Article Text

Download PDFPDF

Extended report
Can flare be predicted in DMARD treated RA patients in remission, and is it important? A cohort study
  1. Benazir Saleem1,2,
  2. Andrew K Brown3,4,
  3. Mark Quinn3,4,5,
  4. Zunaid Karim1,2,
  5. Elizabeth M A Hensor1,2,
  6. Philip Conaghan1,2,
  7. Charles Peterfy6,
  8. Richard J Wakefield1,2,
  9. Paul Emery1,2
  1. 1Divsion of Musculoskeletal Disease, University of Leeds, Leeds, UK
  2. 2Leeds Teaching Hospital NHS, NIHR Leeds Muscoloskeletal Biomedical Research Unit, Leeds, UK
  3. 3Hull and York Medical School, University of York, York, UK
  4. 4Department of Rheumatology, York Teaching Hospital NHS Foundation Trust, York, UK
  5. 5Hull York Medical School, University of York, York, UK
  6. 6Synarc Inc, San Francisco, California, USA
  1. Correspondence to Paul Emery, Division of Musculoskeletal Disease, University of Leeds, Woodhouse Lane Leeds LS2 9JT, UK; p.emery{at}


Objectives The treatment target for patients with rheumatoid arthritis (RA) is remission. Imaging techniques and remission criteria may identify patients at risk of flare and associated consequences. This study aimed to determine the clinical, functional and imaging associations of disease flare in patients with RA in remission and any effect on long-term outcomes.

Methods RA patients in clinical remission as determined by their treating rheumatologist were assessed using clinical, remission criteria, imaging, functional and quality of life measures over 12 months. Flare was defined as any increase in disease activity requiring a change in therapy.

Results 26% of patients (24/93) in remission experienced a flare within 1 year. Fulfilment of remission criteria was not associated with a reduced likelihood of flare. Increased baseline ultrasound power Doppler (PD) activity (unadjusted OR (95% CI) 4.08 (1.26 to 13.19), p=0.014) and functional disability (Health Assessment Questionnaire Disability Index (HAQ-DI) per 0.1 unit1.27 (1.07 to 1.52), p=0.006) were independently associated with risk of flare. Patients who had a flare had significantly worse long-term clinical (Disease Activity Score 28; mean (95% CI) 2.90 (2.55 to 3.24) vs 2.26 (2.06 to 2.46), p=0.002) and functional outcomes (HAQ-DI; 0.412 (0.344 to 0.481) vs 0.322 (0.282 to 0.362), p=0.029) at 12 months compared with patients in sustained remission.

Conclusion The presence of PD activity was the most accurate determinant of flare in RA patients in remission. Flare was associated with worse clinical and functional outcomes. These results suggest ultrasound could form an important part of remission assessment in RA.

Statistics from

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.

European League Against Rheumatism (EULAR) recommendations state that remission is the goal of therapy for patients with rheumatoid arthritis (RA) and with modern therapeutic strategies this target disease state can be achieved more frequently.1,,4 Remission has been described as a ‘state of absent disease activity’ and is the antithesis of flare which is a ‘substantial increase in disease activity.’5

Studies have compared patients in sustained remission with those with continued disease activity6 and in different disease activity states.7 High disease activity has been associated with worse functional and radiographic outcomes when compared with remission or low disease activity.8

Once patients are in remission, the duration of this state is important. Patients with just short-term remission are more likely to experience radiographic structural progression compared with those achieving longer term remission.7 Therefore, a more appropriate target disease state should be sustained remission (the absence of disease activity and flare over the longest possible period of time). This may be more difficult to achieve as data suggest that approximately half of RA patients in remission may experience a disease flare within 24 months.9 Flare has been shown to be associated with more radiographic progression when compared with patients in sustained remission.9 Once remission is achieved, the clinician needs to be confident that the patient will maintain this state in order to ensure the best possible outcomes. Predictors of sustained tumour necrosis factor inhibitor free remission have also been published, for example, short symptom duration prior to commencing therapy;10 however, predictors of disease flare have not been described.

The American College of Rheumatology (ACR) and the EULAR have published criteria for remission.5 However, these criteria were not evaluated as to their ability to predict flare of disease.

Studies have previously demonstrated that ultrasound detected power Doppler (PD) activity (implying active synovial inflammation) may be present in a proportion of patients in clinical remission11 ,12 and this can be used to accurately predict subsequent radiographic structural progression.13 It may be hypothesised that such continued subclinical inflammation despite apparent remission as determined by traditional criteria may trigger a subsequent disease flare.

This study aimed to determine the clinical, functional and imaging factors associated with disease flare and sustained remission in a cohort of disease-modifying antirheumatic drug (DMARD) treated RA patients in clinical remission. In addition, we aimed to determine the impact of disease flare versus sustained remission on long-term, clinical, functional, quality of life and radiographic outcomes.


Consecutive RA patients deemed to be in clinical remission by their assessing consultant rheumatologist were recruited. All patients were taking conventional DMARDs. Patients also had to fulfil the following criteria: (1) RA classified according to the 1987 criteria, (2) age >18 years, (3) no flares of disease in the last 6 months, (4) stable treatment for 6 months and (5) no indication for a change in treatment.

Ethical approval was obtained from the Leeds Teaching Hospitals NHS Trust and written informed consent was obtained from all patients.

Clinical assessments

Clinical assessments were performed at baseline, every 3 months and at the time of flare over a period of 1 year. Clinical assessments included duration of morning stiffness (minutes); visual analogue scale assessments (0–100) for physician and patient global assessment of health and disease activity; number of tender and swollen joints determined by an independent trained metrologist. Joints included in the standard 28-joint count were examined. Routine laboratory tests for C reactive protein, anticyclic citrullinated protein antibody and rheumatoid factor were performed. The Health Assessment Questionnaire Disability Index (HAQ-DI)14 and Rheumatoid Arthritis Quality of Life score (RAQoL)15 were used to assess function and quality of life, respectively.

New and established clinical remission criteria were applied with disease remission being defined as follows: 28-joint count Disease Activity Score (DAS28) <2.616 ,17; Simplified Disease Activity Score (SDAI) ≤3.35 ,7 ,18; 1981 ACR remission criteria (at least five out of six criteria satisfied for two consecutive months19; and 2010 ACR/EULAR criteria (all≤1: tender 28-joint count, swollen 28-joint count, C reactive protein (in mg/dl) and patient global visual analogue scale (0–10 cm)).5

Definition of disease flare

Flare was defined as any increase in disease activity that required an initiation, change or increase in therapy.20 Disease activity was assessed clinically and biochemically.

Ultrasonography assessment

Ultrasound examination of the dominant hand and wrist was performed in all patients using the Phillips ATL HDI 3000 machine with a 10–5 MHz ‘hockey stick’ transducer. Grey Scale (GS) and PD ultrasound were used to assess synovial hypertrophy and synovial vascularity, respectively. The presence and location of any synovial pathology were recorded with reference to standardised definitions agreed by Outcome Measures in Rheumatology Clinical Trials (OMERACT).21 ,22 Ultrasound was performed by a single experienced ultrasonographer (ZK) who was blinded to all other study findings. The dominant hand metacarpophalangeal joints2,,5 and wrist (inter-carpal, radio-carpal, ulnar carpal and distal radio-ulnar compartments) were scanned from dorsal and palmer aspects. Individual joints were scored for GS synovial hypertrophy and PD activity using a validated semiquantitative method (0–3),23 ,24 that is, GS synovial hypertrophy 0: no synovial hypertrophy; 1: mild; 2: moderate; 3: severe; and PD: 0: normal/minimal vascularity; 1: mild hyperaemia; 2: moderate; 3: marked. Ultrasound scores were expressed per patient as a total joint count (where all joints scoring >0 were counted to produce total GS and PD joint counts for each patient). Each patient evaluation took approximately 30 min and representative images were archived. Satisfactory levels of interobserver and intraobserver reliability data have previously been published using this methodology.11 ,12

Radiographic assessment

Postero-anterior radiographs of the hands, wrists and feet were obtained at baseline and 12 months using standardised techniques. Radiographic joint damage was scored according to the Genant-modified Sharp scoring method.25 A single experienced reader (CGP) who was blinded to all other imaging and clinical findings and the chronological sequence of the baseline and follow-up radiographs performed this assessment.

Statistical analysis

In this exploratory analysis no corrections have been made for multiple testing; therefore, emphasis is placed on descriptive statistics throughout. Binary logistic regression was used to identify associations between the various clinical and imaging assessments at baseline and the odds of subsequent flare. Ordinal HAQ-DI and RAQoL scores were Rasch-transformed to interval scaling prior to analysis. The HAQ-DI per 0.1 unit was used to reflect the low HAQ-DI scores observed in this remission cohort. Clinical and imaging variables were compared at 12 months between patients who showed a flare and those who did not using analysis of variancefor normally-distributed continuous interval variables, Mann–Whitney U tests for ordinal or non-normally distributed interval variables, and Pearson's χ2 for categorical variables. All analyses were carried out in SPSS V.17.0.2.


Baseline parameters

Ninety-three patients were recruited (table 1). The median duration of disease was 7 years and the median duration of remission was 22 months. The majority of the patients were on methotrexate monotherapy (n=37, 40%). Fifty-two patients (56%) fulfilled DAS28 remission criteria; thirty-one patients (33%) fulfilled SDAI remission criteria; fifty patients (54%) fulfilled the modified 1981 ACR remission criteria and thirteen patients (14%) fulfilled 2010 ACR/EULAR remission criteria. Eighty-three (89%) patients had GS synovial hypertrophy and 58 fifty-eight (62%) patients had PD activity.

Table 1

Baseline demographic, clinical, biochemical, functional, quality of life and imaging characteristics. All values are median (first quartile to third quartile) unless otherwise stated

Characteristics of disease flare

Twenty-four patients (26%) experienced a disease flare during the study period (four at 3 months; two at 6 months; thirteen at 9 months; five at 12 months) and sixty-nine (74%) remained in remission. The DAS28 at time of flare was available on 19 patients measuring median (IQR) 3.81 (3.40–4.81). Disease flare was managed with non-steroidal anti-inflammatory drug therapy (n=3), intra-articular steroid (n=3), intramuscular steroid (n=9, three of these patients also had an increase in DMARDs), an increase in DMARD therapy (n=11) and an increase in biological DMARD therapy (n=1)).

Factors associated with disease flare

Binary logistic regression analyses were performed to investigate the association between baseline factors and disease flare (table 2). No clinical or biochemical variables were associated with an increased likelihood of flare. None of the clinical remission criteria were associated with substantially reduced odds of flare. The largest effect was seen for the 1981 ACR definition, where 20.0% of those in remission went on to a disease flare compared with 32.6% of those not in remission (OR=0.52 (0.20–1.33), p=0.171). All other composite clinical remission criteria (DAS28, SDAI, ACR/EULAR 2011) performed less well and were unable to demonstrate any accurate predictive validity for disease flare.

Table 2

Demographic, clinical, biochemical and imaging factors associated with flare

When imaging criteria were evaluated, it was noted that 20/58 (35%) of patients with increased PD at baseline had a flare compared with 4/35 (11%) of those with no PD activity (OR=4.08 (1.26–13.19), p=0.014), giving a sensitivity of 83% (95% CI 62% to 95%), specificity 45% (33%–57%), positive predictive value (PPV) 35% (23%–48%) and negative predictive value (NPV) 89% (72%–96%). There was also evidence that the number of joints scoring >0 for PD at baseline was associated with subsequent flare (OR=1.43 (1.01–2.02), p=0.044). There were no substantive associations with the presence of GS synovial hypertrophy or the number of joints showing GS synovial hypertrophy at baseline.

Flare was associated with poor baseline function and quality of life measures: HAQ-DI (OR per 0.1 unit=1.27 (1.07–1.52), p=0.006) and RAQoL (OR=1.10 (1.01–1.20), p=0.036).

Sensitivity to change

Mean changes in clinical variables at time of flare were calculated, along with the associated SD for each, to indicate which were the most sensitive to change (table 3).

Table 3

Changes from baseline in clinical variables at time of flare. Data for five patients were excluded because flare had been treated between study assessments

The utility of combining PD activity and clinical remission criteria for predicting flare

It was hypothesised that combining PD data with clinical remission criteria may improve the ability to predict disease flare (table 4). Irrespective of which remission criteria were used, there were no significant relationships when controlling for the presence or absence of PD activity at baseline (table 4). However, the number of patients who experienced a flare numerically increased regardless of remission criteria when PD was present. Of the patients fulfilling the ACR/EULAR 2011 remission criteria, 50% of patients with PD showed a flare compared with 0% patients with absent PD; however, this subgroup was small (n=8).

Table 4

The likelihood of flare in patients in remission: comparing the presence or absence of PD activity and different remission criteria

Multivariable analysis

Given the number of patients who flared was relatively small it was not possible to construct of a fully comprehensive multiple logistic regression model to identify factors that were independently associated with the odds of flare (table 5 and figure 1). Only PD absent/present and HAQ-DI score in combination with the various remission criteria were included. RAQoL was highly correlated with HAQ-DI (Pearson's r=0.81); therefore, it was not included in the multivariable model. Table 5 shows the results of the multivariable analysis for the four remission criteria. While PD and HAQ-DI were independently associated with the odds of flare, there was no clear relationship with remission status irrespective of which remission criteria were used. PD was the strongest independent predictor of disease flare (OR between 4.53 and 7.57, depending on remission criteria used).

Figure 1

The probability of flare from the regression model that included power Doppler (PD), Health Assessment Questionnaire Disability Index (HAQ-DI) and Simplified Disease Activity Index (SDAI) remission. For all four lines, the probability of flare increases with increasing HAQ-DI score. The presence of PD (solid lines) gives an increased probability of flare compared with absence of PD (dotted lines), but the difference between being in SDAI remission (black lines) or not (grey lines) is comparatively much smaller.

Table 5

The odds of flare in patients in remission: results of multiple binary logistic regression models including baseline measures of synovial hyperaemia, functional impairment and remission according to various criteria

Longitudinal outcome in flare versus sustained remission groups

Radiographic data were available at baseline and 12 months for 83 patients. Eleven of the twenty-two patients who showed a flare (50%) experienced a deterioration in radiographic joint damage compared with 20/61 (33%) who remained in remission (Pearson's χ2=2.05, p=0.152).

At baseline, DAS28 scores were similar between the flare versus sustained remission groups (mean (95% CI) no flare 2.47 (2.25 to 2.69), flare 2.66 (2.20 to 3.11)). However at 12 months, having adjusted for baseline values, there were significant differences in DAS28 scores between these two groups; patients who had a flare had higher DAS28 scores (baseline-adjusted DAS28 at 12 months (95% CI) no flare=2.26 (2.06 to 2.46), flare 2.90 (2.55 to 3.24), ANCOVA F=10.15, p=0.002).

At 12 months, there were no differences between patients who had a flare and those who did not in total PD score (median (IQR) no flare 0.00 (0.00–1.00), flare 0.00 (0.00–3), Mann–Whitney U=967.0, p=0.157) or total GS synovial hypertrophy score (no flare 2.00 (0.00–4.00), flare 3.00 (1.25–5.00), Mann–Whitney U=953.0, p=0.218).

Adjusting for baseline differences, patients who had a flare experienced a worse functional outcome at 12 months reflected by higher HAQ-DI scores (adjusted mean (95% CI) 0.412 (0.344 to 0.481)) compared with patients in sustained remission (0.322 (0.282 to 0.362), F=4.92, p=0.029). A similar, but non-significant descriptive trend was observed for RAQoL (adjusted mean (95% CI) no flare 6.37 (5.61 to 7.13), flare 7.65 (6.37 to 8.93), F=2.85, p=0.095).


In RA patients in clinical remission, PD activity has already been shown to be associated with subsequent radiographic structural damage progression;13 however, the clinical relevance of this has been questioned. This study has now provided evidence that in these patients PD activity may be the best marker of adverse clinical outcomes, manifesting as a flare of disease, which has important long-term implications.

In this cohort of patients in sustained remission, 26% of patients experienced a flare of disease during the 12 months of the study. The key finding from this study is that for patients with ultrasound detected PD activity at baseline the odds of a flare of disease were approximately four times as high as the odds for patients with no PD activity. The finding of active PD signal despite clinical remission provides evidence that flares in disease activity in such patients may be related to incomplete suppression of inflammation. Impaired functional status at baseline, as determined by HAQ-DI, was also associated with increased odds of flare.

Moreover, this study demonstrates that a disease flare has important long-term implications, with patients who experienced a flare having more active disease in addition to inferior function and quality of life outcomes compared with patients in sustained clinical remission.

Once patients have achieved clinical remission, a PD ultrasound assessment may indicate the likelihood or not of remaining in a remission state. A negative PD result could provide greater confidence of maintained remission and greater certainty of good long-term outcome, whereas a positive PD result might alert the rheumatologist to the greater likelihood of disease flare and subsequent adverse long-term clinical, functional and quality of life outcomes. This is supported by the high negative predictive value (84%) for PD activity predicting flare. Patients with absent PD signal appear to be in a more stable remission state and therefore may be suitable for reduced frequency of disease monitoring visits and even consideration for therapy reduction.

Our results suggest that imaging criteria (absence of PD signal) may be superior to ongoing fulfilment of SDAI, DAS28, 1981 ACR or the 2011 ACR/EULAR remission criteria at predicting sustained remission. Indeed, neither the SDAI nor DAS28 provided substantial additional information over PD alone. For example in patients with PD activity at baseline (PD>0), 35% of the patients in SDAI remission (≤3.3) had a flare compared with 34% of those not in SDAI remission. However, our results suggest that if a patient satisfies the 2011 ACR/EULAR remission criteria and has absent PD signal, the risk of flare is minimal (0%). In addition, it is reassuring that if a patient has no PD activity and fulfils SDAI remission criteria disease flare is less, and only 7% of patients in this study flared. Further work with more patients is required to explore the relevance of these results.

There are potential limitations of this study. Only 24 patients experienced a flare limiting the statistical modelling. Data from some of this cohort of patients were included in a previous publication;13 however, the data on flare rates and the value of imaging and clinical remission criteria are novel and have not been previously reported. This study used the OMERACT 9 definition of flare which is a working definition, but the best available. This definition does allow a wide range of patients to be defined as flaring, for example, some patients required non-steroidal anti-inflammatory drug therapy while others required DMARD escalation to treat the flare. However, these patients are representative of those seen in general rheumatology clinics and as the current definition of flare does not distinguish severity of flare or provide advice on the management of flare, clinical decisions are made at the discretion of the rheumatologist. The OMERACT 9 working definition does not make allowances for transient increases in disease activity that may have spontaneously resolved between clinic visits. Berthelot et al have designed a questionnaire to identifying flare and disease exacerbations either transient or long lasting.26 Further work on the definition of flare is currently in progress by the OMERACT 10 RA flare workshop.27 ,28

We chose to utilise ultrasound imaging to assess a model system involving eight joint regions in the dominant hand and wrist. This approach has been successfully applied in previous studies and enables a representative number of joints affected by RA to be scanned within a feasible time period and is a pragmatic and practical approach to ultrasound imaging in a busy clinical setting. It is possible that if more joints were scanned then an increased frequency of PD positive joints would have been identified in a greater number of patients. However, our data already suggest that absence of a PD signal may be a reliable means of ensuring sustained remission (NPV 84%). To reflect routine clinical practice, pragmatic entry criteria were selected for this study, which relied on the clinical skills of an experienced rheumatologist to determine a disease remission state. This enabled us to apply and evaluate all available clinical remission criteria in this setting. Our subsequent findings and in particular the absence of a significant relationship between remission criteria and disease flare or sustained remission provide justification for this approach.

To conclude, patients without PD activity at baseline have a low likelihood of flaring. Of all the currently available disease activity measurement techniques, a negative PD examination may enable the rheumatologist to be the most confident of sustained disease remission. Even a single episode of flare in a patient in sustained remission has important immediate and long-term consequences. Consideration should be given to including an imaging assessment as part of RA remission criteria in order to provide optimal patient assessment and management.



  • Funding Arthritis Research UK.

  • Competing interests None.

  • Ethics approval Leeds Regional Ethics Committee.

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