The aim of rheumatoid arthritis (RA) treatment is remission. The combination of new potent treatments, early intervention, and a treatment strategy of tight control makes remission a daily possibility 123. In clinical practice, active disease means that clinicians will increase the treatment to obtain remission; but that also means that clinicians can stop, add to or increase treatments once the goal is achieved. What is the level of activity to amend treatment?

Even with a combination of multiple therapeutic agents, drug-induced remission is difficult to reach as Pinals criteria 4 are difficult to achieve in clinical trials, even when using the most potent new drugs. A better alternative might therefore be to change the remission criteria to other criteria.

For this purpose a new therapeutic goal has been proposed. A state of near remission or partial remission, or of minimal disease activity or low disease activity, has been proposed in recent years, to be used in clinical trials or guidelines with new disease-modifying antirheumatic drug (DMARD) treatments arriving on the market in the 2000s 3567891011.

This area of low disease activity might be limited by two thresholds, an upper threshold and a lower threshold, which are set to help the treatment decision. The upper threshold of disease activity is the level above which treatment prescription in naïve patients or a switch/intensification in previously treated patients is needed and strongly recommended, and the lower threshold is the level below which treatment maintenance could be recommended; some uncertainties remain for practical issues between these two thresholds.

Persistence of moderate to high disease activity (whatever the measurement criteria we use) is the usual standard to decide whether treatment should be considered ineffective and patients should be switched to another treatment. But on the contrary, when treatment is effective, even with persistence of some degree of low disease activity, the question remains how to decide whether it is worth not changing and better maintaining current treatment, to spare further resources.

Several disease activity measures are currently available, and indices such as the 28-joint disease activity score (DAS28) or derivates are currently used for such decisions at the upper bound. Usually DAS28 > 3.2 is accepted to consider the disease active enough and to reinforce the treatment dosage or to switch the DMARD 56789101112. On the contrary, DAS28 < 2.4 defines remission, but there is no recommendation telling the clinician what to do once this level is reached.

Some uncertainties still remain in defining the lower threshold below which treatment maintenance could be recommended. The first uncertainty is that there is no agreement on the best tool to be used to define this state of near remission. Practically, the DAS28 remains the most used in clinical practice and has been validated is trials - with some areas of uncertainty below the threshold of indication for changing therapy (3.2) and around the level of remission (2.6) as proposed by the DAS28 designers. The second issue is that the definition of such low disease activity implies the recognition of a threshold value of activity under which the disease could be considered low enough to keep treatment unchanged (except steroids), and therefore serves as a basis for clinicians to maintain unchanged the treatment.

As everyday practice is different from clinical trials, we aimed to determine such a threshold based on the expert physicians' decision method and to develop recommendations for clinicians to use the DAS28 in clinical practice in the area of uncertainty (DAS28 of 2 to 3.2), in order to guide the decision to maintain unchanged or to not maintain a DMARD treatment; that is, to avoid changing because of sufficient/acceptable effectiveness.

The present study was designed to determine a low disease activity threshold - that is, a DAS28 value - below which DMARD treatment should not be changed in RA patients.

The present study has evidenced some major findings. The discordance in the decision to maintain or change treatment lies within the range of 2.5 to 3.2 for the DAS28 value, suggesting setting a consensual threshold of 2.4 to define maintenance of treatment. Second, panelists make the decision to maintain DMARD treatment (excluding steroids) with excellent intra-rater reliability, but only moderate to good inter-rater reliability. Also, determinants of the decision have been evidenced with remarkable persistence whatever the level of disease activity in this range; namely, the ESR level, the patient's global activity, the number of tender joints and the number of swollen joints are taken into consideration by experts to make their decision. Fourth, these results are independent of experts' characteristics of age, sex, and year of diploma, but not of type of practice. Finally, according to these results, panelists concluded that in clinical practice there is no need to change DMARD treatment (neither molecule nor dose regimen) when a patient with stable clinical status over the past 3 months has DAS28 ≤ 2.4.

Targeting remission is recommended given recent changes and innovations in therapy as well as the evolution of therapeutic strategies. There is no general consensus, however, regarding definition of remission. Achieving Pinals criteria (five out of six criteria: morning stiffness absent or not exceeding 15 minutes, no fatigue, no joint pain by history, no joint tenderness, no joint or tendon sheath swelling, and no elevation of the ESR) was estimated to be rare in the 1990s 41011. Moreover, Pinals criteria are difficult to apply in clinical trials.

The use of different definitions of RA remission leads to different results with regard to remission rates. Analysis of a database of more than 5,800 RA patients showed that the overall remission rate was lowest using the American College of Rheumatology definition of remission (8.6%), followed by the clinical disease activity index (13.8%) and routine assessment of patient index data (RAPID) 3 (14.3%) definitions; the rate was highest when remission was defined using the DAS28 (19.6%) 17.

Using the disease activity score, a state of near remission has been proposed recently 351011 with two thresholds to help the treatment decision: an upper threshold of disease activity, above which intensification is recommended; and a lower threshold, below which treatment maintenance can be recommended. Usually, DAS28 > 3.2 is accepted to consider the disease active enough and to reinforce the treatment dosage or to switch DMARD 8. The lower bound is less obvious and has been investigated using several methods 5678910: the DAS28 value, the OMERACT definition, the patient acceptable symptom state (PASS) or the RAPID score (patient index data scores)

The value of 2.6 has been proposed by developers of the DAS28 12. After calculating the disease activity of patients included in a cohort as well as a modification of the American Rheumatology Association definition for clinical remission, receiver operating characteristic analysis was used to determine the cut-off point with maximum sensitivity and specificity in the DAS28 corresponding with fulfillment of the modified American College of Rheumatology criteria. The optimal cut-off value corresponding to American College of Rheumatology criteria for remission was 2.66 7.

The disease activity score has been criticised for its performances in the spectrum of a low level of disease activity 111819. We found that panelists were in good agreement below a score of 2.5. Over this threshold, other data may be requested and may actually be fully incorporated in clinical practice judgment. Other patient-reported outcomes, in addition to the global patient assessment component, are also important to consider in weighing up the consequences of activity.

One limitation of the disease activity score remains that it does not specify whether swollen joints are ankles or metatarsophalangeal joints, possibly resulting in some unbalanced estimate of activity.

The OMERACT defined minimal disease activity as patients with no tender or swollen joints and ESR < 10 20. If this definition is not reached, patients must have either a DAS28 value ≤ 2.85 or meet five among seven criteria (pain ≤ 2, health assessment questionnaire score ≤ 0.3, tender joint count ≤ 1, swollen joint count ≤ 1, patient global activity ≤ 2, physician global activity ≤ 1.5, ESR ≤ 20).

As low disease activity is closely related to treatment decision-making, others have proposed to define a minimal clinically important improvement and a PASS. The concept of PASS translates the response at the group level (change in mean scores) into clinically meaningful information by addressing the patient level as therapeutic success (yes/no) in order to help the decision. The concept is not fully formulated at the present time 2122.

Pincus and colleagues proposed a continuous quality improvement approach to assess and manage RA patients without formal joint counts, based on quantitative RAPID scores on a multidimensional health assessment questionnaire 23.

Furst and colleagues, using expert opinion and the Research and Development/University of California in Los Angeles (RAND/UCLA) Appropriateness Method, have proposed recently to consider that only the clinically active joint count should be considered the most important decision factor among a literature review of various outcome measures 24. Among 108 scenarios developed to simulate various clinical situations, the panelists recommended that the clinically active joint count should be the most important decision factor, and in patients with no active joints they recommended that, regardless of other factors, treatment should not be changed. Patients with five or more active joints were considered inadequately treated, patients with no active joints had no need to change therapy, and in patients with one to four active joints other variables must be considered.

As some uncertainty still remains, we aimed to determine such a threshold based on the experts' decision method and to develop recommendations for clinicians in order to guide the decision to maintain unchanged or not a DMARD treatment; that is, to avoid changing because of sufficient/acceptable effectiveness. In this particular area below the threshold of indication for changing therapy (3.2) and at about the level of remission (2.6), we have shown that determinants of the decision have been evidenced with remarkable persistence whatever the level of disease activity in this range - namely, the ESR level, patient's global assessment of disease activity, number of tender joints and number of swollen joints are taken into consideration by experts to make their decision. The importance of patient-reported outcomes is now clearly acknowledged, and our results value both patients' and physicians' criteria.

Development of recommendations for clinical practice may use various methodologies and be data driven or expert driven. Elicitation of expert opinion uses the Delphi method, focus groups, consensus, and so forth. The data-driven methodology (DAS28) is based on the data observed and retrospectively analysed. The data are inherently subjected to indication bias (that is, results are dependent on the initial decision of treatment), which is driven by initial judgment of patient status - while expert opinions are dependent on the experts' choice (that is, representativity of experts is sometimes questionable). This latter observation may have some advantage in term of ease to collect data. Expert opinion is now largely developed and accepted as an alternative to produce consensus and informed decision.

One strength of the present work is that the judgments made were unbiased. The OMERACT method is largely used for reaching consensus, with its validated threshold proposed in the literature 21. Our approach, however, asked experts to make their judgment without knowing the DAS28 value of scenarios - therefore, independently.

One important limitation of the present work is that it does not consider the long-term effect of maintaining low disease activity. Recent works have shown that even if patients are maintained under highly effective treatment, with low disease activity, the long-term deterioration of the joint continues to progress, tempering the enthusiasm of the new class of biologic agents. But in the context of spare resources, the aim is still to preserve the future possibility of action.

Another possible limitation is that we did not consider steroids as DMARDs, even if in light of recent data this opinion should be modified in the near future. We consider steroids useful at the very beginning of the disease, awaiting effectiveness of DMARDs or of local injection; but this is not the method utlilised by numerous colleagues. The last point (also shared with the disease activity score) is that the locations of synovitis of swollen joint(s) were not mentioned and could have made a difference between experts' opinions, even if this possibility seems low 18. Indeed, one way to validate the proposition is to apply it to our current own patients in real life. This validation is ongoing with all of the panelists.

According to the presented results, the STPR group recommends maintaining treatment without changing the molecule or the dose regimen when a patient with stable clinical status over the past 3 months has DAS28 ≤ 2.4. Above this threshold, the number of swollen joints should be considered - independently of or in addition to the DAS28 value - for the clinician to make a decision to maintain or to change therapy.

CI: confidence interval; DAS28: 28-joint disease activity score; DMARD: disease-modifying antirheumatic drug; ESR: erythrocyte sedimentation rate; OMERACT: Outcome Measures in Rheumatoid Arthritis Clinical Trials; PASS: patient acceptable symptom state; RA: rheumatoid arthritis; RAPID: routine assessment of patient index data; STPR: stratégie thérapeutique de la polyarthrite.

The authors declare that they have no competing interests.

All authors are members of the STPR group and thus participated in the project and answered interviews. MdB, BF and FG conceived of the project. FG performed the statistical analysis. MdB, BF XLL and FG organised the study. MdB and FG wrote the paper.

JFM, JMB, BC, R-MF, FL, OM, AS, and DW participated in data analysis, interpretation and assisted in manuscript preparation.

STPR members who participated in the study are as follows: E Palazzo, G Streit, O Vittecoq, P Patoz, P Bennet, Y Maugars, Y Laborie, J Gillard, J Morel, MC Legouffre, H Cholvy, B Saint-Marcoux, S Lasbleiz, B De Bie, V Foltz, B Banneville, G Dubourg, P Philippe, F Pouyol, S Muller, H Korn, V Simon, C Collange, Ph Dieudé, M Ballard, Ch Best, S Jousse, J Allain, P Kervarec, B Augé, L Brault, Ch Piroth, F Pascaud, and N Gerard.