-5.0)
Radiographic measures (van der Heijde–modified Sharp score)
Total (0-448)	20.0 (9.0-46.0)	15.0 (7.0-37.5)
Erosion (0-280)	9.0 (4.0-21.0)	7.0 (3.0-15.0)
Joint space narrowing (0-168)	10.0 (3.0-27.0)	9.0 (2.0-22.8)
Magnetic resonance imaging measures (Rheumatoid Arthritis MRI Scoring System)
Synovitis (0-21)	6.0 (3.0-9.0)	5.5 (2.3-8.8)
Osteitis (0-69)	2.0 (0.0-6.0)	2.0 (0.0-5.0)
Erosion (0-230)	4.0 (1.8-10.0)	5.0 (2.0-9.0)
Bone marrow edema present, No. (%)	47 (47)	36 (36)
Exploratory magnetic resonance imaging measures (Rheumatoid Arthritis MRI Scoring System)
Tenosynovitis (0-54)	3.0 (0.0-5.8)	3.0 (0.0-6.0)
Combined inflammation (0-144)f	12.0 (8.0-18.3)	12.0 (6.0-17.0)
Joint space narrowing (0-84)	1.0 (0.0-4.0)	1.0 (0.0-6.0)
Combined damage (0-314)g	6.0 (2.0-12.5)	6.0 (2.0-14.0)
Function and quality of life
Health Assessment Questionnaire (0-3)h	0.44 (0.00-0.77)	0.00 (0.00-0.44)
36-Item Short Form Mental Component Summary score (0-100)i	59.0 (54.0-62.3)	58.0 (54.0-61.0)
36-Item Short Form Physical Component Summary score (0-100)i	47.5 (41.0-51.0)	50.0 (44.0-53.5)
EuroQol-5 dimensions (1 to −0.59)j	0.99 (0.80-0.99)	0.99 (0.82-1.00)
Treatment at inclusion, No. (%)
csDMARD monotherapy < maximum dose	46 (46)	62 (62)
csDMARD monotherapy maximum dose	24 (24)	17 (17)
csDMARD combination therapy < maximum dose	30 (30)	21 (21)
Nonsteroidal anti-inflammatory drug stable dose	10 (10)	6 (6)
Prednisolone stable dose	2 (2)	2 (2)

Abbreviations: CRP, C-reactive protein; DAS28-CRP, disease activity score in 28 joints CRP; csDMARD, conventional synthetic disease-modifying antirheumatic drug.

SI conversion factor: To convert CRP to nmol/L, multiply by 9.524.

^aData are presented as median (interquartile range) unless stated otherwise.

^bComposite score of 28-joint count of swollen and tender joints, patient global assessment of disease activity (on a visual analogue scale) and CRP, with remission defined as DAS28-CRP <2.6.

^cAt any time point, a patient must satisfy all of the following: swollen joint count ≤1, tender joint count ≤1, CRP ≤1 mg/dL, and patient global assessment of disease activity (≤1 on a 0-10 visual analogue scale).

^dComposite score based on a 28-joint count of swollen and tender joints, patient and investigator global assessment of disease activity, and CRP (range, 0-86), with remission defined as Simplified Disease Activity Index ≤3.3.

^eComposite score based on a 28-joint count of swollen and tender joints, patient and investigator global assessment of disease activity (range, 0-76), with remission defined as Clinical Disease Activity Index ≤2.8.

^fSum score of magnetic resonance imaging synovitis, osteitis, and tenosynovitis.

^gSum score of magnetic resonance imaging erosion and joint space narrowing.

^hPatient-related score of physical function (disability), from 0 (best) to 3 (worst).

ⁱPatient-related score of physical and mental function based on 8 health concepts, with a high score defining a favorable health state.

^jPatient-related quality of life from −0.59 (worst) to 1 (best).

Primary Outcomes

There was no difference between the treatment groups for the primary clinical outcome (MRI-guided treat-to-target group [64/75] vs conventional treat-to-target group [83/94]: multiple imputation RD, −4.8% [1-sided 95% CI, −13.6% to + ∞; 1-sided P = .19]; radiographic outcome: MRI-guided treat-to-target group [49/74] vs conventional treat-to-target group [58/93]: multiple imputation RD, 4.7% [1-sided 95% CI, −7.0% to + ∞; 1-sided P = .25]). Data from 1-sided test hypotheses (with 95% CIs) for the 2 co-primary end points and 2-sided test hypotheses (with 95% CIs) for key secondary end points are presented in Table 2 and Figure 2.

Table 2. Primary, Secondary, and Exploratory Outcomes at 24 Months.

	MRI-Guided Treat-to-Target Strategya		Conventional Treat-to-Target Strategya		Adjusted Difference Between Groupsb	P Valueb
	Total No.	No. (%)	Total No.	No. (%)
Primary End Points
Clinical
DAS28-CRP remission (<2.6)	75	64 (85.3)	94	83 (88.3)	−4.8 (−13.6 to + ∞)c	.19
Radiographic
No radiographic progressiond	74	49 (66.2)	93	58 (62.4)	4.7 (−7.0 to + ∞)c	.25
Secondary End Points
Clinical
American College of Rheumatology–European League Against Rheumatism Boolean remission	75	37 (49.3)	94	30 (31.9)	14.0 (0 to 28.1)c	.05
DAS28-CRP	75	1.9 (0.6)	94	2.1 (0.6)	−0.2 (−0.3 to 0.0)	.09
Function and quality of life
Change in Health Assessment Questionnaire	76	−0.05 (0.21)	94	0.09 (0.22)	−0.14 (−0.2 to −0.1)	<.001
Change in EuroQol-5 dimensions	70	0.04 (0.12)	90	0.02 (0.13)	0.02 (−0.02 to 0.06)	.28
Change in 36-Item Short Form Physical Component Summary	69	1.1 (8.3)	89	−0.2 (8.6)	1.3 (−1.3 to 4.0)	.33
Magnetic resonance imaging measures (Rheumatoid Arthritis MRI Scoring System)
Change in osteitis	74	−1.8 (4.8)	93	−0.1 (4.8)	−1.8 (−3.2 to −0.3)	.02
Change in synovitis	71	−0.5 (2.9)	86	0.3 (3.0)	−0.8 (−1.8 to 0.1)	.07
Change in erosion	74	0.5 (1.5)	94	0.6 (1.5)	−0.1 (−0.6 to 0.4)	.66
No progression in magnetic resonance imaging erosion	74	59 (79.7)	93	70 (75.3)	3.3 (−8.7 to 15.3)	.59
Radiographic measures (vdHSS)
Change in total vdHSS	75	1.0 (2.8)	93	1.3 (2.9)	−0.3 (−1.1 to 0.6)	.56
Exploratory End Points
Clinical
Simplified Disease Activity Index remission (≤3.3)	75	48 (64.0)	90	56 (62.2)	−1.4 (−14.6 to 11.9)c	.84
Clinical Disease Activity Index remission (≤2.8)	76	53 (69.7)	91	59 (64.8)	1.3 (−11.7 to 14.3)c	.84
Tender joint count (0-28)	76	0.2 (1.1)	94	0.5 (1.1)	−0.2 (−0.6 to 0.1)	.17
Swollen joint count (0-28)	76	0.0 (0.8)	94	0.3 (0.9)	−0.3 (−0.5 to −0.0)	.04
Patient’s global assessment of disease activity on a visual analogue scale (0-100)	76	15.5 (15.3)	95	21.2 (15.8)	−5.7 (−10.4 to −1.0)	.02
Patient’s assessment of pain, visual analogue scale (0-100)	76	14.2 (14.7)	94	18.7 (15.2)	−4.5 (−9.0 to 0.0)	.05
Patient’s assessment of fatigue, visual analogue scale (0-100)	76	21.8 (16.4)	94	24.4 (16.8)	−2.6 (−7.6 to 2.4)	.31
Investigator’s global assessment of disease activity, visual analogue scale (0-100)	76	4.7 (7.1)	91	6.9 (7.2)	−2.3 (−4.4 to −0.1)	.04
Patient with normal function (Health Assessment Questionnaire score ≤0.5)	76	61 (80.3)	94	75 (79.8)	−2.3 (−13.8 to 9.2)c	.70
Change in 36-Item Short Form Mental Component Summary	69	−0.5 (8.1)	89	−0.9 (8.2)	0.5 (−2.1 to 3.1)	.73
Magnetic resonance imaging measures (Rheumatoid Arthritis MRI Scoring System)
Change in tenosynovitis	70	−0.9 (2.8)	86	0.3 (2.9)	−1.2 (−2.1 to −0.3)	.007
Change in combined inflammation	70	−2.9 (8.8)	85	0.7 (8.9)	−3.6 (−6.4 to −0.8)	.01
Change in joint space narrowing	74	0.1 (1.3)	94	0.4 (1.4)	−0.3 (−0.7 to 0.2)	.24
Change in combined damage	74	0.6 (2.6)	94	1.0 (2.7)	−0.3 (−1.2 to 0.5)	.40
Radiographic measures (vdHSS)
Change in vdHSS erosion score	75	0.6 (1.5)	93	0.8 (1.5)	−0.2 (−0.7 to 0.2)	.29
Change in vdHSS joint space narrowing score	75	0.4 (1.7)	93	0.5 (1.8)	0.0 (−0.6 to 0.5)	.93
Treatment at year 2
csDMARD monotherapy	76	23 (31.6)	95	68 (71.6)	−36.8 (−50.0 to −23.6)c	<.001
csDMARD combination therapy	76	17 (22.4)	95	25 (26.3)	−0.5 (−11.9 to 10.9)c	.94
Biologic treatment	76	35 (46.1)	95	2 (2.1)	37.2 (26.9 to 47.4)c	<.001
First biologic	76	15 (19.7)	95	2 (2.1)	16.8 (8.0 to 25.5)c	<.001
Second or later biologic	76	20 (26.3)	95	0 (0)	21.0 (12.8 to 29.2)c	<.001

Abbreviations: csDMARD, conventional synthetic disease-modifying antirheumatic drug; DAS28-CRP, disease activity score in 28 joints–C-reactive protein; vdHSS, van der Heijde–modified Sharp score.

^aData are presented as No. (%) for dichotomous data and least squares means (SD) for continuous data.

^bFor end points with dichotomous data, adjusted risk differences and 95% CIs between groups were calculated by Wald tests and data for all 200 patients at 24 months were included using multiple imputation. For end points with continuous data, least squares mean differences between groups were calculated based on repeated-measures mixed linear models adjusted for baseline values. For the primary outcomes, confidence intervals were 1-sided 95% CIs and P values 1-sided.

^cData presented as percentages.

^dNo increase (change >0) in vdHSS from baseline to month 24.

Figure 2. Box and Whiskers Plots of the Co-primary Outcomes Stratified by Randomization Group.

Box and whiskers plots display the sample population stratified by group. Circles denote individual patients. Boxes mark quartiles, the band inside the box is the second quartile (the median), while the upper whisker represents the upper adjacent value, the third quartile plus 1.5 times the interquartile range, and the lower whisker represents the lower adjacent value, the lower quartile minus 1.5 times the interquartile range. The dashed line in A represents the threshold for Disease Activity Score in 28 joints C-reactive protein (DAS28-CRP) remission (DAS28-CRP <2.6). The dashed line in B represents the threshold for no radiographic progression (increase in total van der Heijde–modified Sharp score [vdHSS] ≤0). MRI indicates magnetic resonance imaging.

Secondary Outcomes

Among the 10 key secondary outcomes assessed at 24 months, most were not different between the 2 groups (Table 2). Two outcomes showed more improvement in the MRI-guided treat-to-target group than in the conventional treat-to-target group (Table 2). These 2 were change in HAQ (least squares mean, −0.14 [95% CI, −0.2 to −0.1]; P < .001) and change in the MRI target variable osteitis (−1.8 [95% CI, −3.2 to −0.3]; P = .02). There was no significant difference in synovitis scores between groups (−0.8 [95% CI, −1.8 to 0.1]; P = .07) or in other key secondary outcomes. The mean MRI erosive progression was low, with no difference between treatment groups.

Exploratory Outcomes

Of 16 exploratory outcomes, 11 showed no difference between the treatment groups and 5 showed more improvement in the MRI-guided treat-to-target group compared with the conventional treat-to-target group (Table 2). These 5 were swollen joint count, the patient’s global assessment of disease activity, the investigator’s global assessment of disease activity, MRI tenosynovitis, and the MRI combined inflammation score.

Sensitivity Analyses

Baseline characteristics and results of sensitivity analyses are shown in eTables 5-9 and eAppendix in Supplement 2. There were no substantial differences in these results compared with the primary results. However, in the per-protocol analyses, ACR-EULAR remission rates and SF-36 Physical Component Score improvements were higher in the MRI-guided treat-to-target group (eTable 8 in Supplement 2). When missing data were imputed according to a worst-case scenario, ie, in all, missing values were replaced with a negative result (affecting 24 discontinued patients in the MRI-guided treat-to-target group and 5 in the conventional treat-to-target group), there was a higher rate of DAS28-CRP remission in the conventional treat-to-target group compared with the MRI-guided treat-to-target group.

Post Hoc Analyses

Post hoc tipping point analyses suggested that it was unlikely that the MRI-guided treat-to-target group was superior to the conventional treat-to-target group for the outcome of DAS28-CRP remission and it was likely that the groups were similar in terms of radiographic progression (eAppendix in Supplement 2). There were no site-related effects on primary or secondary outcomes at 24 months when stratified by center (eTable 10 in Supplement 2).

Adverse Events

Seventeen patients (17%) in the MRI-guided treat-to-target group experienced 19 SAEs, and 6 patients (6%) in the conventional treat-to-target group experienced 7 SAEs (eTable 11 in Supplement 2). Six patients (3%) experienced serious infections (3 in the MRI-guided treat-to-target group vs 3 in the conventional treat-to-target group), 4 patients developed cancer (3 vs 1), and 2 died (1 vs 1) (eTable 12 in Supplement 2). Adverse events with probable or possible relation to treatment intensification in the MRI-guided treat-to-target group caused 7 (37%) of the 19 reported SAEs and discontinuation of 6 patients.

Treatment Intensification

During the study, based on the treatment dictated by the study algorithm (Box), treatment was intensified 173 times in the MRI-guided treat-to-target group vs 22 times in the conventional treat-to-target group. Corresponding percentages of patients who had their treatment escalated were 73% and 17%, respectively. At 24-month follow-up, 46% of the patients in the MRI-guided treat-to-target-group received biologic treatment vs 2% in the conventional treat-to-target group.

Discussion

Compared with the conventional treat-to-target approach, this trial showed no benefit of using MRI to guide treatment for the primary outcomes of achieving disease activity remission (determined by DAS28-CRP <2.6) and eliminating radiographic progression at 24-month follow-up. To our knowledge, this is the first randomized trial assessing the efficacy of using MRI measures as the primary treatment outcome in a treat-to-target strategy.

One potential explanation for the lack of improved remission rates in the MRI-guided treat-to-target group is that participants randomized to this group had a lower rate of remission at baseline than participants in the conventional treat-to-target group (86% vs 96%). Furthermore, both groups were treated intensively by injecting all swollen joints, which has been shown to provide more prolonged inflammation control,³¹ contributing to low disease activity in both groups.

Approximately one-third of patients had radiographic progression of RA (total vdHSS progression >0), while 17% progressed above the measurement error (smallest detectable change). In a cohort of 535 patients in remission, radiographic damage progression occurred in 20% to 30% of the patients.³² A similar proportion in the present study was seen despite the presence of several factors that would be expected to increase the measured progression: all patients being anti-CCP positive and having erosive disease at baseline (both predictors of radiographic progression³³) and the use of chronological reading, which may give a higher sensitivity to change.²⁰

At 24 months, 2 of 10 key secondary outcomes were statistically significant favoring the MRI-guided treat-to-target group (physical function [HAQ] and osteitis [the target variable in MRI-guided treat-to-target group]), but these results were not adjusted for multiple comparisons and should be interpreted with caution. Consistent with findings reported here, 2 randomized clinical trials in patients with early RA, with the treatment goal of absence of synovitis by ultrasonography, did not show significant improvement in clinical or radiographic outcomes compared with monitoring using clinical treat-to-target strategies.^34,35

More patients in the MRI-guided treat-to-target group had treatment escalations (73% vs 17%), received biologic treatment (46% vs 2%), and experienced SAEs (19 vs 7), which likely was related to the more intensive therapy administered in effort to eliminate MRI measures of inflammation in this group. Six patients (in 3 patients most likely related to biologic treatment) in the MRI-guided treat-to-target group withdrew because of AEs that occurred during the trial. The higher discontinuation rate in the MRI-guided treat-to-target group was related to higher rates of AEs, refusal to intensify treatment, and dropouts. It is possible that patients who felt well were reluctant to repeat MRI scans and/or to intensify treatment.

A strength of the study was the strict design and systematic algorithmic clinical protocols.

Limitations

This study has several limitations. First, the nonblinded study design could have led to bias. Full blinding, however, was not considered feasible or ethically justified because it would have required 4 additional and unused MRI scans per patient in the conventional treat-to-target group. Second, the 2-group study design did not allow investigators to test other potential MRI treatment targets. Third, the between-group imbalance in certain baseline characteristics, such as the presence of remission at baseline, favored the conventional treat-to-target group. Fourth, only patients receiving csDMARDs, not biologic DMARDs, were eligible for inclusion, decreasing generalizability. Fifth, the primary clinical outcome (DAS28-CRP <2.6) may be an unambitious treatment goal. Other more stringent targets, eg, EULAR/ACR remission, could have been preferred because fulfilling such criteria leads to better health-related quality of life.³⁶ Sixth, using 1-sided significance testing at a P < .05 level in a superiority trial is a limitation. When the trial was designed, we assumed no possibility of an inferior result with the MRI intervention and 2-sided testing would, furthermore, have required a study of approximately twice the size, which was not considered feasible for capacity and funding reasons. However, our primary analyses had to demonstrate a statistically significant effect on both co-primary end points.

Conclusions

Among patients with RA in clinical remission, an MRI-guided treat-to-target strategy compared with a conventional treat-to-target strategy did not result in improved clinical disease activity remission or reduce radiographic progression. These findings do not support the use of an MRI-guided strategy for treating patients with RA.

Supplement 1.

Trial Protocol and Statistical Analysis Plan

Supplement 2.

eTable 1. Acquired MRI Sequences in the IMAGINE-RA Trial

eTable 2. MRI Units Used During the Study Period

eTable 3. Composite Disease Activity Scores Used in the IMAGINE-RA Study

eTable 4. Secondary Outcomes at 0-12, 12 and 12-24 Months

eTable 5. Comparison of Primary, Secondary, and Exploratory Outcomes at 24 Months Based on the Full Analysis Set

eTable 6. Baseline Characteristics of Patients Enrolled After the Correct Final Registration at www.ClinicalTrials.gov

eTable 7. Comparison of Primary, Secondary, and Exploratory Outcomes at 24 Months Based on Patients Enrolled After the Correct Final Registration at www.ClinicalTrials.gov

eTable 8. Comparison of Primary, Secondary, and Exploratory Outcomes at 24 Months Based on the Per-Protocol Population

eTable 9. Comparison of Primary Outcomes at 24 Months Based on The Full Analysis Set Applying Descriptive Statistics for Each Group and contrast between groups based on Odds Ratio Applying “Worst Case”, “Best Case Imputation” and Multiple-Imputation (MI) Analyses, Using the “Markov Chain Monte Carlo” (MCMC) Simulation Statement

eTable 10. Primary Outcomes at 24 Months Stratified by Center

eTable 11. Safety Data (Collected Adverse Events; Full Analysis Set)

eTable 12. Summary Narratives for Malignancies and Deaths

eFigure. Proportion of Patients Achieving Co-primary Outcomes at 24 Months, Based on the Intention-to-Treat Population With No Data imputation to replace missing data and Assessed Using Pearson’s Chi-Square Approximation With a 1-Sided Significance Level of 0.05

eAppendix. IMAGINE-RA: Sensitivity Analyses for the Intention To Treat (ITT) Population With Missing Outcome Data (Data Missing Not At Random & Tipping Point Analysis)

Supplement 3.

Data Sharing Statement